Optical properties of materials and their applications [Second edition] 9781119506003, 111950600X, 9781119506058, 1119506050, 9781119506065, 1119506069, 9781119506317

"Optical properties of a material change or affect the characteristics of light passing through it by modifying its

966 150 23MB

English Pages 647 [659] Year 2020

Report DMCA / Copyright

DOWNLOAD FILE

Polecaj historie

Optical properties of materials and their applications [Second edition]
 9781119506003, 111950600X, 9781119506058, 1119506050, 9781119506065, 1119506069, 9781119506317

Table of contents :
Content: List of Contributors xv Series Preface xvii Preface xix 1 Fundamental Optical Properties of Materials I 1S.O. Kasap, W.C. Tan, Jai Singh, and Asim K. Ray 1.1 Introduction 1 1.2 Optical Constants n and K 2 1.2.1 Refractive Index and Extinction Coefficient 2 1.2.2 n and K, and Kramers-Kronig Relations 5 1.3 Refractive Index and Dispersion 7 1.3.1 Cauchy Dispersion Relation 7 1.3.2 Sellmeier Equation 8 1.3.3 Refractive Index of Semiconductors 10 1.3.3.1 Refractive Index of Crystalline Semiconductors 10 1.3.3.2 Bandgap and Temperature Dependence 11 1.3.4 Refractive Index of Glasses 11 1.3.5 Wemple-DiDomenico Dispersion Relation 14 1.3.6 Group Index 15 1.4 The Swanepoel Technique: Measurement of n and

Citation preview

Optical Properties of Materials and Their Applications

Wiley Series in Materials for Electronic and Optoelectronic Applications www.wiley.com/go/meoa Series Editors Professor Arthur Willoughby, University of Southampton, Southampton, UK Dr Peter Capper, Ex-Leonardo MW Ltd, Southampton, UK Professor Sofa Kasap, University of Saskatchewan, Saskatoon, Canada Published Titles Bulk Crystal Growth of Electronic, Optical and Optoelectronic Materials, Edited by P. Capper Properties of Group-IV, III—V and II—VI Semiconductors, S. Adachi Charge Transport in Disordered Solids with Applications in Electronics, Edited by S. Baranovski Optical Properties of Condensed Matter and Applications, Edited by J. Singh Thin Film Solar Cells: Fabrication, Characterization, and Applications, Edited by J. Poortmans and V. Arkhipov Dielectric Films for Advanced Microelectronics, Edited by M. R. Baklanov, M. Green, and K. Maex Liquid Phase Epitaxy of Electronic, Optical and Optoelectronic Materials, Edited by P. Capper and M. Mauk Molecular Electronics: From Principles to Practice, M. Petty Luminescent Materials and Applications, A. Kitai CVD Diamond for Electronic Devices and Sensors, Edited by R. S. Sussmann Properties of Semiconductor Alloys: Group-IV, III—V and II—VI Semiconductors, S. Adachi Mercury Cadmium Telluride, Edited by P. Capper and J. Garland Zinc Oxide Materials for Electronic and Optoelectronic Device Applications, Edited by C. Litton, D. C. Reynolds, and T. C. Collins Lead-Free Solders: Materials Reliability for Electronics, Edited by K. N. Subramunian Silicon Photonics: Fundamentals and Devices, M. Jamal Deen and P. K. Basu Nanostructured and Subwavelength Waveguides: Fundamentals and Applications, M. Skorobogatiy Photovoltaic Materials: From Crystalline Silicon to Third-Generation Approaches, Edited by G. Conibeer and A. Willoughby Glancing Angle Deposition of Thin Films: Engineering the Nanoscale, Matthew M. Hawkeye, Michael T. Taschuk, and Michael J. Brett Physical Properties of High-Temperature Superconductors, R. Wesche Spintronics for Next Generation Innovative Devices, Edited by Katsuaki Sato and Eiji Saitoh Inorganic Glasses for Photonics: Fundamentals, Engineering and Applications, Animesh Jha Amorphous Semiconductors: Structural, Optical and Electronic Properties, Kazuo Morigaki, Sandor Kugler, and Koichi Shimakawa Microwave Materials and Applications, Two volume set, Edited by Mailadil T. Sebastian, Rick Ubic, and Heli Jantunen Molecular Beam Epitaxy: Materials and Applications for Electronics and Optoelectronics, Edited by Hajime Asahi and Yoshiji Korikoshi Metalorganic Vapor Phase Epitaxy (MOVPE): Growth, Materials Properties, and Applications, Edited by Stuart Irvine and Peter Capper

Optical Properties of Materials and Their Applications

Edited by Jai Singh College of Engineering, IT and Environment Charles Darwin University, Darwin, Australia

Second Edition

This edition first published 2020 © 2020 John Wiley & Sons Ltd Edition History John Wiley & Sons Inc. (1e, 2006) All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, except as permitted by law. Advice on how to obtain permission to reuse material from this title is available at http://www.wiley.com/go/ permissions. The right of Jai Singh to be identified as the author of the editorial material in this work has been asserted in accordance with law. Registered Offices John Wiley & Sons, Inc., 111 River Street, Hoboken, NJ 07030, USA John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK Editorial Office The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, UK For details of our global editorial offices, customer services, and more information about Wiley products visit us at www.wiley.com. Wiley also publishes its books in a variety of electronic formats and by print-on-demand. Some content that appears in standard print versions of this book may not be available in other formats. Limit of Liability/Disclaimer of Warranty In view of ongoing research, equipment modifications, changes in governmental regulations, and the constant flow of information relating to the use of experimental reagents, equipment, and devices, the reader is urged to review and evaluate the information provided in the package insert or instructions for each chemical, piece of equipment, reagent, or device for, among other things, any changes in the instructions or indication of usage and for added warnings and precautions. While the publisher and authors have used their best efforts in preparing this work, they make no representations or warranties with respect to the accuracy or completeness of the contents of this work and specifically disclaim all warranties, including without limitation any implied warranties of merchantability or fitness for a particular purpose. No warranty may be created or extended by sales representatives, written sales materials or promotional statements for this work. The fact that an organization, website, or product is referred to in this work as a citation and/or potential source of further information does not mean that the publisher and authors endorse the information or services the organization, website, or product may provide or recommendations it may make. This work is sold with the understanding that the publisher is not engaged in rendering professional services. The advice and strategies contained herein may not be suitable for your situation. You should consult with a specialist where appropriate. Further, readers should be aware that websites listed in this work may have changed or disappeared between when this work was written and when it is read. Neither the publisher nor authors shall be liable for any loss of profit or any other commercial damages, including but not limited to special, incidental, consequential, or other damages. Library of Congress Cataloging-in-Publication Data Names: Singh, Jai, editor. Title: Optical properties of materials and their applications / edited by Jai Singh (College of Engineering, IT, and Environment, Charles Darwin University, Darwin, Australia) Other titles: Optical properties of condensed matter and applications. | Optical properties of condensed matter and applications. Description: Second edition. | Hoboken, NJ : John Wiley & Sons, 2020. | Series: Wiley series in materials for electronic and optoelectronic applications | Previous edition: Optical properties of condensed matter and applications, 2006. | Includes bibliographical references and index. Identifiers: LCCN 2019023895 (print) | LCCN 2019023896 (ebook) | ISBN 9781119506317 (cloth) | ISBN 9781119506065 (adobe pdf ) | ISBN 9781119506058 (epub) Subjects: LCSH: Condensed matter–Optical properties. | Materials–Optical properties. | Electrooptics–Materials. Classification: LCC QC173.458.O66 O68 2020 (print) | LCC QC173.458.O66 (ebook) | DDC 530.4/12–dc23 LC record available at https://lccn.loc.gov/2019023895 LC ebook record available at https://lccn.loc.gov/2019023896 Cover Design: Wiley Cover Images: © mitchFOTO / Shutterstock Set in 10/12pt WarnockPro by SPi Global, Chennai, India 10 9 8 7 6 5 4 3 2 1

v

Contents List of Contributors Series Preface Preface 1 Fundamental Optical Properties of Materials I S.O. Kasap, W.C. Tan, Jai Singh, and Asim K. Ray

1.1 1.2

Introduction Optical Constants n and K 1.2.1 Refractive Index and Extinction Coefficient 1.2.2 n and K, and Kramers–Kronig Relations 1.3 Refractive Index and Dispersion 1.3.1 Cauchy Dispersion Relation 1.3.2 Sellmeier Equation 1.3.3 Refractive Index of Semiconductors 1.3.3.1 Refractive Index of Crystalline Semiconductors 1.3.3.2 Bandgap and Temperature Dependence 1.3.4 Refractive Index of Glasses 1.3.5 Wemple–DiDomenico Dispersion Relation 1.3.6 Group Index 1.4 The Swanepoel Technique: Measurement of n and 𝛼 for Thin Films on Substrates 1.4.1 Uniform Thickness Films 1.4.2 Thin Films with Non-uniform Thickness 1.5 Transmittance and Reflectance of a Partially Transparent Plate 1.6 Optical Properties and Diffuse Reflection: Schuster–Kubelka–Munk Theory 1.7 Conclusions Acknowledgments References 2 Fundamental Optical Properties of Materials II S.O. Kasap, K. Koughia, Jai Singh, Harry E. Ruda, and Asim K. Ray

2.1 2.2 2.3

Introduction Lattice or Reststrahlen Absorption and Infrared Reflection Free Carrier Absorption (FCA)

xv xvii xix 1

1 2 2 5 7 7 8 10 10 11 11 14 15 16 16 22 25 27 31 31 32 37

37 40 42

vi

Contents

2.4 2.5 2.6

Band-to-Band or Fundamental Absorption (Crystalline Solids) Impurity Absorption and Rare-Earth Ions Effect of External Fields 2.6.1 Electro-Optic Effects 2.6.2 Electro-Absorption and Franz–Keldysh Effect 2.6.3 Faraday Effect 2.7 Effective Medium Approximations 2.8 Conclusions Acknowledgments References 3 Optical Properties of Disordered Condensed Matter Koichi Shimakawa, Jai Singh, and S.K. O’Leary

3.1 3.2

Introduction Fundamental Optical Absorption (Experimental) 3.2.1 Amorphous Chalcogenides 3.2.2 Hydrogenated Nano-Crystalline Silicon (nc-Si:H) 3.3 Absorption Coefficient (Theory) 3.4 Compositional Variation of the Optical Bandgap 3.4.1 In Amorphous Chalcogenides 3.5 Conclusions References 4 Optical Properties of Glasses Andrew Edgar

4.1 4.2 4.3 4.4 4.5

Introduction The Refractive Index Glass Interfaces Dispersion Sensitivity of the Refractive Index 4.5.1 Temperature Dependence 4.5.2 Stress Dependence 4.5.3 Magnetic Field Dependence—The Faraday Effect 4.5.4 Chemical Perturbations—Molar Refractivity 4.6 Glass Color 4.6.1 Coloration by Colloidal Metals and Semiconductors 4.6.2 Optical Absorption in Rare-Earth-Doped Glass 4.6.3 Absorption by 3d Metal Ions 4.7 Fluorescence in Rare-Earth-Doped Glass 4.8 Glasses for Fiber Optics 4.9 Refractive Index Engineering 4.10 Glass and Glass–Fiber Lasers and Amplifiers 4.11 Valence Change Glasses 4.12 Transparent Glass Ceramics 4.12.1 Introduction 4.12.2 Theoretical Basis for Transparency

45 48 54 54 55 56 58 61 61 62 67

67 69 69 72 74 79 79 80 80 83

83 84 86 88 90 90 91 92 94 95 95 96 99 102 104 106 109 111 114 114 116

Contents

4.12.3 Rare-Earth-Doped Transparent Glass Ceramics for Active Photonics 4.12.4 Ferroelectric Transparent Glass Ceramics 4.12.5 Transparent Glass Ceramics for X-ray Storage Phosphors 4.13 Conclusions References 5 Concept of Excitons Jai Singh, Harry E. Ruda, M.R. Narayan, and D. Ompong

5.1 5.2

Introduction Excitons in Crystalline Solids 5.2.1 Excitonic Absorption in Crystalline Solids 5.3 Excitons in Amorphous Semiconductors 5.3.1 Excitonic Absorption in Amorphous Solids 5.4 Excitons in Organic Semiconductors 5.4.1 Photoexcitation and Formation of Excitons 5.4.1.1 Photoexcitation of Singlet Excitons Due to Exciton–Photon Interaction 5.4.1.2 Excitation of Triplet Excitons 5.4.2 Exciton Up-Conversion 5.4.3 Exciton Dissociation 5.4.3.1 Conversion from Frenkel to CT Excitons 5.4.3.2 Dissociation of CT Excitons 5.5 Conclusions References 6 Photoluminescence Takeshi Aoki

6.1 6.2

6.3

6.4

Introduction Fundamental Aspects of Photoluminescence (PL) in Materials 6.2.1 Intrinsic Photoluminescence 6.2.2 Extrinsic Photoluminescence 6.2.3 Up-Conversion Photoluminescence (UCPL) 6.2.4 Other Related Optical Transitions Experimental Aspects 6.3.1 Static PL Spectroscopy 6.3.2 Photoluminescence Excitation Spectroscopy (PLE) and Photoluminescence Absorption Spectroscopy (PLAS) 6.3.3 Time Resolved Spectroscopy (TRS) 6.3.4 Time-Correlated Single Photon Counting (TCSPC) 6.3.5 Frequency-Resolved Spectroscopy (FRS) 6.3.6 Quadrature Frequency Resolved Spectroscopy (QFRS) Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique 6.4.1 Overview 6.4.2 Dual-Phase Double Lock-in (DPDL) QFRS Technique

120 121 121 124 124 129

129 130 133 135 137 139 140 141 142 147 148 151 152 153 154 157

157 158 159 160 162 163 164 164 167 168 171 172 173 175 175 176

vii

viii

Contents

6.4.3

Exploring Broad PL Lifetime Distribution in a-Si:H by Wideband QFRS 6.4.3.1 Effects of Excitation Intensity, Excitation, and Emission Energies 6.4.3.2 Temperature Dependence 6.4.3.3 Effect of Electric and Magnetic Fields 6.4.4 Residual PL Decay of a-Si:H 6.5 QFRS on Up-Conversion Photoluminescence (UCPL) of RE-Doped Materials 6.6 Conclusions Acknowledgments References 7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors Jai Singh

7.1 7.2

Introduction Photoluminescence 7.2.1 Radiative Recombination Operator and Transition Matrix Element 7.2.2 Rates of Spontaneous Emission 7.2.2.1 At Nonthermal Equilibrium 7.2.2.2 At Thermal Equilibrium 7.2.2.3 Determining E0 7.2.3 Results of Spontaneous Emission and Radiative Lifetime 7.2.4 Temperature Dependence of PL 7.2.5 Excitonic Concept 7.3 Photoinduced Changes in Amorphous Chalcogenides 7.3.1 Effect of Photo-Excitation and Phonon Interaction 7.3.2 Excitation of a Single Electron–Hole Pair 7.3.3 Pairing of Like Excited Charge Carriers 7.4 Radiative Recombination of Excitons in Organic Semiconductors 7.4.1 Rate of Fluorescence 7.4.2 Rate of Phosphorescence 7.4.3 Organic Light Emitting Diodes (OLEDs) 7.4.3.1 Second- and Third-Generation OLEDs: TADF 7.5 Conclusions Acknowledgments References

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses Sandor Kugler, Rozália Lukács, and Koichi Shimakawa

8.1 8.2 8.3 8.4

Introduction Atomic-Scale Computer Simulations of Photoinduced Volume Changes Effect of Illumination Kinetics of Volume Change

178 179 184 185 189 192 197 198 198

203

203 205 206 211 212 214 215 216 222 223 225 226 228 229 232 233 233 234 235 236 236 237 241

241 243 244 245

Contents

8.4.1 a-Se 8.4.2 a-As2 Se3 8.5 Additional Remarks 8.6 Conclusions References 9 Properties and Applications of Photonic Crystals Harry E. Ruda and Naomi Matsuura

9.1 9.2

Introduction PC Overview 9.2.1 Introduction to PCs 9.2.2 Nanoengineering of PC Architectures 9.2.3 Materials Selection for PCs 9.3 Tunable PCs 9.3.1 Tuning PC Response by Changing the Refractive Index of Constituent Materials 9.3.1.1 PC Refractive Index Tuning Using Light 9.3.1.2 PC Refractive Index Tuning Using an Applied Electric Field 9.3.1.3 Refractive Index Tuning of Infiltrated PCs 9.3.1.4 PC Refractive Index Tuning by Altering the Concentration of Free Carriers (Using Electric Field or Temperature) in Semiconductor-Based PCs 9.3.2 Tuning PC Response by Altering the Physical Structure of the PC 9.3.2.1 Tuning PC Response Using Temperature 9.3.2.2 Tuning PC Response Using Magnetism 9.3.2.3 Tuning PC Response Using Strain 9.3.2.4 Tuning PC Response Using Piezoelectric Effects 9.3.2.5 Tuning PC Response Using MEMS Actuation 9.4 Selected Applications of PC 9.4.1 Waveguide Devices 9.4.2 Dispersive Devices 9.4.3 Add/Drop Multiplexing Devices 9.4.4 Applications of PCs for Light-Emitting Diodes (LEDs) and Lasers 9.5 Conclusions Acknowledgments References 10 Nonlinear Optical Properties of Photonic Glasses Keiji Tanaka

10.1 Introduction 10.2 Photonic Glass 10.3 Nonlinear Absorption and Refractivity 10.3.1 Fundamentals 10.3.2 Two-Photon Absorption

245 246 248 249 249 251

251 252 252 253 255 255 256 256 256 257

257 258 258 258 258 259 260 260 261 262 262 263 265 265 265 269

269 271 272 272 275

ix

x

Contents

10.3.3 Nonlinear Refractivity 10.4 Nonlinear Excitation-Induced Structural Changes 10.4.1 Fundamentals 10.4.2 Oxides 10.4.3 Chalcogenides 10.5 Conclusions 10.A Addendum: Perspectives on Optical Devices References 11 Optical Properties of Organic Semiconductors Takashi Kobayashi and Hiroyoshi Naito

11.1 Introduction 11.2 Molecular Structure of π-Conjugated Polymers 11.3 Theoretical Models 11.4 Absorption Spectrum 11.5 Photoluminescence 11.6 Non-Emissive Excited States 11.7 Electron–Electron Interaction 11.8 Interchain Interaction 11.9 Conclusions References 12 Organic Semiconductors and Applications Furong Zhu

12.1 Introduction 12.1.1 Device Architecture and Operation Principle 12.1.2 Technical Challenges and Process Integration 12.2 Anode Modification for Enhanced OLED Performance 12.2.1 Low-Temperature High-Performance ITO 12.2.1.1 Experimental Methods 12.2.1.2 Morphological Properties 12.2.1.3 Electrical Properties 12.2.1.4 Optical Properties 12.2.1.5 Compositional Analysis 12.2.2 Anode Modification 12.2.3 Electroluminescence Performance of OLEDs 12.3 Flexible OLEDs 12.3.1 Flexible OLEDs on Ultrathin Glass Substrate 12.3.2 Flexible Top-Emitting OLEDs on Plastic Foils 12.3.2.1 Top-Emitting OLEDs 12.3.2.2 Flexible TOLEDs on Plastic Foils 12.4 Solution-Processable High-Performing OLEDs 12.4.1 Performance of OLEDs with a Hybrid MoO3 -PEDOT:PSS Hole Injection Layer (HIL) 12.4.2 Morphological Properties of the MoO3 -PEDOT:PSS HIL

278 280 280 281 283 285 286 288 295

295 296 298 300 304 306 309 314 320 321 323

323 324 325 327 327 328 329 331 333 336 339 340 345 346 347 348 350 353 353 361

Contents

12.4.3 Surface Electronic Properties of MoO3 -PEDOT:PSS HIL 12.5 Conclusions References 13 Transparent White OLEDs Choi Wing Hong and Furong Zhu

13.1 Introduction—Progress in Transparent WOLEDs 13.2 Performance of WOLEDs 13.2.1 Optimization of Dichromatic WOLEDs 13.2.2 J-L-V Characteristics of WOLEDs 13.2.3 Electron-Hole Current Balance in Transparent WOLEDs 13.3 Emission Behavior of Transparent WOLEDs 13.3.1 Visible-Light Transparency of WOLEDs 13.3.2 L-J Characteristics of Transparent WOLEDs 13.3.3 Angular-Dependent Color Stability of Transparent WOLEDs 13.4 Conclusions References 14 Optical Properties of Thin Films V.-V. Truong, S. Tanemura, A. Haché, and L. Miao

14.1 Introduction 14.2 Optics of Thin Films 14.2.1 An Isotropic Film on a Substrate 14.2.2 Matrix Methods for Multi-Layered Structures 14.2.3 Anisotropic Films 14.3 Reflection-Transmission Photoellipsometry for Determination of Optical Constants 14.3.1 Photoellipsometry of a Thick or a Thin Film 14.3.2 Photoellipsometry for a Stack of Thick and Thin Films 14.3.3 Remarks on the Reflection-Transmission Photoellipsometry Method 14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies 14.4.1 Electrochromic Thin Films 14.4.2 Pure and Metal-Doped VO2 Thermochromic Thin Films 14.4.3 Temperature-Stabilized V1-x Wx O2 Sky Radiator Films 14.4.4 Optical Functional TiO2 Thin Film for Environmentally Friendly Technologies 14.5 Application of Tunable Thin Films to Phase and Polarization Modulation 14.6 Conclusions References 15 Optical Characterization of Materials by Spectroscopic Ellipsometry J. Mistrík

15.1 Introduction 15.2 Notions of Light Polarization

363 368 369 373

373 374 374 377 384 386 386 390 395 400 400 403

403 404 404 406 407 408 408 410 412 412 413 414 417 420 424 430 430 435

435 436

xi

xii

Contents

15.3 Measureable Quantities 15.4 Instrumentation 15.5 Single Interface 15.6 Single Layer 15.7 Multilayer 15.8 Linear Grating 15.9 Conclusions Acknowledgments References 16 Excitonic Processes in Quantum Wells Jai Singh and I.-K. Oh

16.1 16.2 16.3 16.4

438 441 442 448 454 458 462 463 463 465

Introduction Exciton–Phonon Interaction Exciton Formation in QWs Assisted by Phonons Nonradiative Relaxation of Free Excitons 16.4.1 Intraband Processes 16.4.2 Interband Processes 16.5 Quasi-2D Free-Exciton Linewidth 16.6 Localization of Free Excitons 16.7 Conclusions References

465 466 467 474 475 479 485 491 499 500

17 Optoelectronic Properties and Applications of Quantum Dots Jørn M. Hvam

503

17.1 Introduction 17.2 Epitaxial Growth and Structure of Quantum Dots 17.2.1 Self-Assembled Quantum Dots 17.2.2 Site-Controlled Growth on Patterned Substrates 17.2.3 Natural or Interface Quantum Dots 17.2.4 Quantum Dots in Nanowires 17.3 Excitons in Quantum Dots 17.3.1 Quantum-Dot Bandgap 17.3.2 Optical Transitions 17.4 Optical Properties 17.4.1 Radiative Lifetime, Oscillator Strength, and Internal Quantum Efficiency 17.4.2 Linewidth, Coherence, and Dephasing 17.4.3 Transient Four-Wave Mixing 17.5 Quantum Dot Applications 17.5.1 Quantum Dot Lasers and Optical Amplifiers 17.5.1.1 Gain Dynamics 17.5.1.2 Homogeneous Broadening and Dephasing 17.5.1.3 Long-Wavelength Lasers 17.5.1.4 Nano Lasers 17.5.2 Single-Photon Emitters 17.5.2.1 Micropillars and Nanowires

503 504 504 505 506 507 508 509 510 513 514 516 517 520 520 522 524 526 527 527 530

Contents

17.5.2.2 Photonic Crystal Waveguide 17.6 Conclusions Acknowledgments References 18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications Junwei Xu, D.L. Carroll, K. Biswas, F. Moretti, S. Gridin, and R.T. Williams

18.1 Introduction 18.1.1 Review 18.1.2 The Structures 18.1.2.1 Simple Cubic Frameworks 18.1.2.2 The Multiplicity of Hybrids 18.1.2.3 Structural Variation 18.2 Hybrid Perovskites in Photovoltaics 18.2.1 Review 18.2.2 The Phenomena Characterized as “Defect Tolerance” 18.3 Light-Emitting Diodes Using Solution-Processed Lead Halide Perovskites 18.3.1 Review 18.3.2 Construction and Characterization of LEDs Utilizing CsPbBr3 Nano-Inclusions in Cs4 PbBr6 as the Electroluminescent Medium 18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects 18.5 Conclusions Acknowledgments References

531 533 534 534

537

537 537 538 538 539 540 544 544 548 549 549

553 562 582 583 583

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures Akihiro Murayama and Yasuo Oka

589

19.1 Introduction 19.2 Quantum Wells 19.2.1 Spin Injection 19.2.2 Study of Spin Dynamics by Pump-Probe Spectroscopy 19.3 Fabrication of Nanostructures by Electron-Beam Lithography 19.4 Self-Assembled Quantum Dots 19.5 Hybrid Nanostructures with Ferromagnetic Materials 19.6 Conclusions Acknowledgments References

589 591 591 594 596 599 604 607 608 609

xiii

xiv

Contents

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors Ruben Jeronimo Freitas and Koichi Shimakawa

611

20.1 Introduction 20.2 A Review of PPC in III-V Semiconductors 20.3 Key Physical Terms Related to PPC 20.3.1 Dispersive Reaction 20.3.2 SEF and Power Law 20.3.3 Waiting Time Distribution 20.4 Kinetics of PPC in III-V Semiconductors 20.5 Conclusions Acknowledgments 20.A On the Reaction Rate Under the Uniform Distribution References

611 613 615 615 616 617 617 623 623 623 625

Index

627

xv

List of Contributors Takeshi Aoki Joint Research Center of High-technology, Department of Electronics and

Information Technology, Tokyo Polytechnic University, Atsugi, Japan K. Biswas Department of Chemistry and Physics, Arkansas State University, Jonesboro,

USA D.L. Carroll Department of Physics and Nanotechnology Center, Wake Forest University,

Winston-Salem, North Carolina, USA Andrew Edgar School of Chemical and Physical Sciences, Victoria University of

Wellington, New Zealand Ruben Jeronimo Freitas Department of Electrical and Electronic Engineering, National

University of Timor Lorosae, Díli, East Timor S. Gridin Department of Physics and Nanotechnology Center, Wake Forest University,

Winston-Salem, North Carolina, USA A. Haché Département de physique et d’astronomie, Université de Moncton, New

Brunswick, Canada Jørn M. Hvam Department of Photonics Engineering, Technical University of Denmark,

Kongens Lyngby, Denmark S.O. Kasap Department of Electrical and Computer Engineering, University of

Saskatchewan, Saskatoon, Canada Takashi Kobayashi Department of Physics and Electronics, Osaka Prefecture University,

Sakai, Japan K. Koughia Department of Electrical and Computer Engineering, University of

Saskatchewan, Saskatoon, Canada Sandor Kugler Department of Theoretical Physics, Budapest University of Technology

and Economics, Hungary Rozália Lukács Norwegian University of Life Sciences, Ås, Akershus, Norway Naomi Matsuura Centre for Nanotechnology, University of Toronto, Canada L. Miao Guilin University of Electronic Technology, Guangxi, P.R. China

xvi

List of Contributors

J. Mistrík Center of Materials and Nanotechnologies, Faculty of Chemical Technology,

University of Pardubice, Czech Republic F. Moretti Lawrence Berkeley National Laboratory, Berkeley, California, USA Akihiro Murayama Graduate School of Information Science and Technology, Hokkaido

University, Sapporo, Japan Hiroyoshi Naito The Research Institute for Molecular Electronic Devices, Osaka

Prefecture University, Sakai, Japan M.R. Narayan College of Engineering, Information Technology and Environment,

Charles Darwin University, Darwin, Australia S.K. O’Leary School of Engineering, The University of British Columbia, Kelowna,

Canada I.-K. Oh College of Engineering, Information Technology and Environment, Charles

Darwin University, Darwin, Australia Yasuo Oka Institute of Multidisciplinary Research for Advanced Materials, Tohoku

University, Sendai, Miyagi, Japan D. Ompong College of Engineering, Information Technology and Environment, Charles

Darwin University, Darwin, Australia Asim K. Ray Department of Electrical & Computer Engineering, Brunel University

London, Uxbridge, UK Harry E. Ruda Centre for Nanotechnology and Electronic and Photonic Materials Group,

Department of Materials Science, University of Toronto, Ontario, Canada Koichi Shimakawa Department of Electrical and Electronic Engineering, Gifu University,

Japan Jai Singh College of Engineering, Information Technology and Environment, Charles

Darwin University, Darwin, Australia W.C. Tan Department of Electrical & Computer Engineering, National University of

Singapore, Kent Ridge, Singapore Keiji Tanaka Department of Applied Physics, Graduate School of Engineering, Hokkaido

University, Sapporo, Japan S. Tanemura Japan Fine Ceramics Centre, Mutsuno, Atsuta-ku, Nagoya, Japan V.-V. Truong Physics Department, Concordia University, Montreal, Quebec, Canada R.T. Williams Department of Physics and Nanotechnology Center, Wake Forest

University, Winston-Salem, North Carolina, USA Choi Wing Hong, Department of Physics, Hong Kong Baptist University, Kowloon Tong,

China Junwei Xu Department of Physics and Nanotechnology Center, Wake Forest University,

Winston-Salem, North Carolina, USA Furong Zhu Department of Physics, Hong Kong Baptist University, Kowloon Tong, China

xvii

Series Preface Wiley Series in Materials for Electronic and Optoelectronic Applications This book series is devoted to the rapidly developing class of materials used for electronic and optoelectronic applications. It is designed to provide much-needed information on the fundamental scientific principles of these materials, together with how these are employed in technological applications. These books are aimed at (postgraduate) students, researchers, and technologists engaged in research, development, and the study of materials in electronics and photonics, and at industrial scientists developing new materials, devices, and circuits for the electronic, optoelectronic, and communications industries. The development of new electronic and optoelectronic materials depends not only on materials engineering at a practical level, but also on a clear understanding of the properties of materials and the fundamental science behind these properties. It is the properties of a material that eventually determine its usefulness in an application. The series therefore also includes such titles as electrical conduction in solids, optical properties, thermal properties, and so on, all with applications and examples of materials in electronics and optoelectronics. The characterization of materials is also covered within the series as much as it is impossible to develop new materials without the proper characterization of their structure and properties. Structure–property relationships have always been fundamentally and intrinsically important to materials science and engineering. Materials science is well known for being one of the most interdisciplinary sciences. It is the interdisciplinary aspect of materials science that has led to many exciting discoveries, new materials, and new applications. It is not unusual to find scientists with a chemical engineering background working on materials projects with applications in electronics. In selecting titles for the series, we have tried to maintain the interdisciplinary aspect of the field, and hence its excitement to researchers in this field. Arthur Willoughby Peter Capper Safa Kasap

xix

Preface The second edition, being published more than 10 years after the first edition, presents state-of-the-art developments in almost all topics related to the optical properties of materials and their applications presented in the first edition. Since the publication of the first edition in 2006, many advances have been made in fields such as the optical properties of materials, electroluminescence in organic light-emitting devices, organic solar cells, opto-electronic devices, etc. It is hence very timely to update all the chapters in the first edition by adding developments since 2006 to produce the second edition. This second edition contains 15 of the original 16 chapters, all of which have been updated, as well as 5 brand new chapters, contributed by very experienced and well-known scientists and groups available on different aspects of the optical properties of materials. The study of optical properties of materials has now become an interdisciplinary field, and scientists of physical, chemical, and biological sciences; nanotechnology engineers; and industry researchers have strong interests in this field. The field offers one of the fastest-growing research platforms in material sciences. The second edition covers many examples and applications in the field of electronic and optoelectronic properties of materials, and in photonics. Most chapters are presented to be relatively independent with minimal cross-referencing, and chapters with complementary contents are arranged together to facilitate a reader with cross-referencing. Books written in this field mostly follow one of the two pedagogies: chapters are either based on (i) physical processes, or (ii) the various classes of materials. This book combines the two approaches by first identifying the processes that should be described in detail, and then introducing the relevant classes of materials. Many books also miss the details of how various optical properties are measured. This book presents a comprehensive review of experimental techniques, including recent advances in ultrafast (femtosecond) spectroscopy of materials. Not many books are currently available with such a wide coverage of the field with clarity and levels of readership in a single volume as this book. In Chapters 1 and 2 by Kasap et al., the fundamental optical properties of materials are reviewed, and as such these chapters are expected to refresh the readers with the basics by providing useful optical relations. In Chapter 3, Shimakawa et al. present an up-to-date review of the optical properties of disordered inorganic solids, and Chapter 4 by Edgar presents an extensive discussion on the optical properties of glasses. Chapter 5 by Singh and co-workers presents the concept of excitons in inorganic and organic semiconductors, both crystalline and non-crystalline variants. In Chapter 6, Aoki has presented a comprehensive review of the experimental advances in the techniques of

xx

Preface

measuring photoluminescence together with updates in luminescence results in amorphous semiconductors, and Chapter 7 by Singh complements the theoretical advances in the field of photoluminescence and photoinduced changes in non-crystalline semiconductors. In Chapter 8 by Kugler et al., recent advances in the simulation of photoinduced bond breaking and volume changes in chalcogenide glasses are presented. In Chapter 9, Ruda and Matsuura present a comprehensive review of the properties and applications of photonic crystals. In Chapter 10, Tanaka has presented an up-to-date review of the nonlinear optical properties of photonic glasses. Chapter 11 by Kobayashi and Naito discusses the fundamental optical properties of organic semiconductors. In Chapter 12, Zhu has presented a comprehensive review of the applications of organic semiconductors, in particular, in developing organic light-emitting diodes (OLEDs). In Chapter 13, Hong and Zhu have reviewed the recent developments in the fabrication of transparent white light-emitting diodes (WOLEDs). This is a new chapter added in the second edition. In Chapter 14, Truong and Tanemura have presented an up-to-date review of the optical properties of thin films and their applications, and Chapter 15 by Mistrik deals with the optical characterization of materials by spectroscopic ellipsometry. This is the second new chapter in the second edition. In Chapter 16, Singh and Oh have discussed the excitonic processes in quantum wells. In Chapter 17, the third new chapter in this edition, Hvam has presented an up-to-date comprehensive review of the optoelectronic properties and applications of quantum dots. Chapter 18 by Xu et al. presents up-to-date developments in the applications of perovskites. This is the fourth new chapter in the second edition. In Chapter 19, Murayama and Oka have presented the optical properties and spin dynamics of diluted magnetic semiconductor nanostructures. In the final Chapter 20, the fifth new chapter in this edition, Freitas and Shimakawa have discussed the kinetics of the persistent photoconductivity in Crystalline III–V semiconductors. Thus, the addition of the five new chapters on transparent WOLELDs, ellipsometry, quantum dots, perovskites, and persistent photoconductivity widens the scope of the second edition to a new level. One of the chapters on the negative index of refraction in the first edition has not been included in the second edition at the request of the authors. The readership of the book is expected to be the senior undergraduate and postgraduate students, and teaching and research professionals in the field. In conclusion, I am very grateful to all the contributing authors of the second edition for their utmost co-operation in meeting the deadlines, without which this project would not have concluded. I also would like to acknowledge the technical support from Drs Stefanija Klaric and Luis Herrera Diaz in preparing my chapters. I would also like to thank my friend Beth Woof for her support throughout the course of preparation of this volume. Darwin, Australia

Jai Singh

1

1 Fundamental Optical Properties of Materials I S.O. Kasap 1 , W.C. Tan 2 , Jai Singh 3 , and Asim K. Ray 4 1

Department of Electrical and Computer Engineering, University of Saskatchewan, 57 Campus Drive, Saskatoon, Canada Department of Electrical & Computer Engineering, National University of Singapore, Kent Ridge, Singapore College of Engineering, IT and Environment, Purple 12, Charles Darwin University, Ellengowan Drive, Darwin, Australia 4 Department of Electrical & Computer Engineering, Brunel University London, Kingston Lane, Uxbridge, UK 2 3

CHAPTER MENU Introduction, 1 Optical Constants n and K, 2 Refractive Index and Dispersion, 7 The Swanepoel Technique: Measurement of n and 𝛼 for Thin Films on Substrates, 16 Transmittance and Reflectance of a Partially Transparent Plate, 25 Optical Properties and Diffuse Reflection: Schuster–Kubelka–Munk Theory, 27 Conclusions, 31 References, 32

1.1 Introduction Optical properties of a material change or affect the characteristics of light passing through it by modifying its propagation vector or intensity. Two of the most important optical parameters are the refractive index n and the extinction coefficient K, which are generically called optical constants, although some authors include other optical coefficients within this terminology. The latter is related to the attenuation or absorption coefficient 𝛼. In Part I, in this chapter, we present the complex refractive index, the frequency or wavelength dependence of n and K, so-called dispersion relations, how n and K are inter-related, and how n and K can be determined by studying the transmission as a function of wavelength through a thin film of the material. Physical insights into n and K are provided in Part II (Chapter 2). In addition, there has been a strong research interest in characterizing the optical properties of inhomogeneous media, such as porous media, in which both light absorption and scattering take place so that the reflectance is not specular but diffuse. The latter problem is now included in this second edition. The optical properties of various materials, with n and K being the most important, are available in the literature in one form or another, either published in journals, books, and handbooks, or posted on websites of various researchers, organizations (e.g. NIST), or companies (e.g. Schott Glass). Nonetheless, the reader is referred to the Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

2

1 Fundamental Optical Properties of Materials I

works of Greenway and Harbeke [1], Wolfe [2], Klocek [3], Palik [4, 5], Ward [6], Efimov [7], Palik and Ghosh [8], Nikogosyan [9], and Weaver and Frederikse [10] for the optical properties of a wide range of materials. Adachi’s books on the optical constants of semiconductors are highly recommended [11–13], along with Madelung’s third edition of Semiconductors: Data Handbook [14]. There are, of course, other books and handbooks that also contain optical constants in various chapters; see, for example, references [15–20]. There are also various books that describe optical properties of solids at the senior undergraduate and introductory graduate levels, such as those by Tanner [21], Jimenez and Tomm [22], Stenzel [23], Fox [24], Simmons and Potter [25], Toyozawa [26], Wooten [27], and Abeles [28], which are highly recommended. A number of experimental techniques are available for measuring n and K, some of which have been summarized by Simmons and Potter [25]. For example, ellipsometry measures changes in the polarization of light incident on a sample to sensitively characterize surfaces and thin films (see Chapter 23 in this volume). The interaction of incident polarized light with the sample causes a polarization change in the light, which may then be measured by analyzing the light reflected from the sample. Collins has also provided an extensive in-depth review of ellipsometry for optical measurements [29]. One of the most popular and convenient optical experiments involves a monochromatic light passing through a thin sample, and measuring the transmitted intensity as a function of wavelength, T(𝜆), using a simple spectrophotometer. For thin samples on a thick transparent substrate, the transmission spectrum shows oscillations in T(𝜆) with the wavelength due to interferences within the thin film. Swanepoel’s technique uses the T(𝜆) measurement to determine n and K, as described in Section 1.4.

1.2 Optical Constants n and K One of the most important optical constants of a material is its refractive index, which in general depends on the wavelength of the electromagnetic (EM) wave, through a relationship called dispersion. In materials where an EM wave loses its energy during its propagation, the refractive index becomes complex. The real part is usually the refractive index, n, and the imaginary part is called the extinction coefficient, K. In this section, the refractive index and extinction coefficient will be presented in detail, along with some common dispersion relations. A more practical and a semiquantitative approach is taken along the lines in [30] rather than a full dedication to rigor and mathematical derivations. More analytical approaches can be found in other texts, such as [25, 26]. 1.2.1

Refractive Index and Extinction Coefficient

The refractive index of an optical or dielectric medium, n, is the ratio of the velocity of light c in vacuum to its velocity v in the medium; n = c/v. Using this and Maxwell’s equations, one obtains the well-known Maxwell’s formula for the refractive index of a √ substance as n = 𝜀r 𝜇r , where 𝜀r is the static dielectric constant or relative permittivity and 𝜇r the relative magnetic permeability of the medium. As 𝜇r = 1 for nonmagnetic √ substances, one gets n = 𝜀r , which is very useful in relating the dielectric properties to optical properties of materials at any particular frequency of interest. As 𝜀r depends on the wavelength of light, the refractive index also depends on the wavelength of light, and

1.2 Optical Constants n and K

this dependence is called dispersion. In addition to dispersion, an EM wave propagating through a lossy medium experiences attenuation, which means it loses its energy, due to various loss mechanisms such as the generation of phonons (lattice waves), photogeneration, free carrier absorption, scattering, etc. In such materials, the refractive index becomes a complex function of the frequency of the light wave. The complex refractive index in this chapter is denoted by n*, with real part n, and imaginary part K, called the extinction coefficient, is related to the complex relative permittivity, 𝜀r = 𝜀′r + i𝜀′′r , by, √ √ (1.1a) n∗ = n + iK = 𝜀r = 𝜀′r + i𝜀′′r where 𝜀′r and 𝜀′′r are, respectively, the real and imaginary parts of 𝜀r . Eq. (1.1a) gives: n2 − K 2 = 𝜀′r and 2nK = 𝜀′′r . In explicit terms, n and K can be obtained as √ ′′2 1∕2 n = (1∕ 2)[(𝜀′2 + 𝜀′r ]1∕2 r + 𝜀r ) √ ′′2 1∕2 K = (1∕ 2)[(𝜀′2 − 𝜀′r ]1∕2 r + 𝜀r )

(1.1b)

(1.2a) (1.2b)

Some books (particularly in electrical engineering) use 𝜀r = 𝜀′r − i𝜀′′r and n* = n − iK instead of 𝜀r = 𝜀′r + i𝜀′′r and n* = n + iK. The preference lies in what was assumed for the propagating electric field, whether it is represented by expi(𝜔t − kx) or expi(kx − 𝜔t), where k is the propagation constant. In a lossy medium, the imaginary part of√ n* must ′′ ′ lead to a traveling wave whose amplitude decays. Notice that, for 𝜀r ≪ 𝜀r , n = 𝜀′r and K = 𝜀′′r ∕2n—that is, the refractive index is essentially determined by the real part of 𝜀r and K is determined by the imaginary part of 𝜀r , which is known to represent losses in a dielectric medium. The extinction coefficient K represents loss from the energy carried by the propagating EM wave by conveniently including this loss as the imaginary part in the complex refractive index. The optical attenuation coefficient 𝛼 gauges the rate of this loss from the propagating EM wave. In the absence of scattering, the attenuation would be due to absorption within the medium. For an EM wave that is propagating along x with an intensity I, 𝛼 is defined by dI (1.3) 𝛼=− Idx We can relate 𝛼 and K quite easily by taking a plane wave traveling along x for which the electric field in the wave propagates as E = Eo expi(kx − 𝜔t), where Eo is a constant, 𝜔 is the angular frequency and k is the complex propagation constant in the medium, related to n* by its definition k = n*𝜔/c = (n + iK)(𝜔/c). In free space k = k o = 𝜔/c = 2𝜋/𝜆, where 𝜆 is the free space wavelength. We can substitute for n* and then use I is proportional to |E|2 to find I ∝ exp[−2(𝜔/c)Kx)]—that is, I decays exponentially with the distance propagated. We can substitute for I in (1.3) to find 2𝜔 K (1.4) 𝛼= c The optical constants n and K can be determined by measuring the reflectance from the surface of a material as a function of polarization and the angle of incidence. For normal incidence, the reflection coefficient, r, is obtained as 1 − n∗ 1 − n − iK r= = (1.5) 1 + n∗ 1 + n + iK

3

4

1 Fundamental Optical Properties of Materials I

The reflectance R is then defined by: | 1 − n − iK |2 (1 − n)2 + K 2 | = R = |r|2 = || . (1.6) | (1 + n)2 + K 2 | 1 + n + iK | Notice that whenever K is large, for example, over a range of wavelengths, the absorption is strong, and the reflectance is almost unity. The light is then reflected, and any light in the medium is highly attenuated (typical sample calculations may be found in [24, 30]). Optical properties of materials are typically presented either by showing the frequency dependences (dispersion relations) of n and K or 𝜀′r and 𝜀′′r . An intuitive guide to explaining dispersion in insulators is based on a single oscillator model in which the electric field in the light induces forced dipole oscillations in the material (displaces the electron shells in an atom to oscillate about the positive nucleus) with a single resonant frequency 𝜔o . The frequency dependences of 𝜀′r and 𝜀′′r are then obtained as: 𝜀′r = 1 +

Nat ′ N 𝛼 and 𝜀′′r = 1 + at 𝛼e′′ , 𝜀o e 𝜀o

(1.7)

where N at is the number of atoms per unit volume, 𝜀o is the vacuum permittivity, and 𝛼e′ and 𝛼e′′ are, respectively, the real and imaginary parts of the electronic polarizability, given respectively by: 𝛼e′ = 𝛼eo

1 − (𝜔∕𝜔o )2 [1 − (𝜔∕𝜔o )2 ]2 + (𝛾∕𝜔o )2 (𝜔∕𝜔o )2

(1.8a)

(𝛾∕𝜔o )(𝜔∕𝜔o ) [1 − (𝜔∕𝜔o )2 ]2 + (𝛾∕𝜔o )2 (𝜔∕𝜔o )2

(1.8b)

and 𝛼e′′ = 𝛼eo

where 𝛼 eo is the DC polarizability corresponding to 𝜔 = 0 and 𝛾 is the loss coefficient that characterizes the EM wave losses within the material system. Using Eqs. (1.1)–(1.2) and (1.7)–(1.8), the frequency dependence of n and K can be studied. Figure 1.1a shows the dependence of n and K on the normalized frequency 𝜔/𝜔o for a simple single electronic dipole oscillator of resonance frequency 𝜔o . (a)

n, K Complex refractive index 8

Figure 1.1 Refractive index n and extinction coefficient K obtained from a single electronic dipole oscillator model. (a) n and K versus normalized frequency, and (b) reflectance versus normalized frequency.

6 4

γ/ωo = 0.1

n K

2 0 0 R

(b)

n

K

1

1

εDC1/2 = 3

2

3

2

3

ω/ωo

Reflectance

0.5 0 0

1

ω/ωo

1.2 Optical Constants n and K

It is seen from Figure 1.1 that √ n and K peak close to 𝜔 = 𝜔o . If a material has a 𝜀′′r ′ ′′ ≫ 𝜀r , then 𝜀r ≈ i𝜀r , and n ≈ K ≈ 𝜀′′r ∕2 is obtained from Eq. (1.1b). Figure 1.1b shows the dependence of the reflectance R on the frequency. It is observed that R reaches its maximum value at a frequency slightly above 𝜔 = 𝜔o , and then remains high until 𝜔 reaches nearly 3𝜔o ; thus, the reflectance is substantial while absorption is strong. The normal dispersion region is the frequency range below 𝜔o , where n falls as the frequency decreases; that is, n decreases as the wavelength 𝜆 increases. Anomalous dispersion region is the frequency range above 𝜔o where n decreases as 𝜔 increases. Below 𝜔o , K is small and, if 𝜀DC is 𝜀r (0), the DC permittivity, then 𝜔2o

; 𝜔 < 𝜔o . (1.9) 𝜔2o − 𝜔2 Since, 𝜆 = 2𝜋c/𝜔, defining 𝜆o = 2𝜋c/𝜔o as the resonance wavelength, one gets: n2 ≈ 1 + (𝜀DC − 1)

𝜆2 ; 𝜆 > 𝜆o . (1.10) − 𝜆2o While intuitively useful, the dispersion relations in Eq. (1.8) are far too simple. More rigorously, we have to consider the dipole oscillator quantum mechanically, which means a photon excites the oscillator to a higher energy level—see, for example, Fox [24] or Simmons and Potter [25]. The result is that we would have a series of 𝜆2 /(𝜆2 − 𝜆i 2 ) terms with various weighting factors Ai that add to unity, where 𝜆i represent different resonance wavelengths. The weighting factors Ai involve quantum mechanical matrix elements. Figure 1.2 shows the complex relative permittivity and the complex refractive index of crystalline silicon in terms of photon energy h𝜈 [31, 32]. For photon energies below the bandgap energy (1.1 eV), both 𝜀′′r and K are negligible and n is close to 3.7. Both 𝜀′′r and K increase and change strongly as the photon energy becomes greater than 3 eV, far beyond the bandgap energy. Notice that both 𝜀′′r and K peak at h𝜈 ≈ 3.5 eV, which corresponds to a direct photoexcitation processes, electrons excited directly from the valence band to the conduction band, as discussed in Chapter 2. n2 ≈ 1 + (𝜀DC − 1)

1.2.2

𝜆2

n and K, and Kramers–Kronig Relations

If we know the frequency dependence of the real part, 𝜀′r , of the relative permittivity of a material, we can, using the Kramers–Kronig relations between the real and the imaginary parts, determine the frequency dependence of the imaginary part 𝜀′′r , and vice versa. The transform requires that we know the frequency dependence of either the real or imaginary part over as wide a range of frequencies as possible, ideally from zero (DC) to infinity, and that the material has linear behavior, that is, it has a relative permittivity that is independent of the applied field. The Kramers–Kronig relations for the relative permittivity 𝜀r = 𝜀′r + i𝜀′′r are given by [33–35] (see also Appendix 1C in [25] as well as [27]) ∞ 𝜔′ 𝜀′′ r (𝜔′ ) ′ 2 d𝜔 (1.11a) 𝜀′r (𝜔) = 1 + P 𝜋 ∫ 0 𝜔′ 2 − 𝜔2 and 𝜀′′r (𝜔) = −

2𝜔 P 𝜋 ∫0



𝜀′r (𝜔′ ) − 1 𝜔′ 2 − 𝜔2

d𝜔′

(1.11b)

5

6

1 Fundamental Optical Properties of Materials I

50 40

8

ε r″

30

Real

Imaginary

7 6

20

εr′

10

n

4

εr″

0

Real

5 Imaginary

3 εr′

–10

2

K

1

–20

K n

0 1.5

2

3 4 5 Photon energy (ћω) (a)

6

0

2

4 6 8 Photon energy (ћω) (b)

10

Figure 1.2 (a) Complex relative permittivity of a silicon crystal as a function of photon energy plotted in terms of real (𝜀′r ) and imaginary (𝜀′′r ) parts. (b) Optical properties of a silicon crystal vs. photon energy in terms of real (n) and imaginary (K) parts of the complex refractive index. Source: Adapted from D. E. Aspnes and A. A. Studna, 1983 [32] and H.R. Philipp and E.A. Taft, 1960 [31].

where 𝜔′ is the integration variable, P represents the Cauchy principal value of the integral, and the singularity at 𝜔 = 𝜔′ is avoided. Similarly, one can relate the real and imaginary parts of the polarizability, 𝛼 ′ (𝜔) and ′′ 𝛼 (𝜔), and those of the complex refractive index, n(𝜔) and K(𝜔), as well. For a complex refractive index written as n* = n(𝜔) + iK(𝜔), n(𝜔) = 1 +

2 P 𝜋 ∫0



∞ 𝜔′ K(𝜔′ ) ′ n(𝜔′ ) − 1 ′ 2 d𝜔 and K(𝜔) = − d𝜔 P 2 ′ 2 𝜋 ∫ 0 𝜔′ 2 − 𝜔2 𝜔 −𝜔

(1.12)

Although it appears, in theory, that one needs to integrate the spectrum of n or K from DC to infinite frequencies, this is obviously not feasible, and is unnecessary. It should be noted that the experimental setup usually has low- and high-frequency limitations that truncate the preceding integrations. Moreover, in many cases, we are interested in the spectrum of n and K in and around an absorption band. Thus, before and after the absorption frequency range, K would be negligibly small, and we can use this absorption frequency range in the preceding integrals in Eq. (1.12). There are numerous studies in the literature that use the preceding Kramers–Kronig relations in extracting the wavelength dependence of n from that of K, and vice versa, especially around clear absorption bands; a few selected examples can be found in [36–40], and there are many others in the literature. There are also several useful approaches in which the absorption spectrum, or K(𝜔), is described in terms of a particular physical model with a particular expression, and the corresponding refractive index n(𝜔) is derived from the Kramers–Kronig transformation for both amorphous and crystalline solids—for examples, see [41, 42]. It should be emphasized that the optical constants n and K have to obey what are called f-sum rules [43]. For example, the integration of [n(𝜔) – 1] over all frequencies must be zero, and the integration of 𝜔K(𝜔) over all frequencies gives (𝜋/2)𝜔p 2 , where 𝜔p = ℏ(4𝜋NZe2 /me )1/2 is the free electron plasma frequency in which N is the atomic

1.3 Refractive Index and Dispersion

concentration, Z is the total number of electrons per atom, and e and me are the charge and mass of the electron, respectively. The f -sum rules provide a consistency check and enable various constants to be interrelated.

1.3 Refractive Index and Dispersion There are several popular models describing the spectral dependence of refractive index n in a material. Most of these are described in the following text, although some, such as the infrared refractive index, is covered in the discussion on Reststrahlen absorption in Part II, since it is closely related to the coupling of the EM wave to lattice vibrations. The most popular dispersion relation in optical materials is probably the Sellmeier relationship, since one can sum any number of resonance-type terms to get as wide a range of wavelength dependence as possible. However, its main drawback is that it does not accurately represent the refractive index when there is a contribution arising from free carriers in narrow bandgap or doped semiconductors. There are many handbooks, books, and websites that now provide empirical equations for the refractive index of a wide range of solids, for example as in references [1–19, 44]. 1.3.1

Cauchy Dispersion Relation

In the Cauchy relationship, the dispersion relationship between the refractive index (n) and the wavelength of light (𝜆) is commonly stated in the following form: C B + 4 (1.13) 2 𝜆 𝜆 where A, B, and C are material-dependent specific constants. Equation (1.13) is known as Cauchy’s formula; it is typically used in the visible spectrum region for various optical glasses, and it applies to normal dispersion, when n decreases with increasing 𝜆 [45, 46]. The third term is sometimes dropped for a simpler representation of n versus 𝜆 behavior. The original expression was a series in terms of the wavelength, 𝜆, or frequency, 𝜔, or photon energy ℏ𝜔 of light as: n=A+

n = a0 + a2 𝜆−2 + a4 𝜆−4 + a6 𝜆−6 + … 𝜆 > 𝜆th ,

(1.14a)

n = n0 + n2 (ℏ𝜔)2 + n4 (ℏ𝜔)4 + n6 (ℏ𝜔)6 + … ℏ𝜔 < ℏ𝜔th ,

(1.14b)

or where ℏ𝜔 is the photon energy; ℏ𝜔th = hc/𝜆th is the optical excitation threshold (e.g. bandgap energy); and a0 , a2 ,… and n0 , n2 ,… are constants. It has been found that a Cauchy relation in the following form [47]: n = n−2 (ℏ𝜔)−2 + n0 + n2 (ℏ𝜔)2 + n4 (ℏ𝜔)4 ,

(1.15)

can be used satisfactorily over a wide range of photon energies. The dispersion parameters of Eq. (1.15) are listed in Table 1.1 for a few selected materials over specific photon energy ranges. Cauchy’s dispersion relations given in Eqs. (1.13)–(1.14) were originally called the elastic ether theory of the refractive index. It has been widely used for many materials,

7

8

1 Fundamental Optical Properties of Materials I

Table 1.1 Cauchy’s dispersion parameters of Eq. (1.15) for Ge, Si, and Diamond from [43]. n2 (eV−2 )

n4 (eV−4 )

2.378

0.00801

0.000104

3.4189

0.0815

0.0125

4.0030

0.220

0.140

Material

ℏ𝝎(eV) Min

ℏ𝝎(eV) Max

n−2 (eV2 )

n0

Diamond

0.0500

5.4700

−1.07 × 10−5

Si

0.0020

1.08

−2.04 × 10−8

0.75

−8

Ge

0.0020

−1.00 × 10

although, in recent years, many researchers have preferred to use the Sellmeier equation, described in the following text. 1.3.2

Sellmeier Equation

The Sellmeier equation [48] is an empirical relation between the refractive index n of a substance and wavelength 𝜆 of light in the form of a series of single dipole oscillator terms, each of which has the usual 𝜆2 /(𝜆2 − 𝜆i 2 ) dependence as in n2 = 1 +

A1 𝜆2 𝜆2 − 𝜆21

+

A2 𝜆2 𝜆2 − 𝜆22

+

A3 𝜆2

(1.16)

𝜆2 − 𝜆23

where A1 , A2 , A3 and 𝜆1 , 𝜆2 , and 𝜆3 are constants, called Sellmeier coefficients, which are determined by fitting this expression to the experimental data. The actual Sellmeier formula is more complicated. It has more terms of similar form, such as Ai 𝜆2 /(𝜆2 – 𝜆i 2 ), where i = 4, 5, ..., but these can generally be neglected in representing n vs. 𝜆 behavior over typical wavelengths of interest and by ensuring that the three terms included in Eq. (1.16) correspond to the most important or relevant terms in the summation [49]. The Sellmeier coefficients for some materials, including pure Silica (SiO2 ) and 86.5 mol% SiO2 –13.5 mol% GeO2 , are given in Table 1.2 as examples. A quantitative analysis of the application of the Sellmeier dispersion relation to a range of materials, from glasses to semiconductors, has been discussed by Tatian [49]. There are two methods for determining the refractive index of silica–germania glass (SiO2 )1-x (GeO2 )x . The first is a simple, but approximate, linear interpolation of the refractive index between known compositions, for example, n(x) − n(0.135) = (x − 0.235)[n(0.135) − n(0)]/0.135, where n(x) is for (SiO2 )1−x (GeO2 )x ; n(0.135) is for 86.5 mol% SiO2 –13.5 mol% GeO2 ; and n(0) is for SiO2 . The second is an interpolation for coefficients Ai and 𝜆i between SiO2 and GeO2 as [50]: n2 − 1 =

{A1 (S) + X[A1 (G) − A1 (S)]}𝜆2 𝜆2 − {𝜆1 (S) + X[𝜆1 (G) − 𝜆1 (S)]}𝜆21

+ …,

(1.17)

where S and G in parentheses refer to silica and germania, respectively. The theoretical basis of the Sellmeier equation lies in representing the solid as a sum of N lossless (frictionless) Lorentz oscillators such that each has the usual form of 𝜆2 /(𝜆2 – 𝜆i 2 ) with different 𝜆i and each has a different strength, or weighting factor; Ai , i = 1 to N [51, 52]. Such dispersion relationships are essential in designing photonic devices such as waveguides. (Note that although Ai weighs different Lorentz contributions, they do not sum to 1 since they include other parameters besides the oscillator strength f i .) The refractive indices of most optical glasses have been extensively modeled by the Sellmeier equation.

1.3 Refractive Index and Dispersion

Table 1.2 Sellmeier coefficients of a few materials, where 𝜆1 , 𝜆2 , 𝜆3 are in μm. Material

A1

A2

A3

𝝀1

𝝀2

𝝀3

SiO2 (fused silica)

0.696749

0.408218

0.890815

0.0690660

0.115662

9.900559

86.5% SiO2 –13.5% GeO2

0.711040

0.451885

0.704048

0.0642700

0.129408

9.425478

GeO2

0.80686642 0.71815848 0.85416831 0.068972606

0.15396605 11.841931

Barium fluoride

0.3356

0.506762

3.8261

0.057789

0.109681

46.38642

Sapphire

1.023798

1.058264

5.280792

0.0614482

0.110700

17.92656

Diamond

0.3306

4.3356

0.175

0.106

Quartz, no

1.35400

0.010

0.9994

0.092612

10.700 11.310

Quartz, ne

1.38100

0.0100

0.9992

0.093505

KTP, no

1.2540

0.0100

0.0992

0.09646

6.9777

KTP, ne

1.13000

0.0001

0.9999

0.09351

7.6710

9.8500 9.5280 5.9848 12.170

Source: From various sources.

Various optical glass manufacturers such as Schott Glass normally provide the Sellmeier coefficients for their glasses [53]. The optical dispersion relations for glasses have been discussed by a number of authors [7, 25, 54]. There are other Sellmeier–Cauchy-like dispersion relationships that inherently take account of various contributions to the optical properties, such as the electronic and ionic polarization and the interaction of photons with free electrons. For example, for many semiconductors and ionic crystals, two useful dispersion relations are, n2 = A +

B𝜆2 D𝜆2 + 2 , −C 𝜆 −E

(1.18)

𝜆2

and n2 = A +

𝜆2

B C + + D𝜆2 + E𝜆4 , 2 2 − 𝜆o (𝜆 − 𝜆2o )2

(1.19)

where A, B, C, D, E, and 𝜆o are constants particular to a given material. Eq. (1.18) is equivalent to the Sellmeier equation. Eq. (1.19) is known as the Herzberger dispersion relation [52]. Table 1.3 provides a few examples. Both Cauchy and Sellmeier equations are strictly applicable in wavelength regions where the material is transparent, that is, the extinction coefficient is relatively small. The refractive index dispersion relations Table 1.3 Parameters of Eq. (1.19) for some selected materials. D (𝛍m)−2

Material

𝝀o (𝛍m)

A

B (𝛍m)2

C (𝛍m)4

Silicon

0.028

3.41983

0.159906

−0.123109

MgO LiF AgCl

0.11951 0.16733 0.21413

2.95636 1.38761 4.00804

0.021958 0.001796 0.079009

0

1.269 × 10−6

−4.1 × 10

−1.951 × 10−9

−2

−2.05 × 10−5

−3

−5.57 × 10−6

−4

−1.976 × 10−7

−1.0624 × 10 −5

0

E (𝛍m)−4

−2.3045 × 10

−8.5111 × 10

Source: Si data from D.F. Edwards and E. Ochoa, Appl. Optics 19, 4130 (1980), others from W. L. Wolfe, The Handbook of Optics, W.G. Driscoll and W. Vaughan, McGraw-Hill, New York, 1978.

9

10

1 Fundamental Optical Properties of Materials I

for a wide range of semiconductors have been compiled by Madelung in [14]. There are many application-based articles in the literature that provide empirical dispersion relations for a variety of materials; a recent example on far infrared substrates (Ge, Si, ZnSe, ZnS, ZnTe) is given in reference [55]. There are both websites and various journal articles in the literature that give the refractive index of numerous materials as a function of wavelength.

1.3.3 1.3.3.1

Refractive Index of Semiconductors Refractive Index of Crystalline Semiconductors

A particular interest in the case of semiconductors is in n and K for photon energies greater than the bandgap Eg for optoelectronics applications. Due to various features and singularities in the E-k diagrams of crystalline semiconductors, the optical constants n and K for ℏ𝜔 > Eg are not readily expressible in simple terms. Various authors, for example, Forouhi and Bloomer [42, 56] and Chen et al. [57], have nonetheless provided useful and tractable expressions for modeling n and K in this regime. In particular, Forouhi–Bloomer (FB) equations express n and K in terms of the photon energy ℏ𝜔 in a consistent way that obey the Kramers–Kronig relations [42], that is K=

q ∑

Ai (ℏ𝜔 − Eg )2

i=1

(ℏ𝜔)2 − Bi (ℏ𝜔) + Ci

and n = n(∞) +

q ∑ i=1

Boi (ℏ𝜔) + Coi , (ℏ𝜔)2 − Bi (ℏ𝜔) + Ci

(1.20)

where (ℏ𝜔) is the photon energy; q is an integer that represents the number of terms needed to suitably model experimental n, K; Eg is the bandgap and Ai , Bi , C i , Boi , C oi are constants; Boi and C oi depend on Ai , Bi , C i , and Eg —only the latter four are independent parameters; and Boi = (Ai /Qi )[−(1/2)Bi 2 + Eg Bi – Eg 2 + C i ], C oi = (Ai /Qi )[(1/2)(Eg 2 + C i ) Bi − 2Eg C i ], and Qi = (1/2)(4C i − Bi 2 )1/2 . Forouhi and Bloomer provide a table of FB coefficients, Ai , Bi , C i , and Eg for four terms in the summation in Eq. (1.20) [42] for a number of semiconductors; an example that shows an excellent agreement between the FB dispersion relation and the experimental data is shown in Figure 1.3. Table 1.4 provides the FB coefficients for a few selected semiconductors. Other useful theoretical or somewhat semiempirical dispersion relationships have also been proposed, for example, by Afromowitz [58], Adachi [59–63], Campi and Papuzza [64], and others [65]. These models have been applied to various semiconductors and their alloys with relative success over certain photon energy ranges. One of the useful and straightforward approaches to modeling the dispersion has been based on writing the complex relative permittivity 𝜀r (ℏ𝜔) as a finite sum of a number of damped harmonic oscillators (the so-called harmonic oscillator approximation), and fitting this expression to the experimental data as in references [66, 67], even though many terms may be needed and the curve fit process has to be carefully chosen to ensure a reliable representation of the data. One of best models considered so far, however, has involved parametric modeling [68–70], in which not only a sum of harmonic oscillators are used but also Gaussian broadened polynomials to represent the dispersion of the complex relative permittivity, and hence n and K.

1.3 Refractive Index and Dispersion

5 SiC 4 n

3 n, K 2

1

K

0 0

2

4

6 8 10 Photon energy (eV)

12

14

Figure 1.3 n and K versus photon energy for crystalline SiC. The solid line is obtained from the FB equation with four terms with appropriate parameters, and the points represent the experimental data. See original reference [42] for the data and details. Source: Reprinted with permission, from Figure 2c, A.R. Forouhi and I. Bloomer, Phys. Rev. B, 38, 1865. Copyright (1988) by the American Physical Society.

1.3.3.2

Bandgap and Temperature Dependence

The refractive index of a semiconductor (typically for ℏ𝜔 < Eg ) typically decreases with increasing bandgap Eg . There are various empirical and semi-empirical rules and expressions that relate n to Eg . Based on an atomic model, Moss has suggested that n and Eg are related by n4 Eg = K = constant [71, 72] (K is about ∼100 eV). In the Hervé–Vandamme relationship [73], ( )2 A 2 , (1.21) n =1+ Eg + B where A and B are constants as A ≈ 13.6 eV and B ≈ 3.4 eV. The temperature dependence of n arises from the variation of Eg with the temperature T, and it typically increases with increasing temperature. The temperature coefficient of refractive index (TCRI) of semiconductors can be found from the Hervé–Vandamme relationship as: [ ] (n2 − 1)3∕2 dEg dB 1 dn =− + (1.22) TCRI = • n dT 13.6n2 dT dT where dB/dT ≈ 2.5 × 10−5 eV K−1 . TCRI is typically found to be positive (n increasing with temperature) and in the range of 10−6 to 10−4 K−1 . Although Eqs. (1.21) and (1.22) are popular, there are other very useful empirical and semiempirical relationships (see, e.g., [74]), some of which are summarized in Table 1.5 with appropriate references. 1.3.4

Refractive Index of Glasses

The Sellmeier equation with three terms have been found to represent the dispersion of n reasonably well for most glasses and ceramics. The coefficients in the Sellmeier equation have been listed at several websites [44] and handbooks.

11

12

1 Fundamental Optical Properties of Materials I

Table 1.4 FB coefficients for selected semiconductors [42] for four terms (i = 1 to 4). Bi (eV)

C i (eV2 )

n(∞)

E g (eV)

0.00405

6.885

11.864

1.950

1.06

0.01427

7.401

13.754

0.06830

8.634

18.812

0.17488

10.652

29.841

0.08556

4.589

5.382

2.046

0.60

0.21882

6.505

11.486

0.02563

8.712

19.126

0.07754

10.982

31.620

0.00652

7.469

13.958

2.070

2.17

0.14427

7.684

15.041

0.13969

10.237

26.567

0.00548

13.775

47.612

0.00041

5.871

8.619

2.156

1.35

0.20049

6.154

9.784

0.09688

9.679

23.803 44.119 1.914

0.65

1.766

1.27

1.691

0.30

1.803

0.12

Ai

Si

Ge

GaP

GaAs

GaSb

InP

InAs

InSb

0.01008

13.232

0.00268

4.127

4.267

0.34046

4.664

5.930

0.08611

8.162

17.031

0.02692

11.146

31.691

0.20242

6.311

10.357

0.02339

9.662

23.472

0.03073

10.726

29.360

0.04404

13.604

47.602

0.18463

5.277

7.504

0.00941

9.130

20.934

0.05242

9.865

25.172

0.03467

13.956

50.062

0.00296

3.741

3.510

0.22174

4.429

5.447

0.06076

7.881

15.887

0.04537

10.765

30.119

Note: First entry in the box is for i = 1, and the fourth is for i = 4.

Gladstone–Dale formula [82, 83] is an empirical equation that allows the average refractive index n of an oxide glass to be calculated from its density 𝜌 and its constituents as: ∑ n−1 pi ki = CGD , = p1 k1 + p2 k2 + · · · = 𝜌 i=1 N

(1.23)

1.3 Refractive Index and Dispersion

Table 1.5 Various selected simple relationships proposed between n and the bandgap E g . Relationship

Comment/Reference

n4 Eg = K; K = constant

Widely used, but has limitations. K ≈ 95 eV; 173 eV for Group IV elements [75]; K = 108 eV [76]. For a theoretical derivation and discussion of Moss’s rule, see [77].

)2

( 2

n =1+

A Eg + B

A ≈ 13.6 eV and B ≈ 3.4 eV. Hervé–Vandamme relationship. See text.

n = 4.084 + 𝛽Eg

𝛽 = −0.62 eV−1 . Proposed by Ravindra et al. [76], but has serious limitations for small and large n.

n = − ln(0.027Eg )

The Reddy Equation [78]. Based on the Duffy relationship Eg = 3.72Δ𝜒 op [79] and Δ𝜒 op = 9.8exp(−n) in [80], where Δ𝜒 op is optical electronegativity.

12.417 n2 = √ Eg − 0.365

See [81]. Equivalent to n4 (Eg − 0.365 eV) = 154. Similar to the Moss relation. Eg > 0.365 eV

Note: Eg is in eV.

where the summation is for various oxide components (each a simple oxide), pi is the weight fraction of the i-th oxide in the compound, and k i is the refraction coefficient that represents the polarizability of the i-th oxide. The right-hand side of Eq. (1.23) is called the Gladstone–Dale coefficient C GD . In more general terms, as a mixture rule for the overall refractive index, the Gladstone–Dale formula is frequently written as: n −1 n − 1 n1 − 1 w1 + 2 w2 + · · · , = 𝜌 𝜌1 𝜌2

(1.24)

where n and 𝜌 are the effective refractive index and effective density, respectively, of the whole mixture; n1 , n2 ,… are the refractive indices of the constituents; 𝜌1 , 𝜌2 ,… represent the density of each constituent; and w1 , w2 … are the weight fractions of the constituents. Gladstone–Dale equations for the polymorphs of SiO2 and TiO2 give the average n, respectively, as: n(SiO2 ) = 1 + 0.21𝜌 and n(TiO2 ) = 1 + 0.40𝜌

(1.25)

It is generally assumed that the refractive index can be related to the polarizability 𝛼 through the well-known Lorentz–Lorenz equation (equivalent to the Clausius–Mossotti equation in dielectrics), which involves the local field as n2 − 1 1 ∑ N𝛼 (1.26) = 2 n + 2 3𝜀o j j j in which 𝛼 j is the polarizability of a given type (species) (j) of atoms in the structure, and N j is the concentration of this species of atoms. Equation (1.26) includes both electronic and ionic polarizability and assumes that the local field is the Lorentz field, which depends on polarization through P/3𝜀o in the standard model for cubic crystals and noncrystalline solids. Ritland [84] assumed that the local field depends on the polarization as bP/𝜀o , where b is a numerical constant, and reformulated Eq. (1.26) in SI units

13

14

1 Fundamental Optical Properties of Materials I

as n2

n2 − 1 b ∑ N𝛼 = − 1 + (1∕b) 𝜀o j j j

(1.27)

and b is kept as a variable fitting parameter to the experimental data. Obviously, b = 1/3 is the usual Lorentz–Lorenz equation, but the best fits do not necessarily lead to 1/3 [84]. Most recent work on the refractive index of glasses has invariably used the Sellmeier equation, given its excellent fit to the dispersion data on glasses, as well as the adoption of the Sellmeier equation by some industrial glass manufacturers such as Schott [85]. Further, in some cases, the temperature dependence of the Sellmeier coefficients are also evaluated, so that dn/dT can be determined at different wavelengths [86–90]. While dn/dT is positive for many semiconductors, this is not generally true for glasses. 1.3.5

Wemple–DiDomenico Dispersion Relation

Based on the single oscillator model, the Wemple–DiDomenico (WD) model is a semi-empirical dispersion relation for determining the refractive index at photon energies below the interband absorption edge in a variety of materials [91, 92]. It is given by n2 = 1 +

Eo Ed Eo2 − (h𝜈)2

,

(1.28)

where 𝜈 is the frequency, h is the Planck constant, Eo is the single oscillator energy, and Ed is the dispersion energy, which is a measure of the average strength of interband optical transition. Ed can be written as Ed = 𝛽N c Za N e (eV), where N c is the effective coordination number of the cation nearest-neighbor to the anion (e.g. N c = 6 in NaCl, N c = 4 in Ge), Za is the formal chemical valency of the anion (Za = 1 in NaCl, 2 in Te, and 3 in GaP), N e is the effective number of valence electrons per anion excluding the cores (N e = 8 in NaCl, Ge; 10 in TlCl; 12 in Te; 91 /3 in As2 Se3 ), and 𝛽 is a constant that depends on whether the interatomic bond is ionic (𝛽 i ) or covalent (𝛽 c ): 𝛽 i = 0.26 ± 0.04 eV (e.g. halides NaCl, ThBr, etc., and most oxides, Al2 O3 , etc.), and 𝛽 c = 0.37 ± 0.05 eV (e.g. tetrahedrally bonded AN B8−N zinc blende and diamond type structures, GaP, ZnS, etc., and wurtzite crystals have a 𝛽 that is intermediate between 𝛽 i and 𝛽 c ). Further, empirically, Eo = CEg (D), where Eg (D) is the lowest direct bandgap and C is a constant, typically C ≈ 1.5. Eo has been associated with the main peak in the 𝜀′′r (h𝜈) versus h𝜈 spectrum. The parameters required for calculating n from Eq. (1.28) are listed in Table 1.6 [91]. It should be apparent that one can improve on the single oscillator model by adding a second oscillator term to Eq. (1.28) [93], and so on; although one loses the simplicity of the model, which the experimentalists like. Since its publication in 1971, the WD model has also been successfully applied to various amorphous semiconductors and glasses in addition to crystals, following the original arguments of Wemple [92]. In their original work, WD provided the single oscillator parameters for an extensive list of materials, some of which are summarized in Table 1.6, with further discussion in [94]. The ratio Eo /Eg has been observed by researchers to be very roughly in the range 1.5–2.5. The simple WD model is expected to hold in the small absorption region of the spectrum, that is, away from absorption bands due to interband transitions (h𝜈 < Eg ) or Reststrahlen absorption, etc. While it is apparent that the

1.3 Refractive Index and Dispersion

Table 1.6 Examples of parameters for Wemple–DiDomenico dispersion relationship in various materials. E o (eV) E d (eV) 𝜷 (eV) 𝜷

Material

Nc

Za

NaCl

6

1

8

10.3

13.6

0.28

𝛽i

Halides, LiF, NaF, etc.

CsCl

8

1

8

10.6

17.1

0.27

𝛽i

CsBr, CsI, etc.

TlCl

8

1

10

5.8

20.6

0.26

𝛽i

TlBr

CaF2

8

1

8

15.7

15.9

0.25

𝛽i

BaF2 , etc. Oxides, MgO, TeO2 , etc.

Ne

Comment

CaO

6

2

8

9.9

22.6

0.24

𝛽i

Al2 O3

6

2

8

13.4

27.5

0.29

𝛽i

LiNbO3

6

2

8

6.65

25.9

0.27

𝛽i

TiO2

6

2

8

5.24

25.7

0.27

𝛽i

ZnO

4

2

8

6.4

17.1

0.27

𝛽i

ZnSe

4

2

8

5.54

27

0.42

𝛽c

II-VI, Zinc blende, ZnS, ZnTe, CdTe

GaAs

4

3

8

3.55

33.5

0.35

𝛽c

III-V, Zinc blende, GaP, etc.

Si (Crystal)

4

4

8

4.0

44.4

0.35

𝛽c

Diamond, covalent bonding; C (diamond), Ge, 𝛽-SiC, etc.

SiO2 (Crystal)

4

2

8

13.33

18.10

0.28

𝛽i

Average crystalline form

SiO2 (Amorphous)

4

2

8

13.38

14.71

0.23

𝛽i

Fused silica

CdSe

4

2

8

4.0

20.6

0.32

𝛽 i –𝛽 c

Wurtzite

Note: Values extracted and combined from tables in [91].

WD relation can only be approximate, it has nonetheless found wide acceptance among experimentalists due to its straightforward simplicity. For example, in 2018 alone, it has been applied nearly 200 times to a wide variety of inorganic and organic material systems, particularly to semiconductor films. 1.3.6

Group Index

Group index is a factor by which the group velocity of a group of waves in a dielectric medium is reduced with respect to propagation in free space. It is denoted by N g and defined by N g = vg /c, where vg is the group velocity, defined by vg = d𝜔/dk, where k is the wave vector or the propagation constant. The group index can be determined from the ordinary refractive index n through [95] dn (1.29) d𝜆 where 𝜆 is the wavelength of light (in free space). Figure 1.4 illustrates the relation between N g and n in SiO2 . The group index N g is the quantity that is normally used in calculating dispersion in optical fibers, since it is N g that determines the group velocity of a propagating light pulse in a glass or transparent medium. It should be remarked that, although n vs. 𝜆 can decrease monotonically with 𝜆 over a range of wavelengths, N g can exhibit a minimum in the same range where the dispersion, dN g /d𝜆, becomes zero. The point dN g /d𝜆 = 0 is called the zero-material dispersion wavelength, which is around 1300 nm for silica, as apparent in Figure 1.4. Ng = n − 𝜆

15

16

1 Fundamental Optical Properties of Materials I

Figure 1.4 Refractive index n and the group index Ng of pure SiO2 (silica) glass as a function of wavelength. Source: Adapted from S.O. Kasap, 2017 [30].

1.49 1.48 Ng

1.47 1.46

n 1.45 1.44 500

700

900 1100 1300 1500 1700 1900 Wavelength (nm)

1.4 The Swanepoel Technique: Measurement of n and 𝜶 for Thin Films on Substrates 1.4.1

Uniform Thickness Films

In many instances, the optical constants are conveniently measured by examining the transmission through a thin film of the material deposited on a transparent glass or other (e.g. sapphire) substrate. The classic reference on the optical properties of thin films has been the book by Heavens [96]; the book is still useful in clearly describing what experiments can be carried out, and has a number of useful derivations such as the reflectance and transmittance through thin films in the presence of multiple reflections. Since then, numerous research articles and reviews have been published. Poelman and Smet [97] have critically reviewed how a single transmission spectrum measurement can be used to extract the optical constants of a thin film. In general, the amount of light that gets transmitted through a thin film material depends on the amount of reflection and absorption that takes place along the light path. If the material is a thin film with a moderate absorption coefficient 𝛼, then there will be multiple interferences at the transmitted side of the sample, as illustrated in Figure 1.5. In this case, some interference fringes will be evident in the transmission spectrum obtained from a spectrophotometer, as shown in Figure 1.6. One very useful method that makes use of these interference fringes to determine the optical properties of the material is called the Swanepoel method [98] which is based on earlier works of Manifacier et al. [99] and Hall and Ferguson [100]. Swanepoel has shown that the optical properties of a uniform thin film of thickness d, refractive index n, and absorption coefficient 𝛼, deposited on a thick substrate with a refractive index s, as shown in Figure 1.5, can be obtained from the transmittance T given by T=

Ax B − Cx cos 𝜑 + Dx2

(1.30)

where A = 16n2 s, B = (n + 1)3 (n + s2 ), C = 2(n2 − 1)(n2 − s2 ), D = (n − 1)3 (n − s2 ), 𝜑 = 4𝜋nd/𝜆, x = exp(−𝛼d) is an absorbance-type parameter, and n, s, and 𝛼 are all

1.4 The Swanepoel Technique: Measurement of n and α for Thin Films on Substrates

Incident monochromatic wave

Monochromatic light T=1

nair = 1 nfilm = n

d

Thin film

Thin film n* = n + iK

α>0

Glass substrate α=0

nsubstrate = s

Substrate T Detector

Figure 1.5 Schematic sketch of the typical behavior of light passing through a thin film on a substrate. On the left, oblique incidence is shown to demonstrate the multiple reflections. In most measurements, the incident beam is nearly normal to the film, as shown on the right. Full Transmission Spectrum from an a-Se Thin Film 1

Transmittance

0.7 0.6 0.5 0.4 0.3

TRANSPARENT REGION

0.8

STRONG ABSORPTION REGION

0.9

Interference Fringes

ABSORPTION REGION

0.2 0.1 0

Expt data Substrate 600

800

1000

1200 λ in nm

1400

1600

1800

2000

Figure 1.6 An example of a typical transmission spectrum of a 0.969-μm-thick amorphous Se thin film that has been vacuum coated on a glass substrate held at a substrate temperature of 50∘ C during the deposition.

17

1 Fundamental Optical Properties of Materials I

functions of wavelength 𝜆. Although x (with values of 1 under no absorption and approaching 0 under strong absorption) has been called “absorbance” in most studies using the Swanepoel technique as in the original papers, the usual definition of absorbance, however, is log10 (1/T), so that x, strictly, is not true absorbance. Eq. (1.30) assumes K 2 < < n2 , where n + iK is the complex refractive index of the film and normal incidence. What is very striking and useful is that all the important optical properties can be determined from the application of this equation; this will be introduced in the subsequent paragraphs. Before the optical properties of any thin film can be extracted, the refractive index of their substrate must first be calculated. For a glass substrate with very negligible absorption, that is, K ≤ 0.1 and 𝛼 ≤ 10−2 cm−1 , in the range of operating wavelengths, the refractive index s is, √( ) 1 1 s= + −1 (1.31) Ts Ts2 where T s is the transmittance value measured from the spectrophotometer. This expression can be derived from the transmittance equation for a bulk sample with little attenuation. With this refractive index s known, the next step is to construct two envelopes around the maxima and minima of the interference fringes in the transmission spectrum, as indicated in Figure 1.7. There will altogether be two envelopes that have to be constructed before any of the expressions derived from Eq. (1.30) can be used to extract the optical properties. This Full Transmission Spectrum 1 TS

TM

0.9

0.5 0.4 0.3

Tm Extreme Points

0.2 0.1 0

600

Minima

800

1000

TRANSPARENT REGION

0.6

T

ABSORPTION REGION

0.7

Maxima STRONG ABSORPTION REGION

0.8

Transmittance

18

Expt data Substrate Maxima Envelope Minima Envelope 1200 λ in nm

1400

1600

1800

2000

Figure 1.7 The construction of envelopes in the transmission spectrum of the thin a-Se film in Figure 1.6.

1.4 The Swanepoel Technique: Measurement of n and α for Thin Films on Substrates

can be done by locating all the extreme points of the interference fringes in the transmission spectrum and then making sure that the respective envelopes, T M (𝜆) for the maxima and T m (𝜆) for the minima, pass through these extremes, the maxima and minima, of T(𝜆) tangentially. From Eq. (1.30), it is not difficult to see that, at cos 𝜑 = ± 1, the expressions that describe the two envelopes are, Ax , B − Cx + Dx2 Ax Minima∶Tm = . B + Cx + Dx2

Maxima∶TM =

(1.32a) (1.32b)

Figure 1.7 shows two envelopes constructed for a transmission spectrum of an a-Se thin film. It can also be seen that the transmission spectrum is divided into three special regions according to their transmittance values: (i) the transparent region, where T(𝜆) ≥ 99.99% of the substrate’s transmittance value of T s (𝜆), (ii) the strong absorption region, where T(𝜆) is typical smaller than 20%, and (iii) the absorption region, in between the two latter regions as shown in Figure 1.7. The refractive index of the thin film can be calculated from the two envelopes, T M (𝜆) and T m (𝜆), and the refractive index of the substrate s through [ ] [ ] TM − Tm s2 + 1 2 2 1∕2 1∕2 n = N + (N − s ) ; N = 2s + , (1.33) TM Tm 2 where N is defined by the second equation on the right hand side of Eq. (1.33). T M and T m are assumed to be continuous functions of 𝜆 and x, so that values have to be at the same wavelength for use in Eq. (1.33). The derivation of Eq. (1.33) is based on considering 1/T m − 1/T M , which is 2C/A, and then substituting for C and A from earlier, and then solving for n. Since the equation is not valid in the strong absorption region, where there are no maxima and minima, the calculated refractive index has to be fitted to a well-established dispersion model for extrapolation to shorter wavelengths before it can be used to obtain other optical constants. Usually either the Sellmeier or the Cauchy dispersion equation is used to fit n vs 𝜆 experimental data in this range. Figure 1.8 shows the refractive indices extracted from the envelopes and fitted to the Sellmeier dispersion model with two terms. With the refractive index of the thin film corresponding to two adjacent maxima (or minima) at points 1 and 2, given as n1 at 𝜆1 and n2 at 𝜆2 , the thickness can be easily calculated from the basic interference equation of waves as follows: dcrude =

𝜆 1 𝜆2 2(𝜆1 n2 − 𝜆2 n1 )

(1.34)

where dcrude refers to the thickness obtained from the maxima (minima) at points 1, 2. As other adjacent pairs of maxima or minima points are used, more thickness values can be deduced, and hence an average value calculated. It is assumed the film has an ideal uniform thickness. The absorption coefficient 𝛼 can be obtained once x is extracted from the transmission spectrum. This can be done as follows: 𝛼=−

ln(x) dave

(1.35)

19

1 Fundamental Optical Properties of Materials I

Refractive index of the a-Se film in Figure 1.7 4 Fitted To Sellmeier Equation Crude Calculation From Envelopes Improved Calculation

3.8 3.6 3.4

2.8 2.6 2.4 2.2 2

TRANSPARENT REGION

3

STRONG ABSORPTION REGION

3.2 n

20

ABSORPTION REGION

800

600

1000

1200 λ in nm

1400

1600

1800

2000

Figure 1.8 Determination of the refractive index from the transmission spectrum maxima and minima shown in Figure 1.7. The solid black curve shows the fitted Sellmeier n vs. 𝜆 curve, which follows n2 = 3.096 + 2.943𝜆2 /[𝜆2 − (402.31)2 ], where 𝜆 is in nm.

EM −



2 EM −(n2 −1)3 (n2 −s4 )

2

s where x = ; EM = 8n + (n2 − 1)(n2 − s2 ); and dave is the average (n−1)3 (n−s2 ) TM thickness of dcrude . The accuracy of the thickness, the refractive index, and the absorption coefficient can all be further improved in the following manner. The first step is to determine a new set of interference orders, represented by m′ , for the interference fringes from the basic interference equation of waves; that is,

m′ =

2ne dave , 𝜆e

(1.36a)

where ne and 𝜆e are values taken at any extreme point, and m′ is an integer if the extremes taken are maxima or a half-integer if the extremes taken are minima. The second step is to get a new corresponding set of thickness values, d′ , from this new set of order numbers m′ , by rearranging Eq. (1.36a) as: d′ =

m′ 𝜆e 2ne

(1.36b)

From this new set of thickness values, d′ , a new average thickness, dnew , must be calculated before it can be applied to improve the refractive index. This can be done by

1.4 The Swanepoel Technique: Measurement of n and α for Thin Films on Substrates

ignoring those d′ that have values very different from the rest during averaging. With this new average thickness, a more accurate refractive index ne can be obtained from the same equation, ne =

m′ 𝜆 e 2dnew

(1.36c)

This new refractive index can then be fitted to the previous dispersion model again, so that an improved absorption coefficient 𝛼 can be calculated from Eq. (1.35). All these parameters can then be used in Eq. (1.30) to regenerate a calculated transmission spectrum T cal (𝜆), so that the root mean square error (RMSE) can be determined from the experimental spectrum T exp . The RMSE is calculated as follows: √ √ q √∑ √ (T − T )2 √ exp cal √ i=1 RMSE = , (1.37) q where T exp is the transmittance of the experimental or measured spectrum, T cal is the transmittance of the regenerated spectrum obtained through the Swanepoel calculation method, and q is the range of the measurement. Figure 1.9 shows the regenerated transmission spectrum of the a-Se thin film that appeared in Figure 1.6 using the optical constants calculated from the envelopes (as quoted in the caption of Figure 1.8). Full Spectrum of the a-Se Film in Figure 1.6 by the Swanepoel Method 1

Transmittance

0.7 0.6 0.5 0.4 0.3

TRANSPARENT REGION

0.8

STRONG ABSORPTION REGION

0.9

RMSE = 1.02%

ABSORPTION REGION 0.2

Expt data Regenerated Spectrum by Sellmeier Fitted n Substrate

0.1 0

600

800

1000

1200 1400 λ in nm

1600

Figure 1.9 Regenerated transmission spectrum of the sample in Figure 1.6.

1800

2000

21

22

1 Fundamental Optical Properties of Materials I

1.4.2

Thin Films with Non-uniform Thickness

The Swanepoel technique in the case of a film with a wedge-like cross-section, as shown in Figure 1.10, involves the integration of Eq. (1.30) over the thickness of the film in order for it to more accurately describe the transmission spectrum [101]. The transmittance then becomes, 𝜑

TΔd =

2 1 Ax dx 𝜑2 − 𝜑1 ∫𝜑1 B − Cx cos 𝜑 + Dx2

(1.38)

with 4𝜋n(d − Δd) 4𝜋n(d + Δd) , and 𝜑2 = , 𝜆 𝜆 where A = 16n2 s, B = (n + 1)3 (n + s2 ), C = 2(n2 − 1)(n2 − s2 ), D = (n − 1)3 (n − s2 ), 𝜑 has been defined after Eq. (1.30), x = exp(−𝛼d) corresponds x to the absorbance-type parameter calculated using the average film thickness d over the illumination region, n and s are the refractive index of the film and substrate, respectively, 𝛼 is the absorption coefficient, and Δd is the thickness variation throughout the illumination area, which has been called the roughness of the film (This nomenclature is actually confusing, since the film may not be truly “rough,” but may just have a continuously increasing thickness as in a wedge from one end to the other.) The first parameter to be extracted before the rest of the optical properties is Δd. Consider the transmission region where absorption is very small, that is, x is almost unity, and the spectrum exhibits clear maxima and minima. Eq. (1.38) can be modified by considering the maxima and minima, which are both continuous functions of 𝜆. In this way, we have, [ ( )] 2𝜋nΔd 𝜆 a 1+b −1 tan tan , (1.39a) Maxima∶TMd = √ √ 2𝜋nΔd 1 − b2 𝜆 2 1 − b [ ( )] 2𝜋nΔd 𝜆 a 1−b −1 , (1.39b) tan tan Minima∶Tmd = √ √ 2𝜋nΔd 1 − b2 𝜆 1 − b2 𝜑1 =

where a =

A , B+D

and b =

C . B+D

Monochromatic light T=1 Nonuniform film n* = n + iK

α>0

Glass substrate α=0 Refractive index = s T Detector

Δd d

Figure 1.10 System of an absorbing thin film with a variation in thickness on a thick finite transparent substrate.

1.4 The Swanepoel Technique: Measurement of n and α for Thin Films on Substrates

Notice that there is no x in these equations, since x = 1 was used. As long as 0 < Δd < 𝜆/4n, the refractive index, n and Δd, can both can be obtained by simultaneously solving Eqs. (1.39a) and (1.39b) numerically at various wavelengths. Since Eqs. (1.39a) and (1.39b) are only valid in the region of zero absorption, the refractive index, outside of the transparent region, must be obtained in another way. Theoretically, a direct integration of Eq. (1.38) over both Δd and x can be performed, although this would be analytically too difficult. Nevertheless, over the wavelength region where absorption is small (where one can still distinguish interference fringes), an approximation to the integration, according to Swanepoel, is as follows: [ ( )] ax 1 + bx 2𝜋nΔd 𝜆 −1 tan tan , (1.40a) Maxima∶TMx = √ √ 2 2𝜋nΔd 1 − b2 𝜆 1 − b x x [ ( )] 1 − bx ax 2𝜋nΔd 𝜆 −1 tan tan , (1.40b) Minima∶Tmx = √ √ 2𝜋nΔd 1 − b2 𝜆 1 − b2 x

Ax B+Dx2

x

Cx . B+Dx2

and bx = As long as 0 < x ≤ 1, numerically, there will only be where ax = one unique solution. Therefore, the two desired optical properties, refractive index, n and x (and hence 𝛼), can both be obtained when Eqs. (1.40a) and (1.40b) are solved simultaneously using the calculated average Δd. Eqs. (1.40a) and (1.40b) are valid for Δd ≪ d. As before, the calculated refractive index can be fitted to a well-established dispersion model, such as the Cauchy or Sellmeier equation, as shown in Figure 1.11, for Refractive Index of a Simulated Sample with Non-Uniform Thickness 4 Fitted To cauchy Equation Crude Calculation From Envelopes Improved Calculation

3.8 3.6

n

3.2 3 2.8 2.6 2.4 2.2 2

STRONG ABSORPTION REGION

3.4

ABSORPTION REGION n(λ) = 2.587 + (2.461 × 10–8 )λ–1 + (2.876 × 10–13)λ–2

600

800

1000

1200 1400 λ in nm

1600

1800

2000

Figure 1.11 The refractive index of a sample with d = 1 μm and Δd = 30 nm, and n fitted to a Cauchy equation in the figure. Eqs. (1.40a) and (1.40b) were used for the determination of n.

23

1 Fundamental Optical Properties of Materials I

extrapolation to shorter wavelengths. The thickness is calculated from any two adjacent maxima (or minima) using Eq. (1.34), and the absorption coefficient can be calculated from ln(xweak ) 𝛼weak = − (1.41) d where 𝛼 weak is the absorption coefficient in the weak and medium absorption region; xweak is the absorbance-like parameter (x) obtained from Eqs. (1.40a) and (1.40b); and d, as before, is the average thickness. According to Swanepoel, in the region of strong absorption, the interference fringes are smaller, and the spectrum approaches the interference-free transmission sooner. Since the transmission spectra in this region are the same for any film with the same average thickness, regardless of its uniformity, the absorption coefficient in the strong absorption region will thus be, 𝛼strong = −

ln(xstrong )

A−



(1.42)

d

(A2 −4Ti2 BD)

2T T

, Ti = T M+Tm , and T M and T m are the envelopes of maxima where xstrong = 2Ti D M m and minima, respectively, constructed from the measured spectrum. The accuracy of the thickness and refractive index can be further improved in exactly the same way as that for a film with uniform thickness for the computation of the new absorption coefficient, using Eqs. (1.40a) and (1.40b). Figure 1.12 shows the regenerated Transmission Spectrum of a Simulated Sample with Non-Uniform Thickness 1

0.8 0.7 0.6 0.5 0.4 0.3

STRONG ABSORPTION REGION

0.9

Transmittance

24

RMSE = 0.25% ABSORPTION REGION

0.2 Expt data Regenerated T Substrate

0.1 0

600

800

1000

1200 1400 λ in nm

1600

1800

2000

Figure 1.12 A regenerated transmission spectrum of a sample with an average thickness of 1 μm and average Δd of 30 nm, and a refractive index fitted to a Cauchy equation in Figure 1.11.

1.5 Transmittance and Reflectance of a Partially Transparent Plate

transmission spectrum of a simulated sample with nonuniform thickness using the optical constants calculated from the envelopes. Marquez et al. [102] have discussed the application of Swanepoel technique to wedge-shaped As2 S3 thin film and made use of the fact that a non-uniform wedge-shaped thin film has a compressed transmission spectrum. Figure 1.13 shows a flow chart that highlights the various steps involved in the extraction of the optical coefficients of thin film for both uniform and a nonuniform film thicknesses. Various computer algorithms based on the Swanepoel technique are available in the literature [103]. Further discussions and enhancements are also available [97, 104–106]. For example, one improvement in the strong absorption region is based on a so-called tangencypoint method as described in [106]. There are numerous useful applications of the Swanepoel technique for extracting the optical constants of thin films; some selected recent examples are given in [107–118].

1.5 Transmittance and Reflectance of a Partially Transparent Plate The transmittance of the thin film in Figure 1.6 is based on the interference of light waves within the thin film; it assumes that waves have a much longer coherence length than the thickness of the film, so that we may suitably add the optical fields of the waves. A film would be too thick if the coherence length of the waves is shorter than the thickness of the film. When we pass light through a transparent (or partially transparent) plate, we typically do not observe interference effects. Even if we have a thin film, we may not observe interference effects if the light source is incoherent. In these cases, we cannot add the electric fields to calculate the reflected and transmitted light irradiances; instead, we need to use reflectance and transmittance of the surfaces. Consider the multiple reflections shown in Figure 1.14. The first transmitted light intensity into the plate is (1 − R), and the first transmitted light out is (1 − R)(1 − R) = (1 − R)2 . However, there are internal reflections, so that the second transmitted light is R2 (1 − R)2 , so that the transmitted intensity through the plate is T plate = (1 − R)2 [1 + R2 + R4 + …], or Tplate =

2nno (1 − R)2 1−R = = 1 − R2 1 + R n2 + n2o

(1.43)

where R = (n – no )/(n + no ), and n is the refractive index of the plate and no that of the surrounding, as in Figure 1.14. One of the simplest ways to determine the refractive index of a medium as a plate is to measure the transmittance T plate , and then use √ n −1 −2 = Tplate + Tplate −1 (1.44) no The overall reflectance is 1 – T plate , so that Rplate =

(n − no )2 n2 + n2o

(1.45)

If the plate is partially transparent with some attenuation coefficient 𝛼, then, each time light traverses the thickness L of the plate, it experiences an attenuation factor exp(−𝛼L),

25

26

1 Fundamental Optical Properties of Materials I Uniform thickness film

Non uniform thickness film

Measure T(λ) for film on substrate

Measure T(λ) for substrate alone

Measure T(λ) for film on substrate

Measure T(λ) for substrate alone

Construct envelopes through extremes Find TM and Tm

Obtain s(λ)

Construct envelopes through extremes Find TM and Tm

Obtain s(λ)

Calculate N from TM, Tm, s and hence n at extremes at λextreme. Fit a dispersion relation, n = f(λ). Use Equation (1.33)

Obtain the refractive index n and Δd simultaneously from Equation (1.39a) and (1.39b) using only TM and Tm from the transparent region of the transmission.

Use adjacent pair of maxima and pair of minima to calculate a set of crude thickness values dcrude from Equation (1.34), then find the average dave

Obtain the refractive index n and x simultaneously from Equation (1.40a) and (1.40b) using only TM and Tm from the weak to medium absorption region (where one can still distinguish interference fringes) and the average Δdave calculated from previous step

Calculate x and hence α in the weak absorption region. From TM at maxima, n and s at λmaxima, find EM and then x. Use dave from above to find α = –ln(x)/dave.

Fit the refractive index, n obtained from the previous two steps (i .e, transparent to medium absorption region) to a dispersion relation, n = f(λ) to get the full spectrum range similar to the measured transmission spectrum

Find interference order m′ from extremes, m′ = 2nedave/λe where ne is n at λe = the extreme point λ. Make m′ integer for maxima and half integer for minima

Use adjacent pair of maxima and pair of minima in the zero to medium absorption region (where one can still distinguish interference fringes) to calculate a set of crude thickness values dcrude from Equation (1.34), then find the average dave

Using integer and half integer m′ find new thickness at each maximum and minimum, d′ = m′λe/2ne where ne is n at λe = λ at an extreme point. Find a new and better average thickness dave

Use x calculated in the weak to medium absorption region above to obtain α from Equation (1.41) in the same region, α in strong absorption region can be calculated from Equation (1.42)

Using this new dave and m′ calculate a new ne at each extreme λe. Obtain a better dispersion fit n = fnew(λ) to the new ne vs. λe data. Non-uniform thickness

Recalculate x using Equation (1.40a) and (1.40b) and use it to obtain a new absorption coefficient α in their respective absorption regions

Uniform thickness Recalculate x and hence α in the weak absorption region. From TM at maxima, new n and s at λmaxima, find EM and then x. Use new dave to find α = –ln(x)/dave

Use these parameters to regenerate the whole spectrum from Equation (1.30) to find the RMSE from Equation (1.37) with respect to the experimental transmission spectrum

All these parameters can then be used in Equation (1.38) to regenerate a transmission spectrum so that the root mean square error (RMSE) can be determined from the experimental spectrum using Equation (1.37)

Figure 1.13 A flow chart highlighting the steps involved in the Swanepoel technique.

1.6 Optical Properties and Diffuse Reflection: Schuster–Kubelka–Munk Theory

Incident beam

R4(1–R)

R3(1–R)

R(1–R)

(1–R) Thick transparent plate

R2(1–R)

no

(1–R)2

n

L

no

R2(1–R)2

Transmitted beam

R4(1–R)2

Detector

Figure 1.14 Transmitted and reflected light through a slab of material in which there is no interference.

resulting in a transmittance T plate = (1 − R)2 e−𝛼L + R2 (1 − R)2 e−3𝛼L + R4 (1 − R)2 e−5𝛼L + …, so that the series sums to (1 − R)2 e−𝛼L (1.46) Tplate = 1 − R2 e−2𝛼L

1.6 Optical Properties and Diffuse Reflection: Schuster–Kubelka–Munk Theory The treatise in Section 1.4 on the analysis of the transmission spectrum T(𝜆) assumes a smooth surface for the thin film and normal incidence. The propagating light wave through the film and its multiple reflections do not experience any scattering. The reflectance spectrum is simply 1 – T(𝜆). The situation is totally different if the film surface is rough or textured, and the bulk of the film exhibits scattering, which would be the case for particulate bulk material. The latter may be due to, for example, the polycrystallinity of the film, or the film may be a pressed powder layer or some composite with a mixture of two phases. Figure 1.15a shows a typical polycrystalline film sample with a rough surface whose optical properties are measured by diffused reflection. The nonspecular portion of light reflection is what constitutes the diffused reflection. There are many examples of diffused reflection from various forms of media, some of the most obvious being diffused reflection from rough surfaces, multiphase media, and multilayered translucent materials (e.g. paint coatings, paper, human skin, clothing, etc.). The instrumental setup is such that all light emitted over a wide solid angle is collected and the secularly reflected light is rejected—and hence the name. The experimental technique for implementing such measurements are well described in the literature (e.g. [119–127]), and there are various commercial instruments available from various vendors. The diffused reflectance spectrum is normally compared to a standard

27

28

1 Fundamental Optical Properties of Materials I

Incident light Diffused reflection cone

A particular incident ray

Specular reflection Diffused reflection

Rough surface Polycrystalline bulk

(a)

L

Inhomogeneous medium (b)

Figure 1.15 (a) Typical polycrystalline sample with a rough surface whose optical properties are measured by diffused reflection. (b) A simplified view of how diffused reflection builds up from scattering processes in the bulk.

by replacing the sample with a standard sample. Figure 1.15b shows a simplified view of how diffused reflection builds up from numerous refraction and reflection processes in the bulk so that the light is effectively scattered within the bulk. Although absorption is not shown, there would also be light extinction in the bulk. The treatise of diffused reflection is mathematically a complicated problem due to the inhomogeneity of the medium. The basic question is, “What are the suitable optical coefficients that can be used to describe the properties of the medium as related to diffused reflection?” This issue has been reviewed and discussed extensively in the literature. The Kubelka–Munk (KM) formalism [128, 129] is a theory that relates the diffuse reflectance R(𝜆) of a substance to multiple scattering and absorption within the substance. The treatise dates back to 1905 when Schuster quantitatively considered radiation traversing a foggy atmosphere [130], and hence the name Schuster–Kubelka–Munk (SKM) theory. The SKM formalism is widely used in various industrial [131–135] and biomedical applications [136, 137]. A recent practical review of diffuse reflectance with numerous references may be found in [126]. The KM theory addresses diffuse reflectance, and assigns two phenomenological coefficients labeled K and S to represent the absorption and scattering in the medium. K and S are used as coefficients that respectively indicate the amount of absorption and diffusion per unit thickness in the medium. K should not be confused with the extinction coefficient of the complex refractive index. It should be emphasized that K and S do not directly correspond to the intrinsic optical constants, that is, the absorption coefficient 𝛼, and the scattering coefficient s of the medium, although they are related. To describe the theory in simple terms, initially consider a medium such as a particulate layer (a layer of pressed powers), and assume that the sample is very thick (the sample thickness L is large, i.e. KL ≫ 1), so that there is no reflection from or transmission through the rear surface of the sample. A remission function (also known as the KM or SKM function) F(R) of diffuse reflectance R at a given wavelength is defined for the radiation absorption and scattering properties the medium of interest as F ≡ F(R) =

(1 − R)2 K = R S

(1.47)

1.6 Optical Properties and Diffuse Reflection: Schuster–Kubelka–Munk Theory

Figure 1.16 Medium with light absorption and scattering: K is the absorption and S is the scattering coefficient, that is, the fraction of light intensity absorbed or scattered, respectively, per unit distance. L is assumed to be large. I0 is the intensity of light entering the medium.

Diffused reflection RI0 measured z=0 z

I0 I

RI0 Forward light flux K, S

dz

J

L

Backward light flux

z=L

where the function F(R) represents a nearly steady-state intensity of diffuse reflectance R corresponding to an incident radiation at a wavelength 𝜆. Figure 1.16 presents a two-flux diffuse-reflectance model of a semi-infinite diffusely reflecting medium in which the incident light I 0 toward the +z direction is infinitesimally below the spectral refection surface medium [138]. The model considers the energy balance in a very thin layer (of thickness dz) in terms of diffuse intensities I and J, which are respectively along the forward (+z) and backward (−z) directions, as shown in Figure 1.16 [139], so that, over dz, dI = −(K + S)I + SJ (1.48) dz and dJ = (K + S)J − SI (1.49) dz Equations (1.48) and (1.49) can be applied as long as the particulate size is comparable to or smaller than 𝜆. K and S have been proposed to be proportional (but not equal) to the usual optical absorption coefficient 𝛼 and the scattering coefficient s, respectively, as discussed by Murphy [140, 141]; for example, K = 𝜀𝛼, where 𝜀 is defined in such a way that it is the average path length traveled by diffuse light in crossing dz. From Eqs. (1.48) and (1.49), it can be shown that (1 − R)2 K = (1.50) S 2R For an infinitely thick medium, R = J/I and Eq. (1.50) provides K/S. Equation (1.50) provides a useful way of using the diffusive reflectance spectrum to extract the absorption spectrum. The rearrangement of Eq. (1.50) gives an explicit expression for R as R = (1 + K∕S) − [(1 + K∕S)2 − 1]1∕2

(1.51)

The remission function F(R) and K for a weakly absorbing powdered medium of thickness L may be written in terms of fundamental optical parameters n and 𝛼 through [139], F=

2n2 𝛼L 3

(1.52)

K=

2n2 𝛼 3

(1.53)

and

29

1 Fundamental Optical Properties of Materials I

The particles are assumed to be spherical with rough surfaces, and the particle dimeter is larger than the wavelength. Notice that K is not the same as 𝛼. Although the relatively simple equations in the preceding text are attractive to experimentalists, a more rigorous treatise shows that the relationships between K and S and the material absorption and scattering coefficients 𝛼 and s involve the ratio 𝜇 of the total optical path length of a photon from its entry point (from one scattering process to the next) to its effective displacement from the entry point [142]. The problem immediately becomes analytically intractable without simplifying assumptions. In this model, K is proportional to 𝛼 through the ratio 𝜇. The Tauc law, which has been widely used for determining the bandgap of various semiconductors, has also been adapted in recent years (e.g. [143–148]) for the diffuse reflectance spectrum in terms of a transformed KM function h𝜈F(R), that is, [ ]p h𝜈F(R) = B(h𝜈 − Eg ) (1.54) where B is a constant (independent of photon energy) and p is an index, typically 1/3 to 2, that depends on the nature of the transition and the states that are involved. (The Tauc law is discussed in Chapter 3 in connection with amorphous semiconductors, but, in the form in Eq. (1.54), it represents a general description of absorption for photon energies just above the bandgap energy.) The rationale for Equation (1.54) is based F being proportional to 𝛼. Figure 1.17 shows a diffuse reflectance spectra within the UV and visible region (400 nm < 𝜆 < 860 nm) for a composite film of methylammonium iodide (CH3 NH3 I) and lead iodide (PbI2 ) in a 1 : 4 weight ratio. The Tauc plot is also shown and points to a bandgap Eg of 1.49 eV. Equation (1.54) has had success for estimating Eg for powder-form nanomaterials yielding experimentally supported values as reported in [150]. Alternative descriptions for characterizing the optical properties of coating materials have also been available for Eq. (1.51) in the literature. For example, the diffuse reflectance from a layer of finite thickness L on a substrate having a reflectance Rg is 80

×103 6 [h v F(R)]2

Diffuse reflectance (AU)

30

20kV X10,000

1µm

1

0000 13 41 SEI

0 0.4

0.6 0.8 Wavelength (µm)

4

1.0

0 1.15 1.20

1.30 1.40 1.50 Photon energy (hv) in eV

Figure 1.17 UV-visible diffuse reflectance spectra (left) and the Tauc-like plot (right) with p = 2 of a Perovskite film (550 nm thick) containing methylammonium iodide (CH3 NH3 I) and lead iodide (PbI2 ) in weight ratios of 1 : 4. The inset shows an SEM image of the microstructure, which consists of a random distribution of pores (voids) (AU is “arbitrary units”). The data were extracted, reanalyzed, and replotted from [149].

Acknowledgments

given as [140]: R=

1 − Rg [a − b coth(bSL)] a + b coth(bSL) − Rg

(1.55)

in which a = 1 + K/S and b = [a2 − 1]1/2 . The assumptions in Eq. (1.55) are that the light arrives from a medium that has the same refractive index as the layer, and that scattering is isotropic within the medium. In the case of a homogenous slab that has a thickness L, the diffuse reflectance R at a given wavelength can be written as R=

(1 − 𝛽 2 ) sinh(𝜅L) (1 + 𝛽 2 ) sinh(𝜅L) + 2𝛽 cosh(𝜅L)

(1.56)

where 𝛽 = [K/(K + 2S)]1/2 and 𝜅 = [K(K + 2S)]1/2 . Fitting this equation to the observed reflectance would yield 𝛽 and 𝜅, and hence K and S. However, Eq. (1.51), in many cases, has been found to be adequate for the interpretation of the spectra from high-emissive coatings with no significant effect of its covering power color building in visible and near-infrared spectral range [151]. The usefulness of the KM theory is due to the fact that the scattering and the absorption constants K and S tend to scale linearly with the concentration of colorants or pigments, and the overall contributions of colorants, inks, and pigments can be found by applying an additive rule, that is, adding the individual KL together for absorption and individual SL for scattering. The interpretation of diffuse reflectance spectra R(𝜆) and the extraction of meaningful optical coefficients continue to be an active research area. There have been several important reviews and useful criticisms of the simplified KM approach described above, which should be considered [141, 142, 152–154].

1.7 Conclusions This chapter has provided a semiquantitative explanation and discussion of the complex refractive index n* = n + iK; the relationship between the real n and imaginary part K through the Kramers–Kronig relationships; various common dispersion (n vs. 𝜆) relationships such as the Cauchy, Sellmeier, Wemple–DiDomenico dispersion relations; and the determination of the optical constants of a material in thin film form using the popular Swanepoel technique. The latter technique is based on the interference of waves in the thin film, and uses the maxima and minima in the transmittance spectrum to extract the dispersion relation for n and K. The transmittance of light through a thick semitransparent plate is also considered. The KM formulism is described in a simplified way to interpret diffuse reflection from inhomogeneous coating and/or rough surfaces. Examples are given to highlight the concepts and provide applications. The optical constants of various selected materials have been also provided in tables to illustrate typical values and enable comparisons.

Acknowledgments The authors are most grateful to Cyril Koughia for his helpful comments and assistance during the preparation of the first edition of this chapter.

31

32

1 Fundamental Optical Properties of Materials I

References 1 Greenaway, D.L. and Harbeke, G. (1968). Optical Properties and Band Structure of

Semiconductors. Oxford: Pergamon Press. 2 Wolfe, W.L. (1978). The Handbook of Optics (eds. W.G. Driscoll and W. Vaughan).

New York: McGraw-Hill. 3 Klocek, P. (ed.) (1991). Handbook of Infrared Optical Materials. New York: Marcel

Dekker. 4 Palik, E.D. (ed.) (1985). Handbook of Optical Constants of Solids. San Diego: Aca-

demic Press (now Elsevier). 5 Palik, E.D. (ed.) (1991). Handbook of Optical Constants of Solids II. San Diego: Aca-

demic Press (now Elsevier). 6 Ward, L. (1994). The Optical Constants of Bulk Materials and Films. Bristol: Insti-

tute of Physics Publishing (reprint 1998). 7 Efimov, A.M. (1995). Optical Constants of Inorganic Glasses. Boca Raton: CRC

Press. 8 Palik, E.D. and Ghosh, G.K. (eds.) (1997). Handbook of Optical Constants of Solids,

vol. 1-5. San Diego: Academic Press (now Elsevier). 9 Nikogosyan, D. (1997). Properties of Optical and Laser-Related Materials: A Hand-

book. New York: Wiley. 10 Weaver, J.H. and Frederickse, H.P.R. (1999). CRC Handbook of Chemistry and

Physics (ed. D.R. Lide). Boca Raton: CRC Press, Ch 12. 11 Adachi, S. (1992). Physical Properties of III-V Semiconductor Compounds. New

York: Wiley. 12 Adachi, S. (1999). Optical Constants in Crystalline and Amorphous Semiconductors:

Numerical Data and Graphical Information. Boston: Kluwer Academic Publishers. 13 Adachi, S. (2005). Properties of Group-IV, III-V and II-VI Semiconductors. Chich-

ester: Wiley. 14 Madelung, O. (2004). Semiconductors: Data Handbook, 3e. New York:

Springer-Verlag. 15 Bass, M. (ed.) (2010). Handbook of Optics, 3e, vol. 2. New York: McGraw-Hill,

Chapters 35 (Properties of Materials), 36 (Optical Properties of Semiconductors). 16 Nalwa, H.S. (ed.) (2001). Handbook of Advanced Electronic and Photonic Materials

and Devices, vol. 1–10. San Diego: Academic Press (now Elsevier). 17 Weber, M.J. (2003). Handbook of Optical Materials. Boca Raton: CRC Press. 18 Martienssen, W. and Walimont, H. (eds.) (2005). Springer Handbook of Condensed

Matter and Materials Data. Heidelberg: Springer, various sections. 19 Kasap, S.O. and Capper, P. (eds.) (2017). Springer Handbook of Electronic and

Photonic Materials, 2e. Heidelberg: Springer Chapter 3. 20 Novak, M. (2013). Optical properties of semiconductors, Ch. 7. In: Silicon Based

Thin Films (ed. R. Murri), 177–242. Bentham Books. 21 Tanner, D.B. (2019). Optical Effects in Solids, University of Florida. Cambridge:

Cambridge University Press. 22 Jimenez, J. and Tomm, J.W. (2016). Spectroscopic Analysis of Optoelectronic Semi-

conductors. Switzerland: Springer International Publishing. 23 Stenzel, O. (2014). Optical Coatings. Heidelberg: Springer-Verlag. 24 Fox, M. (2010). Optical Properties of Solids, 2e. Oxford: Oxford University Press.

References

25 Simmons, J.H. and Potter, K.S. (2000). Optical Materials. San Diego: Academic

Press (now Elsevier). 26 Toyozawa, Y. (2003). Optical Processes in Solids. Cambridge: Cambridge University

Press. 27 Wooten, F. (1972). Optical Properties of Solids. New York: Academic Press (now

Elsevier). 28 Abeles, F. (1972). Optical Properties of Solids. Amsterdam: North Holland (now

Elsevier). 29 Collins, R.W. (2004). Ellipsometery. In: The Optics Encyclopedia, vol. 1 (eds. T.G.

Brown, K. Creath, H. Kogelnik, et al.), 609–670. Weinheim: Wiley-VCH. 30 Kasap, S.O. (2017). Principles of Electronic Materials and Devices, 4e. New York: 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

McGraw-Hill Higher Education Chs 7 and 9. Philipp, H.R. and Taft, E.A. (1960). Phys. Rev. 120: 37. Aspnes, D.E. and Studna, A.A. (1983). Phys. Rev. B27: 985. Kronig, R. (1926). J. Opt. Soc. Amer. 12: 547. H.A. Kramers (1927). Estratto Dagli Atti del Congresso Internazionale de Fisici, 2, 545. Nussenzveig, H.M. (1972). Ch. In: Causality and Dispersion Relations1. New York: Academic Press, Inc. Suzuki, N. and Adachi, S. (1994). Jpn. J. Appl. Phys. 33: 193. Nitsche, R. and Fritz, T. (2004). Phys. Rev. B 70: 195432. Bhattacharyya, D. and Biswas, A. (2005). J. Appl. Phys. 97: 053501. Yim, C., O’Brien, M., McEvoy, N. et al. (2014). Appl. Phys. Lett. 104: 103114. Choi, S.G., Kang, J., Li, J. et al. (2015). Appl. Phys. Lett. 106: 043901. Jellison, G.E. and Modine, F.A. (1996). Appl. Phys. Lett. 69: 371. Forouhi, A.R. and Bloomer, I. (1988). Phys. Rev. B 38: 1865. Smith, D.Y. and Shiles, E. (1978). Phys. Rev. B 17: 4689. https://refractiveindex.info (Accessed, 12 August 2018) Cauchy, A.L. (1830). Bull. Sci. Math. 14: 6. Cauchy, A.L. (1836). Memoire sur la Dispersion de la Lumiere. Prague: Calve. Smith, D.Y., Inokuti, M., and Karstens, W. (2001). J. Phys. Cond. Matt. 13: 3883. Sellmeier, W. (1871). Ann. der Phys. 143: S272. Tatian, B. (1984). Appl. Opt. 23: 4477. Fleming, J.W. (1984). Appl. Opt. 23: 4486. Wolf, K.L. and Herzfeld, K.F. (1928). Handbooch der Physik, vol. 20, Ch. 10 (eds. H. Geiger and K. Scheel). Berlin: Springer Verlag. Herzberger, M. (1959). Opt. Acta 6: 197. Bachs, H. and Neuroth, N. (eds.) (1995). Schott Series on Glass and Glass Ceramics. Heidelberg: Springer. Kreidl, N.J. and Uhlmann, D.R. (eds.) (1991). Optical Properties of Glass. The American Ceramic Society. Hawkins, G. and Hunneman, R. (2004). Infrared Phys. Technol. 45: 69. Forouhi, A.R. and Bloomer, I. (1986). Phys. Rev. B 34: 7018. Chen, Y.F., Kwei, C.M., and Tung, C.J. (1993). Phys. Rev. B 48: 4373. Afromowitz, M.A. (1974). Solid State Commun. 15: 59. Adachi, S. (1982). J. Appl. Phys. 53: 5863. Adachi, S. (1988). Phys. Rev. B 38: 12345.

33

34

1 Fundamental Optical Properties of Materials I

61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99

Adachi, S. (1987). Phys. Rev. B 35: 123161. Adachi, S. (1991). Phys. Rev. B 43: 123161. Adachi, S., Mori, H., and Ozaki, S. (2002). Phys. Rev. B 66: 153201. Campi, D. and Papuzza, C. (1985). J. Appl. Phys. 57: 1305. Rakic, A.D. and Majewski, M.L. (1996). J. Appl. Phys. 80: 5909. Erman, M., Theeten, J.B., Chambon, P. et al. (1984). J. Appl. Phys. 56: 2664. Terry, F.L. (1991). J. Appl. Phys. 70: 409. Johs, B., Herzinger, C.M., Dinan, J.H. et al. (1988). Thin Solid Films 314: 137. Kim, T.J., Ghong, T.H., Kim, Y.D. et al. (2003). Phys. Rev. B 68: 115323. Hwang, S.Y., Kim, T.J., Byun, J.S. et al. (2013). Thin Solid Films 547: 276. Moss, T.S. (1950). Proc. Phys. Soc. B 63: 167. Moss, T.S. (1985). Phys. Status Solidi B 131: 415. Hervé, P.J.L. and Vandamme, L.K.J. (1996). J. Appl. Phys. 77: 5476. Kumar, V. and Singh, J.K. (2010). Ind. J. Pure Appl. Phys. 48: 571–574. Moss, T.S. (1952). Photoconductivity in the Elements. New York: Academic Press. Ravindra, N.M., Auchuluk, S., and Srivastava, V.K. (1979). Phys. Stat. Solidi B 93: K155. Finkenrath, H. (1988). Infrared Phys. 28: 327. Reddy, R.R., Nazeer Ahammed, Y., Rama Gopal, K., and Raghuram, D.V. (1998). Opt. Mater. 10: 95. Duffy, J.A. (2001). Phys. Chem. Glass 42: 151. Reddy, R.R. and Nazeer Ahammed, Y. (1995). Cryst. Res. Technol. 30: 263. Reddy, R.R. and Nazeer Ahammed, Y. (1995). Infrared Phys. Technol. 36: 825. Dale, D. and Gladstone, F. (1858). Philos. Trans. 148: 887. Dale, D. and Gladstone, f. (1863). Philos. Trans. 153: 317. Ritland, H.N. (1955). J. Am. Ceram. Soc. 38: 86. SCHOTT (2016). Technical Information TIE29, Refractive Index and Dispersion (available online at http:// www.schott.com; accessed 12 August 2018 Hoffmann, H., Jochs, W., and Westenberger, G. (1992). SPIE Proc. 1780: 303. Ghosh, G., Endo, M., and Iwasaki, T. (1994). J. Light Wave Technol. 12: 1338. Ghosh, G. (1997). Appl. Opt. 36: 1540. Ghosh, G. (1998). Phys. Rev. B 14: 8178. Englert, M., Hartmann, P., and Reichel, S. (2014); https://doi). SPIE Proc. 9131: 91310H. doi: 10.1117/12.2052706. Wemple, S.H. and DiDomenico, M. Jr., (1971). Phys. Rev. B 3: 1338. Wemple, S.H. (1973). Phys. Rev. B 7: 3767. Henry, C.H., Johnson, L.F., Logan, R.A., and Clarke, D.P. (1985). IEEE J. Quantum Electron. QE21: 1881. Wemple, S.H. (1977). J. Chem. Phys. 67: 2151. Kasap, S.O. (2013). Optoelectronics and Photonics: Principles and Practices, 2e. Upper Saddle River, NJ: Pearson Education Ch 1. Heavens, O.S. (1965). Optical Properties of Thin Solid Films. New York: Dover Publications (Reprint 1991). Poelman, D. and Smet, P.F. (2003). J. Phys. D. Appl. Phys. 36: 1850. and references therein. Swanepoel, R. (1983). J. Phys. E Sci. Instrum. 16: 1214. Manifacier, J.C., Gasiot, J., and Fillard, J.P. (1976). J. Phys. E Sci. Instrum. 9: 1002.

References

100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119

120 121 122 123 124 125 126 127

128 129 130

Hall, J.F. and Ferguson, W.F.C. (1955). J. Opt. Soci. Am. 145: 714. Swanepoel, R. (1984). J. Phys. E Sci. Instrum. 17: 896. Marquez, E., Ramirez-Malo, J.B., Villares, P. et al. (1995). Thin Solid Films 254: 83. Caricato, A.P., Fazzi, A., and Leggieri, G. (2005). Appl.Surf. Sci. 248: 440. and references therein. Chambouleyron, I., Ventura, S.D., Birgin, E.G., and Martínez, J.M. (2002). J. Appl. Phys. 92: 3093. and references therein. Ayadi, K. and Haddaoui, N. (2000). J. Mater. Sci. Mater. Electron. 11: 163. Jin, Y., Song, B., Jia, Z. et al. (2017). Opt. Exp. 25: 440. Marquez, E., Ramirez-Malo, J., Villares, P. et al. (1992). J. Phys. D 25: 535. Márquez, E., González-Leal, J.M., Prieto-Alcón, R. et al. (1998). Appl. Phys. A Mater. Sci. Process. 67: 371. Gonzalez-Leal, J.M., Ledesma, A., Bernal-Oliva, A.M. et al. (1999). Mater. Lett. 39: 232. Marquez, E., Bernal-Oliva, A.M., GonzaÂlez-Leal, J.M. et al. (1999). Mater. Chem. Phys. 60: 231. Moharram, A.H., Othman, A.A., and Osman, M.A. (2002). Appl. Surf. Sci. 200: 143. Bakr, N.A., El-Hadidy, H., Hammam, M., and Migahed, M.D. (2003). Thin Solids Films 424: 296. Gonzalez-Leal, J.M., Prieto-Alcon, R., Angel, J.A., and Marquez, E. (2003). J. Non-Crystalline Solids 315: 134. El-Sayed, S.M. and Amin, G.A.M. (2005). ND&E Int. 38: 113. Tigau, N., Ciupina, V., and Prodan, G. (2005). J. Crystal Growth 277: 529. Fayek, S.A. and El-Sayed, S.M. (2006). ND&E Int. 39: 39. Sanchez-Gonzalez, J., D𝚤az-Parralejo, A., Ortiz, A.L., and Guiberteau, F. (2006). Appl. Surf. Sci. 252: 6013–6017. Fang, Y., Jayasuriya, D., Furniss, D. et al. (2017). Opt. Quant. Electron. 49: 237. Companion, A.L. Theory and applications of diffuse reflectance spectroscopy. In: Developments in Applied Spectroscopy, vol. 4 (ed. E.N. Davis), 221–233. Boston, MA: Springer. Hecht, W.W.M. and Wendlandt, H.G. (1966). Reflectance spectroscopy (Chemical analysis), 91–209. Wiley Interscience Publishers. Kortüm, G. (1969). Reflectance Specroscopy: Principles, Methods, Applications. New York: Springer Verlag Chs 5 and 6. Fuller, M.P. and Griffiths, P.R. (1978). Analy. Chem. 50: 1906. Delgass, W.N., Haller, G.L., Kellerman, R., and Lunsford, J.H. (1979). Spectroscopy in Heterogenous Catalysis. New York: Academic Press Ch 4. Blitz, J.P. Ch 5: Diffuse reflectance spectroscopy. In: Modern Techniques in Applied Molecular Spectroscopy (ed. F.M. Mirabella), 1998. New York: Wiley. Weckhuysen, B.M. and Schoonheydt, A. (1999). Catal. Today 49: 441–451. Jentoft, F.C. (2009). Adv. Catal. 52: 129. and references therein. Monsef Khoshhesab, Z. (2012). Infrared Spectroscopy - Materials Science, Engineering and Technology (ed. T. Theophile). Rijeka, Croatia: InTech (ISBN: 978-953-51-0537-4); Ch 11 “Reflectance IR Spectroscopy”. Kubelka, P. and Munk, F. (1931). Z. Tech. Phys. (Leipzig) 12: 593. Kubelka, P. (1948). J Opt Soc 38: 448. Schuster, A. (1905). Astrophys. J. 21: 1.

35

36

1 Fundamental Optical Properties of Materials I

131 Q. Zeng, M. Cao, X. Feng, F. Liang, X. Chen and W. Sheng (1983). A study of

132 133 134 135 136

137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153

154

spectral reflection characteristics for snow, ice and water in the north of China, Hydrological Applications of Remote Sensing and Remote Data Transmission (Proceedings of the Hamburg Symposium). IAHS Publ. no. 145 Dzimbeg-Malcic, V., Barbaric-Mikocevic, Z., and Itric, K. (2011). Teh. Vjesn. 18: 117. Dingemans, L.M., Papadakis, V.M., Liu, P. et al. (2017). J. Eur. Opt. Soc. Rapid Publ. 13 Article Number: 40. https://doi.org/10.1186/s41476-017-0068-2. Rogers, G. (2016). Color. Res. Appl. 41: 580. Džimbeg-Malˇci´c, V., Barbari´c-Mikoˇcevi´c, Ž., and Itri´c, K. (2011). Technical Gazette (Croatia) 18: 117. Gorpas, D., Alexandratou, E., Politopoulos, K., and Yova, D. (2005). Diagnostic algorithms based on Kubelka-Munk theory to discriminate healthy and Atheromatous aorta in animal model. Proc. SPIE 5268: 235. Reuter, T., Karl, S., Hoffmann, M.B., and Dietzek, B. (2013). Biomed. Eng. 58 (Suppl. 1) https://doi.org/10.1515/bmt-2013-435. Myrick, M.L., Simcock, M.N., Baranowski, M. et al. (2011). Appl. Spectrosc. Rev. 46: 140. Simmons, E.L. (1972). Opt. Acta 19: 845–851. Murphy, A.B. (2006). J. Phys. D. Appl. Phys. 39: 3571. Murphy, A.B. (2007). Appl. Opt. 46: 3133. Yang, L. and Kruse, B. (2004). J. Opt. Soc. Am. 21: 1933. Escobedo Morales, A., Sánchez Mora, E., and Pal, U. (2007). Rev. Mex. Fis. 53: 18. Nowak, M., Kauch, B., and Szperlic, P. (2009). Rev. Sci. Intrum. 80: 046107. Yakuphanoglu, F. (2010). J. Alloys Compd. 507: 184. Lopez, R. and Gómez, L.R. (2012). J. Sol-Gel Sci. Technol. 61: 1. Caglar, Y., Görgün, K., Ilican, S. et al. (2016). Appl. Phys. A 122 (733). Nam-GyuPark (2015). Mater. Today 18: 65. Mahapatra, S.K., Saykar, N., Banerjee, I. et al. (2018). J. Mater. Sci. Mater. Electron. https://doi.org/10.1007/s10854-018-9992-1. Ahsan, R., Khan, M.Z.R., and Basith, M.A. (2017). J. Nanophoton. 11: 046016. He, Y., Zhang, X., and Zhang, Y.J. (2016). Wuhan Univ. Technol. Mat. Sci. Edit. 31: 100. Nobbs, J.H. (1985). Rev. Prog. Coloration 15: 66. R.A. de la Osa, A. Fernandez, Y. Gutierrez, D. Ortiz, F. Gonzalez, F. Moreno, J.M. Saiz (2017). The extended Kubelka-Munk theory and its application to colloidal systems. Third International Conference on Applications of Optics and Photonics, Proceedings of SPIE, 10453, 104531F 8 pp. DOI: 10.1117/12.2272115 Wang, J., Xu, C., Nilsson, A.M. et al. (2019). Adv. Optical Mater 7: 1801315.

37

2 Fundamental Optical Properties of Materials II S.O. Kasap 1 , K. Koughia 1 , Jai Singh 2 , Harry E. Ruda 3 , and Asim K. Ray 4 1

Department of Electrical and Computer Engineering, University of Saskatchewan, 57 Campus Drive, Saskatoon, Canada College of Engineering, IT and Environment, B-Purple 12, Charles Darwin University, Ellengowan Drive, Darwin, Australia Department of Materials Science and Engineering, University of Toronto, 170 College Street, Toronto, Canada 4 Department of Electrical & Computer Engineering, Kingston Lane, Brunel University London, UK 2 3

CHAPTER MENU Introduction, 37 Lattice or Reststrahlen Absorption and Infrared Reflection, 40 Free Carrier Absorption (FCA), 42 Band-to-Band or Fundamental Absorption (Crystalline Solids), 45 Impurity Absorption and Rare-Earth Ions, 48 Effect of External Fields, 54 Effective Medium Approximations, 58 Conclusions, 61 References, 62

2.1 Introduction The optical properties of semiconductors typically consist of their refractive index n and extinction coefficient K or absorption coefficient 𝛼 (or, equivalently, the real and imaginary parts of the relative permittivity) and their dispersion relations, that is, their dependence on the wavelength 𝜆, of the electromagnetic radiation or photon energy h𝜈 and the changes in the dispersion relations with temperature, pressure, alloying, impurities, etc. This chapter is an extension of Chapter 1 on fundamental optical properties with emphasis on optical absorption. (Those general references quoted in Chapter 1 also apply to this chapter.) The elementary descriptions herein, as in Chapter 1, emphasize physical concepts and subsequent results rather than mathematical derivations. A typical relationship between the absorption coefficient and photon energy observed in a crystalline semiconductor is shown in Figure 2.1, where various possible absorption processes are illustrated. The important features in the 𝛼 versus h𝜈 behavior as the photon energy increases can be classified in the following types of absorptions: (a) reststrahlen or lattice absorption, in which the radiation is absorbed by vibrations of the crystal ions; (b) free carrier absorption (FCA) due to the presence of free electrons and holes, an effect that decreases with increasing photon energy; (c) an impurity absorption band Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

2 Fundamental Optical Properties of Materials II

106 Absorption coefficient (1/cm)

38

Fundamental absorption Reststrahlen

105

Eg

104 Absorption edge

103 Impurity

102

Exciton (low T) Free carrier

10 1 0.001

Free carrier

0.01

0.1 1 Photon energy (eV)

10

100

Figure 2.1 Absorption coefficient is plotted as a function of the photon energy in a typical semiconductor to illustrate various possible absorption processes.

(usually narrow) due the various dopants; (d) exciton absorption peaks that are usually observed at low temperatures and are close to the fundamental absorption edge; and (e) band-to-band or fundamental absorption of photons, which excites an electron from the valence to the conduction band (CB). The type (e) absorption has a large absorption coefficient and occurs when the photon energy reaches the bandgap energy Eg . It is probably the most important absorption effect; its characteristics for h𝜈 > Eg can be predicted using the results of Section 2.4. The values of Eg , and its temperature shift, dEg /dT, are therefore important factors in semiconductor-based optoelectronic devices. In nearly all semiconductors, Eg decreases with temperature, hence shifting the fundamental absorption to longer wavelengths. The refractive index n also changes with temperature, as described in Chapter 1. There is a good correlation between the refractive index and the bandgap of semiconductors in which, typically, n decreases as Eg increases (see Chapter 1); wider bandgap semiconductors have lower refractive indices. The refractive index n and the extinction coefficient K (or 𝛼) are related by virtue of the Kramers–Kronig relations, described in Chapter 1. Thus, large increases in the absorption coefficient for h𝜈 near and above the bandgap energy Eg also result in increases in the refractive index n versus h𝜈 in this region. The optical and some structural properties of various semiconductors are listed in Table 2.1. The characteristics of some of these absorptions are described in the following sections. While these topics have been covered in extensive detail in various graduate-level textbooks in the past [1, 2], the approach here is to provide concise descriptions with more insight from an experimentalist’s point of view rather than a rigorous treatise. We have included simple electro-optic effects, since these have become particularly important in the last two decades with the advent of optical communications [3]. The last section in this chapter deals with the effective medium approximation to describe the optical properties of heterogeneous media, that is, mixtures of different phases. Such mixed phases occur frequently in various applications, for example a continuous medium that contains dispersed particulate matter, including air pores.

2.1 Introduction

Table 2.1 Crystal structure, lattice parameter a, bandgap energy E g at 300 K, type of bandgap (D = Direct and I = Indirect), change in E g per unit temperature change (dE g /dT) at 300 K, bandgap wavelength 𝜆g , and refractive index n close to 𝜆g .

Crystal a nm

Eg eV

dE g /dT Type meV K−1 𝝀g (𝛍m) n (𝝀g )

Diamond

D

0.3567

5.48

I

−0.05

0.226

2.74

1.1

Ge

D

0.5658

0.66

I

−0.37

1.87

4

27.6 42.4 (4 μm)

Si

D

0.5431

1.12

I

−0.25

1.11

3.45

13.8 16 (5 μm)

a-Si:H

A

Semiconductors

dn/dT ×10−5 K−1

Group IV

1.7–1.8

0.73

III–V Compounds AlAs

ZB

0.5661

2.16

I

−0.40

0.57

3.2

15

AlP

ZB

0.5451

2.45

I

−0.35

0.52

3

11

AlSb

ZB

0.6135

1.58

I

−0.3

0.75

3.7

GaAs

ZB

0.5653

1.42

D

−0.45

0.87

3.6

15

GaAs0.88 Sb0.12

ZB

1.15

D

GaN

W

0.3190 a 3.44 0.5190 c

D

6.8

1.08 −0.45

0.36

2.6

GaP

ZB

0.5451

2.24

I

−0.54

0.40

3.4

GaSb

ZB

0.6096

0.73

D

−0.35

1.7

4

In0.53 Ga0.47 As on InP ZB

0.5869

0.75

D

1.65

In0.58 Ga0.42 As0.9 P0.1 on InP

ZB

0.5870

0.80

D

1.55

In0.72 Ga0.28 As0.62 P0.38 ZB on InP

0.5870

0.95

D

1.3

32.8

InP

ZB

0.5869

1.35

D

−0.36

0.91

3.4–3.5 9.5

InAs

ZB

0.6058

0.35

D

−0.28

3.5

3.8

2.7

InSb

ZB

0.6479

0.18

D

−0.3

7

4.2

29

ZnSe

ZB

0.5668

2.7

D

−0.72

0.46

2.3

6.3 (4 μm) 6.30 (10 μm)

ZnTe

ZB

0.6101

2.25

D

0.55

2.7

II–VI Compounds

Note: A = Amorphous, D = Diamond, W = Wurtzite, ZB = Zinc blende. Typical values combined from different sources.

39

40

2 Fundamental Optical Properties of Materials II

2.2 Lattice or Reststrahlen Absorption and Infrared Reflection In the infrared wavelength region, ionic crystals reflect and absorb light strongly due to the resonance interaction of the electromagnetic (EM) wave field with the transverse optical (TO) phonons. The dipole oscillator model based on ions driven by an EM wave gives the complex relative permittivity as [4]: 𝜀ro − 𝜀r∞ (2.1) 𝜀r = 𝜀′r + i𝜀′′r = 𝜀r∞ + ( ), ( )2 1 − 𝜔𝜔 − i 𝜔𝛾 𝜔𝜔 T

T

T

where 𝜀ro and 𝜀r∞ are the relative permittivity at 𝜔 = 0 (very low frequencies or dc) and 𝜔 = ∞ (very high frequencies), respectively; 𝛾 is the loss coefficient representing the rate of energy transfer from the EM wave to optical phonons (per unit reduced mass of the crystal ions); and 𝜔T is a transverse optical phonon frequency related to the nature of bonding between the ions in the crystal, that is, 𝜔2T = 𝜔2o (𝜀r∞ + 2)∕𝜀ro + 2), in which 𝜔o 2 = 𝛽/Mr , 𝛽 is the force constant in “restoring force = −𝛽 × displacement,” and Mr is the reduced mass of the negative and positive ions in the crystal. It should be emphasized that the derivation of Eq. (2.1) based on a simple Lorentz oscillator model for ionic polarization, which is then used in the Clausius–Mossotti equation. The loss, 𝜀r ′′ and the absorption are maxima when 𝜔 = 𝜔T , and the wave is attenuated by the transfer of energy to the transverse optical phonons, thus the EM wave couples to the transverse optical phonons. At 𝜔 = 𝜔L , the wave couples to the longitudinal optical phonons. The refractive index n from ionic polarization vanishes, and the reflectance is minimum. Figure 2.2 shows the optical properties of AlSb in terms of n, K, and R vs wavelength [5]. The extinction coefficient K and reflectance R peaks occur over about the same wavelength region, corresponding to the coupling of the EM wave to the transverse optical phonons. At wavelengths close to 𝜆T = 2𝜋/𝜔T , n and K peak, and there is a strong absorption of light that corresponds to the EM wave resonating with the TO lattice vibrations, then R rises sharply. 𝜔T and 𝜔L are related through 𝜔2T = 𝜔2L (𝜀r∞ ∕𝜀ro ) which is called the Lyddane–Sachs–Teller relation. √ √ It relates the high- and lowfrequency refractive indices n∞ = 𝜀r∞ and no = 𝜀ro to 𝜔T and 𝜔L . The complex refractive index, n* = n + iK, then becomes [4] [ ] 2 2 𝜔 − 𝜔 L T n∗2 = 𝜀r = 𝜀r∞ 1 + 2 . (2.2) 𝜔T − 𝜔2 − i𝛾𝜔 (Note that n2 − K 2 = 𝜀′r and 2nK = 𝜀′′r .) Taking CdTe as an example, and substituting the values for 𝜀r∞ , 𝜔T , 𝜔L , and 𝛾 from Table 2.2 into the preceding expression, at 𝜆 = 70 μm or 𝜔 = 2.6909 × 1012 rad s−1 , one gets n = 3.20 and K = 0.00235. Although the preceding expression is usually sufficient to predict n* in the infrared for many compound semiconductors and ionic crystals, for low-bandgap semiconductors, one should also include the contribution from the free carriers.

λL

14

λT

14

12

12

10

10

λL

λT

90 80 70 60

8

8

6

6

4

K n

22

4

24

26 28 30 32 Wavelength (μm)

40 30

R (%)

20

K

2 0 20

n

50

2

34

36

0 38

10 0 16 18 20 22 24 26 28 30 32 34 36 38 40 Wavelength (μm)

Figure 2.2 Infrared refractive index n, extinction coefficient K (left), and reflectance R (right) of AlSb. Note: The wavelength axes are not identical, and wavelengths 𝜆T and 𝜆L corresponding to 𝜔T and 𝜔L , respectively, are shown as dashed vertical lines. Source: Data extracted from W. J. Turner and W. E. Reese, 1962 [5] with 𝜆L = 2𝜋(c/𝜔L ) and 𝜆T = 2𝜋(c/𝜔L ) from Table 2.2.

42

2 Fundamental Optical Properties of Materials II

Table 2.2 Values of the quantities required for calculating n and K from Eq. (2.2) for some selected crystals; values adapted and converted from the Handbook of Optical Constants of Solids, edited by E. D. Palik, 1985. Crystal

𝜺r∞

𝝎T x 1012

𝝎L x 1012

𝜸 x 1012

AlSb

9.6

60.09

63.97

CdTe

7.1

26.58

31.86

0.34 1.24

GaAs

11

50.65

55.06

0.45

InAs

11.7

41.09

45.24

0.75

InP

9.61

57.25

65.03

0.66

SiC

6.7

149.48

182.65

0.90

Note: 𝜔T , 𝜔L , and 𝛾 are in rad s−1 .

2.3 Free Carrier Absorption (FCA) An electromagnetic wave with sufficiently low frequency oscillations can interact with free carriers in a material and thereby drift the carriers. This interaction results in an energy loss from the EM wave to the lattice vibrations through the carrier scattering processes. Based on the Drude model, the complex relative permittivity 𝜀r (𝜔) due to N free electrons per unit volume is given by [6] 𝜀r = 𝜀′r + i𝜀′′r = 1 −

𝜔2p 𝜔2

+ i(𝜔∕𝜏)

; 𝜔2p =

Ne2 , 𝜀 o me

(2.3)

where 𝜔p is a plasma frequency, which depends on the electron concentration, and 𝜏 is the relaxation time of the carrier scattering process, that is, the mean scattering time. For metals where the electron concentration is very large, 𝜔p is of the order of ∼1016 rad s−1 , in the range of UV frequencies, and for 𝜔 > 𝜔p 𝜀r ≈ 1, the reflectance becomes very small. Metals lose their free-electron reflectance in the UV range, thus becoming UV transparent. The reflectance does not fall to zero because there are other absorption processes such as those due to interband electron excitations or excitations from core levels to energy bands. Plasma edge transparency, where the reflectance almost vanishes, can also be observed in doped semiconductors. For example, the reflectance of doped InSb has a plasma edge wavelength that decreases with increasing free carrier concentration [7]. Equation (2.3) can be written in terms of the conductivity 𝜎 o at low frequencies (DC) as: 𝜏𝜎o 𝜎o +i . (2.4) 𝜀r = 𝜀′r + i𝜀′′r = 1 − 2 𝜀o [(𝜔𝜏) + 1] 𝜀o 𝜔[(𝜔𝜏)2 + 1] ′′ In metals, 𝜎 o is high. At frequencies where √𝜔 1/𝜏, 𝛼 becomes proportional to N, the free carrier concentration, and 𝜆2 as: 𝛼 ∼ 𝜎o ∕𝜔2 ∼ N𝜆2 .

(2.8)

Experimental observations on FCA in doped semiconductors are generally in agreement with these predictions. For example, 𝛼 increases with N, whether N is increased by doping or by carrier injection [10, 11]. However, not all semiconductors show the simple 𝛼 ∝ 𝜆2 behavior. A proper account of the field-driven electron motion and scattering must consider the fact that 𝜏 will depend on the electron’s energy. The correct approach is to use the Boltzmann transport equation [12] with the appropriate scattering mechanism. FCA can be calculated by using a quantum mechanical approach based on second-order time-dependent perturbation theory with Fermi–Dirac statistics [13]. Absorption due to free carriers is commonly written as 𝛼 ∝ 𝜆p , where the index p depends primarily on the scattering mechanism, although it is also influenced by the compensation doping ratio, if the semiconductor has been doped by compensation, and the free carrier concentration. In the case of lattice scattering, one must consider scattering from acoustic and optical phonons. For acoustic phonon scattering, p ≈ 1.5; for optical phonon scattering, p ≈ 2.5; and for impurity scattering, p ≈ 3.5. Accordingly, the observed FCA coefficient will then have all three contributions as 𝛼 = Aacoustic 𝜆1.5 + Aoptical 𝜆2.5 + Aimpurity 𝜆3.5 .

(2.9)

Inasmuch as 𝛼 for FCA depends on the free carrier concentration N, it is possible to evaluate the latter from the experimentally measured 𝛼, given its wavelength dependence and p, as discussed by Ruda [14]. FCA coefficient 𝛼 (mm−1 ) for n-type GaP, n-type PbTe, and n-type ZnO are shown in Figure. 2.3. FCA in p-type Ge demonstrates how the FCA coefficient 𝛼 can be dramatically different from what is expected from Eq. (2.9). Figure 2.4a shows the wavelength dependence of the absorption coefficient for p-Ge over the wavelength range from about 2 to 30 μm [18]. The observed absorption is due to excitations of electrons from the spin-off band

43

44

2 Fundamental Optical Properties of Materials II

FCA coefficient, 1/mm 50

p-PbTe (p = 3.7 × 1018 cm–3) 77 K

ZnO (n = 5.04 × 1019 cm–3) ZnO (n = 2.33 × 1019 cm–3)

α ~ λ2

10 p-PbTe (p = 5 × 1017 cm–3) 77 K α ~ λ3 n-GaP (Nd = 3 × 1017 cm–3) 300 K

1

n-PbTe (n = 3.9 × 1017 cm–3) 77 K

ZnO (n = 1.02 × 1019 cm–3)

0.3 0.3

1

10

30

Wavelength (μm)

Figure 2.3 Free carrier absorption coefficient (1/mm) in n-GaP at 300 K [15], p and n type PbTe [16] at 77 K, and in doped n-type ZnO at room temperature [17] (Nd is the donor concentration.). 1 × 104 LH to HH

E k

SO to HH

α, 1/m

VB

SO to LH

HH LH Heavy-hole band Light-hole band

1 × 103 p-Ge at 300 K

SO Spin-off band

5 × 102 1

10

40

Wavelength (μm) (a)

(b)

Figure 2.4 (a) Free carrier absorption due to holes in p-Ge [18]. (b) The valence band of Ge has three bands; heavy hole, light hole, and spin-off bands.

to the heavy hole band, and from spin-off band to the light hole band, and from the light hole band to the heavy hole band, as marked in Figure 2.4b. FCA may also be combined with hot-carrier photodetection in silicon to enable sub-bandgap near infrared light to be detected at room temperature with very high responsivity and minimal noise-equivalent power [19].

2.4 Band-to-Band or Fundamental Absorption (Crystalline Solids)

2.4 Band-to-Band or Fundamental Absorption (Crystalline Solids) Band-to-band absorption or fundamental absorption of radiation occurs due to the photoexcitation of an electron from the valence band (VB) to the conduction band (CB). Thus, the absorption of a photon creates an electron in the CB and a hole in the VB, and requires the energy and momentum conservation of the excited electron, hole, and photon. In crystalline solids, as the band structures depend on the lattice wave vector k, there are two types of band-to-band absorptions corresponding to direct and indirect transitions. In contrast, in amorphous solids, where no long-range order exists, only direct transitions are meaningful. The band-to-band absorption in crystalline solids is described in the following text, and that in amorphous or disordered solids will be presented in Chapter 3. First the direct and then indirect transitions will be described here. A direct transition is a photoexcitation process in which no phonons are involved. As the photon momentum is negligible as compared to the electron momentum when the photon is absorbed to excite an electron from the VB to the CB, the electron’s k-vector does not change. A direct transition on a E-k diagram is a vertical transition from an initial energy E and wavevector k in the VB to a final energy E′ and wavevector k ′ in the CB where k ′ = k as shown in Figure 2.5a. The energy (E′ − Ec ) is the kinetic energy (ℏk)2 /(2me *) of the electron with an effective mass me *, and (Ev − E) is the kinetic energy (ℏk)2 /(2mh *) of the hole left behind in the VB. The ratio of the kinetic energies of the photogenerated electron and hole depends inversely on the ratio of their effective masses. The absorption coefficient 𝛼 is derived from the quantum mechanical transition probability from E to E′ , the occupied density of states at E in the VB from which electrons are excited, and the unoccupied density of states in the CB at E + h𝜈. Thus, 𝛼 depends on the joint density of states at E and E + h𝜈, and we have to suitably integrate this joint density of states. Near the band edges, the density of states can be approximated by a parabolic band, and the absorption coefficient 𝛼 is obtained as a function of the photon (αhν)2 (eV/cm)2

E

(αhν)2 (eV/cm)2

50

CB

× 108

× 109 4

E′ 25

Ec

GaAs

CdTe

2

Photon

Eg

3

1

Ev

E

VB k

k

0 hν 1.35 1.40 1.45 1.50 1.55 Photon energy (eV)

hν 1.4 1.8 2.2 Photon energy (eV)

0

Figure 2.5 (a) A direct transition from the valence band (VB) to the conduction band (CB) by the absorption of a photon. Absorption behavior represented as (𝛼h𝜈)2 vs photon energy h𝜈 near the band edge for single crystals of (b) p-type GaAs. Source: From I. Kudman and T. Seidel, J. Appl. Phys., 33, 771 (1962) (c) CdTe. Source: From A. E. Rakhshani, J. Appl. Phys., 81, 7988 (1997).

45

46

2 Fundamental Optical Properties of Materials II

energy as: 𝛼 h𝜈 = A(h𝜈 − Eg )1∕2 ,

(2.10)

where the constant A ≈ [(e2 /(nch2 me *)](2𝜇*)3/2 , in which 𝜇* is a reduced electron and hole effective mass, n is the refractive index, and Eg is the direct bandgap, minimum Ec – Ev at the same k value. Experiments indeed show this type of behavior for photon energies above Eg and close to Eg as shown Figure 2.5b for a GaAs crystal [20] and in Figure 2.5c for a CdTe crystal [21]. The extrapolation to zero photon energy gives the direct bandgap Eg , which is about 1.40 eV for GaAs and 1.46–1.49 eV for CdTe. For photon energies very close to the bandgap energy, the absorption is usually due to exciton absorption, especially at low temperatures, and it will be discussed later in this chapter. In indirect bandgap semiconductors such as crystalline Si and Ge, the photon absorption for photon energies near Eg requires the absorption or emission of phonons during the absorption process, as illustrated in Figure 2.6a. The absorption onset corresponds to a photon energy of (Eg − h𝜗), which represents the absorption of a phonon with energy h𝜗 (𝜗 is the phonon frequency). In this case, 𝛼 is proportional to [h𝜈 − (Eg − h𝜗)]2 . Once the photon energy reaches (Eg + h𝜗), then the photon absorption process can also occur by phonon emission, for which the absorption coefficient is larger than that for phonon absorption. The absorption coefficients for the phonon absorption and emission processes are given by [22]: 𝛼absorption = A[fBE (h𝜗)][h𝜈 − (Eg − h𝜗)]2 ; h𝜈 > (Eg − h𝜗),

(2.11)

[ ] 𝛼emission = A (1 − fBE (h𝜗) [h𝜈 − (Eg + h𝜗)]2 ; h𝜐 > (Eg + h𝜗),

(2.12)

and

α1/2

E 8 CB

(cm–1/2)

2 Phonon

Photon

kCB

Ec Eg

333 K

6

Phonon emission

4 Ev

hυ12

195 K

Phonon absorption

2

VB 1 k1

kVB

k



0 0

1.0

1.1

1.2

1.3

Photon energy (eV)

Eg – hϑ

Eg + hϑ (a)

(b)

Figure 2.6 (a) Indirect transitions across the bandgap involve phonons. Direct transitions in which dE/dk in the CB is parallel to dE/dk in the VB lead to peaks in the absorption coefficient. (b) Fundamental absorption in Si at two temperatures. The overall behavior is well described by Eqs. (2.11) and (2.12).

2.4 Band-to-Band or Fundamental Absorption (Crystalline Solids)

where A is a constant, and f BE (h𝜗) is the Bose–Einstein distribution function at the phonon energy h𝜗, that is, f BE (h𝜗) = [(exp(h𝜗/k B T) − 1]−1 , where k B is the Boltzmann constant and T is the temperature. As we increase the photon energy in the range (Eg − h𝜗) < h𝜈 < (Eg + h𝜗), the absorption is controlled by 𝛼 absorption and the plot of 𝛼 1/2 vs h𝜈 has an intercept of (Eg − h𝜗). For photon energies h𝜈 > (Eg + h𝜗), the overall absorption coefficient is 𝛼 absorption + 𝛼 emission , but, at slightly higher photon energies than (Eg + h𝜗), 𝛼 emission quickly dominates over 𝛼 absorption , since [f BE (h𝜗)] > > [(1 − f BE (h𝜗)]. Figure 2.6b shows the behavior of 𝛼 1/2 versus photon energy for Si at two temperatures for h𝜐 near band edge absorption. At low temperatures, f BE (h𝜗) is small, and 𝛼 absorption decreases with decreasing temperature, as apparent in Figure 2.6b. Equations (2.11) and (2.12) intersect the photon energy axis at (Eg − h𝜗) and (Eg + h𝜗), which can be used to determine Eg . Examination of the extinction coefficient K or 𝜀r ′′ versus photon energy for Si in Figure 1.2 (Chapter 1) shows that absorption peaks at certain photon energies, h𝜈 ≈ 3.5 and 4.3 eV. These peaks are due to the fact that the joint density of states function peaks at these energies. The absorption coefficient peaks whenever there is a direct transition in which the E versus k curve in the VB is parallel to the E versus k curve in the CB, as schematically illustrated in Figure 2.6a, where a photon of energy h𝜈 12 excites an electron from state 1 in the VB to state 2 in the CB in a direct transition k1 = k2 . Such transitions where E versus k curves are parallel at a photon energy h𝜈 12 result in a peak in the absorption versus photon energy behavior and can be represented by the condition that (∇k E)CB − (∇k E)VB = 0.

(2.13)

This condition is normally interpreted as the joint density of states reaching a peak value at certain points in the Brillouin zone called van Hove singularities. Identification of peaks in K versus h𝜈 behavior involves the examination of all E versus k curves of a given crystal that can participate in a direct transition. In silicon, the 𝜀r ′′ peaks at h𝜈 ≈ 3.5 eV and 4.3 eV correspond to Eq. (2.13) being satisfied at points L, along in k-space, and X along in k-space, at the edges of the Brillouin zone. In degenerate semiconductors, the Fermi level EF lies in a band, for example, it lies in the CB for a degenerate n-type semiconductor. In these semiconductors, electrons in the VB can only be excited to states above EF in the CB rather than to the bottom of the CB. The absorption coefficient then depends on the free carrier concentration since the latter determines EF . Fundamental absorption is then said to depend on band filling, and there is an apparent shift in the absorption edge, called the Burstein–Moss shift. Furthermore, in degenerate indirect semiconductors, the indirect transition may involve a non-phonon scattering process, such as impurity or electron-electron scattering, which can change the electron’s wavevector k. Thus, in degenerate indirect bandgap semiconductors, absorption can occur without phonon assistance, and the absorption coefficient becomes: 𝛼 ∼ [h𝜈 − (Eg + ΔEF )]2 ,

(2.14)

where ΔEF is the energy depth of EF into the band measured from the band edge. Heavy doping of degenerate semiconductors normally leads to a phenomenon called bandgap narrowing and bandtailing. Bandtailing means that the band edges at Ev and Ec are no longer well-defined cut-off energies, and there are electronic states above Ev

47

48

2 Fundamental Optical Properties of Materials II

and below Ec whose density of states falls sharply with energy away from the band edges. Consider a degenerate direct band gap p-type semiconductor. One can excite electrons from states below EF in the VB, where the band is nearly parabolic, to tail states below Ec , where the density of states decreases exponentially with energy into the bandgap, away from Ec . Such excitations lead to 𝛼 depending exponentially on h𝜈, a dependence that is usually called the Urbach rule [23, 24], given by: 𝛼 = 𝛼0 exp[(h𝜈 − Eo )∕ΔE],

(2.15)

where 𝛼 0 and Eo are material-dependent constants, and ΔE, called the Urbach width, is also a material dependent constant. The Urbach rule was originally reported for alkali halides. It has been observed for many ionic crystals, degenerately doped crystalline semiconductors, and almost all amorphous semiconductors. While exponential bandtailing can explain the observed Urbach tail of the absorption coefficient versus photon energy, it is also possible to attribute the absorption tail behavior to strong internal fields arising, for example, from ionized dopants or defects.

2.5 Impurity Absorption and Rare-Earth Ions Impurity absorption can be registered as peaks of absorption coefficient lying below the fundamental (band-to-band) and excitonic absorption (Figure 2.1). They can mostly be related with the presence of ionized impurities or, simply, ions. The origin of these peaks lies in the electronic transitions either between the electronic states of an ion and CB/VB or intra-ionic transitions (e.g. within d or f shells, or between s and d shells, etc.). In the first case, the appearing features are intense and broad lines, while in the latter case their appearance strongly depends on whether these transitions are allowed or not by the parity selection rules. For allowed transitions, the appearing absorption peaks are quite intense and broad, while the forbidden transitions produce weak and narrow peaks. The general reviews of the topic may be found in studies by Blasse and Grabmaier [25], Henderson and Imbusch [26], and DiBartolo [27]. Optical spectra of rare-earth ions in various crystals and glasses have been well treated in a number of books, some of which are [26, 28–30]. In the following sections, we concentrate primarily on the absorption properties of rare earth (RE) ions, which are of prime importance for modern optoelectronics. Consider the absorption of radiation due to dopants or impurities in a material system, for example, Er3+ ions in a glass host such as a germanium–gallium–sulfide glass. Figure 2.7a–c show typical examples of the transmission spectra, absorption bands, and the absorption cross-section, respectively, for the 4 I15/2 –4 I13/2 band of Er3+ in (GeS2 )0.9 (Ga2 S3 )0.1 glass. Let N (m−3 ) be the number of dopants per unit volume, and 𝛼(𝜆) be the absorption coefficient for dopant excitations from a manifold centered at energy E1 to a manifold centered at E2 . The absorbed radiation power per unit area 𝛿I(𝜆) over a small distance 𝛿x is 𝛿I(𝜆) = 𝛿xI𝛼(𝜆). The optical absorption due to electronic transitions between manifolds is very often described by using an absorption cross-section 𝜎 a (𝜆) defined as 𝛿I(𝜆) = 𝛿xIN𝜎 a (𝜆), that is 𝜎a (𝜆) =

1 𝛼(𝜆) N

(2.16)

2.5 Impurity Absorption and Rare-Earth Ions

(c) 60

10

40 20

α, cm–1

8

(a)

0 36 34 6

12

6

2H 11/2

4F

(b)

4F 9/2

7/2

4

4I 13/2

Cross-section, ×10–21 cm2

Transmittance, %

80

4 4I

2

4S 3/2

4I 9/2

600

600

2

11/2

0 0

1000 1500 1600 1450 Wavelength, nm

1500

1550

1600

0

Figure 2.7 (a) Optical transmittance of (GeS2 )0.9 (Ga2 S3 )0.1 glass doped with 0.5 at.% of Er. (b) Optical absorption coefficient of Er3+ ions extracted from data in figure (a). Peaks correspond to optical transition from ground manifold 4 I15/2 to excited manifolds as marked in the figure. The meaning of symbols is explained in Figure 2.8. (c) Absorption cross-section for the 4 I15/2 –4 I13/2 transition.

which basically follows the wavelength dependence of the spectral absorption coefficient, as shown in Figure 2.7b,c. RE is the common name for the elements from lanthanum (La) to lutetium (Lu). They have atomic numbers from 57 to 71 and form a separate group in the periodic table. The specific feature of these elements is the incompletely filled 4f shell. The electronic configurations of REs are listed in Table 2.3. The RE may be embedded in different host materials in the form of divalent or trivalent ions. As divalent ions, REs generally exhibit broad absorption–emission lines related to the allowed 4f → 5d transitions. Table 2.3 Occupation of outer electronic shells for rare earth elements. 57 58

La

4s2

4p2

4d10

Ce

2

2

10

4s

2

4p

2

4d

10

– 1

4f

3

5s2

5p6

5d1

6s2

2

6

1

5d

6s2

6

5s

2

5p

59

Pr

4s

4p

4d

4f

5s

5p



6s2

60

Nd

4s2

4p2

4d10

4f4

5s2

5p6



6s2

Er

4s2

4p2

4d10

4f12

5s2

5p6



6s2

Yb

4s2

4p2

4d10

4f14

5s2

5p6



Lu

2

2

10

14

2

6

… 68 … 70 71

4s

4p

4d

4f

5s

5p

6s2 1

5d

6s2

49

2 Fundamental Optical Properties of Materials II

In the trivalent form, REs lose two 6s electrons and one of the 4f or 5d electrons. As a result of the Coulomb interaction of 4f electrons with the positively charged core and other electrons in this shell, and spin-orbit coupling, the 4f level becomes split into a complicated set of manifolds whose position, in first approximation, is relatively independent of the host matrix, because the 4f level is well screened by 5s and 5p shells from outer influence [31]. Figure 2.8 shows the energy level diagram of the low-lying 4fN states of trivalent ions embedded in LaCl3 . (This is the so-called Dieke diagram [32].) In the second approximation, the exact construction and precise energy position

40

2F

1S 0

6D 9/2

7/2 5/2

38

3P 2

4G 9/2

36

6I 7/2

2H 11/2

34

6P

32

5D 4

3M

3P 1 3P 0 1I 6

5/2 7/2

8

2K 13/2

10

2P 3/2

3L

3/2 5/2 7/2

2D

0.3

30

22

18

2G

4F 4S 3/2

9/2

2

6F 1/2 3/2 5/2 7/2 6H

9/2

9/2

2 3

4

3H 6 3H 2F 5 7/2

8

15/2

6 3/2 H15/2 1/2

7

13/2

6

11/2

13/2 11/2

11/2

5/2

9/2 5

2F 3 4 5/2 H4 I9/2 Ce Pr Nd

5I

4 Pm

7F

7F

2F 5/2 3H

6

13/2 11/2

5 4

4

13/2

7

0.8

4

11/2

9/2

0 1 2 3

6

3H

4I

5

7/2

7/2

5I

4I

3F

9/2

11/2

7/2

0

5I

4F 9/2

5/2

4 3

0.5

5S 2 5F 5

6F

1G 4

1G 4

2H 11/2

5

3/2

4F 5F 3/2 1 5/2 3K8 6 4F 2 7/2

4 4S3/2

3 2 1

5/2

0.4

3

5S 2

2

4

4F 9/2

5D 4

4

2H 7/2 9/2

3F

15/2

0

5F

1D 2

5

1

5/2

2H 11/2

12

6

3/2 7/2

2G 7/2

1D 2

14

8

2

7/2

5G 2H 4 9/2 4G 11/2 4I

4G 7/2 4F 3/2 4G 5/2

2G

2K15/6 5G′ 4 9/2 5 G11/2 3K 7

5D 3

5/2

3/2 1 0 4G9/2

16

10

5D 3

3P 2P 1/2 1I 2 4G11/2 6

20

3H6

5I–10

2D

24

2

2

2P 3/2

26

5G 3

5D

5/2 3/2

Wavelength (μm)

4D 1/2

28

Energy (103 cm–1)

50

3F

5

1.0 1.3 1.6

4

2.0 2.5 5.0

3 2 1

6H 7 5/2 F0 Sm Eu

8S 7F6 6H15/2 5I8 4I15/2 3H6 2F7/2 Dy Ho Er Tm Yb Gd Tb

Figure 2.8 Energy level diagram of the low-lying 4fN states of trivalent ions doped in LaCl3 . Source: From G. H. Dieke, Spectra and Energy Levels of Rare Earth Ions in Crystals, Wiley, New York, 1968.

2.5 Impurity Absorption and Rare-Earth Ions

of manifolds depend on the host material via the crystal field and via the covalent interaction with ligands surrounding RE ion. The ligand is an atom, molecule, radical, or ion with one or more unshared pairs of electrons that can attach to a central metallic ion (or atom) to form a coordination complex. Examples of ligands include ions (F− , Cl− , Br− , I− , S2− , CN− , NCS− , OH− , NH2 − ) or molecules (NH3 , H2 O, NO, CO). Some ligands that share electrons with metal ions or atoms form very stable complexes. The optical transitions between 4f manifold levels are forbidden by the parity selection rule which states that, for a permitted atomic (ionic) transition, the wave functions of initial and final states must have different parity, that is, parity must change in the transition. The parity is a property of a quantum mechanical state that describes the function after a mirror reflection. Even functions (states) are symmetric—identical after reflection (like a cosine function)—while functions odd (states) are antisymmetric (like a sine function). For an ion (or atom), which is embedded in host material, the parity selection rule may be partially removed due to the action of crystal field giving rise to “forbidden lines.” The crystal field is the electric field created by a host material at the position of an ion. The resulting absorption–emission lines are characteristics of individual RE ions and quite narrow because they are related to forbidden inner-shell 4f transitions. The intensities of absorption–emission lines are related to the so-called oscillator strengths. Judd–Ofelt (JO) analysis is based on calculating the oscillator strengths of an electric dipole (ED) transition between states of a trivalent RE ion embedded in different host lattices and fitting them to experimentally observed oscillator strengths by using so-called intensity or force parameters Ω2 , Ω4 , and Ω6 that depend on the host material [33, 34]. Experimental oscillator strength for a particular absorption band is given by fexp =

me c 2 𝛼(𝜈)d𝜈 𝜋e2 N ∫Band

(2.17)

where N is the concentration of RE ions in the host, 𝜈 is the frequency, and 𝛼(𝜈) is the experimentally observed absorption spectrum as given in Figure 2.7b. The possible states of RE ions are often referred to terms and written as 2S + 1 LJ , where L = 0, 1, 2, 3, 4, 5, 6... determines the electron’s total orbital angular momentum and is conventionally represented by letters S, P, D, F, G, I. The term (2S + 1) is called spin multiplicity and represents the number of spin configurations, while J is the total angular momentum, which is the vector sum of the overall (total) angular momentum and overall spin (J = L + S). This description in the literature is based on the assumption that Russell–Saunders coupling is approximately valid, even though RE ions are not expected to rigidly obey this rule (their atomic numbers are more than 40). The value (2 J + 1) is called the term’s multiplicity and is the number of possible combinations of overall orbital angular momentum and overall spin, which yield the same J. Thus, the notation 4 I15/2 for the ground state of Er3+ corresponds to term (J, L, S) = (15/2, 6, 3/2), which has a multiplicity 2 J + 1 = 16 and a spin multiplicity 2S + 1 = 4. The energy separation between different terms (different L- and S-values) is due to the Coulombic repulsion between the electrons in this 4f subshell. The latter energy separation depends on L and S, but not J. The energy split for a given L and S—for example, energy separation between 4 I15/2 , 4 I13/2 , and 4 I11/2 , etc., are due to the spin-orbit coupling, which depends on J and MJ . The energy levels that are finely separated within a given 2S+1 LJ are due to the Stark effect.

51

52

2 Fundamental Optical Properties of Materials II ′

If (2S +1 LJ′ ) represents the initial and (2S+1 LJ ) the final state of an RE ion, then the electric dipole transition has the oscillator strength that is in the form of fcal =

∑ 8𝜋me c 𝜒ed Ωk |⟨SLJ|U (k) |S′ L′ J ′ ⟩|2 , 2 3h𝜆(2J + 1)n k=2,4,6

(2.18)

where Ωk are the previously mentioned intensity parameters representing the influence of host material; 𝜒 ed = n(n2 + 2)2 /9 is the so-called local field correction factor for electric dipole moment transitions for a medium with a refractive index n; and the reduced ′ ′ ′ matrix operator components |⟨SLJ|U (k) |S L J ⟩|2 are calculated using the so-called intermediate coupling approximation, and are tabulated because they are practically independent of the host material. Some values of reduced matrix operator components for Er3+ ions are presented in Table 2.4, for which data were taken from [35, 36]. The key idea of JO analysis is to minimize the discrepancy between experimental and calculated values of oscillator strengths, f exp and f cal , respectively, by the appropriate choice of coefficients Ωk . As a possible way, oscillator strengths f exp and f cal are the measured and calculated strengths, respectively, for all possible optical absorption bands ∑ (shown in Figure 2.7b) and the integral error Bands |(f exp 2 – f cal 2 )/f exp 2 | is minimized by choosing Ω2 , Ω4 , and Ω6 . The values Ωk are then used to calculate the probabilities of radiative transitions and appropriate radiative lifetimes of excited states, which are very useful for numerous optical applications. For example, the radiative lifetime 𝜏 r of a transition may be calculated as 𝜏 r = 1 /A, where A is the probability of spontaneous emission per unit time, which can be expressed as A(J ′ ; J) =

( ) 64𝜋 4 e2 × [𝜒ed Sed (J; J ′ ) + 𝜒md Smd J; J ′ ] ′ 3 3h(2J + 1)𝜆

(2.19)

Table 2.4 The reduced matrix elements of U(k) (k = 2, 4, and 6) and typical wavelengths 𝜆 for the Er3+ transitions. (S,L)J

(S′ ,L′ )J′

(U[2] )2

(U[4] )2

(U[6] )2

4

4

I13/2

0.0188

0.11176

1.4617

4

F9/2

0

0.5655

0.4651

1

652

4

S3/2

0

0

0.2285

2

521

2

H11/2

0.7056

0.4109

0.0870

4

F7/2

0

0.1467

0.6273

3

487

4

F5/2

0

0

0.2237

4

450

4

F3/2

0

0

0.1204

2

H9/2

0

0.078

0.17

5

407

2

G11/2

0.9178

0.5271

0.1197

2

G9/1

0

0.2416

0.1235

6

2

K15/2

0.0219

0.0041

0.0758

2

G7/2

0

0.0174

0.1163

378 1520 652

I15/2

Band

𝝀 (nm)

1520

Source: Data extracted from M.J. Weber, Phys. Rev. 157, 1967, 262, 1967 and W.T. Carnall, P.R. Field and B.G. Wybourne, J. Chem. Phys., 49, 4424, 1968.

2.5 Impurity Absorption and Rare-Earth Ions

where Sed (J′ ; J) and Smd (J′ ; J) are the electric and magnetic line strengths of this transition, respectively. The factors 𝜒 ed = n(n2 + 2)2 /9 and 𝜒 md = n3 are local field correction factors for the electric and magnetic dipole moment transitions, respectively. Sed and Smd are given by ∑ Sed (J; J ′ ) = Ωk |⟨SLJ|U (k) |S′ L′ J ′ ⟩|2 (2.20) k=2,4,6

and Smd (J; J ′ ) =

(

ℏ 2mc

)2 ∑

Ωk |⟨SLJ||L + 2S||S′ L′ J ′ ⟩|2 .

(2.21)

k=2,4,6

The magnetic dipole contributions can be easily calculated according to [37] as {[ [ ]]}1∕2 2 2 2 2 (J + 1) − (L − S) ⟨SLJ||L + 2S||S L J ⟩ = (S + L + 1) − (J + 1) 4(J + 1) (2.22) ′ ′ ′

Further, since |⟨SLJ|U (k) |S L J ⟩|2 are given as in Table 2.4, Eq. (2.20), for example, for the 4 I13/2 to 4 I15/2 transition at around 1.55 μm becomes, ′





Sed = 0.0188Ω2 + 0.1176Ω4 + 1.4617Ω6

(2.23)

The intensity parameters Ω2 , Ω4 , and Ω6 are used to characterize and compare different materials. Ω2 is considered to be of a prime importance because it is the most sensitive to the local structure around the rare-earth ion and material composition, and it has been correlated with the degree of covalency [38]. Some values of Ωk (k = 2, 4, and 6) for Er3+ ions in different hosts are given in Table 2.5 as typical examples. Notice that Ω2 is lower for fluoride glasses that are more ionic. More Ωk values for different ions and host materials may be found, for example, in references [38, 46]. Reviews of Table 2.5 Selected examples of intensity parameters Ω2 (k = 2, 4 and 6) for Er3+ in various hosts from various sources. 𝛀2 ×10−20 cm2

𝛀4 ×10−20 cm2

𝛀6 ×10−20 cm2

RE ion

Host

Er3+

Fluorochlorozirconate

1.92 ± 0.3

0.88 ± 0.16

0.59 ± 0.08

[39]

Er3+

Fluoride glasses

2.91

1.27

1.11

[40]

Er3+

ZBLAN

2.91

1.78

1.0

[41]

Er

Silicate glasses

4.23

1.04

0.61

[40]

Er3+

Phosphate glasses

6.65

1.52

1.11

[40]

Er3+

Aluminate glass (66.7CaO)(33.3Al2 O3 )

7.33

1.78

0.47

[42]

Er3+

Silica glass (SiO2 > 96%)

8.15

1.43

1.22

[43]

Er3+

Ge-Ga-S chalcogenide glass Ge0.30 Ga0.04 S0.655 :Er0.005

11 ± 2

2.9 ± 0.2

1.6 ± 0.1

[44]

Er3+

Ge-Ga-Se chalcogenide glass Ge0.23 Ga0.08 Se0.67 S0.014 Er0.006

13 ± 2

3.4 ± 0.2

1.3 ± 0.1

[45]

3+

Ref.

53

54

2 Fundamental Optical Properties of Materials II

the Judd–Ofelt theory and its applications have been given by Walsh [47] (one of the most extensive reviews), Quimby and Miniscalo [48], Tanabe and coworkers [49–51], Goldner and Auzel [52] (includes the modified Judd–Ofelt), Hehlen et al. (comprehensive review) [53], Adam [54], as well as others—for example, [55, 56].

2.6 Effect of External Fields 2.6.1

Electro-Optic Effects

The application of an external field (E or B) can change the optical properties in a number of ways that depends not only on the material but also on the crystal structure. Electro-optic effects refer to changes in the refractive index of a material induced by the application of an external electric field, which therefore “modulates” the optical properties; the applied field is not the electric field of any light wave, but a separate external field. The application of such a field distorts electronic motions in atoms or molecules of a substance and/or the crystal structure, resulting in changes in the optical properties. For example, an applied external field can cause an optically isotropic crystal such as GaAs to become birefringent. Typically, changes in the refractive index are small due to the applied electric field. The frequency of the applied field has to be such that it appears to be static over the time a medium takes to change its properties, as well as the time the light requires to pass through the substance. The electro-optic effects are classified according to first- and second-order effects. If we consider the refractive index as a function of the applied electric field E, that is, n = n(E), we can, of course, expand it in a Taylor series. Denoting the electric field-dependent refractive index by n′ , one can write: n′ = n + a1 E + a2 E2 + … ,

(2.24)

where n represents the electric field-independent refractive index, and the coefficients a1 and a2 are called the linear and second-order electro-optic effect coefficients, respectively. Although one may consider including even higher terms in the expansion of Eq. (2.24), these are generally very small and have negligible effects within highest practical fields. The change in n due to the linear E-dependent term is called the Pockels effect, and that due to the E2 term is called the Kerr effect. The coefficient a2 is generally written as 𝜆K, where K is called the Kerr coefficient. Thus, the contributions to n′ from the two effects can be written as: (2.25a)

Δn = a1 E 2

Δn = a2 E = (𝜆K)E

2

(2.25b)

All materials exhibit the Kerr effect but not Pockels effect, because only a few crystals have non-zero a1 . For all noncrystalline materials (such as glasses and liquids), and those crystals that have a center of symmetry (centrosymmetric crystals such as NaCl), a1 = 0. Only crystals that are noncentrosymmetric exhibit the Pockels effect. For example, an NaCl crystal (centrosymmetric) exhibits no Pockels effect, but a GaAs crystal (noncentrosymmetric) does. The Pockels effect involves examining the effect of the field on the indicatrix, that is, the index ellipsoid, and requires the electro-optic tensor. For example, the change in the

2.6 Effect of External Fields

principal refractive index n1 along the principal axis x (where D and E are parallel) of the indicatrix is written as: 1 1 1 Δn1 = − n31 r11 E1 − n31 r12 E2 − n31 r13 E3 (2.26) 2 2 2 where Ej is the field along j; j = 1–3 corresponding to x, y, and z; and rij are the elements of the electro-optic tensor r, a 6 x 3 matrix. The Pockels coefficients for various crystals can be found in references [57–59]. The Pockels effect has found important applications in optical communications in Pockels cell modulators that typically use lithium niobate (LiNbO3 ). 2.6.2

Electro-Absorption and Franz–Keldysh Effect

Electro-absorption is a change of the absorption spectrum caused by an applied electric field. There are fundamentally three types of electro-absorption processes: the Franz–Keldysh effect (FKE), field-injected FCA, and the confined Stark effect. In FKE [60, 61], a strong constant applied electric field induces changes in the band structure, which change the photon-assisted probability of an electron tunneling from the maximum of the VB to the minimum of CB. FKE can be detected as a variation of absorption and/or reflection of light with photon energies slightly less than the corresponding bandgap. This effect was first observed in CdS, where the absorption edge was red-shifted with the applied electric field, causing an increase of absorption that could be detected even visually [62]. Later, FKE was observed and investigated in Ge [63], Si [64], and other semiconducting materials. In the dynamic Franz–Keldysh effect (DFKE), ultrafast band structure changes are induced in a semiconductor in the presence of a strong laser electric field. These changes include absorption below the band edge and oscillatory behavior in the absorption above. The steady-state effect is normally quite small, but the dynamic effect can be up to 40%, as found in thin films of GaAs [65]. In the presence of an electric field, the absorption coefficient 𝛼 can be written as [1]: [ √ 3∕2 ] 4 2m∗ (Eg − ℏ𝜔) , (2.27) 𝛼(𝜔) ∝ 𝛼0 (𝜔) exp − 3 ℏ2 eF where 𝛼 0 (𝜔) is the absorption coefficient in the absence of electric field, m* is the effective mass, Eg is the optical gap, e is the electron charge, and F is the applied electric field. It should be mentioned that the electric field–induced change in the absorption coefficient implies a change in the refractive index as discussed by Seraphin and Bottka [66]. In the FCA, the concentration of free carriers N in a given band is changed (modulated) by an applied voltage, to change the extent of photon absorption. In this case, the absorption coefficient is proportional to N and to the light wavelength 𝜆 raised to some power, typically 2–3. In the confined Stark effect, the applied electric field modifies the energy levels in a quantum well (QW). A QW is a thin crystalline semiconductor between two crystalline semiconductors as barriers, which can confine electrons or holes in the dimension perpendicular to the layer surface, while the movement in the other dimensions is not restricted. It is often realized with a thin crystalline layer of a semiconductor medium, embedded between other crystalline semiconductor layers of wider bandgap, for example, GaAs embedded in AlGaAs. The thickness of such a QW is typically about

55

56

2 Fundamental Optical Properties of Materials II

p

MQW i

Relative transmission n

100 %

0 E

0

Vr

Figure 2.9 A schematic illustration of an electroabsorption modulator using the quantum confined Stark effect in MQWs. The i-region has MQWs. The transmitted light intensity can be modulated by the applied reverse bias to the pin device, because the electric field modifies the exciton energy in the QWs.

Vr

5–20 nm. Such thin crystalline layers can be fabricated by the molecular beam epitaxy (MBE) or metal–organic chemical vapor deposition. The energy levels are reduced by an amount proportional to the square of the applied field. Without any applied bias, light with photon energy just less than the QW exciton excitation energy will not be significantly absorbed. When a field is applied, the energy levels are lowered and the incident photon energy is now sufficient to excite an electron and hole pair in the QWs. Therefore, the relative transmission decreases with the reverse bias, as shown in Figure 2.9 for a reverse biased pin device in which the i-region has multiple quantum wells (MQWs). The reverse bias increases the electric field in the i-region and which enhances photogeneration in this region. In practice, the effect of electroabsorption is used to construct an electroabsorption modulator (EAM), which is a semiconductor device to control the intensity of a laser beam via applied electric voltage. Most EAMs are made in the form of a waveguide with electrodes for applying an electric field in a direction perpendicular to the modulated light beam. For achieving a high extinction ratio, one usually exploits the quantum confined Stark effect in a QW structure. They can be operated at very high speeds; a modulation bandwidth of tens of gigahertz can be achieved, which makes these devices useful for optical fiber communications. With the recent interest in two-dimensional materials has come the realization that these monolayer systems can offer strong quantum confinement effects and, in particular, with regards to the quantum confined Stark effect in them. This is discussed by Sie et al. [67]. 2.6.3

Faraday Effect

The Faraday effect, originally observed by Michael Faraday in 1845, is the rotation of the plane of polarization of a light wave as it propagates through a medium subjected to a magnetic field parallel to the direction of propagation of light. When an optically inactive material such as glass is placed in a strong magnetic field and then a plane polarized light is sent along the direction of the magnetic field, it is found that the emerging light’s plane of polarization has been rotated, as shown in Figure 2.10. The magnetic field can be applied, for example, by inserting the material into the core of a magnetic coil. The induced specific rotatory power, given by 𝜃/L, where L is the length of the medium and 𝜃 is the angle by which the rotation occurs, is found to be proportional to the magnitude of the applied magnetic field, B, which gives the amount of rotation as, 𝜃 = VBL,

(2.28)

where V is the proportionality constant, the so-called Verdet constant, which depends on the material and wavelength of light.

2.6 Effect of External Fields

I θ

Input light E

Output light E′ θ

B Beam cross-section L

I Output light E″

Input light E′



θ

B

Figure 2.10 The Faraday effect. The sense of rotation of the optical field E depends only on the direction of the magnetic field for a given medium (given Verdet constant). If light is reflected back into the Faraday medium, the field rotates a further 𝜃 in the same sense to come out as E′′ with a 2𝜃 rotation with respect to E.

The Faraday effect is typically small. For example, a magnetic field of ∼0.1 T causes a rotation of roughly 1∘ through a glass rod of length 20 mm. It seems to appear that an “optical activity” has been induced by the application of a strong magnetic field to an otherwise optically inactive medium. There is, however, an important difference between the natural optical activity and Faraday effect. The sense of rotation 𝜃 in the Faraday effect, for a given material (Verdet constant), depends only on the direction of the magnetic field B. If V is positive, for light propagating parallel to B, the optical field E rotates in the same sense as an advancing right-handed screw pointing in the direction of B. The direction of light propagation does not change the absolute sense of rotation of 𝜃. If we reflect the wave to pass through the medium again, the rotation increases to 2𝜃. The Verdet constant depends not only on the wavelength 𝜆 but also on the charge-to-mass ratio of the electron and the refractive index, n(𝜆), of the medium through: (e∕me ) dn 𝜆 . (2.29) 2c d𝜆 The Verdet constant values for some materials are listed in Table 2.6. Notice that dn/d𝜆, and hence V is large over wavelengths near a polarization resonance peak but small far away from a resonance peak. The Faraday effect has found useful applications in photonics, for example, in optical isolators [68]. V =−

Table 2.6 Verdet constants of some selected materials from various sources, including, M. J. Weber, The Handbook of Optical Materials, CRC Press, 2003.

Material

Quartz 589 nm

Tb3 Ga5 O12 633 nm

ZnS 589 nm

ZnTe 633 nm

NaCl 589 nm

Crown glasses 633 nm

Dense flint glass (SF57) 633 nm

V (rad m−1 T−1 )

4.0

−134

65.8

188

10

4–6

20

57

58

2 Fundamental Optical Properties of Materials II

2.7 Effective Medium Approximations Effective medium approximations (EMAs) are commonly employed to describe the optical constants, such as the effective refractive index neff and effective extinction coefficient K eff of a composite medium containing a mixture of two or more phases. The mixture is an inhomogeneous medium. Quite often, the mixture has dispersed particles (second phase) in a host matrix material (the first phase), and one is interested in describing the composite medium in terms of neff and K eff . The dispersed phase may be of definite shape and size (such as dispersed spheres), or both the shape and size may follow some distribution. There are numerous examples of composite media such as porous dielectrics, dielectrics containing a mixture of two or more different ceramics, cermets, nanoparticles dispersed in a solution, mesocrystals, highly polycrystalline media (e.g. thin films), composite epoxies used in electronic packaging, polymer blends used in various engineering applications, polymer oxide composites in optics, etc. Polymer blends can be engineered just right to have the so-called “smart optical properties.” The optical properties of nanomaterials and nanoporous materials invariably involve a description that is based on treating the medium as a mixture of dielectrics, each of which has well-defined n and K, so the objective is to find neff and K eff for the whole medium, that is, “homogenize the heterogeneous medium.” What is the correct mixture rule? The physics of macroscopically inhomogeneous media has attracted much interest, and there are several treatises and reviews on the subject that include dielectric and optical properties (e.g. [69–81]). There are several mathematical descriptions of the optical properties of composite media, including the well-known Maxwell Garnett and Bruggeman models. These models have several implicit assumptions, which have been highlighted in a collection of works in reference [82] and others in [73, 74, 83]. Typically, for example, the size of an inclusion is assumed to be smaller than the light wavelength, so that we only need to consider the interaction of the electric dipole with the electromagnetic wave. There are at least three important EMAs: Lorentz–Lorenz, Maxwell Garnett, and Bruggeman formulations. The Lorentz–Lorenz refractive index mixture rule is based on summing the polarizabilities of the constituents of a mixture so that one can write the effective refractive index neff of a mixture to be determined from its constituents as [84, 85] n2eff − 1 n2eff + 2

= f1

n21 − 1 n21 + 2

+ f2

n22 − 1 n22 + 2

+···

(2.30)

where n1 and f 1 are the refractive index and the volume fraction of component 1, respectively, and n2 and f 2 are the refractive index and the volume fraction of component 2, respectively, and so on, and f 1 + f 2 + … = 1. The mixture rule in Eq. (2.30) is equivalent to using the Clausius–Mossotti equation in the field of dielectrics. The volume fraction can be easily obtained from the mole fraction. This model has been widely and successfully used by many researchers. For example, in a recent paper, Eq. (2.30) has been applied to the measured optical properties of a medium consisting of Pb(II) ions in an aqueous solution and then extracting the Pb(II) concentration [86]. The latter work involved developing a sensor, which consisted of a silicon-on-insulator (SOI) microring resonator coated with tetrasulfide-functionalized mesoporous silica (MPS) (S4-MPS) to detect Pb(II) ions in aqueous solutions from 10 ppb to 1 ppm.

2.7 Effective Medium Approximations

The Bruggeman effective medium approximation [81, 87] for a heterogeneous or composite material system having two components with refractive indices n1 and n2 is given by f1

n21 − n2eff n21 + 2n2eff

+ f2

n22 − n2eff n22 + 2n2eff

=1

(2.31)

where f 1 and f 2 are the volume fractions of phases 1 and 2, and f 1 + f 2 = 1. The composite medium need not have dispersed particles in a continuous matrix and, in general, can be considered to be an aggregate system in which any phase can touch any other phase, that is, there is no condition imposed on whether a phase should be dispersed with a spherical geometry, etc. Porous semiconductors exhibit physically interesting optical properties such as a bandgap shift, increased luminescence intensity, and improved photoresponsivity. Because of their potential applications in optoelectronics and chemical sensors, porous silicon layers have been widely investigated as a composite mixture of three components, namely silicon, silicon dioxide, and voids. Equation (2.31) with three components has been applied to calculate neff for nanoporous silicon over the wavelength range 2–5 μm where the extinction coefficient is small [88]. The results are shown in Figure 2.11 as neff vs porosity from experiments and the Bruggeman theory in Eq. (2.31) with three terms for Si, SiO2 , and air. It can be seen that the agreement is excellent over the entire 30–70% porosity range investigated. While the agreement is indeed very good in this particular case, the fits are not always this close as highlighted by the original authors for the case of mesoporous silicon and in [89]. Although the preceding description specifically considered the real part of the complex refractive index, it is possible to use the mixture rules for the complex refractive index and hence extract both the real and imaginary (extinction coefficient) parts, that is, neff and K eff . For example, in a recent study, gallium nitride films on sapphire substrates were thermally oxidized to nanoporous gallium oxide (Ga2 O3 )

Effective refractive index

2.4

Experiment

2.2 2.0 1.8 1.6 1.4

Bruggeman

1.2 30

40

50 60 Porosity (%)

70

80

Figure 2.11 Plot of the experimental effective refractive index vs porosity and as predicted from the Bruggeman theory for nanoporous silicon, which has been taken as a three-component medium with Si, SiO2 , and pores (air). Source: After M. Khardani, M. Bouaïcha, B. Bessaïs, phys. stat. sol. (c) 4, 1986–1990 (2007).

59

60

2 Fundamental Optical Properties of Materials II

under wet conditions. The ellipsometric spectra of Ga2 O3 films were then analyzed in terms of the Bruggeman effective medium approximation. Figure 2.12 shows the variation of effective refractive index neff and effective extinction coefficient K eff with the free space wavelength 𝜆. The value of neff for nanoporous Ga2 O3 film is 1.92 at a wavelength of ∼400 nm, and the effective extinction coefficient is negligibly small in the wavelength range of 300 nm < 𝜆 < 840 nm, supporting the use of this material as a transparent coating in optoelectronic devices [90]. Equation (2.31) can easily be extended to multicomponent composite material systems. The Maxwell Garnett equation [81, 91] applies to a composite medium where one can identify a host as a continuous medium in which there are dispersed spherical dielectric particles. Suppose that the dispersed spherical dielectric particles have an index nd and volume fraction f d in a continuous host matrix with an index nm ; then, this mixture rule is given by n2eff − n2m n2eff + 2n2m

= fd

n2d − n2m

(2.32)

n2d + 2n2m

in which f d is assumed to be small. Further, the dispersed particles do not touch each other to form a continuum. This equation has been modified by Polder and van Santen [92] to take into account that the dispersed particles may have an ellipsoid shape. The effect of the shape and anisotropy of the microscopic inclusion has been also discussed in [70, 93–96]. Further, the inclusion need not be dielectric—for example, metal particles dispersed in an insulating dielectric [97]. Mixture rules provide a convenient method of calculating the effective n and effective K for a mixture of two or more phases, and hence characterizing the composite medium. The equations can be used to search for the best composite medium for a particular application. Mirkhani et al. [98] have theoretically investigated the change in the refractive index of nanocomposites composed of Ga-doped ZnO nanoparticles (Ga-ZnO NPs) dispersed in a matrix that is composed of conductive polymer Poly(3,4-ethylenedioxythiophene) and Poly(styrene sulfonic acid) (PEDOT:PSS). The aim was to match the refractive index of GaZnO/PEDOT:PSS with that of an

Refractive index

Extinction coefficient

2.4

0.4 Bulk Ga2O3

neff

Keff

Bulk Ga2O3

0.2

2.0 Nanoporous Ga2O3

Nanoporous Ga2O3 1.6

0 300

400 500 600 700 Wavelength (nm)

800

300

400 500 600 700 Wavelength (nm)

800

Figure 2.12 Refractive index and the extinction coefficient vs wavelength for bulk and porous Ga2 O3 . Source: Data extracted and replotted from S. Kim, M. Kadam, J-H Kang, and S-W Ryu, Electron. Mater. Lett., 12, 596–602 (2016).

Acknowledgments

organic emissive layer (1.7–1.8), and the effective medium calculations showed that the volume fraction of the Ga-ZnO should be within the 45–71% range when the concentration of Ga in ZnO nanoparticles is between 0 and 4% (Chapter 15) [98]. A suitably chosen composite medium with a desired effective index of refraction, as determined by the mixture rules, can be used for the enhancement of light extraction and absorption efficiency in optical devices such as organic light emitting diodes (OLEDs) and organic solar cells (OSCs). In these applications, it is crucial to match the refractive indices of conductive anode electrode and the emissive layer. It is important to mention that the preceding description has neglected light scattering that would occur at interphase interfaces. The introduction of scattering losses need to also be considered in an accurate design of a composite medium. Various other formulae have been proposed for calculating the dielectric and optical properties of composite media, some of which are due to Looyenga [99], Monecke [100], and others, as reviewed in [74]. Nearly all EMA formulae have one common feature—that they are based on Maxwell’s equations in the static limit. Theoretical descriptions of composite media have always interested and intrigued researchers and will continue to do so, given their technological importance. The chapter on Ellipsometry (Chapter 15) has further discussion on the effective medium approach to the optical properties of inhomogeneous media.

2.8 Conclusions Selected optical properties of solids are briefly and semi-quantitatively reviewed in this chapter. The emphasis has been on physical insight rather than mathematical rigor. Where appropriate, examples have been given with typical values for various constants. A classical approach is used for describing lattice or reststrahlen absorption, infrared reflection, and FCA, whereas a quantum approach is used for band-to-band absorption of photons. Direct and indirect absorption and the corresponding absorption coefficients are discussed with typical examples, but without invoking the quantum mechanical transition matrices. Impurity absorption that occurs in RE doped glasses is also covered with a brief description of the Judd–Ofelt analysis and how this theory can be used to extract the lifetime of an upper level from absorption spectra measurements. Due to their increasing importance in photonics, the Kerr, Pockels, Franz–Keldysh, and Faraday effects are also described, but without delving into their difficult mathematical formalisms. Important mixture rules that describe the optical properties of composite media in terms of their constituents, such as the Lorentz–Lorenz, Bruggeman, and Maxwell Garnet equations, have been discussed with typical applications. Media composed of nanoparticles can be most easily described by mixture rules such as the Bruggeman rule.

Acknowledgments One of the authors (S.O. Kasap) thanks the University of Saskatchewan Centennial Enhancement Chair program for financial support.

61

62

2 Fundamental Optical Properties of Materials II

References 1 Boer, K.W. (1990). Survey of Semiconductor Physics. New York: Van Nostrand

Reinhold. 2 Balkanski, M. and Wallis, R.F. (2000). Semiconductor Physics and Applications.

Oxford: Oxford University Press. 3 Kasap, S.O. (2004). Optoelectronics. In: The Optics Encyclopedia, vol. 4

(eds. T.G. Brown, K. Creath, H. Kogelnik, et al.), 2237. Weinheim: Wiley-VCH. 4 Mitra, S.S. (1985; pp. 213–275). Chapter 11: Optical properties of nonmetallic

5 6 7 8 9 10 11 12

13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

solids for photon energies below the fundamental band gap. In: Handbook of Optical Constants of Solids (ed. E.D. Palik). San Diego: Academic Press. Turner, W.J. and Reese, W.E. (1962). Phys. Rev. 127: 126. Fox, M. (2010). Optical Properties of Solids, 2e. Oxford: Oxford University Press; Ch 7. Spitzer, W.G. and Fan, H.Y. (1957). Phys. Rev. 106: 882. Hagen, E. and Rubens, H. (1904). Ann. Phys. 14: 986. Elliott, R.J. and Gibson, A.F. (1974). An Introduction to Solid State Physics and Its Applications. London: MacMillan Press Ltd. Ashcroft, N.W. and Mermin, N.D. (1976). Solid State Physics. Philadelphia, USA: W. B. Saunders. Ch. 27. Briggs, H.B. and Fletcher, R.C. (1953). Phys. Rev. 91: 1342. Pidgeon, C.R. (1980). Free carrier optical properties of semiconductors. In: Handbook on Semiconductors, vol. 2 (ed. M. Balkanski), 223–328. Amsterdam, Ch. 5: North Holland Publishing. Ruda, H.E. (1992). J. Appl. Phys. 72: 1648. Ruda, H.E. (1987). J. Appl. Phys. 61: 3035. Wiley, J.D. and DiDomenico, M. (1970). Phys. Rev. B1: 1655. Riedl, H.R. (1962). Phys. Rev. 127: 162. Weihler, R.L. (1966). Phys. Rev. 152: 734. Kaiser, W., Collins, R.J., and Fan, H.Y. (1953). Phys. Rev. 91: 1380. Tanzid, M., Ahmadivand, A., Zhang, R. et al. (2018). ACS Photonics 5: 3472–3477. Kudman, I. and Seidel, T. (1962). J. Appl. Phys. 33: 771. Rakhshani, A.E. (1997). J. Appl. Phys. 81: 7988. Bube, R.H. (1974). Electronic Properties of Crystalline Solids. San Diego: Academic Press Ch. 11. Urbach, F. (1953). Phys. Rev. 92: 1324. Pankove, J. (1965). Phys. Rev. 140: A2059. Blasse, G. and Grabmaier, B.C. (1994). Luminescent Materials. Berlin: Springer-Verlag. Henderson, B. and Imbusch, G.F. (1989). Optical Spectroscopy of Inorganic Solids. Oxford: Clarendon Press. DiBartolo, B. (1968). Optical Interactions in Solids. New York: Wiley. Becker, P.C., Olsson, N.A., and Simpson, J.R. (1999). Erbium Doped Fiber Amplifiers. Fundamentals and Technology. San Diego, USA: Academic Press, Chapter 4. Hüfner, S. (1978, 1978). Optical Spectra of Rare Earth Compounds. New York: Academic Press.

References

30 Miniscalco, W.J. (1993). Rare-Earth-Doped Fiber Lasers and Amplifiers, Second,

Revised and Expanded (ed. M.J. Digonnet). New York: Marcel Dekker. Chapter 2. 31 Carnall, W.T., Goodman, G.L., Rajnak, K., and Rana, R.S. (1989). J. Chem. Phys. 90:

3443. 32 Dieke, G.H. (1968). Spectra and Energy Levels of Rare Earth Ions in Crystals. 33 34 35 36 37 38 39 40 41 42 43 44 45 46

47

48 49 50 51 52 53 54 55 56 57 58 59 60

New York: Wiley. Judd, B.R. (1962). Phys. Rev. 127: 750. Ofelt, G.S. (1962). J. Chem. Phys. 37: 511. Weber, M.J. (1967). Phys. Rev. 157: 262. Carnall, W.T., Field, P.R., and Wybourne, B.G. (1968). J. Chem. Phys. 49: 4424. Shortley, G.H. (1940). Phys. Rev. 57: 225. Jorgensen, C.K. and Reisfeld, R. (1983). J. Less Common Metals 93: 107–112. Soundararajan, G., Koughia, C., Edgar, A. et al. (2011). J. Non-Cryst. Solids 357: 2475. Zou, X. and Izumitani, T. (1993). J. Non-Cryst. Solids 162: 68. Wetenkamp, L., Weast, G.F., and Tobben, H. (1992). J. Non-Cryst. Solids 140: 35–40. Tanabe, S., Ohyagi, T., Hanada, T., and Soga, N. (1993). J. Ceram. Soc. Jpn. 101: 73. Ninga, D., Iv-yun, Y., Ming-ying, P. et al. (2006). Mater. Lett. 60: 1987. Koughia, K., Saitou, D., Aoki, T. et al. (2006). J. Non-Cryst. Solids 352 (23). Munzar, M., Koughia, C., Tonchev, D. et al. (2006). Opt. Mater. 28: 225. Gschneidner, K.A. Jr., and Eyring, L.R. (eds.) (1998). Handbook on the Physics and Chemistry of Rare Earths, vol. 25. Amsterdam, Lausanne, New York, Oxford, Shannon, Singapore, Tokyo: Elsevier. Walsh, B. (2006). Judd–Ofelt Theory principles and practices I. In: Advances in Spectroscopy for Lasers and Sensing (eds. B. Di Bartolo and O. Forte). Dordrecht: Springer. Quimby, R.S. and Miniscalo, W.J. (1994). J. Appl. Phys. 75: 613. Tanabe, S., Ohyagi, T., Soga, N., and Hanada, T. (1992). Phys. Rev. B 46: 3305. Tanabe, S. (2015). Int. J. Appl. Glass Sci. 6 (4, Special Issue: Glass and Light – Part II): 305. Tanabe, S. (2002). C. R. Chimie 5: 815. Goldner, P. and Auzel, F. (1996). J. Appl. Phys. 79: 7971. Hehlen, M.P., Brok, M.G., and Kramer, K.W. (2013). J. Luminescence 136: 221. Adam, J.-L. (2002). Chem. Rev. 102: 2461. Ebendorff-Heidepriem, H., Ehrt, D., Bettinelli, M., and Speghini, A. (1998). J. Non-Cryst. Solids 240: 66. T. Schweizer 1998. Rare-Earth-Doped Gallium Lanthanum Sulphide Glasses for Mid-Infrared Fibre Lasers, PhD, Southampton, UK and Universität Hamburrg. Yariv, A. (1985). Optical Electronics, 3e. New York: Holt Rinehart and Winston Inc Ch. 9. Yariv, A. (1997). Optical Electronics in Modern Communications, 5e. Oxford: Oxford University Press. Davis, C.C. (1996). Laser and Electro-Optics. Cambridge: Cambridge University Press Chs. 18, 19. W. Franz, Z. Naturforsch (1958). 13a, 484

63

64

2 Fundamental Optical Properties of Materials II

61 Keldysh, L.V. (1958). Zh. Eksperim. i Teor. Fiz. 34: 1138; [English transl.: Soviet 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78

79

80 81 82

83 84 85 86 87 88 89

Phys.—JETP 7, 788 (1958)]. Böer, K.W. (1969). IBM J. R&D 13: 573. Seraphin, B.O. and Hess, R.B. (1965). Phys. Rev. Lett. 14: 138. Frova, A., Handler, P., Germano, F.A., and Aspnes, D.E. (1966). Phys. Rev. 145: 575. Srivastava, A., Srivastava, R., Wang, J., and Kono, J. (2004). Phys. Rev. Lett. 93: 157401. Seraphin, B.O. and Bottka, N. (1965). Phys Rev. 139: A565. Sie, E.J., McIver, J.W., Lee, Y.-H. et al. (2016). Proc. SPIE 9835: 983518. Kasap, S.O. (2012). Optoelectronics and Photonics: Principles and Applications, 2e. Upper Saddle River, NJ, USA: Pearson Education, Ch. 6. Apnes, D.E. (1982). Thin Solid Films 89: 249. Sihvola, A.H. and Kong, J.A. (1988). IEEE Trans. Goesci. Remote Sens. 26: 420 m. Bergman, D.J. and Stroud, D. (1992). Solid State Phys. 46: 147. Nan, C.-W. (1993). Prog. Mater. Sci. 37: 1. Sihvola, A. (1999). Electromagnetic Mixing Formulas and Applications. London, UK: IET (ISBN: 978-0-85296-772-0). Sihvola, A. (2000). Subsurf. Sens. Technol. Appl. 1: 393. Sihvola, A.H. (2002). IEEE Trans. Goesci. Remote Sens. 40: 880. Voshchinnikov, N.V., Il’in, V.B., and Henning, T. (2005). Astron. Astrophys. 429: 371. Liu, Y. and Daum, P.H. (2008). J. Aerosol Sci. 39: 974. Scheller, M., Jansen, C., and Koch, M. (2010). Chapter 12: Applications of effective medium theories in the terahertz regime. In: Optical and Photonic Technologies (ed. K.Y. Kim), 231–250. Croatia: Intech. Humlicek, J. (2013). Data analysis for Nanomaterials: effective medium approximation, its limits and implementations. In: Ellipsometry at the Nanoscale (eds. M. Losurdo and K. Hingerl). Berlin, Heidelberg: Springer. Carvalho Araújo, M., Costa, C.M., and Lanceros-Méndez, S. (2014). J. Non-Cryst. Solids 387: 6. Markel, V. (2016). J. Opt. Soc. Amer. A33: 1244. M. Losurdo, D. E. Aspnes, G. Bruno, E. Irene, K. Hingerl, J. Humlicek Eds (2010). Defining and Analysing the Optical Properties of Materials at the Nanoscale A collection of thoughts, opinions, ideas and data that have matured over years on exploiting ellipsometry for a range of characterisation needs, ManoCharM (A NanoCharM Publication, www.nanocharm.org). Chylek, P. and Videen, G. (1998). Opt. Commun. 146: 15. Lorentz, H.A. (1880). Ueber die Beziehung zwischen der Fortpflanzungsgeschwindigkeit des Lichtes und der Körperdichte. Ann. Phys. 245: 641. Lorenz, L. (1880). Ueber die Refractionsconstante. Ann. Phys. 247: 70. Chen, H., Saunders, J.E., Borjian, S. et al. (2019). Adv. Sustain. Syst. 3: 1800084. https://doi.org/10.1002/adsu.201800084. Bruggeman, D.A.G. (1935). Ann. Phys. (Leipzig) 416: 636. Khardani, M., Bouaïcha, M., and Bessaïs, B. (2007). Phys. Stat. Sol. C 4: 1986. Sohn, H. (2014). Refractive index of porous silicon. In: Handbook of Porous Silicon (ed. L. Canham), 1–12. Switzerland: Springer https://doi.org/10.1007/978-3-31904508-5_25-1.

References

90 Kim, S., Kadam, M., Kang, J.-H., and Ryu, S.-W. (2016). Electron. Mater. Lett.

12: 596. https://doi.org/10.1007/s13391-016-6028-y. 91 Maxwell Garnett, J.C. (1904). Philos. Trans. Royal Soc. 203: 385. and 205, 237

(1906). Polder, D. and van Santen, J.H. (1946). Physica 12: 257. Jones, S.B. and Friedman, S.P. (2000). Water Resour. Res. 36: 2821. Giordano, S. (2003). J. Electrost. 58: 59. Sushko, M.Y. (2009). J. Phys. D 42: 155410. Salski, B. (2012). Prog. Electromagn. Res. Lett. 30: 173. Koledintseva, M.Y., DuBroff, R.E., and Schwartz, R.W. (2006). Prog. Electromagn. Res. 63 (223). 98 Mirkhani, V., Tong, F., Song, D. et al. (2016). J. Nanosci. Nanotechnol. 16: 7358. 99 Looyenga, H. (1965). Physica 31: 401. 100 Monecke, J. (1994). J. Phys. CM 6: 907. 92 93 94 95 96 97

65

67

3 Optical Properties of Disordered Condensed Matter Koichi Shimakawa 1 , Jai Singh 2 , and S.K. O’Leary 3 1 2 3

Department of Electrical and Electronic Engineering, Gifu University, Gifu 501-1193, Japan College of Engineering, IT and Environment, Purple 12, Charles Darwin University, Darwin, NT 0909, Australia School of Engineering, The University of British Columbia, Kelowna, Canada

CHAPTER MENU Introduction, 67 Fundamental Optical Absorption (Experimental), 69 Absorption Coefficient (Theory), 74 Compositional Variation of the Optical Bandgap, 79 Conclusions, 80 References, 80

3.1 Introduction In a defect-free crystalline semiconductor, there exists a well-defined energy gap between the valence and conduction bands. In contrast, in an amorphous semiconductor, the distributions of conduction- and valence-band electronic states do not terminate abruptly at the band edges. Instead, some of the electronic states, referred to as tail states, encroach into the otherwise empty gap region [1]. In addition to tail states, there are other localized states deep within the gap region [2]. These localized tail states in amorphous semiconductors arise as a consequence of defects. The defects in amorphous semiconductors are considered to be all cases of departure from the normal nearest-neighbor coordination (or normal valence requirement). Examples of defects are broken and dangling bonds (typical for amorphous silicon), over- and under-coordinated atoms (such as “valence alternation pairs” in chalcogenide glasses), voids, pores, cracks, and other macroscopic defects. As these tail and deep defect states are localized, there exist mobility edges, which separate these localized states from their extended counterparts [3–5]. These localized tail and deep defect states are responsible for many of the unique properties exhibited by amorphous semiconductors. Despite years of intensive investigation, the exact form of the distribution of electronic states associated with amorphous semiconductors remains a matter for debate. While there are still some unresolved issues, there is general consensus that the tail states arise as a consequence of the disorder (weak and dangling bonds) present within the amorphous network, and that the breadth of these tails reflects the amount of disorder present Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

68

3 Optical Properties of Disordered Condensed Matter

[6]. Experimental results from Tiedje et al. [7] and Winer and Ley [8] suggest exponential distributions for the valence- and conduction-band tail states in hydrogenated amorphous silicon a-Si:H, although other possible functional forms [9] cannot be ruled out. Singh and Shimakawa [5, 10] have derived different effective masses for the charge carriers in their extended and tail states. This implies that the density-of-states (DOS) of the extended and tail states can be represented in two different parabolic forms. A representation of the current understanding of the distribution of electronic states, for the case of a-Si:H, is schematically presented in Figure 3.1 [5, 11–13]. The existence of tail states in amorphous solids has a profound impact upon the band-to-band optical absorption. Unlike the case of a crystalline solid, the absorption of photons in an intrinsic amorphous solid can also occur for photon energies below the optical gap, that is, ℏ𝜔 ≤ E0 , due to the presence of the tail states in the forbidden gap, E0 , denoting the optical gap, which is usually close to the mobility gap, that is, the energy difference between the conduction- and valence-band mobility edges. In a crystalline semiconductor, the energy and crystal momentum of an electron involved in an optical transition must be conserved. In an amorphous semiconductor, however, only the energy needs be conserved [4, 5]. As a result, for optical transitions caused by photons of energy ℏ𝜔 ≥ E0 in amorphous semiconductors, the joint density of states approach is not applicable [5, 14]. One must consider the product of densities of both the conduction and valence bands in calculating the corresponding absorption coefficient [5, 15]. Two approaches are presented here that are used for calculating the absorption coefficient in amorphous semiconductors. In the first approach, one assumes that the transition matrix element is independent of the photon energy. In the second approach, contrary to the first one, using the constant dipole approximation, the transition matrix element is found to be photon-energy dependent. Applying the first approach, one obtains the well-known Tauc’s relation for the absorption coefficient of amorphous semiconductors, that is, (𝛼ℏ𝜔)1/2 ∝ (ℏ𝜔 − E0 ), where 𝛼 is the absorption coefficient. However, applying the second approach, one obtains (𝛼/ℏ𝜔)1/2 ∝ (ℏ𝜔 − E0 ), that is, a slightly different functional dependence. The first approach has been used widely and successfully to interpret the experimental measurements corresponding to a wide variety of amorphous semiconductors of interest, some of which support Tauc’s relation, and others deviating from it. For the latter, as described in the following text, the concept of fractals and effective medium have been used to account for the deviations from the traditional Tauc relationship. In this chapter, we first review the application of the first approach and then consider those of the second. It will be shown that, through the first approach, both the fractal and

E2 Ec

Ect

Tail states Ev

Evt Ev2

Figure 3.1 A schematic illustration of the electronic energy states, E 2 , E c , E ct , E vt , E v , and E v2 , in amorphous semiconductors. The shaded region represents the extended states. Energies E 2 and E v2 correspond to the center of the conduction band and valence band extended states, E ct and E vt representing the end of conduction band and valence band tail states, respectively.

3.2 Fundamental Optical Absorption (Experimental)

effective medium theories are useful in explaining the optical properties of disordered forms of condensed matter.

3.2 Fundamental Optical Absorption (Experimental) Here we will consider the first approach, where it is assumed that the transition matrix element is independent of the photon energy. As the application of this approach to amorphous semiconductors, such as hydrogenated amorphous silicon (a-Si:H), is well known and can be found in many books [4, 5], its application to amorphous chalcogenides (a-Chs) and hydrogenated nanocrystalline silicon (a-ncSi:H) will be presented here. 3.2.1

Amorphous Chalcogenides

Analysis of the optical absorption spectra is one of the most useful tools for understanding the electronic structure of solids in any form, crystalline or amorphous. As found in a free electron gas, the DOS for both the conduction and valence bands are expected to be proportional to the square root of the energy in 3D materials [5]. Applying this to amorphous structures within the framework of the first approach leads to the well-known Tauc plot for the optical absorption coefficient 𝛼 as a function of photon energy ℏ𝜔, giving (𝛼ℏ𝜔)1/2 ∝ (ℏ𝜔 − E0 ) [16]. However, this quadratic energy dependence of the absorption coefficient on the photon energy is not always observed. For example, a linear energy dependence [(𝛼ℏ𝜔) ∝ (ℏ𝜔 − E0 )] for amorphous Se (a-Se) and a cubic energy dependence (𝛼ℏ𝜔)1/3 ∝ (ℏ𝜔 − E0 ) for multicomponent chalcogenide glasses have also been observed [4]. The deviations from the simple Tauc relation may be regarded as arising from the deviations of the DOS functions from a simple power law. The DOS in disordered matter, in general, may be described by taking into account the fractals that are known to dominate many physical properties in amorphous semiconductors [17, 18]. We revisit the classical problem for interpreting the optical properties of amorphous semiconductors on the basis of the form of DOS applicable to amorphous chalcogenides. We find that the fundamental optical absorption in amorphous chalcogenides can be written as ((𝛼ℏ𝜔)n ∝ (ℏ𝜔 − E0 )), where the value of n deviates from the Tauc value of 1/2 [19]. Typical examples are shown in Figures 3.2 and 3.3 for obliquely deposited amorphous As2 S3 and Se (hereafter a-As2 S3 and a-Se), respectively, which are given in the plot of (𝛼h𝜈)n vs h𝜈). The fitting to the experimental data for a-As2 S3 (Figures 3.2a,b) produces n = 0.70 before annealing (as deposited) and n = 0.59 after annealing near the glass transition temperature for the film. Note that similar values of n (0.73 for as deposited and 0.58 after annealing) are also obtained for a-As2 Se3 . The fitting to the data for a-Se (Figure 3.3) produces n = 1 before annealing (as deposited), and it remains unchanged after annealing at 30∘ C for 2 hours. Here, a brief derivation of the fundamental optical absorption coefficient is given in order to facilitate the interpretation of the experimental results given earlier. For inter-band electronic transitions, the optical absorption coefficient through the first approach can be written as [4]: 𝛼(𝜔) = B



Nv (E − ℏ𝜔)Nc (E)dE ℏ𝜔

(3.1)

69

3 Optical Properties of Disordered Condensed Matter

6 × 103 Before annealing n = 0.70

(𝛼hν)n [eV cm–1]n

5 × 103 4 × 103 3 × 103 2 × 103 1 × 103 0 26

27

28

29

30

hν [ev] (a)

(𝛼hν)n [eV cm–1]n

1500

After annealing n = 0.59

900

300 0 2.6

2.7

2.8

2.9

3.0

hν [ev] (b)

Figure 3.2 The optical absorption spectra of obliquely deposited a-As2 S3 , where (𝛼h𝜈)n vs h𝜈 is plotted: (a) n = 0.70 before annealing, and (b) n = 0.59 after annealing at 170∘ C for 2 hours. Figure 3.3 The optical absorption spectra of obliquely deposited a-Se, where (𝛼h𝜈)n vs h𝜈 is plotted with n = 1, and it remains unchanged before and after annealing.

3 × 105 Before annealing n = 1 (𝛼hν)n [eV cm–1]n

70

2 × 105

1 × 105

0 2.0

2.1

2.2 hν [ev]

2.3

2.4

3.2 Fundamental Optical Absorption (Experimental)

where B is a constant which includes the square of the transition matrix element as a factor, and the integration is over all pairs of states in the valence N v (E) and conduction band states N c (E). If the density of states for the conduction and valence states is assumed to be Nc(E) = const (E − EC )s and Nv(E) = const (EV − E)p , respectively, then Eq. (3.1) produces [19]: 𝛼(𝜔)ℏ𝜔 = B′ (ℏ𝜔 − E0 )p+s+1

(3.2)

where B′ is another corresponding constant. This gives [𝛼(𝜔)ℏ𝜔]1∕n = B′

1∕n

(ℏ𝜔 − E0 )

(3.3)

where 1/n = p + s + 1. If the form of both N c (E) and N v (E) is parabolic, that is, p = s = 1/2 for 3D, then the photon-energy dependence of the absorption coefficient obtained from Eq. (3.3) becomes: [𝛼(𝜔)ℏ𝜔]1∕2 = B′

1∕2

(ℏ𝜔 − E0 )

(3.4)

which is the well-known Tauc’s relation for the absorption coefficient. Let us consider Eq. (3.3) and first discuss the simple case of a-Se, where n = 1 is obtained. In this case, the sum of ( p + s) should be zero for n = 1. This is only possible if the product of the DOS functions is independent of the energy. The origin of such DOS functions was argued a long time ago, but was unclear. A chain-like structure is basically expected in a-Se. The top of the valence states is known to be formed by p-lone pair (LP) orbitals (lone-pair interaction) of Se atoms. The interaction between lone-pair electrons should be 3D in nature, and, therefore, the parabolic DOS near the valence band edge can be expected, that is, p = 1/2. The bottom of the conduction-band states, on the other hand, is formed by the anti-bonding states of Se atoms. If the interaction between chains is ignored, the DOS near the conduction-band states may be 1D in nature, that is, s = −1/2. Hence, we obtain n = 1/( p + s + 1) = 1, producing a linear dependence of energy, that is, (𝛼ℏ𝜔) ∝(ℏ𝜔 − E0 ) [19]. Next, we discuss As2 S(Se)3 binary systems. These systems are suggested to have layered structures. The tops of valence-band states are formed from the LP band, and, hence, the parabolic DOS near the top of the valence band can also be expected in these systems, since LP–LP interactions occur in 3D space, as was already mentioned. Unlike a-Se, however, the bottom of the conduction band arises from a 2D structure in nature, if the layer–layer interactions can be ignored for the anti-bonding states. This means that the corresponding DOS function is independent of energy (s = 0). The value of n, in this case, should be given by 2/3, since p = 1/2 and s = 0 are predicted from the argument of space dimensions, and it is close to those observed for as-deposited oblique films of a-As2 S(Se)3 . It should be noted, however, that the layer–layer interactions are not ignored in the DOS of the conduction band [20]. The deviations from n = 2/3 or n = 1/2 may be attributed to the fractional nature of the DOS functions, that is, p or s cannot be given only values such as 1/2 (3D), 0 (2D), and − 1/2 (1D). In obliquely deposited As2 Se3 , for example, n = 0.70 (before annealing) produces p + s = 0.43. In order to interpret this result, we may need to discuss the DOS for fractal structures. The DOS for the extended states with energy E on d-space dimension in the usual Euclidean space perspective is given as: N(E)dE ∝ 𝜌d−1 d𝜌

(3.5)

71

72

3 Optical Properties of Disordered Condensed Matter

where 𝜌 is defined as (2m∗e E)1∕2 ∕ℏ, instead of the wave vector k, which is not a good quantum number in disordered materials, m∗e being the electronic effective mass. In a fractal space, on the other hand, a fractal dimension, D, is introduced, instead of d. The DOS for the extended states in the fractal space D can be given by: N(E)dE ∝ 𝜌D−1 d𝜌 ∝ E

D−2 2

dE

(3.6)

Note that D is introduced as M(r) ∝ rD, where M is the “mass” in a space, and hence D can take any fractional value (even larger than 3). A similar argument of fractional dimensionality on interband optical transitions has also been presented in anisotropic crystals applicable to low-dimensional structures [21]. As we have discussed already, the energy dependence of the DOS for the conduction band is expected to be different from that for the valence band in amorphous chalcogenides, because usually the space dimensionality for the valence band is larger than that for the conduction band. Therefore, we introduce Dv and Dc for the dimensionality of the valence band and the conduction band, respectively. Then p + s + 1 in Eq. (3.2) is replaced by Dv + Dc − 2 (3.7) 2 for fractional-dimension systems. From the value of n, that is, 2/(Dv + Dc − 2), Dv + Dc can then be deduced. In As2 S3 , for example, Dv + Dc is 4.86. After thermal annealing, the value of n tends to 0.5 for both the oblique and flat samples, that is, Dv + Dc ≈ 6, indicating that each Dv and Dc approach three dimensions. This is due to the fact that thermal annealing produces a more ordered and dense structural network. A similar argument has also been given for flatly deposited materials. A cubic energy dependence, n = 1/3, is often observed for multi-component materials, such as Ge–As–Te–Si [4], which gives Dv + Dc = 8 according to the preceding analysis. This higher fractal dimension may be related to “branching” or “cross-linking” between Te chains by introducing As, Ge, and Si atoms. The “branching” may be equivalent to a “Bethe lattice” (or “Cayley tree”), resulting in an increase in the spatial dimensions [18]. In summary, the fundamental form of the optical absorption spectrum, empirically presented by the relation(𝛼ℏ𝜔)n ∝ (ℏ𝜔 − E0 ), where n deviates from 1/2 in amorphous chalcogenides, is interpreted by introducing the DOS of fractals. The energy-dependent DOS form is not the same for the conduction band and the valence band. The presence of disorder can greatly influence the nature of the electronic DOS even for the extended states. Accordingly, the concept of DOS on fractal structures has successfully been applied to interpret the fundamental optical absorption spectra. Finally, we should briefly discuss the validity of Eq. (3.1) itself in which the transition matrix element is assumed to be independent of energy, that is, B is independent of energy. Dersch et al. [22], on the other hand, suggested that the transition matrix element is energy dependent and has a peaked nature near the bandgap energy. The absorption coefficient with and without the energy-dependent matrix element will be briefly discussed in Section 3.3. p+s+1=

3.2.2

Hydrogenated Nano-Crystalline Silicon (nc-Si:H)

In this section, we discuss the optical properties of nanocrystalline Si prepared by the plasma-enhanced chemical vapor deposition (PECVD) method. Since this material

3.2 Fundamental Optical Absorption (Experimental)

𝛼 (cm–1)

Figure 3.4 A comparison of the optical absorption coefficient of a-Si:H, nc-Si:H, and c-Si. The EMA results are given by solid circles.

106

a-Si:H nc-Si:H

104

c-Si

EMA (Xc = 0.8)

102

100 1.2

1.6

2.0

2.4

2.8

Energy [eV]

consists of both amorphous and crystalline phases, its structure is very complex and nonuniform. Therefore, it may be expected that the optical absorption cannot be described by Tauc’s relation (Eq. (3.4)) [16]. One of the interesting properties these materials have is the excess of optical absorption in the fundamental absorption region. The optical absorption coefficient for nc-Si:H is larger than that for crystalline silicon (c-Si), as has been reported in the infrared to blue region. This points to an advantage of using nc-Si:H to fabricate solar cells, because many more photons can be absorbed in the films. An example of this difference is shown in Figure 3.4. Three solid lines represent the experimental data for a-Si:H, nc-Si:H, and c-Si. According to Figure 3.4, a-Si:H has the highest optical absorption coefficient and c-Si has the lowest in the energy range E > 1.7 eV. Therefore, a sample of nc-Si:H (a mixture of both) should be expected to have an absorption in between the two. However, how to calculate the absorption in hydrogenated microcrystalline silicon (μc-Si:H) is still a matter of debate. Although the scattering of light is suggested to be an origin for the enhanced optical absorption, Shimakawa [23] took the effective medium approximation (EMA), in an alternative way, to explain the excess of absorption. Here, first we briefly introduce the EMA before proceeding with the discussion on enhanced absorption within nc-Si:H. The EMA predicts that the total network conductance 𝜎 m for composite materials in D dimensions follows the following condition: ⟨ ⟩ 𝜎 − 𝜎m =0 (3.8) 𝜎 + (D − 1)𝜎m where 𝜎 is a random variable of conductivity and ⟨. . .⟩ denotes spatial averaging. Assuming that a random mixture of particles of two different conductivities, for example, one volume fraction, C, has conductivity of 𝜎 0 , and the remainder has conductivity of 𝜎 1 , 𝜎 1 being substantially less than 𝜎 0 , simple analytical expressions for the dc conductivity and the Hall mobility as a function of C have been derived (see the pioneering works by Kirkpatrik [24] and Cohen and Jortner [25]). EMA has also been extended to calculate the ac conductivity, in which case 𝜎 in Eq. (3.1) becomes a complex admittance, that is, 𝜎* = 𝜎 1 + i𝜎 2 [26]. As the dielectric constant, 𝜀* = 𝜀1 − i𝜀2 = 𝜎 2 /𝜔 − i𝜎 1 /𝜔, is closely related to 𝜎*, the optical absorption coefficient 𝛼(𝜔) can be calculated using 𝛼(𝜔) = 4𝜋𝜎(𝜔)/cn, where c is the speed of light and n is the refractive index [23] of the material. The results obtained from the EMA for the crystalline volume fraction X c = 0.8 and D = 3 are shown by open circles in Figure 3.4. The frequency (energy)-independent

73

74

3 Optical Properties of Disordered Condensed Matter

refractive index, n0 = 3.9 for c-Si and n1 = 3.2 for a-Si:H, which gives the square root of the corresponding real part of their optical dielectric constants, are used in the calculation. The calculated results agree very well with the experimental data, except at an energy of around 1.7 eV (see Figure 3.4). This suggests that a mean field constructed through a mixture of amorphous and crystalline states dominates optical absorption within nc-Si:H. It may be noted that the multiple light scattering suggested earlier seems to be not so important in this energy range.

3.3 Absorption Coefficient (Theory) In the previous section, we discussed the absorption coefficient in relation to the band-to band absorption under the assumption that the transition matrix element was independent of the absorbed photon energy. The constant coefficient B′ in Eqs. (3.3) and (3.4) was not derived, and it could only be determined from fitting of the experimental data. In this section, the theoretical derivation of the absorption coefficient from both approaches will be given. An effort is made to clearly identify the differences between the two approaches and their applications. The absorption coefficient in Eq. (3.1), derived from the rate of absorption, can be written as in reference [5]: )2 E𝜐 +ℏ𝜔 ( 2𝜋e 1 |pcv |2 Nc (E)Nv (E − ℏ𝜔)dE (3.9) 𝛼= ∫Ec ncV𝜔 m∗e where n is the real part of the refractive index and V is the illuminated volume of the material, and pcv is the transition matrix element of the electron–photon interaction between the conduction and valence bands. Thus, the integrand consists of a product of three factors, all three of which depend on the photon energy and integration variable of energy E. Therefore, it becomes very difficult to evaluate the integral analytically. For crystalline materials, pcv is evaluated under the dipole approximation, but for amorphous solids, the first approximation introduced was to consider pcv as being independent of the photon energy, ℏ𝜔 [4], as described in Section 3.2. Then Eq. (3.9) reduces to the form of Eq. (3.1) and the whole analysis of the absorption properties presented in Section 3.2 is based on this approach. However, the resulting constant coefficient, B′ , which is an unknown in Eqs. (3.3) and (3.4), can now be determined. For the constant pcv , a popular form used is that shown in reference [4]: pcv = −iℏ𝜋(L∕V )1∕2

(3.10)

where L is the average interatomic spacing in the sample. Assuming that both the valence band and conduction band DOS functions have square-root dependencies on energy (as in a free-electron gas), one gets Tauc’s relation given in Eq. (3.4), with B′ as given in reference [5]: )] [ ( )2 ( ∗ ∗ 3∕2 m ) L(m e e 1 h B′ = (3.11) nc𝜀0 m∗e 2 2 ℏ3 where m∗h is the effective mass of a hole in the valence band. The advantage of Eq. (3.11) is that, if one knows the effective masses, then the so called Tauc’s coefficient, B′ , can be determined theoretically without having to fit Eq. (3.4) to any experimental data. As

3.3 Absorption Coefficient (Theory)

stated in Section 3.2, the absorption coefficient of many amorphous semiconductors, including chalcogenides, fit to Eq. (3.4) very well, but not all. Deviations from Tauc’s relation have been observed, and if one assumes a constant transition matrix element, pcv , in Eq. (3.9), then the only way these deviations can be explained is from the deviations in the squared-root form of the density of states, N c and N v , as discussed in Section 3.2 for various examples of chalcogenides. For instance, some experimental data for a-Si:H fit much better to a cubic root relation given by a formula in reference [4]: (𝛼ℏ𝜔)1∕3 = C(ℏ𝜔 − E0 )

(3.12)

and therefore the cubic root has been used to determine the optical gap E0 . Here, C is another constant. If one considers that the optical transition matrix element is photon-energy independent [5], one finds that the cubic root dependence on photon energy can be obtained only when the valence band and conduction band DOS depend linearly on energy. Using such DOS functions, the cubic root dependence has been explained by Mott and Davis [4]. Another approach to arrive at the cubic root dependence has been suggested by Sokolov et al. [27]. Using Eq. (3.12), they have modeled the cubic root dependence on photon energy by considering the fluctuations in the optical bandgap due to structural disorders. For arriving at the cubic root dependence, they finally assume that the fluctuations are constant over the range of integration, and then the integration of Eq. (3.12) over the optical gap energy produces a cubic root dependence on the photon energy. Although their approach shows a method of achieving the cubic root dependence, as the integration over the optical gap is carried out by assuming constant fluctuations, Sokolov et al.’s model is a little different from the linear DOS model suggested by Mott and Davis [4]. Let us now consider the second approach, where pcv is not assumed to be a constant. As stated earlier, for studying atomic and crystalline absorption, one uses the dipole approximation to evaluate the transition matrix element. This yields (see reference [5]): pcv = im∗e 𝜔reh

(3.13)

where reh is the average separation between the excited electron-and-hole pair. It may also be noted that, in the case where an electron-and-hole pair is excited by the absorption of a photon, m∗e should be replaced by their reduced mass, 𝜇, where 𝜇−1 = m∗(−1) + e mh∗(−1) . Let us use Eq. (3.13) for amorphous solids as well. Inserting Eq. (3.13) into Eq. (3.9), Cody [28] has derived the absorption coefficient in amorphous semiconductors. Accordingly, the absorption coefficient is obtained, as shown in reference [5]: [𝛼ℏ𝜔] = B′′ (ℏ𝜔)2 (ℏ𝜔 − E0 )2 where e2 B = nc𝜀0 ′′

[

(m∗e m∗h )3∕2 2𝜋 2 ℏ7 𝜈𝜌A

(3.14)

] 2 reh

(3.15)

Here, v denotes the number of valence electrons per atom, and 𝜌A represents the atomic density per unit volume. Eq. (3.14) suggests that [𝛼ℏ𝜔] depends on the photon energy in the form of a polynomial of order 4. Then, depending on which term of the polynomial may be more

75

76

3 Optical Properties of Disordered Condensed Matter

significant for which material, one can get square, cubic, fourth, or any other root of the dependence of [𝛼ℏ𝜔] on the photon energy. In this case, Eq. (3.14) may be expressed as: [𝛼ℏ𝜔]x ∝ (ℏ𝜔 − E0 )

(3.16)

where x ≤ 1/2. Thus, in a way, any deviation from the square root or Tauc’s plot may be attributed to the energy-dependent matrix element [5, 10]. However, this is a rather difficult issue to resolve unless one can determine the form of the DOS functions associated with the conduction and valence bands unambiguously. As a result, as the first approach of the constant transition matrix element has been successful for many samples over the last few decades, as has been demonstrated by many experimental groups, there appears to be a kind of prejudice in the literature in its favor. Let us now discuss how to determine the constants B′ in Eq. (3.11) and B′′ in Eq. (3.15), which involve the effective masses of electrons and holes, m∗e and m∗h , respectively. Recently, a simple approach [5, 29] has been developed to calculate the effective mass of the charge carriers in amorphous solids. Accordingly, different effective masses of the charge carriers are obtained in the extended and tail states. The approach applies the concepts of tunneling and effective medium, and one obtains the effective mass of an electron in the conduction band extended states, denoted by m∗ex , and in the tail states, denoted by m∗et , as in references [5, 29]: m∗ex ≈

EL m 2(E2 − Ec )a1∕3 e

(3.17)

m∗et ≈

EL m 2(Ec − Ect )b1∕3 e

(3.18)

and

where: EL = Here, a = N2 N

ℏ2 me L2

N1 N

(3.19)

< 1, N1 is the number of atoms contributing to the extended states,

< 1, N2 is the number of atoms contributing to the tail states, such that a + b = 1 b= (N = N1 + N2 ), and me is the free-electron mass. The energy E2 in Eq. (3.17) corresponds to the energy of the middle of the conduction extended states at which the imaginary part of the dielectric constant becomes a maximum and Ect is the energy corresponding to the end of the conduction tail states (see Figure 3.1). Likewise, the hole effective masses and m*hx in the valence extended and tail states are obtained, respectively, as: m∗hx ≈

EL m 2(Ev − Ev2 )a1∕3 e

(3.20)

m∗ht ≈

EL m 2(Evt − Ev )b1∕3 e

(3.21)

and

where Ev2 and Evt are the energies corresponding to the half width of valence extended states and the end of the valence tail states, respectively (see Figure 3.1).

3.3 Absorption Coefficient (Theory)

Table 3.1 The effective mass of electrons in the extended and tail states of a-Si:H and a-Ge:H, calculated using Eqs. (3.17) and (3.18) for a = 0.99, b = 0.01, and E ct = E vt = E c /2. E L is calculated from Eq. (3.18). All energies are given in eV. Note that, as the optical absorption coefficient is measured in cm−1 , the value for the speed of light is given in cm/s. L(nm)

E2

Ec

EL

a-Si:H

0.235a)

3.6b)

1.80c)

1.23

a-Ge:H

0.245a)

3.6

1.05d)

1.14

m∗ex

m∗et

0.9

0.34me

6.3me

0.53

0.22me

10.0me

E c –E ct

Note: a) Morigaki [30], b) Ley [31], c) Street [1], and d) Aoki et al. [32].

Using Eqs. (3.18) and (3.19) and the values of the parameters involved, different effective masses for an electron are obtained in the extended and tail states. Considering, for example, the density of weak bonds contributing to the tail states as 1 at.%, that is, b = 0.01 and a = 0.99, the effective mass and energy EL thereby calculated for hydrogenated amorphous silicon (a-Si:H) and germanium (a-Ge:H) are given in Table 3.1. For sp3 hybrid amorphous semiconductors, such as a-Si:H and a-Ge:H, the energies Ect and Evt can be approximated as: Ect = Evt = Ec /2. According to Eqs. (3.17), (3.18), (3.20), and (3.21), for sp3 hybrid amorphous semiconductors, such as a-Si:H and a-Ge:H, the electron and hole effective masses are expected to be the same. In these semiconductors, as the conduction and valence bands are two equal halves of the same electronic band, their widths are the same, and that gives equal effective masses for electron and hole [5, 29]. This is one of the reasons for using Ec t = Evt = Ec /2, which gives equal effective masses for electrons and holes in the tail states as well. This is different from crystalline solids, where m∗e and m∗h are usually not the same. This difference between amorphous and crystalline solids is similar to, for example, having direct and indirect crystalline semiconductors but only direct amorphous semiconductors. Using the effective masses from Table 3.1 and Eq. (3.15), B′ can be calculated for a-Si:H and a-Ge:H. The values thus obtained with the refractive index n = 4 for a-Si:H and a-Ge:H are B′ = 6.0 × 106 cm−1 eV−1 for a-Si:H and B′ = 4.1 × 106 cm−1 eV−1 for a-Ge:H, which are an order of magnitude higher than those estimated by fitting the experimental data [4]. However, considering the quantities involved in B′ (Eq. (3.11)), this can be regarded as a good agreement. In a recent paper, Malik and O’Leary [33] have studied the distributions of conduction and valence-band electronic states associated with a-Si:H. They have noted that the effective masses associated with a-Si:H are material parameters which are yet to be experimentally determined. In order to remedy this deficiency, they have fitted square-root DOS functions to experimental DOS data and found m∗h = 2.34 me and m∗e = 2.78 me . The value of the constant B′′ in Eq. (3.15) can now be calculated theoretically, provided reh is known. Using the atomic density of crystalline silicon and four valence electrons 2 2 = 0.9 Å , which gives reh ≈ 0.095 nm, less than half per atom, Cody [28] has estimated reh

77

78

3 Optical Properties of Disordered Condensed Matter

of the interatomic separation of 0.235 nm in a-Si:H, but of the same order of magnitude. 2 2 = 0.9Å , and extended-state effective masses, we get Using v = 4, 𝜌A = 5 × 1028 m−3 , reh ′′ 3 −1 −3 B = 4.6 × 10 cm eV for a-Si:H and 1.3 × 103 cm−1 eV−3 for a-Ge:H. Cody has estimated an optical gap, E0 = 1.64 eV, for a-Si:H, from which, using Eq. (3.14), we get 𝛼 = 1.2 × 103 cm−1 at a photon energy of ℏ𝜔= 2 eV. This agrees reasonably well with the value 𝛼 = 6.0 × 102 cm−1 used by Cody. If we use the interatomic spacing, L, in place of reh in Eq. (3.14), we get B′′ = 2.8 × 104 cm−1 eV−3 , and then the corresponding absorption coefficient becomes 3.3 × 103 cm−1 . This suggests that, for an estimate, one may use the interatomic spacing in place of reh , if the latter is unknown. Thus, both the constants B′ and B′′ can be determined theoretically, a task not possible earlier due to our lack of knowledge regarding the effective masses in amorphous semiconductors. The absorption of photons of energy less than the band gap energy, ℏ𝜔 < E0 , in amorphous solids involves the localized tail states, and hence follows neither Eq. (3.4) nor Eq. (3.12). Instead, the absorption coefficient depends on the photon energy exponentially, as shown in Chapter 2, giving rise to an Urbach tail. Abe and Toyozawa [34] have calculated the interband absorption spectra in crystalline solids, introducing the Gaussian site diagonal disorder and applying the coherent potential approximation. They have shown that an Urbach tail occurs due to static disorder (structural disorder). However, the current stage of understanding is that Urbach’s tail in amorphous solids occurs due to both thermal and structural disorders [28]. More recent issues in this area have been addressed by Orapunt and O’Leary [35]. Keeping this discussion in mind, the optical absorption in amorphous semiconductors near the absorption edge is usually characterized by three types of optical transitions corresponding to transitions between tail and tail states, tail and extended states, and extended and extended states. The first two types correspond to ℏ𝜔 ≤ E0 , while the third corresponds to ℏ𝜔 ≥ E0 . Thus, the absorption coefficient vs photon energy (𝛼 vs. ℏ𝜔) dependence has the corresponding three different regions A, B, and C that correspond to these three characteristic optical transitions shown in Figure 3.5. In the small optical absorption coefficient range A (also called the weak absorption tail [WAT]), where 𝛼 < 10−1 cm−1 , the optical absorption is controlled by optical transitions from tail-to-tail states. As stated earlier, the localized tail states in amorphous semiconductors arise from defects. To some extent, the absolute value of the absorption in region A may be used to estimate the density of defects in the material. In region B, where typically 10−1 cm−1 < 𝛼 < 104 cm−1 , the optical absorption is related to transitions from the localized tail states above the valence band edge to extended states in the conduction band, and/or from extended states in the valence band to localized tail states below the conduction band. Usually, the spectral dependence in this region follows the so-called Urbach dependence, given in Eq. (2.15) of Chapter 2. For many amorphous semiconductors, ∇E has been related to the breadth of the valence- or conduction-band tail states, and may be used to compare the “breadth” of such localized tail states in different materials; ∇E typically ranges from 50 to 100 meV for the case of a-Si:H. In region C, the optical absorption is controlled by optical transitions from extended states to the extended states. For many amorphous semiconductors, the 𝛼 vs. ℏ𝜔 behavior follows the traditional Tauc relation given in Eq. (3.4). The optical bandgap, E0 , determined for a given material from the 𝛼 vs. ℏ𝜔 relations obtained in Eqs. (2.15) (Chapter 2), (3.4) and (3.12), can vary, as shown in Table 3.2 for a-Si:H alloys.

3.4 Compositional Variation of the Optical Bandgap

𝛼 (cm–1) Tauc absorption

105

(ћω𝛼)1/2

104

C

103 102

Tauc absorption

Urbach edge

101 1 10–1 10–2 10–3

B WAT A

ћω

ћω Eo

Eg

03

Figure 3.5 Typical spectral dependence of the optical absorption coefficient in amorphous semiconductors. In the A and B regions, the optical absorption is controlled by the optical transitions between tail and tail, and tail and extended states, respectively; in the C region, it is dominated by transitions from extended to extended states. In domain B, the optical absorption coefficient follows Urbach rule (see Eq. (2.15) in Chapter 2). In region C, the optical absorption coefficient follows Tauc’s relation (Eq. (3.4)) in a-Si:H, as shown on the right-hand-side figure. Source: Reproduced by permission of Professor S. Kasap. Table 3.2 The optical bandgap of a-Si1–x Cx :H films obtained from Tauc’s (Eq. (3.4)), Sokolov et al.’s (Eq. (3.12)), and Cody’s (Eq. (3.14)) relations (3.27). E g at 𝜶 = 103 cm–1

E g at 𝜶 = 104 cm–1

Eg = Eo (Tauc)

Eg = Eo (Cody)

Eg = Eo (Sokolov)

𝚫E meV

a-Si:H

1.76

1.96

1.73

1.68

1.60

46

a-Si0.88 C0.18 :H

2.02

2.27

2.07

2.03

1.86

89

3.4 Compositional Variation of the Optical Bandgap 3.4.1

In Amorphous Chalcogenides

In Sections 3.2 and 3.3, the behavior of the optical absorption coefficient vs photon energy has been discussed. In this section, we discuss the effect of the compositional variation on the optical gap, E0 , of amorphous chalcogenide alloys. The bandgap varies with the composition and often exhibits extrema at certain stoichiometric compositions, for example, a minima in As2 S(Se)3 and a maximum in GeSe2 (see, e.g., ref. [36]). Applying the virtual crystal approach proposed by Phillips [37], Shimakawa has accounted for such compositional variations in the following form [36]: E0 = xE0 (A) + (1 − x)E0 (B) − 𝛾x(1 − x)

(3.22)

where A and B are composite elements in an Ax B1 − x alloy, and 𝛾 is referred to as the bowing parameter.

79

80

3 Optical Properties of Disordered Condensed Matter

Through the study of the optical gap in composite chalcogenides, amorphous chalcogenides can be classified into three types: (i) random bond network (RBN) type, (ii) chemically ordered bond network (CON) type, and (iii) indefinite network type. For the RBN type, A and B are taken as composite elements, for example, A = Sb and B = Se for Sbx Se1 − x . For the CON type, A and B are taken as the stoichiometric composition and the element is in excess, respectively—for example, A = As2 Se3 and B = Se in a Asx Se1 − x system. It is known that the optical gap of a-Gex Si1 − x : H can also be represented by Eq. (3.22) with 𝛾 = 0 [38]. Note also that the validity of Eq. (3.22), proposed by Shimakawa [36], has been confirmed in many a-Chs [39]. The main conclusions here are as follows: (i) the optical bandgap for amorphous semiconducting alloys is determined by the volume fraction and the optical gap of each element of the alloy, leading to the conjecture that a modified virtual crystal approach for mixed crystals is acceptable for amorphous systems; (ii) the classification into the three types presented earlier is supported by the effective medium approach (EMA), which has been applied to electronic transport in amorphous chalcogenides [40].

3.5 Conclusions The current understanding of the fundamental optical properties in some disordered semiconductors are briefly reviewed. Fundamental optical absorption in amorphous semiconductors, including the chalcogenides, cannot always be expressed by the well-known Tauc relationship. The deviation from the Tauc relationship has been discussed through two approaches—first with the energy-independent transition matrix element, and second with the energy-dependent matrix element. In the first approach, deviations from the Tauc relationship are overcome by introducing the density-of-states in fractal structures of amorphous chalcogenides. In the second approach, the deviations from Tauc’s relation are obtained through the use of the energy-dependent transition matrix element. This fractal nature may be attributed to the clustered layer structures in amorphous chalcogenides. The applicability of an EMA to the fundamental optical absorption spectra of nano-crystalline silicons is confirmed, and the effect of the compositional variation on the optical gap in amorphous semiconductors can also be connected to the validity of EMA.

References 1 Street, R.A. (1991). Hydrogenated Amorphous Silicon. Cambridge: Cambridge Uni-

versity Press. 2 Papaconstantopoulos, D.A. and Economou, E.N. (1981). Phys. Rev. B 24: 7233. 3 Cohen, M.H., Fritzsche, H., and Ovshinsky, S.R. (1969). Phys. Rev. Lett. 22: 1065. 4 Mott, N.F. and Davis, E.A. (1979). Electronic Processes in Non-crystalline Materials.

Oxford: Clarendon Press. 5 Singh, J. and Shimakawa, K. (2003). Advances in Amorphous Semiconductors. London

and New York: Taylor & Francis. 6 Sherman, S., Wagner, S., and Gottscho, R.A. (1996). Appl. Phys. Lett. 69: 3242.

References

7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23

24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

Tiedje, T., Cebulla, J.M., Morel, D.L., and Abeles, B. (1981). Phys. Rev. Lett. 46: 1425. Winer, K. and Ley, L. (1987). Phys. Rev. B 36: 6072. Webb, D.P., Zou, X.C., Chan, Y.C. et al. (1998). Solid State Commun. 105: 239. Singh, J. (2003). J. Mater. Sci. Mater. Electron. 14: 171. Jackson, W.B., Kelso, S.M., Tsai, C.C. et al. (1985). Phys. Rev. B 31: 5187. O’Leary, S.K., Johnson, S.R., and Lim, P.K. (1997). J. Appl. Phys. 82: 3334. Malik, S.M. and O’Leary, S.K. (2004). J. Non-Cryst. Solids 336: 64. Elliott, S.R. (1998). The Physics and Chemistry of Solids. Sussex: Wiley. Singh, J. (2002). Nonlinear Optics 29: 119. Tauc, J. (1979). The Optical Properties of Solids (ed. F. Abeles), 277. Amsterdam: North-Holland. Mandelbrot, B.B. (1982). The Fractal Geometry of Nature. New York: Freeman. Zallen, R. (1983). The Physics of Amorphous Solids, 135. New York: Wiley. Nessa, M., Shimakawa, K., Ganjoo, A., and Singh, J. (2000). J. Optoelectron. Adv. Mater. 2: 133. Watanabe, Y., Kawazoe, H., and Yamane, M. (1988). Phys. Rev. B 38: 5677. He, X.-F. (1990). Phys. Rev. B 42: 11751. Dersch, U., Grunnewald, M., Overhof, H., and Thomas, P. (1987). J. Phys. C Solid State Phys. 20: 121. (a) Shimakawa, K. (2000). J. Non-Cryst. Solids 266–269: 223. (b) Shimakawa, K. (2004). Encyclopedia of Nanoscience and Nanotechnology, vol. 4, 35. Valencia: American Scientific Publishes. (c) Shimakawa, K. (2004). J. Mater. Sci. - Mater. Electron. 15: 63. Kirkpatrick, S. (1973). Rev. Mod. Phys. 45: 574. Cohen, M.H. and Jortner, J. (1973). Phys. Rev. Lett. 30: 699. Springett, B.E. (1973). Phys. Rev. Lett. 31: 1463. Sokolov, A.P., Shebanin, A.P., Golikova, O.A., and Mezdrogina, M.M. (1991). J. Phys. Condens. Matter 3: 9887. Cody, G.D. (1984). Semiconductors and Semimetals B 21: 11. Singh, J., Aoki, T., and Shimakawa, K. (2002). Philos. Mag. B 82: 855. Morigaki, K. (1999). Physics of Amorphous Semiconductors. London: World Scientific. Ley, L. (1984, 1984). The Physics of Hydrogenated Amorphous Silicon II (eds. J.D. Joannopoulos and G. Lukovsky), 61. Berlin: Springer. Aoki, T., Shimada, H., Hirao, N. et al. (1999). Phys. Rev. B 59: 1579. (a) Malik, S.M. and O’Leary, S.K. (2005). J. Mater. Sci. Mater. Electron. 16: 177; (b) O’Leary, S.K. (2004). J. Mater. Sci.: Mater. Electron. 15: 401. Abe, S. and Toyozawa, Y. (1981). J. Phys. Soc. Jpn. 50: 2185. Orapunt, F. and O’Leary, S.K. (2004). Appl. Phys. Lett. 84: 523. Shimakawa, K. (1981). J. Non-Cryst. Solids 43: 229. Phillips, J.C. (1973). Bond and Band in Semiconductors. New York: Academic Press. Bauer, G.H. (1995). Solid-State Phenomena, vol. 365, 44–46. Tichy, L., Triska, A., Barta, C. et al. (1982). Philos. Mag. B 46: 365. Shimakawa, K. and Nitta, S. (1978). Phys. Rev. B 17: 3950.

81

83

4 Optical Properties of Glasses Andrew Edgar School of Chemical and Physical Sciences, Victoria University of Wellington, New Zealand

CHAPTER MENU Introduction, 83 The Refractive Index, 84 Glass Interfaces, 86 Dispersion, 88 Sensitivity of the Refractive Index, 90 Glass Color, 95 Fluorescence in Rare-Earth-Doped Glass, 102 Glasses for Fiber Optics, 104 Refractive Index Engineering, 106 Glass and Glass–Fiber Lasers and Amplifiers, 109 Valence Change Glasses, 111 Transparent Glass Ceramics, 114 Conclusions, 124 References, 124

4.1 Introduction Historically, glass was first valued for jewelry and decoration. As the last millennium developed, however, it became apparent that glass had two key attributes that made it an especially valuable material: (1) it could be worked into a variety of shapes by processes such as casting, drawing, molding, polishing, and blowing; and (2) it was transparent to visible radiation. This combination permitted the development of key technologies: manufacturing of containers with transparent walls, building and vehicle windows, and optical instruments and appliances such as spectacles, telescopes, cameras, and microscopes. It is difficult to appreciate from the context of the twenty-first century what a difference having a transparent, weather-proof, durable window material must have made to daily living in the Middle Ages, or to the ability to correct long- and short-sightedness with spectacles. In the twentieth century, the discovery of methods of making low-loss silica optical fiber revolutionized communications and computing technologies. In the twenty-first century, we can expect further developments based on these two key attributes of glass as a material: optical transparency (which can be

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

84

4 Optical Properties of Glasses

extended from the visible to the ultraviolet [uv] and the infrared [IR]), and the ability to shape or engineer glass into an arbitrary geometry or structure. This chapter first outlines the scientific and technical background of optical glass and its applications before presenting a discussion of some topics of current interest. Sections 4.2–4.5 discuss the complex refractive index, its dispersion, and factors that affect it, such as temperature, stress, chemical composition, and magnetic field. In Sections 4.6 and 4.7, we describe the origins of color in glass, and the effects of doping with open-shell ions from the 3d and 4f series of the periodic table, before outlining the fluorescence properties of rare-earth-doped glasses. In Section 4.8, we outline the optical fiber application of glass. Finally, in Sections 4.9–4.12, we describe work on refractive index micro-engineering in glasses, glass fiber lasers and amplifiers, and on transparent glass ceramics.

4.2 The Refractive Index Well-annealed glass is generally an optically homogeneous (on the scale of the wavelength of light), non-magnetic, and isotropic material. The behavior of an electromagnetic wave propagating through glass can be described by a complex refractive index n-ik, or equivalently a complex dielectric constant 𝜀′ − i𝜀′′ , which are related through (n − ik)2 = 𝜀′ − i𝜀′′

(4.1)

or equivalently 𝜀′ = n2 − k 2

(4.2)

𝜀′′ = 2nk

(4.3)

and In Figure 4.1, we show how the refractive index for a representative glass, silica, varies with wavelength, with data taken from a handbook of optical constants [1]. The key features are a broad region where n is almost (but not quite) wavelength-independent (the transmittance “window” of the glass), bounded by a sharp rise in k at short wavelengths (the “uv edge”), and a sharp rise in k at longer wavelengths (the IR edge). It is shown in many texts [2] that, for a plane monochromatic electromagnetic wave of frequency 𝜔, polarized in the x direction and propagating through a non-conducting glass in the z direction, the electric field strength Ex is given by, Ex (z) = E(0)exp(−kz𝜔∕c)exp[i𝜔(t − nz∕c)]

(4.4)

The wavelength in the glass is given by 𝜆 = 𝜆0 /n, where 𝜆0 is the vacuum wavelength, and the amplitude of the wave decreases exponentially in the propagation direction, with an intensity decaying as exp(−𝛼z), where 𝛼 = 2𝜔k/c is called the absorption coefficient. In Figure 4.2, we show the absorption coefficient in silica glass as a function of wavelength. Clearly, the abrupt rise in k at the two edges gives rise through 𝛼 = 2𝜔k/c to strong absorption (note the logarithmic scale in Figure 4.2) marking the perimeter of the transmission window. We also note that the well-known absorption features at 1.39 and 2.7 μm due to hydroxyl impurities in silica glass are too small to appear in Figure 4.2, the 1.39 μm absorption peak being of the order of 10−5 cm−1 in high-purity SiO2 . Although these bands are very weak, over a long path length of the order of kilometers, they and other effects (Section 4.8) have a dominant influence on the choice of wavelength used for fiber-optic communications within the broad silica window. The sharp rise in k also

4.2 The Refractive Index

3

2.5

n k

n,k

2

1.5

1

0.5

0 0.1

1 Wavelength (µm)

10

Figure 4.1 Complex refractive index (n, k) versus wavelength for silica glass SiO2 . 106

Absorption Coefficient (cm–1)

105

104

1000

100

10

1 0.1

1 Wavelength (µm)

10

Figure 4.2 Absorption coefficient plotted as a function of wavelength for silica SiO2 .

contributes to strong reflections from glass surfaces at the window boundaries. For normal incidence, the intensity reflection coefficient from an air–glass interface is given by | (n − 1)2 + k 2 | | R = || | | (n + 1)2 + k 2 | and, in Figure 4.3, we show the effect of the edges on the reflection coefficient.

(4.5)

85

4 Optical Properties of Glasses

1

0.1 Reflection Coefficient

86

0.01

0.001

0.0001 0.1

1 Wavelength (µm)

10

Figure 4.3 Reflection coefficient at normal incidence versus wavelength for an air–silica glass interface.

4.3 Glass Interfaces The complex refractive index of glasses is dominated in the window of transparency by the real part. In that case, the reflection and transmission coefficients for normal incidence at the interface between two glass media are readily calculated from Eq. (4.5) using the relative refractive index m = n2 /n1 for the refractive indices for light in the medium of incidence (n1 ) and transmission (n2 ); the result is shown in Figure 4.4. For light, which makes an angle of incidence 𝜃 𝜄 with the normal to the interface, the reflection and transmission coefficients depend upon the polarization of the incident light. Light with its electric field vector polarized in the plane defined by the incident, and reflected ray and the normal is referred to as p-polarized. Light with its electric field vector polarized perpendicular to that plane is called s-polarized. The reflection and transmission coefficients for intensity follow from Fresnel’s equations, and are given by [ ] −cos𝜃i + mcos𝜃t 2 , Ts = 1 − Rs (4.6) Rs = cos𝜃i + mcos𝜃t [ ] mcos𝜃i − cos𝜃t 2 , Tp = 1 − Rp (4.7) Rp = mcos𝜃i + cos𝜃t where m is the relative refractive index (assumed real) defined earlier, and 𝜃 t is the angle the transmitted ray makes with the normal, related to 𝜃 t through Snell’s law. Equations. (4.6) and (4.7) reduce to Eq. (4.5) for normal incidence, n1 = 1, and k = 0. The reflection and transmission coefficients are shown in Figure 4.5 for the case when the relative refractive index is 1.5. The reflection coefficient (Rp ) for the p-polarized wave

4.3 Glass Interfaces

Reflection (R) or Transmission (T) Coefficient

1

R T

0.8

0.6

0.4

0.2

0 10

1 Relative Refractive Index m = n2/n1

0.1

Figure 4.4 Reflection (R) and transmission (T) coefficients for intensity at normal incidence for light incident from medium 1 onto medium 2 plotted as a function of the relative refractive index.

Reflection (R) and Transmission (T) Coefficients

1.2

1

Rs Ts Rp Tp

0.8

0.6

0.4

0.2

0

0

20

40 60 Angle of Incidence (degrees)

80

Figure 4.5 Plot of reflection (R) and transmission (T) coefficients versus the angle of incidence in degrees for light incident from medium 1 onto medium 2 for a relative refractive index m = 1.5.

87

4 Optical Properties of Glasses

100

Brewster Critical Angle (degrees)

88

80

Brewster Angle Critical Angle

60

40

20

0 0.2

0.4

0.6 0.8 1 Relative Refractive Index m

3

5

Figure 4.6 Brewster and critical angles as a function of relative refractive index m.

goes to zero at the Brewster angle 𝜃 B , which from Eq. (4.7) and Snell’s law can be calculated to be tan 𝜃B = m

(4.8)

If the refractive index on the incident side is greater than that on the “transmitted” side of the interface, then the incident beam will be fully reflected from the interface if the angle of incidence exceeds the critical angle 𝜃 c where sin 𝜃c = m

(4.9)

In Figure 4.6, we show the Brewster and critical angles as a function of the relative refractive index. In many applications of glasses involving interfaces, these two angles play a crucial role. For example, most gas lasers use windows cut at the Brewster angle to eliminate reflection losses for one polarization. The critical angle is the defining parameter for the propagation of guided waves down fiber optic cables (strictly for cables whose diameter is much greater than the wavelength of light).

4.4 Dispersion The dispersion is the derivative of the refractive index with respect to wavelength, dn/d𝜆, and plays a critical role in the design of many optical instruments. A change in refractive index results in a change of focal length in lenses or deviation in prisms, and so a non-zero dispersion gives rise to chromatic aberrations (e.g. colored fringes of images) in optical systems. To describe the dispersion quantitatively, a figure of merit known as the Abbe number is generally used, and is given by: n −1 𝜈d = d (4.10) nF − nc

4.4 Dispersion

where nd is the refractive index at the helium d-line (587.6 nm), nF at the blue line of hydrogen (486.13 nm), and nC (656.27 nm) at the red line of hydrogen. We note that sometimes the sodium D-line at 589.6 nm is substituted for the helium line, in which case the Abbe number is denoted by vD . An alternative figure of merit, 𝜈 e , is based on the e-line of mercury (546.07 nm), and the F′ and C′ lines of cadmium (479.99 and 643.85 nm, respectively). The fraction (Eq. (4.10)) expresses the refracting power of the glass relative to the dispersion. The Abbe number is a basic parameter for optical instrument designers who have to deal with problems of chromatic aberrations, and is commonly presented as an “Abbe diagram,” where the Abbe number is plotted as a function of refractive index for different glass families. An Abbe diagram for oxide glasses, based on data for the Schott range [3], is shown in Figure 4.7. Usually, glasses with a large refractive index have high dispersion; it is the latter which dominates in Eq. (4.10), and so such glasses have a low Abbe number and appear in Figure 4.7 to the upper right. Glasses with small refractive indexes usually have low dispersion and high Abbe numbers, and so appear in the lower left of Figure 4.7. For achromatic lens design, a second parameter called the relative partial dispersion, Pg,F , defined as (ng – nF )/(nF – nC ), is also used to achieve a more comprehensive match of glass dispersion [3] for different optical elements. The refractive index ng is that at the blue line of mercury (435.83 nm). The Abbe number is an indirect measure of the first derivative of the refractive index, and the partial dispersion is an indirect measure of the second derivative. The refractive index nd , Abbe number 𝜈 d , and partial dispersion Pg,F have been historically used as a three-parameter summary of the refractive index and its dispersion, and are still widely used for comparing different glasses, but for analytical work it is advantageous to return to the refractive index itself. Glass manufacturers generally measure the refractive index at a variety of wavelengths corresponding to atomic 2

1.9

nd

1.8

BAF/BASF/BALF/KZF SF/F/LF LAF/LASF/LAFN LAK SK/SSK K/BAK/ZK PK/FK

1.7

1.6

1.5

1.4 90

80

70

60

50

40

30

20

vd

Figure 4.7 Abbe diagram for common optical glasses. Note that the Abbe number increases from right to left. The boundary at 𝜈 d = 50 divides crown glasses (left) from flint glasses (right).

89

90

4 Optical Properties of Glasses

emission lines and fit the data to a curve, which represents the sum of classical damped oscillator responses to an electromagnetic field. Assuming several such oscillators are responsible for the absorption process near the absorption edge of a material, the refractive index is often written as: ∑ B j 𝜆2 (4.11) n2 − 1 = 2 2 j (𝜆 − 𝜆j ) with the expectation that the formula (known as the Sellmeier formula) would only be a good representation for n for wavelengths well away from the resonance wavelengths 𝜆j . In that case, Eq. (4.11) can also be expressed as an (even) series in inverse powers of 𝜆 by expanding the functions ( )2 ( )4 𝜆j 𝜆j 𝜆2 + −… (4.12) = 1 + 𝜆 𝜆 (𝜆2 − 𝜆2j ) so that, n2 − 1 = A + A1 𝜆2 + A2 𝜆−2 + A3 𝜆−4 + A4 𝜆−6 + A5 𝜆−8 + …

(4.13)

Here, the notation follows that used by Schott [3], and the 𝜆 term (with a negative coefficient) is added to improve the fit toward the IR region of the spectrum. Manufacturers will generally quote the coefficients Ak , or Bj and C j = (𝜆j )2 , so that the refractive index can be computed for any wavelength, or they may simply quote the values of n at several wavelengths. The accuracy of Eq. (4.11) for interpolating the refractive index in the visible region is good to within about 1 part in 105 for a six-parameter fit. Clearly, the Abbe number, being defined in terms of visible wavelengths, is of less relevance in the IR range, and so equivalent alternatives are used for IR-transmitting glasses. The refractive index and Abbe number are used in a common commercial numeric designation of optical glasses, which summarizes their properties in a six- or nine-digit code. For example, a glass described as 805 254 would have a refractive index nd of 1.805 and an Abbe number of 25.4. Schott appends the density in g cm−3 to give a nine-digit code. The chemical composition is indicated by a letter code, which commonly contains “K” for crown glasses or “F” for flint glasses. Historically, crown glasses with a low dispersion were based on SiO2 –CaO–Na2 O, while flint glasses were based on SiO2 –PbO. The current spectrum of glasses contains many other combinations of elements, but they are still labeled as “low-dispersion crown glasses” if the Abbe number is greater than 50, and “high-dispersion flint glasses” if the Abbe number is lower than 50. There are few systematics to the rest of the letter labeling system, and individual manufacturers have their own systems. Nevertheless, as an example, the common Schott BK-7 borosilicate glass is a crown K glass; here, the “B” indicates the boron content, whereas the “7” is a manufacturer’s running number. 2

4.5 Sensitivity of the Refractive Index 4.5.1

Temperature Dependence

The refractive index is a function of variables other than wavelength, notably temperature. From the Sellmeier formula, assuming only a single resonance in the uv region, the

4.5 Sensitivity of the Refractive Index

refractive index variation with temperature may be parameterized as: ) ( )( E n(𝜆, T0 )2 − 1 + 2E ΔT dn 0 1 = D0 + 2D1 ΔT + 3D2 ΔT 2 + dt 2n(𝜆, T0 ) 𝜆2 − 𝜆20

(4.14)

where T 0 is a reference temperature, and the D, E, and 𝜆0 values are experimentally determined parameters. The significance of the temperature dependence of the refractive index lies in the implications for precision refractive optics. It is also important in high-powered laser applications where defocusing can occur in glass materials that have a significant absorption coefficient at the laser wavelength, and so show substantial localized heating in the focal zone. It is also important for applications such as glass etalons where temperature-independent conditions for interference are desirable. The refractive index in a sample of bulk glass must be homogeneous for many optical applications, but inhomogeneities known as Striae can arise during the casting process, arising from convective flow and differential cooling rates. Entrapped gas bubbles, undissolved crystalline inclusions, platinum particles from the melt crucible, and crystals arising from mild devitrification all give rise to light scattering, which is also detrimental to glass performance. The catalogs of glass manufacturers can be consulted for specifications describing Striae and scattering. 4.5.2

Stress Dependence

The refractive index is a function of the stress on a glass sample. The stress may be of internal origin, such as, for example, after a poorly executed annealing process, or external. Sometimes, applied external stress is an undesired consequence of an experimental situation, as in windows for cryogenic apparatus, which are subject to stress from the thermal contraction of the support structure. The key attribute of stress is that it makes the material optically birefringent—that is, the refractive index depends upon the polarization of the optical electromagnetic (EM) field. Since a stress field is described by a tensor, the birefringence is a tensorial quantity; however, in the simple case of uniaxial stress 𝜎, the refractive indices for light polarized parallel and perpendicular to the stress direction may be written as n∥ = n +

dn∥ d𝜎

𝜎

(4.15)

and dn⟂ 𝜎 (4.16) d𝜎 The derivatives in Eqs. (4.15) and (4.16) are often referred to as the photo-elastic coefficients. An alternative pair of birefringence parameters can be defined in terms of strain and related to those discussed earlier [3] through the theory of elasticity. Birefringence polarization modulators based on exciting a block of silica glass into longitudinal resonant oscillation at ultrasonic frequencies are available commercially. A three-component resonator system that uses a feedback loop to keep the oscillation amplitude constant, based on the design first used for internal friction measurements by Robinson and Edgar [4], is used commercially in ellipsometers. n⟂ = n +

91

92

4 Optical Properties of Glasses

4.5.3

Magnetic Field Dependence—The Faraday Effect

In the presence of a magnetic field, normally optically isotropic materials such as glasses become birefringent. The refractive index depends on the relative orientation of the magnetic field, the propagation direction for the light, and the polarization. The simplest geometry to analyze and the most important in practical applications is that where the propagation direction of linearly polarized light is parallel with the magnetic field, and in this case the manifestation of the effect is that the plane of polarization is rotated through an angle 𝜃, proportional to the field strength H, after traversing a path length s through the glass. This effect is known as the Faraday effect, after its discoverer Michael Faraday. The proportionality constant is called the Verdet constant, V , and is defined through 𝜃 = VsH

(4.17)

Plane polarized light may be decomposed into two counter-rotating circularly polarized waves traveling in the same direction. The Faraday effect arises through a difference in the refractive indices n+ and n− for the right-hand and left-hand rotating components, respectively. The relationship in Eq. (4.17) follows from the more fundamental definition ) 2c VH (4.18) 𝜔 The Faraday effect has both diamagnetic and paramagnetic contributions; the former is present in all glasses, but it is small, almost temperature independent, and characterized by a positive value for V , while the latter only occurs when the glass contains paramagnetic ions, is relatively large, negative, and has a large temperature coefficient. Both effects are strongly wavelength dependent. The diamagnetic effect arises from the Zeeman splitting of excited states. For example, the splitting of an excited state which is twofold degenerate results in a relative shift of the two associated transitions from the ground state; each transition is allowed only for one or other beam polarization. From Eq. (4.11), the value of 𝜆j for each transition is different, and so therefore are the refractive indices n+ and n− . The general form of the wavelength and temperature dependence of the Verdet constant is generally written [5] in terms of the frequency of the transition 𝜈, the number of ions per unit volume N, and the absolute temperature T, as: [ ] ∑ B(n, g) C(n, g) A(n, g) 2 V = 4𝜋N𝜈 + (4.19) + 2 2 2 2 (𝜈 2 − 𝜈n,g ) (𝜈 2 − 𝜈n,g ) T(𝜈 2 − 𝜈n,g ) n,g (

n− − n+ =

where the sum runs over all the states g of the ground multiplet and excited state levels n between which transitions are possible. The T −1 dependence for the C term reflects the temperature dependence of the level populations within the ground multiplet, which mirrors that of the Curie law for magnetic susceptibility, and is a reasonable approximation near room temperature. However, deviations from Curie’s law, arising from crystal field splittings in the ground multiplets, are often observed for rare-earth ions [6], and the same is also true for Verdet constants. In Figure 4.8, we show the (reciprocal) Verdet constant for a phosphate glass [5] plotted as a function of temperature; the relationship may be fitted with a Curie–Weiss law which shows that V varies with temperature T as (T + 65 K)−1 , not as T −1 . Van Vleck and Hebb [7] proposed that Faraday rotation should

4.5 Sensitivity of the Refractive Index

3.5

–1/V (Oe cm/min)

3 2.5 2 1.5 1 0.5 0

0

50

100

150 200 Temperature (K)

250

300

350

Figure 4.8 Reciprocal of the Verdet constant for Ce3+ ions in a phosphate glass using data taken from Borelli [5]. The line is a least squares fit to the Curie Weiss law, which gives a variation as (T + 65 K)−1 .

be proportional to magnetic susceptibility, implying that their temperature dependencies should be the same even if they are not Curie-like, but a test of this hypothesis by Edgar et al. [8] in the case of Ce3+ ions in fluoride glasses found significant deviations, which were explained in terms of an inappropriate assumption in Van Vleck and Hebb’s analysis—that the excited state splitting of the Ce3+ ions was relatively small. Nonetheless, the temperature dependence of the magnetic susceptibility is a good indicator of what to expect for the Faraday rotation. The experimental wavelength dependence of the Faraday rotation for undoped and rare-earth-doped ZBLAN fluorozirconate glass [8] is shown in Figure 4.9 and illustrates several aspects of Eq. (4.19). (Members of the fluorozirconate glass family are commonly labelled by an acronym comprising the chemical symbols for their constituent cations, so that ZBLAN glass comprises zirconium, barium, lanthanum, aluminum, and sodium fluorides). First, the glass itself has a small positive Verdet constant arising from 10 Verdet Constant (Rad/T.m)

Undoped 0 –10

1I

6

3P 3P0 1

1D

8% Pr 2

10% Ce

–20 –30 –40 400

450

500

550 600 650 Wavelength (nm)

700

750

800

Figure 4.9 Verdet constants for undoped and rare-earth-doped fluorozirconate glass recorded at room temperature as a function of wavelength. Source: Reproduced with permission from A. Edgar et al., J. Non-Cryst Solids, 231, 257 (1998).

93

94

4 Optical Properties of Glasses

optical transitions that make up the uv edge near 200 nm. For Ce3+ doping, the Verdet constant is large and negative, and shows the smooth wavelength dependence predicted by the C term in Eq. (4.19) that arises from the intense 4d-5f Ce3+ transitions in the uv around 270–350 nm. Finally, for Pr3+ , the main effect is again from the same 4d-5f transitions, which are further toward the uv region, but there are also weaker 4f-4f transitions in the visible region that cause deviations from Eq. (4.19). Both the deviation and the absorption at those wavelengths must be considered in any application. Yamane and Asahari [9] give a useful compilation of Verdet constants for diamagnetic and paramagnetic glasses. Diamagnetic glasses generally have smaller Verdet constants than paramagnetic ones, but they offer the advantage for practical applications that they have no absorption lines across the visible region of interest, and the Faraday rotation is nearly temperature independent. Paramagnetic glasses can have much larger Verdet constants, but they have a significant temperature dependence, and there can be absorption lines or bands in the area of interest. Glasses doped with 3d ions show strong broad absorption, and so are of little interest as Faraday rotation materials. For the rare earths, trivalent cerium and divalent europium do not generally have any absorption in the visible region, and have intense 4d-5f transitions in the near uv region, and so give a large Verdet constant. Terbium has a large ground state g-value, resulting in large ground state splittings, and so also has a large Verdet constant, but with the complication of weak 4f absorption lines in the visible region. Glasses that can accommodate a large fraction of terbium display a correspondingly large Verdet constant—for example, Suzuki et al. [10] report a borate glass containing 60 mol% Tb2 O3 with a Verdet constant of −234 rad/Tm at 633 nm. The very high fraction of terbium was achieved through containerless processing to inhibit crystallization. However, for fiber applications, a lower terbium fraction is necessary to avoid crystallization during drawing, and so smaller Verdet constants result: for example, Sun et al. [11] describe an all-fiber magnetic field sensor system based on terbium-doped silicate glass fiber with a Verdet constant of −24.5 rad/Tm at 1053 nm. The Faraday effect finds applications in optical isolators, modulators, and in electric current sensing. In the latter application [12], a fiber optic sensor is typically used to measure the large current flowing in overhead power lines. A few turns of the fiber are wrapped around the cable, and the current detected through the surface magnetic field it generates; the advantage over competing electrical means of current sensing is the optical isolation provided by the fiber optic cable from the high voltage. 4.5.4

Chemical Perturbations—Molar Refractivity

It has been known for centuries that the refractive index of a glass increases when heavy elements are added to it—heavy in the sense of those having a high atomic number, such as lead, lanthanum, or barium. These elements are commonly found in flint glasses. The underlying relationship between the bulk property and the atomic character can be discussed using the Clausius–Mossotti (or Lorentz–Lorenz) relationship, which is derived in most standard texts on electromagnetism (e.g. Griffiths [13]) 1 ∑ n2 − 1 N𝛼 (4.20) = 2 n + 2 3 𝜀0 i i i

4.6 Glass Color

where n is the refractive index, 𝜀0 is the permittivity of free space, and N i is the density of atoms or ions of atomic polarizability 𝛼 𝜄 . If the glass is composed of compounds whose refractive indices can be measured separately, then the Clausius–Mossotti relation implies that the glass refractive index can be computed from the “molar refractivities” defined as ( 2 ) nj − 1 MWj (4.21) MRj = 𝜌j n2j + 2 where MW j is the molecular weight of the compound and 𝜌j is the density, so that the glass refractive index can be computed as 𝜌glass ∑ n2 − 1 f MRj (4.22) = 2 n + 2 MWglass j j where f j is the mole fraction of compound j in the glass. In practice, the ions or atoms in a glass do not have a unique polarizability; for example, for oxygen ions, the polarizability depends on whether they are in bridging or non-bridging positions. Nonetheless, Eq. (4.22) does give a useful first-order guide as to the effect of different chemical perturbations on the refractive index.

4.6 Glass Color Glasses may appear colored in transmitted light due to a variety of mechanisms, including absorption by transition metal ions, band-edge cut-off, and colloidal precipitates. Band edge coloration simply implies that the fundamental edge for absorption has moved into the visible region of the spectrum, and many glasses based on anions other than oxygen, such as sulfur, selenium, and tellurium, qualify in this regard. Transition metal ion doping by elements from the 3d series with their open valence shell electronic structure results in absorption bands in the visible and infra-red regions arising from 3d-3d electronic transitions. In principle, rare-earth ion doping can also give rise to coloration, but the absorption per ion for the usual 4f-4f transitions is much weaker and narrower than for 3d-3d transitions. 4.6.1

Coloration by Colloidal Metals and Semiconductors

It has long been known that the addition of minute quantities of metals such as Cu, Ag, and Au to a glass can, under certain preparation conditions, give rise to a strong coloration. The classic example is “ruby glass,” which is a silicate glass containing colloidal gold. The effect arises from the strong perturbation that small particles of metal make to the propagation of EM waves through the glass, due to the huge difference in the dielectric constants of the glass and metal. The effect is to scatter and absorb the light, with the balance between these two effects, and the wavelength and angular dependence (for scattering) depending on the particle size and the wavelength. The theoretical basis for discussing the subject is Mie scattering theory, as described, for example, by Van der Hulst [14], Kerker [15], or Bohren and Hoffman [16]. The latter authors, for example,

95

96

4 Optical Properties of Glasses

show that the absorption and scattering cross-sections are, respectively, given for small spherical particles of radius a by ] [ kV 27 Cabs = (4.23) 𝜖 ′′ 3 (𝜖 ′ + 2)2 + 𝜖 ′′ 2 and [ ] k4V 2 27 2 (4.24) Cscat = |𝜀 − 1| 18𝜋 (𝜖 ′ + 2)2 + 𝜖 ′′ 2 where k is the wavenumber in the glass, 𝜀 = 𝜀′ − j𝜀′′ is the dielectric constant of the particle relative to the glass, and V (= 4 /3 𝜋a3 ) is the particle volume. The extinction coefficient is given by 𝛼 = N(Cscat + Cabs ),

(4.25)

where N is the particle concentration, and the light intensity I after traversing a thickness z of glass is I = I0 exp(−𝛼z).

(4.26)

The relationships in Eqs. (4.23) and (4.24) are only valid for particles which are small relative to the wavelength, in the sense that |𝜀ka| < 6 have matrix elements which are identically zero within the 4fn configuration, and so they are usually omitted from Eq. (4.27). For crystals, some of the remaining Bl m parameters are necessarily zero as a consequence of the crystal symmetry, but for glasses there is no such constraint, and furthermore the ligand field varies from site to site. The ligand field is a small perturbation as compared to the spin orbit or electrostatic energies, and this, together with the site-to-site distribution in ligand fields, means that resolved ligand field levels within an [S, L, J] multiplet are not generally observed in glasses, unlike in crystals. The effect is generally just to broaden the

97

4 Optical Properties of Glasses

Figure 4.10 Energy level diagram for Pr3+ ions.

4f2 levels

40

Energy (104cm–1)

98

20

4fn–1 5d1 and/or charge transfer levels

1I 6

1S

0

3P 2 3P 0 1D 2

3P 1

1G 3F 3

4 3F 4 3F 2 3H 6

3H 5

0

3H 4

transitions associated with each [S, L, J] free ion level, albeit with some unresolved structure. Thus, the absorption spectrum of any particular rare-earth ion does not vary substantially from one glass to another, because it is the unresolved ligand field splittings which are glass-specific. The absorption is also quite weak. This is also a direct result of the shielding effect of the outer 5s2 , 5p6 shells, which has two consequences with regard to transition intensities. Firstly, electric dipole transitions within a pure 4f configuration are strictly forbidden by Laporte’s parity rule [19, 20]. However, the odd parity (l odd) components in Eq. (4.27) have a special role in breaking this selection rule. They have no matrix elements within states drawn from a 4f configuration, and so do not give rise in the first order to any crystal field splitting, but they do admix states from excited configurations with the opposite parity into the 4f states. To the extent that the states now contain a small admixture of opposite parity states, electric dipole transitions between the “impure” 4f states are now permitted. Thus, electric dipole transitions are weakly allowed through configuration mixing. (We note that magnetic dipole transitions are allowed within a 4f configuration, but they are approximately six orders of magnitude weaker than allowed electric dipole transitions, and so the dominant effect is usually the electric dipole transitions induced by configuration mixing, as discussed here.) The second key effect is that the electron-vibrational interaction between the 4f electronic states and the dynamic glass environment is also weak due to the outer shell shielding. In the case of the 3d ions, the electron phonon coupling to odd parity

ABSORPTION COEFFICIENT (cm–1)

4.6 Glass Color

5

3P

ZBLA:Pr3+ (2 mol.%) T = 300K

3F 3

2

4 3P ,1I 1 6

3 3F 2

3P

3F 4

0

2 1

3H

5

3H

1D 2

6 1G

0

4

10 15 WAVENUMBER (103 cm–1)

5

20

25

Figure 4.11 Optical absorption spectrum of Pr3+ ions in ZBLAN glass Source: Reproduced from J.L. Adam and W.A. Sibley, J. Non-Cryst. Solids, 76, 267 (1985) by permission of Elsevier.

vibrations results in allowed “vibronic” transitions, as explained in the next section, giving rise to broad vibronic sidebands which are more intense than the purely “zero-phonon” electronic transitions. For the rare earths, this effect is much weaker, and consequently the optical absorption is much less than for the same concentration of a 3d transition metal ion. In Figure 4.11, we show the experimental optical absorption spectrum for Pr3+ ions in ZBLAN glass [21] which can be compared with the energy level scheme shown in Figure 4.10. The transitions from the 3 H4 ground state are labeled by the final state. It is evident that the ligand field splittings of the multiplets cannot be resolved. The combination of the transitions to the 1 D2 state in the yellow, and the 3 P states in the blue, results in the glass having a light green tinge. 4.6.3

Absorption by 3d Metal Ions

The 3d series of metal ions are characterized by progressive filling of the 3d shell as shown in Table 4.1. The unshielded outer shell electrons interact strongly with the ligands and the glass environment, unlike the case for the rare-earth ions. The effect of the crystal field is greater than that of spin orbit coupling for the 3d ions, and so we consider the effects of the crystal field first. In a crystal, the ligand field has a well-defined symmetry which markedly reduces the number of parameters in the Hamiltonian in Eq. (4.27) (in addition, the summation only contains terms up to and including Table 4.1 Electronic configurations for ions of the 3d series. d1

d2

d3

d4

V2+ 3+

Ti

V

3+

3+

Cr

Mn4+

d5

Cr2+ 3+

Mn

Mn2+ 3+

Fe

d6

d7

d8

d9

Fe2+

Co2+

Ni2+

Cu2+

99

4 Optical Properties of Glasses

l = 4 since the matrix elements are identically zero for l > 4 within the states of a 3d configuration), and the powerful formalism of point group theory may be brought to bear on the problem. For glasses, there is no such symmetry, but nonetheless we find that dopant ions are typically found in environments which are a distribution about an “average” environment. For the purpose of discussion, let us suppose that the average nearest-neighbor environment is octahedral with six identical ligands. The symmetry of a perfect octahedron then limits the ligand field to have the form ( ) √ 5 4 ) (4.28) (C 4 + C−4 H = B4 C04 + 14 4 so that only one parameter is required. Tanabe and Sugano [22] have calculated the eigenvalues of the Hamiltonian (Eq. (4.28)) for the electrostatic and ligand field terms for this case, and for the similar cases of tetrahedral and cubic coordination, assuming particular values for the intra-shell coulombic interaction specified by the Racah parameters B and C [20]. Their plots of energy versus cubic crystal field are a standard aid in understanding 3dn absorption spectra. A modern account is given by Figgis and Hitchman [20]. Ligand field theory abounds in different normalizations, and Tanabe and Sugano use an equivalent but proportional parameter Dq to parameterize the ligand field rather than B4 . In Figure 4.12, we show the Tanabe–Sugano diagram for d7 configuration in an octahedral field for C/B = 4.63 [23]. The predicted absorption spectra, ignoring the effects of non-octahedral ligand field perturbations, spin-orbit coupling, and vibrational interactions, may be read off the diagram with the known typical Dq and B values for different ligand coordinations. B(MgF2) = 975 cm–1 B(ZBLA) = 920 cm–1

Co2+ 60 ZBLA

MgF2

86K 2A

50

2g

40 2F

E/B

100

2A

20

1g

4A

30

2g 4T 1g

2T 1g 2T 2g

2G

4P

4T 2g

10 2E

g 4T 1g

4F

0

1.0

0.5 Dq/B

1.5

Figure 4.12 Energy levels for the d7 configuration. Source: Reprinted from Y. Suzuki et al. Phys Rev. B, 35, 4472. (1987) by permission of the American Physical Society.

4.6 Glass Color

The implicit assumption here is that the crystal field is dominated by the ligands, but this should be a better approximation for glasses than for crystals due to the lack of long-range order in glasses. Fuxi [24] gives a table of estimated values of Dq and B for different ions in different glasses. It should be noted that the observed spectra are not the sharp lines which might be expected by running a vertical line up Figure 4.12 at the expected value of Dq/B. At the very least, the distribution in crystal fields is about the average, and the spin–orbit coupling splits and shifts term sub-levels, resulting in substantial inhomogeneous broadening of the order of hundreds of cm−1 . An additional and substantial homogeneous broadening effect for the 3d ions arises from the effect of strong electron–phonon coupling. The broadening effect is best explained through the “configuration coordinate” model (e.g. see Figgis and Hitchman [20]), which accounts for transitions that involve both electronic transitions and phonon excitations. This means that substantial absorption occurs on the high-energy side of the purely electronic transition, and in fact the model shows that the most probable transition is typically not the “zero-phonon” transition but one which corresponds to the creation of several phonons. The overall result is that the absorption band is approximately a broad Gaussian envelope, comprising many individual electron–phonon excitations, whose center is displaced to the high-energy side of the purely electronic transition. (The reverse occurs for fluorescence; the emission band lies on the low-energy side of the zero-phonon line.) The strength of the optical absorption for 3d ions is greater than that of 4f ions since the configuration admixture effect is larger (because the 3d outer shell is unshielded); in addition, odd-parity vibrations enhance the transition probability also through configuration admixture effects. A selection rule which is important in determining the relative strengths of electronic transitions for the 3d series is the spin selection rule that ΔS = 0, which arises because the electric dipole operator in the transition probability expression does not involve spin. However, spin–orbit mixing of pure LS terms means that such spin-forbidden transitions can occur, albeit with a reduced intensity. In Figure 4.13, we WAVELENGTH (nm) 1000 600

ABSORPTION COEFFICIENT (cm–1)

2000 1.5

MgF2:Co2+ 2.0 × 1020cm–3

1.0

4T

2g

4A

0.5 0 8 6 4

0

2A 1g

4T 1g

86K

2g 4A 2g

2

86K

2g

ZBLA:Co2+ 3.3 × 1020cm–3 4T

400 4T 1g

2A

1g

× 1/4 5

10 15 20 WAVE NUMBER (103cm–1)

25

Figure 4.13 Optical absorption spectrum of Co2+ in ZBLA glass and MgF2 . Source: Reprinted from Y. Suzuki et al. Phys Rev. B, 35, 4472 (1987) by permission of the American Physical Society.

101

102

4 Optical Properties of Glasses

show the observed absorption spectrum for Co2+ ions in a ZBLAN glass and MgF2 . The major spin-allowed transitions can be directly interpreted with the aid of Figure 4.12. The strong absorption in the red–green region leaves most Co-doped glasses with a deep blue coloration. Figures 4.11 and 4.13 show the typical behavior that rare-earth ions show narrower sets of absorption lines than the 3d transition metal ions. Figure 4.13 also shows the typical behavior of glasses vs crystals: that ligand field fine structure is frequently resolved in crystals but rarely in glasses. In summary, transition metal ions give rise to strong broad absorption bands in glasses. The best guide to the interpretation and prediction of spectra is a knowledge of the average local coordination together with the energy level diagrams of Tanabe and Sugano [22].

4.7 Fluorescence in Rare-Earth-Doped Glass Fluorescence in rare-earth-doped glass is the basis for many applications such as neodymium-doped glass lasers, up-converters, erbium-doped fiber amplifiers for telecommunications, and fiber lasers. Fluorescence can be excited by optical pumping via the parity-allowed transition into the 4fn-1 5d configuration or charge-transfer excited states lying in the ultraviolet region indicated in Figure 4.10. From there, relaxation to the excited sub-states of the 4fn configuration either via radiative or nonradiative (phonon-assisted) mechanisms can occur, followed by radiative (fluorescent) decay to the ground state either directly, or via a cascade through intermediate states. Alternatively, the fluorescence can be excited by pumping into the higher-lying states of the 4fn configuration directly, but this is a weaker process as it is parity forbidden. As with the absorption, the fine structure due to crystal field splittings of the free ion multiplets is not resolved; the glassy environment again gives rise to a distribution of crystal fields, and this, together with the small magnitude of the crystal field, gives rise to an inhomogeneous broadening of the lines rather than a splitting, although unresolved structure can sometimes be observed. The unresolved structure can be probed further using techniques such as fluorescent line narrowing (FLN), which rely on the power and narrow linewidth of modern lasers to isolate transitions due to a particular glass site from within a broad envelope. Since the ligand field splittings are rarely resolved for rare-earth ions in glasses, it is not usually possible to parameterize the effects with a small set of ligand field parameters as is typically done for rare-earth ions in crystalline solids. However, there is great interest in the relative magnitudes of the transition probabilities for the various possible fluorescent transitions, since these are essential inputs to any consideration of practical device performance. Judd [25] and Ofelt [26] have shown that it is possible to predict the intensity of any allowed transition from just three parameters, Ω2 , Ω4 , and Ω6 , which summarize the effect of admixtures of excited state configurations into the ground state, induced by odd-parity components of the ligand field. The electric dipole oscillator strength for absorption or emission is given by [22]: 8𝜋 2 m𝜐𝜒ed ∑ fed (a, b) = Ω |⟨a‖U t ‖b⟩|2 (4.29) 3h(2J + 1)n2 t=2.4.6 t

4.7 Fluorescence in Rare-Earth-Doped Glass

where J is the total angular momentum of the initial level, n is the refractive index of the glass, 𝜈 is the frequency of the transition, and 𝜒 ed = n(n2 + 2)2 /9 is a local field correction for electric dipole transitions. The quantities U 2 , U 4 , U 6 are double-reduced matrix elements of the unit tensor operator U t that are tabulated, for example, by Carnall et al. [27]. From the oscillator strengths, the radiative lifetime can be calculated as the probability per unit time for the decay as: 8𝜋 2 𝜐2 e2 n2 fed (a, b). (4.30) mc3 The Judd–Ofelt parameters Ωt thus serve as a three-parameter base from which all the electric dipole transition probabilities [L, S, J] ⇒ [L′ , S′ , J ′ ] within a given configuration may be calculated, albeit for the entire multiplet rather than individual ligand field levels; but, since these are usually unresolved, this is not an important limitation. The parameters Ωt are specific to a given configuration and given glass host, since they parameterize the electron–ligand interaction. They can themselves be estimated from the absorption spectrum. In principle, measurements of the absorption cross-section for just three transitions suffice; but, in practice, least squares fitting to measurements from several lines is usually performed. The measurements yield the absorption cross-section 𝜎 abs , which is related to the absorption oscillator strength by: A=

f =

mc 𝜎 (𝜈)d𝜈 𝜋e2 ∫ abs

(4.31)

For transitions that have a significant magnetic dipole character, the oscillator strength for magnetic dipole allowed transitions is given by: fmd =

h𝜈𝜒md |⟨a‖L + 2S‖b⟩|2 6(2J + 1)n2 mc2

(4.32)

where 𝜒 md = n3 is the local field correction for magnetic dipole transitions. The magnetic dipole oscillator strength should first be subtracted from the observed oscillator strength f abs to yield f ed = f abs − f md for use in determining the Judd–Ofelt parameters from Eq. (4.29). Tables of Judd–Ofelt parameters for rare-earth ions in heavy-metal fluoride glasses have been given by Quimby [28], and for Nd3+ and Er3+ ions in a variety of glasses by Fuxi [24]. While this discussion has focused on the radiative process of de-excitation, one should be aware that there are parallel non-radiative processes involving energy migration between the rare-earth ions (in sufficiently concentrated systems) and ubiquitous phonon–assisted decay, which act to reduce the lifetime estimated from purely radiative considerations. In Figure 4.14, we show the fluorescence from Pr3+ ions in ZBLA glass. This spectrum was generated by pumping from the ground state into the 3 P2 state. It is noticeable that the low-temperature spectrum shows slightly narrower bands, but the residual inhomogeneous ligand field broadening still prevents resolution of any ligand field splittings. Fluorescence from 3d ions is infrequently observed in glasses, with the notable exception of the 3d3 (eg Cr3+ ) and 3d5 (eg Mn2+ ) ions.

103

4 Optical Properties of Glasses

ZBLA:Pr3+ (2 mol.%) λexc = 441 nm

6

T = 297K

Figure 4.14 Luminescence spectrum for Pr3+ ions in ZBLA glass. Source: Reproduced from J.L. Adam and W.A. Sibley, J. Non-Cryst. Solids, 76, 267 (1985) by permission of Elsevier.

4 LUMINESCENCE INTENSITY (arb.units)

104

2

3F

4

3F

3

3H

6

3H

5

0 3H 6

30 3F

T = 13K 2

20

3H 3F

3

3F

4

5

3H

4

10

0 11

13 15 17 19 21 PHOTON ENERGY (103 cm–1)

4.8 Glasses for Fiber Optics The glasses used to make fiber optics for long-distance communications are necessarily characterized by a low extinction coefficient in the transmission window, so low that, in the case of silica, it is too small to appear in Figure 4.1. As a practical engineering matter, the attenuation is measured in units of dB/km rather than as an extinction coefficient in cm−1 , with the relation between the two being ( ) ) ( P(L) 1 dB = log10 = 43429𝛼 (4.33) attenuation per unit length km L P(0) where P(x) is the power at distance x from the origin of a fiber optic, and the numerical value is for the attenuation coefficient 𝛼 in units of cm−1 . Practical values of the minimum attenuation for silica are in the region of 0.2 dB/km or 5 × 10−6 cm−1 . This implies that the intensity is reduced by a factor of 10 after about 50 km of cable. The attenuation in the window region for fiber optic glasses contains contributions from several effects, which are shown schematically in Figure 4.15. The so-called “V-curve” [29] is defined by electronic edge absorption and Rayleigh scattering on the short wavelength side of the minimum, and multi-phonon absorption on the long wavelength side; these effects can be described semi-empirically by [30]: 𝛼 = A exp(−a∕𝜆) +

B + C exp(c∕𝜆) 𝜆4

(4.34)

4.8 Glasses for Fiber Optics

Extinction Coefficient (cm–1)

1

0.1

0.01

Extrinsic Absorption Multiphonon (eg OH–)

0.001

0.0001

10–5 0.1

Electronic absorption

Rayleigh Scattering 1 Wavelength (µm)

10

Figure 4.15 Schematic diagram for the spectral dependence of extinction coefficient of silica.

The first term of Eq. (4.34) describes the absorption associated with the band edge. In crystalline solids, the absorption associated with electronic band gap transition should, in principle, commence abruptly when the photon energy first exceeds the band gap energy. However, there is a weak tail (the “Urbach Tail”) extending into the band gap zone that is thought to arise principally from phonon-assisted electronic transitions, and from localized defect-related electronic states. The involvement of phonons relaxes the electron wavevector selection rules that otherwise confine transitions to be vertical on the usual energy band diagram. In glasses, the structural disorder means that the concept of an abrupt onset of electronic states must be abandoned, and there is a continuous transition from high-mobility bulk states, which lie well within the effective band, to localized low-mobility states, which lie in the nominal band gap. It is these latter states, together with phonon-assisted transitions, which give rise to the Urbach tail in glasses. However, this contribution is usually dominated, at least in wide-gap materials, closer to the V-curve minimum by the Rayleigh scattering arising from chemical and structural perturbations on a molecular scale in the glass. We can expect from Eq. (4.24) (specifically in the form of Eq. (4.40)) that this effect will scale as inverse of the fourth power of the wavelength, as the square of the refractive index deviation from the average, and as the volume of the micro-regions that show refractive index deviation. On the long-wavelength side of the V-curve, the attenuation is determined by multiple phonon absorption. In a crystal, such transitions are permitted through anharmonicity of the lattice vibrational potential energy, giving rise to a long-wavelength tail in the absorption associated with the infrared absorption bands. For different materials, the position of the infrared bands shifts, depending on the atomic masses involved. In a simple estimation, the frequency of the IR bands is given by 𝜔20 = K∕𝜇

(4.35)

105

106

4 Optical Properties of Glasses

where K is the force constant and 𝜇 is the reduced mass for the vibration. Thus, for glasses made from higher-mass atoms and/or with weaker bonds, we expect that the IR bands will move to longer wavelengths (lower frequencies), and that the multi-phonon edge will move with them. Thus, the V curve for heavy-atom glasses moves to longer wavelengths along with the IR bands, and so we observe minima at longer wavelengths for the IR-transmitting heavy-metal fluoride and chalcogenide glasses. Since the Rayleigh scattering contribution is not directly dependent on the atomic mass, the effect is that the V-curve minima are expected to be deeper for these glasses than for silica. So far, this potential for lower fiber-optic attenuation than for silica has remained just that, due to extrinsic scattering by impurities and inclusions. For the heavy-metal fluoride glasses, residual transition metal ion concentrations, particularly of Fe2+ , Co2+ , and Cu2+ , give rise to electronic transitions near the V-curve minima, and there are also wavelength-independent scattering contributions from metallic inclusions such as Pt particles originating from the crucible used for glass melting, and from inadvertent crystallization in the fiber drawing process. All of these dominate the contributions in Eq. (4.34) near the V-curve minimum, and so silica retains its position as the fiber optic glass of choice. Even for silica, there is the well-known hydroxyl impurity vibrational overtone near 1.39 μm (with weaker overtones at 1.25 and 0.95 μm), shown schematically in Figure 4.15, which lies very close to the V-curve minimum at 1.55 μm, so that, in practice, it is the two minima on either side of this impurity absorption which are of practical use, resulting in the two telecommunications windows 1.2–1.3 μm and 1.55–1.6 μm. While the operation of a fiber optic cable at the wavelength of minimum attenuation is clearly desirable for long-distance communication, it is also important that the dispersion be minimal, since otherwise distortion of a pulse shape occurs in the time domain, limiting the maximum bit rate. Since the physical factors that determine the position of the V-curve minimum and the wavelength for zero dispersion are quite different, there is no reason to expect the wavelengths of minimum extinction and zero dispersion to be the same. If we use the Sellmeier formula with just two terms to express the refractive index due to one electronic resonance in the uv region and a vibrational resonance in the infrared as B e 𝜆2 B v 𝜆2 − (4.36) n2 − 1 = (𝜆2 − 𝜆2e ) (𝜆2 − 𝜆2v ) then we find, by computing the dispersion dn/d𝜆, that there is a zero in the dispersion at approximately ( )1∕4 Be 𝜆2e 𝜆2v (4.37) 𝜆= Bv In the case of silica fiber, this zero is at 1.27 μm as compared to the extinction minimum at 1.55 μm.

4.9 Refractive Index Engineering By systematically manipulating the refractive index within a volume of glass, it is possible to fabricate devices that show useful optical properties. The classic example is the

4.9 Refractive Index Engineering

optical fiber with a cladding that has a lower refractive index than the core, giving a structure which guides light rays within the core by total internal reflection. In this case, the refractive index variation is achieved, in concept at least, by simple fusion of a cylinder of core material and a pipe of cladding material, followed by stretching. The most common fiber-optic cables for telecommunications are of the so-called “single mode” construction where the core is of such small diameter than only a single mode can propagate, which eliminates modal dispersion and thus optimizes bit rates. The success of fiber optics has generated interest in extending the manipulation of refractive indices to other situations. GRadient INdex (GRIN) fibers, where the refractive index is engineered to vary radially with a gradual rather than a step transition, have improved transmission properties as compared to large-core multi-mode step fibers with regard to modal dispersion. The gradation is generally on the scale of microns, but it is also possible to produce index gradients over a larger size scale of millimeters, giving rise to a family of GRIN optical elements [31]. For example, the simplest GRIN device is the rod lens, essentially a scaled-up version of the graded index fiber, in which a radial index gradient results in the periodic focusing of light along the axis of the rod. GRIN lenses are typically a few millimeters in diameter, with focal lengths that can be of the same scale, and are used as relay lenses or in endoscopes. One advantage of GRIN over conventional lenses is that they have planar faces, which assists in fabrication and systems integration. Other applications include fiber optic couplers and imaging, where arrays of GRIN micro-lenses can be used to image large objects at short object-detector distances [9]. The index profile can be produced in a variety of ways. In the ion exchange method, a borosilicate glass containing a highly polarizable ion such as thallium or cesium is immersed in a bath of a molten salt containing an alkali ion of lower polarizability—for example, sodium or potassium [9]. Ion exchange results in a change in the refractive index of the glass, which can be understood through the Clausius–Mosotti relation (Eq. (4.20)) as a consequence of the replacement of an ion of one polarizability with another. The radial variation in refractive index directly reflects the ionic diffusion profile. Thallium salts are toxic, and so thallium-free processes (e.g. Hornschuh et al. [32]) based on exchanges between the ions Li, K, Na, and Ag have been developed. The ion exchange process can also be used to generate thin planar waveguide structures on glass surfaces that are more suited to opto-electronic integration than conventional fibers. Photolithography can be used to define waveguide geometry. GRIN structures have also been produced in polymers, with the advantage that larger diameters are possible; for example, Koike et al. [33] have made GRIN contact lens structures from polymer GRIN material. Spatial variations of the refractive index on a much smaller size scale are generated in the inscription of Bragg gratings [34] into optical fibers. The motivation here is to produce a periodic variation in refractive index, that is, a grating, along the fiber length that can act as an interference filter at wavelengths related in the usual way to the grating period. In Figure 4.16, we show the reflectivity of a Bragg grating inscribed in silica [35]. Such filters find widespread use in wavelength-dependent multiplexing in fiber optic communications, and in fiber optic sensors [36]. In the limiting case, the grating acts as a mirror, and so can be used as an in-situ reflector for fiber lasers. The grating is usually generated by an interference pattern from a high-intensity uv laser (typically at 244 nm from a continuous wave (CW) frequency-doubled 488 nm argon ion laser), and

107

4 Optical Properties of Glasses

20 REFLECTIVITY (%)

108

16

BRAGG GRATING RESPONSE ANDREW D-TYPE FIBER L = 0.9 mm E = 200 mJ/cm2/pulse T = 20 min at 50 pps

16% REFLECTIVITY

12 8 4

FABRICATION BY CONTACT PHOTOLITHOGRAPHY Using 1.06 μm Pitch Zero-Order Nulled Phase Mask and KrF Excimer Laser (249 nm)

0 1525

1527

1529 1531 WAVELENGTH (nm)

1533

1535

Figure 4.16 Reflectivity of a Bragg grating inscribed in a silica optical fiber. Source: Reproduced from K.O. Hill et al., Appl. Phys.Lett., 62 1035 (1993), by permission of the American Institute of Physics.

the interference pattern produces a permanent or temporary refractive index change in the glass at the positions of the interference antinodes. The interference patterns can be generated using a phase mask [35], or by an arrangement such as a Lloyd’s mirror configuration. In the common silica-based fiber, the photosensitivity is established by germanium doping and enhanced by hydrogen loading [34]. Refractive index changes of the order of 10−4 –10−2 can be produced. The germanium doping introduces an absorption band at 242 nm, and it is clear that the basic interaction with the uv light is through the resulting breaking and reconstruction of Ge–O bonds. But the exact mechanism of refractive index change is still a matter of discussion, with two interpretations. One is that a periodic densification effect is responsible, with the refractive index being periodically modulated either directly due to the changed particle density (Eq. (4.20)), or indirectly, for example, by the resulting periodic core-cladding stress modulating the refractive index through the photo-elastic effect [37]. The other model involves microscopic defects with changed atomic polarizability such as color centers [38]. Bragg gratings have been observed in other glasses (e.g. Ce-doped ZBLAN), but the mechanisms can be expected to be quite different from those proposed for silica. However, the best prospects for refractive index engineering come with the micro-machining technique [39]. In this method, a high-energy and high-repetition-rate femto-second laser (e.g. Ti: Saphire) or oscillator is focused via a microscope objective onto a micrometer-scaled volume inside a material that is normally transparent at the laser wavelength. Within the focal volume, the intensity is so high that electron-hole pairs are generated through non-linear interactions, and the resulting bond disruptions can lead to permanent refractive index changes of the order of 10−3 through mechanisms that are probably similar to those applied to Bragg gratings. (Although we note that color centers have been discounted by Streltsov and Borelli [40] in the case of silica and borosilicate glass.) By scanning the focused spot parallel to the surface, and also normal to the surface (to an extent limited by the focal length of the microscope objective), three-dimensional structures can be written in the glass. If the change in refractive index is positive, these structures act as light waveguides.

4.10 Glass and Glass–Fiber Lasers and Amplifiers

1.518

n

1.516

1.514

1.512 –20

–10

0 x (μm)

10

20

Figure 4.17 Variation of refractive index across the width of light waveguide produced by laser micromachining. Source: Courtesy of Professor Mazur.

If the energy dumped into the focal spot is increased, the concentration of electronhole-pair multiplies in an avalanche process. Their energy is transferred to the lattice through the electron–phonon interaction, and rapid, highly localized heating occurs. The material within the focal volume is essentially rendered into a plasma that first expands and then cools, leaving behind a permanent structural modification of approximately spherical geometry in which the refractive index varies radially [41]. The refractive index is smaller at the core than at the outer perimeter, which has been radially compressed, and so is just the opposite to that described earlier. In front of and beyond the focal volume, the electric field strength is inadequate to trigger the effect. By tuning the energy, focusing, and repetition rate of the laser or oscillator, the diameter of the microsphere and the refractive index can be adjusted. Again, by scanning the material perpendicular to the beam with an XY stage, or along the axis of the beam, or all of these, two- or three-dimensional light-guiding structures can be machined into the glass. In Figure 4.17, we show how the refractive index varies over the width of a simple linear waveguide that has a tubular structure [42]. The light guiding is achieved within the walls of the tubular structure. The technique is of commercial interest for micro-machining in a range of transparent solids; multiplexers, Bragg gratings, and filters have all been fabricated using this technique. The technique can also be used for write-once mass storage [43]. Schaffer has recently reviewed femtosecond micromachining [44].

4.10 Glass and Glass–Fiber Lasers and Amplifiers Lasers and optical amplifiers based on crystalline materials find widespread applications in modern society. Many, though not all, are based on the spectroscopic properties of rare-earth ions in crystalline hosts—for example, the high-powered Nd:YAG laser.

109

110

4 Optical Properties of Glasses

Since the energy level structure of rare-earth ions are relatively insensitive to the local atomic environment, it is not surprising that rare-earth-doped glass is also widely used as a lasing and optically amplifying medium, with the advantages of ease and versatility of fabrication, for example, into fibers. Long lengths of amplifying medium are also possible, which leads to high gains and efficiencies. Fiber lasers and optical amplifiers, in particular, have undergone extensive development in the past two decades, driven by the telecommunications and materials processing industries. Lasers and amplifiers based on glass rods or plates, rather than fibers, find more limited applications, such as the high-powered Nd:phosphate glass amplifiers operating at 1.053 μm used at the National Ignition Facility in the USA. A number of monographs [45–47] have appeared that describe the basis and operation of fiber lasers and amplifiers in particular, so only an overview will be given here. Optical amplifiers are required for long-distance fiber optic communication to restore signal levels after propagation over some tens or hundreds of kilometers. A typical system in silica-based fiber operates at the wavelength corresponding to the dip in attenuation near 1.55 μm (see Figure 4.15), and so any amplifying element should preferably show narrow-band amplification centered around this wavelength. Fortunately, trivalent erbium shows a fluorescent transition in the infrared between the excited 4 I13/2 multiplet and the ground 4 I15/2 multiplet that is very close to this wavelength, and the levels are very suited to three or four level amplification (depending on the temperature) by stimulated emission. At room temperature, three-level amplification is the basis for the so-called erbium-doped fiber amplifier (EDFA), where the amplifying medium is the erbium-doped core of a silica fiber. The amplifier is usually pumped by an InGaAs laser diode operating at 980 nm, which excites the Er3+ ions into the 4 I11/2 levels, from which the 4 I13/2 multiplet is populated by radiative or non-radiative decay. However, supplementary and direct pumping into the 4 I13/2 multiplet using 1480 nm light from a InGaSP diode laser is also frequently used. The emission spectrum for the resulting 4 I13/2 to 4 I15/2 fluorescence transition, which is the basis for optical amplification, is broadened to around 30 nm by inhomogeneous (crystal field) and homogeneous (phonon) effects, which permits wavelength-division multiplexing within the bandwidth, and hence optimal bit rates. EDFA amplifiers can have gains of over 50 dB and output powers of the order of hundreds of watts, and are essential components of any long-distance fiber optic communications network. Nonetheless, while silica is an excellent material for optical transmission, it is in many ways a sub-optimal host for rare-earth ions and optical amplification—the solubility of rare-earth ions is very low in this material and so only a small atomic fraction of dopant ions can be incorporated. Secondly, the phonon frequencies associated with the SiO2 network are very high, which means that few phonons are required to bridge any energy gap with correspondingly fast non-radiative relaxation processes that bypass radiative relaxation, reducing radiative efficiency. Consequently, other ions such as Pr3+ (which would otherwise be a suitable material for amplification in the 1.3 μm window [Figure 4.15]) are primarily used in other host glasses such as ZBLAN or tellurite glasses, where the low phonon frequencies corresponding to heavy host ions mean that radiative efficiencies are much higher. However, the difficulties of splicing different fibers together and the different pump wavelengths required mean that the silica-based EDFA amplifier remains dominant in telecommunications systems.

4.11 Valence Change Glasses

Turning to optical fiber lasers, these can also be constructed from rare-earth-doped glasses, but now with feedback typically provided by UV-inscribed Bragg gratings at either end of the fiber. A wide variety of wavelengths from the visible to the infrared are available by choosing different combinations of rare earths and transitions, and matching the Bragg grating period to the transition of interest. The long path lengths available with fibers results in high gains, and power levels of tens of watts in single-mode CW can be attained, and hundreds of watts in multimode operation. This leads to applications ranging from optical sources in communications and illumination, to welding and cutting, including surgical procedures. High-power infrared fiber lasers based on Yb3+ (∼1 μm) or Tm3+ (∼2 μm) find widespread use in industrial processing. The choice of host glass is governed by the same issues as for optical amplifiers: transparency in the visible or infrared, rare-earth solubility, non-radiative relaxation rates, and energy migration. Silicate and phosphate glasses have a higher solubility for rare-earth ions, while fluoride glasses such as ZBLAN offer, in addition, excellent transparency in the infrared. A second type of amplifier or fiber laser based on stimulated Raman scattering (SRS) rather than simulated emission has also been developed [48]. These devices operate on the Raman scattering sidebands of a pump laser and so depend upon the vibrational spectrum of the glass host. Although not as efficient, SRS fiber devices have the advantage that the operating wavelength is not limited by the energy level structure of the rare-earth ion in the simulated emission scheme. Rather, the emission wavelength is just that of the pump displaced by the Raman shift to the peak of the Raman sideband spectrum, or to a wavelength within the sideband determined by the Bragg grating reflectors. Thus, the emission or amplifying wavelength is determined primarily by the pump wavelength, and is not fixed by a rare-earth transition energy, a considerable advantage in some practical applications. In addition, SRS devices can be cascaded so that the output of one acts as the input to the next, so that multiple upshifts in wavelength can be achieved. Alternatively, multiple Bragg gratings within a single fiber construction can be used to produce the same effect. It is also possible to engineer the device so that multiple wavelength-shifted outputs are available with a comb-like output spectrum. Thus, SRS devices are particularly useful in wavelength division multiplexing technologies. The glass types that are used are typically silica with modified phonon spectra through high phosphorus or germanium doping [48]. Phosphorus doping is used with Bragg reflectors that select out the phonon peak at 13.2 THz, resulting, for example, in a pump laser wavelength at 1117 nm producing gain or an output at 1175 nm. In the case of Ge doping, a secondary peak in the phonon spectrum at 40 GHz is observed, so that the same pump wavelength would result in an output at 1313 nm. By fabricating the Bragg grating period appropriately, the SRS for this wavelength shift can be selected rather than on the main peak. This has the advantage that fewer cascaded units are necessary to achieve a given wavelength shift from the pump.

4.11 Valence Change Glasses Some rare-earth ions and transition metal ions can occur in different valence states. In the rare-earth series, europium and samarium can occur either as the trivalent or the divalent ion, depending on the host glass. For a given glass, it is possible to achieve

111

4 Optical Properties of Glasses

partial or complete conversion from one form to another by oxidation or reduction processes, which are in general irreversible, or by irradiation with either uv or X-rays, which can in some cases be reversed by irradiation with a different wavelength, or thermally. These processes give rise to a useful method of imaging for X-ray or uv radiography. For example, samarium-doped fluorophosphate or fluoroaluminate glass as prepared in argon [49] shows exclusively the Sm3+ valence state in the fluorescence spectrum (Figure 4.18). Sm3+ 1.0 X-ray Intensity, a.u.

(a)

2 sec 0.8 0.6 0.4

1.0 0.8 0.6 0.4 0.2 0.0

0.2

30 40 50 60 70 80 hv, keV

0.0

(c)

1.0 50 sec 0.8 0.6

Sm2+

0.4 0.2 0.0 1.0

2.0 2000 sec

0.8

1.5

0.6 1.0 0.4 0.5

0.2 0.0

550

600

650 700 750 Wavelength, nm

800

Absorbance, arb. un.

(b) Photoluminescence, arb. un.

112

0.0 850

Figure 4.18 The evolution of PL spectra of samarium-doped fluorophosphate glass under synchrotron X-ray irradiation. The irradiations times are (a) 2 s, (b) 50 s, and (c) 2000 s, respectively. Experimental data are shown by symbols. Emissions of Sm3+ and Sm2+ ions are shown by thin solid lines. The inset in (a) shows the spectrum of the X-ray irradiation. The broken line in (c) represents the X-ray-induced absorbance. Source: Reproduced from Okada et al., Applied Physics Letters 2011, 99, 121 105–8 (2011) by permission of the American Institute of Physics.

4.11 Valence Change Glasses

Irradiation with X-rays results in the partial conversion of Sm3+ to Sm2+ , as shown in the figure. By exciting the Sm2+ fluorescence selectively, and with appropriate filtration preceding the detection device, it is possible to record only the induced Sm2+ fluorescence. The intensity of the fluorescence directly reflects the incident X-ray dose. Thus, the material can be used for X-ray dosimetry, or for X-ray imaging. If an object is placed between an X-ray source and a flat plate made of Sm-doped fluorophosphate glass, the post-irradiation spatial variation of Sm2+ intensity displays the shadow or radiographic image of the object of interest. The stored image is stable for periods of weeks or more, but can be erased either with uv light or by heating. This radiation-induced valence change effect is potentially very useful for checking X-ray beam profiles in a new technique for the treatment of cancer, known as microbeam radiation therapy (MRT). In MRT, the X-ray beam passes through a venetian-blind structured collimator, so that the emerging X-rays are in the form of thin slices, some 50 μm thick, separated by typically 400 μm. It is critically important for the safety of the patient and the treatment efficacy that the resulting beam intensity cross-section closely follow the thickness profile of the collimator. In a beam quality control check, a sheet of the glass is placed behind the collimator, and, after exposure, the spatial variation of the Sm2+ fluorescence intensity is recorded using the a high-resolution scanning technique based on confocal microscopy, which can record the variation on a micron scale. Figure 4.19 shows the read-out from the glass after a 20 kGy exposure. The rare-earth-doped glass is thus a very viable 2-d dosimeter material for MRT beam quality assurance. Competing optical methods based on radiation-sensitive powders or polymers cannot achieve this because of light scattering in the detector, while single crystal methods are prohibitively expensive due to the cost of large area detectors. 1.6 X-rays

R(300 s), arb.un.

1.4 1.2

Air W

1.0

Glass

0.8 0.6 0.4 0.2 0.0 100

200

300

400 500 600 Distance, µm

700

800

Figure 4.19 The Sm2+ /Sm3+ conversion pattern induced by X-rays through a microslit collimator and read-out by photoluminescence confocal microscopy as the ratio R(300 s) = PL(Sm2+ )/PL(Sm3+ ). The line is a guide to eye. The inset shows the geometry of the experiment. The irradiation time is 5 min, which corresponds to ∼20 Gy. Source: Reproduced from Okada et al., Applied Physics Letters 99, 121 105 (2011) by permission of the American Institute of Physics.

113

114

4 Optical Properties of Glasses

Figure 4.20 X-ray image of a semiconductor IC recorded in Sm-doped fluorophosphate glass. The image shows the square semiconductor IC, the heat sink adhesive, the surrounding track pads, and the fine bonding wires linking the pads to the IC. The white marks are scattering from scratches in the glass surface.

The X-ray doses required for a readily measurable effect with the Sm-doped fluorophosphate or fluoroaluminate glass are quite high, which is appropriate for MBT, but inadequate for routine medical radiography. However, for industrial radiography, the dose is not a problem, and very-high-resolution images may be obtained with quite simple equipment such as a professional-standard digital camera. Figure 4.20 shows an image of a semiconductor IC recorded with an Sm-doped fluorophosphate glass using a charge-coupled device (CCD)–based digital camera (intended for astronomy) equipped with suitable filters. The 25-μm-thick bonding wires leading to the IC are clearly visible. The active process in the fluorophosphate glass is deduced to be one involving ionization of the phosphate bonding network, with subsequent trapping of the electron at the Sm3+ ion: PO → POHC + e− Sm3+ + e− → Sm2+ where PO represents a phosphorus–oxygen bond in the un-irradiated glass, POHC (phosphorous-oxygen hole centre) represents a hole in that bond, and e− is the photo-ejected electron. The POHC gives rise to color center absorption, which partially overlaps the Sm3+ and Sm2+ emission bands, as seen in Figure 4.18, but this overlap can be compensated for in any measurement. In the fluoroaluminate glass, the degree of overlap with the equivalent color center bands is much less of a problem. Valence conversion after uv irradiation has been observed for a number of other polyvalent ions in the 3d, 4d, and 5d series [50].

4.12 Transparent Glass Ceramics 4.12.1

Introduction

Glass ceramics comprise glassy matrices containing crystallites, which may range in size from nanometers to microns, and in volume fractions up to several tens of percent. They

4.12 Transparent Glass Ceramics

are produced by thermally controlled nucleation and growth of crystals, often with the aid of a nucleating agent, from a pre-quenched glass. The crystals can impart improved mechanical properties to the host glass through their inhibition of crack propagation. In addition, the balance between the thermal expansion coefficients of the crystals (sometimes negative) and the host glass permits adjustment of the composite expansion coefficient, and it is possible to design materials such as the Schott product Zerodur , which has zero expansion coefficient at room temperature. The outstanding mechanical and thermal properties of these glass ceramics find applications in areas as widespread as artificial joints, machineable ceramics, magnetic disk substrates, and telescope mirror blanks. However, the major interest in the present context is in “nanophase” or “nanocrystalline” glass ceramics, where crystals of size ≤ 50 nm are uniformly dispersed in a transparent host glass, resulting (for favorable conditions, discussed in the following text) in a transparent glass ceramic. Beall and Pinckney [51] give a review of these materials. Generally, the crystals are themselves transparent, but it is worth pointing out that the first optical application for nanophase glass ceramics was based on metallic crystallites. Stookey et al. [18] pioneered the work on photochromic glasses where the reversible darkening and/or coloration in silver- and copper-doped borosilicate glasses is based on photochemical reactions involving silver and copper ions, and colloidal silver particles [3]. In this case, the optical effects are due to Mie scattering from the metallic particles described earlier (Section 4.6.1). The first major application of glass ceramics containing transparent crystals evolved from the earlier development of low-thermal-expansion coefficient glass ceramics containing “stuffed” 𝛽-quartz when it was found that these could be made in transparent form if the particle size was sufficiently small. The term “stuffed” here refers to 𝛽-quartz, where some of the Si4+ ions have been replaced by Al3+ ions and monovalent cations such as Li+ ; in the limit, this is the crystal 𝛽-eucryptite LiAlSiO4 . This immediately gives rise to the possibility of a transparent material able to survive high temperature without cracking, and this type of transparent glass ceramic now finds applications in fire-doors, stove tops, and cookware. However, in the past 20 years, there has been growing interest in transparent glass ceramics for photonic applications. These materials combine the ease of formation and fabrication of a glass with the desirable properties of a crystalline environment for photo-active dopant ions, such as those from the rare-earth series. Critical to most applications is the transparency of the glass, since in general each crystallite will act as a scattering center, and, in applications such a lasers and optical amplifiers, it is absolutely critical that any mechanism (scattering, absorption, etc.) that removes light intensity from the beam be minimal. In 1993, Wang and Ohwaki [52] triggered the current interest when they reported that a glass ceramic comprising Er3+ - and Yb3+ -doped fluoride crystals in an oxyfluoride glass was as transparent as the precursor glass, and showed 100 times the up-conversion efficiency when pumped at 0.97 μm, a result comparable with the best achieved in single crystals. The potential advantages of this new class of photonic material were appreciated by Tick et al. [53], who adapted the material for Pr3+ amplifier operation at 1.31 μm by replacing Yb with Y and Zn fluorides, and the Er3+ with Pr3+ . They showed that the glass ceramic had a greater quantum efficiency than the competing Pr3+ -doped ZBLAN fluoride glass, and remarkably low scattering, raising the potential of this material as a fiber amplifier. In a later paper, Tick [54]

®

115

116

4 Optical Properties of Glasses

went further to suggest that, under appropriate conditions of particle size and volume fraction, glass ceramics could be ultra-transparent—that is, have a transparency similar to that of pure glass. These three papers acted as the catalyst for the substantial interest over the past decade in transparent glass ceramics for photonics applications. In the absence of an established theory, qualitative ideas guided the development of transparent glass ceramics. The ideas of effective media suggest that, if the size scale of the refractive index perturbations are much smaller than the wavelength of light in the medium, then the material should be described by a single effective dielectric constant (given, e.g., by the Maxwell–Garnett theory), and so there should be no scattering. Tick [54] has suggested the following four empirical criteria for a transparent glass ceramic: 1) 2) 3) 4)

The particle size must be less than 15 nm. The inter-particle spacing must be comparable with the crystal size. Particle size distribution must be narrow. There cannot be any clustering of the crystals.

One might also add that the refractive index difference must be small. If there is no refractive index mismatch, then the material is of course fully transparent. It is worth pointing out that this can only occur for cubic (i.e. non-birefringent) crystals, but can explain why some glass ceramics involving a single crystalline phase can be transparent for large crystallite sizes. Experimentally, particle sizes are often measured from the X-ray diffraction pattern line widths using the Scherrer formula [55]. This is necessarily a mean size estimate, gives no information on the size distribution, and is limited by the instrumental resolution to particle sizes less than about 100 nm. Transmission electron microscopy (TEM) measurements are better in this regard, and can give particle size distributions, mean spacings, and volume fractions, but are much more time-consuming. Estimates of volume fraction from the starting compositions are likely to be overestimates because of loss of volatile components during the melting process. For example, the oxyfluoride glass ceramics are susceptible to loss of fluoride through SiF4 volatilization. As an example of the insight which can be gained from TEM work, Dejneka [56] has found agreement with Rayleigh scattering for 7% volume fraction LaF3 nano-crystals of average size 15 nm, with a refractive index mismatch of 0.07 to an oxyfluoride host glass. However, a sample that contained 300-nm particles when quenched, and which was quite opaque, became more transparent on annealing. TEM studies showed that the effect of annealing was simply to fill in the spaces between the large crystals with small 20–30-nm crystals. This cannot be explained by Rayleigh scattering, and suggests that a cooperative effect is involved in the scattering process. 4.12.2

Theoretical Basis for Transparency

An understanding of light scattering processes in glass ceramics is clearly important to optimize the applications. We begin by examining the regime in which the particles act as independent scatterers. Mie first developed a theory of independent scattering from spherical particles, which reduces to the Rayleigh theory for particles much smaller than the wavelength of light. The Mie theory is exact in the sense that it is based on solutions of Maxwell’s equations resulting from the matching of electromagnetic waves at the surface of the sphere, but those solutions are in the form of approximations to the resulting

4.12 Transparent Glass Ceramics

infinite series. There is also the Rayleigh–Gans (or Rayleigh–Debye) theory, which can be applied to particles somewhat larger than those for which Rayleigh theory is appropriate, and for non-spherical geometries. It works by dividing a volume into Rayleigh scattering micro-volumes, and then adding the scattering from each micro-volume taking into account position-dependent phase shifts. The scattering cross-section for extinction is defined as the area which, if it were completely “black,” that is, completely absorbing, would remove an equivalent intensity from the beam per scattering center. For a spherical particle of refractive index n1 and radius a, immersed in a medium of refractive index nm , the scattering cross-section Csca for light of vacuum wavelength 𝜆 is ( )2 24𝜋 3 V 2 n2 − 1 Csca = (4.38) (𝜆∕nm )4 n2 + 2 where V is the particle volume and n is the relative refractive index n1 /nm . This is actually Eq. (4.24) expressed for a real dielectric constant and in terms of refractive indices. A light beam of initial intensity I 0 that passes through a thickness z of scattering particles with concentration N m−3 will emerge with intensity I where I = I0 exp(−NCsca z) = I0 exp(−az)

(4.39)

If the difference in refractive index |n1 -nm | is much less than the average n = (n1 + nm )∕2, then Eq. (4.38) can be approximated in the expression for the extinction coefficient 𝛼 = NC sca by: ) ( 𝜋 4 a3 128 (NV ) 4 (Δnn)2 NCsca = (4.40) 9 𝜆 which illustrates some of the key dependencies. The extinction coefficient depends linearly on the volume fraction, on the cube of the particle size, and on the square of the difference in refractive indices. Thus, the designer of transparent glass ceramics should seek to minimize all of these quantities, particularly the particle size and refractive index mismatch, to achieve transparency in the Rayleigh regime. Tick et al. [53] pointed out that, for the estimated volume fraction of particles (∼20–30%) present in the oxyfluoride glass ceramics, the Rayleigh or Mie theories for dilute and weak scatterers give results for the extinction coefficient that are orders of magnitude larger than those observed. Since a theory for nucleated glass ceramics was not available, Tick et al. [53] turned to the work of Hopper [57, 58] for light scattering in glasses that showed spinodal decomposition. Hopper showed that the extinction coefficient should be 𝛼 = (6.3x10−4 )k04 𝜃 3 (Δnn)2

(4.41)

where 𝜃 is the mean phase width and k 0 is the wavevector in free space. We note first that the Rayleigh-like wavelength dependence of the extinction coefficient as the inverse fourth power is preserved in this theory. Making the large conceptual step of applying the theory to nucleated crystals, we can equate 𝜃 to (a + W /2), where a is the particle radius and W the interparticle spacing. If we take a value of 50% for the volume fraction, Hopper’s theory predicts an extinction coefficient about two orders of magnitude lower than that obtained from the Rayleigh theory [53] . However, the validity of this theory applied to a glass showing nucleated rather than spinodal decomposition remains questionable.

117

118

4 Optical Properties of Glasses

Hendy [59] has presented a more recent analysis of light scattering for the late-stage phase-separated materials and has deduced an extinction coefficient given by ) ( 14 Δn 2 8 7 𝛼= k a (4.42) ⟨𝜙⟩⟨1 − 𝜙⟩ 15 n where φ is the volume fraction of the phase, k = 2𝜋/𝜆, and the other terms have their usual meaning. This formula has a very different dependence on wavelength and particle size to that of the Rayleigh and Hopper models; in particular, note the very high orders for the wavelength and particle size dependence. As far as the author is aware, there have been no reports of a 1/𝜆8 wavelength dependence for the extinction coefficient observed experimentally. This would be expected to give rise to visible coloration effects, more pronounced than those for Rayleigh scattering. In fact, there have been very few attempts to investigate the wavelength dependence of the scattering and confront it with the theories, but this is difficult in the case of rare-earth or other doping in glass ceramics of practical interest, since electronic absorption can dominate the extinction coefficient in the uv region where the 𝜆−8 effect would be most pronounced. We have examined [60] the second and fourth of Tick’s criteria, namely that the interparticle spacing must be comparable with the particle size, and that there should be no clustering, by making use of the discrete dipole approximation (DDA), [61], which has been used for a variety of scattering simulations. The DDA models the response of a dielectric material by a set of dipoles located on the lattice points of a simple cubic lattice. The response of the dipoles to an incident electromagnetic field is calculated self-consistently, including dipole–dipole interactions. The polarizability of the individual dipoles can be chosen to model any dielectric structure. For a glass–ceramic simulation, we populated a 64 × 64 × 64 array with 2 × 2 × 2 nanocubes whose dipole polarizabilities were chosen to represent BaCl2 nanoparticles in a fluorozirconate glass matrix, and the entire supercube was taken to be embedded in a host glass whose dielectric constant was estimated by the Maxwell–Garnett formula for an effective medium [16]. The total nanocube volume fraction was varied, but the nanocube distribution was constrained to be either: (i) at random locations without overlap, but with at least one lattice spacing between particles; or (ii) also at random, but permitting clustering without overlap. The results of the simulation show that the scattering varies almost linearly with particle concentration up to 25% volume fraction. If transparency were to be associated with small inter-particle separation (equivalent to a high-volume fraction of particles), we would expect to see some sub-linearity as particle concentration increased. The effect of clustering was to increase the scattering in a supra-linear fashion, as would be expected from the high-power law dependence on particle size when some particles cluster together to form an effectively larger composite particle. As a general observation, these calculations do not support the notion that a high-volume fraction of scattering particles is a key factor in determining transparency in glass ceramics, but support the idea that agglomeration is detrimental. A new explanation of high transparency in glass ceramics was subsequently put forward by the present author [62] based on a core-shell model for the nanoparticles. The basic assumptions are that, in the nucleation and growth process, some ions move into the volume of the particle core from the surrounding shell, while others from the core move into the surrounding shell, such that the number and type of particles is conserved. It is also assumed for simplicity that the final core and shell are (separately)

4.12 Transparent Glass Ceramics

of homogeneous composition, and that the scattering is dominated by single-particle scattering events. The analysis is based on the Rayleigh–Debye approximation [15], and shows that the scattered intensity for unpolarized light can be expanded in a power series of k4 , where k = 2𝜋/𝜆 is the wavevector of the incident light. The first term in the expression for the scattered intensity I(𝜃) at distance r and angle 𝜃 to the incident direction is given by [ ] |2 | ∑ | (1 + cos2 𝜃)k 4 || S 2 I = I0 4𝜋s ΔNi (s)𝛼i ds|| (4.43) | 2 2r |∫0 | i | | where s is the radial distance from the center of a particle of shell radius S, ΔN i is the change in concentration of ion species i from the homogeneous glass, and 𝛼 i is the polarizability of that ion species. Under the assumption of conservation of ions within the core-shell particle, this term is identically zero, giving rise to the prediction of ultra-transparency for any glass ceramic that satisfies the conditions of the model. The basic reason for the predicted ultra-transparency is that the scattering is unaffected (to lowest order) by the specific distribution of ions within the core shell structure, so that a core shell has the same scattering as the uncrystallized glass—that is, no scattering. The second term in the series is ( ) [ ] |2 2(1 + cos2 𝜃)sin4 𝜃2 k 8 || S ∑ | 4 | I = I0 4𝜋s ΔNi (s)𝛼i ds|| (4.44) | 2 9r |∫0 | i | | and varies as k 8 in the same way as predicted by the Hendy model. The angular dependence shows strong backscatter, as shown in Figure 4.21. As far as the author is aware, this predicted wavelength dependence in a highly transparent glass ceramic has yet to be tested. The experimental difficulties in testing such a relation relate to the difficulty [60] of separating any weak scattering contribution to the Figure 4.21 Polar plot showing the angular dependence (dashed line) of the scattering predicted by the second term in the expansion series for the core-shell scattering analysis.

90

2

120

60

1.5 30

Intensity

1

Intensity

0.5 0

0

210

330

240

300 270

119

120

4 Optical Properties of Glasses

extinction coefficient of the type predicted by Eq. (4.44), which would only be prominent at short wavelengths, from the large residual electronic absorption of rare-earth or transition metal impurities or dopants. At long wavelengths, the wavelength-independent scattering from a small fraction of large particles dominates. We note that there is some evidence, for other materials, for an inverse eighth-power dependence on wavelength, and for strong back scatter in early work on “anomalous scattering,” which is summarized by Goldstein [63]. The core-shell model analytical predictions have been checked in two ways: by a parallel analytical calculation based on Mie scattering from a coated sphere, and by a DDA calculation of the scattering from a single core-shell structure embedded in a binary composition glass [64]. These alternative analyses confirm the basic premise—that, if the crystallization process comprises nucleation and growth from the local environment resulting in a core-shell nano-crystallite, the resulting scattering is much less than would be expected from Rayleigh scattering from the core embedded in a uniform glass. 4.12.3

Rare-Earth-Doped Transparent Glass Ceramics for Active Photonics

Rare-earth-doped transparent glass ceramics have been an area of substantial activity, primarily driven by telecommunication applications. Dejneka [65] and Goncalves et al. [66] have presented reviews of photonic applications of rare-earth-doped transparent glass ceramics. These materials offer the following advantages for applications such as fiber amplifiers and up-conversion: 1. They can be almost as transparent as simple glasses. 2. The active rare-earth ions may selectively partition into the crystals. 3. The rare-earth ions are in a crystalline environment where their radiative decay characteristics are superior to those typically found in a glassy environment. This is especially the case for heavy-metal halides such as LaF3 and Pbx Cd1-x F2 . 4. The glass ceramics include silicate-based varieties that are compatible with bonding to silica fiber. A great deal of work has been done on the oxyfluorides containing Pbx Cd1-x F2 and LaF3 crystallites. In the latter case, it is perhaps not surprising that the nano-crystalline phase readily accommodates a variety of trivalent rare-earth ions, since there is excellent charge and ionic radius match. However, Dejneka [56] has made an interesting observation: that oxyfluorides doped with GdF3 , rather than LaF3 , show crystallization in a hexagonal tysonite phase that has not been previously reported; similarly, YF3 and TbF3 are also found in the hexagonal phase, although this is not the room temperature stable phase for these compounds. We have also found hitherto unreported phases, and phases unstable at room temperature and pressure in ZBLAN glasses containing chloride and bromide crystals. It seems likely that the increased role of surface as against volume energies for nanoscale crystallites stabilizes phases that would normally be unstable at STP, and that one can, in general, “expect the unexpected” in nanoscale glass ceramics. For the case of Pbx Cd1-x F2 , the nano-crystalline phase was identified from the X-ray diffraction (XRD) pattern as a solid solution of CdF2 and PbF2 . It is surprising that trivalent rare-earth ions selectively partition into these nano-crystals, given the mismatch of charge and ionic radius. Tihkomirov et al. [67] proposed that, in erbiumdoped oxy-fluorides, the Er3+ ions are involved in nucleating centers that comprise

4.12 Transparent Glass Ceramics

mixed-phase orthorhombic 𝛼-PbF2 : ErF3 nucleation centers for the subsequent growth of 𝛽-PbF2 nano-crystals. This mechanism can explain the selective uptake of rare-earth ions, and has considerable implications for the quantum efficiencies, since one can expect at least a region of high rare-earth concentration near the core of the nano-crystal that may be subject to concentration quenching. One would therefore expect a marked dependence of the optical performance of the material on the nucleation and growth conditions. Just this effect was observed by Mortier et al. [68] in Er3+ -doped germanate glasses, with clear evidence in the 4 I11/2 –4 I13/2 fluorescent transition for concentration-quenching effects. In contrast to these results, for sol-gel-derived oxyfluoride glasses, the rare-earth ion seems to limit nucleation and growth of the crystals [66], and so the role of the rare-earth ion is clearly process dependent. 4.12.4

Ferroelectric Transparent Glass Ceramics

Ferroelectric crystals from the ABO3 family such as LiNbO3 are widely used in non-linear electro-optics. But the expense of single crystal production has prompted a search for glass ceramic materials containing ferroelectric crystals that may be more cost-effective, and which also offer the usual glass advantages of fabrication versatility. Such materials can be used directly as an optically isotropic medium in Kerr birefringence modulators, or can be electrically poled at an elevated temperature, resulting in a linear electro-optic effect as required for a Pockels modulator and for second harmonic generation. A number of transparent glass ceramics containing ferroelectric crystals have been reported [69–71], and in a niobium–lithium–silicate glass containing sodium niobate crystals, a Kerr effect comparable with nitrobenzene, a standard Kerr cell liquid, has been observed [72]. Nonlinear properties of glasses are the subject of Chapter 10. 4.12.5

Transparent Glass Ceramics for X-ray Storage Phosphors

X-ray storage phosphor (XRSP) imaging plates [73, 74] are solid-state replacements for the photographic film used in medical, dental, and industrial radiography. The currently favored materials used in imaging plates are powdered crystalline BaFBr or CsBr doped with ∼1000 ppm divalent europium. The image is stored as a spatial variation of radiation-induced trapped electrons and holes, and is read out by stimulating their recombination with a raster-scanned red laser beam. The recombination energy is transferred to the europium ions and appears as blue photo-stimulated luminescence (PSL). Although the imaging plate technology offers many advantages over the traditional photographic film process such as re-usability, wider dynamic range, freedom from chemical developers, and direct digital image read-out, the spatial resolution is not as good as that of fine-grained photographic film. The reason is that the focused laser beam used in the read-out process in BaFBr is scattered by the powder grains, so that the PSL occurs not only from the region of the focusing spot, but also from the surrounding material, limiting the resolution to about 100 μm. While this is adequate for most applications, such as chest X-rays, it is inadequate for applications such as mammography and crack detection in materials testing. It may be thought that the problem could be overcome by embedding the material in a refractive index matching binder, but BaFBr is birefringent, so matching is not possible. For cesium bromide, a preparation process that results in

121

122

4 Optical Properties of Glasses

an oriented needle-like structure in the image plate gives a somewhat improved spatial resolution through the light-guiding properties of the needles, but the improvement factor is no better than about two [75]. Consequently, the idea of a transparent XRSP such as a glass is very attractive for minimizing optical scattering and improving the spatial resolution in imaging plate radiography, but unfortunately it seems that the density of stable trapping centers in simple glasses examined so far is insufficient for a viable XRSP. However, it has been shown [76–79] that fluorozirconate glass ceramics containing europium-doped BaBr2 , Rb2 BaBr4 , RbBa2 Br5 , or BaCl2 nano-crystals show a significant XRSP effect, in the latter case up to 80% of that for BaFBr:Eu. The magnitude of the effect depends on the crystallite size, and it was suggested by Secu et al. [76] that there was a surface layer on the BaCl2 crystallites of about 7 nm in thickness which was not PSL active. Large crystallites of size ≅30 nm are therefore optimal for a large PSL effect. However, an increasing crystallite size also results in more light scattering [80], as shown in Figure 4.22, and so there is a trade-off between PSL efficiency and transparency, as shown in Figure 4.23. For samples that are still just transparent to the eye, the relative efficiency falls to about 10%. The spatial resolution of the glass ceramic, as specified by the modulation transfer function, is significantly better than for BaFBr [81, 82]. High-resolution images taken with a Eu-doped fluorochlorozirconate glass ceramic have been presented by Schweizer et al. [82], who also demonstrate that the same material can be used as a high-spatial-resolution phosphor/scintillator. The effect of light scattering for scanned laser beam image read-out from a glass ceramic X-ray storage phosphor has been simulated numerically [83]. The model used for the read-out process includes scattering of both read-out light and PSL, and absorption. Scattering was characterized by a macroscopic scattering length μ, such that the probability of a photon being scattered after traversing a length 𝜉 is given by ( ) −𝜉 (4.45) P(𝜉) = 1 − exp 𝜇 and the anisotropy in the scattering was approximated by a cone-like Henyey– Greenstein function [84]. The major conclusions were that the modulation transfer function is bimodal, as observed experimentally, with contributions from the effects of both the scattering and the finite laser beam diameter. The spatial resolution is optimized when the scattering length is much greater than the thickness, corresponding to a highly transparent glass ceramic, and is limited only by the beam diameter, but this case results in low read-out intensity, since most photons pass straight through the sample without interaction. Hence, a fully transparent X-ray storage phosphor material is not necessarily a practical solution. The other extreme, where the scattering

270

275

277

280

283

Figure 4.22 Visual appearance of fluorochlorozirconate glass ceramic storage phosphor annealed at various temperatures (∘ C). Source: Reproduced from A. Edgar, G.V.M. Williams, M. Secu, S. Schweizer, J-M Spaeth. Radiat Meas. 2004;38(4–6):413–6 by permission of Elsevier.

4.12 Transparent Glass Ceramics

35 30

Transmittance –0.5

25 20

–1 15

Relative PSL (%)

Log (Transmittance at 633 nm)

0

10 –1.5 5 PSL –2

0 220

230 240 250 260 270 Annealing Temperature (°C)

280

Figure 4.23 PSL efficiencies relative to BaFBr: Eu(100 ppm) for annealed glass. The measured transmittance at 633 nm is also plotted for a 1.45-mm-thick sample. The lines are guides to the eye. Source: Reproduced from A. Edgar, G.V.M. Williams, M. Secu, S. Schweizer, J-M Spaeth. Radiat Meas.; 38(4–6):413–6 (2004) by permission of Elsevier.

length is much less than the sample thickness, also results in high spatial resolution, since the interactions are concentrated in a small volume near the surface, but again a low read-out intensity results, because only a thin layer of material near the surface is being activated. The optimal practical solution is a compromise between spatial resolution and read-out efficiency, corresponding to the case when the scattering length is approximately the same as the thickness. This results in a bimodal modulation transfer function as observed [83], which can be reproduced by an appropriate choice of parameters in the simulation. It was also shown that selective choice of additives (such as Co2+ ) to the glass matrix to absorb red light, but not blue, would result in an improved modulation transfer function by limiting the spatial spread of the red stimulating light. In an effort to understand the processes that lead to the nucleation and growth of barium chloride nano-crystals in this fluorozirconate glass host, Hendy et al. [85] and Edgar and Hendy [86] undertook molecular dynamics simulations of the quenching process for glass melts. In a ZBLN glass with up to 30% of the fluoride ions replaced by chloride ions, it was found that, for chlorine concentrations above about 20%, there was clustering into zirconium-rich/chlorine-poor and sodium- and barium-rich/chlorine-rich regions. Figure 4.24 shows the fluorine and chlorine ions for a glass containing 20% chlorine anions, showing the predicted phase separation for the anions. In a subsequent study [86] based on both chlorine-doped ZBLN and ZBLLi fluoride glasses, a similar propensity to phase separation was again found, but was most pronounced in the ZBLN glass. Although direct precipitation of a solid crystalline phase was not predicted and is, in any case, not expected with this sort of calculation, the existence of a phase separation is a clear portent of crystallization by a chloride phase. Given that ternary barium chlorides involving sodium or lithium are not known, this would most probably be barium chloride, as observed. The simulations do give a guide

123

124

4 Optical Properties of Glasses

Figure 4.24 Anion distribution for a ZrBaLaNa glass showing fluorine distribution (left) and chlorine distribution (right). A chlorine-rich, fluorine-poor region can be seen at the lower left. Source: Reproduced from S.C. Hendy and A. Edgar, J Non-Cryst Solids;352(5):415–22 (2006) by permission of Elsevier.

to the optimal chlorine concentration, since the phase separation reached a maximum in the latter calculation at around 12–14% of the total anion content.

4.13 Conclusions The optical properties of glasses is a subject that has both mature and developing aspects; the historical use of glass in optical instruments means that a great deal of information has been gathered and analyzed regarding the bulk optical properties of traditional glass in the visible region. Yet, there are new and fascinating applications that have emerged in the past 30 years, dominated by fiber optics and the telecommunications industries. The revolution in human communications brought about by the Internet and the resources that it delivers are, of course, based on the high-speed, high-capacity optical links provided by optical pulse propagation on glassy silica fibers. But there have been other new developments: novel glass families optimized for optical performance in the infra-red and ultraviolet spectral regions, glass fiber lasers and amplifiers, a whole new class of glassy materials in the form of transparent glass ceramics, new physical phenomena such as PSL, and new technologies for refractive index engineering. The next few years promise to deliver equally intriguing developments in optical glass science.

References 1 Palik, E.D. (ed.) (1985). Handbook of Optical Constants of Solids. Orlando: Academic

Press. 2 Garbuny, M. (1965). Optical Physics. New York: Academic Press.

References

3 Bach, H. and Neuroth, N. (eds.) (1995). The Properties of Optical Glass. Berlin:

Springer-Verlag. 4 Robinson, W.H. and Edgar, A. (1974). The piezo-electric method of determining

5 6 7 8

9 10

11

12 13 14 15 16 17 18 19 20 21 22 23

24 25

mechanical damping at frequencies of 30 to 200 kHz. IEEE Trans. Sonics Ultrason. 21: 98–105. Borelli, N.F. (1964). Faraday rotation in glasses. J. Chem. Phys. 41: 3289–3293. Carlin, R.L. (1986). Magnetochemistry. Berlin: Springer-Verlag. Van Vleck, J.H. and Hebb, M.H. (1934). On the paramagnetic rotation of tysonite. Phys. Rev. 46: 17–21. Edgar, A., Giltrap, D., and MacFarlane, D.R. (1998). Temperature dependence of Faraday rotation and magnetic susceptibility for Ce3+ and Pr3+ ions in fluorozirconate glass. J. Non-Cryst. Solids 231: 257–267. Yamane, M. and Asahara, Y. (2000). Glasses for Photonics. Cambridge: Cambridge University Press. Suzuki, F., Sato, F., Oshita, H. et al. (2018). Large Faraday effect of borate glasses with high Tb 3+ content prepared by containerless processing. Opt. Mater. 76: 174–177. Sun, L., Jiang, S., and Marciante, J. (2010). All-fiber optical magnetic-field sensor based on Faraday rotation in highly terbium-doped fiber. Opt. Express 18 (6): 5407–5412. Silva, R.M., Martins, H., Nascimento, I. et al. (2012). Optical current sensors for high power systems: a review. Appl. Sci. 2 (3): 602. Griffiths, D.J. (1999). Introduction to Electrodynamics. Upper Saddle River: Prentice Hall. Van der Hulst, H.C. (1957). Light Scattering by Small Particles. New York: Wiley. Kerker, M. (1969). The Scattering of Light. New York: Academic Press. Bohren, C.F. and Huffman, D.R. (1983). Absorption and Scattering of Light by Small Particles. New York: Wiley. Edgar, A. (1997). Strong red-light scattering from colloidal copper in ZBLAN fluoride glass. J. Non-Cryst. Solids 220: 78–84. Stookey, D.S., Beall, G.B., and Pierson, J.E. (1978). Full-colour photosensitive glass. J. Appl. Phys. 49: 5114–5123. Dieke, G.H. (1968). Spectra and Energy Levels of Rare Earth Ions in Crystals. New York: Interscience Publishers. Figgis, B.N. and Hitchman, M.A. (2000). Ligand Field Theory and its Applications. New York: Wiley-VCH. Adam, J.L. and Sibley, W.A. (1985). Optical transitions of Pr3+ ions in fluorozirconate glass. J. Non-Cryst. Solids 76: 267–279. Tanabe, Y. and Sugano, S. (1954). On the absorption spectra of complex ions II. J. Phys. Soc. Jpn. 9: 766–779. Suzuki, Y., Sibley, W.A., El Bayoumi, O.H. et al. (1987). Optical properties of transition-metal ions in zirconium-based metal fluoride glasses and MgF2 crystals. Phys. Rev. B. 35: 4472–4482. Fuxi, G. (1992). Optical and Spectroscopic Properties of Glass. Berlin: Springer-Verlag. Judd, B.R. (1962). Optical absorption intensities of rare-earth ions. Phys. Rev. 127: 750.

125

126

4 Optical Properties of Glasses

26 Ofelt, G.S. (1962). Intensities of crystal spectra of rare earth ions. J. Chem. Phys. 37:

511. 27 Carnall WT, Crosswhite H, Crosswhite HM (1977). Energy level structure and

28 29 30 31 32 33 34 35

36 37

38 39 40 41 42 43 44

45 46 47

transition probabilities of the trivalent lanthanides in LaF3. Argonne, IL: Argonne National Laboratory Report. Quimby, R.S. (1991). Active phenomena in doped halide glasses. In: Fluoride Glass Fibre Optics (eds. I.D. Aggarwal and G. Lu). Boston: Academic Press. Aggarwal, I.D. and Lu, G. (1991). Fluoride Glass Fiber Optics. Boston: Academic Press. Sanghera, J.S. and Aggarwal, I.D. (1998). Infrared Fiber Optics. Boca Raton: CRC Press. Moore, D.T. (1980). Gradient-index optics: a review. Appl. Opt. 19: 1035–1038. Hornschuh, S., Messerschmidt, B., Possner, T. et al. (2004). Silver ion exchange in glasses of the system Na2O/Al2O3/B2O3/SiO2. J. Non-Cryst. Solids 347: 121–127. Koike, Y., Asakawa, A., Wu, S.P., and Nihei, E. (1995). Gradient-index contact lens. Appl. Opt. 34: 4669–4673. Albert, J. (1998). Permanent photoinduced refractive index changes for Bragg gratings in silicate glass waveguides and fibers. Mater. Res. Bull. 23: 36–41. Hill, K.O., Malo, B., Bilodeau, F. et al. (1993). Bragg gratings fabricated in monomode photosensitive optical fiber by UV exposure through a phase mask. Appl. Phys. Lett. 62: 1035–1036. Othonos, A. and Kalli, K. (1999). Fiber Bragg Gratings: Fundamentals and Applications in Telecommunications and Sensing. London: Artech House Books. Limberger, H.G., Fonjallaz, P.-Y., Salathe, R.P., and Cochet, F. (1996). Compaction – and photoelastic-induced index changes in fiber Bragg gratings. Appl. Phys. Lett. 68: 3069–3071. Kristensen, M. (2001). Ultraviolet-induced processes in germania-doped silica. Phys. Rev. 64: 144201/1–144201/12. Hirao, K. and Miura, K. (1998). Writing waveguides and gratings in silica and related materials by a femtosecond laser. J. Non-Cryst. Solids 239: 91–95. Streltsov, A.M. and Borelli, N.F. (2002). Study of femtosecond-laser-written waveguides in glasses. J. Opt. Soc. Am B. 19: 2496–2504. Glezer, J.N. and Mazur, E. (1997). Ultrafast-laser driven microexplosions in transparent materials. Appl Phys Letters. 71: 882–884. Mazur, E. (2005). Femtosecond laser micromachining of glass for photonics applications. J. Phys. Chem. Glasses. Glezer, J.N., Milosavljevic, M., Finlay, R.J. et al. (1996). 3-D optical storage inside transparent materials. Opt. Lett. 21: 2023–2025. Schaffer, C.B., Brodeur, A., and Mazur, E. (2001). Laser-induced breakdown and damage in bulk transparent materials induced by tightly focused femtosecond laser pulses. Meas. Sci. Technol. 12 (11): 1784–1794. Dutta, N.K. (2015). Fiber Amplifiers and Fiber Lasers. World Scientific Publishing Co. Digonnet, M.J. (2001). Rare-Earth-Doped Fiber Lasers and Amplifiers, 2e. Boca Ratan: CRC press. Ter-Mikirtychev, V. (2014). Industrial Applications of Fiber Lasers, vol. 2014. Cham: Springer.

References

48 Headley, C.M.M. and Bouteiller, J.C. (2004). Raman fiber lasers. In: Raman Ampli-

fiers for Telecommunications (ed. N. IM), 2. New York, NY: Springer. 49 Okada, G., Morrell, B., Koughia, C. et al. (2011). Spatially resolved measurement of

50 51 52 53 54 55 56 57

58 59 60 61 62 63 64

65 66 67

68

high doses in microbeam radiation therapy using samarium doped fluorophosphate glasses: development of a high-resolution high-dose dosimetric detector. Appl. Phys. Lett. 99: 121105–121108. Ehrt, D. (2010). Photoionization of Polyvalent Ions (ed. D. Moncke). New York, USA: Nova Science Publishers Inc. Beall, G.H. and Pinkney, L.R. (1999). Nanophase glass-ceramics. J. Am. Ceram. Soc. 82: 5–15. Wang, Y. and Ohwaki, J. (1993). New transparent vitroceramics codoped with er3+ and Yb3+ for efficient frequency upconversion. Appl. Phys. Lett. 63: 3268–3270. Tick, P.A., Borrelli, N.F., Cornelius, L.K., and Newhouse, M.A. (1995). Transparent glass ceramics for 1300nm amplifier applications. J. Appl. Phys. 78: 6367–6374. Tick, P.A. (1998). Are low-loss glass-ceramic optical waveguides possible? Opt. Lett. 23: 1904–1905. Klug, H.P. and Alexander, L.E. (1974). X-Ray Diffraction Procedures, 2e. New York: Wiley. Dejneka, M. (1998). The luminescence and structure of novel transparent oxyfluoride glass-ceramics. J. Non-Cryst. Solids 239: 149–155. Hopper, R.W. (1985). Stochastic theory of scattering from idealised spinodal structures II. Scattering in general and for teh basic late stage model. J. Non-Cryst. Solids 70: 111–142. Hopper, R.W. (1982). Stochastic theory of scattering from idealised spinodal structures I: structure and autocorellation function. J. Non-Cryst. Solids 49: 263–285. Hendy, S.C. (2002). Light scattering in transparent glass ceramics. Appl. Phys. Lett. 81: 1171–1173. Edgar, A., Williams, G.V.M., and Hamelin, J. (2006). Optical scattering in glass ceramics. Curr. Appl. Phys. 6 (3): 355–358. Draine, B.T. and Flateau, P.J. (1994). Discrete-dipole approximation for scattering calculations. J. Opt. Soc. Am. A 11: 1491–1499. Edgar, A. (2006). Core-Shell particle model for optical transparency in glass ceramics. Appl. Phys. Lett. 89:041909. Goldstein, M. (1963). Theory of scattering for diffusion-controlled phase separations. J. Appl. Phys. 34 (7):1928. Edgar, A. (2007). The core-shell particle model for light scattering in glass-ceramics: Mie scattering analysis and discrete dipole simulations. J. Mater. Sci.-Mater. Electron. 18: S335–S338. Dejneka, M. (1998). Transparent oxyfluoride glass ceramics. MRS Bull.: 57–62. Clara Goncalves, M., Santos, L.F., and Almeida, R.M. (2002). Rare-earth-doped transparent glass ceramics. C. R. Chem. 5: 845–854. Tikhomirov, V.K., Furniss, D., Seddon, A.B. et al. (2002). Fabrication and characterisation of nanoscale, Er3+−doped, ultratransparent oxyflyuoride glass ceramics. Appl. Phys. Lett. 81: 1937–1939. Mortier, M., Goldner, P., Chateau, C., and Genotelle, M. (2001). Erbium-doped glass ceramics: concentration effect on crystal structure and energy transfer between active ions. J. Alloys Compd. 323-324: 245–249.

127

128

4 Optical Properties of Glasses

69 Takahashi, Y., Benino, Y., Fujiwara, T., and Komatsu, T. (2003). Formation mech-

70 71

72

73 74 75 76

77 78

79 80

81 82 83 84 85 86

anism of ferroelastic Ba2TiGe2O8 and second order optical non-linearity in transparent glasses. J. Non-Cryst. Solids 316: 320–330. Zhilin, A.A., Petrovsky, G.T., Goulbkov, V.V. et al. (2004). Phase transformations in Na2O-K2O-Nb2O5-SiO2 glasses. J. Non-Cryst. Solids 345&346: 182–186. Benino, Y., Takahashi, Y., Fujiwara, T., and Komatsu, T. (2004). Second order optical non-linearity of transparent glass ceramic materials induced by an alternating field. J. Non-Cryst. Solids 345&6: 422–427. Biswas, A., Maciel, G.S., Friend, C.S., and Prasad, P.N. (2003). Upconversion properties of a transparent Er3+-Yb3+ co-doped LaF3-SiO2 glass ceramics prepared by sol-gel method. J. Non-Cryst. Solids 316: 393–397. Rowlands, J.A. (2002). The physics of computed radiography. Phys. Med. Biol. 47: R123–R166. Schweizer, S., Secu, M., Spaeth, J.-M. et al. (2004). New developments in X-ray storage phosphors. Radiat. Meas. 38 (4–6): 633–638. Winch, N.M. and Edgar, A. (2011). X-ray imaging using a commercial camera. Nucl. Instrum. Methods A A654: 308–313. Secu, M., Schweizer, S., Spaeth, J.-M. et al. (2003). Photostimulated luminescence from a fluorobromozirconate glass-ceramic and the effect of crystallite size and phase. J. Phys. Condens. Matter. 15: 1097–1108. Schweizer, S., Hobbs, L., Secu, M. et al. (2005). Photostimulated luminescence from fluorochlorozirconate glasses. J. Appl. Phys. 97:0835522/1–8. Edgar, A., Williams, G.V.M., and Appleby, G.A. (2004;108/1–4). Photoluminescence and photo-stimulated luminescence from europium-doped rubidium barium bromide in fluorozirconate glass ceramics. J. Lumin.: 19–23. Schweizer, S. and Johnson, J.A. (2007). Fluorozirconate-based glass ceramic X-ray detectors for digital radiography. Radiat. Meas. 42 (4–5): 632–637. Edgar, A., Williams, G.V.M., Secu, M. et al. (2004). Optical properties of a high-efficiency glass ceramic x-ray storage phosphor. Radiat. Meas. 38 (4–6): 413–416. Edgar, A., Williams, G.V.M., Schweizer, S., and Spaeth, J.M. (2006). Spatial resolution of a glass-ceramic X-ray storage phosphor. Curr. Appl. Phys. 6 (3): 399–402. Schweizer, S., Henke, B., Miclea, P.T. et al. (2010). Multi-functionality of fluorescent nanocrystals in glass ceramics. Radiat. Meas. 45 (3–6): 485–489. Winch, N. and Edgar, A. (2009). Light scattering in glass ceramic x-ray storage phosphors. J. Appl. Phys. 105: 023506–1/8. Fasbender, R., Li, H., and Winnacker, A. (2003). Monte Carlo modeling of storage phosphor plate readouts. Nucl. Instrum. Methods Phys. Res. Sect. A 512 (3): 610–618. Hendy, S.C. and Edgar, A. (2006). Structure of fluorochlorozirconate glasses using molecular dynamics. J. Non-Cryst. Solids 352 (5): 415–422. Edgar, A., Hendy, S.C., and Bayliss, D. (2006). Phase separation in mixed halide Ba-La-Li/Na zirconate glasses: a molecular dynamics study. Phys. Chem. Glasses-Eur. J. Glass Sci. Technol. Part B. 47 (2): 254–258.

129

5 Concept of Excitons Jai Singh 1 , Harry E. Ruda 2 , M.R. Narayan 1 , and D. Ompong 1 1

College of Engineering, Information Technology and Environment, Charles Darwin University, Darwin, NT 0909, Australia Centre for Nanotechnology, and Electronic and Photonic Materials Group, Department of Materials Science, University of Toronto, Toronto, Ontario, Canada 2

CHAPTER MENU Introduction, 129 Excitons in Crystalline Solids, 130 Excitons in Amorphous Semiconductors, 135 Excitons in Organic Semiconductors, 139 Conclusions, 153 References, 154

5.1 Introduction An optical absorption in semiconductors and insulators can create an exciton, which is an electron–hole pair excited by a photon and bound together through their attractive Coulomb interaction. The absorbed optical energy remains held within the solid for the lifetime of an exciton. Because of the binding energy between the excited electron and hole, excitonic states lie within the band gap near the edge of the conduction band. There are two types of excitons that can be formed in nonmetallic solids: Wannier or Wannier–Mott excitons and Frenkel excitons. The concept of Wannier–Mott excitons is valid for inorganic semiconductors such as Si, Ge, and GaAs, because, in these materials, the large overlap of interatomic electronic wave functions enables electrons and holes to be far apart but bound in an excitonic state. For this reason, these excitons are also called large-radii orbital excitons. Excitons formed in organic crystals are called Frenkel excitons. In organic semiconductors/insulators or molecular crystals, the intermolecular separation is large and hence the overlap of intermolecular electronic wave functions is very small, and electrons remain tightly bound to individual molecules. Therefore, the electronic energy bands are very narrow and closely related to individual molecular electronic energy levels. In such solids, the absorption of photons occurs close to the individual molecular electronic states, and excitons are also formed within the molecular energy levels (see, e.g., reference [1]). Such excitons are therefore also called molecular excitons. For details of the theory of Wannier and Frenkel excitons, readers may like to refer to the book by Singh [1]. Another way of understanding excitons in Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

130

5 Concept of Excitons

inorganic and organic semiconductors is on the basis of the dielectric constant of these two solids. For example, in silicon (Si), the static dielectric constant is 12, and that in organic solids and polymers is about 3–4. This gives the binding energy of excitons in organic solids to be nearly four times larger than that in Si, leading to the excitonic Bohr radius in Si to be four times larger than that in organic semiconductors. Thus, the excitonic Bohr radius is larger in inorganic semiconductors, and hence termed as large-radii orbital excitons. The excitonic concept from the theory point of view was initially developed only for crystalline solids, and it used to be believed that excitons cannot be formed in amorphous semiconductors. However, several experiments on photoluminescence measurements have revealed the existence of excitons in amorphous semiconductors as well, and the theory of excitons in such solids was subsequently developed. Here, the concept of excitons in crystalline solids is reviewed briefly first as it is well established, and then amorphous semiconductors will be considered.

5.2 Excitons in Crystalline Solids In Wannier–Mott excitons, the Coulombic interaction between the hole and electron can be viewed as an effective hydrogen atom with, for example, the hole establishing the coordinate reference frame about which the reduced-mass electron moves. If the effective masses of the isolated electron and hole are m∗e and m∗h , respectively, their reduced mass, 𝜇x , is given by: 𝜇x−1 = (m∗e )−1 + (m∗h )−1

(5.1)

Note that, in the case of so-called hydrogenic impurities in semiconductors (i.e. both shallow donor and acceptor impurities), the mass of the nucleus takes the place of one of the terms in Eq. (5.1), and hence the reduced mass is given to a good approximation by the effective mass of the appropriate carrier. In the case of an exciton as the carrier, effective masses are comparable, and hence the reduced mass is markedly lower—accordingly, the exciton binding energy is markedly lower than that for hydrogenic impurities. The energy Ex of a Wannier–Mott exciton is given by (e.g., see reference [1]): ℏ2 K 2 (5.2) − En 2M where Eg is the bandgap energy; ℏK the linear momentum; M (= m∗e + m∗e ) the effective mass associated with the center of mass of an exciton; and En is the exciton binding energy, given by: Rxy 𝜇 e4 𝜅 2 1 (5.3) En = x 2 2 2 = 2 2ℏ ε n n 1 , ε is the static dielectric constant of the solid, where e is the electronic charge, 𝜅 = 4𝜋ε 0 and n is the principal quantum number associated with the internal excitonic states n = 1 (s), 2 (p), etc., as shown in Figure 5.1. Rxy is the so-called effective Rydberg constant of an exciton, given by Ry (𝜇x /me )/ε2 , where Ry = 13.6 eV. According to Eq. (5.3), as stated earlier, the excitonic states are formed within the bandgap near the conduction band edge. However, as the exciton binding energy is very small—for example, a few meV in bulk Si and Ge crystals—exciton absorption peaks can be observed only at very low Ex = Eg +

5.2 Excitons in Crystalline Solids

E

Figure 5.1 Schematic illustration of excitonic bands for n = 1 and 2 in semiconductors. E g represents the energy gap.

Conduction Band n=2 n=1 Eg

Excitonic Band

K

Valence Band

temperatures. For bulk GaAs, the binding energy (n = 1) corresponds to about 5 meV. Following from the hydrogen atom model, the extension of the excitonic wave function can be found from an effective exciton Bohr radius ax , given in terms of the Bohr radius as a0 = h2 /𝜋e2 ; that is, ax = a0 (ε/𝜇x ). For GaAs, this corresponds to about 12 nm or about 21 lattice constants—that is, the spherical volume of the exciton radius contains ∼(ax /a)3 or ∼9000 unit cells, where a is the lattice constant of GaAs. As Rxy ≪ Eg and ax ≫ a, excitons in GaAs are large-radii orbital excitons, as stated in the preceding text. It should be noted that the binding energy of excitons in semiconductors tends to be a strong function of the bandgap. In Figure 5.2a,b are shown the dependencies of Rxy (exciton binding energy) and ax /a on the bandgap of semiconductors, respectively. In the case of excitons having large binding energy and correspondingly small radius (i.e. approaching the size of about one lattice parameter), the excitons become localized on a lattice site, as observed in most organic semiconductors. As stated earlier, such excitons are commonly referred to as Frenkel excitons or molecular excitons. Unlike Wannier–Mott excitons, which typically are dissociated at room temperature, Frenkel excitons are stable at room temperature. For the binding energy of Frenkel excitons, one may refer to Singh [1], for example. Excitons may recombine radiatively, emitting a series of hydrogen-like spectral lines as described by Eq. (5.3). In bulk crystalline (3D) semiconductors such as Si, Ge, and GaAs, exciton lines can be observed only at low temperatures. As the binding energy is small, typically ≤10 meV, excitons in bulk are easily dissociated by thermal fluctuations. The preceding discussion refers to the so-called free excitons formed between the conduction band electrons and valence band holes of a crystalline inorganic semiconductor or insulator. According to Eq. (5.2), such an exciton is able to move throughout a material with a center-of-mass kinetic energy (second term on the right-hand side). It should be noted, however, that free electrons and holes move with a velocity ℏ(dE/dk), where the derivative is taken for the appropriate band edge. To move through a crystal, both electron and hole must have identical translational velocity, thereby restricting the regions in k-space where these excitations can occur with (dE/dk)electron = (dE/dk)hole , commonly referred to as critical points.

131

5 Concept of Excitons

Exciton binding energy (meV)

1000

100

10

1

0.1 0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

3.0

3.5

4.0

Bandgap (eV) (a) Exciton radius/lattice constant

132

100.00

10.00

1.00 0

0.5

1.0

1.5

2.0

2.5

Bandgap (eV) (b)

Figure 5.2 Dependence of (a) exciton binding energy (R∗y ) (Eq. (5.3)) and (b) size in terms of the ratio of the excitonic Bohr radius to lattice constant (a∗B ∕a0 ) as a function of the semiconductor bandgap. Exciton binding energy increases concurrently with marked diminishment of exciton spreading as the bandgap increases. Above a bandgap of about 2 eV, the Wannier-based description is not appropriate.

In quantum wells and other structures of reduced dimensionality, the spatial confinement of both the electron and hole wave functions in the same layer ensures strong excitonic transitions of a few meV below the bandgap, even at room temperature. The binding energy of excitons, Ebx (α, n), in confined systems of dimension α, is given by references [2, 3]: Ebx (α, n) =

Rxy [n + (α − 3)∕2]2

(5.4)

Thus, the binding energy of excitons in quantum wells (α ≈ 2) and n = 1 increases to four times the value in 3D (α= 3). For quantum wires, α= 1, the binding energy becomes infinitely large, and for quantum dots (α= 0), it becomes the same as in 2D. It is this enhancement in the binding energy due to confinement that allows excitonic absorption and photoluminescence to be observed even at room temperature in quantum wells.

5.2 Excitons in Crystalline Solids

Furthermore, the observation of biexcitons [3, 4] and excitonic molecules also becomes possible due to the large binding energy. The ratio of the binding energy of biexcitons, Ebxx to that of excitons, Ebx for n = 1, is usually constant in quantum wells, and, for GaAs quantum wells, one gets Ebxx = 0.228Ebx [2, 3]. 5.2.1

Excitonic Absorption in Crystalline Solids

As exciton states lie below the conduction-band edges in crystalline solids, absorption to excitonic states is observed below the conduction-band edge. According to Eq. (5.2), the difference of energy between the bandgap and excitonic absorption gives the binding energy. As the exciton–photon interaction operator and pair of excited electron and hole and photon interaction operator depend only on their relative motion momentum, the form of these interactions is the same for band-to-band and excitonic absorption. Therefore, for calculating the excitonic absorption coefficient, one can use the same form of interaction as that for band-to-band absorption in Chapter 2, but one has to use the joint density of states for crystalline solids. Thus, the absorption coefficient associated with the excitonic states in crystalline semiconductors is obtained as (e.g. [5]): ( )1∕2 2 4e2 |pxv |2 (ℏ𝜔 + En − Eg )1∕2 , n = 1, 2, … (5.5) αℏ𝜔 = √ 2 𝜇 x εcℏ where ℏ𝜔 is the energy of the absorbed photon, pxv is the transition matrix element between the excitonic state and the valence band, and En is the exciton binding energy corresponding to a state with n = 1, 2, etc. (see Eq. (5.3)). Equation (5.5) is similar to the case of direct band-to-band transitions discussed in Chapter 2, and it is valid only for the photon energies, ℏ𝜔 ≥ Eg . There is no absorption below the excitonic ground state in pure crystalline solids. The absorption of photons to the excitonic energy levels is possible either by exciting electrons to higher energy levels in the conduction band and then by their non-radiative relaxation to the excitonic energy level, or by exciting an electron directly to the exciton energy level. Excitonic absorption occurs in both direct as well as indirect semiconductors. If the valence hole is a heavy hole, the exciton is called a heavy-hole exciton; conversely, if the valence hole is light, the exciton is a light-hole exciton. For large exciton binding energy, say corresponding to n = 1, the excitonic state will be well separated from the edge, Eg , of the conduction band, and then one can observe a sharp excitonic peak at the photon energy ℏ𝜔 = Eg − E1 . The absorption to higher excitonic states, corresponding to n > 1, may not be observed in materials with small binding energies, as these states will be located within the conduction band. The excitonic absorption and photoluminescence in GaAs quantum wells are shown in Figure 5.3a,b, respectively [3]. In Figure 5.3b are also shown the biexcitonic photoluminescence peaks (with superscript XX) observed in GaAs quantum wells. In the absorption spectra (Figure 5.3a), both light-hole (LHX ) and heavy-hole (HHX ) excitons are observed in quantum wells of different well widths indicated by the corresponding subscript. For example, HH100 X and HH100 XX mean heavy-hole exciton and biexciton peaks, respectively, in a quantum well of width 100 Å. It may be noted that the transition matrix element, pxv , in Eq. (5.5) is assumed to be photon-energy independent. It is the average of the linear relative momentum between

133

5 Concept of Excitons

(a) 2.5 HHx130

2.0

αd

1.5

HHx100 HHx80

HHx160

LHx130

LHx100

LHx160

1.0 0.5 0.0

(b) HHxx 160

HHx160

HHxx 130

PL intensity [Arb. units]

134

HHx130 HHx100

1.52

1.53

1.54

1.55

HHx80

1.56

1.57

1.58

Photon energy (eV)

Figure 5.3 (a) Low-temperature absorption, and (b) photoluminescence spectra in GaAs quantum wells of different well widths. (HHx100 and HHxx 100 denote heavy-hole exciton and biexcition, respectively, in a quantum well of width 100 Å.) The photoluminescence data are obtained using HeNe laser excitation [3].

electron and hole in an exciton. However, if one applies the dipolar approximation, the transition matrix element thus obtained depends on the photon energy [5, 6], which gives a different photon-energy dependence in the absorption coefficient. This aspect of the absorption coefficient is described in detail in Chapter 3. Excitonic absorption is spectrally well located and very sensitive to optical saturation. For this reason, it plays an important role in nonlinear semiconductor devices (nonlinear Fabry–Perot resonator, nonlinear mirror, saturable absorber, and so on). For practical purposes, the excitonic contribution to the overall susceptibility 𝜒 ex around the resonance frequency 𝜈 ex can be written as: 𝜒ex = −A0

(𝜐 − 𝜐0 ) + jΓex (𝜐 − 𝜐0 )2 + Γ2ex (1 + S)

(5.6)

with Γex being the line width and S = I/I S , the saturation parameter of the transition. For instance, in GaAs multiple quantum wells (MQWs), the saturation intensity I S is as low

5.3 Excitons in Amorphous Semiconductors

as 1 kW cm−2 , and Γex (≈3.55 meV at room temperature) varies with the temperature according to: Γ1 (5.7) Γex = Γ0 + ) ( ℏ𝜔LO exp 𝜅T − 1 where ℏΓ0 is the inhomogeneous broadening (≈2 meV), ℏΓ1 the homogeneous broadening (≈5 meV), and ℏ𝜔LO the longitudinal optical-phonon energy (≈36 meV). At high carrier concentrations (provided either by electrical pumping or by optical injection), an efficient mechanism of saturating the excitonic line is the screening of the Coulombic attractive potential by free electrons and holes. A number of more complex pairings of carriers can also occur, which may also include fixed charges or ions. For example, for the case of three charged entities, with one being an ionized donor impurity (D+ ), the following possibilities can occur: (D+)(+)(−), (D+)(−)(−), and (+)(+)(−) as excitonic ions, and (+)(+)(−)(−) and (D+)(+)(−)(−) as biexcitons or even bigger excitonic molecules (see, e.g., references [2, 3, 7]). Complexity abounds in these systems as each electronic level possesses a fine structure corresponding to allowed rotational and vibrational levels. Moreover, the effective mass is often anisotropic. Note that, when the exciton or exciton complex is bound to a fixed charge such as an ionized donor or acceptor center in the material, the exciton or exciton complex is referred to as a bound exciton. Indeed, bound excitons may also involve neutral fixed impurities. It is usual to relate the exciton in these cases to the center binding them—thus, if an exciton is bound to a donor impurity, it is usually termed a donor-bound exciton.

5.3 Excitons in Amorphous Semiconductors As stated earlier, the concept of excitons is traditionally valid only for crystalline solids. However, several observations in the photoluminescence spectra of amorphous semiconductors have revealed the occurrence of photoluminescence associated with the singlet and triplet excitons (see, e.g., reference [5]). Applying the effective-mass approach, a theory for the Wannier–Mott excitons in amorphous semiconductors has been developed in real coordinate space [7–10]. The energy W x of an exciton thus derived is obtained as: P2 (5.8) − En (S) Wx = Eg + 2M where Eg is the optical gap, P is the linear momentum associated with the exciton’s center of mass motion, and En (S) is the binding energy of excitons, given by: En (S) = where

𝜇x e 4 𝜅 2 2ℏ2 ε′ (S)2 n2

[ ]−1 (1 − S) ε (S) = ε 1 − A ′

(5.9)

(5.10)

with S being the spin of an exciton (S = 0 for singlet and S = 1 for triplet), and A is a material-dependent constant representing the ratio of the magnitude of the Coulomb

135

136

5 Concept of Excitons

and exchange interactions between the electron and hole of an exciton. Equation (5.9) is analogous to Eq. (5.3) obtained for excitons in crystalline solids for S = 1. This is because Eq. (5.3) is derived within the large-radii orbital approximation, which neglects the exchange interaction and hence is valid only for triplet excitons [1, 11]. As amorphous solids lack long-range order, the exciton binding energy is found to be larger in amorphous solids than in their crystalline counterparts; for example, in hydrogenated amorphous silicon (a-Si:H), the binding energy is higher than in crystalline silicon (c-Si). This is the reason why it is possible to observe the photoluminescence of both singlet and triplet excitons in a-Si:H [12] but not in c-Si. According to Eq. (5.9), the singlet exciton binding energy corresponding to n = 1 becomes: (A − 1)2 𝜇x e4 𝜅 2 (5.11) E1 (S = 0) = 2A2 ℏ2 𝜖 2 and the triplet exciton binding energy becomes: 𝜇x e 4 𝜅 2 (5.12) 2ℏ2 ε2 From Eqs. (5.11) and (5.12), we get the relation between the singlet and triplet exciton binding energies as: E1 (S = 1) =

(A − 1)2 E1 (S = 1) (5.13) A2 For hydrogenated amorphous silicon (a-Si:H), A is estimated to be about 10 [5], which gives from Eq. (5.13) the result that the singlet exciton binding energy is about 81% of the triplet exciton binding energy. As 𝜇x is different in the extended and tail states, there are four possibilities through which an exciton can be formed in amorphous solids: (i) both the excited electron and hole are in their extended states; (ii) the excited electron is in the extended states and the hole is in the tail states; (iii) the excited electron is in the tail states and the hole is in the extended states; and (iv) both the excited electron and hole are in their tail states. Using the expressions of effective mass of charge carriers derived in the extended and tail states in Chapter 3, we get the effective mass of the electron and hole in the extended states in a-Si:H as m∗ex 0.34 me and in the tail states as m∗et = 7.1 me . This gives 𝜇x = 0.17me for possibility (i) when both the excited carriers are in their extended states; 𝜇x = 0.32me for possibilities (ii) and (iii) when one of the excited charge carriers is in the tail and the other in the extended states; and 𝜇x = 3.55 me for possibility (iv) when both electron and hole are in their tail states in a-Si:H. Using ε= 12, one obtains from Eq. (5.12) the triplet exciton binding energies in a-Si:H as 16 meV for possibility (i); 30 meV for possibilities (ii) and (iii); and 0.34 eV for possibility (iv). The corresponding singlet exciton binding energies in a-Si:H is obtained from Eq. (5.13) as E1 (S = 0) = 13 meV for possibility (i), 24 meV for possibilities (ii) and (iii), and 0.28 eV for possibility (iv). These energies are measured from the conduction band edge or the conduction band mobility edge, Ec , and are schematically shown in Figure 5.4. Accordingly, the triplet exciton states lie below the singlet exciton states. E1 (S = 0) =

5.3 Excitons in Amorphous Semiconductors

Conduction states Eb (n = 1; s = 0)

ΔEex

Eb (n = 1; s = 1)

ESinglet state ETriplet state

Eopt

Valence states

Figure 5.4 Schematic illustration of singlet and triplet excitonic states in amorphous solids. ΔE ex represents the energy difference between singlet and triplet excitonic states.

The excitonic Bohr radius is also found to be different for singlet and triplet excitons in amorphous semiconductors. Writing the exciton energy as: E1 (S) = −

𝜅e2 2ax (S)

(5.14)

we get from Eqs. (5.11)–(5.13): ax (S = 0) =

A2 a (S = 1) (A − 1)2 x

(5.15) m

which gives for a-Si:H with A = 10, ax (S = 0) ≈ 54 ax (S = 1), where ax (S = 1) = 𝜇 e a0 , x and a0 = 0.0529 nm is the Bohr radius. Accordingly, for triplet excitons in a-Si:H, we get ax (S = 1)= 3.73 nm for possibility (i), 1.98 nm for possibilities (ii) and (iii), and 0.18 nm for possibility (iv); and, for the corresponding excitonic Bohr radius for a singlet exciton, we get for possibility (i) ax (S = 0) = 4.66 nm, for possibilities (ii) and (iii) ax (S = 0) = 2.5 nm, and for possibility (iv) ax (S == 0) ≈ 0.233 nm. The excitonic Bohr radius plays a very significant role in the radiative recombination of excitons, because this is the average separation between the electron and hole in an exciton prior to their recombination. Therefore, the rates of spontaneous emission depend on the excitonic Bohr radius (see Chapter 7). 5.3.1

Excitonic Absorption in Amorphous Solids

In amorphous semiconductors, the excitonic absorption and photoluminescence can be quite complicated. According to Eq. (5.8), the excitonic energy level lies below the optical bandgap by an energy equal to the binding energy given in Eq. (5.9). However, as stated in the previous section, there are four possibilities of transitions for absorption in amorphous semiconductors: (i) valence-extended to conduction-extended states, (ii) valence-tail to conduction-extended states, (iii) valence-extended to conduction-tail states, and (iv) valence-tail to conduction-tail states. These possibilities will have different optical gap energies, Eg , and different binding energies. Possibility (i) will give rise

137

138

5 Concept of Excitons

to absorption as in the free exciton states, possibilities (ii) and (iii) will give absorption in the bound exciton states because one of the charge carriers is localized in the tail states, and the absorption through possibility (iv) will create localized excitonic geminate pairs. This can be visualized as follows: if an electron – hole pair is excited by a high-energy photon through possibility (i) and forms an exciton, initially its excitonic energy level and the corresponding Bohr radius will have a reduced mass corresponding to both charge carriers being in the extended states. As such an exciton relaxes downward nonradiatively, the binding energy and excitonic Bohr radius will change because the effective mass changes in the tail states. When both charge carriers reach the tail states, possibility (iv), although the pair is localized, its excitonic Bohr radius will be maintained. In this situation, the excitonic nature breaks down as both charge carriers are localized, and the excitonic wave function cannot be used to calculate any physical property of these localized carriers. One has to use the individual localized wave functions of the electron and hole. For calculating the excitonic absorption coefficient in amorphous semiconductors, one can use the same approach as presented in Chapter 3, and similar expressions, such as Eqs. (3.4) and (3.14), will be obtained. This is because (i) the transition matrix element remains the same for excitonic absorption and for band-to-band free-carrier absorption, and (ii) the concept of the joint density of states applied to excitonic absorptions in crystalline solids is not applicable to excitonic absorptions in amorphous solids. Therefore, by replacing the effective masses of charge carriers by the excitonic reduced mass and the distance between the excited electron and hole by the excitonic Bohr radius, one can use Eqs. (3.4) and (3.14) for calculating the excitonic absorption coefficients for transitions corresponding to the preceding four possibilities in amorphous semiconductors. Thus, one obtains two types of excitonic absorption coefficients for amorphous solids. The first type is obtained by assuming that the transition matrix element is independent of the photon energy but depends on the excitonic Bohr radius as: 1∕2

[αℏ𝜔]1∕2 = Bx [ℏ𝜔 − Eg ]

(5.16)

where Bx =

𝜇x e 2 ax 4ncε0 ℏ3

(5.17)

The coefficient Bx in Eq. (5.17) is obtained by replacing m∗e and m∗h in Eq. (3.11) by 𝜇x , and L by ax . The absorption coefficient derived in Eq. (5.16) has the same photon-energy dependence as Tauc’s relation obtained for the band-to-band free carrier absorption (see Chapter 3). The second expression is obtained by applying the dipole approximation as: [αℏ𝜔] = B′x (ℏ𝜔)2 (ℏ𝜔 − Eg )2

(5.18)

where B′x =

3 𝜇xa 2 e2 x ncε0 2𝜋 2 ℏ7 𝜐𝜌A

(5.19)

Here, again, reh in Eq. (3.15) has been replaced by ax . Thus, the excitonic absorption coefficient depends on the photon energy in the same way as does the band-to-band absorption coefficient described in Chapter 3. This is probably the reason why distinct excitonic absorption peaks in amorphous semiconductors have not yet been observed

5.4 Excitons in Organic Semiconductors

to the best of our knowledge, but excitonic photoluminescence peaks have been observed [13].

5.4 Excitons in Organic Semiconductors The organic semiconductors consist of organic molecules, which are usually hydrocarbons. A hydrocarbon molecule usually consists of hydrogen and carbon atoms bonding sometimes with other atoms of oxygen and nitrogen, depending on the structures. The organic solids are formed by the weak Van der Waals forces, leading to weak bonding caused by the weak overlap of the electronic wave functions between neighboring molecules. The result of this weak bonding is that the inter-molecular separation in organic solids is usually much larger than those in inorganic solids, leading to much narrower electronic bands than in inorganic solids, and hence the energies of the valence and conduction bands of solids can be well approximated by those of the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbitals (LUMOs) of individual molecules, respectively. In the earlier studies on excitons in organic solids, these used to be referred as molecular excitons, because both the excited electron and hole are excited on the same molecule to form Frenkel excitons [1, 14]. Although approximate, it is now fully accepted that the HOMO and LUMO of individual molecules represent the valence and conduction bands, respectively, and the corresponding energy gap represents the energy gap of an organic semiconductor. As described in the following text, if not all, most optical, electronic, and opto-electronic properties of organic semiconductors are studied using the HOMO and LUMO energies for the energy band energies. There are two classes of organic semiconductors: (i) low-molecular-weight compounds, and (ii) polymers. The common feature of both classes is that the constituent molecules are planar, and their carbon atoms are bonded through sp3 hybridization. In such a bonding, the 𝜎 bonds are formed between the planar s, px , and py orbitals, and these bonds are much stronger, contributing to the stability of the planar structure. The pz components of the p-bonds, known as the 𝜋-bonds, are loosely bonded and form delocalized conjugated bonds, which contribute to the electronic and optical properties of molecules. As a result, the low-energetic electron excitations of conjugated molecules occur through π-π*–type transitions from the HOMO to the LUMO energy levels. As described earlier, the energy gap between the LUMO and HOMO is identified as the energy band gap in organic semiconductors, which is usually between 1.5 and 3 eV. Such transitions cause the absorption or emission in the visible region of solar radiation. The energy gap may be controlled by the degree of conjugation of the individual systems, which opens various possibilities for the modification of optoelectronic properties of organic semiconductors [15, 16]. Another common aspect in the electronic and opto-electronic properties of organic semiconductors is that, as soon as electronic excitation occurs from HOMO to LUMO energy levels, either through optical excitations or charge carrier injections, the excited electron and hole pairs in organic semiconductors form Frenkel excitons instantaneously. This is because the dielectric constant of most organic solids is about 3–4, which is much lower than that of inorganic semiconductors—for example, it is 12 in Si. 2 This means that the binding energy EB = 𝜅eεr between the excited pair of electron (e)

139

140

5 Concept of Excitons

and hole (h) is at least four times larger than that in Si. (Here, k = (4𝜋εo )−1 = 9 × 109 , e is electronic charge, ε is the dielectric constant of the organic material, and r is the average separation between the excited e-h pair.) This is the reason why organic semiconductors are also known as excitonic semiconductors, and this is the key difference between inorganic semiconductors such as Si and Ge and organic semiconductors such as poly(thiophene) (P3HT) and poly(fluorene-alt-pyridine) (PFPy). In this section, a theoretical study of the optical properties of organic semiconductors is examined with a view to review its applications in devices such as organic solar cells (OSCs) and organic light-emitting devices (OLEDs). From the theory point of view, the operation of OSCs is opposite to that of OLEDs based on electroluminescence (EL): one shines light on OSCs to generate current, and one passes current in OLEDs to generate light. In practice, however, it is not a device with an identical structure on which if you shine light it will generate current and if you pass current it will generate light. The device structure of the organic layers in OSCs and OLEDs is very different from each other. For understanding the operation of OSCs and OLEDs, it is very important to understand the opto-electronic properties, such as photon absorption, photoluminescence, electroluminescence, and charge carrier generation in organic semiconductors. In Section 5.4.1, the theory of optical absorption is presented, and exciton generation, photoluminescence, and electroluminescence are presented in Section 5.4.2, followed by the exciton dissociation in Section 5.4.3. 5.4.1

Photoexcitation and Formation of Excitons

Absorption of light of energy larger than or equal to the bandgap of an organic solid results in the production of excited electron (e) and hole (h) pairs. An electron is excited to the LUMO, leaving a hole in the HOMO, which instantly forms an exciton due to the strong Coulomb interaction between e and h caused by the low dielectric constant (ε = 3–4) of organic solids. In organic solids, as stated earlier, the energy bandgap is equal to the energy difference between the LUMO and HOMO energy levels. As the electronic intermolecular interaction is weak in organics, the formation of such excitons is usually of Frenkel type [17–20]. There are two possible spin configurations for the formation of excitons: (i) singlet excitons, and (ii) triplet excitons, as shown in Figure 5.5. Absorption of a photon in any organic (inorganic) semiconductor can excite an electron from the HOMO (valence band) to the LUMO (conduction band), leaving a hole behind in the HOMO (valence band). Such an excited state will usually be a

(a)

+

,

,



(b)

Figure 5.5 Schematic diagrams of (a) one-spin configuration for the singlet; and (b) three-spin configurations for the triplet excitons.

5.4 Excitons in Organic Semiconductors

singlet because of the spin conservation. In organic semiconductors, usually a triplet state lies at a lower energy than a singlet state, and hence it can be excited from the singlet excited state through intersystem crossing (ISC) [21]. The mechanism of ISC will be described in the following text (Section 5.4.1.2). The triplet state can also be excited directly by photons of energy larger than or equal to the triplet mediated by the exciton–spin–orbit–photon interaction operator, as described in the following text. The mechanism of absorption of a photon and the formation of a singlet exciton due to exciton–photon interaction is presented in Section 5.4.1.1, and for a triplet exciton due to the exciton-spin-orbit-photon interaction in Section 5.4.1.2. 5.4.1.1

Photoexcitation of Singlet Excitons Due to Exciton–Photon Interaction

We consider an organic solid in which an electron is excited directly from HOMO to LUMO upon photon absorption without any change in its spin configuration. The inter̂Is , between radiation and a pair of electron and hole is given by [6]: action operator, H ) ( e e HIs = − (5.20) p − p .A m∗e e m∗h h where the subscript I and s denote interaction and singlet, respectively; m∗e and m∗h are the effective masses; and pe and ph are the linear momenta of the excited electron and hole, respectively; e is the electronic charge; and A is the vector potential of photons, given by: ∑ A0s (ε̂𝜆 e−i𝜔𝜆 t C𝜆+ + ε𝜆 C𝜆 ei𝜔𝜆 t ) (5.21) A= 𝝀

)1∕2 ( where A0s = 2ε nℏ2 V 𝜔 , n is the refractive index, V is the illuminated volume of the 0 𝜆 material, 𝜔𝜆 is the photon frequency, ε0 is the vacuum permittivity, ℏ is the reduced Planck’s constant, and ε̂𝜆 is the unit polarization vector of photons. C𝜆+ (C𝜆 ) is the creation (annihilation) operator of a photon in mode 𝜆. The first term of A represents the emission, while the second term relates to absorption of a photon. In this chapter, only the absorption term is considered from here onward, and emission will be considered in Chapter 6 on photoluminescence. Using the center of mass coordinate, Rx = (m∗e re + m∗h rh )∕M, where M = m∗e + m∗h and ̂as for singlet, can relative coordinate r = (re − rh ), the absorption interaction operator, H be written as [22, 23]: ( )1 2 e ∑ ℏ Has = − ei𝜔𝜆 t (ε𝜆 .p)C𝜆 (5.22) 2 𝜇x 𝜆 2ε0 n V 𝜔𝜆 where the H as denotes the absorption interaction operator, with only the second term of Eq. (5.21), and p = − iℏ∇r is the relative momentum between electron (e) and hole (h). The interaction operator in Eq. (5.22) can be written in the second quantized form as [22]: ( )1 ∑ 2 e ℏ ̂as = − ei𝜔𝜆 t Zlm𝜆 B+LHlm C𝜆 (5.23) H 2 𝜇x 𝜆 2ϵ0 n V 𝜔𝜆 [ ( ) ( ) ( ) ( )] + + − 12 + a+Ll − 12 dHm + 12 [5] is the creation operwhere B+LHlm = √1 a+Ll + 12 dHm 2 ator of a singlet exciton by exciting an electron in LUMO at site l and a hole in HOMO at

141

142

5 Concept of Excitons

site m, and Zlm𝜆 = ∫ 𝜙Ll (re )ε𝜆 .p𝜙H m (rh )dre drh . Applying Fermi’s golden rule and two-level ̂ approximation with Has in Eq. (5.23) as the perturbation operator, the rate of absorption of photons to generate singlet excitons in organic semiconductors is obtained as [22]: √ 4𝜅e2 εa2s (ELUMO − EHOMO )3 Ras = (5.24) 3ℏ4 c3 ε1.5 where 𝜅 = (4𝜋ε0 )−1 = 9 × 109 , c is the speed of light, n2 = ε is used to write Eq. (5.24) in terms of the static dielectric constant ε, and as is the excitonic Bohr radius of singlet excitons. 5.4.1.2

Excitation of Triplet Excitons

The triplet excitons can be excited in two ways in organic semiconductors—directly through strong exciton–spin–orbit–photon interactions, and indirectly through the ISC, as described in the following text. Direct Excitation to Triplet States Through Exciton–Spin–Orbit–Photon Interaction By absorp-

tion of a photon of adequate energy, an electron can be directly excited to a triplet spin configuration by flipping its spin through the electron–spin–orbit–photon interaction. Using the time-dependent exciton–spin–orbit–photon interaction operator derived in [24, 25] as a perturbation operator and Fermi’s golden rule, the rate Rat of triplet exciton absorption is obtained as [22, 24]: Rat =

32e6 Z2 𝜅 2 ε(ELUMO − EHOMO ) 𝜇x4 c7 𝜀0 a4t

(5.25)

where Z is the heaviest atomic number present in the organic molecule/solid, and at is the excitonic Bohr radius of triplet excitons. The rates of singlet and triplet absorption derived in Eqs. (5.24) and (5.25), respectively, are sensitive to the energy gap, (ELUMO − EHOMO ), of the organic material. For various organic materials used as donors in the fabrication of OSCs, the rates of formation of singlet and triplet excitons are calculated using their (ELUMO − EHOMO ) obtained from the experimental studies and presented in Table 5.1. The inverse of these rates gives , i = s or t). the time of formation of the corresponding exciton, (𝜏ai = R−1 ai According to Eqs. (5.24) and (5.25), Ras does not depend on Z, but Rat depends on Z2 . Thus, in the case of a triplet excitation, the higher the Z value, the faster will be Rat . Normally, organic materials consist of hydrocarbons, where carbon, C, has the highest atomic number. In experimental studies [5, 26, 27, 36, 37], it is found that incorporating the heavy metal atoms such as Ir in the donor organic materials enhances the photon–electron conversion efficiency of OSCs. Therefore, we have calculated the rates and corresponding formation times of singlet and triplet excitons with Z = 6 for C and Z = 77 for Ir, for several organic materials used as donors in OSCs, and the results are presented in Table 5.1. For all organic materials used as donors and listed in Table 5.1, the rate of excitation of singlet excitons is five orders of magnitude larger than that of triplet excitons. This is because Ras is highly sensitive to the absorption energy/energy gap Eg3 = (ELUMO − EHOMO )3 , while Rat is only linearly dependent on Eg . According to Table 5.1, TFB has the highest absorption energy of 3.59 eV and provides the fastest rate of absorption of 6.36 × 1010 s−1 for singlet excitons and 5.28 × 105 s−1 for triplet excitons

5.4 Excitons in Organic Semiconductors

Table 5.1 Rates of absorption of photons to form singlet excitons due to exciton–photon interaction and triplet excitons due to exciton–photon–spin–orbit interaction, calculated using Eqs. (5.24) and (5.25), respectively, at the listed energy gaps E g = E LUMO − E HOMO (eV) of various donor organic materials. Organic material

Eg

Ras 1010

𝝉 as 10−11

Rat (C) 103

𝝉 at (C) 10−4

Rat (Ir) 105

𝝉 at (Ir) 10−6

(eV)

(s−1 )

(s)

(s−1 )

(s)

(s−1 )

(s)

Ref.

PCBM

2.40

1.90

5.26

2.14

4.67

3.53

2.83

[26]

PFPy

2.94

3.49

2.86

2.62

3.81

4.32

2.31

P3HT

2.10

1.27

7.85

1.87

5.34

3.09

3.24

[27]

α-NPD

3.10

4.10

2.44

2.77

3.61

4.56

2.19

[28]

P3OT

1.83

0.84

11.9

1.63

6.12

2.69

3.72

[29]

Pt(OEP)

1.91

0.96

10.4

1.70

5.87

2.81

3.56

[30]

MEV-PPV

2.17

1.41

7.12

1.94

5.16

3.19

3.13

[31]

PPV

2.80

3.02

3.31

2.50

4.00

4.12

2.43

[32]

MDMO-PPV

2.20

1.46

6.83

1.96

5.09

3.23

3.09

[33]

PEOPT

1.75

0.74

13.6

1.56

6.40

2.57

3.89

[34]

PTPTB

1.70

0.68

14.8

1.52

6.59

2.50

4.00

BBL

1.90

0.94

10.6

1.70

5.90

2.79

3.58

F8BT

1.80

0.80

12.5

1.61

6.22

2.65

3.78

PFB

2.81

3.05

3.28

2.51

3.99

4.13

2.42

TFB

3.59

6.36

1.57

3.20

3.12

5.28

1.89

[35]

Abbreviations: PCBM, [6,6]-phenyl-C61 -butyric acid methyl ester; PFPy, poly(fluorene-alt-pyridine); P3HT, poly(thiophene); α-NPD, N,N′ -diphenyl-N,N′ -bis(1-naphthyl)-1-1′ biphenyl-4,4′′ diamine; P3OT, poly(3-octylthiophene-2,5-diyl); Pt(OEP), platinum octaethyl porphyrin; MEV-PPV, poly(2-methoxy-5-(2′ -ethylhexyloxy)-1,4-phenylenevinylene); MDMO-PPV, poly[2-methoxy-5-(3′ ,7′ -dimethyloctyloxy)-1,4-phenylenevinylene]; PEOPT, poly(3-(4′ -(1′′ ,4′′ ,7′′ -trioxaoctyl)phenyl)thiophene); PTPTB, poly-(N-dodecyl-2,5-bis(2′ -thienyl)pyrrole-(2,1,3-benzothiadiazole)); BBL, poly(benzimidazobenzophenanthroline ladder); F8BT, poly(9,9-dioctylfluorene-cobenzothiadiazole); PFB, poly(9,9-dioctylfluorene-co-bis-N,N-(4-butylphenyl)-bis-N,N-phenyl-1,4-phenylenediamine); TFB, poly(9,9-dioctylfluoreneco-N-(4-butylphenyl) diphenylamine).

as compared to other organic materials. Therefore, utilizing wide-band-gap materials can be expected to greatly enhance the singlet and triplet absorption rates in OSCs. According to Table 5.1, the rate of triplet excitation by incorporating Ir atom (Z = 77) increases by two orders of magnitude in all donors than without it. This clearly shows that the incorporation of heavy-metal atoms enhances the rate of triplet excitation in OSCs due to enhanced exciton–spin–orbit–photon interaction. Indirect Excitation of Triplet Excitons Through Intersystem Crossing and Exciton–Spin–Orbit– Phonon Interaction In this case, an exciton is already excited in the singlet exciton state

in an organic molecule of an organic solid, and the molecule also has a triplet state at a lower energy by ΔE such that the molecular vibrational energies of singlet and triplet overlap as shown in Figure 5.3. If the singlet exciton is excited at a higher molecular vibrational energy state of a molecule overlapping with a molecular vibrational energy

143

144

5 Concept of Excitons

state of the triplet state, then the singlet can get transferred to the triplet; however, the spin configuration will not flip to a triplet configuration. The flipping of the spin can only occur through the exciton–spin–orbit interaction. Accordingly, the mechanism of ISC can occur through the combination of excitons, molecular vibrations, and spin–orbit interactions. It is interesting to note that, although the ISC has been known for very long, no such interaction operator was known to exist in the literature until recently. We have derived a new exciton–spin–orbit–phonon interaction operator suitable for ISC in organic solids [21], as described in the following text. The stationary part of the spin–orbit interaction for an exciton in a molecule consisting of N atoms can be written as [21, 24]: ( ) N Zn e2 g𝜅 ∑ Zn (5.26) s ⋅ l + 3 sh ⋅ lhn , HSO = − 2 3 e en 2𝜇x c2 n ren rhn where se (sh ) is the electron (hole) spin, len = ren × pe is the electron angular momentum, and ren (pe ) is the position vector (orbital momentum) of the electron from the nth nucleus. Similarly, lhn = rhn × ph is the hole angular momentum, and rhn (ph ) is the position vector (orbital momentum) of the hole from the nth nucleus. For a non-rigid structure, Eq. (5.26) can be expanded in Taylor series about the equilibrium positions of molecules. Terminating the expansion at the first order, we get: 0 1 + HSOv HSO = HSO

(5.27)

0 where HSO is the zeroth-order term and represents the interaction in a rigid structure, 1 and HSOv is the first-order term that gives the interaction between exciton–spin– orbit–molecular vibration interactions, and it is obtained as: ( ) 3e2 g𝜅Z ∑ se ⋅ le sh ⋅ lh 1 Rnv , + 4 (5.28) HSOv = − 2𝜇x2 c2 n,v re4 rh

where Rnv is the molecular displacement from the equilibrium position due to the intramolecular vibrations. In Eq. (5.28), the subscript n on ren and rhn is dropped using −4 the approximation that the quantity within parentheses in Eq. (5.28) depends on ren −4 and rhn ; thus, the nearest and heaviest nucleus to the electron and hole is expected to play the dominant influence, and, as such, the presence of other nuclei may be neglected. This approximation helps in reducing the summation to only one nucleus for each electron and hole, and hence the subscript n will be dropped here onward. In carrying out the Taylor series expansion, it is further assumed that the distances ren and rhn of the electron and hole with reference to the individual nuclei of a molecule can be replaced by their distances re and rh , respectively, with reference to the equilibrium position of the individual molecules. This approximation may be regarded to be quite justified within the Born–Oppenheimer approximation regime. Applying the preceding approximation, Rnv = Rv can be expressed in the second quantization as [21, 38]: Rv = (qv0 − q00 )(b+v + bv ),

(5.29)

where b+v (bv ) is the vibrational creation (annihilation) operator in vibrational mode v.

5.4 Excitons in Organic Semiconductors

For expressing the operator in Eq. (5.28) in the second quantization, we can write the field operator for an electron in the LUMO and that of a hole in the HOMO, respectively, as: ∑ 𝜑LUMO aL (𝜎e ), (5.30a) 𝜓 ̂e = 𝜎

e ∑ 𝜑HOMO dH (𝜎h ), 𝜓 ̂h =

𝜎h

dH (𝜎h ) = a+H (−𝜎h ),

(5.30b)

where 𝜑LUMO and 𝜑HOMO are the wavefunctions of the electron in the LUMO and hole in the HOMO, respectively. Using Eqs. (5.30), the interaction operator in Eq. (5.28) can be expressed in the second quantization as: 24e2 g𝜅Z ∑ ̂1 ≈ − H (qv0 − q00 )aL (𝜎e )dH (𝜎h )𝛿𝜎e ,𝜎h (se .le + sh .lh )(b+v + bv ) (5.31) SOv 𝜇x2 c2 rx2 v,𝜎e ,𝜎h where rx is the average separation between the electron and hole in the exciton, and it is approximated as: ( r )−4 ⟨𝜑HOMO ∣ re−4 ∣ 𝜑LUMO ⟩ ≈ ⟨𝜑HOMO ∣ rh−4 ∣ 𝜑LUMO ⟩ ≈ x . (5.32) 2 Using the interaction operator in Eq. (5.28) as a perturbation operator and Fermi’s golden rule, the rate of intersystem K isc crossing is obtained as [21]: Kisc =

3072𝜋 2 𝜀6 𝜅 2 Z2 e4 ℏ3 (ΔE)2 𝜇x4 c4 a6x (ℏ𝜔v )3

s−1

(5.33)

where ΔE is the exchange energy difference between singlet and triplet energy states (see Figure 5.6), ax is the excitonic Bohr radius, and ℏ𝜔v is the molecular vibrational energy. It may be noted that it is the singlet excitonic Bohr radius ax that should be used in the calculation, which is the state prior to the transition. The derivation of the rate in Eq. (5.33) also clarifies how the phenomenon of ISC occurs. An exciton is first excited to the singlet exciton state, which is higher in energy than the triplet state. The vibrational Figure 5.6 Schematic illustration of a singlet excited state whose molecular vibrational energy states overlap with those of a triplet excited state that is at a lower electronic energy by ΔE.

ΔE Triplet excited state

145

146

5 Concept of Excitons exp Table 5.2 The calculated intersystem crossing rate K isc from Eq. (5.33) and experimental rates (Kisc ) for some OSC materials along with their highest atomic number (Z) and singlet–triplet energy difference (ΔE). Here we have used ε = 3, ax = 4.352nm, and 𝜔v = 8 × 1014 s−1 .

Organic material

Z

𝚫E (eV)

K isc (s−1 )

exp −1 Kisc (s )

NPD (Ir doped)

77

0.90

1.1 × 1011

[39]

CBP (Ir doped)

77

0.90

1.1 × 1011

[39]

P3HT

16

0.80

3.7 × 109

SubPc

9

0.71

9.2 × 108

9.1 × 108

[41]

F8BT

16

0.70

2.8 × 109

1.2 × 107

[42]

8

6

Ref.

[40]

Toluene

6

0.70

4.0 × 10

8.5 × 10

[43]

Naphthalene

6

1.47

1.8 × 109

5.0 × 106

[44]

1-Bromonaphthalene

35

1.30

4.7 × 1010

≈×109

[45]

Benzophenon

16

0.30

5.2 × 108

≈×1010

[44]

>×10

[46]

Platinum-acetylide

78 ′

10

0.80

8.8 × 10

11



Abbreviations: NPD, N,N -bis (naphthalen-1-yl)-N, N -bis(phenyl)-benzidine; CBP, 4,4′ -bis(9-carbazolyl)-1,1′ -biphenyl; P3HT, poly(3-hexylthiophene); SubPc, boron subphthalocyanine chloride; F8BT, poly(9,9-dioctylfluorene-cobenzothiadiazole).

states of both singlet and triplet overlap in energy as shown in Figure 5.6. Thus, the exciton–spin–orbit–molecular vibration interaction flips the spin to the triplet state, which then gets transferred to the triplet state. This is the reason why K isc in Eq. (5.33) vanishes if ΔE = 0. It is unlikely that the singlet and triplet states in any molecule can become isoenergetic, and, if that is the case, then the singlet absorption will dominate. The rate of intersystem crossing K isc in Eq. (5.33) is calculated for some organic materials used in the fabrication of OSCs and listed in Table 5.2. Also listed in Table 5.2 are the corresponding experimental values of K isc for comparison. According to Table 5.2, the calculated rates K isc are found to be in reasonable agreement with experimental results, and minor discrepancies may be attributed to the approximations used in deriving Eq. (5.33). The rate in Eq. (5.33) can be applied to calculate the ISC rate in any molecular solids. Comparing the rates derived in Eqs. (5.24), (5.25), and (5.33) as listed in Tables 5.1 and 5.2, we find that, in most organic solids, the rate of singlet absorption Ras is of the order of 1011 s−1 and formation time 𝜏 as ≥ 10 ps, and the rate of the direct triplet absorption is Rat ≈ 103 s−1 and 𝜏 at ≈ 10−4 s without the involvement of any heavy metal atoms, but with the incorporation of Ir as the heavy metal atom, we get Rat ≈ 105 s−1 and 𝜏 at ≈ 1 μs. In comparison, the rate of intersystem crossing K isc ≥ 108 s−1 (𝜏 isc ≈ 1 ns) without the metal atom and K isc ≥ 1011 s−1 (𝜏 isc = 1 ps) with Ir atoms involved. It is interesting to note that the rate of ISC to form triplet excitons is as fast as the rate of formation of singlet excitons. If the organic solids do not have any heavy metal atoms, then the rate of formation of triplet excitons via direct transitions is three orders of magnitude lower than that via the ISC. It may however be noted that there is some energy loss in exciting triplet excitons through the ISC, which is lost as thermal vibrational energy and may heat a device.

5.4 Excitons in Organic Semiconductors

5.4.2

Exciton Up-Conversion

In the preceding text, we have described the theory of ISC, where an exciton in singlet state converts into a triplet state through the exciton–spin–orbit interaction, which flips the spin and vibrational energy overlap that moves the exciton from singlet to triplet state. The reverse process of converting a triplet exciton into a singlet exciton is also possible and is applicable in the operation of OLEDs, where electrons and holes are injected from the opposite electrodes (see Chapter 7). As shown in Figure 5.5, there is statistically only one possible spin configuration for forming singlet excitons, but there are three spin configurations possible for forming triplet excitons. However, the radiative recombination of triplet excitons is not spin allowed unless one has strong spin–orbit interaction, which is usually not the case in organic solids. In such a situation, only the singlet excitons emit light, and that limits the internal quantum efficiency of OLEDs to 25%, because triplets cannot recombine and hence cannot emit radiation. If the triplets could easily be converted into singlets, then the internal quantum efficiency of 100% can be achieved. This is the principle of operation of OLEDs based on thermally activated delayed fluorescence (TADF) [47]. In organic solids, it is relatively easy to substitute another molecule (a donor) whose singlet energy may be close to the triplet state of the host solid (acceptor) and also to introduce heavy metal atoms to enhance the exciton–spin–orbit interaction. If the energy difference is very small, the up-conversion can occur even at room temperature. The thermal energy will be adequate to raise the vibrational energy of the triplet exciton to overlap with the singlet state of the donor to enable the transfer by flipping the spin through the exciton–spin–orbit interaction, as schematically shown in Figure 5.7. The rate of up-conversion also known as the reverse intersystem crossing (K RISC has recently been derived [48], and it should be the same as the rate of ISC (Eq. (5.33))), multiplied by the Boltzmann factor as: ( ) ΔEST , (5.34) KRISC = KISC exp − kb T where k b is the Boltzmann constant and T is temperature. For an efficient reverse ISC, the radiative lifetime of singlet state must be shorter than that of the triplet state, and the energy gap between singlet and triplet excited states must be smaller than 100 meV [47]. An example of thermally activated up-conversion material is 2-biphenyl-4,6-bis (12-phenylindolo [2,3-a]carbazole-11-yl)-1,3,5-triazine (PIC-TRZ), containing an indolocarbazole donor unit and a triazine acceptor unit [47]. Figure 5.7 Schematic illustration of thermally activated up-conversion or reverse intersystem crossing from a triplet exciton to singlet exciton state with energy difference ΔE ST . S0 , S1 , and T 1 are the ground, first excited singlet, and first excited triplet exciton states of an organic molecule.

∆EST S1 T1

S0

147

148

5 Concept of Excitons

Following the up-conversion to a singlet state, fluorescence occurs due to the radiative recombination of thus up-converted singlet excitons, known as the TADF. Using TADF, in an OLED, the generated triplet excitons, which cannot contribute to electroluminescence due to unfavorable spin configurations, can be made to recombine radiatively, and thus the internal quantum efficiency can be enhanced through the up-conversion. The synthesized molecule of PIC-TRZ exhibits a very small ΔEST , providing both efficient up-conversion from T1 to S1 levels and intense fluorescence that leads high electroluminescence (EL) efficiency. Both the processes of intersystem and reverse ISCs are nonradiative. The fluorescence- or radiative-recombination-related processes are described in more detail in Chapter 6. 5.4.3

Exciton Dissociation

As described in the preceding text, photo-excitations in organic semiconductors form excitons instantly, and excitons are the excited pairs of e and h bound in hydrogen-like electronic states. These are not free excited pairs of charge carriers, and, therefore, the formation of excitons in devices such as OSCs is not good for their operation, because one needs to dissociate excitons into free electron (e) and hole (h) excited pairs that can be transported to the opposite electrodes of OSCs to produce photocurrents. On the contrary, formation of excitons is good for OLEDs, because it enables radiative recombination and emission of light. In this section, the dissociation of excitons in the bulk heterojunction (BHJ) OSCs is described in detail. In the development of OSCs, first a single layer of organic semiconductor sandwiched between anode and cathode was prepared. In such a structure, there is no force other than the electrical force available due to the difference in work functions of the two electrodes to dissociate the excitons excited within the organic layer. The built-in electric field due to the difference in the work functions of the electrodes is given by: 𝜙c − 𝜙a , (5.35) r where 𝜙c and 𝜙a are the work functions of the cathode and anode, respectively, and r is the separation between them, which is about the total thickness of an OSC. However, as the exciton is an electrically neutral entity, the built-in electric field is expected to have little contribution in the exciton dissociation. This built-in electric field can only contribute in charge separation and collection after the excitons have dissociated. As a result, the single-layered OSCs have very poor power conversion efficiency (PCE). Then, in 1986, a bilayer concept was invented by Tang [49] with one layer of a donor material, which can easily give away its electrons, and the second layer of an acceptor material, which can easily accept electrons. The two layers have an interface between them and sandwiched between two electrodes. Many combinations of donor and acceptor materials have been invented and tried for OSCs. The important point in selecting the donor and acceptor materials is that the LUMO and HOMO of a donor molecule are at higher energies than those of an acceptor molecule, as shown in Figure 5.8. It is generally accepted that the generation of photo charge carriers in a bilayer OSC occurs through the following five processes in sequence [38, 50]: (i) photon absorption from the sun in the donor and/or acceptor excites electron–hole pairs that instantly form neutral Frenkel excitons; (ii) diffusion of the excited excitons to the donor–acceptor F=

5.4 Excitons in Organic Semiconductors

∆ELUMO

LUMO

LUMO

CT exciton (1) CT exciton (2) HOMO ∆EHOMO Donor

HOMO Acceptor

Figure 5.8 Schematic illustration of the formation of CT excitons at the D-A interface in a BHJ OSC. CT exciton (1) is formed when Frenkel exciton is excited in the donor, and CT exciton (2) is formed when Frenkel exciton is excited in the acceptor.

(D-A) interface; (iii) formation of charge transfer (CT) excitons at the D-A interface by transferring the electron to the acceptor, from excitons excited in the donor and/or by transferring the hole to the donor from excitons excited in the acceptor [51]; (iv) dissociation of the CT excitons at the D-A interface; and (v) transport and collection of the dissociated free charge carriers at their respective electrodes to generate photocurrent, which is the main purpose of any solar cell. In a bilayer structure, excitons are required to diffuse to the D-A interface to form CT excitons, leading to the subsequent dissociation. How exactly the dissociation of CT excitons takes place will be discussed later in text, but it may be noted that, an exciton being electrically neutral, it cannot be directed to move in any particular direction by any external or built-in electric field. Therefore, an exciton can only diffuse from one point to another in a random motion, and when it reaches the D-A interface, it will form a CT exciton. This requires that the exciton diffusion length LD be larger than the thickness of the donor or acceptor layers. In organic semiconductors, LD is short (∼10 nm), and hence the bilayer-structured OSCs also have very poor PCE. One way forward is to make the active organic layer from a blend of the donor and acceptor materials, and the structure thus obtained is called BHJ OSC. BHJ OSCs are one of the most promising alternative photovoltaic technologies due to the advantages of high absorption coefficient, light weight, flexibility, and the potential of low-cost solution process capability, etc. In a BHJ OSC also, the photogeneration of charge carriers occurs through the above five processes, but excitons do not have to diffuse to a D-A interface at a fixed distance. The BHJ OSCs have reached PCE more than 10% [43–46], with an expectation of achieving 15% in the very near future [45]. The processes 1–3 and 5, listed earlier for the operation of BHJ OSCs, have been quite well investigated and understood. The heterojunction structure is the only possible way of letting excitons diffuse and change into CT excitons at a nearby D-A interface. However, the process (iv) of the dissociation of CT excitons at the D-A interface assisted only by the built-in electric field can be very inefficient [52], as discussed in the following text, and hence the mechanism of dissociation has not been fully understood. Therefore, for an efficient dissociation of CT excitons into free electron and hole pairs, one must consider other possibilities. As both the donor and acceptor materials are organic semiconductors with similar dielectric constants, the formation of CT excitons from Frenkel excitons, excited either

149

150

5 Concept of Excitons

in the donor or acceptor material, neither makes the CT excitons loosely bound, nor does the electron and hole in a CT exciton become farther separated. This is because excitons excited in the donor molecules form CT excitons: (i) by electron transfer from the LUMO of the donor to the LUMO of the neighbor acceptor molecules, having a lower energy by ΔELUMO , across the D-A interface, and those excited in the acceptor molecules form CT excitons; and (ii) by hole transfer from the HOMO of the acceptor to the HOMO of the donor molecules, having a lower energy by ΔEHOMO , as clearly illustrated in Figure 5.8. For an efficient formation of CT excitons, it is required that both HOMO and LUMO levels of the donor be at higher energies than those of the acceptor. As the CT excitons are created by transferring the electrons and holes to lower energy states, it may be expected that the CT exciton states are even more stable than their predecessor Frenkel excitons, and hence cannot be dissociated in any way easier than the corresponding Frenkel excitons. The binding energy of singlet excitons is about 0.06 eV and that of a triplet is 0.7 eV in most organic solids. Therefore, the binding energy of the corresponding CT excitons is expected to be at least the same, if not larger. As the formation of a CT exciton involves two molecules—electron excited on an acceptor molecule and hole on a donor molecule—it is speculated that the electron and hole become farther apart in a CT exciton and hence can easily be dissociated due to the built-in electric field generated by the difference in the electrode work functions. A study on the dissociation of CT excitons by Devizis et al. [50] reveals that only charge pairs with an effective electron–hole separation distance of less than 4 nm are created during the dissociation of Frenkel excitons, which is about the same as the excitonic Bohr radius of singlet Frenkel excitons in organic solids [30]. Therefore, an exciton with a separation of 4 nm between their charge carriers is not dissociated yet, but it may dissociate after the formation of a CT exciton. As already described earlier, it is also to be noted that the built-in electric field due to the difference in work functions of electrodes cannot act efficiently on an electrically neutral particle, such as a CT exciton, and hence it cannot dissociate it. Therefore, the cause of dissociation of a CT exciton is puzzling and needs further investigation. In view of this, although the formation of a CT exciton is a prerequisite intermediate state for its dissociation, as it has been identified earlier [31, 53, 54], its dissociation requires some excess energy to overcome its binding, which is about the same as that of an exciton binding energy as explained earlier. It is usually assumed that, if the built-in electric field given in Eq. (5.35) is adequate to offer an energy larger than or equal to the binding energy of a CT exciton, it may be expected to dissociate. However, if CT excitons can be dissociated by this built-in electric field, then Frenkel excitons can also be dissociated, whether excited in the donor or acceptor, because, as discussed earlier, they have similar or even lower binding energies. However, that would mean that the formation of CT excitons at the D-A interface in BHJ OSCs plays no role, which is contrary to the observed higher efficiency in BHJ OSCs, and hence cannot be accepted. It has also been suggested that a CT exciton will subsequently dissociate following Onsager’s theory [55]. However, a CT exciton is a discrete quantum state holding a quantum of solar energy; it is not exactly a geminate pair of varying distance between the two charge carriers (e and h). In this situation, whether Onsager’s theory can be applied for the dissociation of an exciton becomes, on its own, a topic of debate and controversy. Although, there is a significant amount of work done in studying the morphology of the BHJ OSCs for efficient charge generation [52], not much has been discussed on the

5.4 Excitons in Organic Semiconductors

mechanism of dissociation of CT excitons. We have recently proposed [38, 56] that a CT exciton can be dissociated only if it is given an excess energy equivalent to or greater than its binding energy from an external source. The only possible source of such energy available in OSCs is due to the formation of CT excitons, which releases excess energy from the conversion of a Frenkel exciton (at higher energy) to a CT exciton (lower energy), as it has been modeled recently [28]. Thus, the dissociation of an exciton at the D-A interface can be regarded as a two-step process as described in the following text. 5.4.3.1

Conversion from Frenkel to CT Excitons

As an exciton is electrically neutral, it cannot be directed to move in any particular direction to reach the D-A interface. So while it is excited within a material sandwiched between an anode and cathode, it can diffuse in any direction randomly. To direct an exciton to move in any particular direction, one has to introduce an interface with an acceptor (another organic material with a lower LUMO energy) in that direction. In this way, a cascade kind of structure of LUMO and HOMO of donor and acceptor with their interfaces parallel to the electrodes can direct an exciton to move to the nearest D-A interface by forming a CT exciton, and then to the next D-A interface, and thus forming and remaining in a CT exciton state at each interface, as shown in Figure 5.9 for a cascade of three materials. In this case, in a blended structure of ternary materials, the excitons can be excited in each of the three materials. If the exciton is excited in the first donor, then the CT exciton can be formed at the first interface by transferring the electron to the LUMO of the first acceptor and thus releas1 . This CT exciton can then change to another CT exciton such ing an energy ΔELUMO that the electron is transferred to the second acceptor material by releasing an energy, 2 ΔELUMO , but the hole is still in the first donor material. In case the exciton was excited in the first acceptor material (middle donor/acceptor), then it has the possibility of forming

LUMO ∆E 1LUMO ∆E 2LUMO

LUMO

∆E 1HOMO HOMO

∆E 2HOMO HOMO

1 Figure 5.9 Schematic illustration of the formation of CT excitons in a ternary BHJ OSC, where ΔELUMO 2 1 2 and ΔELUMO represent the LUMO energy off-sets, and ΔEHOMO and ΔEHOMO represent the HOMO energy offsets at interfaces 1 and 2, respectively, counted from left to right.

151

152

5 Concept of Excitons

a CT exciton in two ways: (i) electron goes down to the third acceptor’s LUMO by releas2 ; and/or (ii) the hole goes up to the HOMO of the first donor by ing an energy ΔELUMO 1 releasing an energy ΔEHOMO . Finally, if the exciton is excited in the second acceptor (third material), it has only one way of forming a CT exciton at the second D-A interface by transferring its hole to the 2 , which may acceptor of HOMO of the middle material and releasing an energy ΔEHOMO then move to the first interface and form a CT exciton by transferring itself to the first 1 . donor and releasing an energy ΔEHOMO 5.4.3.2

Dissociation of CT Excitons

The dissociation of a CT exciton can only occur due to the mechanism of its formation, as discussed in our earlier work [28, 57]. Accordingly, as the formation of a CT exciton involves release of energy due to either transfer of an electron to a lower energy state of LUMO of the acceptor at a D-A interface or transfer of a hole to a higher HOMO level of the donor. If this released energy, usually in the form of molecular vibrational energy, impacts back on the CT exciton, it may dissociate, provided this energy is larger or at least equal to the binding energy of CT excitons. Thus, the formation of a CT exciton also involves its dissociation if the energy offsets are greater than or equal to the binding energy of the CT exciton. This may be the reason for assuming that the formation of a CT exciton leads to automatic dissociation of excitons, and hence the formation of CT state is a prerequisite state for the dissociation [50, 51]. Accordingly, the condition of dissociation of CT excitons can be given by [28]: ΔELUMO or ΔELUMO ≥ Eb , D ELUMO

(5.36)

A ELUMO

where ΔELUMO = − is the difference between the energy of donor LUMO, D A D A ELUMO , and that of acceptor LUMO, ELUMO ; ΔEHOMO = EHOMO − EHOMO is the differD A ence between the energy of donor HOMO, EHOMO , and that of acceptor HOMO, EHOMO ; and Eb is the binding energy of excitons (Figure 5.4). Based on the condition in Eq. (5.36) and using the newly derived exciton–molecular vibration interaction operator as a perturbation, the rate of dissociation of CT excitons is obtained as [28]: i,j

RD =

1 96𝜋 2 ℏ3 ε2 Ebi

[(EjD − EjA ) − Ebi ]2 (ℏ𝜔v )𝜇x a2xi

(5.37)

where i denotes parameters associated with S (singlet) and T (triplet) excitons, and j denotes LUMO when CT excitons are formed from excitons excited in the donor and HOMO when CT excitons are formed from excitons excited in the acceptor material. The rate given in Eq. (5.37) is applicable to the dissociation of both singlet and triplet excitons, but one has to use the corresponding parameters. In deriving the dissociation rate in Eq. (5.37), it is assumed that a singlet Frenkel exciton changes to a singlet CT exciton, and a triplet Frenkel exciton changes to a triplet CT exciton at the D-A interface. Accordingly, the exciton dissociation at a D-A interface may be regarded as a two-step process: (i) the formation of CT excitons at the interface; and (ii) if the condition in Eq. (5.36) is met, they may dissociate efficiently. Although the CT exciton formation at the D-A interface is a pre-requisite for its dissociation, it is not yet dissociated. As stated earlier, a CT exciton is no different in the binding energy than a Frenkel exciton, and,

5.5 Conclusions

similar to a Frenkel exciton, it is also an electrically neutral entity and hence cannot be influenced by the built-in electric field caused by the electrodes. Therefore, for an efficient dissociation of a CT exciton at the D-A interface, it is important that the condition in Eq. (5.36) be met. This is also supported by the experimental result by He et al. [53], who fabricated a BHJ OSC using PTB7 as donor and PC71 BM as acceptor, and reported a PCE of 9.2% for an inverted structure. We will not discuss here the effect of inverted structure on the PCE, but it is to be noted that the LUMO energy of PTB7 as donor is at −3.31 eV and that of PC71 BM as acceptor is at −4.3 eV, giving ΔELUMO = 0.99 eV, which is much larger than the binding energy of both singlet and triplet CT excitons, and hence satisfies the condition in Eq. (5.36) very well for dissociating excitons excited in the donor material. Likewise, for the dissociation of excitons excited in the acceptor PC71 BM with HOMO energy at −5.15 eV and HOMO of donor PTB7 at −6.1 eV, ΔEHOMO = 0.95 eV, which is also more than the binding energy of both singlet and triplet excitons, and satisfies the condition in Eq. (5.36). The dissociation rates calculated from Eq. (5.37) for excitons excited in the donor are 1.89 × 1014 s−1 for singlet and 4.82 × 109 s−1 for triplet excitons. The rates of dissociation for excitons excited in the acceptor are 1.64 × 1014 s−1 for singlet excitons and 3.16 × 109 s−1 for triplet excitons. These high rates imply faster dissociation of singlet than triplet excitons in PTB7: PC71 BM BHJ OSCs, leading to faster free charge carrier generation, resulting in enhanced photocurrent and hence higher power conversion efficiency. The concept of having ΔEji ≥ Ebi , I = S or T, and j = LUMO or HOMO for an efficient dissociation of CT excitons is very important, although it has not yet been fully realized.

5.5 Conclusions In this chapter, concepts of excitons in crystalline and amorphous solids, both inorganic and organic, are presented. Excitonic absorption in crystalline solids is reviewed. It is shown that, in amorphous solids, the excitonic absorption spectrum is similar to the band-to-band absorption spectrum. For example, the excitonic absorption in amorphous semiconductors also satisfies Tauc’s relation, as does band-to-band absorption. This is because: (i) the transition matrix element remains the same for excitonic absorption and for band-to-band free-carrier absorption, and (ii) the concept of the joint density of states applied to excitonic absorptions in crystalline solids is not applicable to excitonic absorptions in amorphous solids. In organic semiconductors, mainly Frenkel excitons are created, and the absorption of photons creating Frenkel excitons, both singlet and triplet, is described in detail. Two processes of exciting triplet excitons are described: (i) direct excitation to the triplet through the exciton–spin–orbit–photon interaction; and (ii) indirect excitation through the ISC caused by the exciton–spin–orbit–molecular vibration interaction. Likewise, two processes for the excitation of singlet excitons is described in organic solids: (i) direct excitation through the exciton–photon interaction; and (ii) indirect excitation through the up-conversion or reverse ISC. The process of dissociation of excitons in organic semiconductors at the donor– acceptor interface of BHJ OSCs is also described.

153

154

5 Concept of Excitons

References 1 Singh, J. (1994). Excitation Energy Transfer Processes in Condensed Matter. New

York: Plenum. 2 Singh, J., Birkedal, D., Lyssenko, V.G., and Hvam, J.M. (1996). Phys. Rev. B 53: 15909,

and references therein. 3 Birkedal, D., Singh, J., Lyssenko, V.G. et al. (1996). Phys. Rev. Lett. 76: 672. 4 Miller, R., Kleinman, D., Gossard, A., and Monteanu, O. (1982). Phys. Rev. B 25:

6545. 5 Singh, J. and Shimakawa, K. (2003). Advances in Amorphous Semiconductors. London 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38

and New York: Taylor & Francis. Singh, J. and Oh, I.-K. (2005). J. Appl. Phys. 95: 063516. Singh, J. (1997). Nonlinear Optics, vol. 18, 171. Singh, J., Aoki, T., and Shimakawa, K. (2002). Philos. Mag. B 82: 855. Singh, J. (2002). J. Non-Cryst. Solids 444: 299–302. Singh, J. (2002). Nonlinear Opt. 29: 111. Singh, J. (2003). J. Mater. Sci. 14: 171. Elliot, R.J. (1962). Polarons and Excitons (eds. K.G. Kuper and G.D. Whitfield), 269. Edinburgh and London: Oliver & Boyd. Aoki, T., Koomedoori, S., Kobayashi, S. et al. (2002). Nonlinear Opt. 29: 273. Davydov, A.S. (1971). Theory of Molecular Excitons. New York: (Plenum. Brutting, ̀ W. (2005). Physics of Organic Semiconductors. Wiley-VCH. Swist, A. and Soloducho, J. (2012). Chemik 66: 289. Narayan, M.R. and Singh, J. (2012). Phys. Status Solidi C 9: 2386. Roncali, J. (2009). Acc. Chem. Res. 42: 1719. Kippelen, B. and Brédas, J.L. (2009). Energy Environ. Sci. 2: 251. Brédas, J.L., Norton, J.E., Cornil, J., and Coropceanu, V. (2009). Acc. Chem. Res. 42: 1691. Ompong, D. and Singh, J. (2016). Phys. Status Solidi C 13: 89. Narayan, M.R. and Singh, J. (2013). J. Appl. Phys. 114: 154515. Singh, J. and Williams, R.T. (eds.) (2015). Excitonic and Photonic Processes in Materials. Singapore: Springer, Ch. 8. Singh, J. (2007). Phys. Rev. B 76: 085205. Singh, J. (2011). Phys. Status Solidi A 208: 1809. Schulz, G.L. and Holdcroft, S. (2008). Chem. Mater. 20: 5351. Yang, C.-M., Wu, C.-H., Liao, H.-H. et al. (2007). Appl. Phys. Lett. 90: 133509. Singh, J. (2010). Phys Status Solidi C 7: 984. Shafiee, A., Salleh, M.M., and Yahaya, M. (2011). Sains Malays. 40: 173. Singh, J., Baessler, H., and Kugler, S. (2008). J. Chem. Phys. 129: 041103. Li, Y., Cao, Y., Gao, J. et al. (1999). Synth. Met. 99: 243. Da Costa, P.G. and Conwell, E. (1993). Phys. Rev. B 48: 1993. Thompson, B.C. and Fréchet, J.M. (2007). Angew. Chem. Int. Ed. 47: 58. Winder, C. and Sariciftci, N.S. (2004). J. Mater. Chem. 14: 1077. Bittner, E.R., Ramon, J.G.S., and Karabunarliev, S. (2005). J. Chem. Phys. 122: 214719. Xu, Z., Hu, B., and Howe, J. (2008). J. Appl. Phys. 103: 043909. Huang, J., Yu, J., Guan, Z., and Jiang, Y. (2010). Appl. Phys. Lett. 97: 143301. Narayan, M.R. and Singh, J. (2013). J. Appl. Phys. 114: 73510.

References

39 Luhman, W.A. and Holmes, R.J. (2009). Appl. Phys. Lett. 94: 153304. 40 Gautam, B.R. (2013). Magnetic Field Effect in Organic Films and Devices. Ann Arbor:

The University of Utah. 41 Medina, A., Claessens, C.G., Rahman, G.M.A. et al. (2008). Chem. Commun.:

1759–1761. 42 Ford, T.A., Avilov, I., Beljonne, D., and Greenham, N.C. (2005). Phys. Rev. B 71:

125212. 43 Cogan, S., Haas, Y., and Zilberg, S. (2007). J. Photochem. Photobiol., A 190: 200. 44 Donald, L.P., Gary, M.L., George, S.K., and Engel, R.G. (2011). A Small Scale

Approach to Organic Laboratory Techniques, 415. Belmont, CA: Cengage Learning. 45 Klessinger, M. and Michl, J. (1995). Excited States and Photochemistry of Organic

Molecules, 255. New York: VCH. 46 Guo, F., Kim, Y.G., Reynolds, J.R., and Schanze, K.S. (2006). Chem. Commun.:

1887–1889. Endo, A., Sato, K., Yoshimura, K. et al. (2011). Appl. Phys. Lett. 98: 093302. Usman, S. and Singh, J. (2018). Org. Electron. 59: 121–124. Tang, C.W. (1986). Appl. Phys. Lett. 48: 183. Devizis, A., Jonghe-Risse, J.D., Hang, R. et al. (2015). J. Am. Chem. Soc. 137: 8192. Wright, M. and Uddin, A. (2012). Sol. Energy Mater. Sol. Cells 107: 87. Peumas, P., Yakimov, A., and Forest, S.R. (2003). J. Appl. Phys. 93: 3393. He, Z., Zhong, C., Su, S. et al. (2012). Nat. Photonics 6: 591. Lu, L., Zheng, T., Wu, Q. et al. (2015). Chem. Rev. 115: 12666. R.F. Service (2011). Science 332 (6027): 293. Singh, J., Narayan, M.R., Ompomg, D., and Zhu, F. (2017). J. Mater. Sci. - Mater. Electron. 28: 7095. 57 Narayan, M.R. and Singh, J. (2017). J. Mater. Sci. - Mater. Electron. 28: 7070–7076. 47 48 49 50 51 52 53 54 55 56

155

157

6 Photoluminescence Takeshi Aoki Joint Research Center of High-technology, Department of Electronics and Information Technology, Tokyo Polytechnic University, Atsugi 243-0297, Japan

CHAPTER MENU Introduction, 157 Fundamental Aspects of Photoluminescence (PL) in Materials, 158 Experimental Aspects, 164 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique, 175 QFRS on Up-Conversion Photoluminescence (UCPL) of RE-Doped Materials, 192 Conclusions, 197 References, 198

6.1 Introduction A radiative emission process in condensed matter is called luminescence, and such a luminescent material is sometimes called phosphors. Luminescence can occur through a variety of electronic processes unrelated to heat, among which photoluminescence (PL) and electroluminescence (EL) are most popularly used. The radiative process requires a non-equilibrium carrier concentration in the electronic band of solids or in the electronic state of an impurity or defect. If the non-equilibrium state (excited state) is created by photoexcitation (PE), the resulting luminescence is called PL, and if it is obtained by carrier injection through an electric field, it is called EL [1–6]. The study of luminescence from condensed matter is not only of scientific but also technological interest, because it forms the basis of solid-state lasers. Also, it is important for display panels in electronic equipment; lighting, such as in fluorescent lamps and phosphor-converted (white) light-emitting diodes (pcWLEDs); fluorescent paints; bioimaging; and photodynamic therapy (PDT). Undoubtedly, PL is frequently used as a non-destructive technique for material characterization or research in material science as well. PL spectroscopy is a sensitive tool for investigating both intrinsic electronic transitions between energy bands and extrinsic electronic transitions at impurities and defects of organic molecules, semiconductors, and insulators [6]. As a comprehensive review of PL spectroscopy used on all the condensed matter systems is beyond the scope of this chapter, it mainly focuses on a limited number of Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

158

6 Photoluminescence

topics related to PL recombination in disordered materials. In particular, the quadrature frequency resolved spectroscopy (QFRS) of PL is not popular, despite it being a powerful tool for PL lifetime analysis. Therefore, the QFRS of PL of amorphous semiconductors and up-conversion photoluminescence (UCPL) of rare-earth (RE) ion–doped glasses are addressed specifically in this chapter, and readers interested in other topics or further details should refer to the relevant references herein. Section 6.2 provides a brief outline of the fundamental aspects of PL in condensed matter. Section 6.3 focuses on the experimental aspects of PL, which ends with the fundamentals of the QFRS used for PL lifetime analyses. Section 6.4 deals with the QFRS of PL of amorphous semiconductors, in particular, hydrogenated amorphous Si (a-Si:H), including a-Ge:H. And, finally, Section 6.5 presents recent developments made by QFRS in studying the UCPL of RE-doped chalcogenides glass (ChGs).

6.2 Fundamental Aspects of Photoluminescence (PL) in Materials Conceptually, PL is the inverse of the absorption process of photons, where photons absorbed into a material transfer their energy to electrons in the ground state and energize them to excited states. Applying the old quantum theory and statistical mechanics to the two-energy-level model, Einstein predicted the occurrence of two radiative emission processes: stimulated photon emission and spontaneous photon emission after the photon absorption [5, 7, 8]. It is noteworthy that the former process is induced by an incoming photon and is the basis of lasing, whereas the latter occurs without any photon stimulation. In a strict sense, therefore, the photon absorption process is the stimulated photon absorption. If there are multiple number of excited states, then electrons excited to higher excited states rapidly relax non-radiatively to the lowest excited state S1 by thermalization, or by emitting phonons, and subsequently relax radiatively to the ground state S0 by emitting photons, as shown in Figure 6.1. Therefore, the absorbed photons are usually of higher energy than the emitted photons by the energy of thermalization or emitted phonons. The energy difference between the absorbed and emitted photons is called Stokes’ shift. PL includes fluorescence and phosphorescence, which are preferably used in the field of organic chemistry as well as biochemistry. Fluorescence occurs when a radiative transition from the excited state to the ground state is spin-allowed, that is, both the ground and excited states have the same spin multiplicity (singlet–singlet S1 –S0 or sometimes triplet–triplet states), and phosphorescence occurs when the transition is spin-forbidden, that is, the ground and excited states have different spin multiplicity (triplet–singlet T1 –S0 states). The spin-forbidden transitions can only occur through the spin–orbit coupling, which is usually weak, and hence such a radiative recombination has longer lifetime. As shown in Figure 6.1 for organic molecules, fluorescence and phosphorescence arise from radiative recombination of singlet and triplet Frenkel excitons, respectively [5, 9–11]. The fluorescence lifetime is usually much shorter (0.1 ∼ 10 ns) than the phosphorescence lifetime (1 ms ∼ 10 s) due to the spin selection rule. However, a delayed fluorescence possesses exceptionally longer lifetime similar to the phosphorescence lifetime, examples of which are semiconductor nanocrystals (NCs) such as CdSe, Cu+ :CdSe, and

6.2 Fundamental Aspects of Photoluminescence (PL) in Materials

S1

Intersystem Crossing Exchange Energy

EX

T1

Fluorescence ћω Phosphorescence ћω

S0

Figure 6.1 Illustration of radiative recombination processes in organic materials. S0 and S1 denote the ground and first excited singlet states, respectively. T1 denotes the first excited triplet state. E X is photoexcitation energy, and ℏ𝜔, fluorescence or phosphorescence emission energy. Here, competing nonradiative recombination processes are omitted.

CuInS2 [12]. However, it should be noted that the distinction between fluorescence and phosphorescence is not always clear [5, 13]. Sometimes, delayed fluorescence is also called phosphorescence. Metal–ligand complexes (MLCs), which contain a transition metal atom and one or more organic ligands, possess rather large intersystem crossing coefficients, owing to the large spin–orbit coupling caused by the heavy metal atoms, and therefore exhibit strong phosphorescence due to mixed singlet-triplet states (Figure 6.1). Accordingly, these MLCs display intermediate lifetimes of 10 ns ∼ 10 μs [13] and, in particular, fac-tris (2-phenylpyridine) iridium [Ir(ppy)3 ] demonstrates a very high internal quantum efficiency (QE) due to the emission from threefold degenerate triplet excitons [4, 14, 15], which work as organometallic triplet emitters in phosphorescent organic light-emitting diodes (OLEDs) with very high QE in green emission [16] (for theory, see Chapter 7). However, achieving the same level of QE in blue phosphorescent emitter still seems to be challenging [17]. 6.2.1

Intrinsic Photoluminescence

In inorganic semiconductors, PL is classified into two categories: intrinsic PL and extrinsic PL [1, 2, 6, 18, 19]. The intrinsic PL occurs mainly due to the band-to-band radiative transitions in a highly pure semiconductor even at a relatively high temperature, where, by absorbing a photon of energy higher than the band-gap energy EG , an electron is excited to the conduction band (CB), leaving a hole behind in the valence band (VB), and then they radiatively recombine to give rise to intrinsic PL, as shown in Figure 6.2a. The band-to-band transition that occurs in indirect-gap semiconductors such as Si and Ge is called an indirect transition, in contrast to the direct transition in direct-gap semiconductors such as GaAs. The latter transition does not need phonon-assistance, where the radiative recombination probability (rate) is much larger (high QE), and such materials are extensively applied to LEDs and semiconductor laser diodes. At low temperatures, an excited electron–hole (e–h) pair can form a free exciton (FE), or Wannier exciton through their Coulomb attractive force, as shown in Figure 6.2d, and then the PL occurs through the excitonic recombination in place of the band-to-band transition in the pure semiconductors. Beside the PL from FE, recombination of various exciton types—such as bound exciton (BE), excitonic polaron, self-trapped

159

160

6 Photoluminescence

Conduction band (CB) D0 (or D+)

D0 EG

D0

A0 r

A0 (a)

(b)

(c)

Deep level

(d)

(e)

(f)

➀ ➁ (g)

Valence band (VB)

Figure 6.2 Schema of radiative recombination processes in semiconductors. (a) band-to-band e–h recombination; (b) neutral donor (D0 ) to valence band (VB) transition; (c) conduction band (CB) to neutral acceptor (A0 ) transition; (d) radiative recombination of free exciton (FE); (e) radiative recombination of bound exciton (BE), which is bound to D0 —recombination of BE bound to ionized donor D+ is also possible; (f ) donor–acceptor pair (DAP) recombination with separation r; (g) 1 and  2 is radiative. deep-level defect luminescence; either one of two transitions 

exciton (STE), and excitonic molecule—take place at low temperatures. Figure 6.2e shows an example of the radiative recombination of BE bound to neutral donor (D0 ) or ionized donor (D+ ). Excitons play even more important roles in semiconductor quantum wells, quantum wires, and quantum dots (QDs), where the excitonic PL can be observed even at much higher temperatures because the quantum confinement increases exciton-binding energy [4, 5, 20–22] (see Chapter 5). 6.2.2

Extrinsic Photoluminescence

An extrinsic PL is caused mostly by metallic impurities or defects (activators) intentionally incorporated into materials such as ionic crystals and semiconductors, and materials made luminescent in this way are called phosphors. The extrinsic PL occurs either between free carriers and impurity or defect states, or within the impurity or defect states (localized luminescence centers). The most important impurities are donors (Ds) and/or acceptors (As) in ionic crystals and semiconductors. In extrinsic semiconductors, where Ds or As are intentionally incorporated, PL occurs between an electron trapped by a neutral donor D0 and a hole at the top of VB (Figure 6.2b) or between an electron at the bottom of CB and a hole trapped by a neutral acceptor A0 (Figure 6.2c). In compensated semiconductors, where both Ds and As are ionized, or charged as D+ with the ionization energy ED and A− with the ionization energy EA , donor–acceptor pair (DAP) recombination occurs, as shown in Figure 6.2f. It is called distant DAP recombination, when the separation r between a D+ ion and an A− ion is much greater than the larger internal dimension, that is, the effective Bohr radius of D0 or A0 . In the distant DAP recombination, an excited free electron in CB gets bound on a D+ , accompanied by D+ → D0 , and a free excited hole in VB gets bound to an A− , accompanied by A− → A0 after band-to-band excitations. Thus, both the D and A become neutral D0 and A0 , respectively, and then the radiative recombination between the e–h pair bound to the DAP occurs, leaving the DAP charged as D+ and A− (Figure 6.2f ): D0 + A0 → D+ + A− + ℏ𝜔, where ℏ𝜔 is the emitted photon energy, but a potential energy

6.2 Fundamental Aspects of Photoluminescence (PL) in Materials

−e2 /𝜅r due to the Coulomb attractive force between the ionized DAP (D+ and A− ) remains, where 𝜅 is the static dielectric constant of host material [1, 4, 5, 18]. Therefore, ℏ𝜔 should be higher than the energy difference between D and A levels by e2 /𝜅r, and is given by: ℏ𝜔 = EG − ED − EA +

e2 kr

(6.1)

Since the D–A distance r can be multiples of the crystallographic lattice constant, the last term in Eq. (6.1) gives rise to a long series of sharp lines on the higher photon energy side of the PL spectrum. This was experimentally observed in the indirect gap semiconductor GaP doped with S (D) and Si (A) and first analyzed by Hopfield et al. [23]. The transition probability of a DAP recombination is proportional to the square of the overlap of hydrogen-like D and A wave functions. The hydrogen-like wave function is expressed as ∝ exp(−r/a), where a is the effective Bohr radius of a hydrogen-like impurity atom [1, 2, 4, 5]. If either of D0 or A0 has larger a, the square of the overlap proportional to the transition probability is approximately given by exp(−2r/a).Thus, the PL lifetime 𝜏 of DAP recombination depends on the DAP separation r as: ( ) 2r , (6.2) 𝜏 = 𝜏 0 exp a where 𝜏 0 is an electric–dipole transition time usually shorter than 10 ns. The distant DAP recombination shows the following features. First, according to Eq. (6.2), a decrease in D–A separation r increases the distant DAP recombination rate 1/𝜏, and thereby increases PL intensity. However, any decrease in r decreases the number of possible pairings of DAPs, leading to a reduction in PL intensity. Thus, the PL intensity should exhibit a maximum at a certain r. On the other hand, the PL emission energy ℏ𝜔 is a monotonically decreasing function of r (Eq. (6.1)), which is reflected in static PL spectroscopy (Section 6.3.1). It is observed in GaP crystal at 1.6 K that the PL spectrum is broad at r > 4 nm, with a maximum at r ≈ 5 nm and discrete at 1 < r < 4 nm, with a number of peaks corresponding to the D–A separations [24]. Since the recombination rate is slower for a DAP of larger r, PL intensity at lower ℏ𝜔 gets easily saturated and does not increase by increasing the excitation intensity similar to that observed at higher ℏ𝜔. Second, if we observe the temporal PL spectrum of the distant DAP recombination, the PL peak energy shifts to lower ℏ𝜔 as time elapses, since a DAP of smaller r which emits higher ℏ𝜔 will decay faster than that of a larger r which emits lower ℏ𝜔. This implies the importance of both PL lifetime spectroscopy and the static PL spectroscopy for a detailed study on the DAP recombination. The hydrogen-like nature of either D or A is essential to the distant DAP recombination; either D or A should be in a shallow impurity state: ED ≪ EG or EA ≪ EG . Therefore, the occurrence of this type of recombination is limited only to relatively low temperatures, which is not suitable for practical use except for the characterization of semiconductors [6]. In contrast, when both D and A are energetically deep, they are localized, and the PL of this type DAP recombination is strongly affected by phonons, and its mechanism is not entirely clear. Because of the deep impurity levels of D and A, however, the PL intensity

161

162

6 Photoluminescence

is sometimes enough intense at room temperature to serve as practical phosphors, such as blue-emitting ZnS:Ag and green-emitting ZnS:Cu, Cl [4, 5]. Generally, defects and metallic impurities intentionally incorporated in ionic crystals and semiconductors often act as efficient phosphors (luminescence centers) [1, 2, 4, 5]. Color- or F-centers are optically active vacancies in ionic crystals such as alkali halides. Paramagnetic metal ions incorporated into host materials act as localized activators; actually, these are transition metal ions such as Mn2+ , Cr2+ , etc., and the rare earth (RE), or lanthanide ions of Nd3+ , Eu3+ , Tb3+ , Er3+ , etc. For the transition metal ions, optical transitions occur between the unfilled d-d states, and, for the RE ions, these occur between the unfilled f -f states (see Chapters 2 and 4; also [4, 5]). Though both d-d and f-f transitions are forbidden by the selection rules for electric dipole transitions, their forbidden character is altered by the crystalline electric field, and then these transitions become more or less allowed. Since 3d orbitals of the transition metal ions are the valence orbitals, any environmental perturbation is influential to the d-d transitions; the energy (color) of PL is rather dependent on the host materials. On the contrary, RE ions are formed when the outermost 6s electrons are removed, leaving the optically active 4f orbitals inside the filled 5s and 5p shells. This makes the unfilled 4f orbitals smaller in radius and less sensitive to the crystalline field, so that the optical spectra of doped rare RE ions are generally similar to those of free ions having narrow emission lines. A good example of this type of luminescence is from phosphors of pcLEDs, for which RE-doped materials such as (Y,Gd)3 (Al,Ga)5 O12 :Ce3+ (YAG:Ce), (Ba,Sr,Ga)2 SiO2 :Eu2+ , etc., have been developed [4, 5]. YAG laser is also an example, where Nd3+ ion is incorporated into the host of yttrium aluminum garnet (Y3 Al5 O12 , or YAG) as an activator for the stimulated emission at 1064 nm [8]. Instead of phosphors, the term fluorophores is used for organic dyes, biological fluorescent proteins such as green fluorescent protein (GPF), quantum dots (QDs) of CdS, InP, PbS, etc., and nanoparticles (NPs) activated by RE ions. The fluorophores serve as tracers in fluid measurement and as sensors for metal ions, gas molecules, pesticide residue, etc., in environmental measurements [13, 25, 26]. More importantly, they are used as stains, markers, and probes in bioimaging and medically as photosensitizers of PDT and photo-thermal therapy (PTT) [13, 27, 28]. 6.2.3

Up-Conversion Photoluminescence (UCPL)

Since 4f orbitals of the RE ions possess ladder-like energy levels and some of them have similar energy difference, an anti-Stokes process of UCPL occurs with two or more sequential photon absorption processes followed by a photon emission at an energy higher than that of the absorbed photons in RE-doped crystals, glasses, proteins, etc. [4, 29, 30]. The typical UCPL from RE ions can be categorized into three mechanisms: (i) GSA/ESA: the excitation of a single ion by ground state absorption (GSA) (Figure 6.3a), and successively followed by excited state absorption (ESA) (Figure 6.3b) to excite the uppermost state; (ii) GSA/ETU: one of two neighboring ions excited by GSA relaxes to the lower state (not necessarily the ground state) by transferring its energy: energy transfer (ET) to another ion to excite the uppermost state by energy transfer up-conversion (ETU) (Figure 6.3c); and (iii) PA: photon avalanche via cross-relaxation process (CRP) (Figure 6.3d), where an ion excited to its uppermost state relaxes to

6.2 Fundamental Aspects of Photoluminescence (PL) in Materials

(a)

(b)

(c)

(d)

Figure 6.3 Basic mechanisms important for up-conversion of photoluminescence (UCPL): (a) ground state absorption (GSA), (b) excited state absorption (ESA), (c) energy transfer up-conversion (ETU), and (d) cross-relaxation process (CRP).

the intermediate state by simultaneously exciting the lower state of a proximate ion to the intermediate state via ET and thereby the intermediate state doubles by a single transition from the uppermost to the intermediate state, giving rise to PA. However, the condition for the occurrence of PA is rather special [4, 30, 31]. Thus, the GSA/ESA, GSA/ETU, and composite GSA/(ESA + ETU) processes are commonly accepted as UCPL mechanisms of RE ions. Since the GSA/ESA occurs with two or more sequential photons’ absorption at a single RE ion, it is almost insensitive to RE ion concentration. In contrast, the GSA/ETU process arises from ET between two neighboring RE ions and thus depends on the RE ion concentration [4, 29, 30]. The RE-doped nanoparticles (NPs) that can emit the visible UCPL, excited through “optical transparent window of biological tissue: 700–1100 nm” by near infrared (NIR) light, serves as a promising fluorophore for biochemistry, molecular biology, PTT, and PTD [27, 28, 32]. Another notable potential of RE-doped NPs is modification of the solar-spectrum by up-conversion and/or down-conversion to improve the efficiency of solar cells [33, 34]. In addition, up-conversion can also be assisted thermally as applied in OLEDs, which is described in Chapter 7. 6.2.4

Other Related Optical Transitions

Intense light irradiation to some condensed matters induces two phenomena similar to UCPL; one is two photon absorption (TPA or 2PA), and the other is second-harmonic generation (SHG). TPA occurs in materials such as atomic vapor, chromatic molecules, and semiconductors, with simultaneous absorption of the two photons having their sum energy equal to the energy difference between a ground state and a real excited state, where the electronic excitation follows via a virtual intermediate state in the absence of a resonant intermediate state [35]. In contrast, the SHG occurs from simultaneous absorption of two photons in nonlinear optical crystal materials such as LiNbO3 and KH2 PO4 , followed by the electronic excitation from a ground state to a virtual excited state via a virtual intermediate state [8]. However, both phenomena are low in the conversion efficiencies due to the involvement of virtual intermediate states. Raman scattering is the scattering of a light by phonons in condensed matters. When a laser light of narrow linewidth with the excitation energy Ex enters the condensed matter having the phonon energy ℏ𝜔ph , a Stokes shifted emission appears at the energy ℏ𝜔 = Ex − ℏ𝜔ph in the spectrum of the scattered light, and an anti-Stokes light appears at the energy ℏ𝜔 = Ex + ℏ𝜔ph . When Ex is close to the energy of a real excited

163

164

6 Photoluminescence

state, the Raman signals are enhanced, which is called resonant Raman scattering. Quantum-mechanically, in Raman scattering, a laser light excites electrons from the ground state to virtual excited states at Ex , which then relax radiatively to the ground states by emitting phonons (Stokes shift) or absorbing phonons (anti-Stokes shift). In resonant Raman scattering, the excited states are not virtual but real, like band-gap, d-d, or f-f electronic transition states. Since ℏ𝜔ph ≪ Ex , Raman spectroscopy needs narrow-linewidth excitation laser and high-resolution spectrometer. Occasionally in Raman spectroscopy, PL coexists with Raman signals, causing some confusion. This is solved by changing Ex in Raman spectroscopy and plotting the spectrum in an absolute wavelength [36]. Furthermore, mapping in conjunction with photoluminescence excitation (PLE) and PL spectroscopies can identify the difference between PL and Raman signals [6].

6.3 Experimental Aspects 6.3.1

Static PL Spectroscopy

PL contains information about emitted photon density (PL intensity), photon energy (wavelength), photon polarization, photon-emitting position, and photon lifetime. Except for emitted photon lifetime obtained from temporal PL measurements, the valuable information is available fundamentally from static PL measurements. The apparatus used for static PL spectroscopy [37, 38] is conventionally assembled as shown in Figure 6.4. The PL excitation source can be a laser beam, monochromatic light of a Xenon lamp, or Halogen lamp monochromatized by an additional (excitation) monochromator (MC) and LED light, and the PL excitation energy Ex is normally higher than the PL emission energy ℏ𝜔 due to Stokes shift. Here, it should be noted that the unit of photon energy is usually expressed in eV in physics and electronics, but frequently in wavenumber cm−1 in physics and chemistry; according to quantum mechanics, an NIR photon of wavelength 𝜆 = 1 μm in free space possesses the wavenumber 1/𝜆 = 104 cm−1 and the energy ℏ𝜔 = hc/𝜆 ≈ 1.24 eV with Planck’s constant h and light velocity c in free space. For a thin film on a substrate, the substrate surface should be roughened and the thickness d of the film should be as thick as possible, because the internal PL reflection causes interference effects on the PL spectrum. Interference at the film interfaces will occur at m𝜆 = 2nd with the refractive index n of the film and an interference multiple number m. If we plot PL spectrum in ℏ𝜔 = hc/𝜆, an interference fringe appears with a fringe distance hc/(2nd). Hence, the large d reduces the fringe distance. For example, in a-Si:H with n ≈ 3.6, hc/(2nd) amounts to ∼17 meV for d = 10 μm. The fringe is thereby obscured by the monochromator of resolution ∼30 meV; see Section 6.4.3 (details on the interference spectrum of thin films are given in Chapter 1). A PL-absorbing substrate should be avoided, since it will significantly affect the PL spectrum and reduce the intensity [39]. In some cases, the self-absorption of PL in the film also modifies the spectrum, which needs some corrections [1]. A grating monochromator is used to measure the PL spectrum: intensity vs. wavelength 𝜆, but a light of wavelength in higher orders such as 𝜆/2, 𝜆/3, etc., must be blocked by a long pass filter (LPF). However, care must be taken in the ultraviolet (UV) photoexcitation, where the blocking filter as well as other filters such as a neutral density filter

6.3 Experimental Aspects

Cryostat

Sample

Laser BPF

NDF

C1

C2

Lock-in Amp

Lens LPF Lens

Monochromator λ

PC

GPIB Bus

Detector

Figure 6.4 Experimental set-up of static PL measurement. C1, optical chopper at normal position; C2, optical chopper for residual PL decay measurement; BPF, band-pass filter; NDF, neutral density filter; LPF, long-pass filter; PC, personal computer.

(NDF) often fluoresces by a stray UV light. This is also the case even with UV-laser notch filters. The situation becomes more serious with a glass filter, which fluoresces even with a visible light excitation. Sometimes such an effect can be avoided or reduced by putting the filters at the output side of the monochromator. The PL sensitivity depends largely on the throughput of the monochromator as well as the sensitivity of the detector; a monochromator of a low f -number has high throughput at the expense of the resolution. Usually, focusing the light source at an input slit with the same f -number, we get the maximum throughput (optical matching). However, note that PL is not always a point source of light, but it has a finite size. If it is collected by a lens of a small f -number, for example, f/1.0, and incident on the slit of the monochromator with f/4.0 under the optical matching, the image of PL at the slit is magnified by 4 in this example. The PL image larger than the slit area loses some intensity, which is called a vignetting loss. Since we must trade between the throughput and the spectral resolution by adjusting the slit width as well, a grating monochromator between sample and detector often decreases the throughput by nearly three orders, even if it has low f -number. We can vary the 𝜆 of a monochromatized light by rotating the grating of the monochromator, manually or automatically with a stepping motor and a personal computer (PC)–controlled drive. However, a set of band-pass filters (BPFs) is an inexpensive and more efficient alternative for this, but sacrifices the arbitrary choice of 𝜆. Today, a charge-coupled device (CCD) is installed in a position of the output slit of the monochromator: a multi-channel spectrometer with electronically scanned 𝜆 by the CCD. Various handheld multi-channel spectrometers, equipped with USB interface and a coupler for optical fiber leading to low stray light, are commercially available from Ocean Optics, StellarNet, Hamamatsu, and so on.

165

166

6 Photoluminescence

Another alternative in the range of NIR and far infrared (FIR) is the Fourier transform infrared (FTIR) spectrometer based on a Michelson interferometer, which is commercially available from Bruker Optics, Oriel Instruments (Newport). The method is called Fourier transform photoluminescence (FTPL) technique [5, 38, 40, 41]. Unlike the dispersive spectrometer, the FTIR spectrometer collects all wavelengths simultaneously (Felgett Advantage) and does not use the entrance and exit slits. Hence, the PL sensitivity is increased, and the measuring time is saved by the FTPL. Bignazzi et al. [42] have, however, given some warnings about the limitations of the FTPL. The PL can be detected by various detectors such as photomultiplier tube (PMT), photodiode (PD), and photoconductor. A PMT is the most sensitive detector with high-speed response, and is hence usable for time resolved spectroscopy (TRS), as described in the following text. Its spectral response is usually limited below a wavelength of ≈900 nm. However, NIR PMT: R5509-72 (Hamamatsu Photonics) and its series of InGaAsP/InP photocathode having sensitivity up to 𝜆 ≈ 1700 nm is superior to other NIR detectors such as Ge PD in sensitivity and time response. PMT output may be directly measured with an electronic dc current meter, but usually we modulate the excitation light at rather low frequency, typically from a few Hz to 10s of kHz by setting an optical chopper C1 in front of the sample (Figure 6.4), and we thereby detect the PL by a lock-in amplifier with a significantly high sensitivity and high signal-to-noise ratio (S/N). However, care must be taken in choosing the chopping frequency, so that the chopping period may last longer than the PL lifetime for the reason described in Section 6.3.5. Incidentally, by setting the chopper C2 between the sample and the detector, we can avoid the preceding problem, and also can observe the residual decay of PL with high sensitivity and S/N after turning off the PL excitation light (Figure 6.4) as described in Section 6.4.4, where we need a complete darkroom and must strictly avoid stray NIR light arising from an optical switch of the chopper and the temperature T-control heater of the cryostat when a sample is installed into a cryostat to measure the T-dependence of PL. At a very low PL intensity, one can operate PMT in the digital mode, called photon counting method [5, 13, 43]. At the very low light intensity, photons incident on the photocathode of PMT are separated in time, and thus the PMT output consists of separated electric pulses corresponding to the respective incident photons. Since the number of PMT output pulses is proportional to the amount of incident light, we can measure the light intensity by electronically counting the output pulses, in contrast to the analogue mode, where the PMT output pulses overlap each other and eventually can be regarded as electric current of superposed shot noises. Even in the dark, however, the PMT has dark current (noise pulses) arising from various sources. The amplitude of a noise pulse is generally lower than that of the incident photon, and thus the noise pulse can be removed by an electronic discriminator, which significantly enhances the S/N of the photon counting method. We may also apply an avalanche photodiode (APD), a sort of PDs [5, 8], to the photon counting instead of the PMT, but its gain is at most ∼100 in the linear mode operation below the bias voltage of avalanche breakdown, whereas that of the PMT amounts to ∼107 . However, the APD operated above the avalanche breakdown voltage works as a single-photon avalanche diode (SPAD) or Geiger-mode APD, and has a typical gain of 105 ∼ 106 . Once the SPAD is precipitated into the breakdown state by the absorption of a photon, that is, the ON state, it cannot recover to the OFF state by itself,

6.3 Experimental Aspects

because it is biased above the avalanche breakdown. Therefore, quenching by a resistor or more sophisticated active circuitry is necessary to restore it to the OFF state for a subsequent photon. Nowadays, a sort of solid-state photomultiplier composed of SPAD pixels, implemented with the number of 102 ∼ 104 in a matrix 106 and surpasses conventional PMT in several points: the internal QE (>0.8) of SPAD is higher than the external QE ( 1. Therefore, in the static PL spectroscopy with lock-in detection of the chopper frequency 𝜔, one should set 𝜔 such that it satisfies 𝜔𝜏 ≪ 1. However, by using the optical chopper in the position of C2 between the sample and the detector (Figure 6.4), we can avoid such a limitation, provided that the PL detecting system is completely shielded in a perfect darkroom from unwanted light—for example, NIR light from the optical switch to control the chopper frequency 𝜔 and provide reference signal, and that from a cryostat heater (Section 6.3.1). More precisely, 𝜏 is deduced by fitting R(𝜔) and 𝜃(𝜔) to data of amplitude and phase delay as functions of 𝜔. Materials having the multi-exponential decay as given by Eq. (6.12) are usually analyzed by nonlinear least squares method for fitting [63]. Such FRS with varying frequency 𝜔 is called variable-frequency fluorometry, and is used for organic and biological molecules and a variety of condensed matters [13]. If a luminescent material has a broad lifetime distribution, such as amorphous semiconductors, it will be more tedious to obtain the lifetime distribution by the FRS. In order to overcome this difficulty of FRS as well as disadvantages of the biased TRS, Depinna and Dunstan [64] devised the modified FRS, called quadrature frequency resolved spectroscopy (QFRS), as described in the following text. The theoretical principle of QFRS was fully established by Stachowitz et al. [56].

6.3.6

Quadrature Frequency Resolved Spectroscopy (QFRS)

As stated in the preceding text, QFRS [56, 64] is more suitable for analyzing the broad PL lifetime distributions of disordered (amorphous) materials and the UCPL of RE-ion-doped materials having multi-lifetimes. Here again, we employ the alternating-current circuit theory in Figure 6.5; by substituting s = i𝜔 and separating real and imaginary parts, the transfer function of Eq. (6.9) I(s) becomes: I(iω) =

1 𝜔𝜏 1 −i = 1 + iω𝜏 1 + (𝜔𝜏)2 1 + (ω𝜏)2

,

(6.16)

where the real part gives the in-phase output and the imaginary part gives the quadrature output iQ (𝜔) against the input of modulated PL excitation: G¯aei𝜔t .

173

174

6 Photoluminescence

Then, as done in Eq. (6.13), here again we introduce the probability density P(𝜏) to write the quadrature part of the modulated PL as: ∞

iQ (𝜔) = Ga

P(𝜏)

∫0

𝜔𝜏 dτ, 1 + (𝜔𝜏)2

(6.17)

The function 𝜔𝜏/[1 + (𝜔𝜏)2 ] is single-peaked with maximum at 𝜔 = 𝜏 −1 , but it cannot be approximated as a delta function, because its integral from 𝜔 = 0 to ∞ diverges, and it is not a sharply peaked function. In order to solve this problem, we set 𝜔 = 10−x and 𝜏 = 10u (d𝜏 = ln 10 ⋅ 10u du) to get: ∞

ln10 ⋅ sech(ln10 ⋅ (x − u)) 𝜋 iQ (𝜔) = Ga 10u P(10u ) du ∫ 2 𝜋 −∞ ∞

=

𝜋 Ga 10u P(10u )𝜉(x − u)du , ∫ 2

(6.18)

−∞

where ξ(x) = ln10⋅sech(ln10⋅x) is the QFRS kernel for the logarithmic inverse angular fre𝜋 quency x = log10 𝜔−1 , and normalized for its integration from x = −∞ to ∞ to be unity, and peaks at x = 0 with FWHM of ≈1.14 (Figure 6.6). Thus, if the distribution function 10x P(10x ) is broad enough for 𝜉(x) to be regarded as a delta function, we may approximate the lifetime distribution in the logarithm of reciprocal 𝜔: x = log10 𝜔−1 as [56]: 𝜋 (6.19) iQ (𝜔) = iQ (10−x ) ≈ Ga10x P(10x ). 2 If this is not the case, we must recover the true lifetime distribution 10x P(10x ) by deconvoluting Eq. (6.18) to be reconciled with the QFRS data. However, the deconvolution is not always easy, because it often exaggerates the experimental noise. For this reason, two approximate methods of deconvolution are proposed in the following text. Method 1: Since the integration of sech x is tan−1 (sinh x), we can analytically convolute a simple rectangular function of x with the kernel. Assuming the lifetime distribution 10x P(10x ) to be a superposition of rectangular functions, similar to piled-up building blocks, we get an analytical expression of its convolution with𝜉(x). Then, fitting it to the QFRS data, we get approximate 10x P(10x ), again similar to piled-up building blocks [65]. Method 2: This is completely numerical. We assume the true lifetime distribution to be a linear combination of Gaussians, each having three unknown parameters, and then determine the parameters by a nonlinear regression, so that its numerical convolution with 𝜉(x) can reconcile with the data. Strictly, the true lifetime distribution is not always guaranteed to be the linear combination of Gaussians (Section 6.4.3). More specifically, we have derived the analytical form of the distant-pair (DP) recombination lifetime distribution with a single parameter of steady-state carrier concentration obtained from the theoretical impulse-response of the biased TRS given by Dunstan and Levin et al. [58, 59] separately and verified for the QFRS of a-Si:H (for details, see [66]). As mentioned in the preceding text, PL measurements for absolute QE of a film on opaque substrate is not established yet, but the deconvolution of QFRS spectrum gives the relative QE for a deconvoluted component by dividing its area by the total area of all the deconvoluted components. TRS and QFRS methods using the respective kernels 𝜆(x) and 𝜉(x) are mathematically equivalent [56], but QFRS has a decisive advantage in

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

the availability of a conventional lock-in amplifier with extreme sensitivity and S/N in the frequency range from 1 Hz to 100 kHz, corresponding to a five-decade lifetime from ∼1 μs to ∼0.1 s.

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique 6.4.1

Overview

Using the QFRS technique, Depinna and Dunstan [64] first demonstrated the PL lifetime distribution of a-Si:H more precisely than that using TRS measurements, and also observed emissions from singlet and triplet bound-excitons in GaP crystals, well-separated in lifetime as well as emission energy. Although the static PL spectroscopy of a-Si:H was nearly established in the 1970s and early 1980s [39, 67], considerable debates over its detailed mechanisms still continue even today [68–70]. It has been generally agreed that, in an intrinsic a-Si:H at a low temperature T, photoexcited electrons and holes immediately thermalize in extended and band-tail states by the hopping process, and intrinsic PL arises from the radiative tunneling (RT) transition between electrons and holes localized in their respective tail states [39, 67, 70]. The PL lifetime is governed by the RT lifetime identical to Eq. (6.2) of the DAP recombination as: ( ) 2R , (6.20) 𝜏 = 𝜏0 exp a where R is the e–h separation. The pre-exponential factor 𝜏 0 is the electrical-dipole transition time usually expected to be ∼10−8 s, and a is the extent of the larger of the electron and hole wave functions (localization length): usually, the electron Bohr radius of a-Si:H ∼1 nm. However, as stated in the preceding text, whether the recombination is geminate (recombination of e–h pair created by a single photon) or nongeminate (recombination of e–h pair created by different photons) is still a controversial issue. This comes from the featureless distribution of the PL lifetime 𝜏 as well as the PL emission energy ℏ𝜔 due to the disorder in amorphous semiconductors. In fact, the optical absorption due to an exciton being a typical geminate e–h pair has not been observed in a-Si:H even at low temperatures, unlike in crystalline semiconductors. Thus, the precise measurement of lifetime is necessary to investigate the PL of a-Si:H at a sufficiently low generation rate G for photocarriers, which can be done using the QFRS technique [56, 64]. Using this technique, Bort et al. [71] have observed a transition in the recombination kinetics of PL in a-Si:H at low T at a generation rate G ≈ 1019 cm−3 s−1 ; the peak lifetime 𝜏 p , that is, the peak position of QFRS spectrum, is constant under the condition G < 1019 cm−3 s−1 , whereas 𝜏 p decreases as G increases for G > 1019 cm−3 s−1 . They have proposed a geminate recombination model at G < 1019 cm−3 s−1 and the distant-pair (DP), or nongeminate recombination model at G > 1019 cm−3 s−1 . By contrast, the light-induced electron spin resonance (LESR) intensity, known to be proportional to steady state density of photoexcited e–h pair n in the band-tail states, depends on G sublinearly for the whole range of 1015 < G < 1020 cm−3 s−1 [71–74]. If the PL is simply governed by the DP recombination based on Eq. (6.20), since n ≈ G𝜏 p holds,

175

176

6 Photoluminescence

the contentious decrease of 𝜏 p with increasing G should give rise to the same sublinear G-dependence of n as the LESR intensity in the preceding whole range of G; however, this is not the case for G ≤ 1019 cm−3 s−1 [58, 71, 75]. Meanwhile, under the geminate condition (G < 1019 cm−3 s−1 ), Boulitrop and Dunstan [76] were the first to identify a double-peaked lifetime distribution of PL consisting of short-lived (∼μs) and long-lived (∼ms) components in a-Si:H by QFRS. The double-peak QFRS spectrum of a-Si:H was studied in further detail by Ambros et al. [77], expanding the high frequency limit to 2 MHz, corresponding to a lifetime of 𝜏 ≈ 0.1 μs. It is difficult to identify the two lifetime components on the basis of the RT model [76, 77]. Stachowitz et al. [78] have proposed the exciton involvement in the double-peak phenomenon, attributing the short- and long-lived components to singlet- and triplet-excitons, respectively. We have also observed a double-peak lifetime in the similar tetrahedral amorphous semiconductor of a-Ge:H, supporting the exciton model [69, 79, 80]. However, identifying the short-lived component of ∼𝜇s with singlet exciton recombination is a problem, in that the lifetime of a singlet exciton, or fluorescence lifetime, is normally less than 10 ns. In addition, a disadvantage of the QFRS is the minimum lifetime limited to the μs order due to the upper-limit frequency of the conventional lock-in amplifiers. Using the rf digital lock-in amplifier of the frequency range from 25 kHz to 200 MHz (SR844, Stanford Research System), we have developed a nanosecond QFRS system named dual-phase double lock-in (DPDL) QFRS, as described in the following text. 6.4.2

Dual-Phase Double Lock-in (DPDL) QFRS Technique

As illustrated in Figure 6.8a, we have developed the DPDL technique [65, 81, 82] to measure PL lifetime distribution from 2 ns to 5 μs, employing the SR844 rf digital lock-in amplifier and an electro-optic modulator (EOM, Conoptics) to modulate the laser beam from 0.5 Hz to 80 MHz. Since electromagnetic cross-talk between the EOM driver of high rf power (D) and the rf lock-in amplifier becomes serious at rf frequency 𝜔/(2𝜋) > 10 MHz, the laser beam was chopped at a low frequency of 𝜔m ≪ 𝜔 to discriminate the PL signal from the cross-talk by the double lock-in detection [65, 81]. Figure 6.8b represents the signal flow of DPDL. Instead of a complex exponential input ei (t) ∝ ei𝜔t in the real-coefficient Eq. (6.11) (Figure 6.5), employing its imaginary part, ei (t) ∝ sin(𝜔t), we obtain the output eo (t) ∝ R(𝜔) sin(𝜔t − 𝜃(𝜔)) from the imaginary part of Eq. (6.15), and then the doubly modulated PL signal becomes: S(t) = R(𝜔) sin(𝜔t − 𝜃(𝜔))• sin(𝜔m t),

(6.21)

where R(𝜔) and 𝜃(𝜔) are the PL amplitude and phase at 𝜔, respectively, in Eq. (6.15). When 𝜔m ≪ 𝜔, the quadrature part of Eq. (6.21) is given by R(𝜔)sin𝜃(𝜔) against the sinusoidal excitation, sin(𝜔t), which is proved to be proportional to iQ (𝜔) in Eq. (6.16). The time constant of LPF of the rf lock-in amplifier is set much greater than 𝜔−1 and much less than 𝜔m −1 . Thus, the rf lock-in amplifier outputs in-phase signal X(t) on the X-channel and quadrature signal Y (t) on the Y -channel, which are sinusoidal at the frequency of 𝜔m as: } X(t) = 1∕2R(𝜔) cos 𝜃(𝜔) sin(𝜔m t) . (6.22) Y (t) = 1∕2R(𝜔) sin 𝜃(𝜔) sin(𝜔m t)

Chopper BPF NDF 8Hz

sin(ωmt–ψ)

Cryostat

C

Sample Laser

Laser

EOM

FG DC-80MHz ~

D

Expander

XX Delay Line Lock-in Amp. X-out Ref. in XY RF Lock-in Amp. Ref. in YX Y-out Lock-in Amp. YY

sin ωt × sin ωmt

EOM

Sample

PL C

NIR-PMT

Lens NDF LPF

XY YX

DC-100MHz Amp

FG

XX

YY

Lock-in Amp.

PA RF Lock-in Amp.

Lock-in Amp.

Filter or Monochromator

sin ωt

X(t)

Y(t)

PMT S(t) = R(ω)sin(ωt – θ(ω))·sin(ωmt)

PC

PC GPIB bus (a)

(b)

Figure 6.8 (a) Experimental set-up of DPDL QFRS system. C, optical chopper; D, electronic driver for EOM (electro-optic modulator); FG, function generator; BPF, band-pass filter; LPF, long-pass filter; NDF, neutral density filter; NIR-PMT, near-infrared photomultiplier tube; PC, personal computer. Delay line: for compensation of instrumental phase-difference between signal and reference inputs of RF lock-in amp. (b) Block diagram of DPDL QFRS system. C, optical chopper with chopping frequency of 𝜔m and phase delay of 𝜓; FG, function generator; PA, preamplifier; PMT, photomultiplier tube; PC, personal computer. Solid line, electrical signal line; dotted line, optical pass. Source: Reproduced from T. Aoki, T. Shimizu, D. Saito, and K. Ikeda, J. Optoelectron. Adv. Mater., 7, 137 (2005) with permission from INOE.

178

6 Photoluminescence

These signals are again synchronously detected at 𝜔m using the two digital lock-in amplifiers (SR830) with LPF time constants much greater than 𝜔m −1 . An instrumental phase shift 𝜓 is inserted between the chopped light and the synchronous signal of the chopper driver (Figure 6.8b). Hence, the X- and Y -channel outputs of the two lock-in amplifiers (X X , X Y , Y X , and Y Y ) are given by: XX XY YX YY

= = = =

1 R(𝜔) cos 𝜃(𝜔) cos 𝜓 ⎫ 4 ⎪ 1 R(𝜔) cos 𝜃(𝜔) sin 𝜓 ⎪ 4 ⎬. 1 R(𝜔) sin 𝜃(𝜔) cos 𝜓 ⎪ 4 1 R(𝜔) sin 𝜃(𝜔) sin 𝜓 ⎪ ⎭ 4

(6.23)

Eliminating 𝜓 in Eq. (6.23), we obtain R(𝜔) and 𝜃(𝜔) as follows: √ ⎫ R(𝜔) = 4 XX2 + XY2 + YX2 + YY2 ⎪ ( )⎬ . 1 −1 YX +XY −1 YX −XY 𝜃(𝜔) = 2 tan X −Y + tan X +Y ⎪ X Y X Y ⎭

(6.24)

However, as R(𝜔) and 𝜃(𝜔) include components of instrumental responses due to the EOM and its driver, the PL detecting system, and optical and electrical lengths, the quadrature signal should be calibrated. It should be noted that a delay-line (RF coaxial cable) is intentionally inserted between the function generator (FG) and the reference input (Ref. in) of the RF lock-in amplifier in order to reduce the instrumental phase-difference between the signal and reference inputs in Figure 6.8a. By measuring the instrumental values of the amplitude R′ (𝜔) and phase 𝜃 ′ (𝜔) of the modulated laser light reflected by a roughened Al plate instead of the sample as a reference, the intrinsic QFRS signal of PL is given by [82]: R(𝜔) sin(𝜃(𝜔) − 𝜃 ′ (𝜔))∕R′ (𝜔).

(6.25)

The nanosecond order of lifetime resolution of the DPDL QFRS has been confirmed by measuring the fluorescence lifetime ∼4 ns at room temperature of Rhodamine 6G in 1 × 10−6 mol l−1 aqueous solution in agreement with that obtained by TCSPC [81, 83]. 6.4.3

Exploring Broad PL Lifetime Distribution in a-Si:H by Wideband QFRS

In order to explore the wide lifetime-distribution of amorphous semiconductors, by using an acousto-optic modulator (AOM) and the lock-in amplifier (Stanford SR830), we have also expanded the lower-limit frequency of QFRS down to 1 mHz (𝜏 ≈ 160 s); it is operated in internal reference mode together with synchronous filtering in order to reduce the phase noise, where a single QFRS measurement of 𝜏 ≥ 30 s needs 6 h by setting its time constant at 3 ks [84]. Thereby, the wideband QFRS system combining the DPDL QFRS with the internal reference mode makes it possible to analyze lifetimes over almost 11 decades from 2 ns to 160 s [84]. In the following, we show that our wideband QFRS results reveal the exciton involvement as well as DP recombination in the triple-peaked lifetime structure of the intrinsic a-Si:H. Similar phenomena observed in chalcogenide amorphous semiconductors as well as a-Ge:H indicate that the triple-peak QFRS spectrum, or the coexistence of

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

geminate and nongeminate recombination, is universal among amorphous semiconductors [85]. In addition, the residual PL decay of a-Si:H persisting for more than 104 s is presented; the DP recombination kinetics as well as steady-state photocarrier concentration obtained from QFRS results agrees with those of LESR [86]. Films of intrinsic a-Si:H were deposited on roughened Al substrates, with thickness of 1 ∼ 9 μm and defect density ≤ 2.0 × 1016 cm−3 , and intrinsic a-Ge:H with thickness of ≈1 μm and defect density ∼1016 cm−3 . PL signals were detected by a Hamamatsu R5509-42 NIR PMT at photon energies ranging from 0.9 to 1.7 eV for a-Si:H of bandgap energy EG ≈ 1.8 eV and R5509-72 NIR PMT from 0.7 to 1.5 eV for a-Ge:H. An optical system of f /1.0 ∼ 2.0 optics was designed to collect PL emission into the PMTs. The QFRS spectra of dispersed PL were measured by a 10 cm and f/3.0 monochromator with a resolution ∼30 meV. 6.4.3.1

Effects of Excitation Intensity, Excitation, and Emission Energies

Figure 6.9a shows G-evolved QFRS spectra of PL of a-Si:H from 2 ns to 160 s excited at the above-bandgap excitation energy Ex = 2.33 eV for generation rate G from 2.5 × 1015 to 5.0 × 1022 cm−3 s−1 [87]. We can see that the peak-positions of the longand short-lived components are fixed at 𝜏 T ≈ 3 ms and at 𝜏 S ≈ 2 μs, respectively, even though G changes from 2.5 × 1015 to 4.1 × 1019 cm−3 s−1 ; this is the well-known τS G [cm–3s–1] 5.0 × 1022

τT

a-Si:H EX = 2.33[eV] T = 3.7[K]

2.0 × 1022

106 104

1.9 × 1021 4.1 × 1020 1.2 × 1020 τD

4.1 × 1019 1.3 × 1018 2.8 × 1017

EX = 1.81[eV] τD

T = 3.7[K] a = 10[A] τ0 = 10–8[s]

100 10–2 τT 10–4

6.5 × 1016

10–6

1.3 × 1016 2.5 × 1015 10–8

EX = 2.33[eV]

a-Si:H

102

PL Lifetime τ [s]

QFRS Signal [a.u.]

5.0 × 1021

10–6

10–4

10–2

100

102

10–8 1012

τS 1014

1016

1018

1020

1022

PL Lifetime τ [s]

Generation rate G [cm–3s–1]

(a)

(b)

1024

Figure 6.9 (a) G-evolved QFRS spectra from 2 ns to 160 s for a-Si:H at 3.7 K and E X = 2.33 eV with various G from 2.5 × 1015 to 5.0 × 1022 cm−3 s−1 . The two data at G of 2 × and 5 × 1022 cm−3 s−1 were taken by laser light condensed through a lens. Source: Reproduced from T. Aoki, J. Non-Cryst. Solids, 352, 1138 (2006) by permission of Elsevier. (b) Peak positions 𝜏 S , 𝜏 T , and 𝜏 D as functions of G at 3.7 K at E X = 2.33 eV (⚬) and 1.81 eV (•). Solid line is calculated from balance equation. Source: Reproduced from T. Aoki, T. Shimizu, D. Saito, and K. Ikeda, J. Optoelectron. Adv. Mater., 7, 137 (2005) with permission from INOE.

179

180

6 Photoluminescence

double-peak lifetime distribution observed under the so-called geminate condition G ≤ 1019 cm−3 s−1 [76, 77]. By extending the longer lifetime limit with the internal reference mode, a third peak higher than the other two peaks appears at 𝜏 D ≈ 20 s for G ≈ 2.5 × 1015 cm−3 s−1 (see Figure 6.9a). This peak might have been overlooked earlier due to the lack of the very-low-frequency QFRS. As G increases, 𝜏 D continuously shortens, and the peak merges with the 𝜏 T -component at G ≈ 1.3 × 1018 cm−3 s−1 . Figure 6.9b shows the peak-positions 𝜏 D , 𝜏 T , and 𝜏 S for 1013 < G < 1023 cm−3 s−1 at the above-bandgap excitation of Ex = 2.33 eV (denoted by ⚬) and the bandgap excitation of Ex = 1.81 eV (•); the continuous shortening of 𝜏 D with increasing G is a salient feature of DP recombination based on the RT model and the plot of 𝜏 D vs. G fits the curve calculated 1 from the balance equation based on Eq. (6.20) with n ≈ G𝜏D , R ≈ n− 3 , 𝜏 0 = 10−8 s, and a = 1 nm [82, 88–90]. Thus, the PL of intrinsic a-Si:H is triple-peaked in the lifetime distribution at a low temperature of 3.7 K and Ex = 2.33 eV under the geminate condition G ≤ 1019 cm−3 s−1 , with the third peak-position of lifetime 𝜏 D decreasing from ∼20 to ∼0.1 s as G increases from ∼1015 to ∼1018 cm−3 s−1 . The peak lifetimes 𝜏 D (•), shorter by nearly two orders of magnitude at Ex = 1.81 eV in Figure 6.9b, indicate the effect of reduced thermalization of electrons excited by Ex close to EG ≈ 1.8 eV [82]. At sufficiently low G, the three peaks have very distinct lifetimes, suggesting that the recombination events at 𝜏 S , 𝜏 T , and 𝜏 D occur via three independent channels as non-competing recombination. However, when the 𝜏 D -component begins to merge with the 𝜏 T -component as G approaches ∼1018 cm−3 s−1 (Figure 6.9a), the two recombination events at 𝜏 T and 𝜏 D are no longer independent. Indeed, further increasing G to ∼1.2 × 1020 cm−3 s−1 shifts the combined component of 𝜏 T and 𝜏 D to shorter lifetimes and merges it with the 𝜏 S -component at G ≈ 1022 cm−3 s−1 , leading to a single-peak structure. Similar triple-peaked lifetime structures are also observed in the G-evolved QFRS spectra of a-Ge:H excited by Ex = 1.81 eV under the geminate condition G < 1019 cm−3 s−1 at T = 3.7 K [84]. In Figure 6.10, plots (⚬) show the steady-state carrier concentration nD = 𝜂 D G𝜏 D of a-Si:H as a function of G, where 𝜏 D and the relative QE 𝜂 D of the 𝜏 D -component are obtained by deconvoluting the G-evolved QFRS spectra (Figure 6.9a). The metastable carrier density nmet denoted by × and + deduced by integrating the residual PL decay of Figure 6.20 described in Section 6.4.4 and LESR spin densities are denoted by ∇ [71], Δ [73], and • [74]. All results shown in Figure 6.10 agree with the sublinear G-dependence ∝ G0.2 [86, 87, 91]. The steady-state carrier concentrations nS = 𝜂 S G𝜏 S denoted by ◽ and nT = 𝜂 T G𝜏 T denoted by ▴ with the respective relative QEs 𝜂 S and 𝜂 T calculated from Figure 6.9a are plotted as functions of G in Figure 6.10. In this case, nT increases almost linearly with G under the geminate condition G ≤ 1019 cm−3 s−1 , while 𝜏 T and 𝜂 T remain nearly constant with G (Figure 6.9b). When G exceeds 1019 cm−3 s−1 , the plots of nT coalesce with the sublinear curve of nD ∝ G0.2 with nT ≈ nD ≈ 1017 cm−3 . At around G ≈ 1019 cm−3 s−1 , the two components are observed to merge, as shown in Figures 6.9a and 6.10. Similarly, an extrapolation of the plot of nS vs. G intersects the sublinear curve at around nS ≈ nD ≈ 1018 cm−3 (G ≈ 1023 ∼1024 cm−3 s−1 ); the coalescence of all the components is also seen at around G ≈ 5.0 × 1022 cm−3 s−1 in Figure 6.9a. Now, by attributing 𝜏 S - and 𝜏 T -components to singlet and triplet excitons, respectively—in particular, STEs [52, 68, 92] and the 𝜏 D component to the DP recombination, we shall explain all the results. The values of nS ≈ 1018 cm−3 and

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

1020

Metastable Carrier Density [cm–3]

Figure 6.10 The plots ⚬, ◽, and ▴ indicate steady-state carrier densities vs. generation rate G: nD, nS , and nT , respectively, obtained from the QFRS spectra of Figure 6.9. The plots × and + denote metastable carrier density nmet vs. G deduced by integrating residual PL decay for high-quality film A and defective film B of a-Si:H (Figure 6.20), respectively. The plots ∇, Δ, and • denote LESR densities obtained from References [71, 73, 74], respectively. The unit of plots Δ is arbitrary [73]. The dotted line indicates the sublinear G-dependence of nmet ∝ G0.2 . Source: Reproduced from T. Aoki, J. Optoelectron. Adv. Mater., 11, 1044 (2009) with permission from INOE.

1018

1016

PL Decay (high quality) PL Decay (defective) DP (QFRS)

a-Si:H

G0.2

1014

1012

1010

108 1013

Triplet Exciton Singlet Exciton 1015 1017 1019 1021 Generation Rate G [cm–3s–1]

1023

nT ≈ 1017 cm−3 at the coalescences give the average inter-pair distance between the singlet excitons as ∼0.5nS −1/3 , amounting to ∼5 nm, and that between triplet excitons ∼0.5nT −1/3 , amounting to ∼10 nm. The spatial wavefunction of a singlet exciton is symmetric due to the anti-parallel spins, while that of a triplet exciton having parallel spins is antisymmetric; the exciton radius is smaller on average for the symmetric spatial function than the antisymmetric function [10, 87]. Thus, the triplet exciton coalesces at a lower concentration nT as compared to the singlet exciton. If a singlet exciton radius aex is close to its inter-pair distance ∼5 nm, the exciton binding energy e2 /2𝜅aex amounts to be ∼5 meV with a dielectric constant of a-Si:H: 𝜅 ≈ 12, which explains the disappearance of the 𝜏 S component at the low T ≈ 25 K (see Section 6.4.3.2). Figure 6.11a shows the QFRS spectra of dispersed PL of a-Si:H at 3.7 K, Ex = 2.33 eV, and G ≈ 2.3 × 1017 cm−3 s−1 ; the two lifetime peaks 𝜏 S and 𝜏 T get shorter as PL emission energy ℏ𝜔 increases, whereas 𝜏 D remains constant. The plots of recombination rates 𝜏 S −1 and 𝜏 T −1 vs. ℏ𝜔 are approximately proportional to (ℏ𝜔)3 in contrast to 𝜏 D −1 (Figure 6.11b). The (ℏ𝜔)3 dependence of 𝜏 S −1 and 𝜏 T −1 suggests the involvement of STEs in the 𝜏 S - and 𝜏 T -recombination according to References [86, 93–96]; details are available in the following text. On the other hand, 𝜏 D −1 remains constant with ℏ𝜔, which is explained by the involvement of DP recombination governed by Eq. (6.20) being independent of ℏ𝜔 (Figure 6.11b) [86, 91]. The reason why the observed lifetime 𝜏 S of singlet exciton is of the μs order is explained by the difference in e–h orbital sizes of the STEs [93, 97]. Kivelson and Gelatt [93] gave the radiative transition rate 𝜏 S −1 of the singlet exciton with a Bohr radius aB *: [ ][ 3] 1 4 1 e2 𝜔 = 𝜅− 2 (6.26) a2h S, 𝜏s 3 ℏc c2 where e2 /(ℏc) ≈ 1/137 is the fine structure constant, c is the speed of light, and S is an overlap term ∼(2ah /aB *)3 . Hence, this equation corroborates the proportionality between 𝜏 S −1 and (ℏ𝜔)3 . Omitting S reduces the right side of Eq. (6.26) to the radiative

181

6 Photoluminescence

a-Si:H G = 2.3 × 1017[cm–3s–1] τT Ex = 2.33[eV] Eћω[eV] τ T = 3.7[K] S 1.60 τD 1.55 1.50 1.45 1.40 1.35 1.30 1.25 1.20 1.15 1.10 1.05 1.00 0.95 0.90 10–8 10–6 10–4 10–2 100 102 PL Lifetime τ [s] (a)

107 Ex = 2.33eV 1/τS Emission Rate 1 / τ [s–1]

QFRS Signal [a.u]

182

105

(ћω)3

103

1/τT

101

1/τD G = 2 × 1017cm–3s–1

10–1 0.9 1.0

1.5 ћω [eV] (b)

Figure 6.11 (a) QFRS spectra of the monochromatized PL for emission energy ℏ𝜔 from 0.90 to 1.60 eV under the geminate condition G = 2.3 × 1017 cm−3 s−1 for a-Si:H at 3.7 K and E x = 2.33 eV. Source: Reproduced from T. Aoki, J. Non-Cryst. Solids, 352, 1138 (2006) by permission of Elsevier. (b) Double-log plots of emission rates for three recombination processes (𝜏 S −1 , 𝜏 T −1 , 𝜏 D −1 ) vs. PL emission energy ℏ𝜔 obtained by deconvoluting (a). The straight line indicates the power law 1/𝜏 ∞ (ℏ𝜔)3 on double-logarithmic scale. Source: Reproduced from T. Aoki, K. Ikeda, S. Kobayashi, and K. Shimakawa, PML, 86, 137 (2006) by permission of Taylor & Francis.

electric dipole transition rate 𝜏 0 −1 of an electric dipole moment eah [94]. Since aB * ≈ 2.4 nm for a-Si:H by Kivelson and Gelatt [93] and aB * ≈ 5 nm by us as mentioned earlier, 𝜏 S ranges from 1.2 to 11 μs, with the photon energy ℏ𝜔 = 1.5 eV, 𝜅 = 12, and ah = 0.2 nm. Considering an order of magnitude uncertainty in Eq. (6.26), our experimental result of 𝜏 S ≈ 2 μs agrees quite satisfactorily with the estimated value. By fixing the frequency at 39 kHz (𝜏 S ≈ 4.1 μs), 58 Hz (𝜏 T ≈ 2.7 ms), and 1.1 Hz (𝜏 D ≈ 0.14 s), the PL spectra of QFRS signals of the three components were taken at G ≈ 2.8 × 1017 cm−3 s−1 and T = 3.7 K for a sample of thickness d ≈ 9.3 μm to avoid the interference effects (Figure 6.12). Though fixing 𝜏 S and 𝜏 T is not strictly correct due to their dependence on ℏ𝜔 (Figure 6.11b), the PL spectrum of the QFRS signal at 𝜏 S ≈ 4.1 μs is similar to that of 𝜏 T ≈ 2.7 ms, except for a shift to higher ℏ𝜔 by ∼40 meV. This has been confirmed by more correct plots of areas of the deconvoluted 𝜏 S - and 𝜏 T -components from Figure 6.11a vs. ℏ𝜔 [88]. The spectrum of the 𝜏 D -component is the lowest in ℏ𝜔, suggesting that this component arising from deeper states (Figure 6.12). An increase of Ex does not alter the QFRS spectra for 𝜏 S and 𝜏 T so much, but enhances the 𝜏 D -component as well as 𝜏 D at G ≈ 2.0 × 1015 cm−3 s−1 (Figure 6.13a). Figures 6.12 and 6.13a suggest that both the 𝜏 S - and 𝜏 T -components originate from similar recombination processes except the difference ∼40 meV in ℏ𝜔, which is attributed to the exchange energy between singlet- and triplet-excitons [22, 64, 80, 87]. The increase in 𝜏 D and 𝜂 D by increasing Ex is attributed to the thermalization (diffusion) of DPs, since

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

1.0

τD

τS

2.33 1.94 1.81 1.70 1.63 1.58 1.53 1.49

τD

0.6

τT

0.4

10–1

1

1.2

1.4 ћω [eV]

10–5 10–3 10–1 PL Lifetime τ [s] (a)

101

1.8

104

ηS

103 α

10–2

102

10–3

101 G = 2.0 × [cm–3s–1]

1015

10–4 10–7

1.6

ηD ηT

T = 3.7[K]

1.46 10–9

τS

0.2

1 Quantum Efficiency η [a.u.]

QFRS Signal [a.u]

Ex[eV]:

τT

Resolution

0.8

0.0

a-Si:H

Ex = 2.33[eV] G = 2.8 × 1017[cm–3s–1]

a-Si:H T = 3.7[K]

1.4

1.5 1.6 1.7 1.8 1.9 Excitation Energy Ex [eV]

Absorption Coefficient α [cm–1]

Normalized QFRS Signal [a.u.]

Figure 6.12 PL spectra of QFRS signals at fixed lifetime 𝜏 S ≈ 4.1 μs, 𝜏 T ≈ 2.7 ms, and 𝜏 D ≈ 0.14 s for a-Si:H of thickness ∼9.3 μm at 3.7 K, E X = 2.33 eV, and G = 2.8 × 1017 cm−3 s−1 . All the peaks of spectra are normalized to unity: actual peaks for 𝜏 S and 𝜏 D are, respectively, ∼30% and ∼80% of that for 𝜏 T . Source: Reproduced from T. Aoki, J. Non-Cryst. Solids, 352, 1138 (2006) by permission of Elsevier.

100 2.0

(b)

Figure 6.13 (a) QFRS spectra of a-Si:H excited at various PL excitation energy E x with G ≈ 2.0 × 1015 cm−3 s−1 at 3.7 K. (b) PL excitation spectra (PLE) for three QFRS components: 𝜂 S (◾), 𝜂 T (▴), and 𝜂 D (•); each component is deconvoluted and normalized by PL excitation intensity; 𝛼 denotes absorption coefficient (cm−1 ) at 3.7 K; the sample is the same as in Figure 6.12. Source: Reproduced from T. Aoki, K. Ikeda, N. Ohrui, S. Kobayashi, and K. Shimakawa, J. Optoelectron. Adv. Mater., 9, 70 (2007) with permission from INOE.

an increase in Ex will extend the DP distance in band-tail states as noted for Figure 6.9b [82, 84]; also see Section 6.4.3.2. Figure 6.13b shows the PLE spectra for the three relative QEs: 𝜂 S , 𝜂 T , and 𝜂 D , together with the absorption coefficient 𝛼 (cm−1 ), where the relative QEs are the deconvoluted components from Figure 6.13a divided by the PL excitation intensity at each Ex . Fast decline in 𝜂 D compared to that in 𝜂 S and 𝜂 T by lowering Ex below the bandgap EG ≈ 1.83 eV (for 𝛼 = 103 cm−1 at 3.7 K), as well as the lowest PL spectrum of the 𝜏 D -component (Figure 6.12), indicates that the PL of the 𝜏 D -component has the largest Stokes shift due to DPs deeply trapped in tail states [87, 98]. The theory of such a large Stokes shift is given in Chapter 7.

183

6 Photoluminescence

6.4.3.2

Temperature Dependence

Figure 6.14 shows the temperature (T)-evolved QFRS spectra of the a-Si:H film (S163) having a defect density of ∼1 × 1016 cm−3 excited at Ex = 2.33 eV and G ≈ 7.0 × 1016 cm−3 s−1 [91, 99, 100]. As T is raised from 3.7 K to 100 K, the 𝜏 T -component shifts to shorter lifetimes and disappears at T ≈ 75 K. The third peak at 𝜏 D , on the other hand, remains persistent up to ∼133 K, and 𝜏 D shortens continuously as T increases. The 𝜏 S -component disappears at T ≈ 25 K, but, concomitantly, another shoulder denoted by 𝜏 G ≈ 10−5 ∼ 10−4 s emerges between 𝜏 S and 𝜏 T , growing into a hump at higher T. Thus, a new double-peak structure with maxima at 𝜏 D and 𝜏 G appears in the QFRS spectra of a-Si:H at ∼80 K [91, 99, 100]. The reason for the disappearing 𝜏 S -component at T ≈ 25 K and 𝜏 T -component at T ≈ 75 K is explained by singlet- and triplet-excitons, respectively. This is because the binding energy of singlet excitons is smaller than that of the triplet-excitons by the exchange energy of ∼40 meV, as mentioned in Section 6.4.3.1. Moreover, T ≈ 75 K for the disappearance of the 𝜏 T -component corresponds to the onset of instability of

F = 0[kV/cm] a-Si:H(S163) Ex = 2.33[eV] G ≈ 7 × 1016[cm–3s–1]

Normalized QFRS Signal [a.u.]

184

τT τD

τD

10 14

20 25

τG

20

30 40 47 50 55 60 70 80 85 90 100 10–9

τT

7.5

τS

T [K] 3.7 15

G ≈ 4.3 × 1017 [cm–3s–1]

a-Ge:H Ex = 1.81eV τS T [K] 3.7

30 40 50 60 70 τG

10–7

10–5 10–3 PL Lifetime τ [s] (a)

85 10–1

10

10–9

10–7

10–5 10–3 10–1 PL Lifetime τ [s]

101

(b)

Figure 6.14 (a) Temperature T-evolved QFRS spectra of undoped a-Si:H film (S163) without application of electric field (F = 0 kV cm−1 ). PL was excited at E X = 2.33 eV with G ≈ 7 × 1016 cm−3 s−1 . Source: Reproduced from T. Aoki, N. Ohrui, C. Fujihashi, and K. Shimakawa, PML, 88, 9 (2007) by permission of Taylor & Francis. (b) a-Ge:H photoexcited at E x = 1.81 eV with G ≈ 4.3 × 1017 cm−3 s−1 for various temperature T. Source: Reproduced from T. Aoki, T. Shimizu, S. Komedoori, S. Kobayashi, and K. Shimakawa, J. Non-Cryst. Solids, 338–340, 456 (2004) by permission of Elsevier.

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

self-trapped holes forming the STE [92]. In contrast, the 𝜏 D -component gets enhanced, and 𝜏 D is shortened by elevating T, which is explained in detail in Section 6.4.3.3. A similar T-dependence of QFRS spectra is observed in a-Ge:H at Ex = 1.81 eV and G ≈ 4.3 × 1017 cm−3 s−1 , as shown in Figure 6.14b [84]; the 𝜏 S - and 𝜏 T -components disappear at ∼14 K, whereas the 𝜏 D -component survives up to ∼85 K. Another peak becomes noticeable at 𝜏 G ≈ 20 μs between 𝜏 S and 𝜏 T when T ≈ 20 K and the QFRS spectrum of a-Ge:H turns into a double-peak structure with maxima at 𝜏 G and 𝜏 D at T ≈ 20 ∼ 85 K. The lower thresholds of T for the disappearance of the 𝜏 S - and 𝜏 T -components in a-Ge:H may be attributed to the smaller exciton binding energy due to smaller bandgap of a-Ge:H as compared to a-Si:H [80]. Figure 6.15 shows the G-evolved QFRS spectra of a-Si:H at T ≈ 100 K and Ex = 2.33 eV, where 𝜏 G remains constant as G increases up to ≈1019 cm−3 s−1 , whereas 𝜏 D continues to decrease [82, 84, 100]. This suggests that a part of the photogenerated e–h pairs are not in the form of excitons, but remain as geminate pairs at 100 K; presumably, the spin effect on geminate pairs fades out at ∼100 K, but the Coulombic effect persists even above 100 K. The effects of electric field F, T, and Ex on the 𝜏 G component are explained by the classical Onsager model (see Section 6.4.3.3 and Reference [100]). 6.4.3.3

Effect of Electric and Magnetic Fields

Figure 6.16 shows the electric field F-evolved QFRS spectra (at T = 3.7 K) of the a-Si:H film (S163 used for Figure 6.14 in Section 6.4.3.2) deposited on Al2 O3 (sapphire) of high thermal conductivity and equipped with 0.5 mm spaced coplanar electrodes [99, 100]. a-Si:H T = 100[K] G [cm–3s–1] 6.7 × 1014

Ex = 2.33eV τG

τD

2.4 × 1015 Normalized QFRS Signal [a.u.]

Figure 6.15 G-evolved QFRS spectra of a-Si:H excited with various G = 6.7 × 1014 − 1.2 × 1021 cm−3 s−1 at T = 100 K and E x = 2.33 eV. Source: Reproduced from T. Aoki, T. Shimizu, S. Komedoori, S. Kobayashi, and K. Shimakawa, J. Non-Cryst. Solids, 338–340, 456 (2004) by permission of Elsevier.

9.8 × 1015 3.8 × 1016 1.5 × 1017 7.7 × 1017 4.0 × 1018 2.0 × 1019 1.0 × 1020 4.9 × 1020 1.2 × 1021 10–8

10–6

10–4 10–2 PL Lifetime τ [s]

100

102

185

6 Photoluminescence

T = 3.7[K]

Figure 6.16 Electric field F-evolved QFRS spectra of undoped a-Si:H film (S163) at T = 3.7 K. E X and G are the same as in Figure 6.14a. Source: Reproduced from T. Aoki, N. Ohrui, C. Fujihashi, and K. Shimakawa, PML, 88, 9 (2007) by permission of Taylor & Francis.

τT τD

F [kV/cm] Normalized QFRS Signal [a.u.]

186

τS

0 10 20 30 40 50 60 70 80 90 100 10–9

10–7

10–5

10–3

10–1

101

PL Lifetime τ [s]

Increasing the electric field F up to 100 kV cm−1 monotonically decreases the 𝜏 S and 𝜏 T components, but only slightly, and it enhances the DP component and shortens the lifetime 𝜏 D more significantly. In this case, no 𝜏 G component appears at all (Figure 6.16). On the other hand, raising T from 40 to 80 K (at F = 0 kV cm−1 ) increases the DP component and shortens 𝜏 D in Figure 6.14a [99]. Thus, there appears to be a strong resemblance between the T- and F-evolved DP lifetimes 𝜏 D as shown in Figure 6.17, where 𝜏 D is plotted simultaneously as a function of T at F = 0 kV cm−1 and F at T = 3.7 K in the logarithmic T and F scales for the samples S163 and S119 of a defect density ∼2 × 1016 cm−3 . The plots of 𝜏 D vs. T for T ≥ 35 K coincide with those of 𝜏 D vs. F for F ≥ 50 kV cm−1 by shifting along the bidirectional arrow (Figure 6.17), which has been interpreted on the basis of the theory of Shklovskii et al. [89, 101] and Monroe [102]. According to their theory, a photogenerated electron at low T makes energy loss hops (ELHs), and the electron either recombines with a localized hole separated by R in a recombination lifetime given by Eq. (6.20), or makes an ELH to another site at a distance r with a lower energy in a hopping time 𝜏 d , given by: ( ) 2r , (6.27) 𝜏d = 𝜈0−1 exp a where 𝜈 0 is the attempt-to-escape frequency of the order of phonon frequency ∼1012 s-1 . The hopping diffusion dominates the recombination at early steps of ELHs due to the significant difference between 𝜏 0 and 𝜈 0 −1 in respective Eqs. (6.20) and (6.27). After successive ELHs, R becomes of the same order as r, and also becomes larger than the

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

101

(S163)T = 3.7[K] (S119)T = 3.7[K] DP Lifetime τD [s]

Figure 6.17 DP lifetime 𝜏 D as a function of T for the samples (•) S119 and (⚬) S163 at F = 0 kVcm−1 and 𝜏 D as a function of F for the samples (×) S119 and (◽) S163 at T = 3.7 K. Bottom scale is logarithms of both T [K] and F [kVcm−1 ]. Bidirectional arrow indicates that the data for T ≥ 35 K and those for F ≥ 50 kVcm−1 coincide by shifting along it. Source: Reproduced from T. Aoki, N. Ohrui, C. Fujihashi, and K. Shimakawa, PML, 88, 9 (2007) by permission of Taylor & Francis.

100

10–1

F

(S163)F = 0[kV/cm]

T

(S119)F = 0[kV/cm]

10–2 10

20 50 T [K] & F [kV/cm]

100

characteristic length Rc = (a/2)ln(𝜈 0 𝜏 0 ). A computer simulation shows that the recombination competes with the hopping diffusion under a condition r and R > Rc , where it is diffusion-limited recombination with the recombination lifetime 𝜏 being indistinguishable from the hopping time 𝜏 d [89]. Moreover, as T is elevated above ∼20 K, the so-called the transport energy (TE) in the tail state starts to shift upwards to the mobility edge, and electrons localized in tail states deeper than TE hop up to the vicinity of the TE, which increases recombination paths as well as the hopping diffusion, eventually leading to the shortening of 𝜏 D [89, 102, 103]. Similar shortening of 𝜏 D and enhancement of 𝜂 D were observed by applying infrared of energy 0.3 eV and 75 mW cm−2 to a-Si:H at 3.7 K, which supports the hop-up of localized electrons [98]. The hopping length at TE is given by rt = 3𝜀0 a/(2kT), with the exponential decay parameter of the electron tail state 𝜀0 ≈ 25 meV [101]. When T is elevated to T d = 3𝜀0 /k ln(𝜈 0 𝜏 0 ) for rt = Rc , where the condition of the diffusion-limited recombination r and R > Rc no longer holds, one theoretically obtains T d ≈ 94 K for a-Si:H [99, 101]. The interplay between T- and F-dependent 𝜏 D is interpreted on the basis of the effective temperature theory by introducing the effective temperature T eff = 𝛾eFa/k with elementary charge e and shifting the data of either T- or F-dependent 𝜏 D along the bidirectional arrow to coincide with each other in Figure 6.17. Here, we obtain the proportional factor 𝛾 ≈ 0.6 with a = 1 nm, which is surprisingly close to 𝛾 ≈ 0.67 obtained from a hopping transport parameter: photoconductivity (PC) in a-Si:H [104, 105]. Strictly, F-dependent PC was not obtained at T = 0 K, but a finite T, and thus the influence of finite T and F on PC, was parametrized by a single quantity 𝛽: Teff (T, F) = [T 𝛽 + (γeFa∕k)𝛽 ]1∕𝛽

(6.28)

From T- and F-dependences of PC, 𝛽 = 2 was determined in a-Si:H [104, 105]. We have measured 𝜏 D for various values of F at the intermediate temperature T = 40 and 55 K for the samples S119 and S163, and optimized 𝛽 = 1.4, so that all the data of the two samples may conform on a line [99].

187

6 Photoluminescence

In contrast, the 𝜏 G -component decreases by increasing F up to 70 kV cm−1 , as well as by increasing T from 40 to 100 K, which can be expressed by the relevant thermalization length r0 from the classical Onsager theory for e–h pairs [100, 106]. The T- and F- dependences of the 𝜏 G component can be explained by the increase of r0 caused by raising T and/or Ex above EG ≈ 1.8 eV of a-Si:H from 1.95 to 2.33 eV [100]. A magnetic field is found to have little influence on the 𝜏 S -component up to 0.9 T, but it enhances the 𝜏 T -component and reduces the 𝜏 D -component in the QFRS spectra of a-Si:H at 3.7 K, Ex = 2.33 eV, and G ≈ 7.7 × 1016 cm−3 s−1 , as shown in Figure 6.18a,b [82]. Three QFRS signals at fixed lifetimes (𝜏 S ≈ 2 μs, 𝜏 T ≈ 3 ms, and 𝜏 D ≈ 80 ms) vs. the magnetic field intensity are plotted as a function of the magnetic field intensity in Figure 6.18b. The preceding effects of the magnetic field on the PL lifetimes are interpreted as follows. Robins and Kastner [107] observed an enhancement of triplet exciton recombination for both amorphous and crystalline As2 Se3 under a magnetic field by TRS. Although singlet exciton state is absent in these materials [81, 82, 107–109], we have confirmed the same magnetic field effect on a-As2 Se3 by QFRS as well [85]. According to Robins and Kastner’s explanation, the 𝜏 S component assigned to a singlet exciton of total spin S = 0 is unaffected by a magnetic field. In contrast, the transition of triplet exciton of total spin S = 1 to the singlet ground state (S = 0) is spin-forbidden, but only allowed by weak coupling with the singlet excited state, and hence its relaxation rate depends on the coupling strength, as mentioned about Figure 6.1 in Section 6.2. Application of a magnetic field induces Zeeman splitting in the triplet excited state, increasing the coupling strength to enhance the triplet exciton recombination. Thus, the enhancement of the 𝜏 T component under the magnetic field may be attributed to the triplet exciton recombination in a-Si:H as well as in a-As2 Se3 [82, 85]. By a distinct method of pulsed optically

1

1 G = 7.7 × 1016[cm–3s–1] a-Si:H 0[T] 0.9[T]

T = 3.7[K] τT

τS

Ex = 2.33[eV]

τD

0 10–6 10–5 10–4 10–3 10–2 10–1 100 101 102 PL Lifetime τ [s] (a)

τT = 3 × 10–3[s] QFRS Signal [a.u.]

QFRS Signal [a.u.]

188

τD = 8 × 10–2[s]

τS = 2 × 10–6[s]

0 10–4

10–3

10–2

10–1

100

Magnetic Field [T] (b)

Figure 6.18 (a) QFRS spectra of a-Si:H at 3.7 K, E x = 2.33 eV, and G = 7.7 × 1016 cm−3 s−1 with (•) and without (⚬) application of a 0.9 T magnetic field. (b) QFRS signals of a-Si:H at fixed lifetimes 𝜏 S = 2 μs, 𝜏 T = 3 ms, and 𝜏 D = 80 ms as functions of magnetic field intensity. Source: Reproduced from T. Aoki, T. Shimizu, D. Saito, and K. Ikeda, J. Optoelectron. Adv. Mater., 7, 137 (2005) with permission from INOE.

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

STE

DP

~5 meV ~40 meV

EC EG ≈ 1.8 eV

Figure 6.19 Schematic models of three types of radiative recombination in a-Si:H of bandgap E G ≈ 1.8 eV at low temperatures. STE, self-trapped exciton; S, singlet; T, triplet; STH, self-trapped hole; R, intra-distance of a DP; r, an ELH (energy loss hop) distance.

EV

T

S ћω

ћω

ћω STH ~0.4 eV R

r

detected magnetic resonance (pODMR), Lips et al. [110] have found a direct evidence of the triplet exciton in a-Si:H at a rather high G ≈ 2 × 1022 cm−3 s−1 . Meanwhile, the spin directions of the DPs are completely random in the absence of a magnetic field; the total spin S is not a good quantum number because of the lack of correlation in the pair. Application of a magnetic field aligns spins of carriers in average, and thus the number of spin-aligned pairs (spin-forbidden to recombine) exceeds that of antiparallel-spin pairs (spin-allowed to recombine), leading to the paramagnetism of DPs. Hence, the spin-aligned DPs will be prevented from recombination by the magnetic field, which decreases the 𝜏 D component (Figure 6.18a,b) [82]. A recombination model that takes into account all these in a-Si:H at a low T is presented in Figure 6.19. 6.4.4

Residual PL Decay of a-Si:H

The kinetics of spectrally integrated PL intensity I(t) was measured by turning off 2.33 eV excitation laser by an electro-mechanic shutter after sufficiently long time irradiation on a-Si:H films to allow I(t) to reach its steady state value. Subsequently, the residual PL decay I(t) was lock-in detected by chopping PL emission in the position of C2 of Figure 6.4, so that I(t) may be free from chopping frequency, where all experimental procedures were performed electronically [86]. Figure 6.20 shows the double-log plots log10 I(t) vs. log10 t after cessation of PL excitation at 3.7 K with various G from 2 × 1013 to 2 × 1019 cm−3 s−1 for a-Si:H films A and B. The film A has the defect density 2 × 1016 cm−3 , and the more defective film B of defect density 6 × 1016 cm−3 is excited at G = 2 × 1013 cm−3 s−1 for comparison. Except for the I(t) observed at the lowest G of 2 × 1013 cm−3 s−1 , all other PL decay curves of the film A converge asymptotically at ∼104 s, even though they are initially different by five orders of magnitude. As mentioned in Section 6.4.3.1, the metastable carrier density of the film A, nmet (×), was calculated for various G by the integration by substitution of I(t) with t = 10x and adjusting nmet to nD at G = 1016 cm−3 s−1 (inset of Figure 6.20) [86]; both nmet and nD show sub-linear G-dependences (∝ G0.2 ), agreeing with the LESR densities (∇[71], Δ[73], •[74]). These results support that LESR measures the photocarriers localized in band-tail states, which subsequently recombine to give the very-long-lived PL. However, as G decreases below ∼1016 cm−3 s−1 , nmet as well as the LESR densities deviates from the sub-linear dependence of G0.2 ; the deviation is larger for the more defective film B

189

6 Photoluminescence

Figure 6.20 Double-log plots of residual PL decay of a-Si:H at 3.7 K and E x = 2.33 eV with various values of G—(1A) G = 2.0 × 1019 cm−3 s−1 , (2A) 2.0 × 1018 , (3A) 2.0 × 1017 , (4A) 2.0 × 1016 , (5A) 2.0 × 1015 , (6A) 2.0 × 1014 , (7A) 2.0 × 1013 —for film A of defect density 2 × 1016 cm−3 ; (7B) G = 2.0 × 1013 cm−3 s−1 for film B of defect density: 6 × 1016 cm−3 . Plots are fitted to the derivative of a stretched exponential (SE) function (solid red curves). Inset is 10x I(10x ) vs. x = log10 t for G corresponding to (4A), (5A), and (6A). Source: Reproduced from T. Aoki, K. Ikeda, S. Kobayashi, and K. Shimakawa, PML, 86, 137 (2006) with permission from Taylor & Francis.

103 4A

101

10xI(10x) [a.u.]

1A. 2A. 3A. PL Intensity I(t) [a.u.]

190

10–1

2.0

6A 0 –2

4A.

5A

1.0

0

2 X = log10 t

4

5A.

10–3

6A. 7A. 7B.

10–5

10–7 10–2

10–1

100

101 102 Time t [s]

103

104

105

(+) shown in Figure 6.10. Actually, the PL decay curve of the defective film (curve 7B in Figure 6.20) deviates more pronouncedly from the asymptotically converged curves at t ≈ 103 s. Thus, non-radiative recombination at the defect sites probably participates in shortening the fate of metastable carriers, when nmet becomes comparable with the defect density [86, 101, 111]. In order to avoid the contribution of excitonic or geminate recombination to the early stage of the PL decay at times t < 1 s in Figure 6.20, we have fitted the derivative of a stretched exponential (SE) function Kt 𝛽-1 exp[−(t/𝜏 0 )𝛽 ] to each PL decay I(t) for t > 10 s after the cessation of PL excitation. Here, 𝜏 0 is the effective recombination time, and 𝛽 is a dispersion parameter (0 < 𝛽 < 1). It is seen that I(t) after ∼10 s obeys the derivative of SE function for various G (≤2 × 1019 cm−3 s−1 ), and various T (described below in Figure 6.22), which manifests the DP recombination governed by time-dispersive monomolecular reaction. Figure 6.21a,b show, respectively, 𝛽 and 𝜏 0 (I-shape bars) vs. log10 G obtained by the non-linear regress for Figure 6.20 [91]. The lengths of the I-shape bars indicate the extents of 𝛽 and 𝜏 0 , where the nonlinear least square fit keeps the 𝜒 2 value within 1.5 times its minimum value. The squares (◽) in Figure 6.21a,b represent, respectively, the 𝛽 and 𝜏 0 values estimated by Morigaki [112] from the data of LESR kinetics of Yan et al. [74] for various G. Here, we notice a very close similarity in the two kinetic parameters 𝜏 0 and 𝛽 between PL and LESR. Since 𝜏 0 is relevant to the derivative of SE function at t around 𝜏 0 , a measuring time shorter than 𝜏 0 may cause a noticeable error. We have measured the PL decay over t > 104 s, while the LESR decay was measured only until t ≦ 2 × 103 s. Thus, one order difference of 𝜏 0 in magnitude between the PL and

6.4 Photoluminescence Lifetime Spectroscopy of Amorphous Semiconductors by QFRS Technique

0.6 PL 0.5

LESR

β

0.4 0.3 0.2 0.1 0.0 15

16

17

18

19

20

19

20

Log(G) [cm–3s–1]

Log(τ0) [s]

(a) 5 4 3 2 1 0 –1 –2 –3 –4 –5 –6 –7 –8 15

PL LESR

16

17

18

Log(G) [cm–3s–1] (b)

Figure 6.21 Parameters of SE function (I-shape bars) obtained by curve-fitting of Figure 6.20. Squares (◽) obtained by Morigaki from LESR decay [112]. (a) Dispersion parameter of SE function 𝛽 vs. G; (b) effective recombination lifetime 𝜏 0 vs. G. Source: Reproduced from T. Aoki, J. Optoelectron. Adv. Mater., 11, 1044 (2009) with permission from INOE.

LESR kinetics arises from the differences in the measuring time t, Ex , and T, and its exponential dependence on 1/𝛽. For details, see Reference [91]. Figure 6.22 demonstrates the PL decay I(t) at various T under the geminate condition of G ≈ 2.0 × 1016 cm−3 s−1 , where the PL decay can still be represented by the derivative of SE function. At an elevated T, photogenerated electrons diffuse apart by thermally activated hop-ups as well as hop-downs in the tail states and recombine with holes liberated from the self-trapping, which is the so-called dispersive bimolecular recombination. In fact, both 𝛽 and 𝜏 0 decrease with raising T, and I(t) approaches a power law (curve T4). The overall features of the PL decay can be given by the combined effects of radiative and nonradiative recombinations. Nevertheless, Figure 6.23 shows that the plots of integrated I(t), nmet (Δ) vs. T, agree with the LESR data (+) from Reference [113], and (◾) from Reference [114] as well as nD (⚬) obtained from the T-evolved QFRS spectra [87].

191

6 Photoluminescence

PL Intensity I (t) [ a.u.]

Figure 6.22 Double-log plots of residual PL decay of a-Si:H film A with G = 2.0 × 1016 cm−3 s−1 and E x = 2.33 eV at various temperatures T; (T1) 3.7 K, (T2) 100 K, (T3) 130 K, (T4) 150 K. Plots are fitted to the derivative of an SE function (solid red curves). Source: Reproduced from T. Aoki, K. Ikeda, S. Kobayashi, and K. Shimakawa, PML, 86, 137 (2006) by permission of Taylor & Francis.

G = 2.0 × 1016[cm–3s–1]

T1. T2.

10–1

T3. T4. 10–3 3.7[K]

10–5

150[K] 130[K] 100[K]

10–7 10–2 10–1

100

101 102 Time t [s]

103

104

105

1017 a-Si:H Carrier Con. n [cm–3]

192

1016

QFRS PL Decay

1015

LESR [a.u.] 1 LESR [a.u.] 2 1014

0

20

40

60 80 100 120 Temperature T [K]

140

160

Figure 6.23 Plots of steady-state carrier concentrations n vs. T; the plots nmet (Δ) estimated from residual PL decay, (⚬) from QFRS [84, 87] and the LESR intensities in arbitrary units as functions of T for a-Si:H; (+) from Ref. [113], (◾) from [114]. Source: Reproduced from T. Aoki, J. Non-Cryst. Solids, 352, 1138 (2006) by permission of Elsevier.

6.5 QFRS on Up-Conversion Photoluminescence (UCPL) of RE-Doped Materials Unlike amorphous semiconductors with broad PL spectra and lifetime distributions, RE ions in condensed matters possess ladder-like energy levels and emit narrow PL spectra similar to line spectra of isolated atoms due to the f-f transitions. The relaxation time between the two levels of an RE ion ranges from μs to ms.

6.5 QFRS on Up-Conversion Photoluminescence (UCPL) of RE-Doped Materials

We assume an M-level model of the RE ion energy, and the population density N i of i-th (i = 0, 1, 2,…M−1) level of RE ions is obtained from a set of M simultaneous rate equations, which are intrinsically nonlinear, since UCPL involves multi-photoexcitation steps. Hence, a set of nonlinear differential equations for N i cannot be solved in a closed form, and are therefore analyzed either numerically or approximately [29, 30, 115, 116]. As mentioned in Section 6.3.3, a biased TRS is technically difficult to realize, because a small impulse response of PL cannot be obtained by a phase-sensitive (lock-in) technique [58, 59]. However, the QFRS sinusoidally modulates the CW excitation flux density 𝛷 with a perturbation 𝜙, which induces sinusoidal perturbation ni in the time-invariant population density N i at each i-th level, and detects it by the lock-in technique with very high S/N [56, 64]. Here, N i is normalized by the total density of RE ions N ion (cm−3 ) and numerically calculated from the M-1 nonlinear rate equations ∑M−1 with the time-derivative Ṅ i = 0 and i=0 Ni = 1, where i = 0 denotes the ground state. On the assumption |𝜙| ≪ 𝛷, the second-order perturbation terms such as ni 𝜙 and ni nj can be neglected. Accordingly, the equation of the vector n = [n0 , n1 , n2 ,…nM−1 ]T is linearized in the set of M linear differential equations, with T denoting “transpose.” Furthermore, we subject the linearized equation of n to Laplace transform by setting ̃(s) and 𝜙̃ ∝ exp[ist]with i2 = −1 to Laplace transform of n and 𝜙, respectively, and n obtain a set of algebraic equations expressed in the M × M matrix form: ̃ , (sI − A − W − 𝛷P)̃ n(s) = 𝜙PN

(6.29)

where I is an M × M identity matrix; N = [N 0 , N 1 , N 2 , …N M − 1 ]T ; and A, W , and P correspond, respectively, to the radiative and nonradiative relaxation terms, the energy-transfer (ET) term involved with ETU and/or CRP, and the photoexcitation term (PE) consisting of GSA and ESA parameters (for details, see Reference [117]). Hence, we by multiplying Eq. (6.29) by the inverse matrix (sI − A − W − 𝛷P)−1 : can figure out ñ𝜙(s) ̃ ̃(s) n = (sI − A − W − 𝛷P)−1 PN 𝜙̃ Thereby, we have the transfer function

̃ ni (s) 𝜙̃

(6.30) for the modulated UCPL at the i-th level, and ̃ n (s)

the QFRS signal is obtained from its imaginary part –Im[ i𝜙̃ ] by setting s = i𝜔. Now, we shall limit ourselves to the typical three-level model of the UCPL GSA/(ESA + ETU) of Er3+ ions, as shown in Figure 6.24, where level 0 corresponds to the ground state Er3+ : 4 I 15/2 ; level 1, to the intermediate excited state 4 I 11/2 ; and level 2, to the uppermost excited states of three-bundled 4 F 7/2 , 2 H 11/2 /4 S3/2 manifolds [117, 118], and the matrices A, W , and P are given by: ⎡0 k1 ⎤ ⎡−𝜎0 0 0⎤ ⎡0 2wN 1 0⎤ bk 2 ⎥ ⎥ ⎢ ⎥ ⎢ ⎢ A = ⎢0 −k1 (1 − b)k2 ⎥ , W = ⎢0 −4wN 1 0⎥ , P = ⎢ 𝜎0 −𝜎1 0⎥ ⎢0 0 ⎥ ⎢ 0 𝜎 0⎥ ⎢0 2wN 0⎥ −k2 ⎦ ⎦ ⎣ ⎦ ⎣ ⎣ 1 1

(6.31)

with the relaxation rate k 1 (s−1 ) from level 1 to 0; the relaxation rate at level 2, k 2 (s−1 ), with the branching ratio b (e.g. bk 2 is the relaxation rate from level 2 to 0, as shown

193

6 Photoluminescence

0

4I

ETU

15/2

bk2

(1–b)k2

w

Figure 6.24 Schematic of three-level model for green UCPL Er3+ (2 H11/2 / 4 S3/2 → 4 I15/2 ) of mixed GSA/(ESA + ETU) at 975 nm excitation. 0: ground state (4 I15/2 ). 1: intermediate state (4 I11/2 ). 2: uppermost state (three bundled 4 F 7/2 , 2 H11/2 / 4 S3/2 ). The kinetic parameters k1 , k2 , b, w, 𝜎 0 , and 𝜎 1 are described in text (color online).

N1 + n1

k1

4I 11/2

ESA

1

N2 + n2

GSA

4F 7/2 2H 11/2 4S 3/2

σ1 (Φ + ϕ)

2

σ0(Φ + ϕ)

194

N0 + n0

in Figure 6.24); the ETU parameter, w (s−1 ); and the GSA and ESA cross-sections, 𝜎 0 ̃ n (s) and 𝜎 1 (cm2 ), respectively. Thereby, the transfer function of UCPL 2𝜙̃ is expressed as a linear combination of two fractional functions Ri (𝛷)/[s + Ri (𝛷)] and coefficients ai (𝛷) (i = 1, 2) [117, 118]: ̃ (s + k1 + 𝜎0 𝛷)𝜎1 N1 + 𝜎0 𝜎1 N0 𝛷 + 2wN 1 (𝜎0 N0 + 𝜎1 N1 ) n2 (s) = (s + k + 𝜎0 𝛷 + 𝜎1 𝛷 + 4wN 1 )(s + k2 ) − [(1 − b)k2 − 𝜎0 𝛷](𝜎1 𝛷 + 2wN 1 ) ̃ 𝜙 1 R1 (𝛷) R2 (𝛷) = (6.32) a (𝛷) + a (𝛷), s + R1 (𝛷) 1 s + R2 (𝛷) 2 The fractional function Ri (𝛷)/[s + Ri (𝛷)] is the transfer function of the system having the relaxation rate Ri (𝛷), or the relaxation (lifetime) time 𝜏 i = Ri (𝛷)−1 in Section 6.3.3 (Eq. (6.9)); Ri (𝛷) is obtained by setting the denominator of Eq. (6.32) to 0. Thus, the QFRS on the three-level model produces two relaxation rates R1 (𝛷) at level 1 and R2 (𝛷) at level 2, which become k 1 and k 2 , respectively, as 𝛷 → 0. In contrast to the ordinary TRS [120], 𝜏 i of QFRS depends on 𝛷, and R1 (𝛷) is approximated in the first order of 𝛷 for 𝜎 0 𝛷/k 1 ≪ 1 to be: [ }] { 2w𝜎0 (1 + b) 𝛷 = k1 + ̂ S𝛷, (6.33) R1 (𝛷) ≈ k1 + 𝜎0 + 𝜎1 b + k1 𝜎1 with the slope Sˆ and R2 (𝛷) ≈ k 2 due to k 2 ≫ k 1 [117, 118]. Moreover, the ratio of the QFRS peak 𝛾(𝛷) = a2 (𝛷)/a1 (𝛷) gives ETU contribution against ESA expressed by the 2w𝜎 ratio k 𝜎 0 [117–119]. As 𝛾(𝛷) is a complicated function of 𝛷, only 𝛾(0) is given here as: 1 1

( )( ) 2w𝜎0 −1 k1 k1 γ(0) = 1 − − . 1+ k2 k1 σ1 k2

(6.34)

We have applied QFRS on the green UCPL of Er-doped Ge28.1 Ga6.3 S65.6 chalcogenides glass (ChGs) while systematically varying Er-doping from 0.01 to 0.5 at% as well as 975 nm excitation power P∝𝛷 (for experiment details, see References [117, 118]).

6.5 QFRS on Up-Conversion Photoluminescence (UCPL) of RE-Doped Materials

(a)

Normalized QFRS Signal (a.u.)

P (mW)

(b)

P (mW)

42.5

42.5

42.5

15.3

15.3

15.3

11.3

11.3

11.3

9.8

9.8

9.8

7.8

7.8

7.8

5.3

5.3

5.3

2.3

τ2

τ1

10–6 10–5 10–4 10–3 10–2

2.3

τ2

τ1

10–6 10–5 10–4 10–3 10–2 Lifetime τ (s)

(c)

P (mW)

2.3

τ2

τ1

10–6 10–5 10–4 10–3 10–2

Figure 6.25 Pump power (975 nm laser) P-evolved QFRS spectra of green (2 H11/2 /4 S3/2 → 4 I15/2 , ∼550 nm) UCPL in (a) 0.02, (b) 0.1, and (c) 0.5 at% Er-doped GeGaS ChGs. Red curves are best-fit curves of spectra for P = 2.3 mW by nonlinear regression of Eq. (6.32), with red bars indicating the corresponding magnitudes a1 (𝛷) and a2 (𝛷). The dotted lines are guides for eyes of P-evolved lifetimes 𝜏 1 and 𝜏 2 (color online). Source: Reproduced from L. Strizik, V. Prokop, J. Hrabovsky, T. Wagner, and T. Aoki, J. Mater. Sci: Mater. Electron. 28, 7053 (2017) by permission of Springer.

Figure 6.25 demonstrates the P-evolved QFRS spectra of UCPL of the samples for the excitation power P = 2.3–42.5 mW (roughly corresponding to the photon-flux density: 𝛷 ≈ 2.3 × 1021 –4.25 × 1022 cm−2 s−1 ), exhibiting a salient feature of two lifetime components of 𝜏 1 and 𝜏 2 . An increase of P shortens the long lifetime 𝜏 1 from 1.1 to 0.6 ms, in contrast to the short lifetime 𝜏 2 —remaining fairly constant at several tens of μs. The red curves are obtained by the nonlinear regression for the imaginary part of Eq. (6.32) with s = i𝜔 as a model function with unknown parameters k 1 , k 2 , and w to fit the QFRS spectra at the lowest P (2.3 mW) [117]. Thereby, the red bars indicate the amplitudes a1 (𝛷) and a2 (𝛷) corresponding to the 𝜏 1 and 𝜏 2 components, respectively; clearly, the ratio 𝛾(0) decreases as Er concentration N Er increases, predicting an increase of w according to Eq. (6.34) [117, 118]. As P increases, however, the model function [Eq. (6.32)] does not give a good fit to the QFRS spectra. Hence, we have introduced the lifetime distributions of two Gaussians centered around 𝜏 1 and 𝜏 2 at the higher P by the convolution techniques mentioned in Section 6.3.6 [118]. Figure 6.26a shows R1 (𝛷) vs. 𝛷 giving k 1 ≈ 970 s−1 in average. However, similarly obtained k 2 varies almost linearly from ∼1.5 × 104 to ∼2.3 × 104 s−1 with increasing N Er , which is attributed to the ET from the uppermost three-bundled states 4 F 7/2 , 2 H 11/2 /4 S3/2 to the absorption edge of the ChGs host [117, 119]. The GSA cross-section 𝜎 0 was estimated in two ways: 𝜎 0 = 7.2 × 10−21 cm2 by Judd-Ofelt (JO) analysis (see Chapters 2 and 4), and 𝜎 0 = 5.9 × 10−21 cm2 by the absorption spectra (GSA). The values of 𝜎 1 /𝜎 0 = 1.4 and b = 0.7 are determined from the JO analysis [117].

195

Relaxation rate R1(Φ ) = τ1–1(×103 s–1)

6 Photoluminescence

1.7 1.6 1.5 1.4 1.3 1.2 1.1

0.5 % Er 0.1 % Er 0.02 % Er

1.0 0.9 0

1

2

3

4

5

Photon-flux density Φ (×1022cm–2 s–1) (a) 14 Macroscopic ETU parameter WEr (×10–18cm3 s–1)

196

12

Slope R1 (Φ)JO

Ratio γ (0)JO

Slope R1 (Φ)Exp.

Coleman et al.

10 8 6 4 2 1019

1020

Er concentration NEr

(cm–3)

(b)

Figure 6.26 (a) Relaxation rate R1 (𝛷) at level 1 (4 I11/2 ) vs. excitation photon-flux density 𝛷 of 975 nm laser. (⧫) for 0.02, (◽) 0.1, and (⚬) 0.5 at% Er-doped GeGaS ChGs. The lines were obtained by linear fits of Eq. (6.33) to data at low 𝛷. (b) Macroscopic ETU parameter W Er = w/NEr vs. Er concentration NEr (cm−3 ) of the green (∼550 nm) UCPL. (⚬) obtained from slope Sˆ of the R1 (𝛷) vs. 𝛷 at low 𝛷 with 𝜎 0 = 7.2 × 10−21 cm2 and 𝜎 1 /𝜎 0 = 1.4 (JO analysis), (◊) with 𝜎 0 = 5.9 × 10−21 cm2 (GSA) and 𝜎 1 /𝜎 0 = 1.4, (•) from peak ratio 𝛾(0) at P = 2.3 mW and (◽) from Reference [116]. Source: Reproduced from L. Strizik, V. Prokop, J. Hrabovsky, T. Wagner, and T. Aoki, J. Mater. Sci: Mater. Electron. 28, 7053 (2017) by permission of Springer.

Figure 6.26b shows the macroscopic ETU parameter W Er = w/N Er (cm3 s−1 ) plotted as a function of N Er . Here, the symbols ⚬ and ◊ denote W Er obtained from Sˆ in Eq. (6.33) with 𝜎 0 = 7.2 × 10−21 cm2 (JO analysis) and with 𝜎 0 = 5.9 × 10−21 cm2 (GSA), respectively, and the symbol • denotes W Er derived from the ratio 𝛾(0) of Eq. (6.34). The plots ⚬ and • are similar to those of Coleman et al (◽) by TRS in GeGaS:Er ChGs [116], whereas those of ◊ deviate considerably from others. Probably, the error in w∝(Sˆ − 𝜎 0 − b𝜎 1 ) is exaggerated when Sˆ approaches 𝜎 0 + b𝜎 1 , requiring more accurate determination of 𝜎 1

6.6 Conclusions

and b. Furthermore, the differences among the data of W Er are partly addressed to the determination of N Er , which was estimated from volumetric mass density of the glass under the assumption of the homogeneous distribution and full ionization of Er3+ in the GeGaS matrix. Therefore, N Er is not always identical to N ion (cm−3 ), and, moreover, inaccurate N Er causes noticeable error in determining 𝜎 0 = 5.9 × 10−21 cm2 from the GSA spectra. Nevertheless, the obtained w predicts that ETU dominates ESA only for the 2w𝜎 0.5% Er-doped sample by the criterion ETU ≷ ESA according to the ratio k 𝜎 0 ≷ 1 [117]. 1 1 The UCPL of Er3+ ions excited at 975 nm emits red light (4 F 9/2 → 4 I 15/2 , ∼660 nm) as well as green light. We have made QFRS on the red UCPL of the preceding samples and analyzed the data by a 5 × 5 matrix equation on the five-level model (M = 5) by adding two 4 I 13/2 and 4 I 9/2 manifolds, which demonstrates the versatility of Eq. (6.30). For details, please see Reference [119].

6.6 Conclusions The PL of condensed matter is reviewed, and its state of the art is described as far as possible. Though QFRS is a counterpart in the frequency-domain against TRS in the time-domain and a powerful tool to investigate widespread PL lifetime distribution, it is not so prevalent. Thus, this chapter mainly focuses on the QFRS on PL lifetime distribution: the broad lifetime distribution of amorphous semiconductors, in particular, a-Si:H; and the narrow but multi-lifetime distribution of RE-doped ChGs. The wideband QFRS technique has revealed the triple-peak lifetime distribution in PL of a-Si:H as well as a-Ge:H at low temperatures T (∼3.7 K) and low generation rates G ( Eh , but gets localized in the tail states for Ev < Eh . ̂ xp (Eq. (7.3)) can be written Using Eqs. (7.2), (7.4), and (7.6), the interaction operator H in the second quantized form as: ( )1∕2 e ∑ ℏ ̂ Hxp = − Qcv c+𝜆 (7.8) 𝜇x 𝜆 2𝜀0 n2 V 𝜔𝜆

207

208

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

where Qcv = N −1



exp[−ite ⋅ Rel ] exp[−i th ⋅ Rhm ]Zlm𝜆 Bcvlm (S = 0)

(7.9)

l,m

where Bcvlm (S = 0) =

∑ 𝜎e ,𝜎h

acl (𝜎e )dvm (𝜎h )𝛿𝜎e ,−𝜎h is the annihilation operator of a singlet

exciton of spin S = 0 by annihilating an electron in the conduction c states at site l and a hole in the valence v states at site m, and the summations over 𝜎 e and 𝜎 h represent combinations of spins to form singlet and triplet excitons [5]. Zlm𝜆 is given by: Zlm𝜆 =



𝜙∗l (re )𝜀̂𝜆 ⋅ p𝜙m dre drh

(7.10)

It may be noted that it is not necessary to use the exciton operator Bcvlm (S = 0) in ∑ Eq. (7.9); one can also use the product of fermion operators 𝜎e ,𝜎h acl (𝜎e )dvm (𝜎h )𝛿𝜎e ,−𝜎h instead, provided the condition of singlet spin state 𝛿𝜎e ,−𝜎h is maintained. This is because the integral in Eq. (7.10) is non-zero only if the spin configuration of the photoexcited electron and hole is in singlet state; in other words, it is non-zero only for singlet excitons. We now consider a transition from an initial state with one singlet exciton created by exciting an electron at a site, say, l, and hole at site m, and where any other site is assumed to have zero excitons to a final state with one created photons and no excitons. We also assume that there are no photons in the initial state. The transition matrix element is then obtained as [19]: ( )1∕2 ∑ ℏ ̂ xp ∣ i⟩ = − e ⟨f ∣ H pcv (7.11) 𝜇x 𝜆 2𝜀0 n2 V 𝜔𝜆 where pcv = N −1



exp[−ite ⋅ Rel ] exp[−ith ⋅ Rhm ]Zlm𝜆 𝛿l,m

(7.12)

l,m

Following Eq. (7.10), the transition matrix element in Eq. (7.11) is also non-zero only for singlet excitons and vanishes for triplet excitons. Also, the same transition matrix element is obtained even if one considers the electron and hole field operators in Eqs. (7.4) and (7.6), respectively, to be spin independent. That means the transition matrix element of the radiative recombination of a singlet exciton or an excited free electron–hole pair remains the same as given in Eq. (7.11), but that it is zero for a triplet spin configuration. Therefore, the formalism developed here onward for calculating the rates of spontaneous emission is applicable to singlet excitons, type I and singlet type II geminate pairs, and nongeminate pairs. Let us first derive Zlm𝜆 . As stated earlier in Chapter 3, there are two approaches, which are used in amorphous solids, to evaluate this integral. The integral actually determines the average value of the relative momentum between the excited electron–hole pair. In the first approach, it is assumed to be a constant and independent of the photon energy as [1, 5]: ( )1∕2 L Zlm𝜆 = 𝜙∗l (re )𝜀̂𝜆 ⋅ p𝜙m dre drh = Z1 = 𝜋h (7.13) ∫ V

7.2 Photoluminescence

where L is the average bond length in a sample and Z1 denotes the matrix element obtained from the first approach. This form was first introduced by Mott and Davis [1] and has been used widely since then. In the second approach, the integral is evaluated using the dipole approximation as [5, 33–35]: Zlm𝜆 =



𝜙∗l (re )𝜀̂𝜆 ⋅ p𝜙m dre drh = Z2 = i𝜔𝜇x |re−h |

(7.14)

where Z2 denotes the transition matrix element obtained from approach 2; ℏ𝜔 = E′ c − E′ v is the emitted photon energy; and |re−h | = ⟨l|𝜀̂𝜆 ⋅ r||m⟩ is the average separation between the excited electron–hole pair, which also can be assumed to be site independent. Thus, in both approaches, the integral becomes site independent and can be taken out of the summation in Eq. (7.11). In the case of singlet excitons and type II geminate pairs, it can be easily assumed that |re − h | = aex , the excitonic Bohr radius, but for the case of type I geminate and nongeminate pairs, |re − h | is the average separation between e and h. It may be noted here that the value of the integral obtained from the first approach, Z1 , leads to the well-known Tauc’s relation for the absorption coefficient observed in amorphous semiconductors, and that obtained from the second approach is used to explain the deviations from Tauc’s relation in the absorption coefficient observed in some amorphous semiconductors [1, 33, 34] (also see Chapter 3 of this volume). Using Eqs. (7.13) and (7.14), pcv in Eq. (7.12) can be written as: ∑ pcv = Zi N −1 exp[−ite ⋅ Rel ] exp[−ith ⋅ Rhm ]𝛿l,m , i = 1, 2 (7.15) l,m

Now the derivation of pcv in Eq. (7.15) for non-crystalline solids depends on four possibilities, which are considered separately in the following text. (i) Extended-to-extended states transitions Rearranging the exponents in Eq. (7.15) as: exp[−ite ⋅ Rel − ith ⋅ Rhm ] = exp[−ite ⋅ (Rel − Rhm ) − i(te ⋅ th ) ⋅ Rhm ]

(7.16)

and identifying the fact that Rel − Rhl = aex , the excitonic Bohr radius (the separation between e and h in an exciton prior to their recombination, which is assumed to be site independent), the first exponential becomes site independent. It can be taken out of the summation signs, and then the matrix element in Eq. (7.15) becomes: pcv = Zi N −1 exp[−ite ⋅ aex ]𝛿te ,−th , i = 1, 2

(7.17)

where 𝛿te ,−th represents the momentum conservation in the transition. The square of the transition matrix element then becomes: |pcv |2 = |Zi∗ Zi | = |Zi |2 , i = 1, 2

(7.18)

(ii) Transitions from extended to tail states Here, we consider the electron e excited in the extended state and the hole h in the tail state. For this case, Eq. (7.15) can be written as: ∑ exp[−ite ⋅ Rel ] exp[−th′ ⋅ Rhm ]𝛿l,m , i = 1, 2 (7.19) pcv = Zi N −1 l,m

209

210

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

where ∣ th′ ∣= th′ =

√ 2m∗h (Eh − Ev )∕ℏ

(7.20)

Rewriting Eq. (7.19) as: pcv = Zi N −1



exp[−ite ⋅ (Rel − Rhm )] exp[−i(te − ith′ ) ⋅ Rhm ]𝛿l,m , i = 1, 2

l,m

(7.21) and simplifying it in the same way as Eq. (7.16), we get: pcv = Zi exp[−ite ⋅ aex ]𝛿te ,ith , i = 1, 2

(7.22)

Equation (7.16) gives the same expression for |pcv |2 as in Eq. (7.18) for the possibility (i), and in this case the momenta also remain conserved. (iii) Transitions from tail to extended states We consider an electron e excited in the tail and the hole h in the extended states. For this case, the transition matrix element can be derived in a way analogous to that for the extended-to-tail states. pcv is obtained as: pcv = Zi exp[−ith ⋅ aex ]𝛿ite′ ,th , i = 1, 2 where ∣ te′ ∣= te′ =

√ 2m∗e (Ec − Ee )∕ℏ

(7.23)

(7.24)

Here, again, |pcv |2 remains the same as in Eq. (7.18). (iv) Transitions from tail-to-tail states Here, we consider that both electron e and hole h are excited in the tail states. In this case, an exciton loses its usual excitonic character and behaves like a type II geminate pair, as explained earlier. This is because the localized form of wave functions of e and h does not give rise to an exciton. For this case, using the localized form of the electron and hole wave functions, pcv is obtained as: ∑ pcv = Zi N −1 exp[−te′ ⋅ Rel ] exp[−th′ ⋅ Rhm ]𝛿l,m , i = 1, 2 (7.25) l,m

Here again, one can rearrange the exponential as: ∑ pcv = Zi N −1 exp[−te′ ⋅ (Rel − Rhm )] exp[−(te′ + th′ ) ⋅ Rhm ]𝛿l,m , i = 1, 2

(7.26)

l,m

which becomes: pcv = Zi exp[−te′ ⋅ aex ]𝛿te′ ,−t′ , i = 1, 2 h

(7.27)

Assuming that the relative momentum of the electron, ℏt′ e , will be along the direction of aex at the time of recombination, |pcv |2 from Eq. (7.27) becomes: |pcv |2 = |Zi |2 exp[−2te′ aex ] 𝛿te′ ,−t′

h

(7.28)

7.2 Photoluminescence

7.2.2

Rates of Spontaneous Emission

Using Eq. (7.11) and applying Fermi’s golden rule, the rate Rsp (s−1 ) of spontaneous emission can be written as [19, 36–40]: ( ) ∑ 1 2𝜋e2 |p |2 f f 𝛿(E′ − Ev′ − ℏ𝜔𝜆 ) (7.29) Rsp = 2 2 2𝜀0 n V 𝜔𝜆 E′ ,E′ cv c v c 𝜇x c

v

where f c and f v are the probabilities of occupation of an electron in the conduction state and a hole in the valence state, respectively. The rate of spontaneous emission in Eq. (7.29) can be evaluated under several conditions, which will be described in the following text. (i) Rate of spontaneous emission under two-level approximation Here, only two energy levels are considered as in atomic systems. An electron takes a downward transition from an excited state to the ground state; no energy bands are involved. In this case, f c = f v = 1, and then, denoting the corresponding rate of spontaneous emission by Rsp12 , we get: Rsp12 =

𝜋e2 |p12 |2 𝛿(ℏ𝜔 − ℏ𝜔𝜆 ) 𝜀0 n2 𝜔𝜇x2

(7.30)

where ℏ𝜔 = E2 − E1 , which is the energy difference between the excited (E2 ) and ground (E1 ) states. In this case, p12 is derived using the dipole approximation (Eq. (7.14)) as: p12 = i𝜔𝜇x ⟨𝜀̂𝜆 .r⟩

(7.31)

where r is the dipole length, and ⟨…⟩ denotes integration over all photon modes l. Substituting Eq. (7.31) into Eq. (7.30), and then integrating over the photon wave vector k using 𝜔 = kc/n, we get [8, 37]: √ 4𝜅e2 𝜀𝜔3 |re−h |2 (7.32) Rsp12 = 3ℏc3 where 𝜀 = n2 is the static dielectric constant, ∣re − h ∣ is the mean separation between the electron and hole, and 𝜅 = 1/(4𝜋𝜀0 ). It may be noted that, for deriving the rate of spontaneous emission within the two-level approximation, pcv with Zlm𝜆 in the form of Eq. (7.13) has not been used to the author’s knowledge. The expression of the rate of spontaneous emission obtained in Eq. (7.32) is well known [8, 11, 36], and it is independent of the electron and hole masses and temperature. As only two discrete energy levels are considered, the density of states is not used in the derivation. Therefore, although by using |re − h | = aex , the excitonic Bohr radius, the rate Rsp12 has been applied to a-Si:H [8, 11], it should only be used for calculating the radiative recombination in isolated atoms, not in condensed matter. (ii) Rates of spontaneous emission in amorphous solids In applying Eq. (7.29) for any condensed-matter system, it is necessary to determine f c and f v and the density of states. On the one hand, it may be argued that short-time PL can occur before the system reaches thermal equilibrium, and that therefore no equilibrium distribution functions can be used for the excited charge carriers. In this case, one should use f c = f v = 1 in Eq. (7.29) for condensed matter as well. On the other hand, as the carriers are excited by the same energy photons, even in a

211

212

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

short time delay, they may be expected to reach thermal equilibrium among themselves, but not necessarily with the lattice. Therefore, they will relax according to an equilibrium distribution. As the electronic states of amorphous solids include the localized tail states, it is more appropriate to use the Maxwell–Boltzmann distribution for this situation. We will consider here radiative recombination under both nonequilibrium and equilibrium conditions. 7.2.2.1

At Nonthermal Equilibrium

Considering f c = f v = 1 and substituting |pcv |2 (Eq. (7.12)) derived above from the two approaches in Eq. (7.29), we get the rates, Rspni , for the possibilities (i)–(iii) as: ( ) ∑ 2𝜋e2 1 Rspni = 2 𝛿(Ec′ − Ev′ − ℏ𝜔𝜆 ), i = 1, 2 (7.33) |Zi |2 2 2𝜀0 n V 𝜔𝜆 𝜇x E′ ,E′ c

v

where the subscript spn in Rspni denotes rates of spontaneous emission derived at the non-equilibrium. For evaluating the summation over Ec′ and Ev′ in Eq. (7.33), the usual approach is to convert it into an integral by using the excitonic density of states, which can be obtained as follows. Using the effective mass approximation, the electron energy in the conduction band and hole energy in the valence band can be, respectively, written as: p2 (7.34) Ec′ = Ec + e ∗ 2me and Ev′

= Ev −

p2h

(7.35)

2m∗h

Subtracting Eq. (7.35) from Eq. (7.34) and then applying the center of mass and relative coordinate transformations (see Eq. (7.3)), we get the excitonic energy Ex in the parabolic form as: p2 P2 (7.36) + Ex = Eo + 2M 2𝜇x where E0 = Ec − Ev is the optical gap; P and M = m*e + m*h are linear momentum and mass of an exciton associated with its center of mass motion, respectively; and the last term is the kinetic energy of the relative motion between e and h with linear momentum p, which contributes to the exciton binding energy through the attractive Coulomb interaction potential between them [41]. The exciton density of states g x then comes from the parabolic form of the second term associated with the center of mass motion as: ( ) V 2M 3∕2 (Ex − E0 )1∕2 (7.37) gx = 2𝜋 2 ℏ2 Using Eq. (7.37), the summation in Eq. (7.33) can be converted into an integral as: ∑ 𝛿(Ex′ − ℏ𝜔𝜆 ) = Ij = gx 𝛿(Ex′ − ℏ𝜔𝜆 )dE′x (7.38) ∫ ′ E E′ x

x

which gives Ij =

( ) V 2M 3∕2 (ℏ𝜔𝜆 − E0 )1∕2 2𝜋 2 ℏ2

(7.39)

7.2 Photoluminescence

Substituting Eq. (7.39) into Eq. (7.33) we get: ( ) 2𝜋e2 1 j |Zi |2 Ij Θ(ℏ𝜔𝜆 − E0 ), Rspni = 2 2𝜀0 n2 V 𝜔𝜆 𝜇x

i = 1, 2

(7.40)

where Θ(ℏ𝜔𝜆 − E0 ) is a step function used to indicate that there is no radiative recombination for ℏ𝜔𝜆 < E0 , and j denotes that these rates are derived through the joint density of states. Having derived the rates of spontaneous emission using the excitonic density of states, it is important to remember that the use of such joint densities of states for amorphous semiconductors does not give the well-known Tauc’s relation [1, 5] in the absorption coefficient [34]. Therefore, by using it in calculating the rate of spontaneous emission, one would violate the Van Roosbroeck and Shockley relation [42] between the absorption and emission. For this reason, the product of individual electron and hole densities of states is used in evaluating the summation in Eq. (7.27) for amorphous semiconductors. This approach has proven to be very useful in amorphous solids as it gives the correct Tauc’s relation [1, 5, 33, 43]. Denoting the summations over E′ c and E′ v in Eq. (7.33) by I and using the product of individual densities of states, I can be evaluated by converting the summations into integrals as: Ev +ℏ𝜔𝜆

I=

∫Ec

Ev

∫E′ −Ev

gc (Ec′ )gv (Ev′ )𝛿(Ec′ − Ev′ − ℏ𝜔)dE′c dE′v

(7.41)

c

where g c (E′ c ) and g v (E′ v ) are the densities of states of the conduction and valence states, respectively, and these can be written within the effective-mass approximation as: ( ) V 2m∗ 3∕2 1∕2 Eq , q = c, v (7.42) gq (E′ ) = 2𝜋 2 ℏ2 where m* is the effective mass of the corresponding charge carrier, and q = c (conduction) and q = v (valence) states. Substituting Eq. (7.42) into Eq. (7.41), the integral can be evaluated analytically to give [5]: V 2 (m∗e m∗h )3∕2

(ℏ𝜔 − E0 )2 (7.43) 4𝜋 3 ℏ6 Using Eqs. (7.39) and (7.43) in Eq. (7.33) and substituting the corresponding Zi, we get the two rates in nonthermal equilibrium for possibilities (i)–(iii) as: I=

Rspn1 =

e2 L(m∗e m∗h )3∕2 4𝜀0 ℏ3 n2 𝜇x2 (ℏ𝜔)

(ℏ𝜔 − E0 )2 Θ(ℏ𝜔𝜆 − E0 )

(7.44)

and Rspn2 =

(m∗e m∗h )3∕2 e2 |re−h |2 2𝜋 2 𝜀0 n2 ℏ7

V ℏ𝜔(ℏ𝜔 − E0 )2 Θ(ℏ𝜔𝜆 − E0 )

(7.45)

For excitonic transitions, it is more appropriate to replace m∗e and m∗h in Eqs. (7.44) and (7.45) by the excitonic reduced mass 𝜇x and use ∣re − h ∣ = aex . It may also be noted that the volume V is appearing in the formula for Rspn2 (Eq. (7.45)). This is inevitable through the second approach, and it has been tackled earlier by Cody [35] by defining 2N 0 /V = 𝜈𝜚A , where N 0 is the number of single spin states in the valence band, and thus 2N 0 becomes the total number of valence electrons occupying N 0 states, 𝜈 is the

213

214

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

number of coordinating valence electrons per atom, and 𝜌A is the atomic density per unit volume. Thus, replacing V by V/N 0 in Eq. (7.45), one can get around this problem. We thus obtain: e2 L𝜇x Rspn1 = (7.46) (ℏ𝜔 − E0 )2 Θ(ℏ𝜔𝜆 − E0 ) 4𝜀0 ℏ3 n2 𝜇x2 (ℏ𝜔) and Rspn2 =

𝜇x3 e2 a2ex ℏ𝜔(ℏ𝜔 − E0 )2 Θ(ℏ𝜔𝜆 − E0 ) 2𝜋 2 𝜀0 n2 ℏ7 𝜈𝜌A

(7.47)

It may be emphasized here that E0 , defined as the energy of the optical gap, is not always the same in amorphous solids. It depends on the lowest-energy state within the conduction band from where the radiative recombination occurs. As the rates derived in Eqs. (7.46) and (7.47) do not have any peak value, it is not possible to determine E0 from these rates. Likewise, using |pcv |2 (Eq. (7.28)) in Eq. (7.33) for possibility (iv), we get the rates of spontaneous emission from the two approaches for the tail-to-tail states transitions as: Rspnti = Rspni exp(−2te′ ∣ re−h ∣), i = 1, 2

(7.48)

where the subscript spnt of Rspnt stands for the spontaneous emission at nonequilibrium from tail-to-tail states. re−h is the average separation between e and h; for excitons, re−h = aex , where aex is the excitonic Bohr radius in the tail states, and for a singlet exciton it is given by [6, 10]: aex =

5𝜇𝜺 a 4𝜇x 0

(7.49)

where m is the reduced mass of an electron in the hydrogen atom, and a0 = 0.529 Å is the Bohr radius. Results for the rates of spontaneous emission obtained in the preceding text are valid in the nonthermal equilibrium condition, as stated earlier. However, unless one knows the relevant effective masses of charge carriers and the value of E0 , these rates cannot be applied to determine the lifetime of PL. The effective mass of charge carriers in their extended and tail states can be determined [5], as also described in Chapter 3 of this volume, but not E0 . For this reason, it is useful first to derive these rates under thermal equilibrium as well, as shown in the following text. 7.2.2.2

At Thermal Equilibrium

Assuming that the excited charge carriers are in thermal equilibrium among themselves, the distribution functions, f c and f v , can be given by the Maxwell–Boltzmann distribution function, as given in reference [44]: [ ] (E − EFn ) fc = exp − e (7.50) 𝜅B T and

[ ] (EFp − Eh ) fv = exp − 𝜅B T

(7.51)

where Ee and Eh are the energies of an electron in the conduction and a hole in the valence state, respectively, and EFn and EFp are the corresponding Fermi energies. 𝜅 B is

7.2 Photoluminescence

the Boltzmann constant and T temperature of the excited charge carriers. The product f c f v is then obtained as: [ ] (ℏ𝜔 − E0 ) (7.52) fc fv ≈ exp − 𝜅B T where Ee − Eh = ℏ𝜔 and EFn − EFp = E0 are used. It may be noted that Eq. (7.52) is also an approximate form of the Fermi–Dirac distribution obtained for Ee > EFn and Eh < EFp , and it has been used widely [36, 38–43] for calculating the rate of spontaneous emission in semiconductors. Substituting Eq. (7.52) in the rate in Eq. (7.29), the integral in Eq. (7.41) is obtained as: [ ] V (m∗e m∗h )3∕2 (ℏ𝜔 − E0 ) 2 I= (ℏ𝜔 − E ) exp − (7.53) 0 𝜅B T 4𝜋 3 ℏ6 Using Eq. (7.53), the rates of spontaneous emission at thermal equilibrium for possibilities (i)–(iii) are obtained as: [ ] (ℏ𝜔 − E0 ) , i = 1, 2 (7.54) Rspi = Rspni exp − 𝜅B T And for the possibility (iv), we get: [ ] (ℏ𝜔 − E0 ) , Rspnti = Rspni exp − 𝜅B T

i = 1, 2

(7.55)

Rates in Eqs. (7.54) and (7.55) have a maximum value, which can be used to determine E0 , as described in the following text. 7.2.2.3

Determining E0

For determining E0 , we assume that the peak of the observed PL intensity occurs at the same energy as that of the rate of spontaneous emission obtained in Eqs. (7.54) and (7.55). The PL intensity as a function of ℏ𝜔 has been measured in a-Si:H. From these measurements, the photon energy corresponding to the PL peak maximum can be determined. By comparing the experimental energy thus obtained with the energy corresponding to the maximum of the rate of spontaneous emission, we can determine E0 . For this purpose, we need to determine the energy at which the rates in Eqs. (7.54) and (7.55) become maximum. Defining x = ℏ𝜔 − E0 , (x > 0) and 𝛽 = 1/𝜅 B T and then setting dRspi /dx = 0 (i = 1, 2), we get x01 and x02 from Eq. (7.54) and Eq. (7.55), respectively, at which the rates are maximum, as [19]: √ 1 − 𝛽E0 (1 − 𝛽E0 )2 2E0 + ± (7.56) x01 = 2𝛽 4𝛽 2 𝛽 and x01

3 − 𝛽E0 = ± 2𝛽



(3 − 𝛽E0 )2 2E0 + 4𝛽 2 𝛽

(7.57)

where only the + sign produces x > 0. It may be noted that both Eqs. (7.54) and (7.55) give the same expression for x01 and x02 as derived in Eqs. (7.56) and (7.57). Using

215

216

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

x0 = Emx − E0 , where Emx = ℏ𝜔mx is the emission energy at which the PL peak intensity is observed, the corresponding E0 value is obtained from Eqs. (7.56) and (7.57) as: E01 =

Emx (−1 + 𝛽Emx ) 1 + 𝛽Emx

(7.58)

E01 =

Emx (−3 + 𝛽Emx ) 1 + 𝛽Emx

(7.59)

and

Using this in Eqs. (7.54) and (7.55), one can calculate the rates of spontaneous emission from the two approaches for all four possibilities. The time of radiative recombination or radiative lifetime, 𝜏 ri , is then obtained from the inverse of the maximum rate (𝜏 ri = 1/Rspi , I = 1, 2) obtained from Eqs. (7.54) and (7.55), calculated at ℏ𝜔mx = Emx . 7.2.3

Results of Spontaneous Emission and Radiative Lifetime

Applying the preceding theory, rates of spontaneous emission are calculated in this section for a-Si:H, (i) within two-level approximation, (ii) under nonequilibrium with the joint and product density of states, and (iii) under equilibrium with the product density of states. Although the rate derived within the two-level approximation is strictly valid only for isolated atomic systems, for applying it to a condensed-matter system, one requires 𝜔 and |re − h |. For applying it to the excitonic radiative recombination, one may be able to assume |re − h | = aex , but it is not clear at what value of 𝜔 one should calculate the rate to determine the radiative lifetime. For this reason, the rate Rsp12 in Eq. (7.32) derived within the two-level approximation will be calculated later at the end of this section. Rates in Eqs. (7.46)–(7.48), derived under nonthermal equilibrium with the product density of states, cannot be used to calculate the radiative lifetime unless one knows the value of E0 , which cannot be determined from these expressions. Therefore, here we can only present the results for the rates of spontaneous emission derived under the thermal equilibrium in Eqs. (7.54) and (7.55). We will use Eq. (7.54) to calculate the rates, Rspi , I = 1, 2, for possibilities (i)–(iii), and Eq. (7.55) to calculate Rspt1 and Rspt2 for possibility (iv) in a-Si:H. As the non-radiative relaxation is very fast (in ps), only excitonic recombinations can be expected to contribute to PL from possibilities (i)–(iii). All free excited carriers (type I and II geminate pairs and nongeminate pairs) are expected to have relaxed to their tail states before recombining radiatively, and hence they can contribute to PL only through possibility (iv). For calculating the effective masses of charge carriers, we consider a sample of a-Si:H with 1 at% weak bonds contributing to the tail states (i.e. a = 0.99 and b = 0.01), using L = 0.235 nm [4], E2 = 3.6 eV [45, 46], Ec = 1.80 eV [2], and Ec − Ect = 0.8 eV [47]. Using these in Eqs. (3.17)–(3.21), we arrive at m∗ex = 0.34 me and m∗et = 7.1 me , respectively, for a-Si:H. For determining E0 from Eqs. (7.58) and (7.59), we need to know the values of Emx and the carrier (exciton) temperature T. Wilson et al. [8] have measured PL intensity as a function of the emission energy for three different samples of a-Si:H at 15 K and at two different decay times of 500 ps and 2.5 ns. Stearns [9] and Aoki [12, 49, 50] have also measured it at 20 K and 3.7 K, respectively. Emx estimated from these spectra is given in Table 7.1.

7.2 Photoluminescence

Table 7.1 Values of E mx estimated from the maximum of the observed PL intensity from three different experiments (Figure 7.2) and the corresponding values of E 0 calculated from Eqs. (7.58) and (7.59) (note that both expressions produce the same value). E mx (eV)

E 0 (eV)

Experiment

T (K)

500 ps

2.5 ns

500 ps

2.5 ns

Sample 1 [11]

15

1.428

1.401

1.425

1.398

Sample 2 [11]

15

1.444

1.400

1.441

1.397

Sample 3 [11]

15

1.448

1.405

1.445

1.402

Sample 4 [9]

20

1.450

1.447

Sample 5 [50]

3.7

1.360

1.359

The next problem is to find the temperature of the excited carriers before they recombine radiatively. This can also be done on the basis of the three experimental results on PL [9, 11, 48–50]. The measured energy, Emx, of maximum PL intensity is below the mobility edge Ec by 0.4 [8, 11] to 0.44 eV [50]. This means that most excited charge carriers have relaxed down below their mobility edges and are not hot carriers anymore before the radiative recombination. It is therefore only logical to assume that the excited charge carriers in these experiments are in thermal equilibrium with the lattice, and E0 should be calculated at the lattice temperature. The assumption is very consistent with the established fact that the carrier–lattice interaction is much stronger in a-Si:H [5] than in crystalline Si. Using the experimental values of Emx and the corresponding lattice temperature, the values of E01 calculated from Eq. (7.58) and those of E02 calculated from Eq. (7.59) are also listed in Table 7.1. Both the approaches, Eqs. (7.58) and (7.59), produce identical results for E0 . It should be noted, however, that E0 increases slightly with the lattice temperature. Having determined the effective mass and E0 , we can now calculate the rates of spontaneous emission for transitions through possibilities (i)–(iv). (i) Possibility (i)—Both e and h excited in extended states As explained in the preceding text, possibilities (i)–(iii) are only applicable to the singlet excitonic recombinations. In possibility (i), the excitonic reduced mass is obtained as 𝜇x = 0.17 me and the excitonic Bohr radius as 4.67 nm. Using n = 4, the calculated Rsp1 is plotted as a function of the emission energy, and at a temperature T = 15 K for one of the samples of a-Si:H from Wilson et al. [11] in Figure 7.1a and Rsp2 for the same sample in Figure 7.1b. Similar curves are obtained for all the samples at other temperatures as well, but the magnitudes of the rates vary. The radiative lifetime can be obtained from the inverse of the maximum value of the rate at any temperature and E0 . The maximum value of the rates is obtained at the emission energy, Emx , at a given temperature. The rates and the corresponding radiative lifetimes thus calculated are listed in Table 7.2. As can be seen from Table 7.2, the radiative lifetime is found to be in the ns time range at 15 and 20 K and in the ms range at 3.7 K. (ii) Possibilities (ii)–(iii)—Extended-to-tail states recombinations For possibilities (ii) and (iii), where one of the charge carriers of a singlet exciton is in its extended state and the other in its tail state, we get 𝜇x = 0.32me and

217

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

Figure 7.1 Rates of spontaneous emission plotted as a function of the emission energy for E 0 = 1.445 eV at a temperature of 15 K and calculated from (a) method 1 and (b) method 2 for calculating the matrix elements are listed in Table 7.2. As can be seen from Table 7.2, the radiative lifetime is found to be in the ns time range at 15 and 20 K, and in the ms range at 3.7 K.

Rate (108 s–1)

1.50 1.25 1.00 0.75 0.50 0.25 0 1.445

1..45 1.455 1.46 Emission energy (eV)

1.465

(a) 2.00 Rate (108 s–1)

218

1.50 1.00 0.50 0 1.445

1.455 1.46 1..45 Emission energy (eV)

1.465

(b)

aex = 2.5 nm. Using these and the other quantities as in possibility (i), we get from Eq. (7.54) (i = 1), Rsp1 = 2.94 × 108 s−1 , which gives 𝜏 r1 = 3.41 ns at 15 K, 5.22 × 108 s−1 giving 𝜏 r1 = 1.92 ns at 20 K, and at 3.7 K we get Rsp1 = 1.47 × 107 s−1 , which gives 𝜏 r1 = 0.1 μs. Likewise, from Eq. (7.54) (i = 2), we get Rsp2 = 3.45 × 108 s−1 and 𝜏 r2 = 2.9 ns at 15 K, and Rsp2 = 1.08 × 109 s−1 giving 𝜏 r2 = 0.92 ns at 20 K. At 3.7 K, we obtain Rsp2 = 1.62 × 107 s−1 and 𝜏 r2 = 0.1 μs. Thus, the radiative lifetimes for all the three possibilities (i)–(iii) are in the ns time range at 15 and 20 K, but in the μs time range at 3.7 K. It is not possible from these results to determine where exactly the PL is originating from in these experiments—that is, whether it is from the extended or tail states. In order to determine the origin of PL, one also needs to know the Stokes shift in the PL spectra. For example, if the excitonic PL is occurring through radiative transitions from extended-to-extended states, then the observed Stokes shift observed in the PL spectra should only be equal to the exciton binding energy. For this reason, it is important to determine the exciton binding energy corresponding to all the four transition possibilities. The ground-state singlet exciton binding energy Es in a-Si:H is obtained as [5, 10]: Es =

9𝜇x e4 20(4𝜋𝜀0 )2 𝜀2 ℏ2

(7.60)

This gives Es ∼ 16 meV for possibility (i), 47 meV for possibilities (ii) and (iii), and 0.33 eV for possibility (iv). The known optical gap for a-Si:H is 1.8 eV [2], and the

7.2 Photoluminescence

E0 estimated from experiments is about 1.44 eV at 15 K [11], 1.45 eV at 20 K [9], and 1.36 eV at 3.7 K [49] (see Table 7.1). Considering that the PL in a-Si:H originates from excitonic states (possibility (i)), we find a Stokes shift of 0.36 eV, 0.35 eV, and 0.44 eV at temperatures 15 K, 20 K, and 3.7 K, respectively. Such a large Stokes shift is not possible due only to the exciton binding energy, which at most is in the meV range for possibilities (i)–(iii), as given earlier. As the nonradiative relaxation of excitons is very fast, not much PL may occur through possibility (i) even at a time delay of 500 ps measured by Wilson et al. [11] and Stearns [9]. It may possibly occur from transitions only through possibilities (ii)–(iv), which means at least one of the charge carriers has relaxed to the tail states before recombining radiatively. This can be easily explained from the following text: as stated earlier, the charge carrier–lattice interaction is much stronger in a-Si:H than in crystalline Si [5]. As a result, it is well established [4] that an excited hole gets self-trapped very fast in the tail states in a-Si:H. Thus, a Stokes shift of about 0.4 eV observed experimentally at 3.7, 15, and 20 K is due to the relaxation of holes in excitons to the tail states plus the excitonic binding energy, which is in the meV range and hence plays a very insignificant role. The radiative lifetimes of such transitions calculated from the present theory and listed in Table 7.2 fall in the ns time range at temperatures 15–20 K, which agrees very well with those observed by Wilson et al. [8, 11] and Table 7.2 The maximum rates, Rsp1 and Rsp2 , of spontaneous emission and the corresponding radiative lifetime calculated using the values of E mx and E 0 for the five samples in Table 7.1. Sample

T (K)

Rsp1 (s−1 )

1a)

15

1.58 × 108

6.34 × 10−9

2.05 × 108

4.88 × 10−9

15

8

−9

8

4.82 × 10−9

8

2a)

Rsp2 (s−1 )

𝝉 r1 (s)

1.56 × 10

6.42 × 10

𝝉 r2 (s)

2.07 × 10

3a)

15

1.55 × 10

6.43 × 10

2.00 × 10

5.00 × 10−9

4a)

20

2.77 × 108

3.62 × 10−9

3.72 × 108

2.69 × 10−9

6

−6

6

0.12 × 10−6

8

2.93 × 10−9

8

5a) 1b)

3.7 15

8

7.77 × 10

8

2.97 × 10

−9

0.13 × 10

−9

3.37 × 10

8.38 × 10 3.43 × 10

2b)

15

2.93 × 10

3.41 × 10

3.46 × 10

2.89 × 10−9

3b)

15

2.92 × 108

3.42 × 10−9

3.47 × 108

2.89 × 10−9

20

8

−9

9

0.92 × 10−9

7

0.62 × 10−7

7

4b) 5b)

3.7

8

5.22 × 10

7

1.47 × 10

−9

1.92 × 10

−7

0.68 × 10

1.08 × 10 1.62 × 10

1c)

15

1.04 × 10

0.96 × 10

1.18 × 10

0.85 × 10−7

2c)

15

1.03 × 107

0.97 × 10−7

1.19 × 107

0.84 × 10−7

15

7

−7

7

0.84 × 10−7

7

0.47 × 10−7

6

1.82 × 10−6

3c) 4c) 5c)

20 3.7

7

1.03 × 10

7

1.84 × 10

6

0.51 × 10

−7

0.97 × 10

−7

0.55 × 10

−6

1.96 × 10

1.19 × 10 2.14 × 10 0.55 × 10

Note: Results are given for transitions involving extended–extended states (possibility (i)), extended–tail states (possibilities (ii) and (iii)), and tail–tail states (possibility (iv)). a) Extended–extended states transitions (possibility (i)). b) Extended–tail or tail–extended states transitions (possibilities (ii)–(iii)). c) Tail–tail states transitions (possibility (iv)).

219

220

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

Stearns [9]. At 3.7 K, the calculated radiative lifetime is found to be in the μs range, which also agrees very well with Aoki et al.’s experimental results of the singlet exciton’s radiative lifetime [49] (see Figures 6.11, 6.13, and 6.14). (iii) Possibility (iv)—Tail-to-tail states recombinations For possibility (iv), as a type II singlet geminate pair, when both e and h are localized in their tail states, we get 𝜇x = 3.55 me , aex = 0.223 nm, and t ′ e = 1.29 × 1010 m−1 . Using these in Eq. (7.55), with i = 1, one gets Rspt1 = 1.03 × 107 s−1 , which gives 𝜏 r1 = 0.97 × 10−7 s at 15 K, Rspt1 = 1.84 × 107 s−1 , which gives 𝜏 r1 = 0.55 × 10−7 s at 20 K, and Rspt1 = 0.51 × 106 s−1 with the corresponding 𝜏 r1 = 2.0 μs at 3.7 K. Using Eq. (7.55) with i = 2, we get Rspt2 = 1.18 × 107 s−1 , which gives 𝜏 r2 = 0.85 × 10−7 s at 15 K, and Rspt2 = 2.14 × 107 s−1 , which gives 𝜏 r2 = 0.47 × 10−7 s at 20 K. The corresponding results at 3.7 K are obtained as Rspt2 = 0.55 × 106 s−1 and 𝜏 r2 = 2.0 μs. Wilson et al. [11] have also observed the lifetime in the μs time range, which may be attributed to transitions through possibility (iv). Aoki et al. [49] have observed another PL peak at 3.7 K (𝜏 G , Figure 6.14) with a radiative lifetime in the μs range but slightly longer than the singlet exciton radiative lifetime (𝜏 s ) at 3.7 K. According to the radiative lifetime calculated here (Table 7.2), the PL peak at 𝜏 G may be attributed to transitions from the tail-to-tail states from a type II singlet pair (possibility (iv)), because the corresponding radiative time is about 2 μs, longer than the radiative lifetime for possibilities (i)–(iii). A geminate pair of type I may be expected to have a larger separation and hence longer radiative lifetime. Indeed, in Figure 6.13, a peak, denoted by 𝜏 D , appears at much larger lifetime of ms, and Aoki et al. have attributed this peak to the distant or nongeminate pairs. According to the present theory, this peak may be attributed to the type I geminate pairs or distant pairs, but it is difficult to distinguish between the two (see the following text). For possibility (iv), we have calculated the rates using the same value of E0 as for possibilities (ii) and (iii), which may not be correct. However, as E0 cannot be measured experimentally, it is difficult to find a correct value for it for possibility (iv). One way to find a value of E0 for possibility (iv) is from the expected corresponding Stokes shift, when both the charge carriers have relaxed to their tail states. This can be done as follows [31]. The measured Stokes shift in a-Si:H is found to be 0.440 at 3.7 K [49] and 0.350 eV at 20 K [9] for possibilities (ii)–(iii). Here, we consider only these two different experimental values as examples. For possibility (iv), when both e and h are in their tail states, a double Stokes shift can be assumed. Accordingly, one gets the Stokes shift of 0.880 eV at 3.7 K and 0.700 eV at 20 K. Then, considering EC = 1.8 eV in a-Si:H, the PL peak is expected to occur at Emx = 1.8 − 0.880 = 0.920 eV at 3.7 K and 1.1 eV at 20 K, which give, from Eqs. (7.58) and (7.59), E0 = 0.918 eV at 3.7 K and 1.094 eV at 20 K. Using these values of E0 , the maximum value of rates at ℏ𝜔 = Emx are calculated, and the inverse of these reveals the corresponding radiative lifetimes (𝜏 r = 1/Rsp ). For type I geminate pairs, the lifetime is calculated for e–h separation, reh = 2L and 3L, where L = 0.235 nm, the average interatomic separation in a-Si:H; and, for the type II geminate pairs, reh = aex is used. The radiative lifetimes thus calculated in a-Si:H using the experimental value of Emx at 3.7 K from Aoki et al.’s experiment [49] and at 20 K from Stearns’ experiment [9] are listed in Table 7.3. At 3.7 K, for the type II geminate pairs, the radiative lifetime

7.2 Photoluminescence

Table 7.3 Rates of spontaneous emission and corresponding radiative lifetimes calculated from Eq. (7.55) (with i = 2) for geminate pairs of types I and II in a-Si:H. Radiative lifetime (𝝉 r ) Geminate I E 0 (eV) E mx (eV)

T (K)

Geminate II

reh = 2L

reh = 3L

reh = aex

0.918

0.920

3.7

1.10 ms

0.15 s

15.00 μs

1.094

1.100

20

6.20 μs

0.87 ms

88 ns

Note: Rates for type I are calculated for the average e–h separation, re–h = 2L and 3L (L = 0.235 nm) at the same E0 and Emx values.

is obtained in the ms time range, which agrees with the lifetime 𝜏 G observed by Aoki et al. At 20 K, the same lifetime is in the ns time range, which agrees with the lifetime observed by Stearns. Here, it is important to note that the input values used to calculate the rate of spontaneous emission at 3.7 and 20 K are obtained from two different experiments, namely references [9, 49], and therefore are different. Aoki et al. have not observed any radiative lifetime in the ns time range at any temperature, including 20 K as observed by Stearns. The value of Emx is not available to the author at 3.7 and 20 K from the same experiment. The results for the type I geminate pairs calculated at reh = 2L and 3L are in the ms to s time range, which agree with the radiative lifetimes, 𝜏 D , measured by Aoki et al. [48] but attributed to nongeminate pairs (see Figure 6.13). It is not possible to distinguish between the lifetime of type I geminate pairs and nongeminate pairs unless one can distinguish the corresponding separation between e and h in each. Therefore, it is possible that the measured lifetime is actually due to type I geminate pairs, because the type II geminate pairs were not known then [50]. In this view, the lifetime 𝜏 G measured in the ms time range can be associated to the type II geminate pairs, 𝜏 D in the ms to s range to type I geminate pairs, and any other longer lifetimes to nongeminate pairs. We have discussed here only the type II singlet geminate pairs, because, without consideration of the spin–orbit coupling, the transition matrix element (Eq. (7.11)) vanishes for a triplet spin configuration. As the magnitude of the spin–orbit coû xp in Eq. (7.3), the radiative lifetime of triplets pling is usually smaller than that of H is expected to be longer. The excitonic Bohr radius of triplet excitons and triplet type II geminate pair may not be very different, and therefore the lifetime 𝜏 T (see Figure 6.13) may be attributed to both. As the rates of recombination of nongeminate pairs or distant pairs have the same expression as in Eq. (7.55), it is not possible to distinguish between nongeminate and geminate pairs of the type I. The only difference can be the average separation between carriers, as it is well established that nongeminate pairs are usually formed from those geminate pairs excited by higher-energy photons, the separation between e and h in nongeminate pairs being larger than the critical separation given by rc ≤ n−1/3 ≤ 2rc for small generation rate [29, 30]. According to the present theory, a larger separation will yield an exponentially smaller recombination rate and hence larger radiative lifetime in the time range of seconds [50] (see Figure 6.13).

221

222

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

The radiative lifetime for reh ≥ 3L indeed produces such longer lifetimes at 3.7 K. One may therefore conclude that the excited pairs with a separation of 3L or more may be classified as nongeminate pairs. The longer lifetime obtained for possibility (iv) is because of the exponential factor exp(−2te′ |re−h |) appearing in the rate of recombination in Eq. (7.55) through Eq. (7.48), which is proportional to the probability of quantum tunneling a barrier of height Ec for a distance |re−h |. Therefore, the expression also implies that the rate of spontaneous emission reduces, and hence the radiative lifetime is prolonged in possibility (iv) due to the quantum tunneling through a distance of |re−h | before the radiative recombination. In evaluating the transition matrix element of the exciton–photon interaction operator (Eq. (7.3)), which depends only on the relative momentum between e and h in an exciton, the wave functions of e and h are used instead of the( wave) function of an exciton [10, 41]—for example, in the form of Ψ(R, r) = exp i Pℏ. R 𝜙(r). From this point of view, the theory presented here appears to address the recombination between an excited pair of an electron and a hole, not that between an electron and a hole in an exciton. The fact that the interaction operator is independent of the center of mass momentum P means that the integral Zlm𝜆 (Eq. (7.10)) always gives the average value of the separation between e and h. In an exciton, the separation is the excitonic Bohr radius, and in an e-and-h pair, it would be their average separation. As far as the recombination in amorphous solids is concerned, therefore, this is the only difference between an excitonic recombination and free e and h recombination. (iv) Results of two-level approximation As we have determined the energy, Emx , at which maxima of PL peaks have been observed in the five samples of a-Si:H, we can calculate Rsp12 in Eq. (7.32) and the corresponding radiative lifetime at these photon energies for a comparison. Here, we will use |re−h | = aex and 𝜔 = Emx /ℏ in Eq. (7.32). For extended-to-extended state recombination, aex = 4.67 nm, and, using Emx = 1.4 eV for Wilson et al.’s sample, we get w = 2.1 × 1015 Hz, Rsp12 = 7.4 × 1010 s−1 , and 𝜏 r = 1/Rsp12 = 13 ps. For Stearn’s sample, with Emx = 1.45 eV leading to 𝜔 = 2.2 × 1015 Hz, we get Rsp12 = 8.5 × 1010 s−1 and the corresponding 𝜏 r = 12 ps. For Aoki’s sample, with Emx = 1.36 eV and the corresponding 𝜔 = 2.0 × 1015 Hz, we get Rsp12 = 6.2 × 1010 s−1 and 𝜏 r = 16 ps. Thus, within the two-level approximation, one cannot get any results for the radiative lifetime longer than ps, which does not agree with the experimental results. This is a further clear indication of the fact that the two-level approximation is not valid for condensed-matter systems. 7.2.4

Temperature Dependence of PL

The maximum rates calculated from Eqs. (7.54) and (7.55) at the emission energy Emx increase exponentially as the temperature increases. This agrees very well with the model used by Wilson et al. [11] to interpret the observed temperature dependence in their PL intensity. Such a temperature dependence is independent of the origin of the PL, whether the recombination originates from the extended or tail states. However, the maximum rates obtained in Eqs. (7.54) and (7.55) vanish at T = 0, because x01 (Eq. (7.56)) and x02 (Eq. (7.57)) become zero as T → 0, which is in contradiction with the observed

7.2 Photoluminescence

rates [9, 11]. The reason for this is that one cannot derive the rate at 0 K from Eqs. (7.54) and (7.55). At 0 K, the electron-and-hole-probability distribution functions become unity; f c → 1 for Ee < EFn and f v → 1 for Eh > EFp as T → 0, and using this the rates at 0 K become equal to the pre-exponential factors as obtained in Eqs. (7.46)–(7.48). Thus, the temperature dependence of the rates obtained here can be expressed as: Rspi (T) = R0i {[1 − Θ1 (T)] + exp[−(ℏ𝜔 − E0 )∕𝜅B T]Θ2 (ℏ𝜔 − E0 )}, i = 1, 2

(7.61)

where R0i are the pre-exponential factors of Eqs. (7.54) and (7.55) corresponding to the four possibilities. The first step function, Θ1 (T), is used to indicate that the first term of Eq. (7.61) vanishes for T > 0, and the second step function Θ2 (ℏ𝜔 − E0 ) shows that there is no spontaneous emission for ℏ𝜔 − E0 < 0. For an estimate of R0i , we have used the values of Emx and E0 obtained at T = 3.7 K, which gives R01 = 7.77 × 106 s−1 , 1.47 × 107 s−1 , and 0.51 × 106 s−1 for extended–extended, extended–tail, and tail–tail states, respectively, and the corresponding values for R02 are obtained as 8.38 × 106 s−1 , 1.62 × 107 s−1 , and 0.55 × 106 s−1 . Wilson et al. [11] have fitted their observed rates to the following model: 1 = 𝜈1 + 𝜈0 exp(T∕T0 ) 𝜏r

(7.62)

and the best fit has been obtained with v1 = 108 s−1 , v0 = 0.27 × 106 s−1 , and T 0 = 95 K. Stearns [9] has obtained a best fit to his data with v1 = 2.7 × 108 s−1 , v0 = 6.0 × 106 s−1 , and T 0 = 24.5 K. According to Eq. (7.62), 𝜏r−1 = 𝜈1 at T = 0. Comparing these results with Eq. (7.61), one should have R0i = v1 , but R0i is obtained an order of magnitude smaller than v1 . Also, in Eq. (7.62), v1 > v0 , whereas, according to Eq. (7.61), v1 = v0 = R0i . These discrepancies may be attributed to the fact that Eq. (7.62) is obtained by fitting to the experimental data and not by any rigorous theory. As a result, both Eqs. (7.61) and (7.62) produce similar results for the radiative lifetime, but individual terms on the right-hand sides contribute differently. This is also apparent from the fact that, according to Eq. (7.62), v1 contributes to the rates also at nonzero temperatures, but, according to Eq. (7.61), R0i contributes only at T = 0 K. Stearns [9] has fitted the observed rate of emission at the emission energy of 1.43 eV to Eq. (7.62). We have plotted the calculated rates in Eq. (7.54) as a function of the temperature for possibility (ii) at the same emission energy (1.43 eV), and our results are shown in Figure 7.2a,b, respectively. Aoki et al. [49] have measured the QFRS spectra of a-Si:H and a-Ge:H at various temperatures (see Figure 6.14), where all PL peaks appearing at 𝜏 S (μs), 𝜏 G (μs), 𝜏 T (ms), and 𝜏 D (0.1 s) show a shift toward a shorter time scale as the temperature increases. This behavior is quite consistent with the temperature dependence obtained from the present theory as well. 7.2.5

Excitonic Concept

Another question may arise here. Why is the concept of excitons necessary to explain the radiative lifetime when the excited charge carriers can recombine without forming excitons? This can be answered as follows: Instead of using the reduced mass of excitons and the excitonic Bohr radius, if one uses the effective masses of e and h and the distance between them, one can calculate the radiative lifetime of a free excited electron–hole pair (not an exciton) from Eqs. (7.54) and (7.55). The radiative lifetime thus obtained

223

Rate (1010 s–1)

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

Figure 7.2 Rates plotted as a function of temperature at an emission energy of 1.43 eV and calculated for possibility (ii) from Eq. (7.54) in (a) i = 1 and (b) i = 2.

6 4 2 0 0

50

100

150 T (K)

200

250

300

200

250

300

(a)

10 Rate (1010 s–1)

224

8 6 4 2 0 0

50

100

150 T (K) (b)

for such a recombination is also in the ns range for possibilities (i)–(iii). For possibility (iv), of course, as the separation between e and h is expected to be larger, one may get smaller rates and hence a longer radiative lifetime. However, there are two important observations that support the formation of excitons: (i) A free e–h pair will relax down to the tail states in a ps time scale, resulting in PL only from possibility (iv), which is not supported by the observed PL (see Figures 6.11–6.13); and (ii) PL has a double-peak structure appearing in the time-resolved spectra corresponding to the singlet and triplet excitons. No such spin-correlated peaks can be obtained with the recombination of free electron–hole pairs. The theory presented here for the radiative recombination from the tail states is different from the radiative tunneling model [2], where the rate of transition is given by Rspt ∼ R0 exp(−2d/d0 )—d being the separation between e and h; d0 being the larger of the extents of the electron and hole wave functions, usually considered to be 10–12 Å; and R0 being the limiting radiative rate, expected to be about 108 –109 s−1 . Here, d, being the distance between a pair of excited charge carriers (not an exciton), is not a fixed distance, such as aex; it can be of any value. As a result, the radiative tunneling model cannot explain the appearance of the double-peak structure in PL originating from the tail states. Although the rate of recombination from the tail states derived in Eq. (7.55) also has a similar exponential dependence as in the radiative tunneling model, it depends on the excitonic Bohr radius, which is a constant, and hence the exponential factor becomes a constant. This explains very well the peak structure in PL.

7.3 Photoinduced Changes in Amorphous Chalcogenides

Kivelson and Gelatt [7] have developed a comprehensive theory for PL in amorphous semiconductors based on a trapped-exciton model. A trapped exciton considered by them is one in which the hole is trapped in a localized gap state and the electron is bound to the hole by their mutual Coulomb interaction. The model is thus similar to possibility (ii) considered here, but possibilities (i) and (iii) have not been considered by them. The square of the dipole transition matrix element is estimated by them to be ∼(2ah )5 /(aB )3 . This is probably one of the main differences between the present theory and that of Kivelson and Gelatt. The other important difference between the present theory and that of Kivelson and Gelatt is in possibility (iv), involving recombinations in the tail states. For this possibility, Kivelson and Gelatt have followed the radiative tunneling model and used the exponential factor as exp(−2d/a∗B ). As discussed in the preceding text, such an exponential dependence on varying d cannot explain the peak structures observed in PL spectra. It is quite clear from the expressions derived in Eqs. (7.54) and (7.55) that rates do not depend on the excitation density. This agrees very well with the measured radiative lifetime by Aoki et al. [48], who have found that the radiative lifetime of singlet and triplet excitons is independent of the generation rate (see Chapter 5). This provides additional support for the PL observed in a-Si:H to be due to excitons. One of the speculations, as described earlier, has been that the shorter PL time in the ns range observed by Wilson et al. may be due to the high excitation density used in the TRS measurements. The present theory clarifies this point very successfully—that it is due not to the excitation density but to the combined effect of temperature and fast nonradiative relaxation of charge carriers to lower energy states that one does not observe any peak in the ns time range at 3.7 K. Considering distant-pair recombinations, Levin et al. [30] have studied PL in amorphous silicon at low temperatures by computer simulation. They have also used the radiative tunneling model for recombinations in the tail states, but they have not considered radiative recombination of excitons. Therefore, the results of their work also cannot explain the occurrence of the double-peak structure in PL.

7.3 Photoinduced Changes in Amorphous Chalcogenides One of the photoinduced changes observed in amorphous semiconductors, particularly in amorphous chalcogenides (a-Chs), is the reduction in the bandgap by illumination, and it is commonly known as PD [5, 24, 51, 52]. In some cases, PD is also accompanied by photoinduced volume expansion (PVE) in the material (see Chapter 7). There are two kinds of PD observed in a-Chs, one that disappears by annealing but remains even if the illumination is stopped [26], and the other that disappears once the illumination is stopped [27]. The former is called metastable PD and the latter transient PD, which has been observed only recently. Much effort has been devoted to understanding the phenomena of PD and PVE in the last two decades [5, 53–56]; however, no model has been successful in resolving all issues observed. One of the recent models is repulsive electronic interaction [57]. As chalcogenides, similar to As2 Se(S)3 , have layered structures, the charge carriers excited by illumination in these materials move in these layers. The hole mobility is larger than the electron mobility in these materials, and hence photo-generated electrons reside mostly in the conduction-tail states while

225

226

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

holes diffuse away faster in the un-illuminated region through valence-extended and -tail states. Therefore, the layers that absorb photons during the illumination become negatively charged, giving rise to a repulsive inter-layer Coulomb interaction, which increases the inter-layer separation and causes PVE. The same force is assumed to induce an in-plane slip motion that increases the LP–LP interaction and causes PD. However, the repulsive Coulomb force model is qualitative, and more recent calculations [58] indicate that the force is too weak to cause PVE. More recently [59] a unified description of the photo-induced volume changes in chalcogenides based on tight-binding (TB) molecular dynamics (MD) simulations of amorphous selenium has been put forward (see Chapter 7). For causing movements in an atomic network, one requires the involvement of lattice vibrations, of which (surprisingly) none of the models of PD and VE has given any account. However, the involvement of lattice vibrations has recently been considered [60] in inducing photo-structural changes in glassy semiconductors. In 1988, Singh [61] discovered a drastic reduction in the bandgap due to the pairing of charge carriers in excitons and exciton–lattice interaction in nonmetallic crystalline solids. The magnitude of reduction in the bandgap varies with the magnitude of the exciton–phonon interaction, which is different in different materials; the softer the structure is, the stronger the carrier–phonon interaction. It has been established that, due to their planar structure [5], the carrier–phonon interaction in a-Chs is stronger than in a-Si:H, which satisfies the condition for the occurrence of Anderson’s negative-U [62] in a-Chs but not in a-Si:H. This is the basis of the hole-pairing model for the creation of light-induced defects in a-Chs [5]. In this section, based on Holstein’s approach [63], the energy eigenvalues of positively and negatively charged polarons and paired charge carriers, created by excitations due to illumination, are calculated. It is found that the energies of the excited electron (negative charge) and hole (positive charge) polarons become lowered due to the carrier–phonon interaction [64]. Also, the like-excited charge carriers can become paired because of the negative-U effect [62] caused by the strong carrier–lattice interaction, and the energy of such paired states is also lowered. Thus, the hole polaronic state and paired-hole states overlap with the lone-pair and tail states in a-Chs, which expands the valence band and reduces the bandgap energy and hence causes PD. The formation of polarons, as well as the pairing of holes, increases the bond length on which such localizations occur, which causes VE. The energy eigenvalues of such polarons and pairing of charge carriers have been calculated. 7.3.1

Effect of Photo-Excitation and Phonon Interaction

Based on Holstein’s approach [63], we consider here a model amorphous solid in the form of a linear chain of atoms. The electronic Hamiltonian in such a chain can be ̂ ph ), and charge carrier–phonon inter̂ el ), phonon (H written as a sum of charge carrier (H ̂ I ) energy operators in the real coordinate space as [5, 65]: action (H ̂ =H ̂ el + H ̂ ph + H ̂I H

(7.63)

In a-Chs, as the lone-pair orbitals overlap with the valence band, the combined valence band becomes much wider than the conduction band. As the effective mass of charge carriers is inversely proportional to the corresponding band width [5] (see Eqs. (3.17)

7.3 Photoinduced Changes in Amorphous Chalcogenides

and (3.20)), the hole’s effective mass becomes smaller than the electron’s effective mass in a-Chs. As a result, holes move faster than electrons in these materials. However, in the tail states, the charge carriers become localized. In such a system, we consider that a photon is absorbed to excite an electron (e) in the conduction state and hole (h) in the valence state. Considering the effect of strong carrier–phonon interaction, the excited hole may become a positive-charge polaron and an electron may become a negative-charge polaron. Furthermore, two excited electrons and two excited holes can become paired and localized on a bond due to the effect of negative-U, a fact which has already been established [5, 62]. Pairing of holes on a bond breaks the bond and creates a pair of dangling bonds, which is used to explain the creation of light-induced defects in a-Ch [5]. Let us first consider the case of an excited hole in the valence band. The eigenvector of such an excited hole can be written as: | | ∑ | +| Cl d0l (7.64) |0 > |h, 0 >= | | l | | where h denotes the hole and 0 denotes valence band; and the conduction band is + (=a0l ) is the denoted by 1. C l represents the probability amplitude coefficient, d0l creation operator of a hole, a0l is the annihilation operator of an electron in the valence states on site l, and |0> represents the vacuum state with all valence states completely filled and all conduction states completely empty. Using Eqs. (7.63) and (7.64), one can ̂ 0 ≥ Wh |h, 0 >, to get a secular equation such as: solve the Schrödinger equation, H|h, ) ( ∑ ℏ2 ∇2n 1 ∑ h 0 M𝜔m 𝜔n xm xn − Al xl − Eh Cl + Th (Cl−1 + Cl+1 ) Wh Cl = − + 2M 2 m,n n (7.65) where W h is the energy eigenvalue of the hole; the first two terms within the parentheses correspond to the kinetic and potential energies of nuclear vibrations; xn is the nth bond length of a diatomic molecule in the chain vibrating with a frequency 𝜔n ; Ahl is the hole–phonon coupling coefficient; Eh0 = Ehl , a constant of energy of a hole localized at site l and hence site independent; and T h is the hole-transfer energy between nearest neighbors from l to {l ±1} in the chain. Although the vibration frequency 𝜔m has a subscript m, being the intramolecular vibrational frequency of identical molecules, it is site independent. Considering only the diagonal terms in the vibrating potential, multiplying Eq. (7.65) by Cl∗ and then summing over all l, we get: ∑ ℏ2 ∇2n 1 ∑ ∑ M𝜔2m x2m − Al xl |Cl2 | − Eh0 + Th (Cl−1 + Cl+1 )Cl∗ (7.66) Wh = − + 2M 2 n m l ∑ ∗ ∑ 𝜕W 2 where l Cl Cl = l |Cl | = 1 is used. Setting 𝜕x h = 0, the energy eigenvalue, W h , can q

be minimized with respect to the bond length, xq . This gives the bond length at the minimum energy as: j

x0q =

Aq |Cq(0) |2

(7.67) M𝜔2 where the superscript (0) denotes the value of quantities at the minimum energy, and the subscript q denotes the qth bond at which the hole is localized. The subscript q has

227

228

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

been dropped from the frequency 𝜔 as it is independent of bond sites. The bond length increases by x0q due to the localization of the hole on the bond. Substituting Eq. (7.67) into Eq. (7.66), we calculate the minimum energy of the hole, denoted by Wh0 , as: Wh0 = −Eh0 − 2Th + Ehq

(7.68)

where Ehq represents the hole–polaron binding energy obtained as: 2

Ehq =

(Ahq ∕M𝜔2 )2 48Th

(7.69)

This is the energy by which the energy of an excited hole is lowered from the free-hole state energy, which is at = −Eh0 − 2Th . This means that the hole-energy state moves upward in the energy gap by releasing an energy Ehq to phonons. 7.3.2

Excitation of a Single Electron–Hole Pair

The generation of excited electrons and holes occurs in pairs through photoexcitations. The excited electrons may also form polarons in the conduction states, and their energy will also be lowered. For a single pair of excited charge carriers on our linear chain, the secular equation can be written as: ] [ 1∑ l+X l e h 2 2 Weh Cl = (Ee − Eh ) − (Al+X − Al ) + M𝜔 xm − Ueh Cl 2 m − Te (Cl+X+1 + Cl+X−1 ) + Th (Cl+1 + Cl−1 )

(7.70)

where W eh is the energy eigenvalue of the excited pair; Eel = Ee0 represents the energy of an electron at site l and is also site independent; U eh is the Coulomb interaction energy between the excited pair; T e is the energy of electron transfer between nearest neighbors; M is the atomic mass and 𝜔 the frequency of vibration between nearest-neighbor atoms; X is the distance between the excited pair; and xl is the bond length between nearest neighbors of site l. Minimizing the energy with respect to xq , we obtain from Eq. (7.70) as [5]: Weh = Ee0 − Eh0 − Ueh − 2(Te − Th ) − Eeq − Ehq

(7.71)

where Ee0 and Eh0 are the site-independent energies of the excited electron and hole, respectively, without the lattice and Coulomb interactions between the excited charge carriers, and Eeq is the electronic polaron-binding energy, which can be obtained from Eq. (7.69) by replacing the subscript h by e. The energy of an excited pair of charge carriers without the lattice interaction is given by: Weh = Ee0 − Eh0 − Ueh − 2(Te − Th )

(7.72)

In the excitation of a free pair of charge carriers, one usually neglects U eh , and then the 0 energy Weh is close to the optical energy gap in most materials. Subtracting Eq. (7.72) from Eq. (7.71), we get the reduction in the optical gap due to the formation of an excited pair of polarons as: 0 ΔW = Weh − Weh = −Eeq − Ehq

(7.73)

7.3 Photoinduced Changes in Amorphous Chalcogenides

The lowering of an excited pair’s energy also means that the bond on which such polaronic formation occurs will be stretched by x(0) q , and the bond may break. A bond breaking due to such single excitation has been recently demonstrated by numerical simulation ([66]; see also Chapter 7) as well. This can also be seen from Eq. (7.67)—that the bond becomes stretched due to an excitation in which a hole is localized on the bond, and hence the bonding becomes weaker. 7.3.3

Pairing of Like Excited Charge Carriers

Here, we consider that two pairs of electrons and holes are excited in the chain. Assuming that the separation between electron and hole in one excited pair is x and that in the other is x′ , and that the separation between two holes is X, the secular equation analogous to Eq. (7.70) can be written as [5, 64]: W (x, x′ , X)Cl = [(Eel+x − Ehl ) + (Eel−X+x − Ehl−X ) − (Ael+x xl+x − Ahl xl ) 1∑ ′ − (Ael−X+x′ xl−X+x′ − Ahl−X xl−X ) + M𝜔2m x2m + U12 )]Cl 2 m ′

− Te (Cl+x+1 + Cl+x−1 + Cl−X+x′ +1 + Cl−X+x′ −1 ) + Th (Cl+1 + Cl−1 + Cl−X+1 + Cl−X−1 )

(7.74)

where W (x, x′ , X) is the energy eigenvalue of the two excited pairs of electrons and holes, and U ′ 12 is the total Coulomb interaction between the two pairs of excited charge carriers. It is well established that Se-based chalcogenides have linear chain-like structures and hence are flexible, and in such structures the carrier–phonon interaction is considered to be very strong. Such a strong carrier–phonon interaction can induce pairing of like-excited charge carriers on a bond due to Anderson’s U effect [62], known for valence-band electrons. In most solids, usually one of the two interactions, electron–phonon or hole–phonon, is stronger than the other. Therefore, here we first consider the case that the hole–phonon interaction is stronger, and therefore that two holes can be paired on a bond and the two excited electrons will form two polarons on the chain. The pairing of electrons can also be studied in an analogous way, as described in the following text. In solving Eq. (7.74), it is assumed that the energy eigenvalues of two electronic polarons are already derived in an analogous way as for the hole polaron (Eq. (7.68)). Our interest here is to calculate the energy eigenvalue of the excited state with two holes localized on the same bond, that is, X = 0. In this case, the secular equation (Eq. (7.74)) reduces to: ] [ 1∑ 0 l h 2 2 ′ W2h(x,x′ ,0) Cl = 2We − 2Eh + 2Al xl + M𝜔m xm + U12 Cl + 2Th (Cl+1 + Cl−1 ) 2 m (7.75) where the subscript 2h on W 2h(x, x′ , 0) denotes the localization of the two holes, and We0 is the energy of an electronic polaron derived in an analogous way as in Eq. (7.68) for a hole polaron, and is obtained as: We0 = Ee0 − 2Te − Eeq

(7.76)

229

230

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

Following the steps used to derive Eqs. (7.65) and (7.66), here again we can minimize the energy with respect to xq to get: 0 = 2We0 − 2Eh0 + 4Th + Uh − Ehh W2h ′ = 2(Ee0 − Eh0 ) − 4(Te − Th ) + U12 − 2Eeq − Ehh

where

( Ehh

1 = 6Th

(Ahq )2

(7.77)

)2 (7.78)

M𝜔2

The bond length of a bond on which such a pairing occurs becomes twice as large as when only a single-hole polaron is formed, and is given by: 2Ahq Cq∗ Cq

= 2x0q (7.79) M𝜔2 The energy of two excitations without the charge carrier–phonon interaction can be written as: xhh q =

00 ′ W2h = 2(Ee0 − Eh0 ) − 4(Te − Th ) + U12

(7.80)

Thus, the energy of a pair of excitations with two holes localized on a bond is lowered by ΔE, given as: 00 0 ΔE = W2h − W2h = 2Eeq + Ehh

(7.81) 0 W2e ,

In an analogous way, one can derive the energy eigenvalue, of a pair of excited electrons localized on an antibonding orbital of a bond and two-hole polarons localized separately elsewhere as: 0 ′ W2e = 2(Ee0 − Eh0 ) − 4(Te − Th ) + U12 − 2Ehq − Eee

(7.82)

where Eee can be obtained from Eq. (7.78) by replacing the subscript h by e. It may be noted here that, unlike the case of pairing of holes on a bond, pairing of electrons on a bond does not break the bond. In this case, the two-hole polaron’s energy states move upward in the bandgap, and the paired-electron energy states in the conduction band move downward, resulting in a narrowing of the bandgap. We have shown earlier that the localization of an excited hole on a bond increases its bond length, and the bond can break. Such a localization occurs by the formation of a hole polaron due to strong interaction between the hole and lattice vibrations. The hole-polaron state has an energy lower than the free-hole state and moves upward, mixing with the lone-pair orbitals in chalcogenides that widens the valence band and narrows the bandgap. Such a strong charge carrier–phonon interaction is possible in a-Chs because of their linear flexible structure and weak coordination, which can also induce pairing of excited holes on a bond. In this case, the bond length of a bond increases twice as much as in the case of polaron formation, and their binding energy is eight times larger than the polaron binding energy. Such paired-hole states contribute significantly to both PD and PVE. PD is caused by the lowering of the paired-hole state energy, and such a state widens the valence band even further and hence reduces the bandgap. As the bond length expands twice as much as in the case of a single-hole polaron, this contributes significantly to VE as well. Moreover, pairing of holes on a bond breaks the bond because

7.3 Photoinduced Changes in Amorphous Chalcogenides

of the removal of covalent electrons and causes photo-induced bond breaking in a-Chs, as has already been established. Let us arrive at some estimates of the possible reduction in the bandgap due to formation of polarons and bipolarons (paired holes) on a bond. Depending on the material, ΔE (Eq. (7.81)) can be in the region of a fraction of an eV. We have estimated ΔE in As2 S3 as follows. The energy of lattice vibration of a bond can be written as [5]: 1 E(q) = E0 + M𝜔2 (q − q0 )2 (7.83) 2 where E0 and q0 are the energy and the interaction coordinate, respectively, at the minimum of the vibrational energy. The vibrational force along the interaction coordinate can be obtained as: ( ) 𝜕E A= = −M𝜔2 q0 (7.84) 𝜕q q=0 Using this in Eqs. (7.69) and (7.78), we obtain: ( )2 2 2 Ehp Ehh 1 M𝜔 q0 =8 = Th Th 6 Th

(7.85)

For As2 S3 , the phonon energy of the symmetric mode is 344 cm−1 [4]. Using this and applying Toyozawa’s criterion [66] of strong carrier–phonon interaction as Ehp ≥ T h , we get T h = 12.33 meV from Eq. (7.85), which gives Ehp = 12.33 meV, Ehh = 98.40 meV, and ΔE = 0.123 eV. This agrees very well with the observed reduction in the bandgap in As2 S3 of about 0.16 eV [5, 67]. A similar narrowing in the bandgap is expected when two excited electrons become paired on an anti-bonding orbital and two-hole polarons are formed elsewhere. In this case, however, as no bond breaking may occur, the substance will go back to the original state after the exciting energy source (illumination) is stopped. In materials where the excited pairs of charge carriers form only a pair of polarons, without any pairing of like-charge carriers, the reduction in bandgap will be equal to Eep + Ehp ∼ 2Ehp = 25 meV. Reductions in the bandgap in various materials, estimated from experimental data [67], are listed in Table 7.4; accordingly, most observed reductions are found to be in the range of 0.02–0.17 eV. Table 7.4 Bandgap reduction, ΔE estimated from the observed data [67] in various amorphous materials. Amorphous material

𝚫E (eV)

As2 S3

0.16

GeS2

0.17

S

0.12

As2 Se3

0.07

Se

0.06

GeSe2

0.03

As2 Te3

0.06

Sb2 S3

0.02

231

232

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

Thus, it is shown in the preceding text that, in materials with strong carrier–lattice interaction, the excited charge carriers can form polarons; also, like-excited-charge carriers can become paired as self-trapped bipolarons on a bond because a paired like-charge carrier excited state is more energetically stable. The planar structure of chalcogenides with weak coordination enables these materials to be more flexible and hence possess stronger carrier–phonon interaction. The energy states of both polarons and bipolarons are lower than that of excited free-charge carriers. Thus, the energy of a hole polaron and bipolaron (paired holes) moves up further in the lone-pair orbitals and -tail states, which expands the valence band. Likewise, the energy states of electron polaron and bipolaron (paired electrons) in the anti-bonding orbitals lower the conduction mobility edge down. These effects together are responsible for a reduction in the bandgap, and hence PD. It should be noted that all the preceding possibilities might not occur together in the same material. In materials where the electron–lattice interaction is larger than the hole–lattice interaction, the formation of electron polaron and bipolaron will be more efficient, and materials with stronger hole–lattice interaction will have hole polaron and bipolaron formation more efficient. Accordingly, a varying degree of PD is expected to occur in different materials, and this is quite in agreement with the results listed in Table 7.4. It has been established [5] that the pairing of holes on a bond breaks the bond as soon as two excited holes become localized on it. It is also possible for a bond to be broken due to localization of a single hole on a bond. The bond breaks due to the removal of covalent electrons, and a pair of dangling bonds is created. This is the essence of the pairing-hole theory of creating light-induced defects in a-Chs. Such light-induced defects are reversible by annealing. However, the pairing of excited electrons does not break a bond; it only reduces the bandgap, and such an excited state will revert to the original material after the illumination has been switched off. This concept can be applied to explain both metastable and transient PD. The former occurs due to either formation of hole polarons or pairing of holes, or both, such that bonds are broken, and which cannot be recovered by stopping the illumination. It remains metastable, and the material reverses back to its original form by thermal annealing. The latter occurs due to pairing of electrons and/or formation of polarons without any bond breaking, and then the material reverses back to its original form after the illumination is stopped. Usually, transient PD is observed more than is metastable PD. This is because there are three processes contributing to transient PD—pairing of electrons, formation of positive-charge polarons (without bond breaking), and negative-charge polarons—as compared to only two possible channels of formation of positive-charge polarons (with bond breaking) and pairing holes contributing to metastable PD.

7.4 Radiative Recombination of Excitons in Organic Semiconductors The photoexcitation and formation of excitons have been discussed in Section 4.4 in detail. The non-radiative processes of the formation of charge transfer (CT) excitons and dissociation of excitons have also been discussed in Section 4.4. In this section, mechanisms of radiative recombination of singlet and triplet excitons are described with a view of their applications in OLEDs. In organic semiconductors and polymers, when the

7.4 Radiative Recombination of Excitons in Organic Semiconductors

electron and hole pairs are created either by photoexcitation or carrier injection, they usually form Frenkel excitons instantaneously due to the high binding energy between them caused by the low dielectric constant, in the range of 3–4. As stated earlier in Section 4.4, this high binding energy enables excitons to recombine radiatively by emitting light. However, only the radiative recombination of singlet excitons is spin-allowed and that of triplet excitons is spin-forbidden. Therefore, the radiative recombination from triplet excitons can only occur through spin–orbit interaction, which is known to be weak in organic materials. The emission of light due to the radiative recombination of singlet excitons is called fluorescence and due that due to triplet excitons is called phosphorescence. 7.4.1

Rate of Fluorescence

The rate of spontaneous emission of singlet excitons or fluorescence derived within the two-level approximation as given in Eq. (7.32) can also be applied to organic semiconductors. The rate of fluorescence, also known as the prompt fluorescence, denoted by RPF , because this is much prompter than the rate of phosphorescence, can be written as: (7.86)

RPF = Rsp12

where Rsp12 is the rate of spontaneous emission from a singlet exciton derived in Eq. (7.32), and |re−h | = axs ∕𝜀

(7.87) 5axt , 4

with axt as where axs is the Bohr radius of singlet excitons, which is given by axs = the triplet excitonic Bohr radius given by axt = 𝜇𝜀 a , where 𝜇 is the reduced mass of 𝜇x o electron in hydrogen atom, ao = 0.0529 nm is the Bohr radius [68, 69], 𝜀 is the static dielectric constant, and 𝜇x = 0.5 me is the reduced mass of excitons in organic solids Thus, we get the singlet exciton Bohr radius as: axs =

5𝜇𝜀 a 4𝜇x o

(7.88)

In organic semiconductors, the rate of prompt fluorescence is usually of the order of 108 s−1 [70], which gives the fluorescence time, 𝜏 PF = 1/RPF , in the ns range. 7.4.2

Rate of Phosphorescence

The rate of radiative recombination of a triplet exciton can be calculated using the newly invented time-dependent exciton–photon–spin–orbit interaction operator [71]. Within the two-level approximation, applicable to most organic solids, the rate of spontaneous emission from the radiative recombination of triplet excitons is obtained as [68, 72]: Rspt =

32e6 Z2 𝜅 2 𝜀ℏ𝜔12 𝜇x4 c7 𝜀0 a4xt

s−1

(7.89)

where Z is the highest atomic number in the organic molecules, 𝜅 = (4𝜋𝜀0 )−1 , ℏ𝜔12 = ELUMO − EHOMO is the emitted energy equal to the energy difference between the excited triplet (LUMO) and ground state (HOMO) in an organic solid, 𝜀o is the 𝜇 vacuum permittivity, and ℏ is the reduced Planck’s constant. Using 𝜇x = 0.5, 𝜀 = 3, we

233

234

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

get axt = 6a0 (see Eq. (7.88)) and other standard parameter values in Eq. (7.89), and we can write the rate of phosphorescence, RPhos , as: RPhos = Rspt = 25.3Z2 ℏ𝜔12 s−1 ( ℏ𝜔12 in eV)

(7.90)

For emission in the range of 2.00–2.5 eV, in an organic compound doped with iridium (Ir) with Z = 77, we get the rate of phosphorescence RPhos ∼ 105 s−1 , which is two orders of magnitude higher than in any undoped organic solid and three orders of magnitude lower than the rate of prompt fluorescence in Eq. (7.86). The time of phosphorescence 𝜏 Phos = 1/RPhos is in the μs range in organic solids incorporated with Ir, and in a fraction of ms range in undoped ones. 7.4.3

Organic Light Emitting Diodes (OLEDs)

In a basic OLED, a layer of organic material is sandwiched between two electrodes, as shown in Figure 7.3. In organic solids, the valence and conduction bands are presented by the highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO), respectively. When an external voltage is applied in the positive bias, electrons are injected from the cathode in the LUMO and holes from anode to the HOMO of the organic layer, and they instantly form Frenkel excitons due to the lower dielectric constant leading to a higher binding energy (see Figure 7.3). Although the formation of excitons due to higher binding energy is good for their radiative recombination and hence the emission of light, also called the electroluminescence, excitons are formed in both singlet and triplet states. Due to the multiplicity of the triplet configuration, the statistical probability of forming singlet and triplet excitons is in the ratio of 1 : 3, and, as triplet excitons cannot recombine radiatively, the emission efficiency, also called internal quantum efficiency (IQE), of such a simple OLED can at most be only 25%. The rate of spontaneous emission of singlet excitons or fluorescence is given in Eq. (7.86), which has the radiative lifetime in the ns range. Therefore, it is very important to capture the emission of triplet excitons in OLEDs. For this, there are two ways in which the emission from triplet excitons can be achieved: (i) by incorporating heavy metal atoms, such as iridium (Ir), palladium (Pd), or platinum (Pt), in the organic compound, which increases the spin–orbit interaction by increasing the atomic number Z in the rate of phosphorescence described in Eqs. (7.89) and (7.90); and (ii) by converting triplet excitons into singlet excitons through either the reverse intersystem crossing (RISC) described in Chapter 5 (see Eq. (5.34)) or triplet–triplet annihilation (TTA) [41]. Using the first method, the radiative lifetime of triplet excitons LUMO CATHODE e Exciton Light emission h ANODE HOMO Organic layer

Figure 7.3 A schematic structure of a simple OLED and its operation mechanism.

7.4 Radiative Recombination of Excitons in Organic Semiconductors

gets shortened to the μs time range, which is still longer than that of the prompt fluorescence. Thus, the IQE can be increased from 25% to 100% by using one of these two methods; however, through TTA, as two triplets interact and upconvert into one singlet, the maximum contribution can only be 37.5% out of 75% triplet excitons. Through the other two processes, one can achieve 100% IQE. However, it may be noted that both the processes—that is, direct conversion to singlet through spin–orbit interaction (see rates in Eqs. (7.89) and (7.90)) and RISC (Eq. (5.34))—require the incorporation of heavy metal atoms for efficient transfer, but the cost of heavy metal atoms, for example, Ir or Pt, is quite high, and hence the use of such OLEDs becomes expensive [73, 74]. In this regard, the method of conversion of triplet excitons into singlets, known as up-conversion, leading to TADF has been explored extensively in the recent past, and has emerged as one of the most attractive technologies for lighting [73, 75]. As described in Chapter 5, the conversion from triplet to singlet or vice-versa involves two mechanisms: (i) spin flip; and (ii) least possible energy difference between singlet and triplet, denoted by ΔEST , also known as exchange energy. The process may not be very efficient if ΔEST is large and the spin cannot be flipped. TADF OLEDs based on heavy metal atoms are commonly called second-generation OLEDs. 7.4.3.1

Second- and Third-Generation OLEDs: TADF

The mechanism of TADF is schematically presented in Figure 7.4, where first a triplet exciton at T1 gets thermally elevated with energy ΔEST and then converted into a singlet exciton at S1 , through the RISC. The radiative lifetime of the singlet exciton (prompt fluorescence) is assumed to be much faster than that of the nonradiative intersystem crossing to go back to the triplet state. The singlet exciton radiatively recombines with the ground state S0 and emits fluorescence. According to Figure 7.4, the radiative lifetime of TADF can be obtained from the rate of TADF as [70]: −1 −1 τTADF = R−1 TADF = RPF + kRISC

(7.91)

As RPF is much faster than k RISC , the time 𝜏 TADF ≈k RISC , which is usually much slower than the radiative life time of prompt fluorescence (𝜏 PF = k PF −1 ), and hence TADF occurs with a delay in time—this is the reason for calling it delayed fluorescence. However, in 2013, Adachi et al. [76] have proposed the third generation of OLEDs, where two molecules are synthesized, such that ΔEST is very small between them. By blending such organic molecules as donor (with singlet state S1 ) with the acceptor (with triplet state T1 ) in the active layer, Adachi’s group has achieved very efficient TADF OLEDs, as shown in Figure 7.4. Although it is not fully understood yet how the spin −1

Figure 7.4 Schematic presentation of the mechanism of TADF in OLEDs. A triplet exciton gets thermally elevated to singlet S1 , which then recombines radiatively to the ground S0 state.

S1 ΔEST

kRISC T1

RPF RPhos

S0

235

236

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

gets flipped without the involvement of any heavy metal atoms, there are two possible speculations [77]: (i) as ΔEST is very small, the smaller spin–orbit interaction achieved in such organic solids may be adequate to flip the spin; and (ii) the molecular structure of such donor acceptor molecular blending may be such that it is able to enhance the spin–orbit interaction adequately enough to make the spin flipping very efficient.

7.5 Conclusions It is demonstrated that the effective-mass approach can be applied to amorphous structures to understand many electronic and optical properties that are based on the free-carrier concept. Using the effective-mass approximation, it is shown that the excitation-density-independent PL observed in a-semiconductors arises from the radiative recombination of excitons, types I and II geminate pairs, and nongeminate pairs. Both the Stokes shift and radiative lifetimes should be taken into account in determining the PL electronic states. Although the radiative lifetime for possibilities (i)–(iii) are of the same order of magnitude, the Stokes shift observed in PL suggests that these recombinations occur from extended-to-tail states (possibility (ii)) in a-Si:H. The singlet radiative lifetime is found to be in the ns time range at temperatures >15 K and in ms range at 3.7 K, and triplet lifetime in the ms range. In possibility (iv), carriers have to tunnel to a distance equal to the excitonic Bohr radius, and hence the radiative lifetime gets prolonged. The PL from possibility (iv) can arise from two types of geminate pairs, excitonic (type II) and nonexcitonic (type I), and nongeminate or distant pairs. It is also shown that the effective mass of a charge carrier changes in amorphous semiconductors as it crosses its mobility edge. This also influences the radiative lifetime of an exciton as it crosses the mobility edges. A large Stokes shift implies a strong carrier–lattice interaction in a-Si:H, and therefore PL occurs in thermal equilibrium. The results of two-band approximation and nonequilibrium are therefore not applicable for a-Si:H. Finally, using the effective mass-approximation, it is shown quantum mechanically that it is the strong carrier–lattice interaction in a-Chs that causes all three phenomena: creation of the light-induced defects, PD, and VE. The results of the present theory agree qualitatively with those obtained from the molecular dynamics simulation (see Chapter 7). Finally, a theory of electroluminescence from organic semiconductors is presented in Section 7.4, including the calculation of the rates fluorescence and phosphorescence. A theory of light emission in OLEDs is also presented along with the operation of the second- and third-generation OLEDs based on the thermally activated and time-delayed fluorescence (TADF). The theory of TADF is also briefly presented.

Acknowledgments The author has benefited very much from discussions with Professors T. Aoki, K. Shimakawa, Keiji Tanaka, and Sandor Kugler during the course of this work. The work was originally supported by the Australian Research Council’s large grants (2000–2003) and IREX (2001–2003) schemes, and a bilateral exchange grant (2005/06) by the Australian Academy of Science and the Japan Society for the Promotion of Science (JSPS).

References

References 1 Mott, N.F. and Davis, E.A. (1979). Electronic Processes in Non-Crystalline Materials.

Oxford: Clarendon Press. 2 Street, R.A. (1991). Hydrogenated Amorphous Silicon. Cambridge: Cambridge Uni-

versity Press. 3 Redfield, D. and Bube, R.H. (1996). Photoinduced Defects in Semiconductors. Cam-

bridge: Cambridge University Press. 4 Morigaki, K. (1999). Physics of Amorphous Semiconductors. London: World Scientific. 5 Singh, J. and Shimakawa, K. (2003). Advances in Amorphous Semiconductors. 6 7 8 9 10 11 12

13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

London: Taylor and Francis. Street, R.A. (1981). Adv. Phys. 30: 593. Kivelson, S. and Gelatt, C.D. Jr., (1982). Phys. Rev. B 26: 4646. Wilson, B.A., Hu, P., Jedju, T.M., and Harbison, J.P. (1983). Phys. Rev. B 28: 5901. Stearns, D.G. (1984). Phys. Rev. B 30: 6000. Singh, J., Aoki, T., and Shimakawa, K. (2002). Philos. Mag. B 82: 855. Wilson, B.A., Hu, P., Harbison, J.P., and Jedju, T.M. (1983). Phys. Rev. Lett. 50: 1490. Aoki, T., Komedoori, S., Kobayashi, S. et al. (2002). J. Non-Cryst. Solids 299–302: 642. The group has recently notified that an error was made in measuring the Emx in this paper and that it has been corrected in ref. [20]. Deppina, S.P. and Dunstan, D.J. (1984). Philos. Mag. B 50: 579. Boulitrop, F. and Dunstan, D.J. (1985). J. Non-Cryst. Solids 77–78: 663. Stachowitz, R., Schubert, M., and Fuhs, W. (1998). J. Non-Cryst. Solids 227–230: 190. Ishii, S., Kurihara, M., Aoki, T. et al. (1999). J. Non-Cryst. Solids 266–269: 721. Aoki, T., Komedoori, S., Kobayashi, S. et al. (2002). Nonlinear Opt. 29: 273. Aoki, T. (2003). J. Mater. Sci. - Mater. Eng. 14: 697. Singh, J. and Oh, I.-K. (2005). J. Appl. Phys. 97: 063516. Davis, E.A. (1985). Amorphous Semiconductors, 2e (ed. M.H. Brodsky), 41. Berlin), Ch. 3: Springer. Staebler, D.L. and Wronski, C.R. (1977). Appl. Phys. Lett. 31: 527. Stutzmann, M. (1987). Philos. Mag. B 56: 63. Branz, H.M. (1999). Phys. Rev. B 59: 5498. Shimakawa, K., Kolobov, A., and Elliott, S.R. (1995). Adv. Phys. 44: 475. Hisakuni, H. and Tanaka, K. (1994). Appl. Phys. Lett. 56: 2925. Shimakawa, K., Yoshida, N., Ganjoo, A. et al. (1998). Philos. Mag. Lett. 77: 153. Ganjoo, A., Shimakawa, K., Kitano, K., and Davis, E.A. (2002). J. Non-Cryst. Solids 299–302: 917. Dunstan, D.J. and Boulitrop, F. (1984). Phys. Rev. B 30: 5945. Shklovskii, B.I., Fritzsche, H., and Baranovskii, S.D. (1989). Phys. Rev. Lett. 62: 2989. Levin, E.I., Marianer, S., and Shklovskii, B.I. (1992). Phys. Rev. B 45: 5906. Singh, J. (2006). J. Non-Cryst. Solids 352: 1160. Singh, J. (2002). J. Non-Cryst. Solids 299–302: 444. Singh, J. (2002). Nonlinear Opt. 29: 111. Singh, J. (2003). J. Mater. Sci. 14: 171. Cody, G.D. (1984). Semiconductors and Semimetals, vol. 21, part B, 11. NJ: Elsevier.

237

238

7 Photoluminescence, Photoinduced Changes, and Electroluminescence in Noncrystalline Semiconductors

36 Basu, P.K. (1997). Theory of Optical Processes in Semiconductors. Oxford: Clarendon

Press. 37 Grahn, H.T. (1999). Introduction to Semiconductor Physics. Singapore: World Scien-

tific. 38 Dumke, W. (1957). Phys. Rev. 105: 139. 39 Lasher, G. and Stern, F. (1964). Phys. Rev. A 133: 553. 40 Beb, H.B. and Williams, E.W. (1972). Semiconductors and Semimetals, vol. 8 (eds.

R.K. Willardson and A.C. Beer), 18. London: Academic Press. 41 Singh, J. (1994). Excitation Energy Transfer Processes in Condensed Matter. New 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

York: Plenum. Van Roosbroeck, W. and Shockley, W. (1954). Phys. Rev. 94: 1558. Elliott, S.R. (1998). The Physics and Chemistry of Solids. Sussex: Wiley. Shah, J. and Leite, R.C.C. (1969). Phys. Rev. Lett. 22: 1304. Kivelson, S. and Gelatt, C.D. Jr., (1979). Phys. Rev. B 19: 5160. Ley, L. (1984). The Physics of Hydrogenated Amorphous Silicon II (eds. J.D. Joanpoulos and G. Lucovsky), 61. Berlin: Springer-Verlag. Spear, W.E. (1988). Amorphous Silicon and Related Materials (ed. H. Fritzsche). Singapore: World Scientific. Aoki, T., Shimizu, T., Komedoori, S. et al. (2004). J. Non-Cryst. Solids 338–340: 456. Aoki, T., Shimizu, T., Saito, D., and Ikeda, K. (2005). J. Optoelectron. Adv. Mater. 7: 137. Aoki, T. (2006). J. Non-Cryst. Solids 353: 1138. Tanaka, K. (1998). Phys. Rev. B 57: 5163. Tanaka, K. (1980). J. Non-Cryst. Solids 35–36: 1023. Tanaka, K. (1990). Rev. Solid State Sci. 4: 641. Tanaka, K. (2001). Handbook of Advanced Electronic and Photonic Materials (ed. H.S. Nalwa), 119. San Diego: Academic Press. Elliott, S.R. (2001). J. Non-Cryst. Solids 81: 71. Kolobov, A.V., Oyanagi, H., Tanaka, K., and Tanaka, K. (1997). Phys. Rev. B 55: 726. Shimakawa, K. (2005). J. Optoelectron. Adv. Mater. 7: 145. Emelianova, E.V., Adriaenssens, G.J., and Arkhipov, V.I. (2004). Philos. Mag. Lett. 84: 47. Hegedüs, J., Kohary, K., Pettifor, D.G. et al. (2005). Phys. Rev. Lett. 95: 206803. Klinger, M.I., Halpern, V., and Bass, F. (2002). Phys. Status Solidi B 230: 39. Singh, J. (1988). Chem. Phys. Lett. 149: 447. Anderson, P.W. (1975). Phys. Rev. Lett. 34: 953. Holstein, T. (1959). Ann. Phys. 8: 325. Singh, J. and Oh, I.-K. (2005). J. Non-Cryst. Solids 351: 1582. Hegedüs, J., Kohary, K., and Kugler, S. (2005). J. Optoelectron. Adv. Mater. 7: 59. Toyozawa, Y. (1981). J. Phys. Soc. Jpn. 50: 1861. Tanaka, K. (1983). J. Non-Cryst. Solids 59–60: 925. Singh, J. (2010). Phys. Status Solidi C 7: 984–987. Singh, J. (2012). Organic Light Emitting Devices (ed. J. Singh). Rijeka, Croatia: Intech, Ch. 1. Shakeel, U. and Singh, J. (2018). Org. Electron. 59: 121–124. Singh, J. (2007). Phys. Rev. B76: 085205. Singh, J., Baessler, H., and Kugler, S. (2008). J. Chem. Phys. 129: 041103.

References

73 74 75 76 77

Endo, A., Sato, K., Yoshimura, K. et al. (2011). Appl. Phys. Lett. 98: 083302. Tremblay, J.-F. (2016). Electron. Mater. 94: 30–34. Dias, F.B., Penfold, T.J., and Monkman, A.P. (2017). Meth. Appl. Fluoresc. 5: 1. Adachi, C. (2013). SID Symp. Dig. Tech. Pap. 44: 513–514. J. Singh (2018). Private communication with C. Adachi at EXCON, July 2018, Nara, Japan.

239

241

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses Sandor Kugler 1 , Rozália Lukács 2 , and Koichi Shimakawa 3 1 2 3

Department of Theoretical Physics, Budapest University of Technology and Economics, H-1521, Budapest, Hungary Norwegian University of Life Sciences, Ås, Norway Department of Electrical and Electronic Engineering, Gifu University, Gifu, 501-1193, Japan

CHAPTER MENU Introduction, 241 Atomic-Scale Computer Simulations of Photoinduced Volume Changes, 243 Effect of Illumination, 244 Kinetics of Volume Change, 245 Additional Remarks, 248 Conclusions, 249 References, 249

8.1 Introduction Chalcogenide glasses (ChGs), which contain group VI elements S, Se, and Te, but no O, exhibit various reversible and irreversible changes in structural and electronic properties during illumination. These properties include photoinduced volume change, photodarkening (PD), photoinduced crystallization and amorphization, photoinduced fluidity, and anisotropic optomechanical effects, etc. These photoinduced changes are unique to glass phases and are not observed in any crystalline structures of chalcogenide glasses. Films of ChGs can either expand (a-As2S3, a-As2Se3, etc.) or shrink (a-GeSe2, a-GeSe2, etc.). The time scale of photoinduced volume changes is in minutes. These processes are several orders of magnitude slower than those in data storage devices (DVDs) that utilize the properties of phase change materials. Data storage is affected by fast, reversible phase changes between crystalline and amorphous states of chalcogenide alloy systems, such as In–Sb–Te, Ge–Sb–Te, Ag–In–Sb–Te, Ge–Sb–Te, etc. In the recent years, interest in amorphous Se, the model material of chalcogenide glasses, has increased due to its applications as, for example, a good flat-panel X-ray image detector. A selenium atom contains six valence electrons (4s2 4p4 ). From s, px , py , and pz , only four independent orbitals can be constructed, that is, two out of four must be doubly occupied. A hydrogen selenide molecule has a H–Se–H molecular structure, with a bond angle of 90∘ . As a first approximation, we can explain the bonding process in Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses

the following way: two near-perpendicular sigma bonds can be constructed by putting an electron in px and another in py atomic orbitals, which are overlapped by 1s orbital of hydrogen atoms. Other two unshared electrons occupy the pz atomic orbital, forming a non-bonding lone pair (LP) state. The remaining two electrons remain untouched by the s atomic orbital. In the condensed phase, selenium atoms can form a chainlike structure. It can be concluded that only two electrons in px and py orbitals form sigma bonds overlapping with similar orbitals of adjacent atoms. Bond angles are slightly larger than 90∘ , because the s atomic orbital is not untouched; a little contribution of s orbital is needed to form near-perpendicular sigma bonds and an LP. The energy of these non-bonding LP states lies somewhere between electron bonding and antibonding energy levels. It was believed for a long time that photoinduced volume changes (usually called photoinduced volume expansion; PVE) and PD (PD: bandgap decreases during illumination) were two different sides of the same phenomenon in amorphous chalcogenides. It was expected therefore that one-to-one correlation should exist between PVE and PD. Systematic studies of PVE has been first performed by Hamanaka, et al. [1], showing that PD always accompanies PVE. Tanaka [2], however, has experimentally shown, from the time evolution of these two phenomena, that the time constants of PD and PVE in a-As2 S3 are different. It has also been observed that PVE saturates earlier than PD, suggesting that these two phenomena are not directly related to each other. A theoretical description of light-induced metastable defects creation, PD and PVE, in amorphous chalcogenides has been reviewed by Singh [3]. To understand the dynamics of PVE and PD during illumination, in-situ measurements have been performed [4–6]. A real-time in-situ surface height measuring system, called the physical vapor deposition (PVD), has been developed using the Twyman–Green interferometer with image analysis technologies [7, 8]. As shown by the dotted line in Figure 8.1, a small amount of PVE is found in (flatly deposited) a-Se films, and the height increases rapidly by 2.5 nm (Δd/d = 0.5%) with illumination (532 nm in wavelength and 90 mW cm−2 of power density). As soon as the light is switched off after 800 seconds of illumination, the height decreases by 2 nm, and then gradually decreases to the original height in 400 seconds (shown by the dotted line in Figure 8.2). 3.5 Change of thickness [nm]

242

3 a

2.5 2 1.5 1 0.5 0

–0.5

0

50

100

150 Time [s]

200

250

300

Figure 8.1 Time development of measured volume expansion in amorphous Selenium (dotted line) [7] and its fit using our model (solid line).

8.2 Atomic-Scale Computer Simulations of Photoinduced Volume Changes

Change of thickness [nm]

3

a

2.5 2 1.5 1 0.5 0 –0.5 800

850

900

950

1000 1050 1100 1150 1200 Time [s]

Figure 8.2 The measured shrinkage of a-Se (dotted line) [7] and the fitted curve (solid line) after switching off the illumination.

8.2 Atomic-Scale Computer Simulations of Photoinduced Volume Changes Some years ago, we have proposed a description of the photoinduced volume changes based on tight-binding (TB) molecular dynamics (MD) simulations on amorphous selenium. A typical configuration can be seen in Figure 8.3. For the description of atomic interactions between selenium atoms, a TB model developed by Molina et al. [9] has been used. The velocity Verlet algorithm has been applied in order to follow the motion of atoms with a time step equal to t = 2 fs. The temperature was controlled via the velocity-rescaling method. Samples containing 162 atoms were prepared by “cook and quench” sample preparation procedure. They had an initial density of 4.33 g cm−3 . We X

Z

Y

Figure 8.3 Snapshot of a glassy selenium network. The sample can move in z-direction. Arrow shows the thickness of the sample.

243

244

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses

prepared them from liquid phase by rapid quenching. At the beginning, we set the temperature of the system to 5000 K for the first 300 MD steps. During the following 2200 MD steps, we linearly decreased the temperature from 700 to 250 K, driving the sample through the glass transition temperature and reaching the condensed phase. Then we set the final temperature to 20 K and relaxed the sample for 15 ps in 7500 MD steps. The closed box shown in Figure 8.3 was opened in the z-direction at the 3000th MD step. In order to model the photoinduced volume changes, the periodic boundary conditions were lifted along the z direction at this point. This procedure provided us the slab geometry with periodic boundary conditions in only two dimensions. After that, the system was relaxed for another 40 000 MD steps (80 ps) to obtain a stable configuration. Samples prepared at 20 K had final densities in the range of 3.95–4.19 g cm−3 . The number of coordination defects ranged from 3% up to 12%. Most of these defects were located on the surfaces. The structure mainly consisted of branching chains, but some rings could also be found. The sample could expand/shrink in the z-direction. The measure of sample thickness is defined as the distance between centers of masses of 15–15 surface atoms on both sides, as shown by the horizontal arrow in Figure 8.3.

8.3 Effect of Illumination Photoexcitation generates an electron–hole pair when a photon is absorbed. This process can be modeled by transferring an electron from the highest occupied molecular orbital (HOMO) to the lowest unoccupied molecular orbital (LUMO). In computer simulations, it was assumed that, immediately after photon absorption, the electron and the hole became separated in space on a femtosecond time scale [10]. The Coulomb interaction between the electron and hole was neglected. The influences of excited electrons and of holes were treated independently. First, an extra electron was added into the LUMO (excited electron creation), and next, an electron in the HOMO (hole creation) was annihilated. When an additional electron was added in the LUMO, a bond-breaking event occurred. In the majority of cases, a covalent bond between a twofold-coordinated atom and a threefold-coordinated atom was broken. However, sometimes, bond breaking between two twofold-coordinated neighbor atoms was also observed. Our localization analysis reveals that the LUMOs are usually localized at such sites (at a two- or three-fold coordinated atom) before bond breaking. The change of bond lengths alternates between shrinkage and elongation in the vicinity of the broken bond due to bond breaking. If the electron on the LUMO is deexcited, then all the bond lengths are restored to their original values. Reversible thickness change in a-Se sample is observed during several photoexcitations. A very different behavior is observed during hole creations [10]. Interchain weak bonds are formed after creating a hole, thereby causing contraction in the sample. This always happens near the atoms where the HOMO is localized. Since the HOMO is usually localized in the vicinity of a onefold-coordinated atom, the interchain bond formation often takes place between a onefold-coordinated atom and a twofold coordinated atom. Sometimes the formation of interchain bonds between two twofold-coordinated atoms has also been observed. Hartree–Fock ab initio Raman spectra calculations and Raman spectroscopic measurements have been carried out on amorphous selenium in order to identify the

8.4 Kinetics of Volume Change

characteristic vibrational mode due to sigma bonds [11]. In the Raman spectra, the peak occurring around 250 cm−1 corresponds to 0.234 nm covalent bonds vibrational modes in the amorphous Se [12]. As the Raman intensity varies in time due to illumination, a large number of covalent bonds break, and they get re-formed again after stopping the illumination. The difference between the Raman spectra of a 10-minute illuminated sample and after-40-minute relaxed sample was significant [11]. The Raman spectrum measurement provides strong experimental evidence for the photoinduced bond breaking process.

8.4 Kinetics of Volume Change 8.4.1

a-Se

Volume expansion and shrinkage are additive quantities; namely, the expansion by a thickness d+ is proportional to the number of excited electrons ne , so that d+ = 𝛽 + ne ; likewise, the shrinkage d− is proportional to the number of created holes nh , so that d− = 𝛽 − nh . The parameter 𝛽 + (𝛽 − ) is the average thickness change caused by an excited electron (hole). The time-dependent thickness change is given by d(t) = d+ (t) − d− (t) = 𝛽 + ne (t) − 𝛽 − nh (t). Assuming that the number of electrons is equal to number of holes, ne (t) = nh (t) = n(t), we get d(t) = (𝛽+ − 𝛽− )n(t) = 𝛽n(t)d(t) = (𝛽+ − 𝛽− )n(t) = 𝛽n(t)

(8.1)

where 𝛽 = 𝛽 + − 𝛽 − is a characteristic constant of chalcogenide glasses related to the photoinduced volume change, and it is a unique parameter for each sample. The sign of this parameter determines whether the material expands or shrinks. The number of photoexcited electrons and holes is proportional to the time duration of a steady-state illumination. Their generation rate G depends on the photon absorption coefficient and the number of incoming photons. After photon absorption, the excited electrons and holes randomly diffuse and eventually recombine. A phenomenological non-linear rate equation for this process is given as: dn(t) (8.2) = G − Cn(t)2 dn(t)∕dt = G − C n(t)2 dt where C is a constant. Using Eq. (8.1), the solution for the time-dependent thickness change in Eq. (8.2) is given by: √ √ G (8.3) tanh( GC t). d(t) = 𝛽 C The prefactor of tanh(x) (marked by “a” in Figure 8.1) is measurable after long-time illumination (tanh(∞) = 1), and, in this limiting situation, only one fitting parameter √ ( GC) remains to be determined. Figure 8.1 shows the measured time evolution of the surface height (dotted line), and the best fit (solid line) is obtained in the interval of 0–300 seconds illumination. The photoinduced expansion of amorphous selenium films was measured in situ using an optoelectronic interference, enhanced by the image processing [7]. After the light is switched off, the rate Eq. (8.2) reduces to: dn(t) = −Cn(t)2 dn(t)∕dt = −C n(t)2 dt

(8.4)

245

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses

with the solution d(t) = a(aCt𝛽 −1 + 1)−1

(8.5)

where a is equal to d(t) when the illumination is switched off. Figure 8.2 shows the measured time evolution of the surface height (dotted line) [7], and the best fit (solid line) is obtained in the interval of 0–400 seconds after switching off the light. 8.4.2

a-As2 Se3

In flatly deposited films of a-As2 Se3 , the surface height increases by 10 nm (Δd/d = 2%) in 200 seconds of illumination by a laser (532 nm in wavelength and 90 mW cm−2 of power density). After 600 seconds, the illumination was switched off. The surface height started decreasing and settled in 200 seconds at 2 nm less than its height before the light was switched-off. In the case of obliquely deposited films, the surface height decreased by 12 nm (Δd/d = 2.4%) in 3 × 104 seconds [7]. Similar results have also been observed in other amorphous chalcogenides. In the ideal case (a-Se), we assumed that each original local structure was reconstructed after the electron–hole recombination. However, the result of a measured volume change on flatly deposited a-As2 Se3 film is quite different from that in the case of ideal Selenium (see Figures 8.4 and 8.5). To explain the difference, we must take into account a large number of irreversible (metastable) changes in the local atomic arrangement—that is, after turning off the light, a part of local configuration remains the same [13]. The total volume change includes both the reversible (drev (t)) and irreversible (dirr (t)) changes, and it can be written as: (8.6)

dtotal (t) = drev (t) + dirr (t).

The reversible part follows Eqs. (8.2) and (8.4) during and after the illumination, with the corresponding solutions given in Eqs. (8.3) and (8.5), respectively. Let us consider the irreversible component. During the illumination, the generation rate of irreversible microscopic change is time dependent. An upper limit exists for the 10 Change of thickness [nm]

246

a+b b

8

Metastable part

6 4 2 0

Transient part

0

100

200

300 Time [s]

400

a

500

Figure 8.4 Time development of volume expansion of a-As2 Se3 . Thin solid line is the measured curve, and thick solid line is the fitted line. Lower dashed curve is the best fit of reversible part, while upper one is that of the irreversible part.

8.4 Kinetics of Volume Change

Change of thickness [nm]

10 b

8 6 4 2 0

1000

1500

2000

Time [s]

Figure 8.5 The measured decay (thin solid line) and the fitted theoretical curves (thick solid line) for the shrinkage as a function of time after the illumination has stopped.

maximum number of electrons and holes causing irreversible changes, and are denoted by ne,irr,max and nh,irr,max . To simplify the derivation, let us assume that ne,irr,max = nh,irr,max ; then, we can write the electron generation rate for the irreversible component as: Ge (t) = Ge,irr (ne,irr,max − ne (t)).

(8.7)

Note that there is no recombination term in Eq. (8.7). In this case, the irreversible expansion is governed by: dirr (t)∕dt = Girr − Cirr dirr (t).

(8.8)

Girr is the generation rate at the beginning of illumination. The solution of Eq. (8.8) is obtained as: dirr (t) = (Girr ∕Cirr ) ( 1 − exp {−Cirr t }).

(8.9)

Here, at the steady state (t = ∞), we get: d(t = ∞) = (Girr ∕Cirr ) = b.

(8.10)

Here, d(t = ∞) = b is a measurable quantity. Now, we have got only two fitting parameters to determine—a in Eq. (8.5) and b in Eq. (8.10). The parameter a in Eq. (8.5) is associated with when the light is switched off, so it is not obtained at t → ∞ as b is obtained. Using Eqs. (8.3) and (8.9) in Eq. (8.6), the best fit for the total volume expansion (dtotal (t)) from Eq. (8.6), reversible (transient) (drev (t)) in Eq. (8.3), and irreversible (metastable) (dirr (t)) in Eq. (8.9) parts are displayed as a function of the illumination time in Figure 8.4. After illumination, there is no volume change caused by the irreversible microscopic effects. Figure 8.5 shows the initial shrinkage after switching off the illumination, which is similar to what is obtained in Figure 8.2 for the reversible case. This model is a good explanation of a measured volume change on flatly deposited a-As2 Se3 , but the result of obliquely deposited film seems to be quite different, as shown in Figure 8.6, where a permanent volume shrinkage was observed. In order to understand this discrepancy, we have made an additional computer simulation for structures of different preparation methods [14]. The final conclusion of the analysis is that oblique

247

8 Photoinduced Bond Breaking and Volume Change in Chalcogenide Glasses

0

Experimental data –A*tanh(B*t)–C*(1–exp(–D*t))

Height [nm]

–2 –4 –6 –8 –10 –12 –14 0

0.5

1

1.5

2 2.5 Time [s]

3

3.5

4 × 104

Figure 8.6 Volume changes in function of time of obliquely deposited a-AsSe thin film. Symbols are the measured volume changes during the illumination, and solid line is the best fit of a d(t) = (−6.4)*tanh(3*10−4 *t) + (−15.2)*(1 − exp(−1.6*10−5 *t)) function. The prefactors of tanh and (1 − exp) have negative values!

samples of about 1000 atoms contain large voids and have a more porous structure than a flatly deposited film. Large voids can collapse due to bond breakings, which can cause shrinkage during light illumination. If we consider that the signs of prefactors in Eqs. (8.2) and (8.9) are negative, we again get an excellent fit with the measured data [15] as it is displayed in Figure 8.6.

8.5 Additional Remarks Instead of Debye-type normal exponential relaxations, several times stretched exponential functions are fitted to the measured data in disordered systems. These curves usually fit quite well, but sometimes the physical understandings of these approaches are absent. In our case, we also tried to fit a stretched exponential function, and we obtained relatively good correlation, as shown in Figure 8.7. This example demonstrates that two different processes can sometimes be fitted by one stretched exponential function. Experimental Experimentaldata data M*(1–exp(–t/N)^P)) M*(1–exp(–t/N)^P))

0 –2 Height [nm]

248

–4 –6 –8 –10 –12 –14 0

0.5

1

1.5

2 2.5 Time [s]

3

3.5

4 × 104

Figure 8.7 Volume changes of the same obliquely deposited a-AsSe thin film fitted by the d(t) = (−20)*(1 − exp(−(t/3.17*104 )0.5 )) stretched exponential function.

References

8.6 Conclusions The photoinduced covalent bond breaking can be observed in chalcogenide glasses, caused by excited electrons, whereas holes contribute to the formation of inter-chain bonds in the vicinity where these excited electrons and holes are localized. The interplay between photoinduced bond breaking and inter-chain bond formation leads to either volume expansion or shrinkage. We have developed a universal macroscopic model of temporal development that is able to describe the photoinduced expansion as well as the shrinkage in chalcogenide glasses prepared by different methods. It is convenient to follow the time development of photoinduced volume change during and after the illumination.

References 1 Hamanaka, H., Tanaka, K., Matsuda, A., and Iijima, S. (1976). Solid State Commun. 2 3 4 5 6 7 8 9 10 11 12 13 14 15

19: 499. Tanaka, K. (1998). Phys. Rev. B 57: 5163. Singh, J. (2007). J. Optoelectron. Adv. Mater. 9: 50. Ganjoo, A., Ikeda, Y., and Shimakawa, K. (1999). Appl. Phys. Lett. 74: 2119–2122. Ganjoo, A., Shimakawa, K., Kamiya, H. et al. (2000). Phys. Rev. B 62: R14601. Ganjoo, A. and Shimakawa, K. (2002). J. Optoelectron. Adv. Mater. 4: 595–604. Ikeda, Y. and Shimakawa, K. (2004). J. Non-Cryst. Solids 539: 338–340. Kugler, S. and Shimakawa, K. (2015). Amorphous Semiconductors. Cambridge University Press. Molina, D., Lomba, E., and Kahl, G. (1999). Phys. Rev. B 60: 6372. Hegedus, J., Kohary, K., Pettifor, D.G. et al. (2005). Phys. Rev. Lett. 95: 206803-1-4. Lukacs, M.V., Shimakawa, K., and Kugler, S. (2010). J. Appl. Phys. 107: 073517–1-5. Dash, S., Chen, P., and Boolchand, P. (2017). J. Chem. Phys. 146: 224506. Hegedus, J., Kohary, K., and Kugler, S. (2006). J. Non-Cryst. Solids 352: 1587. Lukacs, R., Hegedus, J., and Kugler, S. (2008). J. Appl. Phys. 104: 103512. Lukacs, R. and Kugler, S. (2011). Jpn. J. Appl. Phys. 50: 091401.

249

251

9 Properties and Applications of Photonic Crystals Harry E. Ruda and Naomi Matsuura Centre for Nanotechnology, University of Toronto, 170 College Street, Toronto, Canada

CHAPTER MENU Introduction, 251 PC Overview, 252 Tunable PCs, 255 Selected Applications of PC, 260 Conclusions, 265 References, 265

9.1 Introduction Photonic crystals (PCs) are periodic, dielectric, composite structures in which the interfaces between the dielectric media behave as light scattering centers. PCs consist of at least two component materials having different refractive indices, which scatter light due to their refractive index contrast. The one-, two-, or three- dimensional (1D, 2D, or 3D) periodic arrangement of the scattering interfaces may, under certain conditions, prevent light with wavelengths comparable to the periodicity dimension of the PC from propagating through the structure. The band of forbidden wavelengths is commonly referred to as a “photonic band gap” (PBG). Thus, PCs are also commonly referred to as PBG structures. PCs have great potential for providing new types of photonic devices. The continuing demand for photonic devices in the areas of communications, computing, and signal processing, using photons as information carriers, has made research into PCs an emerging field with considerable resources allocated to their technological development. PCs have been proposed to offer a means for controlling light propagation in small, sub-micron-scale volumes – the photon-based equivalent of a semiconductor chip – comprising optical devices integrated together onto a single compact circuit. Proposed applications of PCs for the telecommunication sector include optical cavities, high-Q filters, mirrors, channel add/drop filters, superprisms, and compact waveguides for use in so-called planar lightwave circuits (PLCs). Practical applications of PCs generally require human-made structures, as photonic devices are designed primarily for light frequencies ranging from the ultraviolet to the Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

252

9 Properties and Applications of Photonic Crystals

near-infrared (IR) regime (i.e. ∼100 nm to ∼2 μm, respectively), and PCs having these corresponding periodicities are not readily available in nature. 1D PCs in this wavelength range may be easily fabricated using standard thin film deposition processes. However, 2D and 3D PC structures are significantly more difficult to fabricate, and remain among the more challenging nanometer-scale architectures to realize with cost-effective and flexible patterning using traditional fabrication methodologies. Recently, there has been considerable interest in PC-based devices that has driven advanced fabrication technologies to the point where techniques are now available to fabricate such complex structures reliably on the laboratory scale. In addition to traditional semiconductor nanostructure patterning methods based on advanced patterning/etching techniques developed by the semiconductor industry, novel synthesis methods have been identified for 2D and 3D periodic nanostructured PC arrays. There are several excellent reports reviewing these fabrication techniques in the literature [1–4], and this growing field has already been the subject of numerous recent reviews, special issues, and books in the area of theoretical calculations (both band structure and application simulations), 2D PC structures, 3D PC structures, and opal-based structures [5–7]. Recently, there has been great interest in exploring the use of PCs for applications in the active field of telecommunications, such as in the area of PLCs (e.g., for optical switching). In such applications, the PC properties should be adjustable to create “tunable” PBGs. This development increases the functionality of all present applications of PCs by allowing the devices in such applications to be adjustable, or tunable. We review here recent developments in the engineering of tunable nanometer-scale architectures in 2D and 3D. This review aims to organize this ever-changing volume of information such that interested theorists can design structures that may be easily fabricated with certain materials, and such that technologists can try to meet existing fabrication “gaps” and problems with current systems.

9.2 PC Overview 9.2.1

Introduction to PCs

The simplest PC structure is a multilayer film, periodic in 1D, consisting of alternating layers of material with different refractive indices. Theoretically, this 1D PC can act as a perfect mirror for reflecting light with wavelengths within its PBG, and for light incident normal to the multilayer surface. 1D PCs are found in nature, as seen, for example, in the iridescent colors of abalone shells, butterfly wings, and some crystalline minerals [4], and in human-made 1D PCs (i.e. also known as Bragg gratings). The latter are widely used in a variety of optical devices, including dielectric mirrors, optical filters, and in optical fiber technology. The center frequency and size (i.e. frequency band) or so-called stop band of the PBG depends on the refractive index contrast (i.e. n1 /n2 , where n1 and n2 represent the refractive indices of the first and second materials, respectively) of the component materials in the system. This multilayer film is periodic in the z-direction and extends to infinity in the x- and y- directions. In 1D, a PBG occurs between every set of bands, at either the edge or at the center of the Brillouin zone – PBGs will appear whenever n1 /n2 is not equal to unity [7]. For such multilayer structures, the corresponding PBG diagrams show that

9.2 PC Overview

the smaller the contrast, the smaller the band gaps [7]. In 1D PCs, if light is not incident normal to the film surface, no PBGs will exist. It is also important to note that at long wavelengths (i.e. at wavelengths much larger than the periodicity of the PC), the electromagnetic wave does not probe the fine structure of the crystal lattice and effectively sees the structure as a homogeneous dielectric medium. The phenomenon of light waves traveling in 1D periodic media was generalized for light propagating in any direction in a crystal periodic in all three dimensions, in 1987, when two independent researchers suggested that light propagation in 3D could be controlled using 3D PCs [8, 9]. By extending the periodicity of the 1D PC to 2D and 3D, light within a defined frequency range may be reflected from any angle in a plane in 2D PBG structures, or at any angle in 3D PBG structures. Since the periodicity of the PCs prevents light of specific wavelengths (i.e. those within the PBG) from propagating through them in a given direction, the intentional introduction of “defects” in these structures allows PCs to control and confine light. Propagation of light with wavelengths that were previously forbidden can now occur through such “defect states” located within the PBG. Defects in such PCs are defined as regions having a different geometry (i.e. spacing and/or symmetry) and/or refractive index contrast from that of the periodic structure. For example, in a 2D PC comprising a periodic array of dielectric columns separated by air spaces, a possible defect would include the removal of a series of columns in a line. Specific wavelengths of light forbidden from propagating through defect-free regions would then be able to propagate through the line defect, but not elsewhere. Indeed, by appropriately eliminating further columns, light may be directed to form optical devices, including, for example, a low-loss 90∘ bend in a 2D waveguide, as shown in the theoretical simulation [7]. Clearly, photons controlled and confined in small structures, of size on the order of the wavelength of light, using the extremely tight bend radii offered by PCs would facilitate miniaturization and the fabrication of PLCs [7]. In addition, since the periodicity of the PC gives rise to the existence of band gaps, which change the dispersion characteristics of light at given frequencies, defect-free PCs give rise to other interesting phenomena, including highly dispersive elements, through the so-called superprism effect [10]. Possible designs of PC-based optical devices have been extensively explored using such properties [1, 11], generating much excitement in the optical telecommunications field [6].

9.2.2

Nanoengineering of PC Architectures

Most of the promising applications of 2D and 3D PCs depend on the center frequency and frequency range for the PBGs. A so-called “complete,” “full,” or “true” PBG is defined as one that extends throughout the entire Brillouin zone in the photonic band structure – that is, for all directions of light propagation for photons of appropriate frequency. An incomplete band gap is commonly termed a “pseudo-gap” or a “stop band,” because it only occurs in the reflection/transmission along a particular propagation direction. A complete gap occurs when stop band frequencies overlap in all directions in 3D. The center frequencies and stop band locations of the PBGs critically depend on the unit cell structure [7–9, 12, 13]. In particular, the PC properties depend on the symmetry of the structure (i.e. the unit cell arrangements), the scattering element shape within the unit

253

254

9 Properties and Applications of Photonic Crystals

b) Scattering center shape

a) Structure symmetry

a

z r y x

d) Topology

c) Fill factor e) Refractive index contrast

Figure 9.1 PBG parameters that affect the frequency and associated stop band. In particular, key PBG parameters for the 2D periodic arrangement of one material (i.e. the rods, in gray) embedded in a second material (i.e. air), periodic in the x-y plane (shown in the central figure) include the structure symmetry, scattering center shape, fill factor, refractive index contrast, and topology.

cell, the fill factor (i.e. the relative volume occupied by each material), the topology, and the refractive index contrast (Figure 9.1). It has been shown that a triangular lattice with circular-cross-section scattering elements in 2D, or a face-centered cubic/diamond lattice with spherical scattering elements in 3D, tend to produce larger PBGs [7]. Also, as discussed above, the dielectric contrast is an important determining factor with 2D and 3D PC structures. The lower the structure dimension, the more readily are PBGs manifested since an overlap of the PBGs in different directions is more likely (i.e. something which is a certainty in 1D structures). In 3D structures, calculations have determined that the minimum dielectric contrast (n1 /n2 ) required to obtain a full PBG is about two [2, 12, 14]. For full PBGs, the ideal structure typically consists of a dielectric–air combination, to obtain both the greatest dielectric contrast as well as reduced losses associated with light propagation in optical materials other than air [7]. The impact of the fill factor, structure symmetry, and scattering element shape on the size and location of the PBG is complex [7, 15–18] and will not be discussed in

9.3 Tunable PCs

detail here. However, it is clear that the ability to adjust, or tune, one or more of these parameters and thus tune the PC properties is a very exciting development for future PC device applications. 9.2.3

Materials Selection for PCs

The physical architecture of a given PC is just one of the design considerations – other important ones are the optical properties of the PC materials. In particular, the refractive index of a given material and its electronic band gap determine the performance and appropriate range of frequencies of PC devices fabricated from such a material. For large PBGs, component materials need to satisfy two criteria; first, a high-refractive-index contrast, and second, a high transparency in the frequency range of interest. It may be a challenge to satisfy both the criteria at optical wavelengths. Suitable classes of PC materials include conventional semiconductors and ceramics, since at wavelengths longer than their absorption edge (or electronic band gap), they can have both high refractive indices and low absorption coefficients. In addition, these materials have other very useful electronic and optical properties that may complement the functions served by the presence of the PBG. Another, often overlooked, consideration for PC material selection is the ability to translate the desired “bulk,” single-crystal properties of well-known materials into nanometer-scale PC properties. Such structures typically have extremely large surface areas, and the microstructure of the constituent PC elements must be controlled during fabrication. Consequently, the final properties of the PC elements often vary from those of the bulk properties. This is extremely relevant when considering the functionality of final PC devices, such as those relying on electronic properties (e.g. lasing). Finally, it is interesting to note that there is a strong correlation between the fabrication methods selected and the component materials. For example, “top-down” dry-etching of semiconductors (e.g. Si, GaAs, InP) into nanostructured 2D arrays is well characterized and relatively commonplace, whereas the same cannot be said for ceramic materials. The opposite is generally true for chemical or “bottom-up” synthesis using sol-gel technology, in which ceramic PCs are relatively easy to fabricate when sol-gel based infiltration techniques are used (e.g. in 3D inverse opal fabrication). It is also clear that the fabrication technique primarily determines the PC structure that can be fabricated, which in turn determines the PC properties.

9.3 Tunable PCs Although conventional PCs offer the ability to control light propagation or confinement through the introduction of defects, once such defects are introduced, the propagation or confinement of light in these structures is not controllable. Thus, discretionary switching of light, for example, or re-routing of optical signals, is not available with fixed defects in PCs. There are two approaches that have been pursued to tune the properties of PCs: (1) tuning the refractive index of the constituent materials or (2) altering the physical structure of the PC. In the latter case, the emphasis has been principally on changing the lattice constant, although other approaches are also relevant (i.e. including, for example, the fill factor, structure symmetry, and scattering element shape). In the text below, we discuss the state of the art in both of these areas.

255

256

9 Properties and Applications of Photonic Crystals

9.3.1 Tuning PC Response by Changing the Refractive Index of Constituent Materials We discuss recent progress in using four approaches within this category – namely, tuning the PC response by using (1) light, (2) applied electric fields, (3) temperature or electrical field, and (4) changing the concentration of free carriers (using the electric field or temperature) in semiconductor-based PCs. 9.3.1.1

PC Refractive Index Tuning Using Light

One approach to modifying the behavior of PCs is to use intense illumination of the PC by one beam of light to change the optical properties of the crystal in a nonlinear fashion; this in turn can thus control the properties of the PC for another beam of light. An example of a nonlinear effect includes the flattening of the photon dispersion relation near a PBG, which relaxes the constraints on phase matching for second- and third-harmonic generation [19]. In addition, light location near defects can enhance a variety of third-order nonlinear processes in 1D PCs [20]. A new approach for PBG tuning has been proposed based on photo-reversible control over molecular aggregation, using the photochromic effect in dyes [21]. This can cause a reversible change over the photonic stop band. Structures that were studied included opal films formed from 275-nm-diameter silica spheres, infiltrated with photochromic dye: two dyes were considered, namely, 1,3-dihydro-1,3,3-trimethylspiro-[2H-indol-2,3′ -[3H]-naphth[2,1-b][1,4]oxazine] (SP) and cis-1,2-dicyano-1,2-bis(2, 4, 5-trimethyl l-3-thienyl)ethene (CMTE). For the SP dye–opal, a reversible 15 nm shift in the reflectance spectrum was observed following UV irradiation, which was ascribed to changes in the refractive index due to resonant absorption near the stop band. Smaller shifts (of about 3 nm) were observed for the CMTE dye–opal. The recovery process on cessation of UV illumination was quite slow, taking about 38 s for the SP dye. Nonlinear changes in the refractive index have also been studied in PCs comprising 220-nm-diameter SA polystyrene spheres infiltrated with water [22]. In these studies, the optical Kerr effect was used to shift the PBG. 40 GW cm−2 of peak pump power at 1.06 μm (35 ps pulses at 10 Hz repetition rate) was used to shift the PBG 13 nm. The large optical nonlinearity originates from the delocalization of conjugated π-electrons along the polymer chains, leading to a large third-order nonlinear optical susceptibility. The time response was measured as a function of the delay time between the pump and the probe, and confirmed a response time of several picoseconds. 9.3.1.2

PC Refractive Index Tuning Using an Applied Electric Field

There have been a number of studies focusing on using ferroelectric materials to form PCs [23–26]. The application of an electric field to such materials can be used to change the refractive index and tune the PC optical response. For example, lead lanthanum zirconate titanate (PLZT) inverse opal structures have been fabricated by infiltration using 350-nm-diameter polystyrene sphere templates and annealing at 750 ∘ C [25]. The films were formed on indium tin oxide (ITO)-coated glass to enable electric-field-induced changes in reflectivity to be measured due to the electro-optical effect. Applying voltages of up to about 700 V across films of thickness of about 50 mm only achieved a few nanometers of peak shift, attributed to the very modest changes in

9.3 Tunable PCs

refractive index from the applied field (i.e. from 2.405 to 2.435, as a result of the bias). It should be noted, however, that the intrinsic response of the electro-optical effect in these materials is known to be in the gigahertz range and hence highly suitable for rapid tuning. Other reports include the formation of inverse opal barium strontium titanate (BST) PCs using infiltration of polystyrene opals [23, 24]. BST is most interesting in that it provides a high-refractive-index material and a factor-of-at-least-two-times-higher breakdown field strength than PLZT, and hence offers a much wider range of the applied field for tuning. Reports of the use of other ferroelectric materials include a high-temperature infiltration process for the ferroelectric copolymer, poly(vinylidene fluoride-trifluoroethane), infiltrated into 3D silica opals with sphere diameters of 180, 225, and 300 nm [26]. 9.3.1.3

Refractive Index Tuning of Infiltrated PCs

In this case, the approach has been to consider modulation of the refractive index of a PC infiltrated with a tunable medium – in particular, the most popular approach has been to use liquid crystals to infiltrate porous 2D and 3D PC structures. Such liquid crystals can behave as ferroelectrics whose refractive index may be tuned using either an applied electric field, or by thermal tuning. Reported results for this approach, using 3D inverse opal structures, have been restricted to infiltration of a ferroelectric liquid crystal material into a silicon/air PC [27]. Reported changes in the refractive index of the liquid crystal are from 1.4 to 1.6 under an applied field [27]. However, since the ferroelectric liquid has a higher refractive index than air, the infiltration results in a significant decrease in the refractive index contrast. This means that the original full PBG of the inverse opal silicon PC no longer exists, and the practical utility of the silicon structure as a PC is effectively lost. Theoretical simulations have shown that partial surface wetting of the internal inverse opal surface can retain the full photonic band in silicon [27], but it is questionable whether such a complex structure can ever be practically fabricated. Finally, the presence of ferroelectric liquid crystal surrenders one of the main advantages of the original concept for PC structures: permitting light propagation in air [28]. The temperature tuning of the liquid crystal material was shown to result in very small changes in refractive index (change in n < 0.01 over a 70 ∘ C change) and thus only provide minimal shifts in the transmittance through the PC over a large temperature range [29]. 9.3.1.4 PC Refractive Index Tuning by Altering the Concentration of Free Carriers (Using Electric Field or Temperature) in Semiconductor-Based PCs

An elegant way to rapidly tune the PBG of semiconductor-based PCs is to adjust the refractive index by modulating the free carrier concentration using an ultra-fast optical pulse [30]. Using this approach, the reflectivity of a two-dimensional silicon-based honeycomb PCs with 412 nm air holes (100 μm in length) in a 500 nm periodic array was studied with a pump-probe approach [30]. By varying the delay between the pump and probe beams, the speed of PBG tuning was measured to be about 0.5 ps. The reflectance relaxation (corresponding to the return of the PBG to its original position) occurred on a timescale of 10–100 ns, characteristic of recombination of excess electrons and holes. Although these results are very encouraging, this approach cannot suppress, or correct for, light scattering losses caused by structural imperfections, which remain important considerations for currently fabricated PCs.

257

258

9 Properties and Applications of Photonic Crystals

9.3.2

Tuning PC Response by Altering the Physical Structure of the PC

The second approach that we discuss for tuning the response of a PC is based on changes to its physical structure. We discuss the following approaches for tuning with this approach using (1) temperature, (2) an applied magnetic field, (3) strain/deformation, (4) piezoelectric effects, and (5) micro-electro-mechanical systems (MEMS) [i.e. actuation]. 9.3.2.1

Tuning PC Response Using Temperature

An example of this approach is a study of the temperature tuning of PCs fabricated from self-assembled polystyrene beads [31]. The PBG in these structures was fine-tuned by annealing samples at temperatures from 20 to 100 ∘ C, resulting in a continuous blue shift in the stop band wavelength from 576 to 548 nm. New stop bands appeared in the UV transmission spectra when the sample was annealed above about 93 ∘ C, the glass transition temperature of the polystyrene beads. 9.3.2.2

Tuning PC Response Using Magnetism

An example of this approach is the use of an external applied magnetic field to adjust the spatial orientation of a PC [32]. This can find application in fabricating photonic devices such as tunable mirrors and diffractive display devices. These authors fabricated magnetic PCs by using monodispersed polystyrene beads self-assembled into a ferromagnetic fluid composed of magnetite particles, with particle sizes smaller than 15 nm [32]. On evaporation of the solvent, a cubic PC lattice was formed with the nanoparticles precipitating out into the interstices between the spherical polystyrene colloids. These authors then showed how the template could be selectively removed by calcination or wet etching, to reveal an inverse opal of magnetite – such structures are being proposed for developing magnetically tunable PCs. 9.3.2.3

Tuning PC Response Using Strain

The concept in this case is quite straightforward – deforming or straining the PC changes the lattice constant or arrangement of dielectric elements in the PC with a concomitant change in the photonic band structure. Polymeric materials would appear to be best suited to this methodology, owing to their ability to sustain considerable strains. However, concerns of reversibility and speed of tuning clearly would need to be addressed. Theoretical predictions for the influence of deformation on such systems include a report on a new class of PCs based on self-assembling cholesteric elastomers [33]. These elastomers are highly deformable when subjected to external stress. The high sensitivity of the photonic band structure to strain and the opening of new PBGs are discussed in this paper [34]. Charged colloidal crystals were also fixed in a poly(acrylamide) hydrogel matrix to fabricate PCs whose diffraction peaks were tuned by applying mechanical stress [35]. The PBG shifted linearly and reversibly over almost the entire visible spectral region (from 460 to 810 nm). Modeling of the photonic band structure of 2D silicon-based triangular PCs under mechanical deformation was also reported [35]. The structures considered comprised a silicon matrix with air columns. The authors showed that while a 3% applied shear strain provides only minor modifications to the PBG, uniaxial tension can produce a considerable shift. Other modeling includes a study of how strain can be used to tune

9.3 Tunable PCs

the anisotropic optical response of 2D PCs in the long wavelength limit [36]. Their calculations showed that the decrease in dielectric constant per unit strain is larger in the direction of the strain than normal to it. Indeed, the calculated birefringence is larger than that of quartz. They suggest that strain tuning of this birefringence has attractive application in polarization-based optical devices. To appreciate the sensitivity of such structures to mechanical tuning, it is instructive to refer to some recent work on PMMA inverse opal PC structures that were fabricated using silica opal templates [37]. Under the application of uniaxial deformation of these PCs, the authors found a blue shift of the stop band in the transmission spectrum – the peak wavelength of the stop band shifted from about 545 nm in the undeformed material to about 470 nm under a stretch ratio of about 1–6. Another practical approach that has been applied to physically tuning PC structures is that of thermal annealing. One such study showed how the optical properties of colloidal PCs composed of silica spheres can be tuned through thermal treatment [38]. This was attributed to both the structural and physiochemical modifications of the material on annealing. A shift in the minimum transmission from about 1000 nm (un-annealed) to about 850 nm (annealed at about 1000 ∘ C) was demonstrated, or a maximum shift in Bragg wavelength of about ∼11%. A quite novel application of strain tuning was recently reported [39]. The authors studied 2D PCs composed of arrays of coupled optical microcavities fabricated from vertical-cavity-surface-emitting laser structures. The influence of strain, as manifested by shifts in the positions of neighboring rows of microcavities with respect to each other, corresponded to alternating square or quasi-hexagonal shear strain patterns. For strains below a critical threshold value, the lasing photon mode locked to the corresponding mode in the unstrained PC. At the critical strain, switching occurred between square and hexagonal lattice modes. Finally, there has been a proposal for using strains in a PC to tune the splitting of a degenerate photon state within the PBG, suitable for implementing tunable PC circuits [40]. The principle they applied is analogous to the static Jahn–Teller effect in solids. These authors showed that this effect is tunable by using the symmetry and magnitude of the lattice distortion. Using this effect, they discussed the design of an optical valve that controls the resonant coupling of photon modes at the corner of a T-junction waveguide structure. 9.3.2.4

Tuning PC Response Using Piezoelectric Effects

In this section, we discuss using piezoelectric effects to physically change PC structures and hence tune them. A proposal was made for using the piezoelectric effect to distort the original symmetry of a two-dimensional PC from a regular hexagonal lattice to a quasi-hexagonal lattice under an applied electric field [41]. The original bands decomposed into several strained bands, dependent on the magnitude and direction of the applied field. In the proposed structures, the application of ∼3% shear strain is shown to be suitable for shifting 73% of the original PBG, which they refer to as the tunable bandgap regime. An advantage of such an approach is that such structures are suggested to be capable of operation at speeds approaching the megahertz level. Another report [42] discusses the design and implementation of a tunable silicon-based PBG microcavity in an optical waveguide, where tuning is accomplished using the piezoelectric effect to strain the PC – this was carried out using integrated piezoelectric

259

260

9 Properties and Applications of Photonic Crystals

microactuators. These authors report a 1.54 nm shift in the cavity resonance at 1.56 μm for an applied strain of 0.04%. There have also been reports of coupling piezoelectric-based actuators to PCs. One such report [43] discusses a poly(2-methoxyethyl acrylate)-based PC composite directly coupled to a piezoelectric actuator to study static and dynamic stop band tuning characteristics – the stop band of this device could be tuned through a 172 nm tuning range, and could be modulated at up to 200 Hz. 9.3.2.5

Tuning PC Response Using MEMS Actuation

There have been a number of interesting developments in this field, including PC-based devices comprising suspended 1D PC mirrors separated by a Fabry Perot cavity (gap) [44]. When such structures are mechanically perturbed, there can be a substantial shift in the PBG due to strain. The authors discuss how a suite of spectrally tunable devices can be envisioned based on such structures – these include modulators, optical filters, optical switches, wavelength-division multiplexing (WDM), optical logical circuits, variable attenuators, power splitters, and isolators. Indeed, the generalization of these concepts beyond 1D was discussed in a recent patent [45] and covers tunable PC structures. This report [45], as well as others [46], considered an extension of these ideas to form families of micromachined devices. The latter authors [46] modeled and implemented a set of micromachined vertical resonator structures for 1.55 μm filters comprising two PC (distributed Bragg reflector [DBR]) mirrors separated by either an air gap or semiconductor heterostructures. Electromechanical tuning was used to adjust the separation between the mirrors and hence fine-tune the transmission spectrum. The mirror structures were implemented using strong-index-contrast InP/air DBRs, giving an index contrast of 2.17, and weak-contrast (i.e. an index contrast of 0.5) silicon nitride/silicon dioxide DBRs. In the former case, a tuning range of over 8% of the absolute wavelength was achieved – varying the inter-membrane voltage up to 5 V gave a tuning range of over 110 nm. Similar planar structures are discussed in other papers [47], except where the mirrors are formed from two slabs of PC separated by an adjustable air gap (also see [45]). These structures are shown to be able to perform as either flat-top reflection or all-pass transmission filters, by varying the distance between the slabs, for normally incident light. Unlike all previously reported all-pass reflection filters, based on Gires–Tournois interferometers using multiple dielectric stacks, their structure generates an all-pass transmission spectrum, significantly simplifying signal extraction and optical alignment. Also, the spectral response is polarization independent owing to the 90∘ rotational symmetry of their structure.

9.4 Selected Applications of PC In this section, a selection of some interesting application areas for PCs are discussed focusing on integrated optics or PLCs, with the notable omission of other important areas such as microwave-based PCs and sensing applications (some of which importantly cover the field of biotechnology). Regretfully, these omissions are necessary, as covering any one of the many other fields would inevitably entail another such special review to do the subject justice.

9.4 Selected Applications of PC

To appreciate the application possibilities within this limited scope, recall that some of the key attractive properties that PCs possess for PLCs include, in particular, an ability to strongly confine light [1], as well as unique dispersive properties [10]. The property of confinement may be exploited to realize compact channel waveguides and sharp bends and also achieve high isolation between adjacent channels (i.e. so-called low cross-talk) [7]. Dispersive properties, on the other hand, readily lend themselves to optical functions including wavelength separation (e.g. the so-called superprism effect) as well as pulse shape modification (e.g. pulse compression) [1]. It should be noted that as far as the maturity of technology is concerned, current PLCs are very simple systems. Typically, they are composed of components having sizes in the millimeter to centimeter range, permitting a very limited number of different functions to be integrated together on a chip, which is very much reminiscent of the state of electronic ICs in the early 1960s. 9.4.1

Waveguide Devices

As discussed above, there are a number of different types of waveguide devices that have been developed in PCs. These include both channel waveguides and coupled resonant optical cavity (CROW) structures [48]. The practical implementation of these structures relies on achieving low losses. Planar PCs are able to achieve ∼4–10 dB/unit cell, and so ∼30 dB can be achieved within a few cells. There are three key loss mechanisms to consider in these structures: (1) scattering losses at structural imperfections in the PCs, (2) out-of-plane losses, and (3) losses caused by TE/TM coupling. When considering scattering losses at imperfections, it should be recognized that intrinsically PC waveguides should not suffer any in-plane losses as such structures are designed to prohibit all propagating modes at the operating wavelength. One therefore expects that scattering from such imperfections should have a much lower impact in PC structures than for ridge waveguides having similar dimensions. On the other hand, in multimode waveguides, this type of scattering can excite higher-order modes and therefore become more significant. PLCs are ideally made as two-dimensional devices, with the third dimension assumed theoretically to be infinite. However, in real finite PLC structures, out-of-plane losses from diffraction and scattering provide the dominant loss mechanism in PC waveguides. In the case of structures formed by etching hole arrays into a solid semiconductor structure, losses originate from poor guiding by holes and their depth being insufficiently long. This causes part of the waveguide mode to be scattered away. Such losses can be minimized by increasing the aspect ratio of the etched holes. In TE/TM mode coupling, such losses can occur in cases such as when the PC surrounding a channel waveguide provides a bandgap for only the TE polarization. In this case, once the confined TE mode is converted to TM, it no longer experiences the photonic bandgap and is therefore free to leak out of the channel waveguide and propagate through the crystal. CROWs are quite distinct from conventional waveguides and have no real analogy with traditional guided wave devices. The waveguide is formed through an array of coupled defects in the PC [48]. In some sense, these structures are analogous to quantum electronic structures such as resonant tunneling structures in which extended states are coupled to standing-wave-type cavity states to control the transmission flux through the structure. In the case of CROWs, by varying the type of defects and their spacing,

261

262

9 Properties and Applications of Photonic Crystals

one can control the propagation of light from defect to defect, and this allows for the tailoring of the wavelength response of the structure as well as its group velocity [48]. Some impressive experimental results have been demonstrated for light propagation in such structures. Importantly, such structures overcome some of the problems of the more conventional PC-based waveguiding devices, showing very little propagation loss as well as providing a means to guide light around sharp bends for ultra-compact optical device and systems applications. Such structures have been successfully demonstrated in the microwave regime, where CROWs are real competitors for conventional PC channel waveguides [49]. Although somewhat unglamorous, the problem of coupling fiber to PCs presents one of the biggest engineering challenges for implementing PC-based PLCs in practice. The problem originates from the fact that fibers possess a circular cross section of ∼5–7 μm diameter as distinct from the PC waveguides, which have cross sections of only several hundreds of nanometers. This leads to a severe mode mismatch between the fiber and PC, and is responsible for insertion losses, which can be ∼20–30 dB. Various approaches to solving this problem have been discussed in the literature. 9.4.2

Dispersive Devices

Light guiding in PC-based structures can be supplemented by unique dispersive devices using the same systems. This again provides a distinct advantage for PCs over traditional alternatives for realizing compact, highly functional optical systems for a variety of applications. One of the most important aspects of this is waveguide dispersion using the superprism effect. That is for performing beam steering, multiplexing, and de-multiplexing (MUX/DEMUX) for dense wavelength-division multiplexing (DWDM), for example. Such dispersion effects in periodic structures have been known for some time [50] and were rediscovered in PCs for application in WDM systems. The superprism effect, as described by Kawakami and coworkers [10], uses the band structural asymmetry near the Γ-point, where the wavevector changes much more dramatically with change in frequency in the Γ-K direction as compared with the Γ-M direction. These workers show that an angular dispersion of 50∘ can be observed for a change in input wavelength from 900 to 1000 nm. As discussed above, PC systems are in principle scalable to any wavelength range. Thus, when such a system is scaled to the 1.55 μm regime for telecommunications applications, this dispersive behavior would correspond to ∼2∘ for the 50 GHz channel spacing in a typical DWDM system. If one assumes a series of output waveguides that are laterally separated by 5 μm, a distance of ∼150 μm would be required to separate the channels. This is an impressively small distance compared with current phased array waveguides, where dimensions are typically on the millimeter to centimeter scale. 9.4.3

Add/Drop Multiplexing Devices

Some of the key components in DWDM systems are the so-called add/drop devices, which allow for the selective removal or addition of a particular wavelength channel to an optical data stream, thereby allowing all the channels to be fully utilized. The first PC-based add/drop device was proposed in 1997 [51] and was based on the resonance created by two defects in intersecting PC waveguide devices, enabling coupling

9.4 Selected Applications of PC

between two channel waveguides at the resonance frequency. It should be noted that such devices are closely related to the CROW structures discussed above. By analogy with quantum electronics, whereas CROWs involve a series of coupled cavities, these devices appear as the equivalent of a single isolated quantum well/dot structure coupled to their waveguides. Unfortunately, experimental realization of this approach has proved to be particularly challenging. The main reason for this is that the resonance condition is strongly dependent on the size of the defect, making it impractical to control the dimensions of the defect to the required tolerance. 9.4.4

Applications of PCs for Light-Emitting Diodes (LEDs) and Lasers

The concept of using microstructured mirrors for developing semiconductor laser structures is fairly mature. Periodic microstructures have been commonly used in distributed feedback (DFB) and DBR lasers, as well as for vertical cavity structures (e.g. vertical-cavity-surface-emitting lasers [VCSELs] using Bragg stacks). However, the dielectric contrast in such structures is typically less than 1% as compared with almost 4:1 in PC-based structures. This is particularly significant as it can lead to much shorter interaction lengths of ∼1 μm as compared with hundreds of micrometers for conventional DFB/DBR structures. Importantly, this reveals the possibility of creating edge emitting laser elements with a very small optical volume, as is discussed below. Our discussions begin by considering how PCs can be fruitfully applied to improve the performance of LEDs. The principal active material for LEDs remains semiconductors with reported internal quantum efficiencies as high as 99.7% [52]. The utility of these materials then depends on how efficiently this generated light can be extracted. As a consequence of Snell’s law, only the radiation falling within a small cone (i.e. ∼16∘ for GaAs) can escape – everything else is totally internally reflected. A simple calculation shows that the extraction efficiency scales as 1/4n2 , where n is the semiconductor refractive index. Thus, the small cone of light allowed by Snell’s law represents only a small percentage (e.g. about ∼2%) of the total available solid angle for GaAs, explaining the low observed external quantum efficiency of ∼2–4% in standard commercial LEDs. High-brightness LEDs have recently come to market and are finding applications in display and lighting applications owing to their superior brightness and lifetime compared with incandescent sources. These devices make use of light extracted from more than one facet. Another approach is to place the active layer on a low-index substrate and then roughen it. The resulting surface scatters a large fraction of the light out of the material, and measured efficiencies of as high as ∼30% have been reported [52]. However, one of the most promising approaches to developing LEDs is based on PCs, as these structures offer the potential to improve both the extraction efficiency of the light as well as control over the direction of light emission from the structures. As regards the latter, PCs enable one to suppress emission in unwanted directions, and enhance it in the desired directions. Such structures have been predicted to be suitable for producing light emitters with external quantum efficiencies exceeding 50% [1]. Two main factors for controlling the emission are (1) increased extraction and photon recycling and (2) altering the fundamental emission process. One of the key motivations for studying PCs is their promise for strong spontaneous emission enhancement – i.e. owing to the fact that the spontaneous emission rate of

263

264

9 Properties and Applications of Photonic Crystals

an excited atom can be increased by placing it in a microcavity. In particular, such a microcavity can be formed by creating a defect within a PC structure, as discussed above. The maximum enhancement factor is achieved when the radiating dipole is so oriented as to experience the maximum interaction with the cavity mode. The cavity enhancement factor f P or Purcell factor is given by (6/𝜋 2 )(Q/V ), where Q is the quality factor of the cavity, and V = v/(l/2𝜋)3 is the effective volume of the cavity taking into account the fact that the emitter is usually embedded in a material of index n and that the number of cavity modes is approximately quantized in integral multiples of half wavelengths, 𝜆/2, in a cavity of volume v. PC microcavities offer high-Q microcavities with an exceptionally low mode volume that is suitable for very high spontaneous emission enhancement factors. Recently, it was shown that one could indeed control the timing of light emission from CdSe nanocrystals (i.e. quantum dots) embedded in a titania inverse opal PC. Londahl et al [53] found that by changing the PC lattice constant, one could accelerate/decelerate the rate of spontaneous emission from the quantum dots by as much as a factor of two. Speeding up of the spontaneous emission could lead to more efficient light sources such as LEDs, while slowing it down could help create more efficient solar cells. The application of PCs for developing lasers also appears to be very promising. Typical semiconductor-based lasers usually have between 104 and 105 modes, of which only one is the desired lasing mode. Clearly therefore, a lot of light is wasted before stimulated emission occurs. In addition, since all of these modes are still there after the onset of lasing, they contribute to noise. PC structures offer a means for reducing the number of available modes and thus reducing both the lasing threshold condition and also dramatically reducing the laser noise characteristics. This may be described by reference to the so-called 𝛽-factor, where 𝛽 = Γ𝜆4 /8𝜋vn3 Δ𝜆. 𝛽 quantifies the mode ratio of the desired mode to all the other modes and determines the laser threshold, and Γ is the confinement factor (the ratio of the gain volume to cavity volume). Defects can be used in the PCs to place modes in an otherwise forbidden spectral regime – for example, one can artificially create a mode within the forbidden bandgap by using a defect, and the 𝛽-factor will then assume a value of unity because the single allowed mode becomes the lasing mode. The principle of realizing a laser using a PC structure is quite straightforward then – one needs to design the PC such that the defect frequency coincides sufficiently closely with the gain peak of the respective emitter material. One other interesting development in PCs is related to harnessing surface plasmons. This may be applied to develop sensors, light emitters, or light detecting structures. In particular, we focus on the application of plasmonic structures for light emitters. Plasmonics rely on plasmon-related effects in metallic structures and can produce exciting phenomena when applied to light-emitting devices. Surface plasmons, quanta of electron oscillations at metal–dielectric interfaces, have been used to explain the super-transmission effect where a transmission of ∼4% was measured for a thin metal film perforated with holes comprising ∼2% area fill factor and size well below the cutoff wavelength (150 nm for 1.55 μm light emission) [54]. The best explanation to date for this is that the incident light excites the top surface plasmons, which then couple to the surface plasmon on the other side of the metal film. This latter surface plasmon mode couples to radiation modes, making the whole process appear as if the incident light was directly transmitted through the metal film. Surface plasmon modes also have been observed in small metal particles with the possibility of enhancing interactions between

References

light and other structures, such as for photodetectors found using such an approach. The concept of developing PCs from such plasmonic systems has obvious attractiveness as a means of using such resonances to transfer light from an active semiconductor structure to the outside world, improving light extraction efficiency – one of the key problems for light source design. The design of such systems would rely on matching the plasma resonance frequency to that of the emission, and by using control over the PC structure, control the coupling.

9.5 Conclusions This chapter discussed the principles and properties of PCs, both passive and tunable, with an emphasis on the application of such structures to PLCs. New phenomena in these systems are offering a glimpse of the far-reaching prospects of developing photonic devices that can, in a discretionary fashion, control the propagation of modes of light in an analogous fashion to how nanostructures have been harnessed to control electron-based phenomena. Analogous to the evolution of electronic systems, one can anticipate a path toward development of compact, active, integrated photonic systems based on the technology outlined in this review. We discuss proposals and indeed demonstrations of a wide variety of PC-based photonic devices with applications in areas including communications, computing, and sensing, for example. In such applications, PCs can offer both a unique performance advantage, as well as potential for substantial miniaturization of photonic systems.

Acknowledgments The authors gratefully acknowledge support from NSERC, OCE, CSA, and AFOSR.

References 1 Krauss, T.F. and de la Rue, R.M. (1999). Photonic crystals in the optical

regime – past, present, and future. Progr. Quant. Electron. 23 (2): 51–96. 2 Berger, V. (1999). From photonic band gaps to refractive index engineering. Opt.

Mater. 11 (2/3): 131–142. 3 Norris, D.J. and Vlasov, Y.A. (2001). Chemical approaches to three-dimensional

semiconductor photonic crystals. Adv. Mater. 13 (6): 371–376. 4 Mizeikis, V., Juodkazis, S., Marcinkevicius, A. et al. (2001). Tailoring and characteri-

zation of photonic crystals. J. Photochem. Photobiol. C 2 (1): 35–69. 5 Di Falco, A., Lapine, M., Tassin, P. et al. (eds.) (2003). Photonics and Nanostruc-

tures – Fundamentals and Applications 1 (1): 1–78: the whole edition is dedicated to fundamentals and applications of photonic crystals. 6 Johnson, S.G. and Joannopoulos, J.D. (2002). Photonic Crystals: Road from Theory to Practice. Boston, MA: Kluwer Academic Publishers. 7 Joannopoulos, J.D., Meade, R.D., and Winn, J.N. (1995). PCs: Moulding the Flow of Light. Princeton: Princeton University Press.

265

266

9 Properties and Applications of Photonic Crystals

8 John, S. (1987). Strong localization in certain disordered dielectric super-lattices.

Phys. Rev. Lett. 58 (23): 2486–2489. 9 Yablonovitch, E. (1987). Inhibited spontaneous emission in solid state physics and

electronics. Phys. Rev. Lett. 58 (20): 2059–2062. 10 Kosaka, H., Kawashima, T., Tomita, A. et al. (1998). Superprism phenomena in

photonic crystals. Phys. Rev. B 58 (16): R10096–R10099. 11 Berger, V. (1999). Photonic crystals and photonic structures. Cur. Opin. Sol. St.

Mater. Sci. 4: 209–216. 12 Ho, K.M., Chan, C.T., and Soukoulis, C.M. (1990). Existence of a photonic gap in

periodic dielectric structures. Phys. Rev. Lett. 65 (25): 3152–3155. 13 Satpathy, S., Zhang, Z., and Salehpour, M.R. (1990). Theory of photon bands in

three-dimensional periodic dielectric structures. Phys. Rev. Lett. 64 (11): 1239–1242. 14 Yablonovitch, E. (1993). Photonic band-gap structures. J. Opt. Soc. Am. B 10 (2):

283–295. 15 Y. Xia (Ed.) (2001). Special issue on photonic crystals, Adv. Mater., 13 (6) 16 Anderson, C. and Giapis, K. (1996). Larger two-dimensional photonic band gaps.

Phys. Rev. Lett. 77 (14): 2949–2952. 17 Meade, R.D., Rappe, A.M., Brommer, K.D., and Joannopoulos, J.D. (1993). Nature of

18 19 20 21 22 23

24

25 26

27 28

the photonic band gap: some insights from a field analysis. J. Opt. Soc. Am. B 10 (2): 328–332. Meade, R.D., Rappe, A.M., Brommer, K.D. et al. (1993). Accurate theoretical analysis of photonic band-gap materials. Phys. Rev. B 48 (11): 8434–8437. Martorell, J., Vilaseca, R., and Corbalan, R. (1997). Second harmonic generation in a photonic crystal. Appl. Phys. Lett. 70 (6): 702–704. Inouye, H. and Kanemitsu, Y. (2003). Direct observation of non-linear effects in a one dimensional photonic crystal. Appl. Phys. Lett. 82 (8): 1155–1157. Gu, Z.-Z., Iyoda, T., Fujishima, A., and Sato, O. (2001). Photo reversible regulation of optical stop bands. Adv. Mater. 13 (7): 1295–1298. Hu, X., Zhang, Q., Liu, Y. et al. (2003). Ultrafast three-dimensional tunable photonic crystal. Appl. Phys. Lett. 83 (13): 2518–2520. Soten, I., Miguez, H., Yang, S.M. et al. (2002). Barium titanate inverted opals – synthesis, characterization, and optical properties. Adv. Funct. Mater. 12 (1): 71–77. Matsuura, N., Yang, S., Sun, P., and Ruda, H.E. (2004). Development of highly-ordered, ferroelectric inverse opal films using sol-gel infiltration. Appl. Phys. A Mater. Sci. Process. Li, B., Zou, J., Wang, X.J. et al. (2003). Ferroelectric inverse opals with electrically tunable photonic band gap. Appl. Phys. Lett. 83 (23): 4704–4706. Xu, T.B., Cheng, Z.Y., Zhang, Q.M. et al. (2000). Fabrication and characterization of three dimensional periodic ferroelectric polymer-silica opal composites and inverse opals. J. Appl. Phys. 88 (1): 405–409. John, S. and Busch, K. (1999). Photonic bandgap formation and tunability in certain self-organizing systems. J. Lightwave Technol. 17 (11): 1931–1943. Joannopoulos, J.D. (1996). The almost-magical world of photonic crystals. Braz. J. Phys. 26 (1): 53–67.

References

29 Yoshino, K., Kawagishi, Y., Ozaki, M., and Kose, A. (1999). Mechanical tuning of the

30

31

32 33 34

35 36 37 38

39 40 41 42 43 44 45 46

47

48

optical properties of plastic opal as a photonic crystals. Jpn. J. Appl. Phys. 38 Part 2 (7A): L786–L778. Leonard, S.W., van Driel, H.M., Schilling, J., and Wehrspohn, R.B. (2002). Ultrafast band edge tuning of a two dimensional silicon photonic crystal via free carrier injection. Phys. Rev. B 66: 161102-1–111102-4. Gates, B., Park, S.H., and Xia, Y. (2000). Tuning the photonic bandgap properties of crystalline arrays of polystyrene beads by annealing at elevated temperatures. Adv. Mater. 12 (9): 653–656. Gates, B. and Xia, Y. (2001). Photonic crystals that can be addressed with an external magnetic field. Adv. Mater. 13 (21): 1605–1608. Bermel, P.A. and Warner, M. (2001). Photonic bandgap structure of highly deformable self-assembling systems. Phys. Rev. E 65 (1), 010702(R) – 1 to 4). Iwayama, Y., Yamanaka, J., Takiguchi, Y. et al. (2003). Optically tunable gelled photonic crystal covering almost the entire visible light wavelength region. Langmuir 19 (4): 977–980. Jun, S. and Cho, Y.-S. (2003). Deformation-induced bandgap tuning of 2D silicon-based photonic crystals. Opt. Express 11 (21): 2769–2774. Kee, C.-S., Kim, K., and Lim, H. (2003). Tuning of anisotropic optical properties of 2D dielectric photonic crystals. Physica B 338: 153–158. Sumioka, K., Kayashima, H., and Tsutsui, T. (2002). Tuning the optical properties of inverse opal photonic crystals by deformation. Adv. Mater. 14 (18): 1284–1286. Miguez, H., Meseguer, F., Lopez, C. et al. (1998). Control of the photonic crystal properties of fcc-packed sub-micrometer SiO2 spheres by sintering. Adv. Mater. 10 (6): 480–483. Pier, H., Kapon, E., and Moser, M. (2000). Strain effects and phase transitions in photonic crystal resonator crystals. Nature 407: 880–883. Malkova, N. and Gopalan, V. (2003). Strain tunable optical valves at T-junction waveguides in photonic crystals. Phys. Rev. B 68, 245115-1 to 6. Kim, S. and Gopalan, V. (2001). Strain tunable photonic band gap crystals. Appl. Phys. Lett. 78 (20): 3015–3017. Wong, C.W., Rakich, P.T., Johnson, S.G. et al. (2004). Strain-tunable silicon photonic bandgap microcavities in optical waveguides. Appl. Phys. Lett. 84 (8): 1242–1244. Foulger, S.H., Jiang, P., Lattam, A. et al. (2003). Photonic crystal composites with reversible high-frequency stop band shifts. Adv. Mater. 15 (9): 685–689. Rajic, S., Corbeil, J.L., and Datskos, P.G. (2003). Feasibility of tunable MEMS photonic crystal devices. Ultramicroscopy 97: 473–479. N. Matsuura, H.E. Ruda and B.G. Yacobi (2015). Configurable photonic device, US patent 6,961,501. Hiller, H., Daleiden, J., Prott, C. et al. (2002). Potential for a micromachined actuation of ultra-wide continuously tunable optoelectronic devices. Appl. Phys. B75: 3–13. Suh, W. and Fan, S. (2003). Mechanically switchable photonic crystal filter with either all-pass transmission or flat-top reflection characteristics. Opt. Lett. 28 (19): 1763–1765. Yariv, A., Xu, Y., Lee, R.K., and Scherer, A. (1999). Coupled-resonator optical waveguide: a proposal and analysis. Opt. Lett. 24: 711–713.

267

268

9 Properties and Applications of Photonic Crystals

49 Bayindir, M., Temelkuran, B., and Ozbay, E. (2000). Propagation of photons

50 51 52

53

54

by hopping: a waveguiding mechanism through localized coupled cavities in three-dimensional photonic crystals. Phys. Rev. B61: R11855–R11858. Russell, P.S.J. (1986). Interference of integrated Floquet-Bloch waves. Phys. Rev. A33 (5): 3232–3242. Fan, S., Villeneuve, P.R., Joannopoulos, J.D., and Haus, H.A. (1998). Channel drop filters in photonic crystals. Opt. Express 3 (1): 4–11. Schnitzer, I., Yablonovitch, E., Caneau, C., and Gmitter, T.J. (1993). Ultrahigh spontaneous emission quantum efficiency, 99.7% internally and 72% externally, from AIGaAs/GaAs/AlGaAs double heterostructures. Appl. Phys. Lett. 62 (2): 131–133. Londahl, P., van Driel, A.F., Nikolaev, I.S. et al. (2004). Controlling the dynamics of spontaneous emission from quantum dots by photonic crystals. Nature 430 (7000): 654–657. Ebbesen, T.W., Lezec, H.J., Ghaemi, H.F. et al. (1998). Extraordinary optical transmission through sub-wavelength hole arrays. Nature 391 (6668): 667–669.

269

10 Nonlinear Optical Properties of Photonic Glasses Keiji Tanaka∗ Department of Applied Physics, Graduate School of Engineering, Hokkaido University, Sapporo, Kita-ku, N13W8, Japan

CHAPTER MENU Introduction, 269 Photonic Glass, 271 Nonlinear Absorption and Refractivity, 272 Nonlinear Excitation-Induced Structural Changes, 280 Conclusions, 285 Addendum: Perspectives on Optical Devices, 286 References, 288

10.1 Introduction In this article, we define photonic glass as a high-purity inorganic glass that can be applied to photonics. There are many types of applications, the most innovative undoubtedly being optical fibers, which now densely surrounds the earth for communication purposes. Optical fiber is also used in functional devices such as Bragg reflectors, optical amplifiers, and lasers. In addition, integrated glass waveguides are extensively developed for optical signal processing such as wavelength divisions. In these applications, not only linear but also nonlinear optical properties play important roles [1–3]. On the one hand, large nonlinearities are needed for all-optical switches, power stabilizers, soliton fibers, supercontinuum generators, etc. [4, 5]. Besides, nonlinear optical excitations are promising for fabricating three-dimensional memories and waveguide devices [4]. On the other hand, we may need a small nonlinearity in high-intensity glass lasers and non-dispersive optical fibers [1, 2]. Despite the wide range of applications, glass science is still less mature than crystal science [6–11]. This is because of three inherent characteristics of the glass . First, due to the lack of long-range structural periodicity, x-ray diffraction patterns are less informative, and explicit atomic structures cannot be determined. Second, also due to the lack of periodicity, electron wave functions and atomic vibrations tend to localize, with the result that the wave numbers are no longer good quantum parameters. Bloch wave functions cannot be postulated, in principle. The conventional one-electron *Emeritus Professor (Email address: [email protected]). Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

270

10 Nonlinear Optical Properties of Photonic Glasses

approximation is largely limited, and only the energy density of states could have physical meanings. Under these circumstances, we may analyze types of many-body problems for electrons and atoms using computer simulations. However, simulated results cannot necessarily provide universal insights because of restricted calculating conditions, such as the enormously fast quenching speeds employed in digitized molecular dynamics. Third, the glass is thermodynamically quasi-stable, having glass-transition temperatures that vary with preparation procedures and post-preparation treatments, as exemplified in the As2 S3 case [12]. The quasi-stability also causes variations in macroscopic properties [7–11]. It is difficult to envisage an ideal glass structure, even in a statistical sense, which is in marked contrast to the situations in ideal crystals and gases, which are perfectly periodic and completely random, respectively [11]. These three features still pose many unresolved problems. As for linear optical properties, for instance, we cannot yet give definite interpretations of the origins of the Tauc optical gap [6, 11], minimal Urbach-edge energy of ∼50 meV [11, 13], midgap absorption [7, 14], etc. For optical nonlinearity in photonic glasses, although we have obtained substantial amount of data for bulk samples, only rough theoretical frameworks have been outlined [4, 5]. Here, we will review three subjects pertaining to optical nonlinearity in photonic glasses. After a brief summary of the photonic glass in Sections 10.2, Section 10.3 focuses on a unified understanding of the third-order nonlinearity. Optical nonlinearity in glasses is analyzed using semiconductor terminology, which may be complementary to the dielectric approach [4] widely employed among glass scientists. In the present scheme, as illustrated in Figure 10.1, the relationship among atomic structures (atom and bonding), electronic structures, and optical absorption (𝛼, Eg , and 𝛽) and refractivity (n0 and n2 ) spectra becomes clearer. Spectral dependence is obtained in a straightforward way. For other nonlinear properties such as the second-order nonlinearity in poled glasses [15] and the large nonlinearity in nano-particle dispersed glasses [16], the reader may refer a recent article by the author [17]. In Section 10.4, we will consider the role of nonlinear excitations in photo-induced phenomena. It is often asserted that when a photo-induced phenomenon is induced by light of photon energy less than the bandgap (ℏ𝜔 < Eg ), nonlinear excitation takes place. However, this may be a hasty conclusion, n0 (3) α, Eg

Boling (6)

Moss

n2 (5)

(7) Sheik – Bahae

(2)

β (4)

electronic structure

bonding

atoms

Figure 10.1 Relations between atomic structure (atoms and bonding), electronic structure, absorption spectra (𝛼 and 𝛽), and refractive index spectra (n0 and n2 ). n0 , 𝛼 and E g are linear properties, and n2 and 𝛽 are nonlinear. Solid arrows represent Eqs. (10.2) and (10.4); double-solid arrows represent linear and nonlinear Kramers–Krönig relations, Eqs. (10.3) and (10.5); and dashed arrows show the Moss rule (E g ∼ n0 ), Boling’s relation, Eq. (10.6), and Sheik-Bahae’s relations including Eq. (10.7).

10.2 Photonic Glass

because it neglects midgap states, whose existence is inherent to the glasses. We will see that photoexcitation mechanisms in glasses are not as simple as those in crystals. Section 10.5 presents conclusions. Finally, the Addendum sketches the personal views of the author on photonic devices from the standpoint of future perspectives.

10.2 Photonic Glass Several types of photonic glasses are available at present, which include oxides [7, 9], chalcogenides [3, 8–11], halides [18], and their mixtures. As is well known, studies on nonlinear effects in halide glasses have been limited due to their smallness [5, 17–20], despite the wider transparency in wavelength, and accordingly, these will not be included in this article. On the other hand, both oxide and chalcogenide glasses consist of the group VIb (16) atoms (O, S, Se, and Te), which are characterized by s2 p4 valence-electron configurations, so that their atomic and electronic structures can be considered in a unified way [3, 6, 11]. Figure 10.2 shows schematic atomic networks in SiO2 and As2 S3 . The structure of such simple glasses can be grasped at four atomic scales [6, 11]. First, it is demonstrated that the so-called short-range structure, which includes the atomic coordination number, bond length, and bond angle, is similar to that in the corresponding crystal. Second, the medium-range order, which encompasses atomic structures of 0.5–3 nm, is believed to exist, but the actual structures have not been elucidated, because of limited experimental methods. Small rings, such as 3-membered Si—O rings (Figure 10.2a), probably exist in SiO2 [6, 7, 21], and distorted layer structures (Figure 10.2b) are proposed for As2 S(Se)3 and GeS(Se)2 glasses [8, 10, 11]. Third, there are point-like defects such as E’ centers (Figure 10.2a) in SiO2 and D0 in GeS2 , which are neutral dangling bonds with typical densities of ppm levels so that their characteristics (a)

(b)

Figure 10.2 Atomic structures of (a) SiO2 and (b) As2 S3 glasses. In (a), Si and O are shown as a solid and an open circle with four- and twofold coordination. In (b), As and S are shown as a solid and as open circles with three- and twofold coordination. The bond lengths are 0.16 nm in (a) SiO2 and 0.23 nm in (b) As2 S3 , so that the side lengths of these illustrations are 2–3 nm. Note that near the centers, (a) includes a small ring and an E’ center and (b) contains wrong bonds (As—As and S—S). In the PbO-SiO2 glass, Si atoms in SiO2 (a) are replaced by Pb with some changes in their atomic coordination.

271

272

10 Nonlinear Optical Properties of Photonic Glasses

have conventionally been analyzed through electron-spin signals [6–11]. In covalent stoichiometric glasses such as As2 S3 , Raman-scattering spectra show unambiguously wrong bonds (Figure 10.2b), i. e. homopolar bonds in stoichiometric glasses, which exist with typical density of a few at.% [6, 8, 11, 12]. Several other defects such as charged defects (D+ and D− ) have been proposed [6–11], which still remain only good working hypotheses, due to difficulties in experimentally confirming their existence. Lastly, it is plausible that the glass structure is inhomogeneous or fluctuating at scales of ∼10 nm or wider [11, 13, 22], specifically in multi-component systems, although such structures have been ruled out in many cases, which may cause intricate problems. Constituent atoms with noncrystalline connectivity determine the electronic structure at three levels [6, 11]. First, short-range bonding governs the optical bandgap energy Eg , which is in the range 4–10 eV for oxides and 1–3 eV for chalcogenides (see Figure 10.6). Accordingly, the latter are regarded as a type of amorphous semiconductors, although the sulfide is electrically insulating. For simple stoichiometric glasses such as SiO2 and As2 S3 , the origins of the valence and conduction bands are known, as will be described in Section 10.3.2. The bandgap energy is a major factor in determining the refractive index n0 of a material, which is ∼1.7 in oxides and ∼2.5 in chalcogenides, which is consistent with the empirical Moss rule, n0 4 Eg ≈ 77 [23]. Second, the medium-range and inhomogeneous structures are assumed to affect the density of states at the band edges, which govern the steepness of the exponential Urbach edges [13]. Third, defects such as dangling bonds and wrong bonds are likely to produce gap states, which are responsible for residual optical absorption [12, 24]. We here emphasize that defect-induced absorption is inherent to glasses. For an ideal dielectric crystal, which is conceptually obtained through infinitesimally slow cooling of a melt to 0 K, we can envisage complete transparency arising ultimately from zero-gap states. By contrast, the glass is prepared through freezing at around the glass-transition temperature, so that it necessarily contains structural disorders such as strained normal bonds and defects. The defective bonds are likely to produce gap states, which give rise to midgap absorption peaks in oxides [6, 7] and weak absorption tails in chalcogenides [11, 14].

10.3 Nonlinear Absorption and Refractivity 10.3.1

Fundamentals

The polarization P induced by an electric field E in a material with electric susceptibility 𝜒 can be written in cgs units as [4, 5]: P = 𝜒 (1) ∶ E + 𝜒 (2) ∶ E ⋅ E + 𝜒 (3) ∶ E ⋅ E ⋅ E,

(10.1)

where the first term 𝜒 :E depicts the conventional linear response, in which the complex parameter 𝜒 (1) represents the linear refractive index n0 and the absorption coefficient 𝛼, and other terms give nonlinear responses. The overall features are illustrated on a frequency axis in Figure 10.3. Here, we should note that the glass is macroscopically isotropic (centrosymmetric) and accordingly cannot provide even-order nonlinear effects arising from 𝜒 (2) :E⋅E, with notable exceptions in poled glasses [4, 5, 15–17]. (Such added nonlinearities are in general smaller than those in selected crystals, while long (1)

10.3 Nonlinear Absorption and Refractivity

χ(1)

χ(2)

ω

0





χ(3)

Figure 10.3 Roles of electric susceptibilities, 𝜒 (1) , 𝜒 (2) , and 𝜒 (3) in the frequency scale upon excitation with light of frequency 𝜔.

and wide samples, including fibers and film forms, may compensate for the smallness.) Then, in the ordinary case, the third-order nonlinearity 𝜒 (3) :E⋅E⋅E becomes the most important nonlinear term, and in some cases the fifth-order effects may appear as well [25–27]. The third-order nonlinearity provides at least three effects [4, 5]. The first effect is third-harmonic generation (see Figure 10.3). Suppose the frequency of a light field E is 𝜔, then E3 contains the term 3𝜔, and this component may be detected if a phase-matching condition for the fundamental and the overtone light, n0 (𝜔) = n0 (3𝜔), could be fulfilled. However, in isotropic materials such as glass, the condition in general cannot be satisfied, resulting in low generation efficiency. In addition, if 𝜔 is at near infrared, 3𝜔 is likely to be located at ultraviolet, at which frequency a glass may not be transparent. Accordingly, such overtone generation has limited applications in practice [28, 29]. The second effect is the so-called optical Kerr effect, including self-(de) focusing, which arises from the intensity-dependent refractive index change n2 I, where I is the light intensity. Such a change appears because the third term of Eq. (10.1), 𝜒 (3) :E⋅E⋅E, can be rewritten as n2 I E, in which n2 is proportional to 𝜒 (3) ; n2 (cm2 /W) ≈ 0.04𝜒 (3) (esu)/n0 2 [5]. The third effect is two-photon absorption, which is schematically illustrated in Figures 10.4b and b’. As is well known, the absorbed light intensity is proportional to the time average of E⋅dP/dt, which contains the term of 𝛽I 2 (∝ E4 ), i.e. two-photon absorption, where 𝛽 is related to the imaginary part of 𝜒 (3) . These two quantities, n2 I and 𝛽I 2 with the proportionality factors of n2 and 𝛽, appear to be very useful in controlling light at optical communication frequencies. Quantitatively, however, 𝜒 (3) in bulk glasses is very small [1–5]. For instance, as described later, n2 ≤ 10−15 m2 W−1 , so the condition n0 = n2 I can be fulfilled only CB

CB

CB

CB

CB

VB

VB

VB

VB

VB

(a)

(a′)

(b)

(b′)

(c)

Figure 10.4 Schematic illustrations of (a) one-photon, (b) two-photon, (b’) resonant two-photon, and (c) two-step absorptions from the valence band (VB) to the conduction band (CB) in a semiconductor. (a’) shows a midgap absorption. Note that the resonant two-photon absorption (b’) is a one-step process, which occurs resonantly with midgap states, while the two-step absorption (c) consists of two successive one-photon processes.

273

274

10 Nonlinear Optical Properties of Photonic Glasses

when incident light is as intense as ∼1015 W m−2 ≈ 102 GW cm−2 , which may damage the glass (see Figure 10.10). To suppress such a breakdown, we could focus ultrashort light pulses onto a small spot (∼μm) and propagate them in long (or thick) samples, such as optical fibers and waveguides. Besides, promising tactics for obtaining efficient nonlinearity are to employ resonant enhancements of the electric field using excitons, plasmons, and/or advanced device structures [1–3, 17]. Alternatively, we may adopt less stringent modulation criteria, an example of which will be described later. Optical responses are formulated in terms of quantum mechanics as follows (see Figure 10.1): The one-photon absorption (Figures 10.4a and a’) coefficient 𝛼(ℏ𝜔) for amorphous materials is conventionally written as [6, 11] 𝛼(ℏ𝜔) ∝ ∣< 𝜑f ∣ H|𝜑i >|2



Df (E + ℏ𝜔)Di (E) dE,

(10.2)

where H is the photon–electron interaction Hamiltonian, the simplest form being proportional to r; 𝜑(r) is the electron wave function, D(E) is the density of states, and the subscripts i and f represent an initial and a final state. Here, the momentum conservation has been neglected taking localized wave functions into account, and the transition matrix amplitude | < 𝜑f |H|𝜑i > | is assumed to be independent of ℏ𝜔 for simplicity. The refractive index spectrum n0 (ℏ𝜔) can then be calculated from 𝛼(ℏ𝜔) using a Kramers–Krönig relation as [30] n0 (𝜔) − 1 = (c∕π)℘



{𝛼(Ω)∕(Ω2 − 2 )} dΩ,

(10.3)

where ℘ denotes the principal value of the integral. Since the absorption spectrum 𝛼(𝜔) is governed by Eg , the Moss rule (n0 4 Eg ≈ 77) [23] works as the simplest approximation of Eq. (10.3) (see Figure 10.8). When photons with ℏ𝜔 ≈ Eg /2 are absorbed, we may envisage two-photon absorption, as illustrated in Figure 10.4b. Under similar assumptions as those employed in deriving Eq. (10.2), the two-photon absorption coefficient 𝛽(ℏ𝜔) can be written as [30, 31] 𝛽(ℏ𝜔) ∝ ∣



< 𝜑f ∣ H ∣ 𝜑n >< 𝜑n ∣ H|𝜑i >∕(Eni − ℏ𝜔) |2

n



Df (E + 2ℏ𝜔)Di (E) dE, (10.4)

for degenerate cases, i.e. when the two photons have the same energy. Here, Eni = En – Ei , where the subscript n represents an intermediate state with energy En . The two-photon absorption governs the intensity-dependent refractive index n2 , which can be calculated using a nonlinear Kramers–Krönig relation for a nondegenerate case. In the present context of degenerate cases, we may express it as a rough approximation in the form [4] n2 (𝜔) ≈ (c∕π)℘



{𝛽(Ω)∕(Ω2 − 𝜔2 )}dΩ,

(10.5)

which appears to be practically useful if 𝛽(Ω) is located in a narrow region at sufficiently higher frequencies than a calculated n2 (𝜔) spectrum, an example of which is shown later. Figure 10.5 shows typical linear and nonlinear spectra of absorption and refractivity for a direct-gap semiconductor with a bandgap energy of Eg [4, 31, 32]. The absorption edge of multi-photon absorption processes is located at Eg /m, where m is the number of photons simultaneously absorbed. Accordingly, two-photon absorption rises at Eg /2.

10.3 Nonlinear Absorption and Refractivity

Figure 10.5 Schematic illustrations of absorption spectra, 𝛼 and 𝛽, and refractivity spectra, n0 and n2 , governed by one- (𝛼 and n0 ) and two-photon (𝛽 and n2 ) processes in an ideal crystal. Excitonic effects are neglected.

n0

α β

n2 0

0

Eg/2

Eg ћω

Note that, in ideal crystalline semiconductors, all the electronic states in the vicinity of band edges are extended, so that a simple band theory can be applied. This situation also applies for indirect-gap semiconductors, although the spectral shape markedly changes [33, 34]. However, the situation becomes intricate in a disordered material. That is, midgap states with localized wave functions influence the absorption in at least three ways. First, the state may work as a site for linear absorption, as illustrated in Figure 10.4a’. The second is through resonant two-photon absorption. If a gap state that satisfies Eni − ℏ𝜔 = 0 in Eq. (10.4) exists, the two-photon absorption at around the gap state can resonantly be enhanced, as in Figure 10.4b’. Third, as illustrated in Figure 10.4c, a midgap state may cause two-step absorption, a successive one-photon absorption process, which can occur if the lifetime of the intermediate states is long enough [35]. The two-step absorption is assumed to occur at the Urbach-edge region in As2 S3 [36] and to govern a defect formation process in SiO2 [24]. It is needless to say that not only two-, but multi-photon absorptions can be envisaged in similar ways. However, theoretical considerations regarding these processes in glasses, including midgap states, still remain to be fully explored. 10.3.2

Two-Photon Absorption

Figure 10.6 shows one- and two-photon absorption (attenuation) spectra of SiO2 [34, 37], As2 S3 [36], and Se [38]. We see that these three glasses possess the same qualitative features. In the one-photon spectra, the Urbach edge appears at ℏ𝜔 ≥ 8.0, 1.9, and 1.4 eV in SiO2 , As2 S3 , and Se, which are consistent with the Tauc gaps at Eg ≈ 9, 2.4, and 2.0 eV [6, 11], respectively. Below the edge, residual absorption (attenuation) seems to exist. However, as exemplified for SiO2 , the spectral shapes have not been reproduced among several studies, probably reflecting varying absorption due to defects and impurities and also light scattering due to structural fluctuations. On the other hand, two-photon absorption appears at ℏ𝜔 ≥ Eg /2, which may have peaks at ∼5.8, ∼2.0, and ∼1.5 eV, respectively. It is interesting to note that, consistent with the absorption spectrum in amorphous Se, Enck [39] has detected two-photon photocurrents using 1.2 eV light pulses.

275

10 Nonlinear Optical Properties of Photonic Glasses

103 x = 68 x = 38

α/cm–1, β/cm·(GW)–1

102 101

SiO2

100

Se

10–1 PbO–SiO2

10–2 10–3

As2S3 0

1

2

3

4 5 6 Photon energy/eV

7

8

9

Figure 10.6 Spectral dependence of the one- and two-photon absorption coefficients, 𝛼 (dashed lines) and 𝛽 (solid lines), respectively, in SiO2 , As2 S3 , Se, and two xPbO-(100-x)SiO2 glasses with x = 38 and 68. 𝛼 in SiO2 at ℏ𝜔 ≤ 8 eV may be influenced by light scattering, where some different spectra have been reported, so that 𝛼 can be read as attenuation.

The spectral features described above can be understood as follows [40]: In SiO2 , as illustrated in Figure 10.7, the top of the valence band is composed of the lone-pair electrons of O 2p states, and the bottom of the conduction band is governed by anti-bonding orbitals consisting of Si 3p and 3d states [6, 7]. The symmetry of the initial and final wave functions is different, i.e. ≠ 0, and accordingly, one-photon absorption can occur from the valence band to the conduction band. The fact that the two-photon spectrum appears at an energy around half of the optical gap (∼9 eV) suggests that the two-photon transition also occurs from the valence band to the conduction band. Similar interpretations are applicable to As2 S3 and Se as well. The spectral shape of the two-photon absorption shown in Figure 10.6 may resemble that of the theoretical curve in Figure 10.5, although the correspondence is not accurate. Specifically, the peak at ∼2.0 eV in As2 S3 may be deceptive, since the decrease in 𝛽 at ℏ𝜔 ≥ 2.0 eV occurs with distortion of light pulses, suggesting two-step absorption [36]. In addition, the observed peaks in SiO2 and Se are substantially sharper than the corresponding theoretical ones shown in Figure 10.5, where the vertical axis is plotted on a linear scale. The reasons for this discrepancy remain to be studied.

Cu 3d

DOS

276

Pb 6s

O 2p

Pb 6p

Na 3s Si 3p 3d

10 eV Energy

Figure 10.7 Density of states (DOS) of the valence (O 2p) and conduction (Si 3p and 3d) bands in SiO2 . Also added are Cu, Pb, and Na states in SiO2 . Occupied states are shaded.

10.3 Nonlinear Absorption and Refractivity

We also see in Figure 10.6 that the 𝛽 spectra are exponential, ∼exp(ℏ𝜔/E𝛽 ), below their peaks. Besides, in As2 S3 at ℏ𝜔 ≈ 1–2 eV, we notice an interesting correlation: the exponential 𝛽 spectrum appears nearly parallel to the weak absorption tail, which has the form 𝛼 ∝ exp(ℏ𝜔/EW ) [11, 13]; E𝛽 (≈ 150 meV) ≈ EW (≈ 200–300 meV) [36]. Note that this comparison is meaningful, since all the data of As2 S3 in the figure have been obtained using a single, high-purity ingot. Although such correlations may also exist in Se and SiO2 , the limited experimental spectra cannot confirm their existence. The nearly parallel exponential spectra in As2 S3 , E𝛽 ≈ EW , suggest the important role of gap states. That is, the two-photon absorption process is likely to occur resonantly with the gap states, which give rise to the weak absorption tail in the linear absorption. We can derive the rough equality as follows: In Eq. (10.4), approximating 1/(Eni − ℏ𝜔) as ∑ 𝛿(Eni − ℏ𝜔), we see that n 1/(Eni − ℏ𝜔) behaves as a density of states, which could imply 𝛽 ∝ Dn (E), where Dn (E) is the density of state of the intermediate (gap) states. On the other hand, the gap state can also govern the final state Df (E) in the linear absorption, given by Eq. (1.2), which may lead to 𝛼 ∝ Df (E). These two relations are consistent with the observation E𝛽 ≈ EW . Figure 10.6 also includes the spectra for PbO-SiO2 glasses, which present three notable features [41, 42]. First, 𝛼(ℏ𝜔) has an exponential Urbach edge at around ℏ𝜔 ≈ 3 eV (𝛼 ≈ 1 ∼ 100 cm−1 ), which shifts nearly in parallel to lower energies with the PbO content. Second, on the other hand, 𝛽(ℏ𝜔) appears to be non-exponential. There exist fairly sharp increases at ∼2.0 eV, which seem to be common to all the compositions. This threshold suggests that two-photon transitions in these glasses occur between the states that are separated in energy by more than ∼4 eV, which is substantially higher than the Urbach edge located at ∼3 eV. Finally, the maximal 𝛽 in PbO-SiO2 increases by ten times from 1 to 10 cm/GW for nearly twice an increase in the PbO content from 38 to 68 at.%. The first and second features of 𝛼(ℏ𝜔) and 𝛽(ℏ𝜔) spectra can be understood as follows [40, 42]: The electronic structure of PbO-SiO2 is illustrated in Figure 10.7, which suggests that the one-photon absorption edge ∼3 eV is governed by intra-atomic 6s → 6p transitions in Pb. This interpretation can explain the red-shifting one-photon absorption edge with an increase in the PbO content, since the 6s and 6p bands arising from the Pb atoms probably become broader-reflecting enhanced interatomic interactions. However, such a transition cannot be assumed for two-photon absorption, because it cannot occur between s and p states in a single atom (see Eq. (10.4); note that H ∝ r). Then, a possible two-photon transition with a small photon energy is assumed to be O (2p) → Pb (6p), which can account for the different threshold in 𝛽(ℏ𝜔) at ∼2 eV from that in 𝛼(ℏ𝜔). This interpretation is also consistent with the dramatic 𝛽 increase with an increase in the PbO content, which is ascribable to resonant two-photon absorption. Two-photon absorption can resonantly be enhanced by the Pb (6s) states, as the term / (EPb6s,O2p − ℏ𝜔) possibly governs the transition probability in Eq. (10.4). Note that since Pb (6s) is an occupied state, it cannot commit to two-step absorption (Figure 10.4c), as is experimentally confirmed [42]. In short, in contrast to the situation in simple glasses such as SiO2 , the three atomic levels (O 2p, Pb 6s, and Pb 6p) seem to play dominant roles in the optical properties of PbO-SiO2 glasses. Comparatively, defective structures are less important in this system. Similar interpretations are probably applicable to other heavy-metal oxide glasses containing Bi2 O3 , etc. [43–46].

277

10 Nonlinear Optical Properties of Photonic Glasses

10.3.3

Nonlinear Refractivity

Since measurements of n2 are relatively difficult, several relations for estimations of n2 using linear optical properties have been proposed [4, 17, 47]. These relations can be grouped into two types. One type is specifically applicable to transparent materials, and the most famous may be Boling’s relation [48], given by n2 (10−13 esu) ≈ 391 (nd − 1)∕𝜈d 5∕4 ,

(10.6)

where nd is the refractive index at the He d-line (𝜆 = 588 nm), and 𝜈 d is the Abbe number. This relation is known to provide satisfactory agreements in materials with a small nd (≤ 1.7). In the other type of relation, which applies to semiconductors, the nonlinear properties are connected with the bandgap energy Eg . For instance, Sheik-Bahae and Van Stryland [31, 32] have shown for many (∼30) crystals with Eg ≈ 1–10 eV that the relation given by n2 n0 = K G(ℏ𝜔∕Eg )∕Eg 4

(10.7)

provides a good approximation, where K is a material-independent constant (∼6 × 10−11 (cm2 W−1 )(eV)7/2 ), and G(ℏ𝜔/Eg ) represents a universal spectral curve; the functional form varies with nonlinear mechanisms. (A corresponding expression for 𝛽 is also given by them.) As shown by the solid line in Figure 10.8, this relation also gives satisfactory fits to many data of oxide and chalcogenide glasses [34, 49, 50], where for Eg we take the 101

15Na2O·85GeO2

PbO–SiO2

GeS2

As2S3

10–2

As2Se3

Ag20As32Se48

10–1

Bi2O3–SiO2

100

35La2S3·65Ga2S3

10

n0

10–3

SiO2

n2n0·(KG)–1

278

10–4

1

5

1 10

Eg/eV

Figure 10.8 Linear and nonlinear refractivity, n0 (open symbols) and n2 (solid symbols), in some oxide (circles), sulfide (triangles), and selenide (squares) glasses as a function of the optical gap E g . The solid and dashed lines depict Sheik-Bahae’s relation (Eq. (10.7)) and the Moss relation n0 4 E g ≈ 77.

10.3 Nonlinear Absorption and Refractivity

Tauc gap or the photon energy at 𝛼 = 103 cm−1 . This result suggests that irrespective of crystalline or noncrystalline solids, n2 is mostly determined by Eg , which is governed by the short-range atomic structure. We can assume that with a scale of optical wavelengths, there is no marked difference between crystals and glasses, provided that anisotropic effects in the former could be neglected. The disordered structure in glasses, including the gap states, tends to cause only secondary effects on n2 , which may explain a large deviation, e.g. in Ag20 As32 Se48 having n2 ≈ 8 × 10−16 m2 W−1 [26], from the line in Figure 10.8. Moreover, as illustrated in Figure 10.8, the Moss rule (n0 4 Eg ≈ 77) [23] also applies to these glasses, which further reinforces the importance of Eg . It may be interesting to examine if we can relate n2 (ℏ𝜔) to 𝛽(ℏ𝜔) in real materials. Since 𝛽(ℏ𝜔) data are limited to sub-gap photon-energy regions, we here approximate the spectra in As2 S3 and SiO2 by Gaussian curves, which are shown by dashed lines in Figure 10.9. Then, n2 (ℏ𝜔) can be calculated using the nonlinear Kramers–Krönig relation, Eq. (10.5), as shown by the solid lines in the figure. We see that the calculated n2 (ℏ𝜔) shapes resemble the theoretical curve for the crystalline semiconductor shown in Figure 10.5. We also see in Figure 10.9 that the agreements between the calculated n2 (ℏ𝜔) and the experimental n2 [5], plotted by open symbols, are surprisingly good. This agreement reinforces the fact that Eq. (10.5) can be practically employed for estimating n2 (ℏ𝜔), and that n2 in the near-infrared region is governed by two-photon absorption, which has also been confirmed in crystalline semiconductors [5]. Similar insights have been obtained also for PbO-SiO2 glasses [42]. In conclusion, irrespective of crystals and non-crystals, the bandgap energy Eg is the primary factor governing the nonlinear absorption and refractivity. The electronic density of states, including gap states, and the form of wave functions play secondary roles, and the details remain to be studied. Besides, since the transition amplitudes of oneand two-photon absorptions have different forms, further studies of these processes will provide valuable insights into related states. The idea described here provides an answer to a frequently asked question: “Why does As2 S3 possess higher n2 than SiO2 at optical communication wavelengths, ℏ𝜔 ≈ 1 eV? ” In the present context, this is because a higher n2 arises from the greater

n2/cm2·(GW)–1, β/cm·(GW)–1

102 100 10–2

As2S3

SiO2

10–4 10–6 10–8

0

1

2

3

4

5

6

7

ћω/eV

Figure 10.9 Two-photon absorption spectra 𝛽(ℏ𝜔) in As2 S3 (solid circles) and SiO2 (solid triangles), fitted Gaussian profiles (dashed lines), calculated n2 (ℏ𝜔) (solid lines), and experimental n2 (open circle and triangle). 𝛽(ℏ𝜔) data are the same as those shown in Figure 10.6 .

279

280

10 Nonlinear Optical Properties of Photonic Glasses

𝛽 (∼10 cm GW−1 ) at a lower photon-energy region in As2 S3 [34]. The greater 𝛽 can then be connected with the smaller Eg , because Eq. (10.4) gives 𝛽 ∝ Eg −2 , provided that the denominator (Eni − ℏ𝜔)2 governs the gross features. (Precise analyses give 𝛽 ∝ Eg −3 [49, 50].) A higher n2 at a smaller value of Eg can also be derived from Eq. (10.7), which leads to n2 ∝ Eg −4 . We can also evaluate the feasibility of n2 in bulk materials at the communication wavelength 𝜆 ≈ 1.5 μm [50]. As Eg corresponds to the electronic cutoff wavelength 𝜆g through a simple equation, Eg [eV]⋅𝜆g [μm] ≈ 1.24, the Sheik-Bahae’s relation (Eq. (10.7)) predicts for materials with Eg ≥ 1.2 eV (𝜆g ≤ 1 μm) that n2 ≤ 10−15 m2 W−1 (= 10 cm2 TW−1 ), which is consistent with reported values [26, 27, 51–55]. This result suggests that for obtaining practical nonlinearity of Ln2 I/𝜆 ≈ 𝜋 in optical waveguides with a light propagation distance L of 1 cm, we need light sources with an intensity I of ∼0.1 GW cm−2 , or 1 W μm−2 , which can be attained using presently available laser diodes.

10.4 Nonlinear Excitation-Induced Structural Changes 10.4.1

Fundamentals

Oxide and chalcogenide glasses are known to exhibit a variety of photo-induced phenomena [7–11]. These phenomena tend to yield metastable structures, whereas it is challenging to identify the atomic structures and transformations occurring in disordered glass networks. Some of the photo-induced changes can be recovered by thermal annealing at the glass-transition temperature T g . It should be noted that in comparison with photo-induced phenomena in crystals, such as F center formation in alkali halides, two features add complexities to such phenomena in glasses. One is that the photo-induced change in a glass depends upon many factors such as the photon energy and (peak) the intensity of illuminating light, continuous wave (cw) or pulse (pulse duration and repetition), light spot size (bulk and interfacial stress effects), light polarization, illuminating temperature T i , and illumination atmosphere (air, vacuum, etc.). For instance, many photo-induced transformations become smaller at higher T i , ultimately disappearing at T i ≈ T g [10, 11, 21]. The other feature that adds complexity is that a glass property varies from sample to sample, reflecting the quasi-stability, which substantially affects photo-induced phenomena. In fact, radiation effects in SiO2 , one of the simplest oxide glasses, cannot universally be grasped due to minute, but crucial, deviations in stoichiometry and impurities such as H, O2 , OH, and Cl [7, 21, 24, 56–58]. In a chalcogenide, As2 S3 , it is known that evaporated and annealed films undergo different photo-induced transformations of polymerization (stabilization) and disorder increases (destabilization), respectively [8–11]. In short, we are confronted with a variety of exposure conditions and samples for a single material, and as a result, it is rather difficult to obtain universal insights. Notable photo-induced effects are produced through linear excitations by bandgap photons with ℏ𝜔/Eg ≥ 1, and also through nonlinear (or linear) optical excitations by photons with ℏ𝜔/Eg < 1. Such low-energy photons are able to penetrate more than ∼1 cm into bulk samples, due to small linear absorption, which is favorable for producing volume-related effects. For instance, in Ge-doped SiO2 fibers with a diameter of ∼100 μm, Bragg reflectors are inscribed using ultraviolet excimer lasers [1]. Also, in

10.4 Nonlinear Excitation-Induced Structural Changes

oxide [58–60] and chalcogenide [3, 53, 61–64] glasses, waveguides can be fabricated by scanning near-infrared laser light. At these photon energies, since linear absorption is relatively small, nonlinear excitations could be responsible for photo-electronic excitations. However, the fundamental mechanisms of the photo-induced phenomena remain to be explored. In general, a photo-induced phenomenon occurs through energy transfer from photons to electron-lattice systems, and to atomic structures. Accordingly, the process may be divided into two parts, photo-electronic excitation and electro-structural change, and the latter can be regarded as either athermal (polaronic) or thermal. In an athermal process, the temperature rise in a material, which inevitably occurs to some degrees under illumination, can be neglected, but it is decisive in the thermal process including heat-spike generation. Otherwise, we may envisage instantaneous photo-electro-structural evolutions in electronic melting under ultrahigh excitations, the concept was originally proposed by Van Vechten et al. [65]. Permanent damage may occur under such excitations. In this section, we will make a comparative study of the athermal photo-induced phenomena induced by one- and multi-photon processes in the oxide SiO2 and the chalcogenide As2 S3 . Nevertheless, at the outset, it may be valuable to estimate the temperature rise induced by light illumination. The temperature rise ΔT can be evaluated very roughly for two extreme cases: Upon single pulse irradiation with a peak intensity I and a pulse duration 𝜏 (≥ 10 ps, which is a typical electron-lattice relaxation time), the temperature rise in a sample can be estimated as ΔT ≈ Q/cV , where Q (≈ 𝛼I𝜏) is the absorbed light energy, c the specific heat, and V (≥ Sd) a related volume, which can be evaluated from the light spot area S, a characteristic distance d (the sample thickness or light penetration length such as 𝛼 −1 or (𝛽I)−1 ), and the thermal diffusion length (k𝜏/c)1/2 , where k is the thermal conductivity. For instance, when an SiO2 bulk sample (k ≈ 10−2 W cm−1 ⋅K and c ≈ 2 J cm−3 ⋅K) is exposed to a light pulse with a width of 1 ns, peak intensity of 1 kW, spot diameter of 10 μm (≈ 1 GW cm−2 ), and ℏ𝜔 = 8 eV (𝛼 −1 ≈ 1 mm), one obtains ΔT ≈ 5 K. For light pulses of duration shorter than 1 ns, the thermal diffusion length (k𝜏/c)1/2 becomes shorter than ∼25 nm, which may be neglected for all practical purposes [66]. On the other hand, a steady-state temperature rise in an absorbing sample exposed to cw or high-repetition pulses with a time-averaged intensity I can be estimated from ΔT ≈ Q/(2𝜋kr) [67], where Q (≈ 𝛼IV ) is the absorbed light power, and r is the radius of the light spot. If Q = 1 mW and r = 10 μm, this gives ΔT ≈ 10 K for SiO2 . Experimentally, the temperature rise can be probed using thin-film thermocouples or infrared thermometers, or it can be estimated from the intensity ratio of the (anti-)Stokes Raman-scattering spectra. 10.4.2

Oxides

Figure 10.10a summarizes the structural responses of SiO2 to light excitations with varied photon energies and intensities. The results are obtained at room temperature, i.e. T i /T g ≈ 300/1500 = 0.2, using three types of light sources: 1) Laser pulses of fs-ps duration and ℏ𝜔 ≈ 1–2 eV (ℏ𝜔/Eg ≈ 0.1–0.2) [57–60, 68–79]. 2) Excimer laser pulses of ∼10 ns and ℏ𝜔 ≈ 5–8 eV (ℏ𝜔/Eg ≈ 0.5) [24, 58, 80–82]. 3) (Super-) bandgap excitations (ℏ𝜔/Eg ≥ 1) including x- and γ-rays [83–85].

281

10 Nonlinear Optical Properties of Photonic Glasses

d

G

d E′& –ΔV

M

+ΔV

K

ablation –ΔV –Δn WB

103

d c–Si

E′

100

β

1 Ti:sapphire

KrF ArF

Eg

F2

m 0

5

(b) T

ablation Δn

G

Δn

Δn

PD

m 0

1

100

+ΔV

Ti:sapphire Nd:YAG

1

103

PD

Δn

β

K

α

Δn Δn WB WB

M

10–3 15

10

ћω/eV

ћω/eV

PD Δn +ΔV

2

α/cm–1, β/cm·(GW)–1

Intensity/W·cm–2

α

d

T

Eg

As–As

α/cm–1, β/cm·(GW)–1

(a)

Intensity/W·cm–2

282

10–3

3

Figure 10.10 Photo-induced phenomena scaled by the excitation photon energy ℏ𝜔 and the (peak) light intensity in (a) SiO2 and (b) As2 S3 . T, G, M, etc. on the left-hand side vertical axis represent TW, GW, MW, etc. Also shown are the one- and two-photon absorption spectra, 𝛼 (solid lines), and 𝛽 (dashed lines), respectively. The scales are shown on the right vertical axis. Photon energies of related laser light are indicated on the horizontal axis as Nd:YAG, Ti:sapphire, KrF, ArF, and F2 . E g represents the bandgap energy. In the abbreviations, E’ indicates an E’-center formation, +ΔV a volume expansion, -ΔV a volume contraction, Δn a refractive index change, WB a wrong-bond formation, PD a photo-darkening, d damage. In the shaded region in (b), photo-darkening, refractive index increase, and volume expansion occur.

Experiments using (super-)bandgap illumination in (3) are limited, probably because of the large bandgap energy of Eg ≈ 9 eV. Takigawa et al. [83] report for synthetic SiO2 windows that a single shot of 9.8 eV laser light (Ar excimer laser, 5 ns) with ∼300 mJ cm−2 (≈ 60 MW cm−2 ) power causes topological surface damage and crystalline Si formation. They also mention that lower-energy shots with ℏ𝜔 ≈ 5 ∼ 8.5 eV do not produce crystalline Si, despite the marked surface damage. Awazu and Kawazoe [21] have demonstrated that super-bandgap excitations with energies 14 and 18 eV, which are obtained from an undulator-equipped synchrotron, produce small Si—O rings in thermally grown amorphous SiO2 films. Akazawa [84] has discovered that under super-bandgap excitations (synchrotron radiation of 100–300 eV), amorphous SiO2 films evaporate while accumulating Si—Si wrong bonds. In short, a (super-)bandgap excitation tends to produce Si wrong bonds and evaporation. These phenomena in SiO2 resemble photo-enhanced vaporization of As2 O3 , which appears efficiently in heated As2 S3 under super-bandgap illumination [86]. By contrast, it seems that a photo-darkening phenomenon (see Section 10.4.3), which is induced in covalent chalcogenide glasses by bandgap illumination, does not exist in SiO2 [10, 24] and GeO2 [87].

10.4 Nonlinear Excitation-Induced Structural Changes

On the other hand, many studies in the above experiments (1) and (2) have demonstrated some changes in small photon-energy excitations with ℏ𝜔/Eg ≈ 0.1–0.8. Specifically, three-dimensional optical fabrications have been explored using intense fs-ps lasers with ℏ𝜔 ≈ 1 eV under various conditions [58, 60, 66, 68, 69, 72–78].When the incident pulse is more intense than the order of J cm−2 levels, or GW/cm2 for 1 ns (or TW/cm2 for 1 ps) pulses, both surface and/or internal damage occur with ablation [21, 58, 59, 68, 72, 73, 76, 78]. However, when the peak intensity is around 1 MW/cm2 , three types of phenomena may appear: (1) creation of defective structures such as Si—Si bonds [79] with small rings [21], (2) volume changes [7, 74, 75, 80], and (3) refractive index changes [7, 58–60, 74–77]. The volume and the refractive index changes have been correlated through the Lorentz-Lorenz relation [7, 74, 75]. Nevertheless, to the best of the author’s knowledge, the polarization effects induced by pulse light have scarcely been taken into account in these changes [76, 77]. Is the multi-photon excitation really responsible in these photo-structural phenomena? It is plausible that in Ge-doped SiO2 , the dopants play the role of color centers, which cause linear absorption [58, 80] as illustrated in Figure 10.10a’. However, in pure SiO2 , it seems still unclear whether the excitation is triggered by one- or by multi-photon processes. For a two-photon excitation to be dominant, at least the condition 𝛽I ≫ 𝛼 must be satisfied. For instance, at ℏ𝜔 ≈ 5 eV, since 𝛼 = 10−2 cm−1 and 𝛽 = 10−1 cm GW−1 (see Figure 10.6), the light intensity must be greater than ∼100 MW cm−2 . This intensity level is located in Figure 10.10a near the boundary between the damage and the nondestructive structural modifications. Here, we should note that 𝛼 is evaluated using a weak probe light, but the condition 𝛽I ≫ 𝛼 should be satisfied under intense illumination that may produce transitory midgap absorption at defective sites [70] accompanying an increase in 𝛼. The intense illumination also produces a temperature rise, which is likely to enhance band-edge absorption. Accordingly, it is not straightforward to examine the 𝛽I-𝛼 criterion under practical conditions. In addition, we should distinguish the two-photon from the two-step absorption (see Figure 10.4), which may exhibit different response times. In fact, Kajihara [24] has proposed that the exposure with 7.9 eV photons to SiO2 causes the formation of E’ centers through a two-step absorption process rather than through two-photon absorption. Finally, how is the excitation converted to structural transformations? To answer this question experimentally, time-resolving photo-structural studies will be valuable for understanding the conversion process from photo-electronic excitation, transitory defect formation, and to the modifications in macroscopic properties [77]. 10.4.3

Chalcogenides

The chalcogenide glasses have an optical gap in the range 1–3 eV, so that visible light can induce several types of prominent structural changes [3, 8–11]. At the outset, however, it is important to distinguish whether the phenomena arise from photo-induced physical (bond alternation) or chemical (oxidation) processes in as-deposited films and annealed samples including bulk glasses and fibers. In the annealed samples, the phenomena of photo-darkening (evaluated as red shifts ΔE in the optical absorption edge, which can be recovered by annealing), related refractive index increase, and volume expansion have attracted continuous interest [8–11, 88]. This is because these changes are simple, athermal, inherent to covalent chalcogenide glasses such as Se and As2 S3 ,

283

284

10 Nonlinear Optical Properties of Photonic Glasses

and promising for optical applications including holographic memories and functional devices [10, 11, 53]. For understanding the fundamental mechanism of photo-darkening and related phenomena, spectral studies are valuable [10, 11]. For instance, in As2 S3 with Eg ≈ 2.4 eV, it is known that bandgap illumination at room temperature (T i /T g ≈ 300/450 ≈ 0.7) produces an edge shift ΔE of ∼50 meV and a refractive index increase of ∼0.02 (see Table 10.1), which are quantitatively related through the linear Kramers–Krönig relation (Eq. (10.3)); and in addition, a volume expansion of ∼0.4%. It has also been demonstrated that even an intense sub-bandgap illumination can produce notable changes. In As2 S3 , continuous illumination of light with a photon energy of 2.0 eV (ℏ𝜔/Eg ≈ 0.8) and an intensity greater than 100 W cm−2 can produce photo-darkening that is comparable to that induced by bandgap illumination, and a seemingly more prominent volume expansion. Pulsed light of energy 1.65 eV (ℏ𝜔/Eg ≈ 0.7) and a peak intensity of 100 MW cm−2 can also induce the optical and volume changes [89, 90]. Tanaka has proposed, taking a marked red shift of the photo-conduction spectrum into account, that these intense sub-gap illuminations excite enough carriers to completely fill band-tail states, which probably exerts similar effects to those by bandgap illumination [89]. However, the half-gap illumination (ℏ𝜔 ≈ 1.2 eV ≈ Eg /2) exhibits different features in As2 S3 [89, 91, 92] as explained in what follows: If the illumination is weak, no detectable changes appear within experimental time scales. But if a sample is illuminated by intense pulses that satisfy the condition 𝛽I » 𝛼, the excitation gives rise to a refractive index increase, but no discernible photo-darkening, 0 ± 5 meV, as shown in Table 10.1 and Figure 10.10b. Besides, no changes in sample thicknesses (≈10–30 μm) are detected with an accuracy of ∼100 nm. On the other hand, Raman-scattering spectroscopy demonstrates that the two-photon excitation causes a density increase (a few percent) in wrong bonds, As—As and S—S. In these experiments performed at ∼300 K, the temperature rise induced by the excitation is estimated to be ∼10 K, which can practically be neglected in comparison with the glass-transition temperature of ∼450 K; the process appears to be athermal. It has also been reported that near-infrared light pulses induce similar refractive index increases in glasses such as As2 Se3 (Eg ≈ 1.8 eV) [93], Ga-La-S (Eg ≈ 2.6 eV) [62, 64, 94], Ge-Sb-Se (Eg ≈ 2 eV) [95], etc. [96]. We may envisage that these observations resemble the fs-ps pulse effects in SiO2 described in Section 10.4.2. Ablation-related phenomena also emerge upon exposure to more intense near-infrared pulses [52, 97]. Table 10.1 Photo-darkening (red shift) ΔE and refractive index increase Δn induced in As2 S3 glass by pulsed light with energies of 1.17 and 2.33 eV and also by continuous-wave (cw) bandgap light of ∼2.4 eV for comparison [89]. ℏ𝝎(eV)

𝚫E(meV)

𝚫n

𝜶 (cm−1 )

𝜷 (cm W−1 )

Intensity (W cm−2 )

1.17

0±5

0.005

10−3

10−10

109

2.33

20

0.003

300

?

107

cw

50

0.02

∼500

∼0.05

Related parameters (𝛼, 𝛽, and excitation light intensity) are also listed. 𝛽 at 2.33 eV cannot be evaluated due to two-step absorption [36]. In the pulse experiments, ΔE and Δn are evaluated at a fixed absorbed photon number of 1023 –1024 cm3 , which may not provide saturated changes.

10.5 Conclusions

Conduction Band

Valence Band

Figure 10.11 Schematic illustrations of the bandgap excitations with one-photon (left) and resonant two-photon (right) processes. One-photon excitations may cause modifications in the medium-range structures, which broaden the valence band, and as a result, photo-darkening occurs. Two-photon excitations may produce localized defect centers such as small As clusters.

To understand the above two-photon excitation effect, we must answer at least two questions: (1) Why does the excitation produce the wrong bonds? And (2) why does the refractive index increase? To answer the first question, we may follow the idea that two-photon absorption can be resonantly enhanced by midgap states if the energies satisfy Eni − ℏ𝜔 ≈ 0 in Eq. (10.4). Otherwise, we envisage two-step absorption. Here, the midgap states are ascribable to the anti-bonding states of As—As bonds [36]. In such cases, resonant two-photon (step) absorption tends to occur spatially selectively around the defective sites, and as a result, As clusters may grow through some presently unspecified mechanisms. On the other hand, bandgap illumination probably causes defect-unrelated excitations resulting in different structural changes, such as bond twisting [8, 10, 11], which probably broadens the valence band. In short, as illustrated in Figure 10.11, reflecting the different transition probabilities between the one- and two-photon excitations, different structural changes may occur. Question (2) can be addressed as follows. Taking the linear Kramers–Krönig relation (Eq. (10.3)) into account, we may assume that the refractive index increase should be accompanied by some absorption increase due to photo-darkening (reduction in Eg ), but experimentally no edge shift has been detected. No volume changes have been detected as well. These facts may suggest the idea that inhomogeneous structures containing As and S clusters produce some stresses, which cause the increase in refractive index through photo-elastic effects [92]. Otherwise, we envisage that it occurs due to the emergence of some midgap absorption processes. In order to examine these notions, further studies such as composition dependence of the photo-induced refractive index change in the As—S system will be valuable.

10.5 Conclusions After briefly reviewing photonic glasses, we have highlighted two topics on nonlinear optical properties of oxide and chalcogenide glasses. One is a unified understanding of the absorptive and dispersive third-order nonlinearity, and the other is the role of nonlinear excitations in photo-induced phenomena.

285

286

10 Nonlinear Optical Properties of Photonic Glasses

Optical nonlinearity in glasses is governed by the bandgap energy and is modified by detailed electronic structures. The role of the bandgap energy is similar to that in crystals, since the short-range structure in a glass is nearly the same as that in its crystalline counterpart. However, a difficult, but case-dependent problem is that the nonlinearity in glasses is affected by the band-tail and midgap states, having spatially localized wave functions, which could cause resonant multi-photon and/or multi-step absorption. Different photo-structural changes appear in a glass depending upon the photon energy and light intensity. When the light intensity is weak and ℏ𝜔 ≈ Eg /2, one-photon absorption by midgap states can be held responsible for polaronic changes. If the intensity is high, we should also take (resonant) two-photon and two-step processes into account. If ℏ𝜔 ≈ Eg , band-to-band excitations become dominant. These photoelectronic excitations seem to cause unique photo-structural changes. In SiO2 , the midgap excitation creates defective structures, which may produce volume contraction and refractive index increase, whereas (super-)bandgap illumination tends to produce Si—Si homopolar bonds. In As2 S3 , pulsed infrared laser light (∼Eg /2) produces wrong bonds, As—As and S—S, probably through localized resonant two-photon excitations, which is in contrast to the fact that bandgap illumination gives rise to photo-darkening, refractive index increase, and volume expansion. These photo-induced structural modifications are inherent to the metastable material, and further studies including polarized-light nonlinear excitations will be interesting for the creation of artificially tuned atomic structures.

10.A Addendum: Perspectives on Optical Devices So far, several types of photonic glass devices, passive and active, have been commercialized, and it would be valuable to shed light on some fundamental functions. At present, most of the devices have been developed for optical communications [1–3, 98], and the rest are for other purposes such as laser components and scientific experiments. For instance, a common passive element is the optically inscribed fiber Bragg filter, which is utilized for wavelength selection, etc. In addition, with regard to active optical devices, some of which are listed in Table 10.A.1, rare-earth-doped fiber amplifiers have widely been employed in wavelength-division multiplexing (WDM) systems. How can we characterize such optical devices? A comparison of the properties of an electron in semiconductors and a photon in photonics glasses reveals contrasting features in three facets; charge, speed, and scale. The electron has the charge, which is the origin of strong interaction with electric and Table 10.A.1 Active optical devices. Device

Nonlinear (bulky)

Others

Source/amplifier

Supercontinuum generator Overtone generator Stimulated Raman-scattering amplifier

Rare-earth doped

Controller/modulator

n2 (Kerr) switch 𝛽 stabilizer (intensity limiter)

(Electro-optical)

10.A Addendum: Perspectives on Optical Devices

magnetic fields. By contrast, the photon (electromagnetic wave) interacts only weakly with a material (atoms and bonding) through the permittivity. Naturally, nonlinear interactions tend to become weaker. Moreover, we can manipulate electrons and holes in semiconductor bipolar devices (or a majority carrier in unipolar devices), while the photonics has no counterpart of holes; i.e. photonic devices are nonpolar. Hence, we cannot produce optical elements such as the pn junction, a basic electronic component. As regards speed, the electron is able to move between zero (trapped) and ∼108 cm s−1 (under an electric field of 105 V cm−1 and a mobility of 103 cm2 (V s)−1 ), while the photon has a nearly fixed speed of ∼1010 cm s−1 . These two characteristics explain a principal outcome: the electron is suitable for signal processing at frequencies below the gigahertz range, and the photon is convenient for teleportation. Otherwise, we may utilize standing light waves in photonic structures, and so forth, although such devices are likely to possess retarded resonant responses. For the scale, an electron can be confined into nanometer wells, while a photon has a wavelength of a micrometer. These scales govern the related device dimensions. For instance, it will be impossible to devise optical circuits with integration scales of nanometers. As regards the macroscopic shape of optical devices, most of them have fiber forms, with functional two- and three-dimensional waveguides remaining to be developed. One of the reasons is that the nonlinearity presently obtained needs light–matter interaction lengths longer than ∼1 m, which are insufficient for compact devices of ∼1 cm or less [49, 50, 54, 55, 99]. As a consequence, the practical usage of integrated nonlinear optical circuits awaits innovative materials and device architectures. For the material, we have understood that the optical nonlinearity in bulk solids is governed by the bandgap energy. Hence, further enhancements of the nonlinearity need more sophisticated materials in combination with novel structures, examples being particle-doped glasses [3, 17, 100], micro-structured fibers [19, 101, 102], tapered fibers [102, 103], nanowires [104], resonant structures including spheres [105], rings [29, 106], photonic crystals [107–109], etc. For instance, it has been demonstrated that a Se-doped zeolite yields three-orders-of-magnitude-higher nonlinearity than that in bulk Se glass [50], with a proposed origin of exciton confinement [38]. Besides, plasmonic enhancements of electric fields offer tempting ideas [17, 110]. We also point out a common feature and a difference in electronic and optical devices: The common feature can be found in many of the component structures, such as diode or triode types. For instance, optical isolators, supercontinuum and overtone generators, and intensity stabilizers using nonlinear absorption can be regarded as a diode type consisting of only an input and an output. On the other hand, as illustrated in Figure 10.A.1, Nonlinear fiber

Figure 10.A.1 A triode-type optical device. INPUT

Isolator

OUTPUT

Filter PUMP

287

288

10 Nonlinear Optical Properties of Photonic Glasses

Figure 10.A.2 A Mach-Zhender–type logic waveguide with nonlinear optical gates G1 and G2.

G1 input G2 output

stimulated Raman-scattering amplifiers, optical switches utilizing n2 , etc. are of the triode type: in the same way as field-effect transistors having gate, source, and drain. Such analogies will be valuable for improving and developing optical devices. For instance, the optical isolator and a pn-junction diode serve a similar function, although the former is still much larger. The intensity stabilizer may be analogous to a Zener diode, although the former function remains comparatively inefficient. We can also directly see that a Mach-Zhender interferometer with nonlinear couplers, illustrated in Figure 10.A.2, could operate as a dual-gate logic element. By contrast, a marked difference, or an advantage of the optical system, is that it can possess spectral (and polarization) multiplicity. Many signals having different wavelengths can be processed simultaneously in a single line, a property that has already been utilized in the WDM system equipped with fiber amplifiers of rare-earth doped and/or stimulated Raman-scattering types. In addition, supercontinuum generators are becoming unique light sources [1–4, 20, 101–104]. Spectral hole-burning memories have been continuously studied [111]. Nevertheless, it should be emphasized that the resonant structure including photonic-crystal devices [107–109] is in principle incompatible with multi-spectral operation. Under these circumstances, we have been exploring ultrafast-response (∼ps), compact, wide-band (Δ𝜆 ≥ 10 nm), nonlinear optical devices with sufficient stability [3, 101, 112]. In future, the optical components may be integrated with electronic (semiconductor) and electro-optical (ferroelectric) circuits on monolithic and/or hybrid chips, the concept originally proposed by Miller nearly a half century ago [113]. Or, the ultimate target will be totally optical systems, in which all the signal processing and transmission are accomplished via multi-spectral photons, and the electricity is fed just for power supply to tunable and/or supercontinuum light sources.

References 1 Agrawal, G.P. (2013). Nonlinear Fiber Optics, 5e. San Diego: Academic Press. 2 Agrawal, G.P. (2008). Applications of Nonlinear Fiber Optics, 2e. San Diego:

Academic Press. 3 Zakery, A. and Elliott, S.R. (2007). Optical properties in Chalcogenide Glasses and

their Applications. New York: Springer. 4 Boyd, R.W. (2008). Nonlinear Optics, 3e. Boston: Academic Press. 5 Sutherland, R.L. (2003). Handbook of Nonlinear Optics, 2e. New York: Marcel

Dekker. 6 Elliott, S.R. (1990). Physics of Amorphous Materials, 2e. Essex: Longman Scientific

& Technical. 7 Pacchioni, G., Skuja, L., and Griscom, D.L. (eds.) (2000). Defects in SiO2 and

Related Dielectrics: Science and Technology. Dordrecht: Kluwer Academic Publishers.

References

8 Popescu, M.A. (2001). Non-Crystalline Chalcogenides. Dordrecht: Kluwer Academic

Publishers. 9 Nalwa, H.S. (ed.) (2001). Handbook of Advanced Electronic and Photonic Materials

and Devices, vol. 5. San Diego: Academic Press, Chapters 2, 3, 4, and 5. 10 Kolobov, A.V. (ed.) (2003). Photo-Induced Metastability in Amorphous Semiconduc-

tors. Weinheim: Wiley-VCH. 11 Tanaka, K. and Shimakawa, K. (2011). Amorphous Chalcogenide Semiconductors

and Related Materials. New York: Springer. 12 Tanaka, K. (1987). Chemical and medium-range orders in As2 S3 glass. Phys. Rev. B 13 14 15

16

17

18 19 20

21

22 23 24 25

26

27

36: 9746–9752. Tanaka, K. (2014). Minimal Urbach energy in non-crystalline materials. J. Non-Cryst. Solids 389: 35–37. Tanaka, K. (2002). Wrong bonds in glasses: a comparative study on oxides and chalcogenides. J. Opt. Adv. Mater. 4: 505–512. Quiquempois, Y., Niay, P., Douay, M., and Poumellec, B. (2003). Advances in poling and permanently induced phenomena in silica-based glasses. Curr. Opin. Solid State Mater. Sci. 7: 89–95. Vijayalakshmi, S. and Grebel, H. (2000). Nonlinear optical properties of nanostructures. In: Handbook of Nanostructured Materials and Nanotechnology, vol. 4 (ed. H.S. Nalwa), 529–560. San Diego: Academic Press, Chapter 8. Tanaka, K. (2017). Optical nonlinearity in photonic glasses. In: Springer Handbook of Electronic and Photonic Materials, 2e (eds. S. Kasap and P. Capper). New York: Springer Chapter 42. Lucas, J. (1997). Fluoride glasses. Curr. Opin. Solid State Mater. Sci. 2: 405–411. Monro, T.M. and Ebendorff-Heidepriem, H. (2006). Progress in microstructured optical fibers. Annu. Rev. Mater. Res. 36: 467–495. Amiot, C., Aalto, A., Ryczkowski, P. et al. (2017). Cavity enhanced absorption spectroscopy in the mid-infrared using a supercontinuum source. Appl. Phys. Lett. 111: 061103. Awazu, K. and Kawazoe, H. (2003). Strained Si-O-Si bonds in amorphous SiO2 materials: a family member of active centers in radio, photo, and chemical responses. J. Appl. Phys. 94: 6243–6262. Boolchand, P., Georgiev, D.G., and Goodman, B. (2001). Discovery of the intermediate phase in chalcogenide glasses. J. Optoelectron. Adv. Mater. 3: 703–720. Moss, T.S. (1959). Optical Properties of Semi-conductors. London: Butterworths Sci. Pub. Kajihara, K. (2007). Improvement of vacuum-ultraviolet transparency of silica glass by modification of point defects. J. Ceram. Soc. Jpn. 115: 85–91. Boudebs, G., Cherukaulappurath, S., Guignard, M. et al. (2004). Experimental observation of higher order nonlinear absorption in tellurium based chalcogenide glasses. Opt. Commun. 232: 417–423. Ogusu, K., Yamasaki, J., Maeda, S. et al. (2004). Linear and nonlinear optical properties of Ag–As–Se chalcogenide glasses for all-optical switching. Opt. Lett. 29: 265–267. Ogusu, K. and Shinkawa, K. (2009). Optical nonlinearities in As2 Se3 chalcogenide glasses doped with Cu and Ag for pulse durations on the order of nanoseconds. Opt. Express 17: 8165–8172.

289

290

10 Nonlinear Optical Properties of Photonic Glasses

28 Baudrier-Raybaut, M., Haidar, R., Kupecek, P. et al. (2004). Random

29 30 31

32

33 34 35

36 37 38 39 40 41 42 43

44

45 46

quasi-phase-matching in bulk polycrystalline isotropic nonlinear materials. Nature 432: 374–376. Carmon, T. and Vahala, K.J. (2007). Visible continuous emission from a silica microphotonic device by third-harmonic generation. Nature Phys. 3: 430–435. Hutchings, D.C., Sheik-Bahae, M., Hagan, D.J., and Van Stryland, E.W. (1992). Kramers-Krönig relations in nonlinear optics. Opt. Quant. Electron. 24: 1–30. Sheik-Bahae, M. and Van Stryland, E.W. (1999). Optical nonlinearities in the transparency region of bulk semiconductors. In: Semiconductors and Semimetals, vol. 58 (eds. E. Garmire and A. Kost). San Diego: Academic Press Chapter 4. Sheik-Bahae, M., Hagan, D.J., and Van Stryland, E.W. (1990). Dispersion and band-gap scaling of the electric Kerr effects in solids associated with two-photon absorption. Phys. Rev. Lett. 65: 96–99. Dinu, M. (2003). Dispersion of phonon-assisted nonresonant third-order nonlinearities. IEEE Quant. Electron. 39: 1498–1503. Tanaka, K. (2005). Optical nonlinearity in photonic glasses. J. Mater. Sci. Mater. Electron. 16: 633–643. Baltraneyunas, R.A., Vaitkus, Y.Y., and Gavryushin, V.I. (1984). Light absorption by non-equilibrium, two-photon-generated, free and localized carriers in ZnTe single crystals. Sov. Phys. JETP 60: 43–48. Tanaka, K. (2002). Two-photon absorption spectroscopy of As2 S3 glass. Appl. Phys. Lett. 80: 177–179. Mizunami, T. and Takagi, K. (1994). Wavelength dependence of two-photon absorption properties of silica optical fibers. Opt. Lett. 19: 463–469. Tanaka, K. and Saitoh, A. (2009). Optical nonlinearity of Se-loaded zeolite (ZSM-5): a moulded nanowire system. Appl. Phys. Lett. 94: 241905. Enck, R.C. (1973). Two-photon photogeneration in amorphous selenium. Phys. Rev. Lett. 31: 220–223. Tanaka, K. (2004). Two-photon optical absorption in amorphous semiconductors. J. Non-Cryst. Solids: 338, 534–340, 538. Tanaka, K., Yamada, N., and Oto, M. (2003). Two-photon optical absorption in PbO-SiO2 glasses. Appl. Phys. Lett. 83: 3012–3014. Tanaka, K. and Minamikawa, N. (2005). Optical nonlinearity in PbO-SiO2 glass: Kramars-Kronig analyses. Appl. Phys. Lett. 86: 121112–121114. Sugimoto, N., Kanbara, H., Fujiwara, S. et al. (1999). Third-order optical nonlinearities and their ultrafast response in Bi2 O3 -B2 O3 -SiO2 glasses. J. Opt. Soc. Am. B 16: 1904–1908. Chen, F., Song, B., Lin, C. et al. (2012). Glass formation and third-order optical nonlinear characteristics of bismuthate glasses within Bi2 O3 -GeO2 -TiO2 pseudo-ternary system. Mater. Chem. Phys. 135: 73–79. Oliveira, T.R., Falcão-Filho, E.L., de Araújo, C.B. et al. (2013). Nonlinear optical properties of Bi2 O3 -GeO2 glass at 800 and 532 nm. J. Appl. Phys. 114: 073503. Besse, V., Fortin, A., Boudebs, G. et al. (2014). Picosecond nonlinearity of GeO2 –Bi2 O3 –PbO–TiO2 glasses at 532 and 1,064 nm. Appl. Phys. B Lasers Opt. 117: 891–895.

References

47 Tichá, H. and Tichý, L. (2002). Semiempirical relation between non-linear suscepti-

48

49 50 51 52

53 54

55

56 57 58

59

60 61

62 63 64

bility (refractive index), linear refractive index and optical gap and its application to amorphous chalcogenides. J. Optoelectron. Adv. Mater. 4: 381–386. Boling, N.L., Glass, A.J., and Owyoung, A. (1978). Empirical relationship for predicting nonlinear refractive index change s in optical solids. IEEE J. Quant. Electron. QE-14: 601–608. Tanaka, K. (2007). Nonlinear optics in glasses: how can we analyze? J. Phys. Chem. Solids 68: 896–900. Saitoh, A. and Tanaka, K. (2011). Optical nonlinearity in chalcogenide glasses for near-infrared all-optical devices. J. Optoelectron. Adv. Mater. 13: 71–74. Sanghera, J.S., Florea, C.M., Shaw, L.B. et al. (2008). Non-linear properties of chalcogenide glasses and fibers. J. Non-Cryst. Solids 354: 462–467. Fedus, K., Boudebs, G., de Araújo, C.B. et al. (2009). Photoinduced effects in thin films of Te20 As30 Se50 glass with nonlinear characterization. Appl. Phys. Lett. 94: 061122. Eggleton, B.J., Luther-Davies, B., and Richardson, K. (2011). Chalcogenide photonics. Nat. Photonics 5: 141–148. Krogstad, M.R., Ahn, S., Park, W., and Gopinath, J.T. (2016). Optical characterization of chalcogenide Ge-Sb-Se waveguides at telecom wavelengths. IEEE Photon. Technol. Lett. 28: 2720–2723. Kuriakose, T., Baudet, E., Halenkoviˇc, T. et al. (2017). Measurement of ultrafast optical Kerr effect of Ge–Sb–se chalcogenide slab waveguides by the beam self-trapping technique. Opt. Commun. 403: 352–357. Saito, K., Ito, M., Ikushima, A.J. et al. (2004). Defect formation and recovery processes in hydrogen-loaded silica fibers. J. Non-Cryst. Solids 347: 289–293. Zoubir, A., Rivero, C., Grodsky, R. et al. (2006). Laser-induced defects in fused silica by femtosecond IR irradiation. Phys. Rev. B 73: 224117. Lancry, M. and Poumellec, B. (2013). UV laser processing and multiphoton absorption processes in optical telecommunication fiber materials. Phys. Rep. 523: 207–229. Saliminia, A., Nguyen, N.T., Nadeau, M.-C. et al. (2003). Writing optical waveguides in fused silica using 1 kHz femtosecond infrared pulses. Appl. Phys. Lett. 93: 3724–3728. Ehrt, D., Kittel, T., Will, M. et al. (2006). Femtosecond-laser-writing in various glasses. J. Non-Cryst. Solids 345&346: 332–337. Efimov, O.M., Glebov, L.B., Richardson, K.A. et al. (2001). Waveguide writing in chalcogenide glasses by a train of femtosecond laser pulses. Opt. Mater. 17: 379–386. Ljungström, A.M. and Monro, T.M. (2002). Light-induced self-writing effects in bulk chalcogenide glass. J. Light Wave Technol. 20: 78–85. Ho, N., Laniel, J.M., Vallee, R., and Villeneuve, A. (2003). Photosensitivity of As2 S3 chalcogenide thin films at 1.5 μm. Opt. Lett. 28: 965–967. McMillen, B., Zhang, B., Chen, K.P. et al. (2012). Ultrafast laser fabrication of low-loss waveguides in chalcogenide glass with 0.65 dB/cm loss. Opt. Lett. 37: 1418–1420.

291

292

10 Nonlinear Optical Properties of Photonic Glasses

65 Van Vechten, J.A., Tsu, R., and Saris, F.W. (1979). Nonthermal pulse laser annealing

of Si; plasma annealing. Phys. Lett. 74A: 422–426. 66 Juodkazis, S., Misawa, H., and Maksimov, I. (2004). Thermal accumulation effect in

three-dimensional recording by picosecond pulses. Appl. Phys. Lett. 85: 5239–5241. 67 Arun, A. and Vedeshwar, A.G. (1997). Temperature rise at laser-irradiated spot in a

low thermal conducting materials. Phys. B Condens. Matter 229: 409–415. 68 Jia, T.Q., Xu, Z.Z., Li, X.X. et al. (2003). Microscopic mechanisms of ablation and

69

70

71

72

73 74

75

76

77 78 79 80

81

82

micromachining of dielectrics by using femtosecond lasers. Appl. Phys. Lett. 82: 4382–4384. Shimotsuma, Y., Kazansky, P.G., Qiu, J.R., and Hirao, K. (2003). Self-organized nanogratings in glass irradiated by ultrashort light pulses. Phys. Rev. Lett. 91: 247405–247409. Fukata, N., Yamamoto, Y., Murakami, K. et al. (2004). In situ spectroscopic measurement of structural change in SiO2 during femtosecond laser irradiation. Appl. Phys. A Mater. Sci. Process. 79: 1425–1427. Mihailov, S.J., Smelser, C.W., Grobnic, D. et al. (2004). Bragg gratings written in all-SiO2 and Ge-doped core fibers with 800-nm femtosecond radiation and a phase mask. J. Lightwave Technol. 22: 94–100. Siegel, J., Puerto, D., Gawelda, W. et al. (2007). Plasma formation and structural modification below the visible ablation threshold in fused silica upon femtosecond laser irradiation. Appl. Phys. Lett. 91: 082902. White, Y.V., Li, X., Sikorski, Z. et al. (2008). Single-pulse ultrafast-laser machining of high aspect nano-holes at the surface of SiO2 . Opt. Express 16: 14411–14420. Ponader, C.W., Schroeder, J.F., and Streltsov, A.M. (2008). Origin of the refractive-index increase in laser-written waveguides in glasses. J. Appl. Phys. 103: 063516. Bressel, L., de Ligny, D., Sonneville, C. et al. (2011). Femtosecond laser induced density changes in GeO2 and SiO2 glasses: fictive temperature effect [Invited]. Opt. Mater. Express 1: 605–613. Beresna, M., Geceviˇcius, M., Kazansky, P.G. et al. (2012). Exciton mediated self-organization in glass driven by ultrashort light pulses. Appl. Phys. Lett. 101: 053120. Mishchik, K., D’Amico, C., Velpula, P.K. et al. (2013). Ultrafast laser induced electronic and structural modifications in bulk fused silica. J. Appl. Phys. 114: 133502. Lebugle, M., Sanner, N., Varkentina, N. et al. (2014). Dynamics of femtosecond laser absorption of fused silica in the ablation regime. J. Appl. Phys. 116: 063105. Rethfeld, B., Ivanov, D.S., Garcia, M.E., and Anisimov, S.I. (2017). Modelling ultrafast laser ablation. J. Phys. D. Appl. Phys. 50 (39pp): 193001. Ikuta, Y., Kikugawa, S., Hirano, M., and Hosono, H. (2000). Defect formation and structural alternation in modified SiO2 glasses by irradiation with F2 laser or ArF excimer laser. J. Vac. Sci. Technol. B 18: 2891–2895. Takahashi, M., Uchino, T., and Yoko, T. (2002). Correlation between macro- and microstructural changes in Ge:SiO2 and SiO2 glasses under intense ultraviolet irradiation. J. Am. Ceram. Soc. 85: 1089–1092. Kajihara, K., Skuja, L., and Hosono, H. (2014). Diffusion and reactions of photoinduced interstitial oxygen atoms in amorphous SiO2 impregnated with 18 O-labeled interstitial oxygen molecules. J. Phys. Chem. C 118: 4282–4286.

References

83 Takigawa, Y., Kurosawa, K., Sasaki, W. et al. (1990). Si-O bond breaking in SiO2 by

vacuum ultraviolet laser radiation. J. Non-Cryst. Solids 116: 293–296. 84 Akazawa, H. (2001). Formation of Si-Si bonds and precipitation of Si nanocrystals

in vacuum-ultraviolet-irradiated a-SiO2 films. J. Vac. Sci. Technol. B 19: 649–658. 85 Shibuya, T., Takahashi, T., Sakaue, K. et al. (2018). Deep-hole drilling of amorphous

silica glass by extreme ultraviolet femtosecond pulses. Appl. Phys. Lett. 113: 171902. 86 Tanaka, K. (1999). Spectral dependence of photoexpansion in As2 S3 glass. Philos.

Mag. Lett. 79: 25–30. 87 Terakado, N. and Tanaka, K. (2007). Photoinduced phenomena in GeO2 -GeS2

glasses. Jpn. J. Appl. Phys. 46: L265–L267. 88 Tanaka, K. (2012). Relation between photodarkening and photoexpansion in As2 S3

glass. Phys. Status Solidi B 249: 2019–2023. 89 Tanaka, K. (2004). Midgap photon effects in As2 S3 glass. Philos. Mag. Lett. 84:

601–606. 90 Tanaka, K. (2000). Sub-gap excitation effects in As2 S3 glass. J. Non-Cryst. Solids

266-269: 889–893. 91 Aleksandrov, A.P., Babin, A.A., Kiselev, A.M. et al. (2001). Formation of

92 93 94 95

96

97

98 99

100

101

102

microstructures in As2 S3 by a femtosecond laser pulse train. Quant. Electron. 31: 398–400. Asobe, M. (1997). Nonlinear optical properties of chalcogenide glass fibers and their application to all-optical switching. Opt. Fiber Technol. 3: 142–148. Ahmad, R. and Rochette, M. (2011). Photosensitivity at 1550 nm and Bragg grating inscription in As2 Se3 chalcogenide microwires. Appl. Phys. Lett. 99: 061109. Li, M., Huang, S., Wang, Q. et al. (2014). Nonlinear lightwave circuits in chalcogenide glasses fabricated by ultrafast laser. Opt. Lett. 39: 693–696. Olivier, M., Tchahame, J.C., Nˇemec, P. et al. (2014). Structure, nonlinear properties, and photosensitivity of (GeSe2 )100-x (Sb2 Se3 )x glasses. Opt. Mater. Express 4: 525–540. Petit, L., Carlie, N., Anderson, T. et al. (2008). Progress on photoresponse of chalcogenide glasses and films to near-infrared femtosecond laser irradiation: a review. IEEE J. Sel. Top. Quantum Electron. 14: 1323–1334. Kadan, V., Blonskyi, I., Shynkarenko, Y. et al. (2017). Single-pulse femtosecond laser fabrication of concave microlens- and micromirror arrays in chalcohalide glass. Opt. Laser Technol. 96: 283–289. Ballato, J. and Dragic, P. (2014). Materials development for next generation optical fiber. Materials 7: 4411–4430. Pelusi, M.D., Ta’eed, V.G., Fu, L. et al. (2008). Applications of highly-nonlinear chalcogenide glass devices tailored for high-speed all-optical signal processing. J. Sel. Top. Quantum. Electron. 14: 529–539. Zakery, A. and Shahmirzaee, H. (2007). Modeling of enhancement of nonlinearity in oxide and chalcogenide glasses by introduction of nanometals. Phys. Lett. A 361: 484–487. Mouawad, O., Amrani, F., Kibler, B. et al. (2014). Impact of optical and structural aging in As2 S3 microstructured optical fibers on mid-infrared supercontinuum generation. Opt. Express 22: 23912–23919. Dai, S., Wang, Y., Peng, X. et al. (2018). A review of mid-infrared supercontinuum generation in chalcogenide glass fibers. Appl. Sci. 8: 707.

293

294

10 Nonlinear Optical Properties of Photonic Glasses

103 Mägi, E.C., Fu, L.B., Nguyen, H.C. et al. (2007). Enhanced Kerr nonlinearity

104 105

106

107 108 109 110

111 112

113

in sub-wavelength diameter As2 Se3 chalcogenide fiber tapers. Opt. Express 15: 10324–10329. Foster, M.A., Turner, A.C., Lipson, M., and Gaeta, A.L. (2008). Nonlinear optics in photonic nanowires. Opt. Express 16: 1300–1320. Elliott, G.R., Hewak, D.W., Murugan, G.S., and Wilkinson, J.S. (2007). Chalcogenide glass microspheres; their production, characterization and potential. Opt. Express 15: 17542–17553. Ferrera, M., Razzari, L., Duchesne, D. et al. (2008). Low-power continuous-wave nonlinear optics in doped silica glass integrated waveguide structures. Nat. Photonics 2: 737–740. Suzuki, K. and Baba, T. (2010). Nonlinear light propagation in chalcogenide photonic crystal slow light waveguides. Opt. Express 18: 26675–26685. Kuznetsov, A.I., Miroshnichenko, A.E., Brongersma, M.L. et al. (2016). Optically resonant dielectric nanostructures. Science 354 (2472). Hussein, H.M.E., Ali, T.A., and Rafa, N.H. (2018). A review on the techniques for building all-optical photonic crystal logic gates. Opt. Laser Technol. 106: 385–397. Ung, B. and Skorobogatiy, M. (2011). Extreme optical nonlinearities in chalcogenide glass fibers embedded with metallic and semiconductor nanowires. Appl. Phys. Lett. 99: 121102. Khan, A.A., Jabar, M.S.A., Jalaluddin, M. et al. (2015). Spectral hole burning via Kerr non linearity. Commun. Theor. Phys. 64: 473–478. Curry, R.J., Birtwell, S.W., Mairaj, A.K. et al. (2005). A study of environmental effects on the attenuation of chalcogenide optical fibre. J. Non-Cryst. Solids 351: 477–481. Miller, S.E. (1969). Integrated optics: an introduction. Bell System Tech. J. 48: 2059–2069.

295

11 Optical Properties of Organic Semiconductors Takashi Kobayashi 1,2 and Hiroyoshi Naito 1,2 1

Department of Physics and Electronics, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka, Japan The Research Institute for Molecular Electronic Devices, Osaka Prefecture University, 1-1 Gakuen-cho, Sakai, Osaka, Japan 2

CHAPTER MENU Introduction, 295 Molecular Structure of π-Conjugated Polymers, 296 Theoretical Models, 298 Absorption Spectrum, 300 Photoluminescence, 304 Non-Emissive Excited States, 306 Electron–Electron Interaction, 309 Interchain Interaction, 314 Conclusions, 320 References, 321

11.1 Introduction Organic materials have fascinating optical properties and have been extensively investigated in many research fields associated with light. Compared to inorganic materials, a great advantage of organic materials is their variety. In inorganic materials, repeating units consist of either the same atoms or a few different atoms at the most, so there is not much room to control one property of the material while keeping the rest unchanged. In organic molecules, there are an infinite number of combinations of atoms, and therefore it is possible to design organic materials to have desirable properties. In addition, molecular arrangement and its dimensionality can also be controlled. Therefore, organic materials have received considerable research attention as a model system for understanding light–material interaction. From the application viewpoint, for fabricating light-emitting devices and nonlinear optical devices, π-conjugated polymers are the most promising materials because of their good processability, high fluorescence yield, large optical nonlinearity, and ultrafast class relaxation time. In this chapter, we will present the fundamental optical properties of π-conjugated polymers, including the underlying physics, and some related optical spectroscopic techniques.

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

296

11 Optical Properties of Organic Semiconductors

11.2 Molecular Structure of 𝛑-Conjugated Polymers Although π-conjugated polymers are mainly composed of only carbon and hydrogen atoms, there are many kinds of polymer backbones. A few representative polymer backbones are shown in Figure 11.1. The most important feature of π-conjugated polymers is that unsaturated carbons lie on a chain, over which π electrons are delocalized one-dimensionally. The side chains provide thermal stability, solubility, hydrophilicity, and many other properties to the polymer backbones, but the fundamental optical (and electrical) properties are determined by the π electrons. In this section, we will first review the chemical structure of π-conjugated polymers, and then introduce some theoretical approaches in the next section. Carbon consists of an atomic nucleus and six electrons, and their electronic configuration is 1s2 2s2 2p2 in the ground state. To form four bonding orbitals, the carbon is first promoted to the configuration of 1s2 2s 2p3 , and then one electron in the 2s orbital and some in the 2p orbitals are hybridized. For instance, in diamond, all of the four electrons in the L-shell form four identical sp3 hybridized orbitals to bond with four carbon atoms. In π-conjugated polymers, instead of forming the sp3 hybridized orbitals, one electron in the 2s orbital and two electrons in the 2p orbitals are hybridized to three identical sp2 orbitals, which are in the same plane and are arranged at an angle of about 120∘ (see Figure 11.2a). Since they do not have angular momentum in the bonding direction, they are called σ orbitals. Each carbon atom is combined with other carbon atoms and hydrogen atoms by the σ orbitals to form the backbone of π-conjugated polymers (see Figure 11.2b). These σ orbitals have maximum amplitude in the bonding direction, so that overlap of the σ orbitals of adjacent carbon atoms is expected to be large. In fact, the σ bonds are strong and cannot be excited by photons in the visible range. On the other hand, the remaining pz orbital in the carbon atom is called the π orbital, and does not have its maximum amplitudes in the bonding direction, so that the π orbital overlaps only slightly with other π orbitals of the neighboring carbon atoms above and below the plane containing the σ bonds to form π bond(s). The π bonds are thus relatively weak. The semiconducting properties of π-conjugated polymers are (a) Polyacetylene

(b) Polydiacetylene H C

H C

(c) Polythiophene S C

C

C

C

H

C H

n

H

H C C

C

C C

C C

H

(e) Polyfluorene H

H

H C C

C

C

H H n

H

n

(d) Poly(p-phenylene vinylene) H

C

C C

C C

C C

H

H

H

C

C C

C C H

H

C

H

n

Figure 11.1 Some backbone structures of π-conjugated polymers.

n

11.2 Molecular Structure of π-Conjugated Polymers

120°

π bond

pz orbital

sp2orbitals

C

C

C

C

C

σ bond (a)

(b)

Figure 11.2 (a) sp2 hybridized orbitals and pz orbital. (b) σ and π bonds in a single chain of polyacetylene. (In this figure, hydrogen atoms and σ bonds connecting carbon and hydrogen atoms are omitted.)

determined by the electrons in the π bonds. In fact, it is possible to extract an electron from and add one into the π bond without breaking up the molecule. Usually, electrons in the σ bonds do not contribute to the semiconducting properties of the polymers. An electron in the linked π orbitals (the π bond) and one in the σ orbitals (the σ bond) are sometimes called the π electron and σ electron, respectively. Let us consider a π-conjugated polymer with an infinite polymer chain, where the linked π orbitals form the π band. Each orbital has two states, i.e., the spin-up and -down states, but supplies only one electron. Thus, one may expect that the π band is a half-filled band, and therefore π-conjugated polymers should show efficient conductivity like metals. However, in the actual polymers, the intervals between carbon atoms are alternately modified to form bond alternation in order to reduce the total energy (see Figure 11.3). This alternation splits the π band into two equal half bands separated by a band gap; the lower half is fully filled, and the upper one is completely empty. Although this bond alternation increases the elastic energy, the energy of the electrons is lowered sufficiently to compensate for the increase. Consequently, π-conjugated polymers H

H

H

H

H

H

C

C

C

C

C

C

C

C

C

C

H

H

H

H

Empty band

Band gap

Half-filled band (a)

Filled band (b)

Figure 11.3 (a) In the case where carbon atoms are equally spaced, the linked π orbitals form a half-filled band. In the above chemical structure, the solid lines are the σ bonds, and the dashed lines indicate the linked π orbitals. (The dashed lines are not π bonds, because each carbon atom has only four bonding orbitals.) (b) In the case where the intervals between carbon atoms are alternately modulated, the half-filled band splits into a filled band and an empty band separated by a band gap.

297

298

11 Optical Properties of Organic Semiconductors

show the electronic properties of semiconductors or insulators. The gap energy usually corresponds to the energy of a photon in the visible light range, and thus the response of π-conjugated polymers to visible light is exclusively due to π electrons. It may be noted that the bond alternation in the polymer backbone is depicted using a single line as well as a double line, but the bond alternation does not include purely single bonds. The interatomic distances between carbon atoms are modified only slightly by the alternation. Several electron diffraction measurements have revealed that the bond lengths are modified by less than 5% from the mean length [1]. Therefore, π electrons are considered to be delocalized on the whole polymer backbone. Although an infinite polymer chain is considered here, the same scenario can be applied to finite chains and small π molecules as well. In such cases, the electric bands are represented by the molecular discrete energy levels: the bottom level of the empty (conduction) band and the top level of the filled (valence) band are represented by the lowest unoccupied molecular orbital (LUMO) and the highest occupied molecular orbital (HOMO), respectively.

11.3 Theoretical Models Although a single chain of π-conjugated polymers can be considered as a onedimensional semiconductor, its theoretical treatment is not simple because of some interactions. In inorganic materials, it is usually assumed that nuclei do not move even after excited states are created. However, in π-conjugated polymers, it is essential to take into consideration coupling between electronic states and bond orders. To do this, the tight binding Hamiltonian including the electron–phonon (or electron–lattice) interaction is used as [2] ∑ 1∑ + K(un+1 − un )2 − tn+1,n (Cn+1 Cn + Cn+ Cn+1 ), (11.1) H= 2 n n where K is the elastic constant, un is the displacement of the carbon atom at site n, and C + n and C n are electron creation and annihilation operators, respectively. The second term in Eq. (11.1) has the same form as in the tight binding model, but the transfer integral, t n+1,n , is defined to be proportional to the interval between the carbon atoms at n + 1 and n using the following relation: tn+1,n = t0 − 𝛼(un+1 − un ),

(11.2)

where 𝛼 is a parameter indicating coupling strength between an electron and nucleus, and t 0 is a constant. This model can be applied to polymers with degenerate ground state, like polyacetylene: in polyacetylene, cis, and trans forms have the same ground state energy (see Figure 11.4a). However, in π-conjugated polymers with ring structures such as polyfluorene and polythiophene, the ground state of geometrical isomers is not degenerate (see Figure 11.4b). In order to include this non-degeneracy effect in such polymers, the following transfer integral is alternatively used: tn+1,n = [1 + (−1)n 𝛿0 ] t0 − 𝛼(un+1 − un ),

(11.3)

where 𝛿 0 is a dimensionless bond alternation parameter and is introduced here to solve the non-degenerate ground states. Calculations using these Hamiltonians have

11.3 Theoretical Models

E

E

Q

Q

H

H

H

H

H

H

H

H

H

H

H

S

S

H H

(a)

H

H

H

(b)

Figure 11.4 Examples of π-conjugated polymers with (a) degenerate and (b) non-degenerate ground states.

succeeded in explaining topological charged excited states, such as solitons, polarons, bipolarons, and molecular deformations due to photoexcitations, which are discussed in later sections. It is known that the electron–electron interaction is also essential in understanding the electronic and optical properties of π-conjugated polymers, and it is taken into account using the following Hubbard-Peierls Hamiltonian [3]: ∑ + + tn+1,n (Cn+1,σ Cn,σ + Cn,σ Cn+1,σ ) H=− n,σ

+

∑ n

U𝜌n↓ 𝜌n↑ +

1 ∑∑ V 𝜌 𝜌 ′, 2 n≠m σ,σ′ m,n n,σ m,σ

(11.4)

where U is the on-site Hubbard repulsion (the nearest-neighbor hopping integral), V is the nearest-neighbor charge density–charge density interaction, 𝜌n,σ = C+ n,σ Cn,σ , and σ indicates spin (up or down). We do not solve this Hamiltonian in this chapter because it is still difficult to solve it for a realistic π-conjugated polymer consisting of more than a few hundred sites and containing significant structural disorders. Here, we would only like to stress that electron–phonon and electron–electron interactions play an essential role in the electronic structure of π-conjugated polymers and that all of their optical and electronic properties cannot be described within the framework of band theory or the effective mass approximation, which are very efficient theoretical approaches to understand the basic optical and electronic properties of inorganic semiconductors. However, it is also true that many similarities exist between inorganic and organic materials. For instance, according to the one-dimensional exciton theory developed for inorganic semiconductors, discrete exciton energy levels appear below the continuum state, and most of the oscillator strength concentrates on the lowest exciton level. These features are observed in π-conjugated polymers as well (see Figure 11.5). Therefore, more familiar and intuitive theories for inorganic materials can be approximately used for organic materials as long as their applicable range is taken into account. The electron–phonon interaction is taken into account throughout this chapter, but the effect of electron–electron interaction is only briefly reviewed in Section 11.7.

299

11 Optical Properties of Organic Semiconductors

Continuum band

Figure 11.5 Schematic energy band structure in π-conjugated polymers.

Exciton levels

Ground state

11.4 Absorption Spectrum

1.2

0.25

1.0

0.20

0.8

0.15

0.6 0.10 0.4 0.05

0.2 0.0 1.5

2.0

2.5

3.0

3.5

4.0

Photon Energy (eV)

4.5

0.00 5.0

PC Yield (arb. units)

Absorption measurement is the most fundamental spectroscopy and is very helpful in understanding the characteristic features of π-conjugated polymers, including large electron–phonon interactions. Figure 11.6 shows the absorption spectrum of spin-coated film of polyfluorene, where a broad and featureless band is observed at 3.2 eV. This band corresponds to the transition from the ground state to the lowest excited state, and the broad width of the band results from significant inhomogeneous broadening. In π-conjugated polymers, ideally, π electrons are delocalized over the whole polymer backbone, but structural disorders, for example, chemical defects and twist of the polymer backbone, limit the delocalization of π electrons and raise the resonance energy of the polymer. Since actual polymers have a large distribution of the delocalization length of π electrons, especially in disordered films and in solution, such a broad and featureless absorption spectrum is often observed. For analyzing the absorption spectrum in π-conjugated polymers, the simple band picture without Coulomb interaction (electron–electron interaction) is not valid; this, in fact, can be confirmed from a comparison between absorption and photoconductivity (PC) yield spectra (see Figure 11.6). The PC yield indicates the probability that an absorbed photon generates a pair of charged carriers. If the Coulomb interaction is negligible in a system, a photoexcitation always produces a pair of oppositely charged carriers, which will contribute to a photocurrent in the system. On the other hand, in a system with electron–electron interaction, the excited pairs of charge carriers can form excitons due to their Coulomb interaction, and then an excess energy will be necessary to separate such a pair from each other. This required excess energy can be estimated

Absorption (arb. units)

300

Figure 11.6 Absorption and photoconductivity (PC) yield spectra of spin-coated film of a polyfluorene derivative. The exciton binding energy can be estimated from the difference between the onsets of absorption and PC yield spectra.

11.4 Absorption Spectrum

from the photon energy difference between the onsets of absorption and PC yield spectra and is called the “exciton binding energy.” The exciton binding energy of the spin-coated film of a polyfluorene derivative is estimated to be about 0.1 eV from the comparison between absorption and PC yield spectra shown in Figure 11.6. In many other π-conjugated polymers, similar exciton binding energies have been reported [4, 5]. Although the band in the PC yield spectrum in Figure 11.6 corresponds to the continuum state shown in Figure 11.5, it is difficult to recognize the corresponding band in the absorption spectrum. This is because most of the oscillator strength concentrates on the lowest exciton level, which is a major feature of one-dimensional systems. Some π-conjugated polymers have a tendency to form an ordered packing structure. While polymer chains usually adopt a randomly twisted or random-coil conformation in disordered films and solutions, they adopt a more planar or sometimes a fully planar conformation(s) in ordered films. An improvement in the chain planarity increases the delocalization length of the π electrons, which reduces the resonance energy and suppresses inhomogeneous broadening. Although ordered films of a polyfluorene derivative can be prepared [6, 7], we here show the results of a polythiophene derivative and its isomer; in these, identical side chains are attached to the polythiophene backbones in two different ways [8]. In the solid state, the side chains attached in one way promote crystallization to form an ordered packing structure, whereas the chains attached in the other way prevent the crystallization. Figure 11.7 shows the absorption spectra of ordered and disordered films prepared by spin-coating from their chloroform solutions. The absorption spectrum of the disordered film is almost the same as that of its chloroform solution. The difference between the two absorption spectra is basically attributed to the difference in the chain planarity. The influence of the interchain interaction is not significant and is discussed in Section 11.8. The absorption spectrum of the ordered film is red-shifted with respect to that in solution and show a well-resolved vibronic structure, which consists of several bands with the same energy interval and is due to the electron–phonon interaction. In order to explain this vibronic structure, we have Disordered

Absorbance

Ordered

H2 C

S

2.0

CH3 O C H2

C H2

*

C H2

CH3

n

2.5 3.0 Photon Energy (eV)

3.5

Figure 11.7 Absorption spectra of ordered and disordered films of a polythiophene derivative. The inset is the chemical structure of the derivative. Source: Reprinted Figure 1 with permission from T. Kobayashi et al., Phys. Rev. B 62, 8580. Copyright (2000) The American Physical Society.

301

11 Optical Properties of Organic Semiconductors

E

E

E

e

e

ћω0

Abs.

g

g Q

Q ∆Q (a)

PL E

302

(b)

Figure 11.8 Schematic illustration of vibronic potential energy curves for (a) absorption and (b) photoluminescence processes. In this figure, e and g indicate electronic excited and ground states. ℏ𝜔0 is the phonon energy of the associated mode, and ΔQ is the difference between the potential minima of the g and e curves. If ΔQ is zero, i.e. the case of no electron–phonon interaction, the transitions to the higher vibrational levels are forbidden.

schematically illustrated the energy potential curves in Figure 11.8, where a few vibrational energy levels of a phonon mode are indicated by the horizontal lines, and the potential minimum of the excited state is slightly shifted from that of the ground state. The magnitude of this shift, ΔQ, represents the strength of electron–phonon interaction in the system. In the system with nonzero ΔQ, transitions from the zero vibrational level in the ground state to the excited vibrational levels in the excited state are allowed, and discrete transition bands, i.e. a vibronic structure, appear in the absorption spectrum. The transition to the zero vibrational level in the excited state is called a “0-0 transition” and corresponds to a purely electronic transition. On the other hand, the other transitions are called “0-1 transition,” “0-2 transition,” etc. After a photoexcitation corresponding to a 0-n transition, an electronic excited state is created, and n phonons are emitted. In many π-conjugated polymers, the associated phonon mode is a C=C stretching mode with a phonon energy of around 0.18 eV. Such a vibronic structure is always observed in their ordered films. Although π-conjugated polymers have many phonon modes, most of their phonon energies are much less than 0.18 eV, and hence their contributions can be included in inhomogeneous broadening. The vibronic structure can be simulated by taking into account the electron–phonon interaction and the associated phonon mode [9]. In a system without electron–phonon interactions, the transition matrix element can be calculated simply from m=



Ψ0e (r)∗ (−er)Ψ0g (r)dr,

(11.5)

where Ψ0 g and Ψ0 e are electronic wave functions for the ground and excited states, respectively, and r is the electronic coordinate. However, in a system with significant electron–phonon interactions, the electronic state is influenced by the displacement of atoms, which can be taken into account by adding the following term as a perturbation in the Hamiltonian: Hint (r, Q) = −u(r) Q,

(11.6)

11.4 Absorption Spectrum

where Q is the generalized coordinate for the associated phonon mode. In Eq. (11.6), we neglect the higher term of Q for simplicity. The perturbed wave function is then written as Φin (r, Q) = Ψi (r, Q) ⋅ 𝜉in (Q),

(11.7)

where i and n indicate the electronic state and vibrational level, respectively, and 𝜉 in describes the vibrational wave function of atoms. This perturbed wave function is referred to as the “vibronic wave function.” Using this vibronic wave function, the transition matrix element of transitions from the ground vibrational level of the ground state to the nth vibrational level of the excited state results in M0n =



=



Φg0 (r, Q)∗ (−er)Φen (r, Q − ΔQ)drdQ Ψg (r)∗ (−er)Ψe (r)dr ×

=m×





𝜉g0 (Q)∗ 𝜉gn (Q − ΔQ)dQ

𝜉g0 (Q)∗ 𝜉gn (Q − ΔQ)dQ,

(11.8)

where we focus on the transition only from the zero vibrational level in the ground state. In Eq. (11.8), the first factor is the same as that given in Eq. (11.5), which involves only transitions between electronic states. The second factor as an integral in Eq. (11.8), which depends on the vibrational levels, determines the vibronic structure. Since the absorption (and photoluminescence (PL)) intensity is proportional to the square of the transition matrix element, the following formula is more practical: | |2 F0n = || 𝜉g0 (Q)∗ 𝜉gn (Q − ΔQ)dQ|| . |∫ |

(11.9)

Equation (11.9) is referred to as the Franck–Condon factor and is denoted by F 0n . Using the harmonic oscillator approximation, Eq. (11.9) can be simplified into the following form: e−S Sn , (11.10) n! where the Huang–Rhys parameter, S, represents the strength of the electron–phonon interaction for an associated phonon mode, and is given by F0n (S) =

ΔQ 2 . (11.11) 2 In Figure 11.9, we show some vibronic structures calculated using Eq. (11.10) for several Huang–Rhys parameters. Equation (11.10) is the same as the Poisson distribution, and always gives the maximum intensity for the 0-S transition. For instance, in a case where S = 2, the 0-2 transition has the maximum intensity. However, it is still difficult to reproduce perfectly the observed absorption spectrum of π-conjugated polymers using Eq. (11.10). This is because the spectrum has a large inhomogeneous broadening even in ordered films and contributions from higher excited states. S=

303

304

11 Optical Properties of Organic Semiconductors

Figure 11.9 Examples of vibronic structure calculated using Eq. (11.10) for several Huang–Rhys parameters, S. S = 0.5

S =1

S =2

S =5

0 1 2 3 4 5 6

11.5 Photoluminescence Many π-conjugated polymers show PL in the visible spectral range. For instance, polyfluorene, poly(p-phenylene vinylene) (PPV), and polythiophene are known to be blue, green, and red emitters, respectively. Some of their derivatives have good fluorescence yield and are expected to be used in fabricating light-emitting devices. In these materials, after excited states are created by photoexcitation, they immediately relax into the lowest excited state (Kasha’s rule) and then emit light. Therefore, their PL straightforwardly reflects the nature of the lowest excited state. In the PL spectrum, a vibronic structure also appears due to transitions from the lowest vibrational level in the excited state to vibrational levels in the ground state (see Figure 11.8b). This process can also be described using Eq. (11.10), and then only the Huang–Rhys parameter determines the vibronic structure under the harmonic oscillator approximation. Therefore, symmetrical absorption and PL spectra are expected to be observed. This symmetry is called “mirror image.” Two examples calculated using Eq. (11.10) are shown in Figure 11.10. The difference between the absorption and the PL maxima is called the Stokes shift, which results from many relaxation mechanisms of the polymer chain occurring after the photoexcitation. However, as shown in the lower figure of Figure 11.10, a vibronic structure with a large Huang–Rhys parameter could be the main reason for the larger Stokes shift. In this case, the Stokes shift is roughly estimated to be 2ℏ𝜔0 S, where ℏ𝜔0 is the associated phonon energy. After the photoexcitation, the bond alternation is slightly modified to reduce the total energy of the excited state by forming a self-trapped excited state. The structure of the self-trapped excited state in polythiophene is shown in Figure 11.11, where

11.5 Photoluminescence

Stokes shift ABS.

S = 0.5

PL

Figure 11.10 Absorption and PL spectra calculated using Eq. (11.10). The difference between their maxima is called the Stokes shift. These are ideal cases where only one excited state appears in the observed spectral range. In actual π conjugated films, the observed absorption spectrum is not entirely in agreement with one expected from the PL spectrum, because the observed absorption spectrum contains inhomogeneous broadening and contributions from higher excited states.

Stokes shift

PL

ABS.

S = 2.5

Photon Energy H

H

H

H

H

H

S S

S H

H (a)

S

H

H e– S e+

S

H

H (b)

Figure 11.11 Bond arrangements (a) in the ground state and (b) in the self-trapped excited state for polythiophene.

an excited state is depicted as localized on one thiophene unit for clarity. However, an actual excited state is still delocalized within several thiophene units, although the delocalization length is shorter than that of the π electrons in the ground state. This self-trapping starts immediately after the photoexcitation and completes within 10 ps [8]. Therefore, the observed PL is emitted from the self-trapped state, and this reduction in energy also contributes to the observed Stokes shift. In disordered films of π-conjugated polymers, the π electrons are not fully delocalized: the twist of the polymer backbone limits delocalization of the π electrons. In such cases, films should be considered as an ensemble of segments with various π-conjugation lengths. The resonance energy of a segment is lowered as its delocalization increases. Therefore, after the photoexcitation, excited states migrate from shorter segments to longer ones prior to emission of PL. This energy migration can be an additional reason for the large Stokes shift observed in disordered films. However, PL does not always originate from the longest segment in disordered films. As an excited state migrates to longer segments, it becomes more difficult to find further longer segments nearby. Within its lifetime, an excited state can only migrate to segments with a certain length of the π conjugation. When a disordered film is excited by photons of high enough energy, the PL spectrum is independent of the excitation photon energy. On the other hand, when the

305

11 Optical Properties of Organic Semiconductors

photoluminescence (arb.units)

306

13

14

15 16 17 18 wavenumber (103 cm–1)

19

20

Figure 11.12 Site-selective fluorescence measurements on a disordered PPV film. The different PL spectra were obtained by varying the excitation energy (indicated by the spikes at the high-energy side), starting at the top far from resonance and moving into resonance going down the figure. Source: Reprinted Figure 4 with permission from H. Bässler and B. Schweitzer, Acc. Chem. Res. 32, 173. Copyright (1999) The American Chemical Society.

photon energy is less than the threshold, PL shows dependence on the photon energy. In this condition, those segments to which an excited state cannot migrate from shorter segments can be directly excited, and PL from these segments can be observed. Such a measurement is called “site-selective fluorescence measurement,” and an example of such measurements is shown in Figure 11.12 [10, 11]. Another factor that increases the Stokes shift is the interchain interaction, which is discussed in Section 11.8.

11.6 Non-Emissive Excited States As in inorganic semiconductors, it is possible to generate charge carriers in π-conjugated polymers also by doping. In π-conjugated polymers, these carriers are called solitons, polarons, or bipolarons, where bond alternation is modified, and charge is delocalized within 10–30 carbon atoms [2] as illustrated in Figure 11.13. Solitons are formed in the π-conjugated polymers that have a degenerated ground state (Figure 11.8a) such as polyacetylene and polydiacetylene, whereas polarons and bipolarons are formed in those π-conjugated polymers with non-degenerated ground states (Figure 11.8b),

11.6 Non-Emissive Excited States

e+ Neutral state

Soliton

e+ Polaron Neutral state

e+ e+ Bipolaron

Figure 11.13 Charged excited states in π-conjugated polymers. In these polymers, charge induces modification of the bond alternation to form stable excited states.

such as polyfluorene and polythiophene. These carriers govern conductivity in doped π-conjugated polymers. In undoped π-conjugated polymers, solitons or polarons can be created by photoexcitations, but they decay nonradiatively within a lifetime of microseconds to milliseconds. Thus, the formation of such charged excitations serves as one of the nonradiative decay channels in some π-conjugated polymers. Generation of triplet excited states via intersystem crossing is also one of the nonradiative decay channels because phosphorescence is extremely weak in conventional π-conjugated polymers [12], which do not contain heavy metals. The lifetime of triplet excited states at a very low temperature is comparable to that of the charged states and is several orders of magnitude longer than that of singlet excitons (around nanoseconds). Such long-lived photoexcitations in π-conjugated polymers can be experimentally observed by continuous-wave (cw) photo-induced absorption (PA) measurements. In PA measurements, a pump beam creates photoexcitations, and a probe beam detects the transitions of photoexcitations to higher excited states. Cw PA measurements are usually conducted using a cw laser and a lock-in-amplifier with a mechanical chopper, so that ultrafast photoexcitations contribute little to cw PA signals, and only long-lived photoexcitations appear in the observed spectrum. In Figure 11.14, we show cw PA spectra of disordered and ordered films of poly(3-hexylthiophene) (P3HT) [13, 14]. In this work, two isomers were used to prepare the ordered and disordered films. The lower figures in Figure 11.14 are photo-induced absorption-detected magnetic resonance (PADMR) spectra [13, 14], from which it is possible to know the spin number of each PA band. In a disordered P3HT film, only a broad PA band is observed, which can be assigned to a triplet–triplet transition because the spin number is 1. In contrast, in an ordered P3HT film, the PA band with spin 1 disappears, and several PA bands with spin 1/2 appear instead. As mentioned above (see Figure 11.3), the band gap in π-conjugated polymers is opened because of bond alternation, which is locally modified after polaron formation. As a result, one localized state is separated from the bottom of the conduction band and another from the top of the valence band. This situation is illustrated in Figure 11.15. In the case of polarons with a positive charge, the lower localized state is occupied by an electron whereas the higher one is empty, and thus new two transitions become possible. These transitions are in fact observed below the band

307

11 Optical Properties of Organic Semiconductors

(a)

3 3 T

2

2

1

1

105(δT/T)

103(–ΔT/T)

PA

0

0

–1

–1 Spin 1

–2

–2 PADMR

–3

–3 4

2 PA

2 DP1 0

(×1/4)

IEX

P2

1 DP2

P1

0 Spin 1/2

–2

–1

–4

104(δT/T)

(b)

103(–ΔT/T)

308

–2 PADMR

–6

0

0.5

1 1.5 Photon Energy(eV)

2

–3

Figure 11.14 (Upper) PA and (lower) PADMR spectra of (a) disordered and (b) ordered P3HT films. In the upper figures, the PA bands correspond to transitions of the long-lived excited states. The lower figures show the spin number of each PA band. Source: Reprinted Figure 5 with permission from O. J. Korovyanko et al., Phys. Rev. B 64, 235122. Copyright (2001) The American Physical Society.

P2

P1

Neutral State

Polaron

Figure 11.15 Energy structure of polarons. Solid lines in the left side correspond to the bottom of the conduction band and the top of the valence band. Polarons have two localized states. Solid arrows represent electrons (spin direction), and broken arrows indicate transitions appearing within the band gap after the polaron formation.

11.7 Electron–Electron Interaction

gap, as shown in Figure 11.14b; the corresponding PA bands are labeled P1 and P2. From the fact that these P1 and P2 bands are also observed in isolated polythiophene chains in an inert polystyrene matrix, these PA bands are attributed to one-dimensional polarons. Ordered P3HT films contain the disordered and crystallized portions of the chains. One-dimensional polarons are considered to form in the disordered portions. In the crystallized portions, on the other hand, polarons delocalizing over several neighboring chains are formed. Such delocalized polarons have two localized states that are slightly shifted due to the interchain interaction, and thus exhibit one PA band lower than the P1 band and the other higher than the P2 band [14]; these bands due to the delocalized polarons are denoted by DP1 and DP2 in Figure 11.14b. Note that in addition to the PA bands with spin 1/2, a PA band with spin zero of the interchain excitons (IEX) is also observed at 1.7 eV, and is attributed to interchain excitations different from conventional singlet excited states. After all, many kinds of long-lived, non-emissive excited states may be created by photoexcitation in films of π-conjugated polymers.

11.7 Electron–Electron Interaction To properly understand the optical properties of π-conjugated polymers, it is also essential to consider the electron–electron interaction. One example showing the importance of the electron–electron interaction is the large exciton binding energy, as discussed in Section 11.4. In this section, we present another example related to the fluorescence yield, along with some experimental techniques used for investigation of the excited-state structure of π-conjugated polymers, which is influenced by the electron–electron interaction. We begin with a brief review of a basic concept of group theory that is applicable to π-conjugated polymers. As a simple model of π-conjugated polymers, let us consider a polyene, C2n H2n+2 (see Figure 11.16a). This molecular structure possesses C2h group symmetry and is invariant under the operation of space inversion at the symmetry center or rotation about the symmetry axis though 180∘ . In group theory, wave functions whose sign changes and remains unchanged after the rotation are labeled “b” and “a,” respectively. Furthermore, wave functions whose sign changes and remains unchanged after the space inversion are labeled “u” and “g,” respectively. The pz orbital has a shape represented by two touching spheres with opposite signs (see Figure 11.16a), and thus the sign of the pz orbital reverses after the space inversion. Considering this fact, the sign of π electrons consisting of 2n pz orbitals can be easily determined. Figures 11.16b and c are two examples of combinations of 2n pz orbitals having different symmetry. In Figure 11.16b, all the “+” spheres are located in the +z direction. In this case, the sign of this π electron level does not change after the rotation but changes after the space inversion. Thus, this π electron level has au symmetry. Similarly, the π electron level shown in Figure 11.16c has bg symmetry. When all combinations of six pz orbitals are examined, you will find that all π electron levels in the molecule can be classified into “au ” or “bg ” states. As illustrated in Figure 11.17, this molecule has 2n energy levels of π electrons, and therefore au and bg states appear alternately from the bottom to the top. Since each level has spin-up and -down states, this molecule has 4n states for 2n π electrons. Thus, the electronic configuration of this molecule is determined by the way that 2n π electrons are arranged to

309

310

11 Optical Properties of Organic Semiconductors

H



H

H



C

+

C

C

+

– –

C

+

+ H

H





+

C

x

C

z

H

+ H

y H

pz orbital

(a) An example of bg state

An example of au state

+ +

+ +

+

+ +

+

– +

– –

The symmetry center (b)

(c)

Figure 11.16 Symmetry of electronic states in C2n H2n+2 (in the case of n = 3). (a) The molecular structure of C2n H2n+2 with the emphasized pz orbitals. The shape of pz orbital is expressed by two touching spheres. In this figure, the touching points are in the plane containing the carbon atoms, and since the two spheres of each pz orbital above and below the plane have opposite signs, they are distinguished by “+” and “–.” (b), (c) Two examples of combinations of 2n pz orbitals. bg au LUMO

bg

HOMO

au bg au Ground Ag state Excited Bu state

Excited Ag state

Figure 11.17 Symmetry of configuration of C2n H2n+2 (in the case of n = 3). The molecule has 2n energy levels of π electrons, and in the ground state, 2n energy levels are occupied by 2n π electrons from the bottom (the left). The middle and left configurations are examples of the excited state of the molecule.

occupy 4n states. In Figure 11.17, three examples of electronic configurations are illustrated. The symmetry of a configuration can be calculated by multiplying the symmetry of 2n π electrons using the following relations: a × a = b × b = a,

g × g = u × u = g,

a × b = a × b = b,

g × u = u × g = u.

(11.12)

From this simple calculation, we find that a system with C2h symmetry has only Bu and Ag states (here, we use capital letters for the symmetry of configurations). In the ground state, 2n π electrons fill only half of the 4n states starting from the bottom, as shown in the left side of Figure 11.17. When one of the π electrons from HOMO is excited into LUMO, the symmetry of the configuration becomes Bu , as shown in the middle of Figure 11.17. Furthermore, when the excited π electron is further excited into the next lowest unoccupied level, the symmetry becomes Ag again. Therefore, the system

11.7 Electron–Electron Interaction

has an electronic energy structure as 1Ag (ground state), 1Bu , 2Ag , 2Bu ,…, nAg , and nBu appear from the lowest to the highest levels. This is always true whenever one electron theory, such as the Hartree–Fock approximation, is valid for the system. However, if electron–electron interaction is not negligible, the electronic configuration in the right side may become a more stable state than that in the middle in Figure 11.17, and then the 2Ag state will appear as the lowest excited state. Since the ground state is the 1Ag state, transitions from the 1Ag state into Bu states are one-photon allowed, and transitions into Ag states are one-photon forbidden. Thus, Ag states are absent in the conventional absorption spectrum. According to Kasha’s rule, photoexcitations into higher excited states immediately relax nonradiatively to the lowest excited state. If the lowest excited state is the 1Bu state, a PL with a radiative lifetime in the picosecond to nanosecond range would be observed. If the lowest excited state has Ag symmetry, the radiative lifetime increases dramatically, and most of the photoexcitations preferably decay nonradiatively. Therefore, the symmetry of the lowest excited state is one of the most important factors in determining the fluorescence yield. One of the experimental techniques that is suitable for a study on Ag states is the two-photon absorption (TPA) technique. When an intense laser pulse is irradiated into a sample, an excited state at an Ag state is created with a very low probability via simultaneous TPA, which is one of the nonlinear optical responses (see Figure 11.18). From the energy conservation law, the separation between the Ag state and the ground state is equal to twice the incident photon energy. In the TPA process, Bu states are forbidden. Thus, with this TPA technique, Ag states are selectively investigated. It is, however, difficult to measure the TPA cross section with a satisfactory signal-to-noise (S/N) ratio. To improve the S/N ratio, the linear dependence of the TPA cross section on the laser intensity is usually used. If a sample has a good fluorescence yield, the two-photon excitation (TPE) coefficient is alternatively measured; the TPE measurements are technically much easier than the TPA ones. In the TPE measurements, excited states are created via TPA, and then fluorescence from the lowest Bu state is measured. In this case, the square dependence of the fluorescence intensity on the laser intensity is utilized to determine the TPE coefficient with a good S/N ratio. Figure 11.19a shows the absorption and TPE spectra of thin films of a polyfluorene derivative [15]. In this figure, the 2Ag state at around 3.9 eV lies above the 1Bu (I) band in the absorption spectrum. A similar excited-state structure with the 1Bu state as the lowest excited state is observed in other luminescent π-conjugated polymers such as PPV [16]. On the other hand, Figure 11.19b shows the absorption and TPA spectra of a single crystal of a polydiacetylene derivative [17]. Polydiacetylene is known to be a non-luminescent π-conjugated polymer and has a fluorescence yield less than 10−5 [18]. In this case (Figure 11.19b), the Ag state is of lower energy than the lowest 1Bu state. Other non-luminescent π-conjugated polymers Figure 11.18 Two-photon absorption process. In the one-photon process, transition between states with the same symmetry is forbidden, and transition between states with opposite symmetries is allowed. In the two-photon process, this selection rule reverses: Transition from the ground (1Ag ) state to a higher Ag state is allowed, but in this process it is required that the energy interval between the two Ag states is equal to the sum of the photon energies of the two incident photons.

2Ag ћω

ћω 1Ag

311

300 1.0 250 0.8

I α2 (cm/GW)

200

0.6

7 Sample 1 Sample 2 6 α (cm–1) 5 4

150

0.4 0.2 0.0 2

×5

3 100

2

50 II 3 4 Photon Energy (eV) (a)

5

0 1.5

α (cm–1 × 105)

11 Optical Properties of Organic Semiconductors

Absorbance

312

1 2 Energy (2hv eV) (b)

0 2.5

Figure 11.19 (a) Squares represent the TPE spectrum of a polyfluorene derivative, whereas the solid line indicates the absorption spectrum. In this figure, “I” and “II” are major absorption bands, i.e. Bu states. Source: Reproduced with permission from S. Ikame et al., Phys. Rev. B 75, 035209. Copyright (2007) The American Physical Society. (b) The solid line is the absorption spectrum of a single crystal of a polydiacetylene derivative (right scale). The empty and filled circles are TPA coefficients obtained from two different samples, and the dashed line serves as a guide for the eye. Below 1.9 eV, the vertical scale is expanded to make clearer the two resonances below the lowest Bu state at 2.0 eV. Source: Reprinted with permission Figure 2 from B. Lawrence et al., Phys. Rev. Lett. 73, 597. Copyright (1994) The American Physical Society.

such as polyacetylene also have such an excited-state structure [19]. Thus, the absence of PL in these polymers is attributed to Ag states being below the lowest Bu state. It should be noted that although the electron–electron interaction is still important in the luminescence of π-conjugated polymers, ring structures in the polymer backbone such as phenyl and five-membered rings are known to stabilize the 1Bu state [20]. The electroabsorption (EA) measurement has also been used as an experimental method to investigate Ag states in π-conjugated polymers. In this measurement, an AC electric field with frequency f is applied to the film, and any change in the absorption coefficient at frequency 2f is detected with a lock-in amplifier. In this process, the applied electric field couples the Ag and Bu states and modifies the energy separation between the ground and excited states. This modification results in a very small absorption spectral change (the absorption coefficient changes by about 0.1% at most). As an example, we show in Figure 11.20 the EA spectrum of thin films of poly(p-phenylene ethynylene) (PPE) [21]. In the EA spectra of many π-conjugated polymers, a signal due to the red shift of the lowest absorption band appears (at around 2.5 eV in the case of Figure 11.20). This red shift is called the Stark shift and results from the strong repulsion between the 1Bu state and higher Ag states; generally, the repulsion between the 1Ag and 1Bu states does not contribute efficiently because of their large energy interval. Theoretically, the EA spectrum is described by the third-order nonlinear susceptibility, which is calculated using the third-order perturbation theory and is proportional to the product of four transition dipole moments [22]. In the two-level model, the dipole transition moment between the two states has to be used four times to calculate the third-order nonlinear susceptibility (see Figure 11.21a) and consequently

11.7 Electron–Electron Interaction

0.4

3

0.2 2 0

10–4 (ΔT/T)

Figure 11.20 Example of EA spectrum of π-conjugated polymers. The solid and dashed lines are the experimental and theoretical ones, respectively. The inset shows in more detail the EA feature at high photon energy. Source: Reproduction with permission from M. Liess et al., Phys. Rev. B 56, 15712. Copyright (1997) The American Physical Society.

1

–0.2 3

3.5

4

0 PPE

–1 2

3 Energy [eV]

4

3.5

(a) 2

1

1Bu 1Ag

0

–1

(b) mAg 1Bu

5 10–4 (ΔT/T)

Figure 11.21 EA spectra simulated using (a) two-level model, (b) three-level model, and (c) three-level model taking into consideration asymmetric inhomogeneous broadening and the vibronic structure. In the insets, energy levels and the ways to choose four transition dipole moments to simulate the EA spectra are depicted. In (c), an example of asymmetric inhomogeneous broadening is also illustrated. Source: Reprinted Figure 6 with permission from M. Liess et al., Phys. Rev. B 56, 15712. Copyright (1997) The American Physical Society.

2.5

1Ag 0

–5

(c)

Energy shift distribution

2

mAg 1Bu

1

1Ag 0

–1 2.0

2.5

3.0 3.5 Energy [eV]

4.0

313

314

11 Optical Properties of Organic Semiconductors

a blue shift in the absorption band is always anticipated. In a three-level model, two kinds of transition dipole moments can be used (see Figure 11.21b), and a term using both transition dipole moments two times each also contributes to the third-order nonlinear susceptibility. This term, which explains the observed red shift, becomes dominant in actual π-conjugated polymers where the energy difference between the 1Ag and 1Bu states is much larger than that between the 1Bu and mAg states. As shown in Figure 11.21b, the highest level (the mAg state) can be found as a small positive peak at 3.3 eV. When inhomogeneous broadening and vibronic replicas are taken into consideration in the simulation using the asymmetric Gaussian function and the Franck–Condon factor, respectively, the experimental EA spectrum of π-conjugated polymers can be reproduced well (see the dashed lines in Figure 11.20) [21]. By fitting the simulated spectrum to experimental one, it is possible to determine the energy levels essential to describe the nonlinear property and dipole moments between these levels.

11.8 Interchain Interaction Some π-conjugated polymers form a well-ordered packing structure in the solid state as mentioned above (see Figure 11.7). In such films, inhomogeneous broadening is suppressed, and their optical properties are expected to be changed due to the interchain interaction. Evidence of the interchain interaction presented so far in this chapter is the transitions due to delocalized polarons (indicated by DP1 and DP2 in Figure 11.14), which are charged states delocalizing among several polymer chains [13]. The influence of the interchain interaction can also be seen in more fundamental optical properties, such as absorption and PL properties. Before discussing the interchain interaction in π-conjugated polymers, we briefly review a theory developed for aggregates or single crystals of small molecules to get a basic picture of the intermolecular interaction, which will be helpful in understanding the interchain interaction in π-conjugated polymers. Each organic molecule has many electronic energy levels that are determined from the chemical structure. Most of these original characters are preserved even in the presence of the intermolecular interactions. Therefore, we can describe the effect of intermolecular interaction using perturbation theory. At first, we consider a simple system consisting of two identical molecules (see Figure 11.22). When the molecules are close enough, each energy level of the isolated molecules split into two levels due to intermolecular interaction. The energy of the interacting system can be written as Esystem = Emolecule + ΔD ± ΔE,

(11.13)

where Emolecule is the energy of the isolated molecule, ΔE is the splitting energy due to the intermolecular interaction, and ΔD indicates an energy shift due to other effects, for example, the effects described by the dielectric constant of the surrounding medium. Generally, ΔD cannot be estimated quantitatively and is treated as a negative constant. If the intermolecular interaction is described by the dipole-dipole interaction, we obtain (M ⋅ r)(M2 ⋅ r) M ⋅M , (11.14) ΔE = 1 3 2 − 3 1 r r5

11.8 Interchain Interaction

2∆E Emolecular

Emolecular+∆D

Ground state Interacting molecules

Isolated molecule (a)

(b)

+

= 0

+

=

+

=

+

=

+

= 0

+

=

(c)

(d)

Figure 11.22 (a) Intermolecular interaction and the resultant energy splitting. On the left are shown the energy levels of an isolated molecule, and on the right are the energy levels of the interacting two molecules. 2ΔE is the splitting energy due to the intermolecular interaction, and ΔD is the energy shift due to other effects. (b)–(d) Molecular arrangements and dipole moments. The ellipses and the inside arrows indicate the interactive molecules and their transition dipole moments. Below the ellipses, their energy levels are depicted. In (b) and (c), the upper and lower energy levels of the split branches, respectively, do not have any transition dipole moment because the two transition dipole moments of individual molecules cancel each other, and only one band is observed in their absorption spectra. On the other hand, in (d), both branches have some transition dipole moments, and two absorption bands would be observed.

where M1 and M2 are transition dipole moments of the molecules, and |r| = r is the distance from one molecule to the other. From this equation, we find that ΔE depends not only on the distance between the two molecules but also on the relative angle between them. Three arrangements of a pair of molecules and their energy diagrams are illustrated in Figures 11.22b–d. In Figure 11.22b, the molecules are in a line, in (c) they are aligned parallel to each other, and in (d) they are in a non-parallel arrangement. In part (b), the lower branch corresponds to a parallel arrangement of two dipole moments, and the higher branch corresponds to an antiparallel arrangement. Since the transition dipole moments are vectors, they cancel each other in the antiparallel arrangement, and then the parallel arrangement has twice the oscillator strength of the isolated molecule. Consequently, in part (b), only the lower branch is observed in the absorption spectrum. In part (c), the antiparallel arrangement appears in the lower branch, and the parallel arrangement appears in the higher branch. Therefore, in this case, the lower branch has no transition dipole moment. In a non-parallel arrangement as shown in Figure 11.22d, the splitting energy may become smaller than the former two cases, and both branches have nonzero transition dipole moments. This means that two branches would be found

315

316

11 Optical Properties of Organic Semiconductors

in the absorption spectrum. The magnitude and direction of the total transition dipole moment of two interacting molecules are calculated by the linear combination of each transition dipole moment. This simple model can also be applied to molecular aggregates where more than two molecules are interacting. When the molecules align in a head-to-tail manner as in Figure 11.22b, an energy band is formed, and the oscillator strength of the system concentrates on the lowest energy level. In such a case, excited states (Frenkel excitons) are delocalized over several molecules and become less sensitive to a slight misalignment of the molecules and local vibrations. This effect is called the motional narrowing effect [23] and reduces inhomogeneous broadening and suppresses the vibronic structure. As a result of the aggregation, a narrow absorption band red-shifted with respect to that of the isolated molecules is observed. This type of aggregate is called J-aggregates. On the other hand, aggregates with a side-by-side molecular arrangement as in Figure 11.22c are called H-aggregates and exhibit an absorption band that is relatively blue-shifted. Frequently, ΔE > ΔD, so that the observed absorption band is blue-shifted with respect to that of the isolated molecule. Another feature of H-aggregate is a suppressed fluorescence yield. In H-aggregates, most of the excited states preferably decay through nonradiative channels, such as by phonon emission, because of the zero transition dipole moment of the lowest excited state. In actual H-aggregates, misalignment of the molecules weakly induces PL from the lowest excited state. In such a case, the observed Stokes shift is equal to 2ΔE. In contrast, in J-aggregates, an enhanced transition dipole moment of the lowest excited state increases the fluorescence yield, and the Stokes shift ideally becomes zero. For a more detailed study on J-aggregates, readers may refer to Kobayashi’s book [24] and a review by Mobius [25]. In this chapter, we present two recent works as examples. The first one is reported by Liess et al. [26], who have synthesized three merocyanine dyes that have the same chromophore backbone but different substituents. Since these merocyanine dyes in dilute solution exhibit very similar absorption spectra, it is confirmed that the substituents do not directly alter their optical properties. However, the substituents significantly influence the packing structures of the molecules, and consequently they exhibit entirely different absorption spectra in the solid state. Interestingly, two of these absorption spectra are typical to J- and H-aggregates. In fact, the red-shifted one is due to the J-aggregate formation, and the blue-shifted one is due to the H-aggregate formation (see Figure 11.23). It should be noted that in the J-aggregates, the molecules are slightly tilted alternately to form a zigzag conformation so that the spectral shift, i.e. ΔE, is relatively small compared to that in the H-aggregates counterpart. The second example is of J- and H-aggregate formations of a nitroazo dye (see Figure 11.24) [27]. When the dye is thermally evaporated onto a glass substrate, an absorption spectrum that can be explained in terms of H-aggregation is observed. Tanaka et al. [27] have also deposited the dye onto a one-dimensionally aligned polytetrafluoroethylene (PTFE) thin layer, which can be prepared by thermal evaporation of PTFE and then rubbing with a cloth or by a friction-transfer method. On the aligned PTFE thin layer, the dye not only aligns macroscopically but also forms J-aggregates. The dye probably has the potential to form two different packing structures in the solid state, and one may be more stable than the other depending on the surface energy and/or the surface structure. These two examples clearly show that intermolecular interaction has a strong impact on the absorption spectrum.

11.8 Interchain Interaction

H

M

500

600

J

λ (nm)

700

Figure 11.23 Absorption spectra of substituted merocyanine dyes in solution (M; monomer) and in solid state (H- and J-aggregates). Double wall arrows represent the dipole moment of the molecule and the molecular arrangements. Source: Reprinted the table-of-contents graphic with permission from A. Liess et al., Nano Lett. 17, 1719. Copyright (2017) The American Chemical Society.

(a)

(b)

(c)

Absorbance (arb. unit)

Figure 11.24 Absorption spectrum of (a) H-aggregates of a nitroazo dye deposited on a glass substrate, (b) the isolated dye in solution, and (c) J-aggregates of the dye formed on an aligned PTFE thin layer. Since the J-aggregates are highly oriented, the spectrum was taken with polarized light. Source: Reprinted Figure 4 with permission from T. Tanaka et al., Langmuir 32, 4710. Copyright (2016) The American Chemical Society.

0 300

400

500 600 Wavelength (nm)

700

800

It should be noted that the concept shown in Figure 11.22 can also be applied to single crystals of small molecules. According to the theory of Frenkel excitons, all unit cells have the same transition dipole moment in exciton levels with k = 0, which is the essential condition for optical absorption. First, we consider the case where a unit cell contains only a single molecule. If the intermolecular interaction between two neighboring molecules aligning in a head-to-tail manner is dominant, the level with k = 0 appears at the bottom of the exciton band. This situation is similar to that illustrated in Figure 11.22b. On the other hand, if the interaction in the face-to-face direction is dominant, the level with k = 0 appears at the top of the exciton band, and the transition dipole moment at the bottom is completely canceled out, as expected from Figure 11.22c. In the case where two equivalent molecules exist in a unit cell, the two energy levels with k = 0 would have nonzero transition dipole moments, as illustrated in Figure 11.22d. In fact, two absorption bands separated by the intermolecular interaction energy ΔE are observed in some molecular crystals. This separation in molecular crystals is called Davydov splitting [28]. In the case of anthracene, for example, the splitting energies of

317

318

11 Optical Properties of Organic Semiconductors

the lowest excited state with a moderate transition dipole moment and the third lowest excited state with a huge transition dipole moment are about 0.03 and 2 eV, respectively [29]. Since the two branches usually have the transition dipole moments with different directions, the relative intensity of the two absorption bands changes depending on the crystal plane. Can this concept be straightforwardly applied to ordered films of π-conjugated polymers? Some conjugated polymers form well-ordered packing structures in the solid state and exhibit clear X-ray diffraction patterns, which indicate that the polymer chains are aligned in a side-by-side manner as shown in Figure 11.22c. π-conjugated polymers usually have a large transition dipole moment between the ground state and the lowest excited state. Therefore, if the interchain interaction is large, it is expected that ordered samples will exhibit optical properties similar to those of the H-aggregates of small molecules. However, the actual ordered films of a polyfluorene derivative, for example, exhibit good fluorescence yield and a very small Stokes shift [6], which must be equal to 2ΔE if H-aggregates were formed. In the case of polythiophene derivatives, their ordered films have suppressed fluorescence yield in the solid state, but show red-shifted absorption spectra and small Stokes shifts with respect to those in solution (see Figure 11.25a). These features are apparently inconsistent with the expectations based on the analogy of the H-aggregates of small molecules. Because of these inconsistencies, the interchain interaction in π-conjugated polymers has remained controversial for many years. In the case of π-conjugated polymers that can form ordered packing structures in the solid state, the chain planarity is significantly improved during the solidification. The improvement in the chain planarity induces a large spectral red shift because of elongation of the delocalization length of the π-electrons (excitons). Therefore, the effect of the interchain interaction cannot be identified from a simple comparison between solution and solid-state samples. This is in striking contrast to aggregates of small molecules, in which the differences in optical properties between solid-state and solution samples can be ascribed to intermolecular interaction. For small molecules, ΔE is given in Eq. (11.14), which is derived from the point dipole approximation. This approximation is valid only when the length of a molecule (or its wave function) is sufficiently short compared to the distance between the two molecules, which is obviously invalid for π-conjugated polymers. From recent studies by Spano et al. [30, 31], it has been revealed that the interchain interaction in π-conjugated polymers is either weak or negligible depending on the interchain distance and the delocalization length of π electrons. Along with the interchain interaction, a topic that has been discussed for a long time is the PL spectral shape (a vibronic structure) of ordered polythiophene films (Figure 11.25a), whose 0-0 transition is much smaller than that expected from the simplified Franck–Condon factor, i.e. Eq. (11.10). In contrast, the PL spectral shape of ordered polyfluorene films can be perfectly reproduced using Eq. (11.10) [7]. Spano et al. [30, 31] have developed a theoretical model that considers interchain (excitonic) and electron–phonon interactions to understand the interchain interaction and the PL spectral shape of ordered films of π-conjugated polymers. Using the theoretical model, it is found that the vibronic structure gets slightly modified when the interchain interaction is in the weak regime. On the other hand, when the interchain interaction is strong enough, vibronic replicas should disappear because of the motional narrowing effect, as observed in the J- and H-aggregates of small molecules (see Figure 11.23, for example). The theoretical model also shows that the change in the vibronic structure

PL (a.u.)

Figure 11.25 (a) PL and absorption spectra of an ordered P3HT film measured at 6 K. The dashed line represents an absorption spectrum of its chloroform solution at room temperature. (b) PL and excitation spectra of ordered thin films of low-MW P3HT at 6 K. The dashed and solid lines indicate the results of Form I and II, respectively. Note that MWs of P3HT used were (a) 13 300 and (b) 3000. Source: Reproduced with permission from T. Kobayashi et al., Nanoscale Res. Lett. 12, 368. Copyright (2017) Springer.

Absorbance (a.u.)

11.8 Interchain Interaction

1.5

2.0 2.5 Photon Energy (eV)

3.0

PL (normalized)

Excitation Intensity (a.u.)

(a)

1.5

2.0 2.5 Photon Energy (eV) (b)

3.0

due to electron–phonon interaction is more noticeable in PL than in absorption. So we here discuss only the PL spectral shape. In the H-aggregates, the bottom of the exciton band appears at k = π so that an optical transition from the bottom to the electronic ground state with zero vibrational level (k = 0) is forbidden. However, transitions to the nonzero vibrational levels are allowed because the momentum is conserved through the involvement of an optical phonon mode (the Einstein model) [30, 31]. As a result, in the absence of any disorder, it is expected that H-aggregates of π-conjugated polymers will show PL without a 0-0 transition. In actual samples, misalignment of the polymer chains and residual slight twists of the polymer backbones increase the 0-0 transition in the PL spectrum. The observed lower fluorescence yield and small Stokes shift are also explained in terms of H-aggregates with the weak interchain interaction (ΔE). Recently, Kobayashi et al. [32] have reported an experimental study on the interchain interaction using polymorphic modifications of P3HT. Low-molecular-weight (MW) P3HT is known to form two different packing structures in the solid state. In one structure (Form I), which is commonly formed by high-MW P3HT, planar polymer chains stack in a face-to-face manner with a stacking (interchain) distance of 3.8 Å, and each stack is separated by around 16 Å by the inert hexyl side chains. The X-ray study has revealed that in the other structure (Form II), the stacking distance increases to 4.4 Å, whereas the separation between the stacks slightly decreases [33]. Since the chain planarity is similarly improved in both packing structures, a comparison between the two polymorphic modifications makes it possible to focus on the effects of the stacking distance in P3HT ordered samples. Kobayashi et al. [32] have succeeded in fabricating

319

320

11 Optical Properties of Organic Semiconductors

thin films that mainly consist of either Form I or II and compared their optical properties (see Figure 11.25b). The Form I sample shows a PL spectrum that is very similar to that known as the PL spectra of ordered film of conventional (i.e. high-MW) polythiophene derivatives. This means that the difference in MW is minor in the PL properties. On the other hand, the Form II sample exhibits a different PL spectrum that is blue-shifted by more than 0.1 eV with respect to that of the Form I sample and has a larger and narrower 0-0 transition. The narrow width indicates that the observed PL is not emitted from a disordered portion contained in the sample. Thus, the blue shift and the enhanced 0-0 transition can be attributed to the further weaker interchain interaction due to the larger stacking distance. The Form II sample has larger fluorescence yield and a smaller Stokes shift compared to those of the Form I counterpart. All of these differences can be consistently explained in terms of the weaker interchain interaction. In other words, the differences in optical properties of the Form I and II samples can be regarded as an experimental evidence of the interchain interaction in π-conjugated polymers. As mentioned above, ordered films of a polyfluorene derivative exhibit very efficient PL with a vibronic structure that can be reproduced with the simplified Franck–Condon factor [7]. In those films, the interchain interaction can be completely ignored. This is because the interchain distance is even larger than that in the P3HT Form II sample [34, 35], and the delocalization length of π electrons is also larger; the latter can be confirmed by the very narrow PL band observed at a very low temperature [6]. A theoretical treatment of the competition between inter- and intrachain excitations in ordered films of π-conjugated polymers can be found in Ref. [31].

11.9 Conclusions π electrons delocalizing on a one-dimensional network of carbon atoms govern the optical properties of π-conjugated polymers in the visible range. To treat the π electrons theoretically, electron–phonon and electron–electron interactions should be included in the Hamiltonian, and the interchain interaction may have to be considered for the case where the polymer chains form a well-ordered packing structure. However, it is still difficult to solve such Hamiltonians because of the huge size of the molecule and the structural disorders (twists of the polymer backbones and misalignment in the packing structure), which also have a large impact on the optical properties. The electron–phonon interaction determines the vibronic structure in the absorption and PL spectra, and the importance of the electron–electron interaction is confirmed by the large exciton binding energy as well as the appearance of a one-photon forbidden excited (Ag ) state below the lowest one-photon allowed (1Bu ) state in certain π-conjugated polymers. In ordered films, π-conjugated polymers may form H-aggregates due to the weak interchain interaction. Low fluorescence yield and a PL spectrum with a suppressed 0-0 transition are signatures of the interchain interaction. In disordered films and solutions, the polymer backbones are randomly twisted. Such backbones should be treated as ensemble of segments with various delocalization lengths of π electrons and thus exhibit different optical properties. Consequently, an improvement in the chain planarity suppresses the inhomogeneous broadening observed in absorption and PL spectra. Along with these fundamental optical properties of π-conjugated polymers, some associated experimental techniques are described in this chapter.

References

References 1 Drenth, W. and Wiebenga, E.H. (1955). Acta Crystallogr. 8: 755. 2 Heeger, A.J., Kivelson, S., Schrieffer, J.R., and Su, W.-P. (1988). Rev. Mod. Phys. 60:

781. 3 Baeriswyl, D., Campbell, D.K., and Mazumdar, S. (1992). An overview of the theory

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

of π-conjugated polymers. In: Conjugated Conducting Polymers (ed. G. Kiess), 7–133. Berlin: Springer-Verlag. Pakbaz, K., Lee, C.H., Heeger, A.J. et al. (1994). Synth. Met. 64: 295. Köhler, A., dos Santos, D.A., Beljonne, D. et al. (1998). Nature 392: 903. Ariu, M., Lidzey, D.G., Sims, M. et al. (2002). J. Phys. Condens. Matter 14: 9975. Asada, K., Kobayashi, T., and Naito, H. (2006). Jpn. J. Appl. Phys. 45: L247. Kobayashi, T., Hamazaki, J., Arakawa, M. et al. (2000). Phys. Rev. B 62: 8580. Henderson, B. and Imbusch, G.F. (1989). Optical Spectroscopy of Inorganic Solids. Oxford: Clarendon Press. Heun, S., Mahrt, R.F., Greiner, A. et al. (1993). J. Phys. Condens. Matter 5: 247. Bässler, H. and Schweitzer, B. (1999). Acc. Chem. Res. 32: 173. Monkman, A.P., Burrows, H.D., Hartwell, L.J. et al. (2001). Phys. Rev. Lett. 86: 13583. Korovyanko, O.J., Österbacka, R., Jiang, X.M. et al. (2001). Phys. Rev. B 64: 235112. Österbacka, R., An, C.P., Jiang, X.M., and Vardeny, Z.V. (2000). Science 287: 839. Ikame, S., Kobayashi, T., Murakami, S., and Naito, H. (2007). Phys. Rev. B 75: 035209. Frolov, S.V., Bao, Z., Wohlgenannt, M., and Vardeny, Z.V. (2000). Phys. Rev. Lett. 85: 2196. Lawrence, B., Torruellas, W.E., Cha, M. et al. (1994). Phys. Rev. Lett. 73: 597. Soos, Z.G., Etemad, S., Galvao, D.S., and Ramasesha, S. (1992). Chem. Phys. Lett. 194: 341. Orlandi, G., Zerbetto, F., and Zgierski, M.Z. (1991). Chem. Rev. 91: 867. Barford, W. (2005). Electronic and Optical Properties of Conjugated Polymers. Oxford: Clarendon Press. Liess, M., Jeglinski, S., Vardeny, Z.V. et al. (1997). Phys. Rev. B 56: 15712. Boyd, R.W. (1992). Nonlinear Optics. San Diego: Academic Press. Knapp, E.W. (1984). Chem. Phys. 85: 73. Kobayashi, T. (ed.) (1996). J-aggregates. Singapore: World Scientific. Mobius, D. (1995). Adv. Mater. 7: 437. Liess, A., Lv, A., Arjona-Esteban, A. et al. (2017). Nano Lett. 17: 1719. Tanaka, T., Ishitobi, M., Aoyama, T., and Matsumoto, S. (2016). Langmuir 32: 4710. Davydov, A.S. (1971). Theory of Molecular Excitons. New York: McGraw-Hill. Schwoerer, M. and Wolf, H.C. (2007). Organic Molecular Solids. Weinheim: Wiley-VCH. Spano, F.C. and Silva, C. (2014). Annu. Rev. Phys. Chem. 65: 477. Spano, F.C. (2006). Annu. Rev. Phys. Chem. 57: 217. Kobayashi, T., Kinoshita, K., Niwa, A. et al. (2017). Nanoscale Res. Lett. 12: 368. Prosa, T.J., Winokur, M.J., and McCullough, R.D. (1996). Macromolecules 29: 3654. Chen, S.H., Chou, H.L., Su, A.C., and Chen, S.A. (2004). Macromolecules 37: 6833. Surin, M., Hennebicq, E., Ego, C. et al. (2004). Chem. Mater. 16: 994.

321

323

12 Organic Semiconductors and Applications Furong Zhu Department of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong, China

CHAPTER MENU Introduction, 323 Anode Modification for Enhanced OLED Performance, 327 Flexible OLEDs, 345 Solution-Processable High-Performing OLEDs, 353 Conclusions, 368 References, 369

12.1 Introduction Silicon-based transistors and integrated circuits are of central importance in the current microelectronics industry, which serves as an engine to drive progress in today’s electronics technology. However, there is a great need for significant new advances in the rapidly expanding field of electronics. New materials and innovative technologies are predicted to lead to developments beyond anything we can imagine today. The demand for more user-friendly electronics is propelling efforts to produce head-worn and hand-held devices that are flexible, lighter, more cost-effective, and more environmentally benign than those presently available. Electronic systems that use organic semiconductor materials offer an enabling technology base. This technology has significant advantages over the current silicon-based technology because it allows for an astonishing amount of electronic complexity to be integrated onto lightweight, flexible substrates for the production of a wide range of entertainment, wireless, wearable-computing, and network-edge devices. For example, organic light-emitting devices (OLEDs) that use a stack of organic semiconductor layers have the potential to replace liquid crystal displays (LCDs) as the dominant flat panel display device. This is because OLEDs have high visibility by self-luminescence, do not require backlighting, and can be fabricated into lightweight, thin, and flexible displays. A combination of organic electronics and existing microelectronics technologies also opens a new world of potential for electronics. The possible uses include a wide variety of industrial, medical, military, and other consumer-oriented applications.

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

324

12 Organic Semiconductors and Applications

Cathode Electron Transport Layer Emissive Layer Hole Transport Layer

Transparent Cathode Electron Transport Layer Emissive Layer

ITO Hole Transport Layer Glass Substrate Anode Si Substrate (a)

(b)

Figure 12.1 (a) A conventional bottom-emitting OLED is made on a transparent substrate, e.g. a glass or clear plastic substrate, and (b) a top-emitting OLED (TOLED) requires a semitransparent top cathode. A TOLED can be made on both transparent and opaque substrates.

12.1.1

Device Architecture and Operation Principle

A typical OLED is constructed by placing a stack of organic electroluminescent and/or phosphorescent materials between a cathode layer that can inject electrons, and an anode layer that can inject “holes.” Polymeric electroluminescent materials have been used in OLEDs, and are referred to as polymer light-emitting devices. A conventional OLED has a bottom-emitting structure, which includes a metal or metal alloy cathode, and a transparent anode on a transparent substrate, enabling light to be emitted from the bottom of the structure (Figure 12.1a). An OLED may also have a top-emitting structure, which is formed on either an opaque substrate or a transparent substrate. A top-emitting OLED (TOLED) has a relatively transparent top cathode so that light can be emitted from the side of the top electrode (Figure 12.1b). Figure 12.1 shows a cross-sectional view of (a) a typical bottom-emitting OLED and (b) a top-emitting OLED. In a multilayered OLED, the organic medium consists of a hole-transporting layer (HTL), a light emissive layer (EML) and an electron-transporting layer (ETL). Indium tin oxide (ITO) is often used as the transparent anode due to its high optical transparency and electric conductivity. The cathode in the OLEDs is made of low-work-function metals or their alloys, e.g. MgAg, Ca, LiAl, etc. A schematic energy level diagram for an OLED under bias is shown in Figure 12.2. When a voltage of proper polarity is applied between the cathode and anode, “holes” injected from the anode and electrons injected from the cathode combine radiatively to release energy as light, thereby producing electroluminescence. The light usually escapes through the transparent substrate. Different organic semiconductor functional layers in an OLED can be optimized separately for carrier transport and luminescence. A large number of conducting small molecular materials and conjugated polymers have been used as the charge-transporting or emissive layer in OLEDs. The molecular

12.1 Introduction

Figure 12.2 Schematic energy level diagram of an OLED, showing the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) of the hole-transporting layer (HTL) and electron-transporting layer (ETL), which are also referred to as affinitive energy level (EA) and ionization potential (IP). ΦITO and Φm represent the work function of the ITO anode and metallic cathode. ΔEh and ΔEe are the barrier heights at the ITO/organic and organic/cathode interfaces. The hole electron current balance in an OLED is set by the size of the barriers at the two electrodes.

m

vacuu

EA

Φm

O

LUM

ΦITO

∆Ee

IP

O HOM ∆Eh ITO

Org

Metal

structures of some commonly used organic semiconductor materials are illustrated in Figure 12.3. A small-molecule-based OLED device has a typical configuration of glass/ITO/HTL/organic emissive layer/ETL/Mg:Ag mixture/Ag. The multilayer thin film device is fabricated using the thermal evaporation method in a vacuum system. The thermal evaporations usually start at the base pressure of 10−6 mbar or lower. The fabrication of polymeric OLEDs involves solution processes, and the devices can be made using spin-coating or inkjet printing methods. After the deposition of organic semiconductor layers, a thin electron injector is evaporated following a 150–300 nm thick Al or Ag layer to inhibit oxidation of the cathode contact. The device operation principle and the fabrication process of an organic-semi conductor-based light-emitting device are different from those known for a conventional light-emitting device made with an inorganic semiconductor. It is very difficult to form a stable organic semiconductor p-n junction, as the organic materials cannot be doped reproducibly to form p-type and n-type semiconductors. The interfaces of organic semiconductor p-n junctions are easily degraded or even destroyed by chemical reaction and/or inter-diffusion. For this reason, OLEDs are usually designed having a p-i-n configuration, where the emissive layer is nominally intrinsic, although in practice it gets automatically doped. The second major difference between an OLED and an inorganic LED is the nature of the charge carrier transport, recombination, and luminescence processes. In an inorganic semiconductor, charge transport is delocalized and described in terms of Bloch states within the single-electron band approximation. However, charge transport in amorphous organic semiconductors is characterized by the localization of electronic states to individual molecules and occurs via thermally activated hopping processes [1]. 12.1.2

Technical Challenges and Process Integration

Organic semiconductors are finding increasing use in organic electronics, including flat panel displays, organic transistors, photodetectors, and solar cells. However, processing

325

326

12 Organic Semiconductors and Applications

N O O N

Al

N

N

N

O

Alq3

NPB

n n

R

PPV

R PFO

O

n

O

S

n SO3H

PEDOT

PSS

Figure 12.3 Molecular structures of some small molecular and polymeric semiconductors commonly used in OLEDs. Tris(8-hydroxyquinoline) aluminum (Alq3 ) is used as an electron-transporting and emissive layer. 𝛼-napthylphenylbiphenyl (NPB) is used as a hole transport layer (HTL). Poly(p-phenylene vinylene) (PPV) and poly(fluorine) (PFO) are typical fluorescent polymers. Poly(styrene sulfonate)-doped poly(3,4-ethylene dioxythiophene) (PEDOT) is often used as a HTL in polymeric OLEDs.

these materials with the desired uniform thin film patterns or multilayer structure is challenging. A spin-coating process is probably the simplest method of producing thin films of a few hundred angstroms, but these films are amorphous with low carrier mobility and may have pinholes. Films produced using the Langmuir–Blodgett method are typically highly ordered, but this method is best suited for thin film formation. Inkjet printing can also be used to produce patterned polymeric thin films, but such films are amorphous. Although the printable transparent conducting and functional organic semiconductor thin film technologies are still in the research stage and have not yet been used in products, the related technologies thus developed are promising for next-generation products. Each currently available thin film deposition method has advantages and disadvantages, depending on the requirements. In addition, for many

12.2 Anode Modification for Enhanced OLED Performance

applications it would be desirable to produce patterned films, but photolithographic techniques cannot be used to pattern these polymeric films. Thus, the challenge is to obtain highly ordered patterned polymeric thin films. New techniques need to be developed for depositing patterned, high-quality pinhole-free ultrathin organic films for electronics.

12.2 Anode Modification for Enhanced OLED Performance Transparent conducting oxide (TCO) thin films have widespread applications due to their unique properties of high electrical conductivity and optical transparency in the visible spectrum range. The distinctive characteristics of the TCO films have been applied in anti-static coatings, heat mirrors, solar cells [2, 3], flat panel displays [4], sensors [5], and OLEDs [6–8]. A number of materials such as ITO, tin oxide, zinc oxide and cadmium stannate are used as TCO in many optoelectronic devices. It is well known that the optical, electrical, structural, and morphological properties of TCO films have direct implications for determining and improving the device performance. The properties of TCO films are usually optimized accordingly to meet the requirements in various applications involving TCO. The light scattering effect due to the usage of textured TCO substrates shows an enhanced absorbance in thin film amorphous silicon solar cells [9, 10]. The ITO contact used in LCDs comprises a relatively rough surface in order to promote the good adhesion of subsequently coated polymeric layer on its surface. However, a rough ITO surface is detrimental for OLED applications. The high electric fields created by the rough anode can cause shorts in thin functional organic layers. 12.2.1

Low-Temperature High-Performance ITO

ITO is one of the widely used materials for TCO. Thin films of ITO can be prepared by various techniques, including thermal evaporation deposition [11, 12], direct current (dc) and radio frequency (rf ) magnetron sputtering [13, 14], electron beam evaporation [15], spray pyrolysis [16], chemical vapor deposition [17], dip-coating techniques [18, 19], and the recently developed pulsed laser deposition method [20, 21]. Among these techniques, magnetron sputtering is one of the more versatile techniques for ITO film preparation. This technique has the advantage of fabricating uniform ITO films reproducibly. Both reactive and non-reactive forms of dc/rf magnetron sputtering can be used for film preparation. ITO films prepared by the dc/rf magnetron sputtering method often require heating the substrate at an elevated temperature during the film deposition or an additional post-annealing treatment at a temperature of over 200 ∘ C. High-temperature processes for ITO preparation is unsuitable in some applications. For instance, the organic-color-filter-coated substrates for flat panel displays, flexible OLEDs made with polyester, polyethylene terephthalate (PET), and other plastic foils are not compatible with a high-temperature plasma process. Therefore, the development of high-quality ITO films with smooth surfaces, low resistivity, and high transmission over the whole visible spectrum range at low processing temperatures for flat panel displays and flexible OLEDs is quite a challenge indeed. A number of techniques have been used to

327

328

12 Organic Semiconductors and Applications

prepare ITO films at low processing temperatures. Ma et al. have deposited ITO films on polyester thin films over a substrate temperature of 80–240 ∘ C by reactive thermal evaporation [22, 23]. Laux et al. have prepared ITO films on glass substrates at room temperature by plasma ion-assisted evaporation [24]. Wu et al. [25] have used pulsed laser ablation to fabricate ITO films on glass substrates at room temperature. ITO films prepared by the radio frequency (rf ) and direct current (dc) magnetron sputtering methods on polycarbonate [26] and glass substrates [27] at low processing temperatures are also reported. ITO films fabricated by the rf and dc magnetron sputtering methods usually require a low oxygen partial pressure in the sputtering gas mixture when both alloy and oxidized targets are used [26, 27]. In the following discussion, a comprehensive study on the morphological, electrical, and optical properties of ITO films fabricated in our laboratory by rf magnetron sputtering using hydrogen–argon mixtures at a low processing temperature will be described. The addition of hydrogen in the sputtering gas mixture affects the overall optical and electric properties of ITO films considerably. Atomic force microscopy (AFM), X-ray photoelectron spectroscopy (XPS), secondary ion mass spectroscopy (SIMS), the four-point probe technique, the Hall effect, and optical measurements were used to characterize the morphological, electrical, and optical properties of the ITO films thus made. The mechanism of ITO film quality improvement due to addition of hydrogen in the sputtering gas mixture is discussed. 12.2.1.1

Experimental Methods

Thin films of ITO were prepared by rf magnetron sputtering on microscopic glass slides using an oxidized ITO target with In2 O3 and SnO2 in a weight ratio of 9:1. The background pressure in the sputter chamber was lower than 1.0 × 10−7 Torr. The deposition rate of the films prepared by rf magnetron sputtering can be controlled by the sputtering power and the substrate temperature [14]. In this study, a fixed power density of about 1.2 W cm−2 for ITO film preparation was used. The deposition process was carried out in a hydrogen–argon gas mixture at low temperature; i.e. the substrate was not heated during and after the film deposition. The total pressure of the sputtering gas was kept constant at 3.0 × 10−3 Torr during the film preparation. The hydrogen partial pressure was varied over the range 1 × 10−5 – 2.0 × 10−5 Torr. The thickness of ITO films deposited on glass substrates for morphological, electrical, optical, and spectroscopic characterizations was maintained at the same value of about 250 nm so that the measured properties of the ITO films are comparable. The thickness of the ITO films was measured by a KLA-Tencor Alpha-Step 500 profilometer. The sheet resistance of the films was determined using a four-point probe method. The charged carrier concentration and mobility of the films were characterized by Hall effect measurements using the van der Pauw technique. Wavelength-dependent absorption and transmission of ITO films were measured by a PerkinElmer spectrophotometer over the wavelength range 0.3–2.0 μm. The surface morphology of the ITO films was investigated in a DI Dimension 3000 Atomic Force microscope. The SIMS depth profiles are acquired using a CAMECA IMS 6f ion microprobe. Cs+ primary ions with an energy of 15 keV and negatively charged secondary ions were used in the SIMS analyses. XPS measurements were performed using a VG ESCALAB 220-i electron spectrometer. The Mg-K𝛼line at 1253.6 eV was chosen as the X-ray source in the

12.2 Anode Modification for Enhanced OLED Performance

XPS measurements. The position of all XPS peaks was calibrated using C 1 s with a binding energy Eb = 284.6 eV. 12.2.1.2

Morphological Properties

The rf magnetron sputtering method to deposit ITO films on glass substrates over the hydrogen partial pressure range 0–2.0 × 10−5 Torr was used. The substrate holder was not heated, and the substrate temperature during the film preparation was observed to be less than 50 ∘ C. The influence of the hydrogen partial pressure on the surface morphological properties of the ITO films was investigated by AFM. Figure 12.4 shows a typical AFM image of ITO films prepared with (a) argon, (b) hydrogen partial pressures of 7.0 × 10−6 Torr, and (c) 2.0 × 10−5 Torr. Figure 12.4 shows the similar grainy surface of the ITO films prepared at different hydrogen pressures. The average granule size of ITO films prepared at the hydrogen pressures of 7.0 × 10−6 Torr (Figure 12.4b) and 2.0 × 10−5 Torr (Figure 12.4c) had smaller dimensions than those of ITO film sputtered with argon shown in Figure 12.4a. The decrease in granule size corresponds to an increase in film smoothness. As the thickness of ITO films prepared at different hydrogen partial pressures was maintained at the same value of about 250 nm, the possible morphological difference due to thickness variations is negligible. The results, shown in Figure 12.4, actually reflect the influence of the hydrogen partial pressure on the morphological property of ITO films. The root mean square (rms) roughness of ITO films was also estimated from AFM images measured over an area of 300 nm × 300 nm as illustrated in Figure 12.4. The corresponding rms values of ITO films prepared with (a) argon, (b) hydrogen partial pressures of 7.0 × 10−6 Torr and (c) 2.0 × 10−5 Torr are 1.44, 1.13, and 0.92 nm, respectively. It is clear that the ITO film prepared with argon gas had a higher rms roughness value than that of ITO films prepared with hydrogen–argon mixtures using the same deposition parameters. This effect of hydrogen partial pressure on film morphology could be used to produce smooth ITO film with suitable opto-electrical properties in applications that are not compatible with the high-temperature process. In the application of flexible panel displays, for example, ITO is often required to be coated on the transparent plastic substrates at low processing temperatures to avoid deformation of the plastic substrates. A smooth ITO anode is also desired in a flexible OLED with a multilayered thin film configuration. An ITO anode with a smooth surface can minimize electrical shorts in the thin functional organic layers in OLEDs that are very often in the range 100–200 nm. ITO films formed by rf magnetron sputtering at low temperatures is usually amorphous. There is no lattice matching for ITO growth on glass substrates. Under this circumstance, three-dimensional (3D) nucleation is the dominant film formation mechanism. The film grown on a glass substrate is most likely formed by the coalescence of islands from nucleation sites [28]. Sun et al. [29] have investigated the initial growth mode of ITO on glass over the substrate temperature range 20–400 ∘ C. They suggested that ITO films with an amorphous structure formed at a substrate temperature below 150 ∘ C are due to a 3D nucleated growth mode similar to the Volmer–Weber mechanism. The layer-by-layer growth mode only occurred at substrate temperatures over 200 ∘ C. In this study, the ITO films were fabricated at a low processing temperature of about 50 ∘ C. The 3D-growth mode from nucleation sites is therefore expected, although the nucleation density may be varied due to the different hydrogen partial pressures

329

12 Organic Semiconductors and Applications

20 nm

(a)

300 μm 200 μm 100 μm

20 nm

(b)

300 μm 200 μm 100 μm

(c)

20 nm

330

300 μm 200 μm 100 μm

Figure 12.4 AFM images obtained over an area of 300 μm × 300 μm for ITO films grown on glass substrates prepared at a low processing temperature with (a) argon gas and hydrogen partial pressures of (b) 7.0 × 10−6 Torr and (c) 2.0 × 10−5 Torr.

12.2 Anode Modification for Enhanced OLED Performance

used. The initial growth mode may affect the ultimate properties of thin films. However, the surface morphological property of the ITO films thus prepared is also related to the energetic ion particles and reactive sputter species in the plasma atmosphere created by the rf magnetron sputtering. In such a plasma environment, the growing film is subjected to various forms of bombardment involving ions, neutral atoms, molecules, and electrons [30]. The morphological changes that occurred in ITO films prepared with hydrogen–argon mixtures, as shown in Figure 12.4, were probably due to the presence of the additional reactive hydrogen species in the sputtering atmosphere. As hydrogen was introduced to the sputtering gas mixture during the film preparation, the growing flux during the magnetron sputtering created a significant amount of energetic hydrogen species with energies over the range 10–250 eV [31]. These reactive hydrogen species could remove weakly bound oxygen from the depositing film [13]. The bombardment of the sputtering particles and hydrogen species on the depositing film could cause the reduction of the indium atoms in the ITO film [32]. The energetic hydrogen species present could also react with the growing clusters containing intermediates such as Inx Oy , adsorbed O, reduced indium atoms, and sub-oxides like In2 O. The weakly absorbed oxygen and possible reduced interstitial metal atoms in the ITO film may be removed or re-sputtered by the reactive hydrogen species in the plasma process. Therefore, the nucleation growth kinetics and surface reaction rates on ITO islands formed via the nucleation sites are altered by reactive hydrogen species. The effect of reactive hydrogen species on the depositing ITO film can reduce the size of clusters in comparison with that of the ITO film prepared with only argon gas under the same conditions. As a consequence, the addition of the hydrogen in the sputtering mixture shows a reduced effect on oxide, and ITO films deposited with hydrogen–argon mixtures exhibited a smoother morphology. 12.2.1.3

Electrical Properties

Figure 12.5 shows the dependence of the electrical property of ITO films on the hydrogen partial pressure in the gas mixture. It can be seen that the resistivity of the ITO films changed considerably over the hydrogen partial pressure range 0–2.0 × 10−5 Torr used in the film preparation. The resistivity of the ITO films decreased initially with the hydrogen partial pressure and reached its minimum value of 4.66 × 10−4 Ωcm at an optimal 6.0 Film resistivity (×10–4 Ωcm)

Figure 12.5 Resistivity of ITO films as a function of hydrogen partial pressure.

5.5

5.0

4.5

4.0 0

5

10

15

20

Hydrogen partial pressure (×10–6 Torr)

331

35

N μ

34

4.5

33 4.0

32 31

3.5

30 29

3.0 0

Carrier concentration, N (1020/cm3)

12 Organic Semiconductors and Applications

Hall mobility, μ (cm2/V sec)

332

5 10 15 20 Hydrogen partial pressure (×10–6 Torr)

Figure 12.6 Carrier mobility and concentration of ITO films as functions of hydrogen partial pressure.

hydrogen pressure of about 7.0 × 10−6 Torr. A further increase in the hydrogen partial pressure above the optimal value was shown to increase the film resistivity. The existence of a minimum resistivity was also observed in ITO films prepared at an elevated temperature of 300 ∘ C [13]. Figure 12.5 shows that the relative minimum resistivity of ITO film prepared under the optimal conditions was about 11% lower than that of the film deposited with argon gas under the same conditions. This shows that the usage of the hydrogen–argon mixture had a direct effect in improving the electrical property of ITO films fabricated by the rf magnetron sputtering method at a low processing temperature. The charge carrier mobility and concentration in ITO films were measured by the Hall effect using the van der Pauw technique. The measured carrier mobility, 𝜇, and concentration, N, in ITO films as functions of the hydrogen partial pressure are plotted in Figure 12.6. It can be seen that both 𝜇 and N are very sensitive to the hydrogen partial pressures used in the film preparation. The results in Figure 12.6 show that ITO films prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr, which produced ITO films with the lowest resistivity as shown in Figure 12.5, exhibited the maximum carrier concentration and minimum mobility values. As the electrical conductivity is proportional to the product of 𝜇 and N, this implies that the low resistivity of ITO films prepared at the optimal hydrogen partial pressure was due to the higher carrier concentration in the conduction mechanism. Figure 12.6 shows that ITO films fabricated at a hydrogen partial pressure of 7.0 × 10−6 Torr have a high carrier concentration of 4.59 × 1020 cm−3 . In ITO films, both tin dopants and ionized oxygen vacancy donors provide the charge carriers for conduction. The number of oxygen vacancies that provide a maximum of two electrons per oxygen vacancy plays a dominant role in determining the charge carrier density in an ITO film with high oxygen deficiency. It is also affected by the deposition conditions such as the sputtering power, substrate temperature, Sn/In composition in the target, and sputtering species in the plasma during the film preparation. Banerjee et al. [33] investigated the effect of oxygen partial pressure on conductivity of the ITO films prepared by electron beam evaporation using a hot pressed powder target with a weight ratio of In2 O3 to SnO2 of 9:1. They found that the increased film conductivity was due to an enhancement in the Hall mobility, but the carrier concentration

12.2 Anode Modification for Enhanced OLED Performance

decreased with the oxygen partial pressure. A similar correlation between oxygen partial pressure and carrier concentration in ITO films prepared by pulsed laser deposition was also observed by Kim et al. [34]. Experimental results reveal that the improved electrical properties of ITO films fabricated at the optimal oxygen partial pressure were due to the increased carrier mobility in the film. The decrease in carrier concentration was attributed to the dissipation of oxygen vacancies when oxygen was used in the gas mixture during the preparation. When the ITO films were prepared in the presence of reactive hydrogen species, however, the carrier mobility of the ITO film had a relative low value. This implies that the mechanism of improvement in the conductivity of ITO films is dependent on the deposition process. The enhancement in the conductivity of ITO films prepared at a presence of hydrogen can be attributed to a high carrier concentration in comparison with ITO films made without hydrogen in the gas mixture. The above analyses are consistent with the previous results obtained from ITO films prepared at high substrate temperatures [13]. The above analyses based on the electrical measurements and morphological results suggest that the film quality improvement was due to the presence of reactive energetic hydrogen species in the sputtering plasma when the hydrogen–argon mixture was used. The relative low film resistivity was attributed to the higher oxygen deficiency and possible good contacts between different domains, as the film was denser when it was prepared with hydrogen. Thin films of ITO with high charge carrier concentration may have some advantages for OLED applications. ITO is an n-type wide bandgap semiconductor. The Fermi level, Ef , of ITO films is located at about 0.03 eV below the conduction band minimum [35]. It has an upward surface band bending with respect to its Ef [36] due to a surface Fermi-level-pinning mechanism. Therefore, the effective barrier for hole injection at the interface of ITO HTL should include both the upward surface band bending of ITO and the band offset between the Ef of ITO and the ionization potential of HTL. It has been reported that the surface band bending of ITO decreases with the increase in the carrier concentration in ITO films [37]. Lesser surface band bending lowers the effective energy barrier for carrier injection when it is used as an anode in an OLED [38]. It was demonstrated that the electroluminescence (EL) efficiency of OLEDs fabricated with ITO having a higher carrier concentration was always higher than that of those fabricated with ITO having a lower carrier concentration [39]. The increase in EL efficiency reflects an enhanced hole injection in the device. It can be considered that the ITO anode with a high carrier concentration has a smaller surface band bending, which lowers the effective energy barrier for hole injection in OLEDs. Therefore, ITO films prepared under an optimal hydrogen partial pressure of 7.0 × 10−6 Torr by the rf magnetron sputtering method at a low processing temperature are preferable in practical applications. 12.2.1.4

Optical Properties

In parallel with the morphological and electrical analyses, the transmission spectra of the ITO films deposited at different hydrogen partial pressures were examined over the wavelength range 0.3–2.0 𝜇m. Figure 12.7 shows the wavelength-dependent transmittance, T(𝜆), of the ITO films prepared with argon gas (solid curve) and hydrogen partial pressures of 7.0 × 10−6 Torr (dashed curve) and 2.0 × 10−5 Torr (dotted curve). Except for obvious deviations in the infrared wavelength region, the T (𝜆) of the films prepared at different hydrogen partial pressures also shows a slight difference over the short wavelength range.

333

12 Organic Semiconductors and Applications

100 80

T(λ) (%)

334

60 40

With argon H2 partial pressure 7.0 × 10–6 Torr

20

H2 partial pressure 2.0 × 10–5 Torr

0 500

1000 1500 Wavelength (nm)

2000

Figure 12.7 Wavelength-dependent transmittance, T(𝜆) of ITO films prepared at a low processing temperature with argon (solid curve) and hydrogen partial pressures of 7.0 × 10−6 Torr (dashed curve) and 2.0 × 10−5 Torr (dotted curve).

Figure 12.7 reveals that the short wavelength cutoff in the T(𝜆) of ITO films prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr shifts toward shorter wavelengths in comparison with that in the T(𝜆) of ITO films prepared with argon gas. The shift of the short wavelength cutoff in T(𝜆) is related directly to the variation in the bandgap in the ITO films. To better understand the shift of the short wavelength cutoff in T(𝜆), the wavelength-dependent absorbance, A(𝜆), of ITO films was measured to estimate their optical bandgaps. Using the absorption coefficient, 𝛼, derived from the measured A(𝜆) of the ITO films, the optical bandgap Eg can be estimated. ITO is an ionic-bound degenerated oxide semiconductor. Usually, the following relation is used to derive Eg for heavily doped oxide semiconductors [3, 13]: 𝛼 2 ≈ (h𝜈 − Eg )

(12.1)

where h𝜈 is the photon energy. Figure 12.8 shows the photon energy dependence of 𝛼 2 for ITO films prepared at different hydrogen partial pressures. Extrapolation of the linear region of the plot to 𝛼 2 at zero gives the value of Eg . It can be seen from Figure 12.8 that Eg values for ITO films prepared with argon at hydrogen pressures of 7.0 × 10−6 Torr and 2.0 × 10−5 Torr are 3.75, 3.88, and 3.73 eV, respectively. Figure 12.8 shows that ITO films prepared with argon and a hydrogen partial pressure of 2.0 × 10−6 Torr have similar optical bandgaps, as both films also have similar carrier concentrations as shown in Figure 12.6. A higher Eg value of 3.88 eV is obtained for the film prepared at the optimal hydrogen pressure of 7.0 × 10−6 Torr. The widening of the bandgap can be attributed to the increase in the carrier concentrations in ITO film prepared at the optimal hydrogen pressure. ITO is a degenerate semiconductor. The conduction band is partially filled with electrons. The Fermi level, Ef , of ITO is very close to the conduction band minimum. The Fermi level shifts upward and overlaps or even is located within the conduction band when the charge carrier concentration increases. As Ef moves to a higher energy in the conduction band, the electronic states near the conduction band minimum are fully occupied. Thus, the energy

12.2 Anode Modification for Enhanced OLED Performance

300

With argon

α2 (1012 cm–2)

H2 partial pressure 7.0 × 10–6 Torr 200

H2 partial pressure 2.0 × 10–5 Torr

100

0 3.00

3.25

3.50

3.75

4.00

4.25

4.50

Photon energy (eV)

Figure 12.8 Square of absorption coefficient, 𝛼 2 , plotted as a function of the photon energy for ITO films prepared under different conditions. Extrapolation of the straight region of plot to 𝛼 2 at zero gives the bandgap E g = 3.88 eV for ITO films prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr.

level of the lowest empty states in the conduction band moves to the higher energy positions as well. Therefore, electrons excited from the valence band to those available electronic states in the conduction band require higher energy. This implies that the effective increase in the energy gap of ITO films prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr was due to an increase in the carrier concentration. The bandgap broadening due to an increased carrier concentration in the ITO film is also known as the Moss–Burstein effect, which means that the lowest states in the conduction band are filled by excess charge carriers [3]. In this case, the increase in charge carriers was due to the increase in the oxygen vacancies in the ITO films. This analysis is in a good agreement with the results obtained from electrical measurements shown in Figure 12.6. The transmission spectrum of ITO films prepared at a hydrogen partial pressure of 7.0 × 10−6 Torr (Figure 12.7) shows a considerable decrease in the near-infrared region in comparison with that measured for ITO films prepared with argon. In this region, the free carrier absorption becomes important for the transmittance and reflectance of ITO films. The optical behavior of ITO films in the infrared region can be explained by the Drude theory for free charge carriers [3, 40]. The appreciable reduction in transmission over the infrared range for ITO films prepared at the optimal hydrogen partial pressure was due to an increased carrier concentration. Since the thickness of ITO films was maintained at the same value, the difference in T(𝜆) observed in Figure 12.5 is an indication of the variation in the refractive index of the ITO films. Bender et al. [41] have calculated the refractive index of ITO films prepared by a dc magnetron sputtering method at different oxygen partial pressures. They found that the refractive index of ITO films increases with increasing oxygen flow used in the film preparation. A similar result was also obtained by Wu et al. [25], who showed that the refractive index of ITO films decreases with increasing carrier concentration in ITO films. Generally, the index

335

336

12 Organic Semiconductors and Applications

of refraction of an ITO film can be represented by [41] n2 = 𝜀opt −

4𝜋Ne2 m∗ 𝜔20

(12.2)

where n is the index of refraction, 𝜀opt is the high frequency permittivity, m* is the effective mass of the electron, and 𝜔0 is the frequency of the electromagnetic oscillations. ITO films prepared with hydrogen–argon mixtures in this work had high carrier concentrations in their conduction mechanism. According to Eq. (12.2), a reduction in the refractive index of ITO films would be expected due to an increase in the carrier concentration, N, in films prepared in the presence of hydrogen. As such, the refractive index of an ITO film prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr would have the lowest n value as it has the maximum carrier concentration, as shown in Figure 12.6. This implies that a film becomes denser as its refractive index decreases. On the basis of the information obtained from the above morphological and electrical studies on ITO films, it can be considered that the average density of the film prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr is higher than that of the film sputtered with argon gas. Then, on the basis of the information obtained from AFM measurements, it may be inferred that a denser ITO film will have fewer internal voids in the bulk and fewer irregularities on its surface when it is fabricated under optimal conditions. 12.2.1.5

Compositional Analysis

The chemical binding energies of In3d5/2 and Sn3d5/2 for different ITO films were examined using XPS measurements. The energy positions of In3d5/2 and Sn3d5/2 peaks measured for the films deposited at different hydrogen partial pressures were all constant at 445.2 and 487.2 eV, respectively. There were no evident shoulders observed at the high-binding-energy side of the In3d5/2 peaks, which may relate to the formation of In-OH–like bonds in the ITO films prepared in the presence of the hydrogen [42]. The same binding energy positions and almost identical symmetric XPS peak shapes of In3d5/2 and Sn3d5/2 observed from different ITO films suggest that the chemical states of indium and tin atoms in the films remained in an ITO form. The atomic concentration of ITO films prepared at different hydrogen partial pressures was also estimated. The result shows that the stoichiometry of ITO films prepared at different hydrogen pressures was very similar. The variation in the hydrogen partial pressure used in this work did not seem to affect the chemical structure of ITO films significantly. The bulk composition of ITO films was also examined by SIMS measurements. Figure 12.9 shows a typical depth profile of ITO films prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr. The x-axis in Figure 12.9 shows the sputter depths of films that were calculated using the product of the sputtering time and an average sputtering rate of ∼0.36 nm s−1 . The depth profile steps occurring at the ITO surface and the boundary between the ITO and the glass substrate were due to the influence of the interfacial effects. It is obvious that the profiles of the ITO elements O, In, and Sn had stable counts throughout the entire depth profile region measured by SIMS. The steady distribution of ITO elements shown in Figure 12.9 confirms that the ITO films thus prepared were very uniform.

12.2 Anode Modification for Enhanced OLED Performance

Counts/sec

106

oxygen

105 ITO deposited under optimal conditions 104 Sn

103 102

Si

In

101 0

50

100

150 200 250 Sputter depth (nm)

300

350

Figure 12.9 SIMS depth profile of typical ITO film prepared at an optimal hydrogen partial pressure of 7.0 × 10 −6 Torr.

It is well known that the tin dopants and ionized oxygen vacancy donors govern the charge carrier density in an ITO film. In order to understand the mechanism of the carrier concentration variations in different ITO films, SIMS is also used to measure the relative concentration of oxygen and tin in films prepared at different hydrogen partial pressures. In order to compare the relative oxygen and tin content in different films, the intensities of oxygen and tin ions measured by SIMS are normalized to the corresponding intensities of indium acquired in the same measurements. Figure 12.10 plots the normalized depth profiles of oxygen in ITO films prepared with (a) argon gas and the hydrogen partial pressures of (b) 7.0 × 10−6 Torr and (c) 2.0 × 10−5 Torr. It can be seen that there is a slight difference in the normalized SIMS oxygen counts measured for the films fabricated at different hydrogen partial pressures. Figure 12.10 shows clearly that the ITO film deposited with only argon gas had a relatively higher oxygen content than those measured from films prepared with hydrogen–argon mixtures. This indicates that the addition of hydrogen to the sputtering gas mixture has the effect of removing the weakly bound oxygen atoms from depositing ITO films and increases the number of oxygen vacancies. This result is similar to the effect of hydrogen on ITO films prepared at a higher processing temperature of 350 ∘ C, as reported in a previous study [13]. SIMS was also used to examine the relative concentration of tin element for ITO films prepared at different hydrogen partial pressures. The corresponding normalized tin concentration in films prepared with argon and at hydrogen partial pressures of 7.0 × 10−6 Torr and 2.0 × 10−5 Torr are plotted as curves (a), (b), and (c) in Figure 12.11, respectively. Figure 12.11 reveals that ITO films fabricated with hydrogen–argon mixtures had a slightly higher relative tin content than that those obtained from an ITO film sputtered with argon gas. In particular, the ITO film prepared at the optimal hydrogen partial pressure of 7.0 × 10−6 Torr – curve (b) in Figure 12.11 – had the highest tin content. A tiny deviation in the normalized tin concentration in different ITO films may not change the chemical structure in ITO films considerably, but it impacts the efficient doping level in the material. During the film formation, tin atoms will substitute for indium atoms in the lattice, leaving the tin atoms with one electron more than the requirement for bonding.

337

12 Organic Semiconductors and Applications

5 × 106

Oxygen

Counts/sec

4 × 106

(a)

3 × 106 (b)

(c)

2 × 106 (a) With argon

(b) H2 partial pressure 7.0 × 10–6 Torr

1 × 106

(c) H2 partial pressure 2.0 × 10–5 Torr 0 0

50

100

150

200

250

300

Sputter depth (nm)

Figure 12.10 Comparison of SIMS depth profiles of normalized relative oxygen concentration in ITO films prepared at a low processing temperature with (a) pure argon gas and hydrogen partial pressures of (b) 7.0 × 10−6 Torr and (c) 2.0 × 10−5 Torr.

Sn

2500

(b)

2000 Counts/sec

338

(c)

(a) 1500

(a) With argon

1000

(b) H2 partial pressure 7.0 × 10–6 Torr (c) H2 partial pressure 2.0 × 10–5 Torr

500 0

50

100

150 200 250 Sputter depth (nm)

300

350

Figure 12.11 Comparison of SIMS depth profiles of normalized relative tin concentration in ITO films prepared at a low processing temperature with (a) pure argon gas and hydrogen partial pressures of (b) 7.0 × 10−6 Torr and (c) 2.0 × 10−5 Torr.

This electron will be free in the lattice and act as a charge carrier. Apart from an increase in the number of oxygen vacancies and hence an increase in the number of charge carriers, the substitution of indium by tin atoms also improves the film conductivity. On the basis of the measured carrier concentration results, as given in Figure 12.6, it can be inferred that an enhanced doping level was achieved in an ITO film prepared at the optimal hydrogen pressure of 7.0 × 10−6 Torr, as shown in curve (b) in Figure 12.11. ITO films fabricated by rf or the dc magnetron sputtering method usually require an annealed substrate at temperatures in the range 200–300 ∘ C when both an alloy and oxidized targets are used. A high processing temperature enhances the crystallinity in the ITO film and hence increases the carrier mobility and film conductivity. Low-processing-temperature ITO is a prerequisite for fabricating flexible OLEDs or

12.2 Anode Modification for Enhanced OLED Performance

top-emitting OLEDs that precludes the use of a high-temperature process. An ITO anode with a smooth surface can minimize electrical shorts in the thin functional organic layers in OLEDs that are very often in the range 100–200 nm. Our results demonstrated that the process thus developed offers an enabling approach to fabricate ITO films with smooth morphology and low sheet resistance on a flexible plastic substrate at a low processing temperature. As such, a process that produces smooth ITO films with high optical transparence and high electric conductivity at a low processing temperature will be of practical and technical interest. 12.2.2

Anode Modification

Much progress has been made in OLEDs since the discovery of light emission from electroluminescent polymers. It is appreciated that the physical, chemical, and electrical properties at both anode and cathode interfaces in OLEDs play important roles, though still not well understood, in determining the device operating characteristics and the stability of these devices. Different substrate treatment techniques, which may involve ultrasonic cleaning of the ITO anode in organic solutions, the exposure of the pre-cleaned ITO to either ultraviolet (UV) irradiation or oxygen plasma treatment, are used for device fabrication [43–45]. It has been reported that oxidative treatments, such as oxygen plasma or UV ozone, can effectively increase the work function of ITO [6, 46] and also form a negative surface dipole facing toward the ITO. This may then lead to enhance hole injection and increase device reliability [8, 47, 48]. Ultraviolet photoelectron spectroscopy and Kelvin probes are often employed to investigate the changes in work function or surface dipole of ITO due to different surface treatments [49, 50]. It shows that an increase in the ITO work function is closely related to the increase in its surface oxygen content due to oxidative treatment [6, 51, 52]. The effect of ITO surface treatments on improvement of OLEDs, although widely investigated, is both important and not fully understood. A better understanding of the mechanism of oxygen plasma treatment on an ITO anode in OLED performance is of practical interest. In addition to the appropriate ITO anode cleaning and treatment, it is also shown that charge injection properties in OLEDs can be modified in different ways. At the cathode side, the use of low-work-function metals [7], and the introduction of electron-transporting materials with high electron affinity [53] or a-few-hundred-angstroms-thick conducting polymer layers have been investigated [50, 54–65]. At the same time, the charge injection can also be improved at the anode side by adding an HTL [57] by various treatments of ITO [6, 8, 58, 59]. An ITO anode with self-assembled monolayers can bring significant enhancements in luminous efficiency [50, 60–62]. Modifying the electrode/organic interface using various insulating layers can also substantially enhance the electroluminescence performance of the devices [63–67]. The following discussion will focus on an understanding of anode modification for enhanced carrier injection properties. In situ four-point probe methods in conjunction with XPS and time-of-flight secondary ion spectrometry (TOF-SIMS) measurements were used to explore the relation between a bilayer ITO/insulating interlayer anode and improvement of OLED performance. It is found that oxidative treatment induces a nanometer-thick oxygen-rich layer on the ITO surface. The thickness of this ultrathin oxygen-rich layer deduced from a dual layer model is consistent with the XPS and

339

340

12 Organic Semiconductors and Applications

TOF-SIMS measurements. The enhancement in OLED performance correlates directly with such an interlayer of low conductivity. It serves as an efficient hole-injecting anode. ITO-coated glass with a thickness of 120 nm and a sheet resistance, Rs, of about 20 Ω/square is used in the device fabrication. The ITO used for OLEDs underwent wet cleaning processes including ultrasonication in the organic solvents, followed by oxygen plasma or UV ozone treatment. In this work, wet-cleaned specimens without undergoing any further oxidative treatment are marked as “non-treated” ITO. Wet-cleaned specimens further treated by oxygen plasma under different conditions prior to OLED fabrication are marked as “treated” ITO. The experiments were carried out in a multi-chamber vacuum system equipped with an ITO sputter chamber, an oxygen plasma treatment chamber, and an evaporation chamber. The system is also connected to a glove box purged by high-purity nitrogen gas to keep oxygen and moisture levels below 1 ppm. The system allows the substrate to be transferred among different chambers without breaking the vacuum. The samples can also be moved from an evaporation chamber to the glove box without exposure to air. The four-point probe is placed inside the glove box for an in situ measurement of any changes in ITO sheet resistance, Rs , due to oxygen plasma treatment. A 15-nm-thin ITO film was also deposited on glass for treatment study. This is to boost the effect of oxygen plasma treatment on the variations in Rs for a more accurate measurement. A phenyl-substituted poly (p-phenylenevinylene) (Ph-PPV) [68] was used as an emissive polymer layer. The single-layer testing device has a configuration of ITO (100 nm)/Ph-PPV (80 nm)/Ca (5 nm)/Ag (300 nm). The change in Rs of ITO films due to different oxygen plasma treatments was monitored by the in situ four-point probe measurement. All plasma-treated ITO films were found to have a higher sheet resistance than non-treated films. The increase in sheet resistance observed in a 15-nm-thick ITO was more apparent than that in a 120-nm-thick ITO treated under the same conditions. The process for oxygen plasma treatment was optimized on the basis of the EL performance of OLEDs, and it was chosen accordingly by varying the oxygen flow rates at a constant plasma power of 100 W and a fixed exposure time of 10 minutes. 12.2.3

Electroluminescence Performance of OLEDs

Figure 12.12 shows the current density–voltage (J–V ) characteristics of a set of identical devices built on ITOs treated with wet cleaning, oxygen plasma, and UV ozone. It shows the effect of different treatments on the device performance. For ITOs treated with oxygen plasma and UV ozone, the devices generally have a similar maximum luminance of ∼50 000 cd m−2 . For a wet-cleaned device, the luminance is ∼20 000 cd m−2 . In terms of the current density, the device treated with oxygen plasma and UV ozone treatments has lower values compared to the wet-cleaned ITO. In this case, the oxygen plasma treatment was not optimized. The J–V characteristics measured for a set of identical OLEDs made on ITOs treated by oxygen plasma at different oxygen flow rates of 0, 40, 60, and 100 standard cubic centimeters per minute (sccm) are illustrated in Figure 12.13. There are obvious differences in the EL performance of OLEDs with regard to the anode treatments. For instance, at a given constant current density of 20 mA cm−2 , the luminance and efficiency of identical devices made with oxygen plasma treatments falls within the range 600–1000 cd m−2

12.2 Anode Modification for Enhanced OLED Performance

Current Density (mA/cm2)

1800 UV ozone Wet-cleaned Oxygen plasma

1500 1200 900 600 300 0 0

2

4

6

8 10 Voltage (V)

12

14

16

18

Current Density (mA/cm2)

Figure 12.12 J–V characteristics of identical devices made on wet-cleaned, oxygen plasma, and UV-ozone-treated ITO substrates.

0 nm (non-treated) 0.6 nm (40 sccm) 0.9 nm (60 sccm) 3.1 nm (100 sccm)

600

400

200

0

0

2

4

6

8

10

12

14

16

18

Operating Voltage (V)

Figure 12.13 J–V characteristics of identical devices made on ITO anodes treated under different oxygen plasma conditions.

and 5.0–11.0 cd A−1 , respectively. These values are 560 cd m−2 and 5.0 cd A−1 , respectively, for the same device fabricated on a non-treated ITO anode. The enhancement in hole injection in the device made on a treated ITO anode is clearly demonstrated.The J–V characteristics, shown in Figure 12.13, indicate that there is an optimal oxygen plasma treatment process. The best EL performance was found in the OLED made on an ITO anode treated with an oxygen flow rate of 60 sccm used in this study. It is considered that the enhancement in EL performance of OLEDs due to oxygen plasma treatment is attributed to the improvement in the ITO surface properties. The improvement in device performance is explained by a low barrier height at the ITO/polymer interface and better ITO/polymer adhesion. This includes an increased work function, an improved surface morphology [6, 69], and a less affected bulk ITO property [8]. In order to explore the correlation between the variations in Rs and the possible compositional changes on the treated ITO surfaces, the surface contents of non-treated and treated ITO films were examined using XPS and TOF-SIMS.

341

12 Organic Semiconductors and Applications

For surface compositional analyses, both non-treated and treated ITO samples were covered with a 5-nm-thick lithium-fluoride (LiF) capping layer before the samples were taken out for XPS and TOF-SIMS measurements. This protective LiF layer was to prevent any possible contamination on the ITO surfaces in air and was removed by argon ion sputtering in the XPS and TOF-SIMS measurements. Therefore, the changes in Rs observed by in situ four-point probe measurements and variations in the surface contents obtained by ex situ spectroscopic analyses on ITO surfaces due to oxygen plasma are being compared. In the XPS measurements, the In3d5/2 , Sn3d5/2 , and O1s peaks of non-treated and treated ITO films were examined. The binding energies of these peaks were found to be 445.2, 487.2, and 530.3 eV, respectively, for all ITO surfaces. This implies that no major chemical changes occur on an ITO surface during the oxygen plasma treatment. However, there was a considerable increase in the ratio of O/(In+Sn) for an ITO as the oxygen flow rate was increased. For example, the ratio of O/(In+Sn) obtained from an ITO surface treated by oxygen plasma with a flow rate of 60 sccm was almost 10% higher than the value obtained from a non-treated ITO surface. This implies that the change in Rs corresponds closely with the increase in the oxygen content on an ITO surface. The increase of the oxygen concentration on an ITO surface was found to correlate strongly with the increase in its work function [43, 44]. The comparison of TOF-SIMS depth profiles of the normalized relative oxygen concentration from the surfaces of non-treated ITO and ITO-treated films with a flow rate of 60 sccm is shown in Figure 12.14. It has been reported that oxygen-plasma-treated ITO has a smoother surface than the surface of the non-treated ITO [6]. In this work, the depth profile for 18 O ion was used for analyses because the intensity of 16 O ion counts was so high that it saturated the machine. As shown in Figure 12.14, from the 18 O ion, the relative oxygen concentration on the plasma-treated ITO surface is obviously higher. On the basis of the sputter rate used in the depth profile measurements, it appears that oxygen plasma treatment can induce a few-nanometers-thick oxygen-rich ITO region

18O

TOF-SIMS depth profile (Arb. Units)

342

Non-treated ITO Oxygen-plasma-treated ITO 0

50

100 150 200 Sputtering time (s)

250

300

Figure 12.14 Comparison of TOF-SIMS depth profiles of relative oxygen concentration on the non-treated and oxygen-plasma-treated ITO surfaces.

12.2 Anode Modification for Enhanced OLED Performance

in the vicinity of the ITO surface. However, the precise thickness of this region is difficult to determine directly by XPS or TOF-SIMS due to the influence of the interfacial effects occurred during the argon ion sputtering. ITO is a ternary ionic-bound degenerate semiconducting oxide. The conductivity of oxygen-deficient ITO is governed by tin dopants and ionized oxygen vacancy donors. In an ideal situation, free electrons can result either from the oxygen vacancies acting as doubly charged donors, providing two electrons each, or from the electrically active tin ionized donor on an indium site [11, 70]. The additional oxidation on an ITO surface by oxygen plasma may cause the dissipation of oxygen vacancies. Therefore, such an oxygen plasma treatment results in a decrease in electrically active ionized donors in a region near the ITO surface, leading to an overall increase in Rs as manifested by the in situ four-point probe measurement. By comparing the highly conducting bulk ITO, it can be considered that the treated ITO anode can be portrayed using a dual layer model. In order to gain an insight into the relation between the variation in sheet resistance and an optimal oxygen plasma process for enhanced hole injection, a set of 15-nm-thick ITO films was coated on glass substrates using the same deposition conditions in the sputter chamber. Each of the thin ITO films was transferred to a connected chamber for oxygen plasma treatments and was then passed to the glove box for an in situ sheet resistance measurement. The changes in the sheet resistance, ΔR, between the treated and non-treated ITO were 19, 30, and 89 Ω/square at the oxygen flow rate of 40, 60, and 100 sccm, respectively. The thickness of the oxygen-plasma-induced low-conductivity layer, x, is estimated to be 0.6, 0.9, and 3.1 nm at oxygen flow rates of 40, 60, and 100 sccm, respectively. Table 12.1 summarizes the changes in ΔR and the corresponding estimated thicknesses of the oxygen-plasma-induced low-conductivity layer, x, obtained for ITO films treated under different conditions. The measured ΔR correlates directly with the thickness of this low-conductivity layer, and both increase with the oxygen flow rate. The oxygen plasma treatment that was optimized for the OLED performance, in this case 60 sccm, induces a nanometer-thick resistive layer on an ITO surface. The above analysis based on the XPS, TOF-SIMS, and the electrical results suggest that the presence of an oxygen-plasma-induced nanometer-thick resistive layer on an ITO surface can also account for the enhanced OLED performance. Identical OLEDs made on ITOs, modified with LiF layers of thicknesses in the range 0–5.0 nm, were fabricated. Figure 12.15 shows the J–V characteristics of devices made with different LiF layer thicknesses. A comparison with the devices made without this Table 12.1 Oxygen-plasma-induced low-conductivity layer thickness and ΔR for the ITO films treated under different conditions.

𝚫R (𝛀/sq)

Oxygen-plasma-induced low-conductivity layer thickness, x (nm)

0



0

40

19

0.6

60

30

0.9

100

85

3.1

Oxygen flow rate (sccm)

343

12 Organic Semiconductors and Applications

1000 Current Density (mA/cm2)

344

Bare ITO 0.5 nm LiF 1.5 nm LiF 5.0 nm LiF

800 600 400 200 0 0

4

8

12

16

Voltage (V)

Figure 12.15 J–V characteristics of devices made with different LiF interlayer thicknesses.

layer demonstrates that the former has a higher EL brightness operated at the same current density. At a given constant current density of 20 mA cm−2 , the luminance and efficiency for devices with 1.5-nm-LiF-coated ITO were 1600 cd m−2 and 7 cd A−1 , respectively. The corresponding values were 1170 cd m−2 and 5.7 cd A−1 , respectively, for the same devices made with only an ITO anode. The results demonstrate that the presence of an ultrathin LiF layer between the ITO and polymer favors efficient operation of LEDs. These improvements are attributed to an improved ITO/polymer interface quality and a more balanced carrier injection that improves the device efficiency. One possible mechanism for the above enhancement in device performance can be obtained from tunneling theory. LiF is an excellent insulator with a large bandgap of about 12 eV. However, the presence of an ultrathin insulating interlayer between ITO and polymer enhances hole injections. This indicates that the potential barrier for hole injection that was present in the device with a 1.5-nm-thick LiF interlayer was thus decreased via tunneling. Recently Zhao et al. [71] reported that insertion of an insulation LiF interlayer between the ITO anode and the HTL induces energy level realignment at the anode/HTL interface. There exists a triangle barrier at the ITO/HTL interface, and when a forward bias is applied on an OLED in the presence of an LiF interlayer, a voltage drop across the LiF lowers the ITO work function and hence also reduces the triangle barrier at the ITO/HTL interface. Although LiF also introduces an additional barrier, the overall effective interfacial barrier height can be reduced under an optimal interfacial modification condition, leading to enhanced carrier injection. The possible chemical reactions at the interface should also be taken into account for a more consistent explanation. They reveal that enhanced hole injection with the bilayer ITO/LiF anode altered the internal electric field distribution in the device. The enhanced injection for holes due to the tunneling effect could induce the change in the potential difference across both electrodes for carrier injection, leading to an improvement in the balance of the hole and electron injections. As the thickness of the insulating interlayer increases, the probability of carrier tunneling decreases, leading to a weaker carrier injection process. To achieve a given luminance, therefore, the applied voltage needs to be increased with increasing interlayer

12.3 Flexible OLEDs

thickness. As it becomes thicker, both the current density and EL brightness decreased because of reduced tunneling, but more balanced hole and electron injections were achieved so that the EL efficiency could reach its maximum value. For instance, an increased EL efficiency of 7.9 cd A−1 for a device with a 1.5-nm-thick LiF indicated a more balanced injection of both types of carriers, which was less optimal in the case of a device with a bare ITO, as shown in Figure 12.15. The devices presented in this study clearly portray this behavior. An ITO anode modified with an insulating interlayer for enhanced carrier injection in OLEDs has been demonstrated [67, 71, 72]. Also, attempts are made to explore the use of ultrathin inorganic insulating layers to modify ITO in OLEDs. Ho et al. [61] have reported that a 1–2-nm-thick insulating self-assembled-monolayer on ITO significantly alters the injection behavior and enhances EL efficiency, which is consistent with our results presented here. These improvements are attributed to an improved ITO/polymer interface quality and a more balanced carrier injection that improves device efficiency. Apart from the current understanding of the formation of a negative dipole facing the ITO due to oxygen plasma or UV ozone treatment, it seems that oxidative treatment modifies an ITO surface effectively by reducing the oxygen deficiency to produce a low-conductivity region. As such, in this case, oxygen-plasma-treated ITO behaves somewhat similar to the specimens where there is an ultrathin insulating interlayer serving as an efficient hole injection anode in OLEDs. The improvement in the EL performance of OLEDs is correlated directly to the thickness of the low-conductivity interlayer between the anode and the polymer layer. The results show that the best EL performance comes from the device with a nanometer-thick low-conductivity layer induced by an optimal oxygen plasma treatment. This is consistent with the use of an ultrathin parylene or LiF layer modified bilayer ITO/interlayer anode for enhanced performance of OLEDs [67, 73].

12.3 Flexible OLEDs Current OLED technologies employ rigid substrates, such as glass, which limits the “moldability” of the device, restricting the design and spaces where OLEDs can be used. The demand for more user-friendly displays is propelling efforts to produce head-worn and hand-held devices that are flexible, lighter, more cost-effective, and more environmentally benign than those presently available. Flexible thin film displays enable the production of a wide range of entertainment-related, wireless, wearable-computing, and network-enabled devices. The display of the future requires that it should be thin in physical dimensions, large-sized, flexible, and full color at a low cost. These demands are sorely lacking in today’s display products and in technologies such as the plasma display and LCD technologies. OLEDs [74–77] have recently attracted attention as display devices that can replace LCDs because OLEDs can produce high visibility by self-luminescence. The OLED stands out as a promising technology that can deliver the above challenging requirements. Next-generation flexible displays are going to be commercially competitive due to their low power consumption, high contrast, light weight, and flexibility. The use of thin flexible substrates in OLEDs will significantly reduce the weight of flat panel displays and provide the ability to bend or roll a display into any desired shape. To date, much

345

346

12 Organic Semiconductors and Applications

effort has been focused on fabricating OLEDs on various flexible substrates [78–82]. The plastic substrates usually do not have negligible oxygen and moisture permeability. The barrier properties of these substrates are not sufficient to protect the electroluminescent polymeric or organic layers in OLEDs due to the penetration of the chemically reactive oxygen and water molecules into the active layers of devices. Therefore, the plastic substrates with an effective barrier against oxygen and moisture penetration have to be fabricated before this simple vision of flexible display can become a reality [83]. Polymer-reinforced ultrathin glass sheet is one of the alternative substrates for flexible OLEDs. In this section, we will discuss the results of OLEDs fabricated on flexible ultrathin glass sheet with polymer reinforcement coating and flexible plastic substrates. 12.3.1

Flexible OLEDs on Ultrathin Glass Substrate

A polymer-reinforced ultrathin glass sheet is one of the alternative substrates for flexible OLEDs. It has been found that the flexibility and handling ability of an ultrathin glass sheet can be improved significantly when it is reinforced. The reinforcement polymer layer has the same shrinkage direction as that of an ITO layer deposited on the opposite side of the substrate. ITO-coated ultrathin glass with a reinforcement polymer layer has high optical transmittance and is suitable for device fabrication. The work presented in Section 12.2.1 in this chapter indicates that the ITO film developed at a low processing temperature is suitable for OLED applications. The improvement in OLED performance correlates directly with the ITO properties. Plastic foils and ultrathin glass sheets with reinforced polymer layers are not compatible with a high-temperature plasma process. Therefore, the successful development of high-quality ITO films with a smooth surface, high optical transparency and electric conductivity at a low temperature provides the possibility for developing flexible displays. We have used flexible ultrathin glass sheets with a 250-nm-thick smooth ITO coating and a sheet resistance of about 20 Ω/square to fabricate phenyl alkoxyphenyl PPV copolymer-based OLEDs. In order to compare the EL performance of the OLEDs thus fabricated, we also fabricated an OLED with an identical structure using a commercial ITO-coated rigid glass substrate and used it as a reference. Both OLEDs were fabricated using the same conditions and had an active emitting area of 4 mm2 on glass substrates with dimensions of 5 cm × 5 cm. Figure 12.16 shows the current and luminance of both types of OLEDs as functions of the operating voltage, i.e. J–V and luminance–voltage (L–V ) characteristics. The solid curves in Figures 12.16a,b represent the device characteristics measured from the reference OLED, and the corresponding dashed curves represent those of the OLED made with an ITO-coated flexible thin glass sheet. As can be seen from Figure 12.16, there are some differences in the J–V curves of the two OLEDs at operating voltages higher than 8 V. This can probably be attributed to the less uniform organic films on ITO-coated ultrathin glass sheets. There was a warp in ITO-coated thin glass sheets due to a mismatch in the coefficients of thermal expansion between the ITO film and the ultrathin glass sheet. The polymer films spun on warped substrates over an area of 5 cm × 5 cm might be less uniform in comparison with those coated on the rigid substrate under the same conditions. This problem can be overcome by attaching the flexible glass sheet on a rigid substrate. In general, the forward current

12.3 Flexible OLEDs

Current density (mA/cm2)

(a)

1800 1600

OLEDs on 1.1 mm glass

1400

OLEDs on UTG

1200 1000 800 600 400 200 0 0

2

4

6 8 10 12 Operating voltage (V)

14

16

14

16

(b) 60000 OLEDs on 1.1 mm glass OLEDs on UTG

Luminance (cd/m2)

50000 40000 30000 20000 10000 0 0

2

4

6 8 10 12 Operating voltage (V)

Figure 12.16 (a) J–V and (b) L–V characteristics measured for the OLEDs fabricated with 50 μm flexible ultrathin glass (UTG) sheets and ITO-coated rigid glass substrate.

density measured from both OLEDs, shown in Figure 12.16a, exhibit similar behavior in the operating voltage range 2.5–7.5 V. A maximum luminance of 4.8 × 104 cd m−2 and an efficiency of 5.8 cd A−1 , measured for OLEDs fabricated on a reinforced flexible glass sheet, at an operating voltage of 7.5 V was obtained. The electroluminescent performance of OLEDs made with 50-μm-thick flexible borosilicate glass sheets is comparable to that of identical devices made with commercial ITO coated on rigid glass substrates. It is envisaged that further improvements in device performance can be achieved by optimizing the fabrication processes and encapsulation techniques. 12.3.2

Flexible Top-Emitting OLEDs on Plastic Foils

In the past decade, the display industries have experienced an extremely high growth rate. In the recent market report by Stanford Research, the compounded annual growth rate of OLEDs from 2003 to 2009 is 56%, which means that it will grow from US$300 million in 2003 to US$3.1 billion in 2009. OLEDs have recently attracted attention as display devices that can replace LCDs because OLEDs can produce high visibility by

347

348

12 Organic Semiconductors and Applications

self-luminescence. OLEDs do not require back lighting, which is necessary for LCDs, and can be fabricated into lightweight, thin, and flexible display panels. A typical OLED is constructed by placing a stack of organic electroluminescent and/or phosphorescent materials between a cathode layer that can inject electrons and an anode layer that can inject holes. When a voltage of proper polarity is applied between the cathode and anode, holes injected from the anode and electrons injected from the cathode combine radiatively and emit energy as light, thereby producing electroluminescence. 12.3.2.1

Top-Emitting OLEDs

A conventional OLED has a bottom-emitting structure, which includes a reflective metal or metal alloy cathode, and a transparent anode on a transparent substrate, enabling light to be emitted from the bottom of the structure. An OLED may also have a top-emitting structure, which is formed on either an opaque substrate or a transparent substrate. Unlike the conventional OLED structure, top-emitting OLEDs can be made on both transparent and opaque substrates. One important application of the top device structure is monolithic integration of a top-emitting OLED on a polycrystalline or amorphous silicon thin film transistors (TFTs) used in active-matrix displays, as illustrated in Figure 12.17. The top-emitting OLED structures therefore increase the flexibility of device integration and engineering. Efficient and durable TOLEDs are required in high-resolution display applications. TOLEDs can be incorporated into an active-matrix display integrated with amorphous or poly-silicon TFTs, which have the important advantage of having on-chip data and scan drivers, allowing for ultrahigh pixel resolution ( 20% has been reported for an OLED using MoO3 HIL, which was achieved due to efficient hole injection at the interface between the HIL and a deep HOMO material, e.g. 4,4′ -Bis(carbazol-9-yl)biphenyl (CBP) [102]. Many studies have revealed that the TMO HIL possesses a high work function and a strong molecular electronegativity, allowing the extraction of electrons from the HOMO level of the neighboring organic layer to its deep conduction band [103–106]. In comparison with the solution-processed PEDOT:PSS HIL, the TMO-based HIL is usually prepared by sputtering or the thermal evaporation method. However, in all-solution fabrication process technologies, e.g. roll-to-roll or inkjet printing, the vacuum-involved fabrication process has a limitation in practical applications. Therefore, a lot of effort has been devoted to developing a solution-processed TMO HIL. Different approaches, e.g. the sol–gel method [107–110], direct dissolution of TMO powder in solvent [111, 112], and synthesis of TMO nanoparticles (NPs) solution [113, 114], have been reported to realize the solution-processed TMO HIL. However, the TMO HIL prepared by the sol–gel method requires a high sintering temperature, which is not suitable for devices made on flexible plastic substrates. For example, the MoO3 sol–gel process requires a sintering temperature of 275 ∘ C [109]. The solubility of the TMO powders in the solvent can be a problem, because most TMO powders do not dissolve in organic solvents easily. The synthesis of TMO NPs solution can be an adequate approach for solution-processed TMO HIL. But the films prepared by TMO NPs have a large surface roughness and a high density of pinholes, causing high leakage current [113, 114]. Recently, the use of hybrid HIL in organic optoelectronic devices has been reported, e.g. blending the PEDOT:PSS and TMOs [111, 115, 116]. High-performance organic solar cells with a hybrid HIL have been demonstrated to show promising results as compared to the device using pure TMO HIL [115–117]. In processing, the hybrid HIL solution can improve the wetting ability and the film formation property on the polymer surface [117]. In addition, a blending film with a pinhole-free morphology can improve interfacial contact and thus enhance the device efficiency and stability [115]. The use of solution-processed hybrid HIL has potential for application in large-area OLEDs. Thus, the solution-processed hybrid HIL plays an important role in the performance of OLEDs. This section discusses the performance of OLEDs with a solution-processed hybrid HIL. The hybrid PEDOT:PSS and MoO3 NPs hybrid HIL show a superior hole injection characteristic at the interface between the HIL and the low-lying HOMO material of CBP, leading to superior device efficiency comparable to the performance of OLED using

355

12 Organic Semiconductors and Applications

–2.5

–2.6

TmPyPB TmPyPB

CBP

Ir(ppy)2acac

LiF/AI MoO3 NPs

MoO3-PEDOT:PSS

–4.9

PEDOT:PSS

ITO

–3.0 Energy (eV)

356

–5.24 –5.6

–5.74 –5.98

Ir(ppy)2acac CBP MoO3-PEDOT:PSS ITO Glass

–6.0 –6.7 (a)

(b)

Figure 12.23 The schematic diagrams of (a) the energy levels of all functional materials used in OLEDs, and (b) OLED’s structural configuration.

the vacuum-evaporated MoO3 HIL. The surface property of the hybrid HIL is determined, and the underlying physics is discussed. MoO3 thin film or the blending film of PEDOT:PSS and MoO3 is used as the HIL to modify the ITO in OLEDs. An HTL of CBP and an ETL of TmPyPB are used as the organic stacks. A simplified trilayer device, having a general configuration of anode/interlayer/HTL/ETL/cathode was used in this experiment. A phosphorescent dopant Bis-(2-phenylpyridine)-(acetylacetonate)-iridium(III) (Ir(ppy)2 acac) with a peak of EL emission at 520 nm was doped in the host material of CBP [102]. The schematic diagrams of the energy levels of the functional materials and the structural configuration of the OLED are presented in Figure 12.23. A bare ITO glass substrate, with a sheet resistance of 15 Ω/sq., was cleaned by ultrasonication sequentially with diluted detergent, deionized water, acetone, and isopropanol each for 20 minutes. The MoO3 NPs solution was synthesized according to the reported procedure [118]. The MoO3 NPs solution (0.1 mol ml−1 , dissolved in ethanol) was mixed with the PEDOT:PSS (CleviosTM P VP Al 4083) solution, with different volume ratios of PEDOT:PSS to MoO3 NPs pf 1:1, 2:1, 3:1, and 4:1. A 30-nm-thick hybrid HIL was formed on the ITO/glass substrate. A 10-nm-thick MoO3 HIL on ITO/glass was also formed using pure MoO3 NP solution for comparison studies. After the annealing at 120 ∘ C for 10 minutes, the samples were loaded into a N2 purged glove box, which is connected to a 10-source evaporator for device fabrication. All functional layers, as shown in Figure 12.23, were deposited by thermal evaporation in a high vacuum system with a base pressure of less than 5.0 × 10−4 Pa. OLEDs with an evaporated MoO3 HIL and a solution-processed MoO3 HIL have the same device configuration of ITO (80 nm)/MoO3 (10 nm)/CBP (50 nm)/CBP:Ir(ppy)2 acac (7%, 15 nm)/TmPyPB (55 nm)/LiF(1 nm)/Al(100 nm). For OLEDs with a hybrid PEDOT:PSS-MoO3 HIL, the devices have a configuration of ITO (80 nm)/hybrid HIL with different volume ratios of PEDOT:PSS to MoO3 (1 : 1, 2 : 1, 3 : 1 and 4 : 1, 30 nm)/CBP (30 nm)/CBP:Ir(ppy)2 acac (7%, 15 nm)/TmPyPB (55 nm)/LiF(1 nm)/Al(100 nm). All OLEDs were encapsulated in the glove box and then characterized in ambient conditions. The current density–voltage–luminance (J–V –L) characteristics of OLEDs were measured by a Keithley source measurement unit (Keithley Instruments Inc.,

12.4 Solution-Processable High-Performing OLEDs

Model 236 SMU), which was calibrated using a silicon photodiode. The EL spectra were measured by a spectra colorimeter (Photo Research Inc., Model 650) spectrophotometer. The surface morphology of HILs was characterized by AFM (Multimode V), which was performed in the tapping mode. The surface electronic properties of different HILs were analyzed using XPS, and UPS measurements were performed in the system equipped with an electron spectrometer (Sengyang SKL-12) and an electron energy analyzer (VG CLAM 4 MCD) operated at the base pressures of 2 × 10−9 mbar. The XPS spectra were measured by an achromatic Mg K𝛼 excitation (1253.6 eV) at 10 kV, with an emission current of 15 mA. The UPS spectra were measured by a He discharge lamp (He I radiation of 21.22 eV), and the samples were biased at −5.0 eV during the measurement in order to observe the low-energy secondary cutoff. The performance of a set of structurally identical OLEDs with different HILs was investigated. Figure 12.24a shows the J–V characteristics of OLEDs with different HILs inserted between ITO and CBP. Without the interlayer at the anode/organic interface, the OLED obviously shows an extremely low current density without emission over the entire operating voltage range from 0 to 9 V. This indicates that the hole injection is not ideal due to a high injection barrier at the ITO/CBP interface [119]. An obvious increase in the injection current in the OLED is obtained after the insertion of a MoO3 HIL between ITO and CBP, as shown in Figure 12.24a. This indicates that the MoO3 HIL makes good contact with CBP [102]. OLEDs made in the presence of a MoO3 layer, prepared by either vacuum evaporation or the solution process, have a comparable current density and a low turn-on voltage of ∼3.0 V. At a low driving voltage, the luminance of OLEDs with a solution-processed MoO3 HIL is slightly lower than that of OLEDs with a vacuum-evaporated MoO3 HIL, as shown in Figure 12.24b, due to a high leakage current. The inset in Figure 12.24a depicts the J–V characteristics of both devices in a semi-log plot. OLEDs with a solution-processed MoO3 NPs HIL have a high leakage current, e.g. about two orders of magnitude higher than that of OLEDs with a vacuum-evaporated MoO3 HIL at an operating voltage 3.2 V). In addition, the efficiency of OLEDs fabricated with different MoO3 -based HILs was also studied. Figure 12.25 shows the power efficiency as a function of the luminance for a set of structurally identical OLEDs made with different HILs. The power efficiency of the OLEDs having a hybrid MoO3 -PEDOT:PSS HIL is the highest compared to the efficiency of OLEDs with either an evaporated or a solution-processed MoO3 HIL, e.g. 89.2 lm W−1 @ 100 cd m−2 and 73.5 lm W−1 @ 1000 cd m−2 . The overall power efficiency of the hybrid HIL-contained OLEDs is almost two times higher than that of OLEDs with a solution-processed MoO3 HIL. Figure 12.26 depicts the EQE as a function of the luminance for a set of structurally identical OLEDs with different HILs. A high EQE

357

12 Organic Semiconductors and Applications

(a) 80

102

Current density (mA/cm2)

101 100 60

10–1 10–2

–3 40 10 10–4

0

2

4

6

8

10

20

0 (b) 105

104 Luminance (cd/m2)

358

103

102 Bare ITO Vacuum-evaporated MoO3

101

Solution-processed MoO3 Hybrid MoO3-PEDOT:PSS

100

2

4

6 Voltage (V)

8

10

Figure 12.24 (a) J–V and (b) L–V characteristics of a set of OLEDs fabricated using different HILs. Inset in Figure 12.24a is the semi-log plot of J–V characteristics of OLEDs with different MoO3 anode interlayers, vacuum-evaporated (green triangle), solution-processed (red circle), and hybrid MoO3 -PEDOT:PSS (inverted blue triangle).

for OLEDs with a hybrid HIL, e.g. >20%, was obtained over a wide luminance range from 10 to 10 000 cd m−2 . The EQE of OLEDs with a hybrid MoO3 -PEDOT:PSS HIL is comparable to the best record of OLEDs with a vacuum-evaporated MoO3 HIL [102]. This implies that the solution-processed hybrid MoO3 -PEDOT:PSS HIL can function as well as the vacuum-evaporated MoO3 anode buffer for application in OLEDs. The inset in Figure 12.26 shows the normalized EL spectra of OLEDs measured at 5.0 V. All OLEDs using different HILs have almost identical EL spectra with a peak position at 520 nm. The performance of a set of OLEDs made with different HILs is summarized in Table 12.2. The effect of the volume ratio of PEDOT:PSS to MoO3 NPs in the mixed solution for making a hybrid HIL on the performance of OLEDs was also investigated. The performance of a set of OLEDs made with a hybrid HIL, prepared using a mixed solution

12.4 Solution-Processable High-Performing OLEDs

100

Power Efficiency (lm/W)

80

60

40

Vacuum-evaporated MoO3

20

Solution-processed MoO3 Hybrid MoO3-PEDOT:PSS 0 1 10

102

103 Luminance (cd/m2)

104

105

Figure 12.25 Comparison of power efficiency as a function of the luminance for a set of structurally identical OLEDs having different HILs. 40 Vacuum-evaporated MoO3

35

Solution-processed MoO3 30

Hybrid MoO3-PEDOT:PSS

20 15

1.0 Normalized EL intensity

EQE (%)

25

10 5

0.6 0.4 0.2 0.0

0 100

0.8

300 400 500 600 700 800 Wavelength (nm)

101

102 103 Luminance (cd/m2)

104

105

Figure 12.26 EQE as a function of the luminance for a set of structurally identical OLEDs having different HILs. The inset in Figure 12.26 is the normalized EL spectra of OLEDs measured at 5.0 V.

having different volume ratios of PEDOT:PSS to MoO3 NPs solutions, e.g. 1:1, 2:1, 3:1, and 4:1, was examined. Figure 12.27 shows the J–V characteristics of the set of OLEDs. It is obvious that OLEDs with different hybrid HILs have a higher injection current than that of OLEDs with an evaporated MoO3 HIL. Apart from the hybrid HIL prepared with a mixed solution having a volume ratio of PEDOT:PSS to MoO3 NPs of 1:1, OLEDs with hybrid HILs having other ratios show a similar injection current.

359

12 Organic Semiconductors and Applications

Table 12.2 Summary of performance of OLEDs prepared with different anode interlayers. CE (cd/A) @ 100/1000/10000 cd m−2

PE (lm/W) @ 100/1000/10000 cd m−2

EQE (%) @ 100/1000/10000 cd m−2

Evaporated MoO3

79.5/76.9/62.3

67.2/55.5/31.2

20.9/20.5/16.5

Solutionprocessed MoO3 NPs

44.7/66.5/53.3

37.1/44.8/26.5

12.1/17.9/13.9

Hybrid MoO3 PEDOT:PSS

92.4/90.3/77.6

89.2/73.5/51.0

25.1/24.5/20.6

HIL

The current efficiency, power efficiency, and external quantum efficiency were measured at different luminance values of 100, 1000, and 10 000 cd m−2 , respectively.

Vacumm-evaporated MoO3 Hybrid MoO3-PEDOTPSS (1:1)

80 Current density (mA/cm2)

360

Hybrid MoO3-PEDOTPSS (2:1) Hybrid MoO3-PEDOTPSS (3:1)

60

Hybrid MoO3-PEDOTPSS (4:1) 40

20

0 2

4

6

8

Voltage (V)

Figure 12.27 J–V characteristics as a function of luminance, measured for a set of OLEDs with hybrid MoO3 -PEDOT:PSS HIL, having different volume ratios of PEDOT:PSS to MoO3 , e.g. 1:1, 2:1, 3:1, and 4:1.

Figure 12.28 presents the EQE as a function of luminance for a set of OLEDs with a hybrid HIL, having different volume ratios of PEDOT:PSS to MoO3 NPs in the mixed solution. The performance of OLEDs with an evaporated MoO3 HIL is plotted as a reference. Compare to the control OLED, there is a remarkable enhancement in the EQE of OLEDs with a hybrid MoO3 -PEDOT:PSS HIL. When the volume ratio of PEDOT:PSS in the MoO3 /PEDOT:PSS mixed solution increased from 1:1 to 3:1, the EQE of the resulting hybrid HIL-contained OLEDs increased from 21.8% to 24.5%, measured at 1000 cd m−2 . The optimal volume ratio of PEDOT:PSS to MoO3 is found to be 3:1. For PEDOT:PSS with a higher volume ratio of PEDOT:PSS in the mixed solution, e.g. 4 : 1, a slight reduction in EQE of the resulting OLEDs was observed. The EQE values obtained for OLEDs made with hybrid HILs, prepared using mixed solution having different volume ratios of PEDOT:PSS to MoO3 , in the mixed solution are summarized in Table 12.3.

12.4 Solution-Processable High-Performing OLEDs

25

EQE(%)

20

15 Vacumm-evaporated MoO3 Hybrid MoO3-PEDOTPSS (1:1)

10

Hybrid MoO3-PEDOTPSS (2:1) Hybrid MoO3-PEDOTPSS (3:1)

5

Hybrid MoO3-PEDOTPSS (4:1) 0 100

101

102 103 Luminance (cd/m2)

104

105

Figure 12.28 EQE as a function of luminance, measured for a set of OLEDs with hybrid MoO3 -PEDOT:PSS HILs, having different volume ratios of PEDOT:PSS to MoO3 in the mixed solutions, e.g. 1:1, 2:1, 3:1, and 4:1. Table 12.3 A summary of the EQE of OLEDs with hybrid MoO3 -PEDOT:PSS HIL, fabricated using mixed solution having different volume ratios of PEDOT:PSS to MoO3 NPs, measured at 100, 1000, and 10 000 cd m−2 , respectively. Mixed solution with different volume ratios of PEDOT:PSS to MoO3 NPs

EQE (%) @ 100/1000/10000 cd m−2

1:1

22.0/21.8/19.4

2:1

23.0/23.5/20.1

3:1

25.1/24.5/20.6

4:1

23.5/23.1/20.0

12.4.2

Morphological Properties of the MoO3 -PEDOT:PSS HIL

The surface topography of HILs, e.g. the morphology and the surface roughness, were analyzed by AFM. AFM images measured for the surfaces of the bare ITO, vacuum-evaporated MoO3 , and solution-processed MoO3 are shown in Figures 12.29a–c, respectively. The bare ITO surface has a root mean square (rms) roughness of 2.3 nm. Figure 12.29b,c are the AFM images measured for the surface of MoO3 films deposited on the ITO, prepared by the vacuum-evaporation and solution-processed approaches respectively. Both images illustrate some large grains formed on the surface, showing the grain size in the sub-nanometer range. In fact, it is quite common for metal oxides to aggregate and form large clusters with an average size ranging from a few nanometers to sub-nanometer [120, 121]. However, there is

361

362

12 Organic Semiconductors and Applications

1μm (a)

1μm (b)

1μm (c)

Figure 12.29 AFM images measured for the surfaces of (a) a bare ITO, (b) a vacuum-evaporated MoO3 , and (c) a solution-processed MoO3 . An area of 5.0 μm × 5.0 μm was characterized in the AFM measurements using tapping mode.

1μm (a)

1μm (b)

1μm (c)

Figure 12.30 AFM images measured for (a) a pristine PEDOT:PSS HIL, (b) a hybrid MoO3 -PEDOT:PSS HIL prepared using mixed solution with a volume ratio of PEDOT:PSS to MoO3 NPs of1:1, and (c) a hybrid MoO3 -PEDOT:PSS HIL prepared using mixed solution with a volume ratio of PEDOT:PSS to MoO3 NPs of 2:1.

a difference in the surface roughness of MoO3 films prepared by vacuum evaporation and the solution process. The evaporated MoO3 film has a low rms roughness of 1.75 nm, while the solution-processed MoO3 film has a slightly higher rms roughness of 3.55 nm. Solution-processed MoO3 HIL with a high rms roughness is responsible for a high leakage current, thereby leading to a poor OLED performance. The surface roughness and the morphology of hybrid MoO3 -PEDOT:PSS HILs were also characterized. A pristine PEDOT:PSS film was fabricated as the reference. Figure 12.30a depicts the AFM image measured for PEDOT:PSS film deposited on the ITO. AFM images measured for a vacuum-evaporated film and a solution-processed MoO3 film are shown in Figures. 12.30b,c. It is clear that PEDOT:PSS has a very smooth surface with an rms roughness of 0.99 nm. The AFM images measured for the surfaces of the hybrid MoO3 -PEDOT:PSS layers, prepared using mixed solution with different volume ratios of PEDOT:PSS to MoO3 NPs, are shown in Figure 12.30. All blending films have a surface morphology similar to that of the pristine PEDOT:PSS film. For a hybrid HIL prepared using a mixed solution with a volume

12.4 Solution-Processable High-Performing OLEDs

Table 12.4 Summary of rms surface roughness measured for ITO, evaporated MoO3 , PEDOT:PSS, and hybrid HILs fabricated using mixed solution having different volume ratios of PEDOT:PSS to MoO3 NPs.

Thin films

rms surface roughness (nm)

a) Bare ITO

2.31

b) Evaporated MoO3

1.75

c) Solution-processed MoO3

3.56

d) PEDOT:PSS and MoO3 NPs (1 : 1)

1.20

e) PEDOT:PSS and MoO3 NPs (2 : 1)

1.17

f ) PEDOT:PSS and MoO3 NPs (3 : 1)

0.99

g) PEDOT:PSS and MoO3 NPs (4 : 1)

1.02

h) PEDOT:PSS

0.99

of PEDOT:PSS to MoO3 NPs of 1:1, the surface rms roughness of the hybrid HIL is ∼1.2 nm. For a hybrid HIL prepared with a mixed solution with other ratios, e.g. 2:1, 3:1, and 4:1, the film roughness approaches that of the surface of the pristine PEDOT:PSS film. In comparison to the solution-processed MoO3 thin film, the hybrid MoO3 -PEDOT:PSS layer on the ITO displays a smoother surface. In addition, no large grains are observed. This demonstrates that the presence of PEDOT:PSS in the hybrid HIL avoids the aggregation of MoO3 NPs, which otherwise would induce a rough surface. This observation is consistent with the improvement in performance of OLEDs with a solution-processed hybrid MoO3 -PEDOT:PSS HIL. The corresponding rms values measured for the bare ITO, surfaces of evaporated MoO3 , PEDOT:PSS, and hybrid HILs prepared with a mixed solution having different volume ratios of PEDOT:PSS to MoO3 NPs are summarized in Table 12.4. 12.4.3

Surface Electronic Properties of MoO3 -PEDOT:PSS HIL

In-depth investigation of the surface electronic properties of the solution-processed MoO3 was carried out by photoelectron spectroscopy. In some TMOs, e.g. MoO3 , WO3 , and V2 O5 , the presence of oxygen vacancy defects can facilitate hole injection at the metal oxide/organic interface [122]. The oxygen deficiency can introduce the lower oxidation state of metal cations and reduce the work function in TMO, and thus influence the injection properties in OLEDs [103]. Therefore, it would be interesting to understand the injection properties by characterizing the work function and the defect states in the solution-processed MoO3 as well as in the hybrid MoO3 -PEDOT:PSS HIL used in the OLED. UPS spectra measured for the solution-processed MoO3 thin film, with secondary cutoff and valence band, are shown in Figures 12.31a,b, respectively. The work function evaluated from the spectra is 5.98 eV. Figure 12.32 presents the XPS spectrum of the solution-processed MoO3 thin film. The core levels of Mo 3d5/2 and 3d3/2 characteristic peaks were found at 232.6 and 235.6 eV, respectively, corresponding to the Mo6+ state. The only oxidation state (Mo6+ ) indicates that no oxygen defects are present in the MoO3 thin film.

363

Intensity (a.u.)

Intensity (a.u.)

12 Organic Semiconductors and Applications

30 28 26 24 22 20 Binding energy (eV)

10

9 8 7 6 5 Binding energy (eV)

(a)

(b)

Figure 12.31 UPS spectra measured for the thin film prepared by solution-processed MoO3 NPs, (a) the secondary-electron cutoff and (b) the HOMO edge near the Fermi level. 5 3d5/2

Mo6+

4 3d3/2 Intensity (a.u.)

364

3 2 1 0 240

235 230 Binding energy (eV)

225

Figure 12.32 XPS spectrum of the core levels of Mo 3d measured for the thin film prepared by solution-processed MoO3 NPs. The deconvoluted XPS peaks of Mo6+ 3d5/2 and Mo6+ 3d3/2 are also plotted.

The electronic properties and the stoichiometric composition of the hybrid MoO3 -PEDOT:PSS HIL were also analyzed using XPS and UPS. Figure 12.33 illustrates the UPS spectra measured for hybrid MoO3 -PEDOT:PSS (ratio of 3:1) HIL. For comparison, the UPS spectra of pristine PEDOT:PSS film was also plotted in the same figure. The work function of PEDOT:PSS film was found at 5.24 eV, which agrees well with the results reported in the literature [92, 95]. The UPS spectra of the blending films show a shift in comparison with the spectra of PEDOT:PSS film, at the regions of the secondary-electron cutoff, and the HOMO edge near the Fermi level, as shown

30

24.0

23.5

23.0

PEDOT:PSS MoO3-PEDOT:PSS

28 26 24 Binding energy (eV) (a)

Intensity (a.u.)

Intensity (a.u.)

12.4 Solution-Processable High-Performing OLEDs

9.5

9.0 A

8.5

PEDOT:PSS MoO3-PEDOT:PSS

10

9 8 7 6 5 4 Binding energy (eV) (b)

Figure 12.33 (a) Secondary-electron cutoff, and (b) the HOMO edge in UPS spectra measured for the thin films of PEDOT:PSS and the hybrid MoO3 -PEDOT:PSS HIL (3:1).

in the inset of Figure 12.33. A higher work function was observed for the hybrid HIL, which is 5.74 eV. The stoichiometric compositions of PEDOT:PSS and hybrid PEDOT:PSS-MoO3 NPs HILs were also analyzed by XPS, as shown in Figures 12.34a,b, respectively. For pristine PEDOT:PSS film, the main contribution of S2 s originating from PSS can be observed at 231.7 and 228.3 eV (Figure 12.34a). For the hybrid MoO3 -PEDOT:PSS films, the main peaks of S2s originating from PSS can still be observed. In addition, the characteristic peaks of the Mo3d doublet can be found, as shown in Figure 12.34b,c. The peak deconvolution of Mo3d in the spectrum can be justified by either one oxidation state revealing Mo6+ only as shown in Figure 12.34c or two oxidation states revealing Mo6+ and Mo5+ as shown in Figure 12.34b. This implies that there are two possible oxidation states for the Mo3d core level XPS peaks. The major contribution is from the Mo6+ located at 235.6 eV (3d3/2 ) and 232.6 eV (3d5/2 ). Another minor contribution is from the Mo5+ located at 234.6 eV (3d3/2 ) and 231.7 eV (3d5/2 ). The atomic concentration of Mo5+ to Mo6+ is about 1:3.25. Thus, the atomic ratio of Mo to O is about 1:2.91. This indicates that oxygen deficiency possibly is present in the hybrid film due to the oxygen defect. The work function of the hybrid MoO3 -PEDOT:PSS HIL prepared using a mixed solution having different volume ratios of PEDOT:PSS to MoO3 NPs were also examined using UPS measurements. It was seen that the work function decreased from 6.15 to 5.52 eV when the volume ratio of PEDOT:PSS to MoO3 NPs in the mixed solution was changed from 1:1 to 4:1. At a high volume ratio of MoO3 in the mixed solution, e.g. 1:1, the work function of the hybrid HIL was close to that of the solution-processed MoO3 NPs film. At a low concentration of MoO3 NPs, e.g. 4:1, the work function of the hybrid HIL was close to that of the pristine PEDOT:PSS film. This demonstrates that the

365

12 Organic Semiconductors and Applications

(a)

Intensity (a.u.)

S2s

(b) S2s Intensity (a.u.)

366

240

Mo6+ Mo5+ 3d3/2

235

3d5/2

230 Binding energy (eV)

225

Figure 12.34 (a) S2 s core level XPS spectrum measured for the pristine PEDOT:PSS thin film. S2 s and Mo3d core level XPS spectra measured for the hybrid MoO3 -PEDOT:PSS (3:1) thin film, with the peak deconvolution for (b) two oxidation states of Mo6+ and Mo5+ .

doping of MoO3 NPs in PEDOT:PSS can improve the work function in the hybrid HIL. Table 12.5 summarizes of the work function of the HILs obtained by the UPS measurements. Figure 12.35 presents the S 2s and Mo 3d core level XPS spectra measured for the hybrid films. It can be clearly seen that the characteristic peaks of Mo 3d are contributed by the major components of Mo6+ located at 235.6 eV (3d3/2 ) and 232.6 eV (3d5/2 ), and the minor components of Mo5+ located at 234.6 eV (3d3/2 ) and 231.7 eV (3d5/2 ). This indicates the presence of oxygen deficiency in all blending films. The Mo to O ratio is roughly 1:2.9. Table 12.5 Values of work function of HILs obtained from UPS measurements. Material

Work function (eV)

a) PEDOT:PSS

5.24

b) Hybrid MoO3 -PEDOT:PSS (4 : 1)

5.52

c) Hybrid MoO3 -PEDOT:PSS (3 : 1)

5.74

d) Hybrid MoO3 -PEDOT:PSS (2 : 1)

5.88

e) Hybrid MoO3 -PEDOT:PSS (1 : 1)

6.15

f ) Solution-processed MoO3 thin film

5.98

12.4 Solution-Processable High-Performing OLEDs

240

235 230 Binding energy (eV)

225

S Mo6+ Mo5+

Intensity (a.u.)

Intensity (a.u.)

S Mo6+ Mo5+

240

235 230 Binding energy (eV) (b)

(a)

235

230

225

S Mo6+ Mo5+

Intensity (a.u.)

S Mo6+ Mo5+

Intensity (a.u.) 240

225

240

235

230

Binding energy (eV)

Binding energy (eV)

(c)

(d)

225

Figure 12.35 S2 s and Mo3d core level XPS spectra measured for the hybrid MoO3 -PEDOT:PSS HILs, prepared using mixed solution having different volume ratios of PEDOT:PSS to MoO3 NPs, (a) 1:1, (b) 2:1, (c) 3:1, and (d) 4:1.

The solution-processed MoO3 NPs film has a high work function, but the surface roughness is not ideal as compared to that of the evaporated MoO3 thin film. The high leakage current in the OLEDs, caused by the pinhole in the solution-processed MoO3 HIL, is responsible for the poor performance of the OLED (as shown in Figures 12.24–12.26). The presence of a hybrid MoO3 -PEDOT:PSS HIL promotes good interfacial contact between the ITO and CBP, favoring efficient hole injection and thus enhancing the OLED performance (as shown in Figures 12.24–12.26). The superior hole injection characteristics at the hybrid HIL/CBP interface can be attributed to three aspects: (1) a smoother interfacial contact, (2) a reduced energy barrier, and (3) improved hole injection capability. AFM measurements reveal a smooth and pinhole-free morphology in the hybrid HILs. The introduction of hybrid MoO3 -PEDOT:PSS helps improve the surface smoothness as compared to the pure solution-processed MoO3 NPs. At an optimal volume ratio of PEDOT:PSS to MoO3 NPs (3:1), the rms surface roughness of the hybrid HIL can be reduced to ∼1.0 nm, which is comparable to that of the pristine PEDOT:PSS film. The smooth and pinhole-free HIL is beneficial in forming the subsequent layer on its surface and is therefore a good contact for hole injection.

367

368

12 Organic Semiconductors and Applications

UPS studies have shown that hybrid films possess a high work function. Under optimal conditions, the hybrid MoO3 -PEDOT:PSS HIL, made with a mixed solution having a volume ratio of PEDOT:PSS to MoO3 NPs of 3:1, has a work function of 5.74 eV, which is close to the HOMO level of CBP. In comparison to the PEDOT:PSS pristine film, an increase of 0.5 eV in the work function of the hybrid HIL is observed. The improved work function can reduce the energy barrier at the HIL/CBP interface, as illustrated in Figure 12.17, thus facilitating the hole injection. Lastly, the XPS spectra have shown the presence of oxygen vacancies in hybrid films. Different studies have demonstrated that oxygen defects in TMOs can modify the Fermi level and facilitate charge transfer at the HIL/organic interface [103–105, 123, 124]. The presence of an oxygen deficiency in TMO results in its Fermi level being located close to the conduction band, making TMO an n-type semiconductor. The low-lying conduction band of the hybrid HIL is capable of extracting electrons from the HOMO level of the adjacent CBP layer. This charge transfer process would generate free hole carriers in the CBP layer. Therefore, this defect-assisted charge transfer can improve the hole injection characteristics at the HIL/CBP interface. An effective solution-processed hybrid MoO3 -PEDOT:PSS HIL has been developed for application in phosphorescent OLEDs. A good HIL/CBP interface facilitates an efficient hole injection characteristic and thus enhances device efficiency. The surface morphological and electronic properties of the hybrid HIL were analyzed using AFM, UPS, and XPS measurements. A smooth pinhole-free surface morphology of the hybrid HIL is beneficial in forming a good contact with the adjacent organic layer. The hybrid HIL also possesses a high work function and thus reduces the interfacial energy barrier at the HIL/CBP interface, favoring efficient hole injection. In addition, the oxygen deficiency induced by the MoO3 NPs in the hybrid HIL can enhance the electron extraction capability, thereby facilitating hole injection at the HIL/CBP interface.

12.5 Conclusions The development of organic semiconductors and related devices is still in its growing stage, particularly the design and optimization of their structures and the performance. Organic semiconductors as active components in devices have many advantages, e.g. thinness, light weight, large area, cost-effectiveness, chemical tenability, and mechanical flexibility. Recently, multilayer heterojunction high-performance organic optoelectronic devices such as multiphoton emission devices, organic solar cells, and organic photodetectors with external quantum efficiencies up to 75% across the visible spectrum and bandwidths approaching 450 MHz have been reported. Optical integrated organic bi-functional matrix arrays and bistable optical switches have also been demonstrated. The advantages of organic semiconductors such as low cost due to possible solution processing technologies make organic semiconductor technology attractive for specialized or cost-sensitive applications. In particular, as signage, flexible information display boards, smart cards, organic logic circuits, and other applications based on organic electronics become widespread, the need for integrated, high-performance organic optoelectronic devices performing specialized functions on the organic “chip” will increase accordingly. There is increasing activity in this area, and the prospects for organic semiconductor integrated circuits provide a realistic goal for future applications.

References

References 1 Pope, M. and Swenberg, C.E. (1982). Electronic Processes in Organic Crystals. New 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28

29 30 31 32 33 34 35 36 37

York: Oxford University Press. Chopra, K.L., Major, S., and Pandya, D.K. (1983). Thin Solid Films 102: 1. Hamburg and Granvist, C.G. (1986). J. Appl. Phys. 60: R123. Lee, B.H., Kim, I.G., Cho, S.W., and Lee, S.H. (1997). Thin Solid Films 302: 25. Luff, B.J., Wilkinson, J.S., and Perrone, G. (1997). Appl. Opt. 36: 7066. Kim, J.S., Granström, M., Friend, R.H. et al. (1998). J. Appl. Phys. 84: 6859. Parker, I.D. (1994). J. Appl. Phys. 75: 1656. Wu, C.C., Wu, C.I., Sturm, J.C., and Kahn, A. (1997). Appl. Phys. Lett. 70: 1348. Schröder, B. (1991). Mater. Sci. Eng. A 139: 319. Zhu, F.R., Fuyuki, T., Matsunami, H., and Singh, J. (1995). Sol. Energ. Mat. Sol. C. 39: 1. Zhu, F.R., Huan, C.H.A., Zhang, K.R., and Wee, A.T.S. (1999). Thin Solid Films 359: 244. Salehi, A. (1998). Thin Solid Films 324: 214. Zhang, K.R., Zhu, F.R., Huan, C.H.A., and Wee, A.T.S. (1999). J. Appl. Phys. 86: 974. Zhang, K.R., Zhu, F.R., Huan, C.H.A. et al. (1999). Surf. Interface Anal. 28: 271. Sheu, J.K., Su, Y.K., Chi, G.C. et al. (1999). Appl. Phys. Lett. 72: 3317. Major, S. and Chopra, K.L. (1988). Sol. Energy Mater. 17: 319. Hu, J. and Gordon, R.G. (1992). J. Appl. Phys. 72: 5381. Takahashi, Y., Okada, S., Tahar, R.B.H. et al. (1997). J. Non-Crystalline Solids 218: 129. Nishio, K., Sei, T., and Tsuchiya, T. (1996). J. Mater. Sci. 31: 1761. Kwok, H.S., Sun, X.W., and Kim, D.H. (1998). Thin Solid Films 335: 299. Kim, H., Piqué, A., Horwitz, J.S. et al. (1999). Appl. Phys. Lett. 74: 3444. Ma, J., Li, S.Y., Zhao, J.Q., and Ma, H.L. (1997). Thin Solid Films 307: 200. Ma, J., Zhang, D.H., Li, S.Y. et al. (1998). Jpn. J. Appl. Phys. 37: 5614. Laux, S., Kaiser, N., Zöller, A. et al. (1998). Thin Solid Films 335: 1. Wu, Y., Marée, C.H.M., Haglund, R.F. Jr., et al. (1999). J. Appl. Phys. 86: 991. Wu, W.F. and Chiou, B.S. (1997). Thin Solid Films 298: 221. Davis, L. (1993). Thin Solid Films 236: 1. L. Mao, R.E. Benoit, and J. Proscia (1994). In: S.M. Yalisove, C.V. Thompson, and D.J. Eaglesham (Eds), Mechanism of Thin Film Evaluation, Boston, U.S.A., November 29–December 3, 1994, Materials Research Society Symposium Proceeding 317, 181. Sun, X.W., Huang, H.C., and Kwok, H.S. (1996). Appl. Phys. Lett. 68: 2663. Greene, J.E. (1987). Solid State Tech. 30: 115. Katiyar, M., Yang, Y.H., and Ableson, J.R. (1995). J. Appl. Phys. 77: 6247. Lan, J.H. and Kanicki, J. (1997). Thin Solid Films 304: 127. Banerjee, R., Das, D., Ray, S. et al. (1986). Sol. Energy Mater. 13: 11. Kim, H., Gilmore, C.M., Piqué, A. et al. (1999). J. Appl. Phys. 86: 6451. Fan, J.C.C. and Goodenough, J.B. (1977). J. Appl. Phys. 48: 3524. Arias, A.C., de Lima, J.R., and Hummelgen, I.A. (1998). Adv. Matter. 10: 392. ven den Meerakker, J.E.A.M., Meulenkamp, E.A., and Scholten, M. (1993). J. Appl. Phys. 74: 3282.

369

370

12 Organic Semiconductors and Applications

38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80

Zhou, X., He, J., Liao, L.S. et al. (1999). Appl. Phys. Lett. 74: 609. Zhu, F.R., Zhang, K., Guenther, E., and Chua, S.J. (2000). Thin Solid Films 363: 314. Meng, L.J. and dos Santos, M.P. (1998). Thin Solid Films 322: 56. Bender, M., Seelig, W., Daube, C. et al. (1998). Thin Solid Films 326: 72. Shigesato, Y., Hayashi, Y., Masui, A., and Haranou, T. (1991). Jpn. J. Appl. Phys. 30: 814. Van Slyke, S.A., Chen, C.H., and Tang, C.W. (1996). Appl. Phys. Lett. 69: 2160. Kim, J.S., Friend, R.H., and Cacialli, F. (1999). Appl. Phys. Lett. 74: 3084. Kurosaka, Y., Tada, N., Ohmori, Y., and Yoshino, K. (1998). Jpn. J. Appl. Phys. 37: L872. Sugiyama, K., Ishii, H., Ouchi, Y., and Seki, K. (2000). J. Appl. Phys. 87: 295. Milliron, D.J., Hill, I.G., Shen, C. et al. (2000). J. Appl. Phys. 87: 572. Steuber, F., Staudigel, J., Stössel, M. et al. (1999). Appl. Phys. Lett. 74: 3558. Park, Y., Choong, V., Gao, Y. et al. (1997). Appl. Phys. Lett. 68: 2699. Campbell, I.H., Kress, J.D., Martin, R.L. et al. (1997). Appl. Phys. Lett. 71: 3528. Mason, M.G., Hung, L.S., Tang, C.W. et al. (1999). J. Appl. Phys. 86: 1688. Chaney, J.A. and Pehrsson, P.E. (2001). Appl. Surf. Sci. 180: 214. Lee, H.M., Choi, K.H., Hwang, D.H. et al. (1998). Appl. Phys. Lett. 72: 2382. Yang, Y. and Heeger, A.J. (1994). Appl. Phys. Lett. 64: 1245. Karg, S., Scott, J.C., Salem, J.R., and Angelopoulos, M. (1996). Synth. Met. 80: 111. Carter, J.C., Grizzi, I., Heeks, S.K. et al. (1997). Appl. Phys. Lett. 71: 34. Greenham, N.C., Moratti, S.C., Bradley, D.D.C. et al. (1993). Nature 365: 628. Nüesch, F., Rothberg, L.J., Forsythe, E.W. et al. (1999). Appl. Phys. Lett. 74: 880. Kim, J.S., Cacialli, F., Cola, A. et al. (1999). Appl. Phys. Lett. 75: 17. Appleyard, S.F.J. and Willis, M.R. (1998). Opt. Mater. 9: 120. Ho, P.K.H., Granström, M., Friend, R.H., and Greenham, N.C. (1998). Adv. Mater. 10: 769. Malinsky, J.E., Jabbour, G.E., Shaheen, S.E. et al. (1999). Adv. Mater. 11: 227. Jabbour, G.E., Kippelen, B., Armstrong, N.R., and Peyghambarian, N. (1998). Appl. Phys. Lett. 73: 1185. Jabbour, G.E., Kawabe, Y., Shaheen, S.E. et al. (1997). Appl. Phys. Lett. 71: 1762. Hung, L.S., Tang, C.W., and Mason, M.G. (1997). Appl. Phys. Lett. 70: 152. Tang, H., Li, F., and Shinar, J. (1997). Appl. Phys. Lett. 71: 2560. Zhu, F.R., Low, B.L., Zhang, K., and Chua, S.J. (2001). Appl. Phys. Lett. 79: 1205. Becker, H., Spreitzer, H., Kreuder, W. et al. (2000). Adv. Mater. 12: 42. Kim, J.S., Cacialli, F., Granström, M. et al. (1999). Synth. Met. 101: 111. Tahar, R.B.H., Ban, T., Ohya, Y., and Takahashi, Y. (1998). J. Appl. Phys. 83: 2631. Zhao, J.M., Zhang, S.T., Wang, X.J. et al. (2004). Appl. Phys. Lett. 84: 2913. Ding, X.M., Hung, L.M., Cheng, L.F. et al. (2000). Appl. Phys. Lett. 76: 2704. Chua, S.J., Ke, L., Kumar, R.S., and Zhang, K.R. (2002). Appl. Phys. Lett. 81: 1119. Hung, L.S. and Chen, C.H. (2002). Mater. Sci. Eng. R 39: 143. Friend, R.H., Gymer, R.W., Holmes, A.B. et al. (1999). Nature 397: 121. Tang, C.W. and Van Slyke, S.A. (1987). Appl. Phys. Lett. 51: 913. Forrest, S.R. (2003). Org. Electron. 4: 45. Fou, C., Onitsuka, O., Ferreira, M. et al. (1996). J. Appl. Phys. 79: 7501. Krasnov, N. (2002). Appl. Phys. Lett. 80: 3853. Gustafsson, G., Treacy, G.M., Cao, Y. et al. (1993). Synth. Met. 57: 4123.

References

81 82 83 84 85 86 87 88 89 90 91

92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119

Gu, G., Burrows, P.E., Venkatesh, S., and Forrest, S.R. (1997). Opt. Lett. 22: 172. Paetzold, R., Heuster, K., Henseler, D. et al. (2003). Appl. Phys. Lett. 82: 3342. Chwang, A.B., Rothman, M.R., Mao, S.Y. et al. (2003). Appl. Phys. Lett. 83: 413. Li, Y.Q., Tang, J.X., Xie, Z.Y. et al. (2004). Chem. Phys. Lett. 386: 128. Zhu, F.R., Hao, X.T., Ong, K.S. et al. (2005). Towards novel flexible displays: design and fabrication of OLEDs on plastic substrates. Proc. IEEE 93: 1440. Ong, K.S., Hu, J.Q., Shrestha, R. et al. (2005). Thin Solid Films 477: 32. Plichta, A., Weber, A., and Habeck, A. (2003). Mater. Res. Soc. Symp. Proc. 769: H9.1. Tummala, R.R. and Rymaszewski, E.J. (1989). Microelectronics Packaging Handbook. New York: Van Nostrand Reinhold, Chapter 10. Zhang, K.R., Zhu, F.R., Huan, C.H.A., and Wee, A.T.S. (2000). Thin Solid Films 376: 255. Reineke, S., Lindner, F., Schwartz, G. et al. (2009). Nature 459: 234. Kanno, H., Sun, Y., and Forrest, S.R. (2005). High-efficiency top-emissive white-light-emitting organic electrophosphorescent devices. Appl. Phys. Lett. 86: 263502. Lee, J., Chopra, N., Eom, S.H. et al. (2008). Appl. Phys. Letts 93: 123306. Wang, H., Klubek, K.P., and Tang, C.W. (2008). Appl. Phys. Lett. 93: 093306. Giebeler, C., Antoniadis, H., Bradley, D.D.C., and Shirota, Y. (1999). J. Appl. Phys. 85: 608. Tse, S.C., Tsang, S.W., and So, S.K. (2006). J. Appl. Phys. 100 : 063708. Harkema, S., Mennema, S., Barink, M. et al. (2009). Proc. of SPIE 7415 : 74150T-1. Kim, W.H., Makinen, A.J., Nikolov, N. et al. (2002). Appl. Phys. Lett. 80: 3844. Matsushima, T., Kinoshita, Y., and Murata, H. (2007). Appl. Phys. Lett. 91: 253504. You, H., Dai, Y., Zhang, Z., and Ma, D. (2007). J. Appl. Phys. 101 : 026105. Kanno, H., Holmes, R.J., Sun, Y. et al. (2006). Adv. Mater. 18: 339. Chu, C.W., Li, S.H., Chen, C.W. et al. (2005). Appl. Phys. Lett. 87 : 193508. Wang, Z.B., Helander, M.G., Qiu, J. et al. (2011). Appl. Phys. Letts 98 : 073310. Greiner, M.T., Chai, L., Helander, M.G. et al. (2012). Adv. Funct. Mater. 22: 4557. Meyer, J., Zilberberg, K., Riedl, T., and Kahn, A. (2011). J. Appl. Phys. 110 : 033710. Krjöger, M., Hamwi, S., Meyer, J. et al. (2009). Appl. Phys. Lett. 95 : 123301. Meyer, J., Krjöger, M., Hamwi, S. et al. (2010). Appl. Phys. Lett. 96 : 193302. Tan, Z., Li, L., Cui, C. et al. (2012). J. Phys. Chem. C 116 : 18626. Lee, H., Kwon, Y., and Lee, C. (2012). J. Soc. Inf. Disp. 20: 640. Girotto, C., Voroshazi, E., Cheyns, D. et al. (2011). ACS Appl. Mater. Interfaces 3: 3244. Yang, T., Wang, M., Cao, Y. et al. (2012). Adv. Energy Mater. 2: 523. Kwon, W., Kim, Y.N., Lee, H.K. et al. (2014). Org. Electron. 15: 1083. Xu, M.F., Cui, L.S., Zhu, X.Z. et al. (2013). Org. Electron. 14: 657. Meyer, J., Khalandovsky, R., Görrn, P., and Kahn, A. (2010). Adv. Mater. 23: 70. Stubhan, T., Ameri, T., Salinas, M. et al. (2011). Appl. Phys. Lett. 98 : 253308. Kim, J.H., Kim, H.J., Kim, G.J. et al. (2014). ACS Appl. Mater. Interfaces 6: 951. Lee, S.J., kim, B.S., Kim, J.Y. et al. (2015). Org. Electron. 19: 140. Wang, Y.L., Luo, Q., Wu, N. et al. (2015). ACS Appl. Mater. Interfaces 7: 7170. Xie, F.X., Choy, W.C.H., Wang, C.D. et al. (2013). Adv. Mater. 25: 2051. Wang, Z.B., Helander, M.G., Qiu, J. et al. (2010). J. Appl. Phys. 108 : 024510.

371

372

12 Organic Semiconductors and Applications

120 121 122 123 124

Castleman, A.W. and Bowen, K.H. (1996). J. Phys. Chem. 100 : 12911. Tang, X., Bumueller, D., Lim, A. et al. (2014). J. Phys. Chem. C 118 : 29278. Ganduglia-Pirovano, M.V., Hofmann, A., and Sauer, J. (2007). Surf. Sci. Rep. 62: 219. Deb, S. and Chopoorian, J. (1966). J. Appl. Phys. 37: 4818. Sian, T.S. and Reddy, G. (2004). Sol. Energy Mater. Sol. Cells 82: 372.

373

13 Transparent White OLEDs Choi Wing Hong and Furong Zhu Department of Physics, Hong Kong Baptist University, Kowloon Tong, Hong Kong, China

CHAPTER MENU Introduction—Progress in Transparent WOLEDs, 373 Performance of WOLEDs, 374 Emission Behavior of Transparent WOLEDs, 386 Conclusions, 400 References, 400

13.1 Introduction—Progress in Transparent WOLEDs White organic light-emitting diodes (WOLEDs) have been extensively studied for applications in flat panel displays and solid-state lighting. Compared to other solid-state lighting technologies, the WOLED is the only device that can provide the diffused light source that is more comfortable to our eyes, which evolved through the white light from solar radiation. To date, much effort has been focused on developing high-performing phosphorescent WOLEDs for applications in lighting [1–5]. Phosphorescent WOLEDs with a power efficiency >40 lm W−1 have been demonstrated by one-sided opaque devices [6–8]. An efficient WOLED with a power efficiency of 55.2 lm W−1 at 100 cd m−2 , the International Commission on Illumination (CIE) coordinates of (0.40, 0.40), and a color rendering index of 81 using a blue iridium carbene complex has been reported [9]. In parallel with the development of new emissive materials, the use of the light out-coupling structure is a promising approach to improving the power efficiency of WOLEDs. A phosphorescent WOLED with a threefold increase in power efficiency to ∼100 lm W−1 at 100 cd m−2 has been demonstrated using a high-refractive-index substrate and a periodic out-coupling structure [7]. Conventional WOLEDs have a bottom-emission configuration, with an opaque metal or metal alloy cathode, and a transparent anode on a transparent substrate, enabling light to be emitted from the bottom of the structure, usually from the transparent anode side. WOLEDs may also have a top-emission configuration, with an opaque cathode/transparent anode pair and can be formed on either an opaque or transparent substrate. A WOLED can also be made transparent when both a transparent anode

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

374

13 Transparent White OLEDs

and cathode are used. Light can be emitted from both anode and cathode sides when both electrodes are relatively transparent, forming transparent WOLEDs [10]. Transparent WOLEDs have the process flexibility for device integration and engineering, opening up a plethora of opportunities and the potential for novel product concepts. They can be integrated easily with any transparent substrate, for example, window panes and glass walls, to improve their illumination functionality, thereby facilitating new features for applications in advanced displays and lighting. Transparent WOLEDs consisting of a transparent indium tin oxide (ITO) anode and a variety of transparent thin metal cathodes of Al, Ag, Au, Ca/Ag, Yb:Ag, Mg:Ag, etc. have been reported [11–20]. When an ultrathin metal cathode is used, the top cathode can be relatively transparent. This arrangement allows the emitted light to escape from both the ITO anode and the top ultrathin metal cathode sides of the WOLEDs. These transparent WOLEDs usually have a preferential one-sided electroluminescence (EL) emission due to the asymmetric emission characteristics at the anode/organic and organic/cathode interfaces. In addition, angular-dependent EL emission is often observed, due to the microcavity effect [21, 22], leading to a change in the emission color quality at different viewing angles. In this chapter, we discuss the development of high-performing transparent phosphorescent WOLEDs using a high-mobility electron transporting layer (ETL) along with a pair of Ag (10 nm)/MoO3 (2.5 nm)-modified ITO anode and a thin Al (1.5 nm)/Ag (15 nm)/index-matching-layer cathode. The hole-electron current balance in a dichromatic white transparent WOLED has been optimized through a combination of theoretical calculation and experimental optimization. A transparent WOLED with a visible-light transparency of >50%, a symmetrical bidirectional EL emission spectra with almost an identical power efficiency of 11 lm W−1 (measured at 100 cd m−2 ) and similar CIE coordinates from both sides has been demonstrated.

13.2 Performance of WOLEDs 13.2.1

Optimization of Dichromatic WOLEDs

An optimized dichromatic WOLED can be readily used for the development of the transparent WOLED. The optimization processes of a dichromatic WOLED are discussed here, including both the theoretical calculation and experimental characterization. A blue phosphorescent emitter bis(3,5-difluoro-2-(2-pyridyl)phenyl(2-carboxypyridyl)-iridium(III) (FIrpic) and an orange phosphorescent emitter bis(2-(9,9-diethyl-fluoren-2-yl)-1-phenyl-1H-benzo-imidazolato)-(actylacetonate)-iridium(III) (Ir(fbi)2 acac) were used in the WOLED for generating white emission. The dopants have photoluminescence (PL) emission peaks at 478 and 562 nm, as illustrated in Figure 13.1. The white emission from the WOLED is expected to cover the spectrum in the range 460 to 610 nm. These two dopants were co-doped in 4,4′ ,4′′ -tris(carbazol-9-yl)-triphenylamine (TCTA) host to form a dual emissive layer with different FIrpic (x%) and Ir(fbi)2 acac) (y%) concentrations. The dopant concentrations are optimized for simultaneous realization of high power efficiency and color quality. The N,N’-bis-(naphthalen-1-yl)-N,N’-bis(phenyl)-benzidine (NPB) hole transporting layer (HTL) and 2,2’,2"- (1,3,5-benzinetriyl)- tris(1-phenyl-1-H-benzimidazole) (TPBi) ETL were used for the carrier transport. To facilitate the carrier injections, the

13.2 Performance of WOLEDs

Figure 13.1 PL spectra measured for thin film emitters of Firpic and Ir(fbi)2 acac. Normalized PL (arb.)

FIrpic Ir(fbi)2acac

400

500 600 Wavelength (nm)

700

MoO3 hole injection layer (HIL) and LiF electron injection layer (EIL) were used to modify the anode and cathode contacts to achieve efficient charge injection. Dichromatic WOLEDs have a layer configuration of MoO3 (10 nm)/NPB (40 nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (x%, y%, 17.5 nm)/TPBi (40 nm)/LiF (1 nm)/Al (100 nm). Figure 13.2 shows the device architecture and the energy diagram of the functional materials used in WOLEDs. The thickness of each layer of functional materials used in the WOLEDs was optimized using a combination of theoretical calculation and experimental optimization to achieve the highest power efficiency. The schematic cross-sectional view of the dichromatic WOLED and the corresponding energy level diagram are also shown in Figure 13.2, where it is expected that the main recombination zone is located at the TCTA/TPBi interface. Since TCTA is an ambipolar material that has a relatively stronger hole transporting capability (𝜇h ∼10−5 cm2 /Vs) [23], holes are expected to accumulate at the TCTA/TPBi interface, due to the lowest highest occupied molecular orbital (HOMO) level (−6.2 eV) of TPBi. The injected electrons travel through the TPBi layer and then drop preferably into the dopant that has a higher lowest unoccupied molecular orbital (LUMO) level rather than the TCTA. As a result, excitons are formed in dopants at the TCTA/TPBi interface. As FIrpic and Ir(fbi)2 acac dopants have a high quantum yield, they are used for high-efficiency –2.3

NPB MoO3 ITO Glass (a)

LiF/Al

–2.6 –2.7

–5.1

–5.4 –6.0

–2.7

TPBi

TCTA

NPB

ITO/MoO3

TCTA

Energy (eV)

Ir(fbi)2 acac

FIrpic

FIrpic

–2.9

TPBi

Ir(fbi)2acac

–2.4

LiF/Al

–5.7eV –6.2 (b)

Figure 13.2 (a) Schematic device architecture and (b) the schematic energy diagram of the functional materials used in the dichromatic WOLED.

375

Power efficiency (lm/W)

13 Transparent White OLEDs (1%,1%) (4%,1%) (9%,1%) (14%,1%) (19%,1%)

10

1

@ 100 cd/m2 16.0 lm/W 17.0 lm/W 17.5 lm/W 18.6 lm/W 20.5 lm/W

10

@1000 cd/m2 10.5 lm/W 13.1 lm/W 14.2 lm/W 14.7 lm/W 15.3 lm/W

100

1000

Figure 13.3 Power efficiency–luminance characteristics of a set of WOLEDs with FIrpic and Ir(fbi)2 acac doped in the TCTA host, with different doping concentrations of (1%, 1%), (4%, 1%), (9%, 1%), (14%,1%), and (19%, 1%). The inset in Figure 13.3 illustrates the values of the power efficiency of the WOLEDs measured at different luminance values of 100 cd m−2 and 1000 cd m−2 .

10000

Luminance (cd/m2)

organic light-emitting diodes (OLEDs) [24, 25]. The concentration of phosphorescent dopants in OLEDs is optimized to prevent the concentration quenching so as to achieve the desired emission properties [26, 27]. The doping concentration of FIrpic in the dichromatic WOLEDs was varied from 1% to 20%; while the concentration of Ir(fbi)2 acac was maintained at 1%. The power–luminance characteristics and the EL spectra, measured for a set of OLEDs having different FIrpic and Ir(fbi)2 acac dopant concentrations in TCTA, are plotted in Figures 13.3 and 13.4, respectively. It was demonstrated that the best power efficiency was at (19%, 1%), i.e. 20.5 lm W−1 at 100 cd m−2 and 15.3 lm W−1 at 1000 cd m−2 (Figure 13.3). However, blue emission of the EL spectra measured for this set of WOLEDs is less satisfactory, with a relatively stronger emission in red as shown in Figure 13.4. The best EL spectrum was observed for the WOLED with a doping concentration of (14%, 1%), but with a slightly lower power efficiency than that with the doping concentration of (19%, 1%). The power efficiency of the WOLED drops significantly at higher FIrpic dopant concentration of >20%. The most appropriate FIrpic concentration is over the range 14% to 19%, which gives rise to a high-efficiency and good spectral distribution in EL spectra. In addition, the Ir(fbi)2 acac concentration At 1000 cd/m2 Normalized EL intensity (a.u.)

376

400

500

(1%,1%) (4%,1%) (9%,1%) (14%,1%) (19%,1%)

600 Wavelength (nm)

CIE: (0.46,0.47) CIE: (0.45,0.47) CIE: (0.44,0.47) CIE: (0.43,0.46) CIE: (0.44,0.46)

700

Figure 13.4 EL spectra of a set of OLEDs with different combinations of FIrpic and Ir(fbi)2 acac concentrations, e.g. (1%, 1%), (4%, 1%), (9%, 1%), (14%,1%), and (19%, 1%) in TCTA host, measured at 1000 cd m−2 .

13.2 Performance of WOLEDs

was also optimized. OLEDs with an Ir(fbi)2 acac concentration above 1% showed poor efficiency and strong red EL spectra that is consistent with what is reported in the literature [27, 28]. Traps in TCTA can be induced at a high Ir(fbi)2 acac concentration, as the dopant has a higher HOMO level of −5.1 eV as compared to that of TCTA (−5.7 eV). It affects the hole transport capability of TCTA and thus reduces the hole current. At a low doping concentration, e.g. 100 cd m−2 ). The improvement in the hole injection current alone does not guarantee an enhancement in power efficiency, due to the unbalanced hole-electron current. The accumulation of the excess holes at the TCTA/TPBi interface is not favorable for efficient operation of such WOLEDs. In addition, the EL spectra measured for both types of WOLEDs are identical, having a CIE of (0.44, 0.46) as shown in Figure 13.7. This demonstrates that the excess holes in these WOLEDs, due to the presence of a non-optimized MoO3 HIL, do not contribute to the emission caused by the unbalanced hole-electron current. An unbalanced hole-electron current in the WOLEDs was observed, resulting in a decrease in the device efficiency. In fact, the power efficiency can be enhanced by suppressing hole injection at the anode side; however this is not a suitable approach to achieve the hole-electron current balance. The ideal method would be to improve the electron current. In addition to the unbalanced hole-electron current, the energy loss due to triplet exciton quenching is another factor that decreases device efficiency. In fact, TPBi is not only a low-mobility material but also is a low-triplet-energy-level material comparable to the FIrpic dopant [30, 33] as shown in Figure 13.8.

13.2 Performance of WOLEDs Triplet energy (T1)

TCTA 2.82

2

FIrpic

Ir(fbi)2acac

2.65

3

TPBi 2.7

(eV)

(eV)

3

Triplet energy (T1)

2.2

1

TCTA 2.82

2

TmPyPB

FIrpic

2.8

Ir(fbi)2acac

2.65

2.2

1

0

0

(a)

(b)

Figure 13.8 Schematic energy diagram illustrating the triplet energy level of the functional materials used in this work, and the energy transfer path of between the triplet excitons (a) in the FIrpic molecules and TPBi ETL, and (b) in the FIrpic molecules and TmPyPB ETL in WOLEDs are depicted.

MoO3 ITO Glass

TPBi

TCTA

NPB

–2.6

–2.7

–2.7

–5.1

–5.4

–6.0eV

–2.7

LiF/Al

NPB

ITO/MoO3

TCTA

Energy (eV)

Ir(fbi) 2 acac

FIrpic

FIrpic

–2.9

TPBi/TmPyPB

TmPyPB

–2.3

Ir(fbi)2acac

–2.4

LiF/Al

–5.7

–6.2 –6.7 (a)

(b)

Figure 13.9 (a) Schematic cross-sectional view of the dichromatic WOLEDs, and (b) the corresponding triplet energy levels of the materials used in the devices.

The triplet-triplet energy transfer from FIrpic to TPBi can occur as shown in Figure 13.8a, and such a transfer will quench luminescence. To resolve these issues, replacement of TPBi with another electron transporting material that possesses higher electron mobility and high triplet energy is desirable to effectively confine the excitons in the FIrpic molecules. We have selected 1,3,5-tri[(3-pyridyl)-phen-3-yl]benzene (TmPyPB) ETL in the WOLED shown in Figure 13.9a, which has a relatively higher electron mobility (𝜇e ∼10−4 cm2 Vs−1 ) and a high triplet energy level (T 1 = 2.8 eV) [34]. The cross-sectional view of this dichromatic WOLED is presented in Figure 13.9a, and the corresponding schematic diagram of the energy levels of the functional materials is shown in Figure 13.9b. Figure 13.10 shows the plot of the power efficiency as a function of luminance of WOLEDs made with two different ETLs: TPBi and TmPyPB. TmPyPB-based WOLEDs show a significant improvement in power efficiency, a yield that represents an almost factor-of-two increase in the maximum power efficiency as compared to that of the WOLEDs with a TPBi ETL (see Figure 13.10). The power efficiencies of the WOLEDs,

379

13 Transparent White OLEDs

Power efficiency (lm/W)

100 with a TmPyPB ETL with a TPBi ETL

10

Maximum 44.1 lm/W 22.0 lm/W

1

1

10

400

Figure 13.10 Power efficiency–luminance characteristics measured for the dichromatic WOLEDs with different ETLs of TmPyPB and TPBi. The power efficiency values obtained for the WOLEDs operated at 100 cd m−2 and 1000 cd m−2 are also listed for comparison.

@ 100 cd/m2 @ 1000 cd/m2 38.2 lm/W 29.1 lm/W 20.0 lm/W 20.2 lm/W

100

1000

Luminance

Normalized EL intensity (Arb. unit)

380

10000 100000

(cd/m2) with a TmPyPB ETL with a TPBi ETL

Figure 13.11 EL spectra measured for the dichromatic WOLEDs with different ETLs of TmPyPB and TPBi, obtained at 1000 cd m−2 .

CIE (0.44,0.46) CIE (0.39,0.45)

500

600

700

Wavelength (nm)

measured at 100 and 1000 cd m−2 , are listed in the inset in Figure 13.10. An obvious improvement in the blue emission is due to the enhanced electron current in the WOLED with a TmPyPB ETL. In addition to the improvement in the hole-electron current balance, the quenching between the triplet excitons in FIrpic to TmPyPB is also suppressed to reduce the energy loss in nonradiative decay processes. EL spectra measured for WOLEDs with different ETLs of TmPyPB and TPBi at 1000 cd m−2 are given in Figure 13.11. Obviously, the EL intensity of the blue component (478 nm) emitted by FIrpic in the WOLED with a TmPyPB ETL is almost two times higher than that of the OLED with a TPBi ETL. This demonstrates that TmPyPB is a promising ETL material for minimizing energy quenching by confining excitons in the FIrpic molecules. In this system, the WOLED incorporating the TmPyPB ETL and TCTA host, both possessing a high triplet energy level (∼2.8 eV), can effectively reduce exciton quenching via the short-range Dexter energy transfer process [7, 33]. The excitons can be well confined in the emission region as illustrated in Figure 13.8b. The device efficiency of the WOLED has been improved by the optimization of the hole-electron current balance. However, the EL spectra of the WOLED are less satisfactory due to a lack of the blue component of the EL emission from FIrpic. The intensity of the blue emission at 478 nm is ∼50% of that of the orange emission at 562 nm, as shown in Figure 13.11. The concentrations of FIrpic (19%) and Ir(fbi)2 acac (1%) in the TCTA host were optimized for achieving both high luminous efficiency and color quality. A further

13.2 Performance of WOLEDs Triplet energy (T1)

TCTA 2.82

TmPyPB FIrpic 2.65

2.2

2

3

2.8

Ir(fbi)2acac

(eV)

(eV)

3

Triplet energy (T1)

TCTA 2.82

TmPyPB

FIrpic

Ir(fbi)2acac

2.65

2.8

2.2

2 1

1

0

0

(a)

(b)

Figure 13.12 Schematic energy levels of the materials and energy transfer path between the triplet excitons in the FIrpic molecules and neighboring materials in the dichromatic WOLEDs.

increase in the concentration of FIrpic was found to reduce device efficiency due to triplet-triplet annihilation (TTA). The incorporation of TmPyPB was able to confine the excitons in the emitters effectively. The excess red emission is due to the energy roll-down from FIrpic to Ir(fbi)2 acac, as shown in Figure 13.12a. In the WOLED structure, FIrpic and Ir(fbi)2 acac in the emissive layer (EML) were co-doped and arranged in the sequence TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 17.5 nm). When the triplet excitons were formed in FIrpic, they preferred to roll down to a lower triplet energy state in Ir(fbi)2 acac through the Dexter energy transfer process (as shown in Figure 13.12a). The triplet-triplet exciton energy transfer takes place due to the short-range electron exchange between molecules, typically in a range of several nanometers (∼5 nm) [35]. If FIrpic and Ir(fbi)2 acac are separated by a larger distance, the energy transfer from FIrpic to Ir(fbi)2 acac can be prevented or avoided. The main recombination zone in the WOLED is located near the TCTA/TmPyPB interface. The emission from FIrpic can be enhanced by increasing the distance (y nm) between the Ir(fbi)2 acac doping region and the TCTA/TmPyPB interface, as shown in Figure 13.13. The EML was

–2.7

LiF/Al

–2.9

–2.7

–2.7

LiF/Al

–2.3

TmPyPB

–2.4

TCTA NPB MoO3 ITO Glass

FIrpic

Ir(fbi)2acac

TCTA

NPB

y nm

ITO/MoO3

FIrpic Ir(fbi)2acac

Energy (eV)

TPBi/TmPyPB

–5.4 –6.0

–5.7

y nm

–6.7

(a)

(b)

Figure 13.13 (a) Device structure and (b) the schematic energy diagram of WOLED.

381

13 Transparent White OLEDs

then fabricated with the structure of TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 17.5-y nm)/TCTA:FIrpic (19%, y nm). This design approach helps control the energy transfer from FIrpic to Ir(fbi)2 acac in order to produce an improved white emission, as shown in Figure 13.12b. EL spectra measured for WOLEDs having different separation distances (y nm) between the Ir(fbi)2 acac doped region in the EML and the TCTA/TmPyPB interface are presented in Figures 13.14a–f. As expected, increasing the distance (y nm) from the main recombination zone can enhance the blue emission from the WOLED. For the WOLED without separation (y = 0 nm), the intensity of the blue emission (at 478 nm) was found to be ∼0.55 of the orange emission at 562 nm. With increasing y, the intensity of blue emission increases, and the intensity of orange emission decreases. With the balanced blue and orange emissions from the WOLED at y = 3 nm, the best white EL having CIE coordinates of (0.34, 0.41) for the dichromatic WOLED was achieved.

(b)

y = 1 nm

(c)

y = 2 nm (d)

y = 3 nm

(e)

y = 4 nm (f)

y = 5 nm

y = 0 nm

(a)

cd/m2

100 1000 cd/m2

Normalized EL intensity (Arb. unit)

382

400

500

600

700 400

500

600

700

Wavelength (nm)

Figure 13.14 EL spectra emission of TmPyPB-based WOLEDs at different separation distances (y nm) between the Ir(fbi)2 acac emission zone in EML and the TCTA/TmPyPB interface, measured at 100 cd m−2 and 1000 cd m−2 .

13.2 Performance of WOLEDs

Upon a further increase in the thickness (i.e. y > 3 nm), the red emission was highly reduced, implying that the Ir(fbi)2 acac-doped orange region moved away from the main recombination zone. Therefore, the recombination zone can be estimated to be located approximately at 1–2 nm apart from the TCTA/TmPyPB interface. At y = 8 nm, there was only a blue emission from FIrpic measured in the WOLED, although orange emission was still observed in the device. The orange emission was suppressed by limiting the Dexter energy transfer between FIrpic and Ir(fbi)2 acac. But this does not decrease the EL efficiency because the confinement of triplet excitons in FIrpic helps enhance the intensity of blue emission. Therefore, no significant reduction in efficiency was observed, as shown in Figure 13.15. A higher power efficiency of 36.6 lm W−1 is obtained at a luminance of 100 cd m−2 from the device. A good white emission WOLED with a high EL efficiency was demonstrated by the control of triplet-triplet energy transfer. The emission characteristics of a multilayered WOLED are also sensitive to the thicknesses of the functional layers, due to the complex optical interference effect [36–39]. Thus, a high-performing WOLED with an optimal configuration is desirable for an improved white EL spectrum. We have analyzed and optimized the emission color and the out-coupling efficiency of WOLEDs using a combination of theoretical calculation and experimental optimizations. The emission behavior of WOLEDs with a layer structure of ITO (90 nm)/MoO3 (5 nm)/NPB (x nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 14.5 nm)/TCTA:FIrpic (19%, 3 nm)/TmPyPB (x nm)/LiF (1 nm)/Al (80 nm) was analyzed using optical admittance analysis [40, 41]. The wavelength-dependent refractive index n(𝜆) and extinction coefficients k(𝜆) of each functional layer were measured by ellipsometry measurements. In the calculation, the location of the emission dipole was assigned at the interface of TCTA/TmPyPB. The thicknesses of the HTL (NPB) and ETL (TmPyPB) were varied respectively over the thickness range 0 to 80 nm while keeping the combined total layer thickness of the HTL and ETL at 80 nm. The optical effect of the HTL/ETL layer thickness on the CIE color coordinates of the WOLED was analyzed using a theoretical simulation. The calculated results are given in Figure 13.16. The corresponding changes in the CIE coordinates (from point A to point B) are depicted in Figure 13.16a. The emission color has a red shift (from point A to B) when the thickness of the HTL/ETL layer increases. The out-coupled radiance, as functions of the HTL/ETL layer thickness, is shown in

100 Power efficiency (lm/W)

Figure 13.15 Power efficiency–luminance characteristics of TmPyPB-based WOLEDs having different separations (y nm) between the Ir(fbi)2 acac emission zone and TCTA/TmPyPB interface.

10 y = 0 nm, y = 1 nm, y = 3 nm, y = 5 nm,

1

1

10

Maximum 44.1 lm/W 45.7 lm/W 44.0 lm/W 43.5 lm/W

100

@ 100 cd/m2 38.2 lm/W 38.0 lm/W 36.6 lm/W 35.8 lm/W

1000

Luminance (cd/m2)

@1000 cd/m2 29.1 lm/W 28.6 lm/W 28.0 lm/W 26.7 lm/W

10,000 100,000

383

384

13 Transparent White OLEDs

(a)

(b)

Figure 13.16 (a) Effect of HTL/ETL thickness, varied from 0 to 80 nm from point A to point B, on CIE coordinates of a WOLED, and (b) plot of radiance as the function of the HTL and ETL layer thicknesses. The vertical bar indicates the magnitude of radiance (arb. unit).

Figure 13.16b. On the basis of the calculation, a pair of 40-nm-thick NPB and TmPyPB was selected by considering an optimal combination of high power efficiency and preferable CIE color coordinates, as indicated by the point X in Figure 13.16. The stack of NPB (40 nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 10 nm)/TCTA:FIrpic (19%, 7.5 nm)/TmPyPB (40 nm) thus optimized was then used for fabrication of transparent WOLEDs. 13.2.3

Electron-Hole Current Balance in Transparent WOLEDs

Following the optimized structure of a high-performance dichromatic opaque WOLED, e.g. with a layer configuration of ITO/organic stacks/Al, the optimal organic light-emitting stack of MoO3 (5 nm)/NPB (40 nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 10 nm)/TCTA:FIrpic (19%, 7.5 nm)/ TmPyPB (40 nm)/LiF (1 nm) was then adopted for fabrication of transparent WOLEDs. When an ultrathin metal cathode is used, the top cathode can be relatively transparent, forming a transparent WOLED. In order to illustrate the point, a transparent cathode consisting of LiF (1 nm)/ Al (1.5 nm)/ Ag (15 nm)/ NPB (50 nm) was used. A dielectric capping layer with an appropriate refractive index was added on top of the transparent cathode to improve the transparency and light output from both sides of the transparent WOLEDs [42–45]. A thin NPB layer on the upper transparent electrode of LiF (1.0 nm)/Al (1.5 nm)/Ag (15 nm) acts as the effective optical index matching layer for enhancing light out-coupling in transparent WOLEDs. A thin NPB layer, prepared by thermal evaporation, has a refractive index of ∼1.7, serving as an optical index matching layer to enhance light emission from the top cathode and also to improve the overall transparency of the WOLEDs. The effect of the NPB index matching layer on the transparency and the efficiency of transparent WOLEDs was analyzed and optimized through theoretical calculation and experimental optimization.

13.2 Performance of WOLEDs

100 Transmittance (%)

Figure 13.17 Calculated transparency of the transparent WOLEDs, with a configuration of bare ITO/a stack of organic layers/LiF (1 nm)/Al (1.5 nm)/Ag (15 nm)/NPB (x nm), as a function of the NPB capping layer thickness over the range 0 to 84 nm.

0 nm 16 nm 34 nm 50 nm 66 nm 84 nm

80 60 40 20 400

500

600

700

Wavelength (nm)

Figure 13.17 presents the calculated transparency of the transparent WOLEDs as a function of the thickness of the NPB layer on the upper transparent cathode. The use of the NPB layer on the upper transparent cathode improves the transparency of the WOLED up to 70% in the visible-light wavelength range. A 50-nm-thick optimal NPB index matching layer that improves the visible-light transparency of the transparent WOLEDs was then selected. As discussed above, transparent WOLEDs usually have a preferential one-sided EL emission due to the asymmetric emission characteristics at the ITO anode/organic and organic/upper cathode interfaces [16, 46, 47]. Figure 13.18a shows the EL spectra of a transparent WOLED measured from the top side and the bottom side at a current density of 10 mA cm−2 . Asymmetric emission in the transparent WOLED is observed; e.g. the device exhibits a strong EL intensity from the ITO anode side and a relatively weaker EL intensity from the transparent cathode side. The luminance measured at the bottom side (ITO anode) is almost two times higher than that from the top side (transparent cathode), as shown in the inset of Figure 13.18b.

@ 1748 cd m–2

EL intensity (a.u.) 400

Bottom emission CIE: (0.32, 0.42)

Top emission CIE: (0.30, 0.43) @ 806 cd

500 600 Wavelength (nm) (a)

m–2

700

Ratio of top to bottom luminance

1.0 Bottom Top

Luminance ratio (top to bottom)

0.8 0.6 0.4 0.2 0.0

10-1

100 101 Current density (mA cm–2) (b)

102

Figure 13.18 (a) EL spectra of a transparent WOLED measured from the bottom and top sides operated at a current density of 10 mA cm−2 . (b) Ratio of luminance measured from the top side to that obtained from the bottom side over the current density range 10−2 to 102 mA cm−2 .

385

386

13 Transparent White OLEDs

13.3 Emission Behavior of Transparent WOLEDs 13.3.1

Visible-Light Transparency of WOLEDs

This section discusses the emission behavior of the dual-sided emissive transparent WOLEDs, e.g. improving the luminous efficiency, achieving color stability, and moderating the angular-dependent EL emission. The design includes consideration of the competing device parameters of color stability at different angles, EL efficiency and transparency, illustrating an approach to fabricating high-performing transparent WOLEDs in terms of achieving weak angular-dependent emission and high transparency. The concept of the avoidance of a spectral overlap between the wavelengths of the EL emission peaks and that of the intrinsic resonant mode, and its impact on the emission behavior of the transparent WOLEDs are examined experimentally and theoretically. A control transparent WOLED, with a layer configuration of ITO (80 nm)/MoO3 (5 nm)/NPB (40 nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 14.5 nm)/TCTA: FIrpic (19%, 3 nm)/TmPyPB (40 nm)/LiF (1 nm)/Al (1.5 nm)/NPB (50 nm), serves as a control device for comparison. The emission behavior of a dual-sided transparent WOLED, with a structure of ITO (80 nm)/Ag (10 nm)/MoO3 (5 nm)/NPB (40 nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19, 1%, 14.5 nm)/TCTA:FIrpic (19%, 3 nm)/TmPyPB (40 nm)/LiF (1 nm)/Al (1.5 nm)/Ag (15 nm)/NPB (50 nm), was analyzed, as shown in Figure 13.19. In order to improve the performance of transparent

NPB NPB LiF/Al/Ag LiF/Al/Ag TmPyPB TmPyPB FIrpic Ir(fbi)2acac

FIrpic Ir(fbi)2acac TCTA

TCTA NPB NPB MoO3 ITO Glass

(a)

MoO3 Ag ITO Glass

(b)

Figure 13.19 (a) Schematic cross-sectional view of a control transparent WOLED having a bare ITO anode, and (b) a structurally identical transparent WOLED with a Ag-modified transparent anode.

13.3 Emission Behavior of Transparent WOLEDs

WOLEDs, a set of structurally identical WOLEDs made with an ultrathin Ag-modified ITO was fabricated. The device has a layer configuration of ITO (80 nm)/Ag (x nm)/ MoO3 (5 nm)/ a stack of organic layers/ LiF (1 nm)/ Al (1.5 nm)/ Ag (15 nm)/ NPB (50 nm), as shown in Figure 13.19b. In parallel with the experiment, the emission behavior of the transparent WOLEDs was also analyzed using admittance analysis. The cavity length of the transparent WOLEDs, including a stack of NPB (x nm)/ TCTA (5 nm)/ TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 14.5 nm)/ TCTA:FIrpic (19%, 3 nm)/ TmPyPB (x nm), was evaluated by optimizing the thickness combination of NPB and TmPyPB for achieving the desired optical and EL properties. An index matching layer with an appropriate refractive index is essential to improve the transparency and light output from both sides of transparent WOLEDs [42–45]. In fact, the transparency is sensitive to both cavity length and the thickness of the index matching layer. The cavity length of the transparent WOLED includes a stack of MoO3 (2.5 nm)/NPB (x nm)/TCTA (5 nm)/TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 14.5 nm)/TCTA:FIrpic (19%, 3 nm)/TmPyPB (x nm). An NPB index matching layer on the transparent cathode of LiF (1 nm)/Al (1.5 nm)/Ag (15 nm) was used, as shown in Figure 13.19. The effect of the NPB thickness on emission behavior and the overall transparency was analyzed. Figure 13.20 depicts the calculated transparency of the transparent WOLEDs as a function of the thickness of the NPB optical index matching layer, for devices with different cavity lengths of (a) 60, (b) 100, and (c) 120 nm. A combination of a 50-nm-thick Figure 13.20 Calculated transparency of the transparent WOLEDs, with a configuration of bare ITO/a stack of organic layers of different thicknesses, (a) 60 nm, (b) 100 nm, and (c) 120 nm/LiF (1 nm)/Al (1.5 nm)/Ag (15 nm)/NPB (x nm), as a function of the NPB capping layer thickness over the range 0 to 84 nm.

100 0 nm 16 nm 34 nm 50 nm 66 nm 84 nm

(a) 60 nm 80 60 40 20

Transmittance (%)

0 (b) 100 nm 80 60 40 20 0 (c) 120 nm 80 60 40 20 0 400

500

600

Wavelength(nm)

700

387

13 Transparent White OLEDs

NPB with an organic cavity length of 100 nm was found to be optimal for the transparent WOLED, leading to a broadband transparency in the visible wavelength region. The measured visible-light transparency of the transparent WOLED, with a configuration of glass/ITO (80 nm)/MoO3 (2.5 nm)/NPB (40 nm)/EML/TmPyPB(40 nm)/LiF (1.0 nm)/Al(1.5 nm)/Ag (15 nm)/NPB (50 nm), is compared with the calculated data. The measured (solid symbols) and calculated (open symbols) visible-light transparency of the same device are plotted in Figure 13.21 as a function of the wavelength. The experimental results agree well with the simulated ones and show that the transparent WOLED has an average transmission of >50% in the visible-light wavelength range. The insets in Figure 13.21 are the photo pictures taken for the transparent WOLED emitting light (left) and illustrating the transparent feature without power (right). In addition to the device transparency, the emission characteristics of transparent WOLEDs are critical for lightning applications. Transparent WOLEDs usually have a preferential one-sided EL emission due to the asymmetric emission characteristics at the ITO anode/organic and organic/upper cathode interfaces [16, 46, 47]. Asymmetric emission in transparent WOLEDs is observed; e.g. devices have a strong EL emission from the ITO anode side and a relatively weaker emission from the transparent cathode side. The emission behavior of transparent WOLEDs can be tuned by adjusting the optical properties of both transparent electrodes and the device structure. Figure 13.22 shows the ratio of the luminance of the EL emission from the anode side to that from the cathode side of different transparent WOLEDs, made with different combinations

Transmittance (%)

0.8

Calculated Measured

0.6 0.4 0.2

400

500

600

Figure 13.21 Comparison of measured (solid symbols) and calculated (open symbols) transparency of the transparent WOLED of glass/ITO (80 nm)/Ag (10 nm)/MoO3 (2.5 nm)/NPB (40 nm)/EML/TmPyPB (40 nm)/LiF(1.0 nm)/Al(1.5 nm)/Ag (15 nm)/NPB (50 nm). The insets are the photo pictures taken for the device emitting light (left) and illustrating the transparent feature without power (right).

700

Wavelength (nm)

Figure 13.22 Ratio of the luminance of EL emission measured from the anode to that obtained from the cathode side of the transparent WOLEDs as a function of the operating current density.

1.4 Ratio of top to bottom luminance

388

1.2 1.0 0.8 0.6 0.4 0.2 0.0 10–3

ITO (control) ITO/Ag (7.5 nm) ITO/Ag (10.0 nm) ITO/Ag (12.5 nm)

10–2

10–1

100

101 2

Current density (mA/cm )

102

13.3 Emission Behavior of Transparent WOLEDs

of top and bottom transparent electrodes, as a function of the current density, ranging from 10−2 to 102 mA cm−2 . In order to evaluate the performance of the transparent WOLEDs, a set of structurally identical thin Ag-modified ITO anodes were evaluated for the transparent WOLEDs, with a device configuration of ITO (80 nm)/Ag (x nm)/ a stack of organic layers/ Al (1.5 nm)/ Ag (15 nm)/ NPB (50 nm), as shown in Figure 13.19b. The thickness of the Ag modification layer was varied over the range 0 to 12.5 nm. For a control transparent WOLED, made with a bare ITO anode, the ratio of the luminance of the EL emission from the cathode side to that from the ITO anode side is about 0.5, indicating that the luminance from the transparent cathode side is about 50% to that from the anode side. Depending on the application, the emission from both sides of the transparent WOLEDs can be tuned by adjusting the optical properties of the transparent electrode. In this work, the improvement in the emissions from both sides of the transparent WOLEDs is obtained by tuning the thickness of the Ag modification layer on the ITO anode. The results of this work demonstrate that symmetrical emission from both sides of the transparent WOLEDs can be realized when a pair of ITO (80 nm)/Ag (10 nm) anode and Al (1.5 nm)/Ag (15 nm)/ NPB (50 nm) cathode is used, as shown in Figure 13.22. Transparent WOLEDs with balanced luminance from both sides can be achieved by modifying the ITO surface with a thin Ag layer. However, the Ag-modified ITO may reduce the transparency of the transparent WOLEDs. Figure 13.23 shows the visible-light transparency measured for a set of transparent WOLEDs made with different front transparent anodes, bare ITO and ITO modified with different Ag thicknesses of 7.5, 10, and 12.5 nm. The transparency of the ITO/glass substrate over the same wavelength range is also plotted for comparison. It is seen that the interposing of an ultrathin Ag interlayer over the thickness range from 7.5 to 12.5 nm does not induce a significant change in the visible-light transparency of the transparent WOLEDs. For transparent WOLEDs with an ITO/Ag (10 nm) anode, there is a slight reduction in the peak transparency from 76% to 70% compared to the control transparent WOLED made with a bare ITO anode, with a similar transparency over the visible-light wavelength region. It can be seen that there is a noticeable red shift in the position of the transmission peak as the thickness of the Ag interlayer increases.

100

Transmittance (%)

Figure 13.23 Visible-light transparency of different transparent WOLEDs fabricated with a bare ITO, and the ITO modified with different Ag interlayer thicknesses of 7.5, 10, and 12.5 nm; the transparency of an ITO/glass is also plotted for comparison.

80 60 40 20 0 400

ITO/glass ITO (control) ITO/Ag (7.5 nm) ITO/Ag (10.0 nm) ITO/Ag (12.5 nm)

500 600 Wavelength (nm)

700

389

13 Transparent White OLEDs

13.3.2

L-J Characteristics of Transparent WOLEDs

In addition to the luminance and the transmittance, the color stability of the transparent WOLEDs is another important factor for application in lighting. The transparent WOLEDs can be considered as a weak microcavity consisting of a stack of organic functional layers sandwiched between two transparent mirrors of an ultrathin Ag-modified ITO anode and a LiF (1 nm)/Al (1.5 nm)/Ag(15 nm) cathode. The resonant mode of the weak microcavity has a relative broader visible transmission behavior as shown in Figure 13.23. The pair of transparent mirrors forms a weak microcavity, which affects the color stability of the EL emissions from both sides of the devices due to the microcavity effect. The resonant mode is directly relevant to the optical path in the microcavity. It is well known that the resonant mode or the resonant wavelength of a microcavity can be described by the Fabry–Perot condition [48, 49]: 4𝜋 (13.1) n (𝜆)dorg cos 𝜃 − 𝜑anode (𝜆, 𝜃) − 𝜑cathode (𝜆, 𝜃) = 2m𝜋, 𝜆 org where 𝜆 is the emission wavelength, 𝜑cathode (𝜆, 𝜃) and 𝜑anode (𝜆, 𝜃) are the wavelengthand angle-dependent phase changes due to the reflection at the organic/anode, organic/cathode, and organic/cathode/interfaces (𝜃 = 0, for the normal incidence). m is the mode number (m = 0 was used in the calculation), and norg (𝜆) and dorg denote the refractive index and thickness of the organic layers in the organic cavity. For the dichromatic white emission system studied in this work, the corresponding resonant wavelength of the transparent WOLEDs changes from 466 to 675 nm as the thickness of the organic stack varies from 60 to 140 nm. Figure 13.24 illustrates the emission spectra of the transparent WOLEDs as a function of the thickness of the organic stack (60, 100, and 140 nm). In this work, a 100-nm-thick organic stack was selected for making transparent WOLEDs, with a corresponding resonant wavelength of 550 nm. The use of a 100-nm-thick organic stack avoids the spectral overlap between the peak position of the organic resonant (550 nm) and the dichromatic white EL emission peaks located at 478 and 562 nm, resulting in a good white emission with stable CIEs of (0.36, 0.43) (anode side) and (0.38, 0.46) (cathode side). Symmetrical and bidirectional emission characteristics, for example, the EL spectra and the power efficiency, can then be realized by optimizing the optical out-coupling characteristics at the anode and cathode side of the transparent WOLEDs. In order Normalized EL intensity (a.u.)

390

400

From anode side

From cathode side Cavity length 60 nm 100 nm 140 nm

500

600

Wavelength (nm)

700

400

500

600

700

Wavelength (nm)

Figure 13.24 Emission spectra calculated for transparent WOLEDs with different cavity lengths of 60, 100, and 140 nm.

13.3 Emission Behavior of Transparent WOLEDs

to illustrate the point, a set of transparent WOLEDs fabricated with the same cathode of LiF (1.0 nm)/Al(1.5 nm)/Ag(15 nm)/NPB (50 nm) but different anode structures of (a) ITO/MoO3 , (b) ITO/Ag (7.5 nm)/MoO3 , (c) ITO/Ag (10.0 nm)/MoO3 , and (d) ITO/Ag (12.5 nm)/MoO3 was fabricated. A transparent WOLED made with an anode of ITO/MoO3 served as a control device for comparison studies. The corresponding EL characteristics of different transparent WOLEDs, measured at a current density of 10 mA cm−2 , are plotted in Figure 13.25. It can be seen clearly that the control transparent WOLED has an obvious preferential emission, with an almost facto-of-two difference in the intensity of EL emission from the anode (1748 cd m−2 ) than the cathode (806 cd m−2 ), due to obvious asymmetrical optical properties at anode and the cathode sides. The EL intensities and the CIE coordinates of the bidirectional illumination in a transparent WOLED can be varied by adjusting the optical out-coupling characteristics from both sides. The experimental results agree with the theoretical calculation in showing that a pair of an ITO (80 nm)/Ag (10 nm)/MoO3 (2.5 nm) anode and a LiF (1.0 nm)/Al (1.5 nm)/Ag (15 nm)/NPB (50 nm) is a suitable transparent electrode choice for symmetrical and bidirectional WOLEDs. As shown in Figure 13.25c, this work yielded comparable CIE color coordinates of (0.35, 0.43) and (0.36, 0.46) and an almost balanced luminance brightness of 1173 and 1158 cd m−2 from the anode and cathode, measured at a current density of 10 mA cm−2 . CIE: (0.32, 0.42) (a) @ 1748 cd / m2

EL intensity (a.u. )

CIE: (0.30, 0.43) @ 806 cd / m2

CIE: (0.34, 0.45) @1149 cd / m2

(d) CIE: (0.35, 0.43) @ 1148 cd / m2

CIE: (0.36, 0.46) @1158 cd / m2

500

Anode Cathode

(c)

CIE: (0.35, 0.43) @ 1173 cd / m2

400

(b)

CIE: (0.34, 0.43) @ 1292 cd / m2

600

700 400

500

CIE: (0.37, 0.46) @ 1391 cd / m2

600

700

Wavelength (nm )

Figure 13.25 CIE coordinates and corresponding EL spectra from both sides of transparent WOLEDs made with different anodes of (a) ITO/MoO3 , (b) ITO/Ag (7.5 nm)/MoO3 , (c) ITO/Ag (10.0 nm)/MoO3 and (d) ITO/Ag (12.5 nm)/MoO3 , measured at a constant current density of 10 mA cm−2 .

391

13 Transparent White OLEDs

100 (a)

(b)

10

Power efficiency (lm/W )

392

Measured from anode side Measured from cathode side

100 (d)

(c)

10

1 100

101

102

103

100

101

Luminance

(cd/m2)

102

103

Figure 13.26 Power efficiency as a function of luminance, measured for a set of transparent WOLEDs with different anodes of (a) ITO/MoO3 , (b) ITO/Ag (7.5 nm)/MoO3 , (c) ITO/Ag (10 nm)/MoO3 and (d) ITO/Ag (12.5 nm)/MoO3 .

Figure 13.26 presents the power efficiency as a function of the luminance for a set of structurally identical transparent WOLEDs, fabricated with different anodes of (a) ITO/MoO3 , (b) ITO/Ag (7.5 nm)/MoO3 , (c) ITO/Ag (10 nm)/MoO3 , and (d) ITO/Ag (12.5 nm)/MoO3 . Figure 13.26a shows that there is an obvious deviation in the power efficiency of the EL emission measured from the anode and cathode sides. For example, at a luminance of 100 cd m−2 , a power efficiency of ∼17.3 lm W−1 was measured from the anode side. However, it yielded only a power efficiency of ∼7.1 lm W−1 from the cathode side measured at the same brightness. The more than a-factor-of-two difference in the power efficiency of the EL emission is mainly due to the use of an optically asymmetrical transparent electrode. A more balanced power efficiency of the bidirectional emission can be realized by optimizing the optical out-coupling characteristics at the anode and cathode sides of the transparent WOLEDs. Similar to the results of the CIE color coordinates, as shown in Figure 13.25, a transparent WOLED fabricated with a pair of an ITO (80 nm)/Ag (10 nm)/MoO3 (2.5 nm) anode and a LiF (1.0 nm)/Al (1.5 nm)/Ag (15 nm)/NPB (50 nm) possesses an almost identical power efficiency as the EL emission

13.3 Emission Behavior of Transparent WOLEDs

Table 13.1 Summary of the power efficiency and CIE color coordinates of a set of transparent WOLEDs with different anodes of (a) ITO/MoO3 , (b) ITO/Ag (7.5 nm)/MoO3 , (c) ITO/Ag (10 nm)/MoO3 , and (d) ITO/Ag (12.5 nm)/MoO3 , measured at 100 cd m−2 and 1000 cd m−2 . Power efficiency (lm W−1 ) at 100 cd m−2 /1000 cd m−2

Transparent WOLEDs

(a) (b) (c) (d)

CIE (x, y) measured at 100 cd m−2 /1000 cd m−2

Anode

17.3/11.3

(0.34, 0.42)/(0.33, 0.42)

Cathode

7.1/4.0

(0.31, 0.43)/(0.30, 0.43)

Anode

12.0/7.5

(0.35, 0.43)/(0.34, 0.43)

Cathode

10.6/6.5

(0.35, 0.45)/(0.34, 0.45)

Anode

10.6/6.6

(0.36, 0.43)/(0.35, 0.43)

Cathode

11.1/6.4

(0.38, 0.46)/(0.37, 0.46)

Anode

10.3/6.2

(0.36, 0.43)/(0.35, 0.43)

Cathode

13.0/8.1

(0.38, 0.46)/(0.37, 0.46)

from both sides, measured in the luminance range 100 to 104 cd m−2 . The transparent WOLEDs with perfect symmetrical and bidirectional illumination characteristics thus developed offer new features and design freedoms for application in planar diffused lighting. A summary of the power efficiency and CIE color coordinates of a set of transparent WOLEDs that were measured at 100 and 1000 cd m−2 is given in Table 13.1. The EQE was determined from the EL measurement. It is found that the EQE (the sum of the EL emission from both sides) measured for a set of transparent WOLEDs, as presented in Figures 13.25a–d, was about 10%. This is comparable to the EQE measured for a control bottom-emission WOLED. This suggests that interposing an ultrathin Ag interlayer over the thickness range 7.5 to 12.5 nm does not induce a significant change in the efficiency of the transparent WOLEDs. The pair of transparent electrodes is optically and electrically favorable for the transparent WOLEDs. In addition to the measurement of EL emission at different angles, color variation of the devices under different biases was also analyzed, as shown in Figure 13.28. Apart from the WOLED operating at a low brightness level of 100 nm) at a large viewing angle. However, the CIE coordinates are angular independent for devices with the thickness of an organic stack less than 100 nm. It is clear that a weak angular-dependent EL emission can be realized if the spectral overlap between the resonant wavelength of the cavity and that of the peak EL emissions in transparent WOLEDs can be avoided, supported by both experimental results and theoretical simulations. The current efficiency (CE), power efficiency (PE), and CIE coordinates of the EL emissions in the normal direction and at 60∘ measured from both sides of transparent WOLEDs are summarized in Table 13.3. The measured and calculated luminance ratios of EL emissions from the anode side to those from the cathode side, and the results of the visible-light transparency of the devices are also listed for comparison. The balanced emission characteristics from both the anode and cathode sides can be obtained. The theoretical calculation agrees with the experimental results in showing that transparent WOLEDs possess an average transparency of 54% over the visible-light wavelength from 380 to 780 nm. EL emissions from both sides of the device exhibited comparable color quality, with CIE coordinates of (0.34, 0.43) and (0.37, 0.46) measured from the anode and cathode sides. This shows that the emission from the cathode side has a weak angular dependency, with a minor change in the CIE coordinates from (0.37, 0.46)

13.3 Emission Behavior of Transparent WOLEDs

0.45

0.50

(a) From anode side

60 nm 120 nm

80 nm 140 nm

100 nm

0.45 0.35 0.40

CIE y-coordinate

CIE x-coordinate

0.40

0.30 Theoretical calculation Measured results 0.25

0.45

–60 –40 –20 0 20 40 Viewing angle (deg)

60

(b) From cathode side

–60 –40 –20 0 20 40 Viewing angle (deg.)

60 nm 120 nm

80 nm 140 nm

60

100 nm

0.35

0.50

0.45 0.35 0.40

CIE y-coordinate

CIE x-coordinate

0.40

0.30 Theoretical calculation Measured results 0.25

–60 –40 –20 0 20 40 Viewing angle (deg)

60

–60 –40 –20 0 20 40 Viewing angle (deg.)

60

0.35

Figure 13.31 CIE coordinates (x, y) as a function of the viewing angle, obtained for transparent WOLEDs with different cavity lengths of 60, 80, 100, 120, and 140 nm, from (a) anode and (b) cathode sides. The experimental results of the angular-dependent emissions (open triangle symbols), measured from both sides of the transparent WOLEDs having an optimized 100-nm-thick cavity length, are also presented.

(normal) to (0.33, 0.43) at a viewing angle of 60∘ . However, the EL emission from the anode side of transparent dichromatic WOLEDs has no observable angular-dependent feature, with almost no change in the CIE coordinates over the viewing angle range from normal to 60∘ . The concept of achieving weak angular-dependent EL emission in a transparent trichromatic white WOLED was also examined theoretically using three emitters of FIrpic (478 nm), Ir(fbi)2 acac (562 nm), and bis(1-phenylisoquinoline)-(acetylacetonate)-iridium(III) (Ir(piq)2 acac) (628 nm). The same device configuration was adopted in the analyses. The emission characteristics of a transparent trichromatic white WOLED, e.g. having a 100-nm-thick cavity length, are shown in Figure 13.32. The color rendering

397

13 Transparent White OLEDs

Table 13.3 Summary of CE, PE, CIE coordinates at 0∘ and 60∘ obtained from the anode (having a luminance of 1173 cd m−2 ) and cathode (having a luminance of 1158 cd m−2 ) sides of the transparent WOLEDs. The measured and calculated luminance ratio of EL emissions from both sides and the results of visible-light transparency of the devices are also listed for comparison. CIE Measured @ 10 mA cm−2

CE (cd/A)

PE (lm/W)

(normal)

(60∘ )

Luminance ratio T (%)

Anode side

11.73

6.30

(0.34, 0.43)

(0.34, 0.43)

1.01a)

54.0a)

(0.33, 0.43)

1.02b)

53.8b)

Cathode side

11.58

6.08

(0.37, 0.46)

a) Obtained from the experimental results. b) Calculated results.

Normalized spectra (arb. unit)

398

400

From anode side From cathode side

500 600 Wavelength (nm)

Figure 13.32 Emission spectra, from anode and cathode sides, for of a transparent trichromatic white WOLED with a 100-nm-thick cavity length.

700

index of the WOLEDs is ∼80. The results from a transparent trichromatic white WOLED are similar to those from a transparent dichromatic white WOLED. The CIE coordinates (x, y) of a transparent trichromatic white WOLED at different viewing angles, calculated for devices with different cavity lengths of 60, 80, 100, 120, and 140 nm, are presented in Figure 13.33. At the anode side, it is found that the emission color is not sensitive to the viewing angle for devices with a cavity length larger than 100 nm. However, the CIE color coordinates have a slight angular-dependent behavior when the cavity length is below 100 nm. At the cathode side, there is a tendency of a blue shift in the EL emission at large viewing angles for devices with a cavity length greater than 100 nm. The CIE coordinates are angular independent for devices with a cavity length less than 100 nm. At a 100-nm-thick cavity length, the emission color is less angular dependent at the anode side, while the emission color is slightly dependent on the viewing angle at the cathode. The concept of avoiding an overlap between the resonant wavelength and that of the emission peaks is adopted to design transparent trichromatic WOLEDs, realizing weak angular-dependent emission and highly transparent WOLEDs. The selection of an appropriate cavity length, e.g. 100 nm with a resonant mode of 550 nm in this case, allows the angular-dependent emission to be reduced from both sides of the

13.3 Emission Behavior of Transparent WOLEDs

0.50

(a) From anode side

60 nm 120 nm

80 nm 140 nm

100 nm

0.45

0.40

0.40

CIE y-coordinate

CIE x-coordinate

0.45

0.35

0.30

0.50

–60 –40 –20 0 20 40 Viewing angle (deg)

60

–60 –40 –20 0 20 40 Viewing angle (deg.)

0.35

60

0.45

(b) From cathode side

0.40 0.40 0.35

CIE y-coordinate

CIE x-coordinate

0.45

0.30 60 nm 120 nm 0.25 –80 –60 –40 –20 0 20 40 Viewing angle (deg)

80 nm 140 nm

60 –80 –60 –40 –20 0 20 40 Viewing angle (deg.)

100 nm 60

0.35 80

Figure 13.33 CIE coordinates (x, y), from (a) anode and (b) cathode sides, as a function of the viewing angle for transparent trichromatic WOLEDs with different cavity lengths of 60, 80, 100, 120, and 140 nm.

devices using this simple approach. The angular-dependent emission behavior, particularly for transparent WOLEDs, can be further reduced by incorporating different microstructures [50]. A combination of 100-nm-thick cavity length e.g. with a resonant mode of 550 nm, and a 50-nm-thick capping layer, in this case allows the angular-dependent emission from both sides of the devices to be reduced. The emission color is less angular dependent at the anode side, while the emission from the cathode side is weakly angular dependent at large viewing angles at the cathode side. The selection of a 100-nm-thick cavity length enables a preferred white emission from both sides of the devices. In this work, high-performing transparent WOLEDs with weak angular-dependent EL emission characteristics have been demonstrated, achieved by avoiding a spectral overlap between the peak wavelengths of the emitting units and the resonant wavelength of

399

400

13 Transparent White OLEDs

the corresponding organic microcavity. The transparent WOLEDs also possess a high visible-light transparency, allowing the surface light source to shine in both directions, an exciting new lighting technology that could bring new device concepts.

13.4 Conclusions A transparent phosphorescent WOLED has been demonstrated using a dichromatic white emissive layer of TCTA:FIrpic:Ir(fbi)2 acac (19%, 1%, 10 nm)/TCTA:FIrpic (19%, 7.5 nm). The transparent WOLED has a visible-light transparency of >50% and an almost identical PE of 11 lm W−1 (measured at 100 cd m−2 ) from both sides. High-performing transparent dichromatic WOLEDs possessing weak angular-dependent EL emission and stable CIE coordinates of (0.34, 0.43) and (0.37, 0.46), measured from both anode and cathode sides, were demonstrated. The results confirmed that the avoidance of a spectral overlap between the resonant wavelength and that of the emitting peaks moderates the angular-dependent emission behavior in transparent WOLEDs. This was realized by optimizing the WOLED structure through a combination of theoretical calculation and experimental optimization. In addition, visible-light transparency and angular-dependent emission behavior of dual-sided transparent trichromatic WOLEDs were also analyzed. The realization of transparent WOLEDs with bidirectional and symmetrical illumination characteristics, including the PE, CIE color coordinates, and weak angular-dependent EL emission spectra, offers new features and design freedoms for application in planar diffused lighting. WOLEDs can be used for a wide variety of industrial, medical, health care, and other consumer-oriented applications. The new transparent WOLEDs increase the flexibility of device integration and engineering, opening up a plethora of opportunities and potential for a new generation of advanced display and lighting technologies.

References Baldo, M.A., Brien, D.F.O., You, Y. et al. (1998). Nature 395: 151. Miao, Y.Q., Wang, K.X., Zhao, B. et al. (2017). J. Mater. Chem. C 5: 12474–12482. Andrade, B.W.D., Thompson, M.E., and Forrest, S.R. (2002). Adv. Mater. 14: 147. Tokito, S., Iijima, T., Tsuzuki, T., and Sato, F. (2003). Appl. Phys. Lett. 83: 2459. Miao, Y.Q., Gao, Z.X., Tao, R. et al. (2016). Sci. Adv. Mater. 8 (2): 401. Su, S.J., Gonmori, E., Sasabe, H., and Kido, J. (2008). Adv. Mater. 20: 4189. Reineke, S., Lindner, F., Schwartz, G. et al. (2009). Nature 459: 234. Sun, Y. and Forrest, S.R. (2007). Appl. Phys. Lett. 91: 263503. Sasabe, H., Takamatsu, J., Motoyama, T. et al. (2010). Adv. Mater. 22: 5003. Choi, W.H., Tam, H.L., Zhu, F.R. et al. (2013). Appl. Phys. Lett. 102: 153308. Gu, G., Bulovic, V., Burrows, P.E. et al. (1996). Appl. Phys. Lett. 68: 2602. Yamamori, A., Hayashi, S., Koyama, T., and Taniguchi, Y. (2001). Appl. Phys. Lett. 78: 3343. 13 Chung, C.H., Ko, Y.W., Kim, Y.H. et al. (2005). Appl. Phys. Lett. 86: 09350. 14 Hsu, S.F., Lee, C.C., Hwang, S.W., and Chen, C.H. (2005). Appl. Phys. Lett. 86: 253508. 1 2 3 4 5 6 7 8 9 10 11 12

References

15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

Lee, C.J., Pode, R.B., Han, J.I., and Moon, D.G. (2007). Appl. Surf. Sci. 253: 4249. Yook, K.S., Jeon, S.O., Joo, C.W., and Lee, J.Y. (2008). Appl. Phys. Lett. 93: 013301. Zhang, T.Y., Zhang, L.T., Ji, W.Y., and Xie, W.F. (2009). Opt. Lett. 34: 1174. Lee, J., Hofmann, S., Furno, M. et al. (2011). Org. Elect. 12: 1383. Cho, H., Choi, J.M., and Yoo, S.H. (2011). Opt. Exp. 19: 1113. Gil, T.H., May, C., Scholz, S. et al. (2010). Org. Elect. 11: 322. Tessler, N., Burns, S., Becker, H., and Friend, R.H. (1997). Appl. Phys. Lett. 70: 556. Juang, F.S., Laih, L.H., Lin, C.J., and Hsu, Y.J. (2002). Jpn. J. Appl. Phys. 41: 2787. Mullen, K. and Scherf, U. (2006). Organic Light Emitting Devices. Wiley-VCH. Tokito, S., Iijima, T., Suzuri, Y. et al. (2003). Appl. Phys. Lett. 83: 569. Huang, W.S., Lin, J.T., Chien, C.H. et al. (2004). Chem. Mater. 16: 2480. Pope, M. and Swenberg, C.E. (1999). Electronic Processes in Organic Crystals and Polymers, 2e. New York: Oxford University Press. Wang, Q., Ding, J.Q., Ma, D.G. et al. (2009). Appl. Phys. Lett. 94: 103503. Wang, Q., Ding, J., Ma, D. et al. (2009). Adv. Mater. 21: 2397. Lee, J., Chopra, N., Eom, S.H. et al. (2008). Appl. Phys. Lett. 93: 123306. Hung, W., Ke, T.H., Lin, Y.T. et al. (2006). Appl. Phys. Lett. 88: 064102. Tong, K.L., Tsang, S.W., Tsung, K.K. et al. (2007). J. Appl. Phys. 102: 093705. Liew, Y.F., Zhu, F.R., Chua, S.J., and Tang, J.X. (2004). Appl. Phys. Lett. 85: 4511. Yeh, H.C., Meng, H.F., Lin, H.W. et al. (2012). Org. Electron. 13: 914. Su, S.J., Chiba, T., Takeda, T., and Kido, J. (2008). Adv. Mater. 20: 2125. Turro, N.J. (1991). Modern Molecular Photochemistry, Chaps. 3, 5 and 9. University Science Books. Li, Y.Q., Tan, L.W., Hao, X.T. et al. (2005). Appl. Phys. Lett. 86: 153508. Fukuda, Y., Watanabe, T., Wakimoto, T. et al. (2000). Synth. Met. 111: 1. Beierlein, T.A., Ott, H.P., Hofmann, H. et al. (2002). Proc. SPIE 4464: 178. Neyts, K., De Visschere, P., Fork, D.K., and Anderson, G.B. (2000). J. Opt. Soc. Am. B 17: 114. Choi, W.H., Tam, H.L., Ma, D.G., and Zhu, F.R. (2015). Opt. Express 23 (11): A471–A479. Wang, X.Z., Ng, G.M., Ho, J.W. et al. (2010). Sel. Top. Quantum Electron. 16: 1685. Riel, H., Karg, S., Beierlein, T. et al. (2003). J. Appl. Phys. 94: 5290. Lee, J.H., Hofmann, S., Furno, M. et al. (2013). Opt. Lett. 37: 2007. Lee, J.H., Cho, H., Koh, T.W. et al. (2013). Opt. Exp. 21: 28040. Chang, H.W., Lee, J.H., Koh, T.W. et al. (2013). Laser Photonics Rev. 7: 1079. Huh, J.W., Moon, J., Lee, J.W. et al. (2012). Org. Electron. 13: 1386. Lee, J., Hofmann, S., Furno, M. et al. (2011). Opt. Lett. 36: 1443. Ji, W.Y., Zhang, L.T., Zhang, T.Y. et al. (2009). Opt. Lett. 34: 2703. Puzzo, D.P., Helander, M.G., O’Brien, P.G. et al. (2011). Nano Lett. 11: 1457. Kim, J.B., Lee, J.H., Moon, C.K. et al. (2013). Adv. Mater. 25: 3571.

401

403

14 Optical Properties of Thin Films V.-V. Truong 1 , S. Tanemura 2,4 , A. Haché 3 , and L. Miao 4 1

Physics Department, Concordia University, 7141 Sherbrooke St W, SP 367.03, Montreal, Quebec, H4B 1R6, Canada Japan Fine Ceramics Centre, Mutsuno 2-4-1, Atsuta-ku, Nagoya 456-8587, Japan Département de physique et d’astronomie, Université de Moncton, Moncton, New Brunswick, E1A 8T1, Canada 4 Guilin University of Electronic Technology, Guilin 541004, P. R. China 2 3

CHAPTER MENU Introduction, 403 Optics of Thin Films, 404 Reflection-Transmission Photoellipsometry for Determination of Optical Constants, 408 Application of Thin Films to Energy Management and Renewable-Energy Technologies, 412 Application of Tunable Thin Films to Phase and Polarization Modulation, 424 Conclusions, 430 References, 430

14.1 Introduction One of the classical treatments of the optical properties of thin films found in the literature remains in the book by O.S. Heavens first published by Butterworths Scientific Publications in 1955 and subsequently reviewed and republished by Dover in 1965 and in 1991 [1]. Besides devoting many sections of the book to the experimental and practical aspects of thin films, the author has also presented Maxwell’s equations to calculate the intensity of light transmitted and reflected by single or multi-layered films. Other comprehensive treatments of optical properties of thin films can also be found in references such as Born and Wolf [2], Azzam and Bashara [3], Palik [4], Klein and Furtak [5], Ward [6], and Bass [7]. Although photometry was frequently used to characterize the optical properties of thin films, ellipsometry, or the study of the state of polarization of the reflected or transmitted light and their relation with material properties, has become a standard and powerful characterization method as more computing power and techniques are readily available. Ellipsometric parameters are very sensitive to changes in material properties, and it is well recognized that ellipsometry is capable of detecting a few angstroms’ change in material layers and can be used to study surface structures in the subnanometer range.

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

404

14 Optical Properties of Thin Films

In this chapter, expressions for reflection and transmission coefficients are first reported for an isotropic thin film on a substrate. Observations are then made regarding the matrix method for layered structures and anisotropic films before presenting an original approach that combines reflection and transmission photoellipsometry for studying a multi-layer structure. Finally, the chapter is completed with a few selected applications of thin films.

14.2 Optics of Thin Films 14.2.1

An Isotropic Film on a Substrate

A situation often encountered is the case of an isotropic film on an infinite isotropic substrate. The film is assumed to have parallel-plane boundaries and a thickness D1 . The ambient (medium 0) is characterized by a real index of refraction N 0 , whereas the film (medium 1) and the substrate (medium 2) have complex indices of refraction, N 1 and N 2 , respectively. Let us consider a plane wave incident at an angle 𝜃 0 from medium 0 that results in a refracted wave at angle 𝜃 1 in medium 1 and a transmitted wave at angle 𝜃 2 in medium 2 (Figure 14.1). By taking into account multiple reflections at the film–ambient and film–substrate interfaces, the reflection and transmission amplitudes can be calculated. Let 𝛽 be the phase change that occurs when the light wave traverses the film once from one boundary to the other: ( ) D1 𝛽 = 2𝜋 (14.1) N1 cos 𝜃1 𝜆 or

( 𝛽 = 2𝜋

D1 𝜆

)

1

(N12 − N02 sin2 𝜃0 ) 2

(14.2)

where 𝜆 is the wavelength of the incident light, and Eq. (14.2) is obtained from Eq. (14.1) by applying Snell’s law, N 0 sin𝜃 0 = N 1 sin𝜃 1 .

θ0 Ambient (0)

Film (1)

Substrate (2)

θ1 θ2

Figure 14.1 Schematic presentation of reflection, refraction, and transmission of an incident wave on an isotropic film deposited on an isotropic substrate.

14.2 Optics of Thin Films

Summation of all partially reflected and transmitted waves within the film leads to an infinite geometric series and the total reflected amplitude R is given by R = r01 +

t01 t10 r12 e−j2𝛽 1 − r10 r12 e−j2𝛽

(14.3)

or R=

r01 + r12 e−j2𝛽 1 + r01 r12 e−j2𝛽

(14.4)

when r10 is replaced by −r10 and (t 01 t 10 ) by (1 − r10 2 ). In a similar way, the total transmitted amplitude can also be obtained as T=

t01 t12 e−j𝛽 1 + r01 r12 e−j2𝛽

(14.5)

where r01 and r12 are, respectively, the Fresnel reflection coefficients at the ambient–film (0–1) and film–substrate (1–2) interfaces, and t 01 and t 12 are the Fresnel transmission coefficients defined in a similar manner. Equations (14.3)–(14.5) are valid when the incident wave is linearly polarized either parallel (p) or perpendicular (s) to the plane of incidence, and a subscript p or s can be added to the formulas in order to account for the polarization of the incident light. Thus, the reflectance amplitude resulting from a p-polarized light will be denoted by Rp and the reflectance amplitude resulting from s-polarized light by Rs . Similarly, the transmission amplitudes taking into account the p and s polarizations will be T p and T s , respectively. The proper expressions for the Fresnel coefficients then have to be used, and they are as follows [3]: r01p =

N1 cos 𝜃0 − N0 cos 𝜃1 N1 cos 𝜃0 + N0 cos 𝜃1

(14.6)

r12p =

N2 cos 𝜃1 − N1 cos 𝜃2 N2 cos 𝜃1 + N1 cos 𝜃2

(14.7)

r01s =

N0 cos 𝜃0 − N1 cos 𝜃1 N0 cos 𝜃0 + N1 cos 𝜃1

(14.8)

r12s =

N1 cos 𝜃1 − N2 cos 𝜃2 N1 cos 𝜃1 + N2 cos 𝜃2

(14.9)

t01p =

2N0 cos 𝜃0 N1 cos 𝜃0 + N0 cos 𝜃1

(14.10)

t12p =

2N1 cos 𝜃1 N2 cos 𝜃1 + N1 cos 𝜃2

(14.11)

t01s =

2N0 cos 𝜃0 N0 cos 𝜃0 + N1 cos 𝜃1

(14.12)

t12s =

2N1 cos 𝜃1 N1 cos 𝜃1 + N2 cos 𝜃2

(14.13)

It is noted that angular quantities and the indices of refraction of the media involved are related through Snell’s law: N0 cos 𝜃0 = N1 cos 𝜃1 = N2 cos 𝜃2

(14.14)

405

406

14 Optical Properties of Thin Films

The above expressions are valid when the substrate is semi-infinite; i.e. reflections at the lower surface of the substrate are not taken into account. In the case of a finite-thickness substrate, the system can be considered to be an ambient medium (medium 0) followed by two films (medium 1 and medium 2, respectively), and a third infinite medium is added (medium 3). By assuming a coherent combination of multiply reflected and transmitted waves within the two films, the summation of those waves will lead to final expressions for the total reflected and transmitted amplitudes. 14.2.2

Matrix Methods for Multi-Layered Structures

Obviously, the summation of multiply reflected and transmitted waves within multiple layers is a very cumbersome process, and alternative methods can be sought for calculating the optical properties of multi-layered structures. Fortunately, the generally known matrix methods offer a very elegant solution [8, 9], and a detailed description of such a method can be found in Reference [3]. It makes use of the fact that the equations for the propagation of light are linear, and the continuity of the tangential fields across an interface can be expressed as a 2 × 2 linear-matrix transformation. Let us consider a multi-layered system consisting of a stack of 1, 2, 3, … , j, … m homogeneous and isotropic layers sandwiched between two semi-infinite ambient (0) and substrate (m + 1) media. It can be shown [3] that the overall reflection and transmission properties of such a structure are represented by a scattering matrix S that can be expressed as the product of each of the interface and layer matrices I and L, which describe the effects of the individual interfaces and layers constituting the whole system. The scattering matrix is thus given by S = I01 L1 I12 L2 … I(j−1)j Lj … Lm Im(m+1)

(14.15)

where I(j−1)j refers to the interface matrix for the interface between medium (j − 1) and medium (j), and Lj is the layer matrix for layer (j). The matrix for a plane interface has the form [ ] ] [ 1∕t r∕t 1 1 r (14.16) = I= t r 1 r∕t 1∕t where r and t are, respectively, the amplitudes of reflection and transmission coefficients. The matrix for a layer of thickness D on the other hand is given by [ ] ej𝛽 0 L= (14.17) 0 e−j𝛽 where the phase shift (layer phase thickness) 𝛽 is expressed as a function of the layer thickness D, the layer refraction index N, the wavelength 𝜆, and the local angle of incidence 𝜃 as 2𝜋DN 𝛽= cos𝜃 (14.18) 𝜆 Let us consider, as an example, the explicit case of two films (1 and 2) between a semi-infinite ambient (0) and a substrate (3) media. According to Eq. (14.14), the scattering matrix for this case is given by S = I01 L1 I12 L2

(14.19)

14.2 Optics of Thin Films

Considering the expressions for the interface and layer matrices expressed by Eqs. (14.16) and (14.17), the scattering matrix is now ][ ][ ][ ] [ ] [ 1 r12 1 r23 ej𝛽1 0 ej𝛽2 0 1 r01 1 × (14.20) S= × t01 t12 t23 r01 1 r12 1 r23 1 0 e−j𝛽1 0 e−j𝛽2 or

S=

ej(β1 +β2 ) t01 t12 t23

(1 + r01 r12 e−j2β1 )r23 ⎤ (1 + r01 r12 e−j2β1 ) ⎡ ⎢+ (r + r e−j2β1 ) × r e−j2β2 + (r + r e−j2β1 ) × e−j2β2 ⎥ 12 01 23 12 01 ⎥ ⎢ ⎥ =⎢ ⎥ ⎢ −j2β −j2β ⎥ ⎢ (r01 + r12 e 1 ) (r01 + r12 e 1 )r23 ⎥ ⎢ −j2β −j2β −j2β −j2β ⎣+ (r01 r12 + e 1 ) × r23 e 2 + (r01 r12 + e 1 ) × e 2 ⎦

(14.21)

As the reflection and transmission coefficients of the multi-layered structure are respectively given by r = (S21 /S11 ) and t = (1/S11 ) [3], expressions for those coefficients can be derived from Eq. (14.21). r and t are thus obtained as r=

(r01 + r12 e−j2𝛽1 ) + (r01 r12 + e−j2𝛽1 )r23 e−j2𝛽2 (1 + r01 r12 e−j2𝛽1 ) + (r12 + r01 e−j2𝛽1 )r23 e−j2𝛽2

(14.22)

t=

t01 t12 t23 e−j(𝛽1 +𝛽2 ) (1 + r01 r12 e−j2𝛽1 ) + (r12 + r01 e−j2𝛽1 )r23 e−j2𝛽2

(14.23)

and

The above formulas apply to both polarizations p and s when the proper subscript is added to r and t. The results in Eqs. (14.22) and (14.23) can be generalized to a larger number of layers, although the expressions derived can be quite lengthy. Matrix calculations for a multi-layer structure can, however, be handled conveniently by a computer. 14.2.3

Anisotropic Films

Many films that are of interest are anisotropic, and this requires a theoretical treatment that is different than the one used for isotropic media. Examples of such films are Langmuir–Blodgett layers, aggregated or island-like structure films, columnar structure films, semiconductor superlattices, and complex oxide films. The optical function of an anisotropic film can be described by a dielectric tensor with three principal components, which are referenced to the crystal coordinate system. Discussions on anisotropic films can be found in various references [3, 10–14]. In the case of aggregated films, we have a uniaxially anisotropic film that has its optical axis perpendicular to its boundaries with the ambient and the substrate. Gold aggregated films, for example, have been studied recently, and their optical properties in the direction perpendicular to the film plane were found to be significantly different from those in the direction parallel to it [15]. Optical thin films composed of a columnar structure also show a large refractive anisotropy. A biaxial model for these films has been proposed in which three main refractive indices correspond to a direct reference with one axis in the direction of the columns and the two other axes in the perpendicular plane [16]. With this model and a

407

408

14 Optical Properties of Thin Films

4 × 4 transfer matrix theory, the reflectance as well as the transmittance of a multi-layer stack of anisotropic films can be computed.

14.3 Reflection-Transmission Photoellipsometry for Determination of Optical Constants As mentioned previously, ellipsometry has become a standard tool for characterizing thin films, and many excellent reviews on ellipsometry are available in the literature [17–20]. In the following, we choose to present a method that combines reflection and transmission photoellipsometry as proposed by Bader et al. [21]. This is a very convenient and powerful way to study practical samples consisting of a combination of thin and thick films. The fact that transmission-mode photoellipsometry is considered together with reflection photoellipsometry eliminates the need to use a completely absorbing substrate or to grind the backside of semitransparent substrates for reflection-only ellipsometric measurements. In this method, the role of the substrate is examined, and contributions from multiple reflections in the substrate are taken into account. The addition of transmission to reflection measurements, on the other hand, can also improve accuracy in determining the optical constants of the thin films studied. For simplicity, all thin films considered here are assumed to be perfect, isotropic, and uniform; i.e. they have completely smooth and parallel surfaces. It is also assumed that the ambient surrounding the samples has no absorption. In this context, a thick film is defined as a layer for which no interference effects can be observed for a given wavelength resolution. As is well known, this situation is obtained when an average of the reflectance and transmittance on the phase shift related to the thick film is taken. Generally, the reason behind the averaging is the existence of a wavelength resolution, but it can also be related to film imperfections or experimental procedures. A substrate can thus be considered as a thick film. For easy reference, the notations adopted in this section are the ones used in Reference [21]. 14.3.1

Photoellipsometry of a Thick or a Thin Film

Let N i , Di , 𝜃 i , and 𝛽 i be the refractive index (complex), the thickness (real), the angle of propagation (complex), and the phase shift related to film number i, respectively. The subscript 0 refers to the ambient medium. Proper care must be taken in the following formulas when calculating cos(𝜃 i ): ( cos 𝜃i = + 1 − (

N02 sin2 𝜃0 Ni2 N02 sin2 𝜃0

)2 )2

=− 1− N2 ( ) i Di 𝛽 = 2𝜋 Ni cos 𝜃i 𝜆

[ ]1 if Im Ni2 − N02 sin2 𝜃0 2 ≥ 0 ]1 [ if Im Ni2 − N02 sin2 𝜃0 2 < 0

where 𝜆 is the light wavelength. Note that Im(𝛽 i ) ≥ 0.

(14.24a) (14.24b)

14.3 Reflection-Transmission Photoellipsometry for Determination of Optical Constants

We now consider an ellipsometer that consists of a rotating polarizer of angle 𝜃 p , one film sample, and a rotating analyzer of angle 𝜃 p (for a practical description of ellipsometry and ellipsometers, readers may refer to Reference [18]). For the case 𝜃 p = 𝜃 a , one can write the reflected and transmitted signals for the detector of such an ellipsometer by using the Jones or Stokes–Mueller formalisms, respectively, as I R = k[Rp cos4 𝜃a + Rs sin4 𝜃a + 2Re(Rps )sin2 𝜃a cos2 𝜃a ]

(14.25a)

I T = k[Tp cos4 𝜃a + Ts sin4 𝜃a + 2Re(Tps )sin2 𝜃a cos2 𝜃a ]

(14.25b)

where k is a normalizing constant that depends on the intensity of the incident light and the detector. The subscripts p and s have the usual meaning of p and s polarizations, respectively, and angle 𝜃 a is referenced to the incidence plane. Here, Rp(s) (real), T p(s) (real), Rp(s) (complex), and T p(s) (complex) refer to reflection and transmission experiments and are defined in the following pages. As in photometry, we can normalize this signal by dividing it by the signal measured by the detector without the sample, and then we obtain I R = Rp cos4 𝜃a + Rs sin4 𝜃a + 2Re(Rps )sin2 𝜃a cos2 𝜃a

(14.26a)

I T = Tp cos4 𝜃a + Ts sin4 𝜃a + 2Re(Tps )sin2 𝜃a cos2 𝜃a

(14.26b)

It is well known that the introduction of a compensator (retarder) in the optical path can replace the real part in Eq. (14.25a-b) by an imaginary part. Thus, complete photoellipsometry permits the determination of the following quantities that can be expressed in terms of the total reflection and transmission amplitudes rp(s) and t p(s) : Rp(s) = |rp(s) |2

(14.27a)

Tp(s) = |tp(s) |2

(14.27b)

Rp(s) = ∣ rp rs∗ ∣

(14.27c)

Tp(s) = |tp ts∗ |

(14.27d)

Rp(s) = ⟨|rp(s) |2 ⟩

(14.28a)

Tp(s) = ⟨|tp(s) |2 ⟩

(14.28b)

Rp(s) = (rp rs∗ )

(14.28c)

Tp(s) = (tp ts∗ )

(14.28d)

and

for a thick film. The quantities rp(s) and t p(s) are the total reflection and transmission amplitudes, respectively, as given in Eq. (14.29a-b) for both polarizations. It must be noted that the ambient-in and ambient-out are supposed to be identical. Otherwise the usual correction factor must be applied to all transmission terms. The fences denote the

409

410

14 Optical Properties of Thin Films

well-known averaging, which deletes the interferences between the multiple beams. The total reflection and transmission amplitudes are given for both polarizations by + r = r0.1 +

t=

t0,1 + t0,1 − r1,2 + exp(2i𝛽1 )

1 − r0,1 − r1,2 + exp(2i𝛽1 ) t0,1 + t1,2 + exp(2i𝛽1 )

1 − r0,1 − r1,2 + exp(2i𝛽1 )

(14.29a) (14.29b)

In Eqs. (14.29a) and (14.29b), the bottom indices refer to the interface, and we have omitted the polarization index. The plus and minus superscript refers to the direction of the incident light on the interface (forward and backward directions). In this simple case, rij+(−) and tij+(−) are the well-known Fresnel amplitudes. It is then easy to show, by using the series expansion, that the averages are obtained for both polarizations in Eqs. (14.28a) and (14.28b) as Rp(s) = ⟨|rp(s) |2 ⟩ = |r0,1,p(s) + |2 +

|t0,1,p(s) + |2 |t0,1,p(s) − |2 |r1,2,p(s) + |2 exp(−4 Im 𝛽1 ) 1 − |r0,1,p(s) − |2 |r0,1,p(s) + |2 exp(−4 Im 𝛽1 ) (14.30a)

Tp(s) = |tp(s) |2 =

|t0,1,p(s) + |2 |t1,2,p(s) + |2 exp(−2Im 𝛽1 ) 1 − |r0,1,p(s) − |2 |r1,2,p(s) + |2 exp(−4Im 𝛽1 )

(14.30b)

For the case of Eqs. (14.27c) and (14.27d), the averages are also easily obtained as Rp(s) = ⟨(rp rs∗ )⟩ = (r0,1,p + r0,1,s −∗ ) (t0,1,p + t0,1,s +∗ )(t0,1,p − t0,1,s −∗ )(r1,2,p + r1,2,s +∗ ) exp(−4Im𝛽1 ) + 1 − (r0,1,p − r0,1,s −∗ )(r1,2,p + r1,2,s +∗ ) exp(−4Im𝛽1 ) (14.30c) Tp(s) = ⟨(tp ts∗ )⟩ +

(t0,1,p + t0,1,s +∗ )(t1,2,p + t1,2,s +∗ ) exp(−2Im𝛽1 ) 1 − (r0,1,p − r0,1,s −∗ )(r1,2,p + r1,2,s +∗ ) exp(4Im𝛽1 )

(14.30d)

It is to be noted that the mathematical structure of Eqs. (14.29a-b) and (14.30a-d) are similar in the sense that an iterative process can be used to compute the reflection and transmission through a stack of thin films, and the same iterative procedure can also be used for a stack of thick films (with appropriate changes). Moreover, this procedure computes not only ⟨|r|2 ⟩ and ⟨|t|2 ⟩ but also ⟨rp rs∗ ⟩ and ⟨tp ts∗ ⟩ for a stack of thick films. This will be shown in the following section. 14.3.2

Photoellipsometry for a Stack of Thick and Thin Films

Here, we consider a layered structure consisting of thick and thin films, as shown in Figure 14.2, with the ambient in denoted by 0 and the ambient out denoted by N. The main idea introduced here is to define pseudointerfaces between thick and thin films. The pseudointerface between thick film (i) and thick film (i + 1) is defined by the reflection and transmission amplitudes through the layered structure of thin films between

14.3 Reflection-Transmission Photoellipsometry for Determination of Optical Constants

Ambient 0

Ambient 0 Thin layers Thick layer 1

Pseudointerface Thick layer 1

Thin layers Pseudointerface Thick layer 2

Thick layer (N–1)

Ambient N

Thick layer (N –1)

Ambient N

Pseudointerface

Figure 14.2 Schematic diagram of a multi-layer structure that consists of thin and thick layers. Thin and thick layers (or substrates) can be replaced by a pseudointerface as defined in the text. Source: Reprinted Fig. 1, Reflection-Transmission Ellipsometry: Theory and Experiments, G. Bader, P.V. Ashrit, F.E. Girouard, and V.-V. Truong, Appl. Optics, 34, 1684 (1995). Copyright (1995) with permission from The Optical Society of America.

the two thick films, and thus for both polarizations and incident directions on the layered thin-film structure we get ri, i + 1, p(s) +(−) , t i, i + 1, p(s) +(−) , where the subscript (i, i + 1) denotes the interface, and the subscript p(s) denotes the polarization. The usual thin-film theory is used to obtain these amplitudes. The computation of the quantities Rp(s), Tp(s), Rps, and Tps uses the same iterative procedure. Let us define Rp(s) (i, N) as the reflectance for incident light coming in the positive direction on a pseudointerface (i, i + 1), reflected on stack (i + 1) to (N + 1) of thick films, and let Tp(s) (i, N) be the square modulus of the transmission amplitude through the same stack for both polarizations. Quantities Rps (i, N) and Tps (i, N) are defined in the same manner. We thus obtain the iterative equations for these quantities as R(i − 1, N) = |ri−1,i + |2 + T(i − 1, N) =

|ti−1,i + |2 |ti−1,i − |2 R(i, N) exp(−4 Im𝛽1 )

1 − |ri−1,i − |2 R(i, N) exp(4 Im𝛽1 ) |ti−1,i + |2 T(i, N) exp(−2 Im𝛽1 )

1 − |ri−1,i − |2 R(i, N) exp(4 Im𝛽1 )

(14.31a) (14.31b)

for both polarizations and the following expressions for Rps and T ps : Rps (i − 1, N) = (ri−1,i,p + ri−1,i,s +∗ ) (ti−1,i,p + ti−1,i,s +∗ )(ti−1,i,p − ti−1,i,s +∗ )Rps (i, N) exp(−4 Im𝛽1 ) + 1 − (ri−1,i,p − ri−1,i,s −∗ )Rps (i, N) exp(−4 Im𝛽1 ) (14.31c) Tps (i − 1, N) =

(ti−1,i,p + ti−1,i,p +∗ )Tps (i, N) exp(−2 Im𝛽i ) 1 − (ri−1,i,p − ri−1,i,s −∗ )Rps (N) exp(−4 Im𝛽i )

(14.31d)

It must be pointed out that Eqs. (14.30a-d) are a direct consequence of Eqs. (14.29a-b), which are easily obtained by considering the series developments of r and t. In the series

411

412

14 Optical Properties of Thin Films

multiplication, the averaging process eliminates all but the diagonal terms. This corresponds to the elimination of the interference terms between different beams. One can construct Eqs. (14.30a-d) in the same way by simply adding a layer iteratively, and then using these iterative equations to evaluate Rp(s) = Rp(s) (0, N) Rps = Rps (0, N)

Tp(s) = Tp(s) (0, N) Tps = Tps (0, N)

(14.32)

by using the starting values Rp(s) (N − 1, N) = |rN−1,N,p(s) + |2

Tp(s) (N − 1, N) = |tN−1,N,p(s) + |2 ∗



Rp(s) (N − 1, N) = (rN−1,N,p + rN−1,N,s + ) Tp(s) (N − 1, N) = (tN−1,N,p + tN−1,N,s + ) (14.33) As already stated, if the ambient in and ambient out are different, the usual factor (N N cos𝜃 N /N 0 cos𝜃 0 ) must be applied to the final results for T p(s) and T ps . It must be stressed that the same iterative procedure as described in Eqs. (14.30a) and (14.30b) can be used to compute the total reflection and transmission amplitudes in both polarizations for a stack of thin films, noting that we now have Fresnel interfaces and provided that |x|2 is replaced by x and −2Im(𝛽) by i𝛽. 14.3.3

Remarks on the Reflection-Transmission Photoellipsometry Method

The above method for reflection and transmission photoellipsometry has been used successfully in the study of thin-film layers on semitransparent substrates [21–23]. Instead of defining the effective ellipsometric parameters when dealing with transparent thick layers, it is simpler and physically more appropriate to derive such quantities as Rs , Rp , Rps , T s , T p , and T ps , which are smoothly varying, have no singularities, and from which effective ellipsometric parameters could be derived [24]. The approach proposed permits the use of any number of semitransparent thick films (or substrates) in a multi-layer system, eliminating the need for an opaque or non-back-reflecting substrate in conventional reflection ellipsometry. The experimental method uses reflection and transmission over a wide range of incident angles. Such a range is critical in detecting any anisotropy in the samples being studied. In the case of isotropic samples, a wide range of incident angles usually permits a fast and unambiguous determination of optical constants and film thicknesses.

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies Applications of thin films on the basis of their specific optical properties abound. One area of particularly great interest is the use of thin films in energy management and renewable-energy technologies. Various applications include selective solar-absorbing surfaces (surfaces absorbing energy in the solar spectrum and having a low emittance in the blackbody range defined by their temperature) [25], heat mirrors (films having a high transmittance in the visible region and a high reflectance in the infrared) [26], photovoltaic films (films that convert light into electricity) [27], and smart window coatings

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

(films with reversible changing optical properties for energy management) [28]. Within the context of this chapter and the authors’ research interests, our review will concentrate on thin films used for smart window applications and some new developments in a sky radiator film and an optical functional film. Coating materials that can change their optical properties upon the application of a stimulus (light, electrical voltage, or heat) are called chromogenic coatings, and they can be used to control the flow of light and heat into and out of a window system in buildings, vehicles, and aircraft. The energy efficiency of such chromogenic windows with a control capacity based on occupancy, temperature, and solar radiation has been investigated, showing a great potential for energy saving and increased comfort [29].

14.4.1

Electrochromic Thin Films

The most-studied thin films for smart window applications are the electrochromic films, whose optical properties change drastically upon the application of an electric field, from the transparent to a colored state. Inorganic materials exhibiting this optical behavior are metal oxides such as WO3 , MoO3 , Nb2 O5 , V2 O5 , NiOx , and IrOx [30]. It is generally accepted that the electrochromic effect is caused by a double injection of electrons and positive ions in the case of cathodic materials (WO3 , MoO3 , Nb2 O5 , V2 O5 ), and by a double ejection of positive ions and electrons in the case of anodic materials (NiOx , IrOx ). For WO3 , the reversible coloring and bleaching process is governed by xM+ + xe− + WO3−y (transparent) ↔ Mx WO3−y (colored) where x is normally less than 0.5, and y less than 0.03 [31]. When the WO3 films are amorphous, the colored state corresponds to an absorbing state, whereas it is a reflective state if the films are originally in the crystalline state. Depending on the nature of the films used, it is thus possible to modulate either the absorption or the reflectance of the films with a proper double-injection mechanism. A thin-film electrochromic device may have the following configuration: Glass substrate∕TC∕CE∕IC∕WO3 ∕TC where TC stands for “Transparent Conductor,” CE for “Counter-Electrode,” and IC for “Ion Conductor.” The transparent conductors serve as electrodes for applying the electric field needed for performing injection of electrons and ions (originally stored in the CE layer) to color the electrochromic WO3 film. By reversing the electric field, the colored WO3 will switch back to its clear state, thus allowing control of the optical transmission of the device. An all-solid-state device has been proposed by us, using amorphous WO3 , indium tin oxide (ITO) films as electrodes, V2 O5 as counter-electrode for lithium-ion storage, and LiBO2 as an ion conductor [32]. Such a device exhibited integrated solar and visible transmittances of 58% and 65%, respectively, in the bleached state, and corresponding values of 9% and 13% in the colored state when 3 V was applied. Many other types of smart window systems using thin films have been studied by different groups. A recent survey of electrochromic coatings and devices can be found in Reference [33].

413

414

14 Optical Properties of Thin Films

14.4.2

Pure and Metal-Doped VO2 Thermochromic Thin Films

Vanadium dioxide (VO2 ) thin films are well known for thermochromism phenomena found in transition metal oxides [34], exhibiting a phase transition upon heating at approximately 67 ∘ C [35]. A substantial change in conductivity is observed as the film changes from semiconductor to metal behavior accompanied by a change in optical properties. Using specific chemical substitutions for the vanadium cation, this transition temperature can be lowered. The substituted cations should behave as electron donors for vanadium V4+ . The doping effect is the origin of the electrical, magnetic, and optical modifications [34]. Although several studies were previously published, most of them were essentially concerned with near-infrared (NIR) and infrared (IR) transmittance potentialities. Applications include smart windows for thermal regulation of buildings (transmittance contrast) [36–53], adjustable IR for laser applications (reflectance contrast) [54–56], optical storage media [57], uncooled micro-bolometer [58], and optical switching [59]. The transmittance and reflectance of VO2 and metal-doped V1-x Mx O2 (M: W or Mo) thin films prepared by reactive radio frequency (RF) magnetron sputtering in our laboratory [39–43, 45] are summarized below. The fabrication apparatus is shown schematically in Figure 14.3. Water-cooled metal targets (V and M: 99.9%) were used. An RF power of 200 W was applied to a V target and a fraction of it to the M target. For the reduction of the sputtering rate for metal M, a grounded cover made of the same metal element was also provided to reduce the actual target area appropriately. With this apparatus, the x-values were controlled to 0–0.26 for the W element and to 0–0.04 for the Mo element. The target was biased as needed and the substrate temperature controlled between room temperature and 900 ∘ C by a SiC back heater. The oxygen flow rate, defined as the ratio between the reactive O2 gas flow rate and the total gas flow rate, and the substrate temperature were identified to be rather important parameters for obtaining good VO2 thin film. The optimal oxygen rate and the substrate temperature were confirmed to be 2.7% ION GAUGE

THERMOCOUPLE THICKNESS MONITOR

HEATER MAIN VALVE

SUBSTRATE SHUTTER

V

Ar O2

TARGETS

TO TMP

GFC

W

Figure 14.3 Schematic representation of a dual-target radio frequency (RF) magnetron sputtering apparatus used to prepare thermochromic VO2 thin films doped with W. Metal mode reactive sputtering is ensured after removing the surface oxides by the presputtering under the shutter application.

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

and 400–500 ∘ C respectively. If the oxygen flow rate was deviated from the optimal value by 0.1–0.2%, mixed phases of VO2 , V2 O7 , V6 O13 , and V2 O5 were grown. The oxygen rate value was valid even for lower substrate temperatures such as 250–350 ∘ C. The deposition rate of pure VO2 thin film was a function of the oxygen flow rate. It decreased from 19 to 17 nm min−1 as the oxygen flow rate was varied from 0% to 2%, then drastically dropped to about 5 nm min−1 with an oxygen flow rate of 3.5% and reached the steady rate. The observed distribution of the horizontal diameter of the oblate VO2 polycrystallites grown vertically in the columnar structure of the films with a thickness of about 60 nm was 1000–1500 and 200–400 nm for samples fabricated at a substrate temperature of 500 ∘ C, and 100–300 nm for samples prepared at 400 ∘ C. The root-mean-square of the surface roughness of those films was smaller than 10 nm. Figures 14.4a,b show the ambient temperature-dependent hemispherical spectral transmittance and/or hemispherical spectral reflectance of a polycrystalline VO2 thin film 65 nm thick on Pyrex glass, respectively. The transmittance was obtained for normal incidence, and the reflectance was taken at an incident angle of 12∘ . Although the change in the transmittance and the reflectance due to the metal–semiconductor transition in the visible and/or solar spectral ranges was not significant, the one observed in the NIR and IR ranges was large enough to produce an optical-switching behavior. The experimental transmittance and reflectance curves were well reproduced by calculations using complex refractive indices obtained by Tazawa et al. [60] for both pure VO2 and V1-x Wx O2 thin films on a glass substrate. In Table 14.1, the numerical 80

Transmittance (%)

20°C 60°C 60 63°C 65°C 66°C

40

67°C 68°C 70°C 72°C 80°C

20

0

Reflectance (%)

60

80°C 72°C 70°C 68°C 67°C 65°C 63°C 20°C

50 40 30 20 10 0 500

1000

1500 2000 Wavelength (nm)

2500

Figure 14.4 Spectral transmittance as a function of an applied external temperature between 20 and 80 ∘ C in NIR and IR regions (a), and spectral reflectance (b).

415

14 Optical Properties of Thin Films

Table 14.1 Difference in solar transmittance (T sol ), luminous transmittance (T lum ), solar reflectance (Rsol ), and luminous reflectance (Rlum ) of VO2 and V0.986 W0.014 O2 between metal and semiconductor phase.

Thickness (nm)

T sol (%) 20 ∘ C/100 ∘ C

Rsol (%) 20 ∘ C/100 ∘ C

T lum (%) 20 ∘ C/100 ∘ C

Rlum (%) 20 ∘ C/100 ∘ C

VO2

65

36.9/30.1

35.5/26.8

37.6/35.5

32.0/28.5

V1-X WX O2

80

32.7/22.7

29.5/21.9

32.7/29.5

15.1/13.3

optical-switching behavior in the luminous and solar spectral ranges for VO2 and V0.986 W0.014 O2 thin films is summarized. The transition temperature 𝜏 c and its width (Δ𝜏 c ) obtained from Figure 14.4a,b at a wavelength of 2000 nm are 67 ∘ C and 12 ∘ C, respectively. The hysteresis loop width of the spectral curves (Δhys ) obtained for an increasing rate of +2 ∘ C min−1 and for a decreasing rate of −2 ∘ C min−1 is 31 ∘ C. When 𝜏 c , Δ𝜏 c , and Δhys are compared to the corresponding values obtained for epitaxially grown VO2 thin films on a sapphire (110) substrate [42, 45], the first parameter is in agreement while the second, and the final ones are larger by 12 and 23 times, respectively. Consequently, the sharp phase transition is confirmed for epitaxial single crystalline films as expected. The decreasing in the transition temperature 𝜏 c that depended on the doping level of W and Mo elements in atomic % for polycrystalline VO2 films is reproduced in Figure 14.5. The decreasing rates for W and Mo are identified as −23 ∘ C/atm and −11 ∘ C/atm, respectively. Figure 14.5 Transition temperature as a function of doping level in atomic %. W-doped samples exhibit larger decreases in the transition temperature as compared to Mo-doped samples.

W–doped VO2 (–24°C/at, %W) Mo–doped VO2 (–11°C/at, %Mo)

Transition Temperature (°C)

416

0

1

2

3

Doping Level (atomic%) Transition temperature vs. doping levels for the metal doped VO2 films.

4

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

Extensive studies to improve the optical-switching behavior to ensure high luminous transmittance and solar modulation have been performed since the middle of the year 2000. Among those, one of our best results [61] is described as follows. In this study, we focused on finding a reproducible economic solution-processed strategy for fabricating VO2 nanocrystals-SiO2 composite films on a borosilicate glass substrate by using the spin-coating method, with the aim of boosting the performance of two designated aspects. We established a single-step chemical process for the fabrication of the green body of the film. The film was annealed at 450 ∘ C for 1 h in air to ensure good crystallinity. Compared to the pure VO2 film, an improvement of 18.9% (from 29.6% to 48.5%) in the luminous transmittance as well as an increase of 6.0% (from 9.7% to 15.7%) in the solar modulation efficiency was achieved when the molar ratio of Si/V became 0.8. The control of the important molar ratios of Si/V was ensured simply by varying the gas flow of nitrogen. We simulated the optical spectra of the composite films by the effective medium theory. Consequently, the best thermochromic property was confirmed under a filling factor of 0.5. The result was consistent with the experimental results. Moreover, the improvement in chemical stability for the composite film against oxidation has been confirmed. Tungsten (molar ratio W/V = 0.015) was also introduced to reduce the phase transition temperature 𝜏 c to the ambient temperature 34.5 ∘ C, and the hysteresis loop width Δhys is also narrowed down by 3.4 ∘ C. It is physically interesting to ensure the maximum doping level x required for the depression of T c in the V1-x Wx O2 thin film. This was performed by Shibuya et al. [62] for epitaxial thin films on a TiO2 (001) substrate for 0 ≤ x ≤ 0.33. The results were as follows: almost a linear reduction of T c in x ≤ 0.07, non-metal-insulator transition (MIT) due to metallic ground state in 0.07 ≤ x ≤ 0.095, and an almost linear increase in T c for x > 0.095. This peculiar feature between T c and x was caused by strongly correlated electron configurations of 3d transition metal oxides, and the physical origin would be based on a “bandwidth controlled Mott MIT” frame [63–69] in quantum phase transition theories [63–69]. Although it was not the case in thin films, we recently achieved a T c of 250 K for x = 0.02 in V1-x Wx O2 ground nanorods assembled samples [70, 71]. This T c value satisfactorily agreed with the above result of Shibuya et al. 14.4.3

Temperature-Stabilized V1-x Wx O2 Sky Radiator Films

The composition of the atmosphere (H2 O, CO2 , etc.) causes a specific spectral range of emittance for the atmosphere in the spectral range from 8000 to 13 000 nm, termed the atmospheric window [72]. The atmospheric window is characterized by a low radiance and a high transmittance of the atmosphere in the mid-IR region, where the peak of blackbody radiation is matched to the ambient temperature. The low radiance is the origin of the nocturnal radiative cooling observed under clear sky radiation. If the humidity and the cloudiness are low, a blackbody may reach a temperature between 15 and 25 ∘ C below the ambient temperature as a result of radiative cooling [73]. Spectral selective radiating materials (SSRMs) can enhance the radiative cooling effect because of their high emittance in the atmospheric window and their high reflectance outside this window. Consequently, the SSRM can be used to obtain a low temperature in a sky radiator for passive cooling in building technology, with no energy consumption [73]. SiO films [72, 74], Si-based films such as SiOx Ny [74], or multi-layered films [75–78] on high reflective substrates are regarded as the most feasible SSRMs. These Si-based

417

14 Optical Properties of Thin Films

60 Radiative cooling power (W/m2)

418

ideal SiO(0.2) SiO(0.4) SiO(0.6) SiO(0.8) SiO(1.0)

50 ideal

40 30

0.4

SiO(x) + Si 1.0 μm Al

0.6

20 0.8

2010

1.0 0.2

0 0

5

10

15

20

25

30

35

40

Ta-Ts (degree)

Figure 14.6 Radiative cooling power of SSRM with composite films of SiO and Si on Al substrate as a function of T a − T s . The film configuration and the SiO ratio are indicated. The ideal case is when the spectral emittance of the SSRM is unity at wavelengths in the atmospheric window and null for other wavelengths.

films with an adequate thickness on metallic substrates usually show an emission only in the atmospheric window, so that an efficient low surface temperature (Ts) from the ambient temperature (Ta) is realized, particularly at nighttime under clear sky conditions. In Figure 14.6, our results from computational simulations of a composite SiO and Si film 1000 nm thick on an Al substrate [75, 77] are depicted. In the simulation, the spectral reflectance and emittance of the multi-layered films were required and were first calculated by following the procedures given in Section 14.2.2 in this chapter. The radiative cooling power depends on the compositional ratio (x) between SiO and Si, and can be increased by about 20% as compared to pure SiO films (when x = 0.4–0.6). This pushes the cross-point value of the power curve on the (Ta-Ts) axis to the higher side by 1–2 ∘ C. Thermochromic materials can control the energy throughputs of the IR and NIR parts of solar radiation as well as environmental radiation in response to the environmental temperature, as shown in the previous section. As the metallic phase of V1-x Wx O2 film has a high reflectance in the IR region, a film made with this material might be a good candidate for the substrate and/or the second layer of SSRMs for particular applications beyond 𝜏 c . Furthermore, the expected changes of its optical properties at 𝜏 c near or below the ambient temperature T a have the potential to add another function to the SSRM as described below. We proposed a unique and interesting SSRM that consisted of a SiO film 1000 nm thick as the top layer and a V1-x Wx O2 film as the second layer on a black substrate for a sky radiator, as shown in Figure 14.7 [79–81]. The designed SSRM was capable of attaining a stable surface temperature with an appropriate adjustment of the transition temperature from a metal to a semiconductor state through the value of x in V1-x Wx O2 . The temperature-stabilized mechanism of this SSRM film is explained by simulation studies as follows.

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

polyethylene cover SiO film V1–xWxO2 film glass substrate insulating material

Figure 14.7 A model of sky radiator with an SSRM consisting of SiO/V1-x Wx O2 /glass. 100 Radiative cooling power [W/m2]

Figure 14.8 Simulated radiative cooling power vs. SSRM temperature for semiconductor and/or metal phases of V1-x Wx O2 films in the modeled sky radiator.

50

0

–50

–100 27

22 17 Temperature of SSRM

12 °C

The simulation to obtain the radiative cooling power was performed using published refractive indexes for SiO [82] and V1-x Wx O2 [63]. Figure 14.8 shows the calculated radiative cooling power vs. the SSRM temperature under the assumption T a = 27 ∘ C. The solid and dotted lines stand for the low-temperature semiconductor phase and the high-temperature metal phase of V1-x Wx O2 , respectively. As the optical properties of the V1-x Wx O2 film change at the transition temperature 𝜏 c , accordingly the radiative cooling power also changes at 𝜏 c . Thus, the radiative cooling power at temperatures lower than 𝜏 c is represented by a solid line and that at higher temperatures by a dotted line. Figure 14.9 shows the change in the radiative cooling power accompanied by a transition occurring at the SSRM surface temperature T s = 19.5 ∘ C (= 𝜏 c ) as indicated by a thin vertical line. The radiative cooling power is positive (cooling potential) at temperatures higher than T s (= 𝜏 c ), and negative (heating potential) at lower temperatures. Therefore, the surface temperature of the SSRM is almost kept constant at 𝜏 c . Strictly speaking, it is relatively difficult to adjust in an accurate way the transition temperature (𝜏 c ) for the V1-x Wx O2 film to a designated value in the usual experimental cases. This interpretation would be reasonable because of the possibility of adjusting the doping level x to the allowed range, and the small dependency of optical constants of the V1-x Wx O2 film on the x-values [79]. This fact was confirmed by our experiments [46]. One of the promising applications of the new SSRM would be in the design of containers for medicines in remote areas, when a refrigerator is not available. If we can adjust x > 2.5%, it will give 𝜏 c < 7 ∘ C.

419

14 Optical Properties of Thin Films

Figure 14.9 Simulated radiative cooling power vs. SSRM temperature, taking into account the metal–semiconductor transition at 𝜏 c for the V1-x Wx O2 film in the modeled sky radiator.

10 Radiative cooling power [W/m2]

420

5

0

–5

–10 27

20 Temperature of SSRM

19 °C

SSRM has been revived as an attractive material for effective passive cooling and wearable energy sources in the year 2010, and several papers that cover the recent studies, including the use of 2D materials/meta-materials, have been published [83–87]. However, a new novel pathway to realize the principle described here to stabilize the SSRM temperature has not been reported yet. 14.4.4 Optical Functional TiO2 Thin Film for Environmentally Friendly Technologies Since the finding of the photo-induced decomposition of water on TiO2 electrodes by Fujishima and Honda [88] in 1971, TiO2 has become a promising material as a photocatalyst [89] whose function is to decompose and oxidize various organic and/or inorganic chemicals in waste and emission with its strong oxidation activity based on the created superoxide anion radical (O2 − ) and OH radical. Its high chemical stability and inexpensive cost are the other benefits that favor its use as a photocatalyst as well as for photocatalytic microbiocidal effects in killing viruses, bacteria, fungi, algae, and cancer cells [90–92]. This material has the additional great potential for application to dye-sensitized photovoltaic cells [93, 94], energy-efficient windows and/or other optical coatings [95–97], and capacitors in LSI [98]. The wide-bandgap semiconductor TiO2 exists in three different crystalline polymorphs: rutile, anatase, and brookite. Among them, rutile (R-titania) and anatase (A-titania) are the most common and widely used phases in applications. Because the optical properties, such as complex refractive indices (n − ik), complex dielectric constants (𝜀r − i𝜀i ) for a certain range of wavelengths between the ultraviolet (UV) and the NIR, and optical bandgap values Eg , are very important criteria for the selection of films for different applications, we have recently reported the simultaneous growth of epitaxial thin films of both A-titania on SrTiO3 (STO) substrates and R-titania on sapphire substrates, and the growth of polycrystalline A- and/or R-titania thin films on silicon and/or glass by helicon RF magnetron sputtering [99–101]. We also reported the optical properties of polycrystalline and/or epitaxial A- and/or R-titania thin films [102], the optical band gaps Eg as extrapolated by the Tauc plot using the obtained

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

Table 14.2 Refractive index n of rutile and anatase TiO2 thin films at the designated wavelength. Rutile

Anatase

Polycrystalline

Epitaxial

Polycrystalline

Epitaxial

n at 500 nm

2.85

2.84

2.66

2.64

Maximum n at 400–800 nm

3.18 (400)

3.17 (400)

2.90 (400)

2.89 (400)

Maximum n at 250–1600 nm

3.95 (330)

4.14 (320)

3.61 (320)

3.79 (330)

extinction coefficient [102], and the interpretation of the observed wider bandgap values [103]. The complex refractive indices from 0.75 to 5 eV (1653–248 nm wavelength range) were obtained by spectroscopic ellipsometry (SE) (Horiba, Jobin-Yvon, UVISE) employing the F&B formula [104] as corrected by Jellison et al. [105] for the optical function. In the present SE, the reflection mode is employed only under the basic mathematical structure slightly modified from that described in this chapter. Important results are reproduced as follows. The typical refractive index n at wavelength 500 nm, the maximum value of this index from 400 to 800 nm, and the maximum value from 248 to 1653 nm are given in Table 14.2. The values higher than the cited ones in recent articles for thin films [106–108] and bulk materials [108–110] reveal the good crystal quality of the films produced. The real and imaginary parts of the complex dielectric function of the rutile are shown in Figures 14.10a,b respectively, and they are compared to Jellison et al.’s recent results for bulk materials in both o and e polarization states [111]. It is noted that these latter used an optical function similar to ours in SE to eliminate the dependency on the different optical functions used. Below the band edge (3.5 eV), the 𝜀r of epitaxial films exhibits a sharp peak at 3.85 eV that agrees with that at 3.9 eV found by Jellison et al. for bulk materials in both o and e states, while the polycrystalline film shows one broad peak at 3.75 eV and a shoulder at 4.3 eV, which are due to the applied two-term model in the F&B formula. This feature is clearly observed in 𝜀i , where sharp peaks are found at 4.1 eV (Epi-rutile), 4.3 eV (Jellison, bulk rutile, e state), and 4.1 eV (Jellison, bulk rutile, o state), and a shoulder and/or a broad peak at 4.0 and 4.5 eV, respectively, for the polycrystalline film. The shape of the 𝜀i curve directly relates to the combined density of states (DOS) for the interband transition. Structures observed at E > 3.5 eV in epitaxial films represent well the DOS, which has a sharp absorptive transition around 4.0 eV as theoretically predicted by Shelling et al. [112]. A similar comparison for the anatase case is made in Figure 14.11a for 𝜀r and in Figure 14.11b for 𝜀i . Here, we compare our results directly with those of anatase thin films studied by Jellison et al. [111]. Two critical points in 𝜀r are observed as a sharp peak at 3.75 eV and a rather broad one at 4.4 eV for an epitaxial film, and a broad peak at 3.8 eV and a shoulder at 4.42 eV for a polycrystalline film. Those peaks are again due to the applied two-term model in the F&B formula. The values for the epitaxial film agree well with those found by Jellison et al. For 𝜀i , only a single broad peak at 4.0 eV and a feature at 4.8 eV are observed with our films, while a shoulder at 4.0 eV and a broad peak at 4.65 eV are observed with a polycrystalline film. Jellison et al.’s results show a behavior

421

14 Optical Properties of Thin Films

25 Poly-rutile film Epi-rutile film

20

Jell-Bulk-rutile-εre Jell-Bulk-rutile-εro

εr

15 10 5 0 1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

4.0

4.5

5.0

Photon energy (eV) (a)

20 Poly-rutile-film Epi-rutile-film

15

Jell-Bulk-rutile εie Jell-Bulk-rutile εio

εi

422

10

5

0 1.5

2.0

2.5

3.0

3.5

Photon energy (eV) (b)

Figure 14.10 (a) Real part and (b) imaginary part of the complex dielectric function of the rutile TiO2 as compared to Jellison et al.’s bulk results for both o and e polarization states.

similar to that of our epitaxial films. The position and the strength of the peaks of 𝜀i for our thin films in the energy range E > 3.0 eV agree satisfactorily with those given by theoretical calculations by Asahi et al. [113], and they correspond to absorptive transitions from valence bands to dxy crystal-field splits levels of the t 2g orbital of the Ti 3d atomic level in the conduction band. The optical band gap Eg of the polycrystalline and epitaxial TiO2 films is determined using the extinction coefficient k from the Tauc expression [114], which is expanded as (E − Eg ) = {(4π𝜅/𝜆)h𝜈/B}1/2 under the ad hoc assumption that the indirect allowed transition [115] is predominant as the optical transition mode. The parameters used are E, photon energy (h𝜈); B, constant; h𝜈, photon energy; 4π𝜅/𝜆, absorption coefficient at wavelength 𝜆; and 𝜅, extinction coefficient, respectively. The bandgap value Eg is given as the intercept of the photon energy axis with the asymptotic linear line of the right-hand

14.4 Application of Thin Films to Energy Management and Renewable-Energy Technologies

14

Poly-anatase-film Epi-anatase-film

12

Jell-Bulk-anatase εre Jell-Bulk-anatase εro Jell-anatase-film εr

εr

10 8 6 4 2 1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

4.0

4.5

5.0

Photon energy (eV) (a) 12 Poly-anatase-film Epi-anatase-film

εi

10 8

Jell-Bulk-anatase εie

6

Jell-Bulk-anatase εio Jell-anatase-film εi

4 2 0 1.5

2.0

2.5

3.0

3.5

Photon energy (eV) (b)

Figure 14.11 (a) Real part and (b) imaginary part of the complex dielectric function of the anatase TiO2 as compared to Jellison et al.’s bulk results for both o and e polarization states. Table 14.3 Bandgap of the anatase and rutile TiO2 films extrapolated by Tauc plot in comparison to the bulk.

Structure

Polycrystalline rutile film

Polycrystalline anatase film

Epitaxial rutile film

Epitaxial anatase film

Bulk rutile

Bulk anatase

Bandgap Eg (eV)

3.34

3.39

3.37

3.51

3.03

3.20

side of the above equation assumed for a certain range of E. The results are summarized in Table 14.3. It is worthwhile to note that the optical bandgap obtained by a Tauc plot always differs from that parameterized in the F&B formula. All values are larger than those for bulk materials [109, 110] by about 10%. It is well known that the optical properties of intrinsic semiconductors can be changed significantly by the application of an external strain to the non-degenerated state [114]. In order to understand the increase in the bandgap value observed in the presently

423

424

14 Optical Properties of Thin Films

strained TiO2 epitaxial films, the ab initio simulation based on density functional theory (DFT) to obtain the DOS and the energy of the electronic orbital that depends on lattice strains was performed by our group [103]. Consequently, the theoretical correlation between the lattice constants of the thin films and the optical bandgap accounted for the increase with respect to the bulk values both for anatase and rutile as listed in Table 14.3.

14.5 Application of Tunable Thin Films to Phase and Polarization Modulation Materials with variable optical properties exhibit unique and useful characteristics when structured in the form of thin layers. Some materials, when excited by heat, strain, electric fields, or optical pulses, undergo phase transitions, causing their refractive indices to change. Modifying the refractive indices changes the material’s Fresnel coefficients, as described by Eqs. (14.6)–(14.13), which in turns alters the reflectance and transmittance of light at the surface of the material. In a thin film, this effect is magnified by the multiple passes of light inside the layer, and with the phase term of Eq. (14.24b) for the accumulation of additional optical phase shifts and absorption. The end result is thin films with highly tunable optical properties: A beam of light interacting with such films will see its amplitude, phase, and polarization altered in a controllable way by the application of external stimuli. Examples of materials suitable for achieving tunable thin films include transition metal oxides, notably vanadium dioxide (VO2 ) and pentoxide (V2 O5 ), as well as chalcogenides such as GeSbTe [116, 117] and AgInSbTe [118, 119], which can exist in two stable phases. One advantage offered by phase-change materials is their large refractive-index changes during the phase transition. Changes on the order of unity have been measured for both the real and imaginary parts of the refractive index; by comparison, electro-optics and nonlinear optical effects typically yield index changes on the order of Δn ≃ 10−4 or less. Liquid crystals offer much larger anisotropy, on the order of Δn ≃ 0.1, but still only one-tenth of that of phase-change materials. For optical applications, a larger index change makes optical phase modulation possible over much shorter distances. For example, while phase and polarization modulation has been demonstrated in VO2 films as thin as 50 nm, the same effect in electo-optic devices (e.g. Pockels cells) require centimeter lengths, and micrometer lengths in liquid crystals. The main applications of phase-change materials have been laser-writeable optical memories, polarization modulators, and spectral filtering. As discussed in Section 14.4, many applications have been found that are based on modifying the amplitude of light reflected or transmitted through the film, with the goal of filtering out certain parts of the spectrum (e.g. infrared and heat filtering), or modifying the spectrum in some ways to obtain dynamic coloration. But there is another class of phenomena that leave the amplitude of light intact but play on the phase of the electromagnetic field, leading to more subtle effects but with new possibilities for applications. This section explores tunable thin films for optical phase and polarization control applications, and provides a theoretical formalism to study them. Although we use recent research on VO2 films as a case study, we should note that the same methods and principles would also apply to films made with other tunable materials.

14.5 Application of Tunable Thin Films to Phase and Polarization Modulation

Broadly speaking, adjustments on the phase of light leads to two types of effects on light beams: (1) interference effects and (2) polarization effects. Interference effects occur when multiple light beams cancel or strengthen each other according to their relative phase, as in Young’s double slit experiment and Michelson’s interferometer. But it can occur on a single beam as well, when various parts of the same light beam experience different phase shifts. For example, since the shape of a wave front dictates its propagation through space, such as convergence (focusing) or divergence, changing the phase of a wave front can lead to focusing or diffraction of a light beam. On the basis of this principle, a physically flat lens was demonstrated using VO2 thin films [120]. Reflecting off a film less than 100 nm in thickness, a laser beam was focused at various distances by locally modifying the refractive index of VO2 . Similarly, the diffraction pattern of a laser beam passing through the edge of a thin VO2 film was shown to be altered when the material underwent a phase transition [121]. Polarization effects, meanwhile, rely on relative phase shifts between the orthogonal components of the electromagnetic field. For this purpose, it is convenient to work with the s- and p-components of the polarization of incident, reflected, or transmitted light. The relative phase between the p- and s-components is 𝛿 = 𝜙p − 𝜙s

(14.34)

And dictates the type of polarization light will have. For example, the condition for linear polarization is 𝛿 = 𝜋n, while circular polarization requires 𝛿 = 𝜋/2 + 𝜋n, where n is 0 or any positive or negative integer. In the context of modulating the polarization by means of tuning the film’s properties, we are specifically interested in changes in the value of 𝛿 when the material undergoes a phase transition. To this end, if we call 𝛿 o the relative phase of light when the film is in one state (e.g. the insulating, non-activated state of VO2 ) and 𝛿 1 the relative phase when the material is in the other state, the quantity Δ = 𝛿1 − 𝛿 0

(14.35)

describes the polarization state. Of particular interest are two cases: (1) with Δ = ± 𝜋 linearly polarized light may rotate by some angle while remaining linearly polarized, and (2) with Δ = ± 𝜋/2 linear polarization may become circular, or vice versa. The first case is particularly interesting for producing large modulation amplitudes in transmission through polarizers. We should note that the value of Δ is generally not the same for transmission and reflection. In fact, previous studies [122] have shown that reflected light tends to experience larger values of Δ. This is not a problem per se, and is actually an advantage, as working in reflection permits the use of opaque substrates or anisotropic substrates that would have otherwise interfered with polarization. If we know the values of the refractive indices of VO2 , we can calculate the total reflection amplitudes given by Eq. (14.29a) for either s- or p-polarizations, and for the two states of VO2 , namely, rp, 0 , rp, 1 , rs, 0 , and rs, 1 . These quantities are complex and contain the phase of each component relative to incident light. We then obtain the phase modulation term Δ = arg(zr )

(14.36)

where zr =

rp,1 rs,0 rp,0 rs,1

(14.37)

425

14 Optical Properties of Thin Films

Figure 14.12 Real (n, solid curves) and imaginary (k, dashed curves) parts of the refractive index of a VO2 film (85 nm in thickness) deposited on soda lime glass. Blue curves are for room temperature (insulating phase of VO2 ) and red curves for 80 ∘ C (metallic phase).

Refractive index

3

2

1

0

0

0.5

1.0 1.5 Wavelength (μm)

2.0

2.5

The main properties of VO2 films have been discussed in Section 14.4.2. Figure 14.12 shows the refractive indices of a typical VO2 film as measured by transmission and reflection ellipsometry. Unlike transmittance and reflectance (see Figure 14.4), we see that the refractive index changes down to wavelengths below 500 nm, suggesting the possibility of using VO2 for phase applications in the visible spectrum. Since the film’s Fresnel coefficients rp, 0 … rs, 1 are implicit functions of the incidence angle 𝜃 o and film thickness D1 , and since the refractive indices depend on the wavelength, the phase modulation term Δ also depends on all three parameters. In order to better visualize the role of each parameter, Figure 14.13 show Δ calculated for many combinations of incidence angles and film thicknesses at wavelengths λ = 600 nm Incidence Angle (°)

426

λ = 1000 nm 90 80 70 60 50 40 30 20 10 0

90 80 70 60 50 40 30 20 10 0 0

100

200

300

0

Film Thickness (nm)

100

200

300

Film Thickness (nm) λ = 1500 nm 1

90 80 70 60 50 40 30 20 10 0

0.8 0.6 0.4 0.2 0

100

200

300

0

Film Thickness (nm)

Figure 14.13 Change in relative phase Δ between s- and p-polarizations (in reflection) from VO2 films deposited on amorphous silica. The color code is normalized to 𝜋.

14.5 Application of Tunable Thin Films to Phase and Polarization Modulation

75 degrees_25C 75 degrees_85C

90

0.6

Intensity (a.u.)

60

120

0.9

150

30

0.3

0.0 180

0

0.3 210

0.6

330

0.9

240

300 270

800 700

Modulation ratio

600 500 400 300 200 100 0 –100 0

50

100

150

200

250

300

350

400

Polarizer angle (degress)

Figure 14.14 Left: Polarization state of reflected light at 633 nm thin VO2 film on a quartz sample (75 nm thickness). Right: The corresponding contrast ratio after passing through a polarizer.

of 𝜆=500, 1000, and 1500 nm. Calculations assume a substrate of amorphous silica and were calculated with the refractive indices shown in Figure 14.12. For these wavelengths (and others in between), there are always some combinations of incidence angle and film thickness for which Δ may be adjusted from 0 to 𝜋. However, film thicknesses of less than 100 nm are usually sufficient, and incidence angles above 50∘ are required. Lower incidence angles are possible by using metallic substrates, as we will see later. Polarization modulation is therefore possible in reflection on thin VO2 films, as Figure 14.14 demonstrates. Here, the linear polarization of a 633 nm (HeNe laser) beam

427

14 Optical Properties of Thin Films

is rotated by 90∘ with little change in amplitude. The right figure shows the modulation amplitude of this beam passing through a polarizer: it is the ratio of signals when VO2 is activated and not activated. In some cases, ratios of several orders of magnitudes have been measured. One limitation of using VO2 films on dielectric substrates is that sizable values of Δ are obtained only for high incidence angles, as Figure 14.13 showed. This is partly because phase shifts for the p-polarization are largest near Brewster’s angle, namely, in the 65–80∘ range in the case of VO2 . While we should not expect polarization rotation to be possible at near-normal incidence angles, where s- and p-polarizations become indistinguishable, in some applications it would be desirable to lower the angle as much as possible. Metallic substrates can alleviate this problem by making polarization modulation not only possible at lowers incidence angles, but also over a wider spectral range as compared to VO2 films on dielectrics. Moreover, reflectance increases significantly when using a metallic back layer. Note that the term “metallic substrate” does not imply a bulk metallic substrate, but only an optically thick substrate. For most metals, a thickness greater than 100 nm is sufficient. Figure 14.15 shows the phase shifts as calculated in Figure 14.13 but with a gold under layer. We see a broadening of possibilities for achieving Δ = 𝜋. Calculations with other metals, including aluminum, silver, copper, and titanium, showed similar enhancement in performance. Figure 14.16 shows an example of polarization rotation of light at 832 nm by a thin VO2 layer on gold. λ = 600 nm Incidence Angle (°)

428

λ = 1000 nm 90 80 70 60 50 40 30 20 10 0

90 80 70 60 50 40 30 20 10 0 0

100

200

300

0

Film Thickness (nm)

100

200

300

Film Thickness (nm) λ = 1500 nm 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

90 80 70 60 50 40 30 20 10 0 0

100

200

300

Film Thickness (nm)

Figure 14.15 Change in relative phase Δ between s- and p-polarizations (in reflection) from VO2 films deposited on gold films. The color code is normalized to 𝜋.

θo = 40° 100 90 80

Cold Hot

70 60

140

40

150

Intensity (a.u)

0.5

170

10 0

0.0 180 0.5 1.0

190

350

200

2.0 2.5 3.0

1.0 0.5

330

210 220 230

320 310 240 250

260 270 280

290

300

70 60 50

140

0.5 1.0

40

2.0

1.5

20

160 170

10 0 350

190

340

200

330 320

220 240

250

260 270 280

290

300

50 40 30

160

20

170

10

0.0 180

0

0.5

350

190

1.0 1.5 2.0

310

230

Cold Hot 60

150

1.0 0.5

90 80 110 100 70

130 140

2.0 30

210 1.5

120

2.5

0.0 180

340

1.5

100 90 80

150

20

160

110

130

1.5 30

1.5 1.0

120

2.0

50

Intensity (a.u)

2.5 2.0

110

130

θo = 60° Cold Hot

Intensity (a.u)

120

3.0

θo = 50°

2.5

Figure 14.16 Polarization state of reflected light at 832 nm on a 72-nm-thick VO2 film on a 50 nm gold film sample.

340

200 210

330 320 310

220 230 240

250

260 270 280

290

300

430

14 Optical Properties of Thin Films

14.6 Conclusions Basic notions on the optical properties of thin films are given in this chapter, introducing the usual tools for treating single layers as well as multiple-layer structures. A section is devoted to presenting a practical method for reflection-transmission photoellipsometry that permits the analysis of samples consisting of thin films combined with semitransparent thick layers or substrates in the form of multi-layer structures. This method has the advantage of being nondestructive and can be used, for example, in studying actual samples of energy-efficient coatings for windows, monitoring useful phenomena such as aging effects. We then illustrate the use of thin films in two domains of interest. The first domain is concerned with applications in energy management and renewable-energy technologies with films that can change their optical properties under certain excitation conditions. The remarkable phase transition materials involved can be also very important in optics and photonics, and the second domain of interest presented here is the potential application of vanadium dioxide films as tunable thin films for phase and polarization modulation.

References 1 Heavens, O.S. (1955). Optical Properties of Thin Solid Films. London: Butterworths

Scientific Publications; Dover, Toronto, 1991. 2 Born, M. and Wolf, E. (1970). Principles of Optics. Oxford: Pergamon. 3 Azzam, R.M.A. and Bashara, N.M. (1977/1984). Ellipsometry and Polarized Light.

Amsterdam: North Holland Publishers. 4 Palik, E.D. (ed.) (1985). Handbook of Optical Constants of Solids. Orlando:

Academic Press. 5 Klein, M.V. and Furtak, T.E. (1986). Optics. New York: Wiley. 6 Ward, L. (1988). The Optical Constants of Bulk Materials and Films. Bristol and

Philadelphia: Adams Hilger. 7 Bass, M. (ed.) (1995). Handbook of Optics, vol. 1. New York: McGraw-Hill. 8 Abeles, F. (1950). Ann. Phys. 5: 595. 9 Hayfield, P.C.S. and White, G.W.T. (1964). Ellipsometry in the Measurement of

10 11 12 13 14 15 16

Surfaces and Thin Films (eds. E. Passaglia, R.R. Stromberg and J. Kruger). Washington DC: National Burean of Standards Miscellaneous Publication 256, U.S. Govt. Printing Office. Schubert, M., Rheinlander, B., Johs, B., and Woollam, J.A. (1997). Proc. SPIE Int. Soc. Opt. Eng. 3094: 255. Alonso, M.I. and Garriga, M. (2004). Thin Solid Films 455–456: 124. Zhokhavets, U., Goldhahn, R., Gobsch, G. et al. (2003). Thin Solid Films 444: 215. Losurdo, M. (2004). Thin Solid Films 455–456: 301. Podraza, N.J., Chen, C., An, I. et al. (2004). Thin Solid Films 455–456: 571. Truong, V.-V., Belley, R., Bader, G., and Hache, A. (2003). Thin Solid Films 212–213: 140. Hodgkinson, I.J., Horowitz, F., Macleod, H.A. et al. (1985). J. Opt. Soc. Am. A 2: 1693.

References

17 Azzam, R.M.A. (1991). Selected Papers on Ellipsometry, SPIE Milestone Series, vol.

MS27. Bellingham, WA: SPIE. 18 Tompkins, H.G. and McGahan, W.A. (1999). Spectroscopic Ellipsometry and Reflec-

tometry, a User’s Guide. New York: Wiley. 19 Theeten, J.B. and Aspnes, D.E. (1981). Annu. Rev. Mater. Sci. 11: 97. 20 Rivory, J. (1995). Thin Films for Optical Systems, Optical Engineering Series No. 49

(ed. F. Flory). New York: Marcel Dekker. 21 Bader, G., Ashrit, P.V., Girouard, F.E., and Truong, V.-V. (1995). Appl. Opt. 34: 1684. 22 Ashrit, P.V., Bader, G., Badilescu, S. et al. (1993). J. Appl. Phys. 74: 602. 23 Bader, G., Ashrit, P.V., and Truong, V.-V. (1995). Proc. SPIE Int. Soc. Opt. Eng.

2531: 70. 24 Zettler, J.-T. and Schrottke, L. (1991). Phys. Status Solidi B 163: K69. 25 (a) Othonos, A., Nestoros, M., Palmerio, D. et al. (1998). Sol. Energy Mater. Sol.

26

27 28

29

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45

Cells 51: 171; (b) Joerger, R., Gampp, R., Heinzel, A. et al. (1998). Sol. Energy Mater. Sol. Cells 54: 351. (a) Tazawa, M., Okada, M., Yoshimura, K., and Ikezawa, S. (2004). Sol. Energy Mater. Sol. Cells 84: 159; (b) Naganuma, T. and Kagawa, Y. (2004). Acta Mater. 52: 5645. (a) Jager-Waldau, A. (2004). Sol. Energy 77: 667; (b) Nelson, J. (2002). Curr. Opin. Solid State Mater. Sci. 6: 87. Lampert, C.M. and Granqvist, C.G. (eds.) (1990). Large Area Chromogenics: Materials and Devices for Transmittance Control, vol. IS4. Bellingham: SPIE Institute Series. (a) See, for example,Karlsson, J. (2001). Control system and energy saving potential for switchable windows. In: Proceedings of the 7th International IBPSA Conference. Rio de Oaneiro, Brazil; (b) Azens, A. and Granqvist, C.G. (2003). J. Solid State Electrochem. 7: 64. Granqvist, C.G. (1995). Handbook of Inorganic Electrochromic Materials. Amsterdam: Elsevier. Svensson, J.S.E.M. and Granqvist, C.G. (1984). Appl. Phys. Lett. 45: 828. Ashrit, P.V., Benaissa, K., Bader, G. et al. (1992). Proc. SPIE Int. Soc. Opt. Eng. 1728: 232. Granqvist, C.G., Avendano, E., and Azens, A. (2003). Thin Solid Films 442: 201. (a) Addler, D. (1968). Solid State Phys. (Academic Press, New York) 21: 1–132. (b) Addler, D. (1968). Rev. Mod. Phys. 40: 714. Goodenough, J.B. (1971). J. Solid State Chem. 3: 490. Granqvist, C.G. (1990). Thin Solid Films 193/194: 730. Rakotoniaina, J.C., Morkrani-Tamelline, R., Gavarri, J.R. et al. (1993). J. Solid State Chem. 103: 81–94. Valmalette, J.C. and Gavarri, J.R. (1994). Sol. Energy Mater. Sol. Cells 33: 135–144. Jin, P. and Tanemura, S. (1994). Jpn. J. Appl. Phys. 33: 1478–1483. Jin, P. and Tanemura, S. (1995). Jpn. J. Appl. Phys. 34: 2459–2460. Jin, P. and Tanemura, S. (1996). Thin Solid Films 281/282: 239–242. Jin, P. and Tanemura, S. (1997). J. Vac. Sci. Technol., A 15: 113–117. Jin, P. and Tanemura, S. (1998). Thin Solid Films 324: 151–158. Valmalette, J.C. and Gavarri, J.R. (1998). Mater. Sci. Eng. B 54: 168–173. Jin, P. and Tanemura, S. (1999). J. Vac. Sci. Technol., A 17: 1817–1821.

431

432

14 Optical Properties of Thin Films

46 Tazawa, M., Jin, P., Miki, T. et al. (2000). Thin Solid Films 375: 100–103. 47 Wu, Z.P., Miyashita, A., and Yamamoto, S. (2000). Mater. Constr. 50 (258): 5–10. 48 Burkhard, W., Christmann, T., Franke, S. et al. (2002). Thin Solid Films 402:

226–231. 49 Lee, M.H. (2002). Sol. Energy Mater. Sol. Cells 71: 537–540. 50 Hanlon, T.J., walker, R.E., Coath, J.A., and Richardson, M.A. (2002). Thin Solid

Films 405: 234–237. 51 Hanlon, T.J., Coath, J.A., and Richardson, M.A. (2003). Thin Solid Films 436:

269–272. 52 Guinneton, F., Sauques, L., Valmalette, J.C. et al. (2004). Thin Solid Films 446:

287–295. 53 Garry, G., Durand, O., and Lordereau, A. (2004). Thin Solid Films 453: 427–430. 54 Konovalova, O.P., Sidrof, A.I., and Shgnanov, I.I. (1995). J. Opt. Technol. 62 (1):

41–43. 55 Sidrov, A.I. and Sosnov, E.N. (1999). Proc. SPIE 3611: 323–330. 56 Mikheeva, O.P. and Sidnov, A.I. (2001). J. Opt. Technol. 68 (4): 278–281. 57 Gal’Prin, V.L., Khakhaev, I.A., Chudovskii, F.A., and Shadrin, E.B. (1996). Proc. SPIE

2969: 270–273. 58 Chen, C., Yi, X., Zhao, X., and Xiong, B. (2001). Sens. Actuators, A 90: 212–214. 59 Becker, M.F., Buckman, A.B., Walser, R.M. et al. (1994). Appl. Phys. Lett. 65: 60 61 62 63 64 65 66 67 68 69 70 71 72

73 74 75

76

1507–1509. Tazawa, M., Jin, P., and Tanemura, S. (1998). Appl. Opt. 37: 1858–1861. Zhao, L.L., Miao, L., Chen, R. et al. (2014). Sci. Rep. 4: 7000. Shibuya, K., Kawasaki, M., and Tokura, Y. (2010). Appl. Phys. Lett. 96: 022102. Mott, N.F. (1990). Metal-Insulator Transition, 2e. London: Taylor & Francis, pp. 123–144 & pp. 185–188. Gebbard, F. (1997). The Mott Metal-Insulator Transition. Berlin, (vol. 137 of “Springer Tracts in Modern Physics”): Springer. Fujimori, A. (1992). J. Phys. Chem. Solids 53: 1595–1602. Fujimori, A., Boucquet, A.E., Saitoh, T., and Mizokawa, T. (1993). J. Electron. Spectrosc. Relat. Phenom. 62: 141–152. Kivelson, S.A., Fradakin, E., and Emery, V.J. (1998). Nature 393: 550–555. Imada, M., Fujimori, A., and Tokura, Y. (1998). Rev. Mod. Phys. 70: 1047–1250. Tokura, Y. and Nagaosa, N. (2000). Science 288 (5465): 462–468. Chen, R., Miao, L., Liang, C.H. et al. (2015). J. Mater. Chem. A 3: 3726–3738. Chen, R., Miao, L., Zhou, J.-H. et al. (2015). Sci. Rep. 5: 14087–14098. Granqvist, C.G. and Erikson, T.S. (1991). Materials for radiative cooling to low temperature. In: Material Science for Solar Energy Conversion Systems (ed. C.G. Granqvist), 168–203. Oxford: Pergamon Press. Martin, M. (1989). Radiative cooling. In: Passive Cooling (ed. J. Cook), 138–196. Cambridge, MA: MIT Press. Granqvist, C.G. and Hjortsberg, A. (1981). J. Appl. Phys. 52: 4205–4220. Tazawa, M., Yoshimura, K., Miki, T., and Tanemura, S. (1993). A computational design of SiO based selectively radiative film. In: Proceedings of the ISES Congress, vol. 2, 333–338. Budapest. Eriksson, T.S., Jiang, S.J., and Granqvist, C.G. (1985). Solar Energy Mater. 12: 319–325.

References

77 Tazawa, M., Jin, P., Tai, Y. et al. (1994). Proc. SPIE 2255: 149–159. 78 Tazawa, M., Miki, T., Jin, P., and Tanemura, S. (1994). Trans. Mater. Res. Soc. Jpn. 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95

96 97 98 99

100 101 102 103 104 105 106 107

18A: 525–528. Tazawa, M., Jin, P., and Tanemura, S. (1995). Proc. SPIE 2531: 326–332. Tazawa, M., Jin, P., and Tanemura, S. (1996). Thin Solid Films 281/282: 232–234. Tazawa, M., Jin, P., Yoshimura, K. et al. (1998). Sol. Energy 64: 3–7. Philipp, H.R. (1985). Silicon monooxide. In: Handbook of Optical Constants of Solids (ed. E.D. Palik), 765–769. New York: Academic Press. Raman, A.P., Anoma, M.A., Zhu, L. et al. (2014). Nature 515: 540–544. Hossain, M.M., Jia, B., and Gu, M. (2015). Adv. Opt. Mater. https://doi.org/10.1002/ adom.20150019. Zhu, L., Raman, A.P., and Fan, S. (2015). Proc. Natl. Acad. Sci. U.S.A. https://doi .org/10.1073/pnas.1509453112. Hossain, M.M. and Gu, M. (2015). Adv. Sci. https://doi.org/10.1002/advs. 201500360. Yi, F., Ren, H., Shan, J. et al. (2018). R. Soc. Chem. https://doi.org/10.1039/ c7cs00849. Fujishima, A. and Honda, K. (1972). Nature 238: 37–38. Fujishima, A. and Honda, K. (1971). Bull. Chem. Soc. Jpn. 44: 1148–1150. Matunaga, T., Tomoda, R., Nakajima, T., and Wake, H. (1985). FEMS Microbiol. Lett. 29: 211–214. Huang, Z., Maness, P.C., Blake, D.M. et al. (2000). J. Photochem. Photobiol., A 130: 163–170. For an example,Miao, L., Tanemura, S., Kondo, Y. et al. (2004). Appl. Surf. Sci. 238: 125–131. Regan, B.O.’. and Gratzel, M. (1991). Nature 353: 737–740. Hagfelt, A. and Gratzel, M. (1995). Chem. Rev. 95: 49–68. Granqvist, C.G. (1991). Energy efficient windows: present and forthcoming technology. In: Materials Science for Solar Energy Conversion Systems (ed. C.G. Granqvist), 106–167. Oxford: Pergamon Press. Brady, G.S. (1971). Materials Handbook, 10e. NewYork: McGraw-Hill. Jin, P., Miao, L., Tanemura, S. et al. (2003). Appl. Surf. Sci. 212–213: 775–781. Lee, Y.H. (1998). Vacuum 51: 503–509. Miao, L., Jin, P., Kaneko, K., and Tanemura, S. (2002). Sputter deposition of polycrystalline and epitaxial TiO2 films with anatase and rutile structures. In: Proceedings of the 8th IUMRS International Conference on Electronic Materials: Advanced Nanomaterials and Nanodevices (eds. H.J. Gao, H. Fuchs and D.M. Chen), 943–963. Xiang, China: IOP. Miao, L., Tanemura, S., Jin, P. et al. (2003). J. Cryst. Growth 254: 100–106. Miao, L., Jin, P., Kaneko, K. et al. (2003). Appl. Surf. Sci. 212/213: 255–263. Tanemura, S., Miao, L., Jin, P. et al. (2003). Appl. Surf. Sci. 212/213: 654–660. Wunderlich, W., Miao, L., Tanemura, S. et al. (2004). Int. J. Nanosci. 3: 439–445. Forouhi, A.R. and Bloomer, I. (1988). Phys. Rev. B 38: 1865–1874. Jellison, G.E. Jr., and Modie, F.A. (1996). Appl. Phys. Lett. 69: 371–373. Ting, C.C. and Chen, S.Y. (2000). J. Appl. Phys. 88: 4628–4633. Suhail, M.H., Rao, G.M., and Mohan, S. (1992). J. Appl. Phys. 71: 1421–1427.

433

434

14 Optical Properties of Thin Films

108 Viseu, T.M.R., Almeida, B., Stchakovsky, M. et al. (2001). Thin Solid Films 401:

216–224. 109 Cronemeyer, D.C. (1952). Phys. Rev. B 87: 876–886. 110 Tang, H., Levy, F., Berger, H., and Schmid, P..E. (1995). Phys. Rev. B 52: 7771–7774. 111 Jellison, G.E. Jr.,, Boatner, L.A., Budai, J.D. et al. (2003). J. Appl. Phys. 93:

9537–9541. 112 Shelling, P.K., Yu, N., and Halley, J.W. (1998). Phys. Rev. B 58: 1279–1293. 113 Asahi, R., Taga, Y., Mannstadt, W., and Freeman, A.J. (2000). Phys. Rev. B 61:

7459–7465. 114 Frouhi, A.R. and Bloomer, I. (1991). Calculation of optical constants, n and k, in

115 116 117 118 119 120 121 122

the interband region. In: Optical Properties of Solids II (ed. E.D. Palik), 163–164. New York: Academic Press. Daude, N., Gout, C., and Jouanin, C. (1977). Phys. Rev. B 15: 3229–3235. Cao, T., Wei, C., Simpson, R.E. et al. (2014). Sci. Rep. 4: 3955. Hosseini, P., Wright, C.D., and Bhaskaran, H. (2014). Nature 511: 206. Jiao, X., Wei, J., Gan, F., and Xiao, M. (2009). Appl. Phys. A 94: 627. Matsunaga, T., Akola, J., Kohara, S. et al. (2011). Nat. Mater. 10: 129. Son, T.V., Ba, C.O.F., Vallée, R., and Haché, A. (2014). Appl. Phys. Lett. 105: 231120. Son, T.V., Zongo, K., Ba, C. et al. (2014). Opt. Commun. 320: 151. Cormier, P., Son, T.V., Thibodeau, J. et al. (2017). Opt. Commun. 382: 80.

435

15 Optical Characterization of Materials by Spectroscopic Ellipsometry J. Mistrík Center of Materials and Nanotechnologies, Faculty of Chemical Technology, University of Pardubice, Pardubice 530 02, Czech Republic

CHAPTER MENU Introduction, 435 Notions of Light Polarization, 436 Measureable Quantities, 438 Instrumentation, 441 Single Interface, 442 Single Layer, 448 Multilayer, 454 Linear Grating, 458 Conclusions, 462 References, 463

15.1 Introduction One of the fundamental properties of light is its polarization. Historically, the first scientific presentation of a phenomenon related to the light polarization is credited to Erasmus Bartholin, who in 1669 studied double refraction of light in a crystal of Iceland spar. Since then, within three centuries, many great scientists have contributed to the description, understanding, and application of light polarization (for a detailed historical survey, refer, for example, to [1–3]). At the end of the nineteenth century, Paul Drude [4–6] extensively studied the change in the polarization upon light reflection from a sample surface and its correlation with sample optical properties. This technique was later named ellipsometry, since elliptic polarization is the most general polarization state of the light. Drude also re-derived Fresnel’s formulas from Maxwell’s equations and thus laid the theoretical foundation of ellipsometry. Furthermore, he constructed the very first ellipsometer – an instrument capable of measuring the polarization change in reflected light. Therefore, it is no wonder that Drude is known as the father of ellipsometry. During the last century, numerous researchers significantly improved both the theoretical formalism and instrumentation of ellipsometry. Simultaneously, materials

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

436

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

databases with accurately determined optical constants were considerably expanded. This effort was particularly triggered by the needs of the semiconductor industry for the production of high-quality microelectronics. The increased performance of personal computers, in turn, made possible further improvements in numerical treatment of ellipsometry data and automation of fast spectroscopic ellipsometers. Presently, ellipsometry is a highly accurate and precise surface characterization tool that is routinely used for determination of bulk, single, and multilayer optical and structural parameters. Moreover, it is worth noting that ellipsometry recently gave evidence of its potentiality also in the characterization of modern nanostructured materials such as gratings, photonic crystals or materials for plasmonics (see [7] and the references therein). This opens promising perspectives for further development of ellipsometry in the future. The aim of the present chapter is to provide a brief introduction to ellipsometry starting with an overview of light polarization, definition of ellipsometric measurable quantities and instrumentation. Subsequent sections cover the applications of ellipsometry to “standard” systems as bulks, single layers, and multilayers. The key theoretical formulas and approaches are supported by selected case studies performed on real samples. The final section deals with ellipsometry (scatterometry) characterization of a linear grating. The advantages and disadvantages together with the complementarity of ellipsometry with respect to conventional techniques are continuously mentioned to underscore the real value of the method.

15.2 Notions of Light Polarization Within the framework of electromagnetic theory, light can be described as a transverse electromagnetic wave. The spatial dependences of electric and magnetic field vectors, which are transverse to each other and are perpendicular to the direction of propagation, are important contributors to light polarization. Several approaches have been developed to describe different polarization states (see, e.g. [1]). Here, we briefly introduce one of them (based on the parameters Ψ and Δ) that is often used in ellipsometry. To keep the formalism (and figures) as simple as possible, we assume a plane monochromatic wave propagating along the z-axes. We usually select the electric field of an electromagnetic wave to represent light behavior. The main reason for this is that the magnetic field vector can always be calculated from the electric field using Maxwell’s equations [2]; also, in optical frequencies, the magnetic field is often less important than the electric field when considering light interaction with the matter. Besides unpolarized light (which is treated in detail, for example, in [3]), the most general case of the spatial dependence of the electric field vector is schematically presented in Figure 15.1a. In a plane perpendicular to the z-axis (let us call it the observation plane), an end point of the electric field vector traces an ellipse as the wave propagates, and therefore we call this electromagnetic wave elliptically polarized. The ellipse of polarization can be described by a pair of parameters Ψ and Δ. For this reason, it is convenient to express the electric field in the wave in terms of its orthogonal components Ex and Ey : E(z, t) = Ex ix + Ey iy

(15.1)

15.2 Notions of Light Polarization

Ex Ex

observation plane

Ex

E0x

z Ey Ey

Ex

z

Δ

E0y

ψ

E0x

E0y

Ey z

Ey (a)

(b)

(c)

Figure 15.1 (a) Spatial dependence of an electric field in an elliptically polarized wave. (b) Its orthogonal components and their relative phase shift as Δ parameter. (c) Ellipse of polarization in the observation plane together with indication of the Ψ parameter.

where ix and iy are base vectors of x and y axes with unit amplitude. The spatial and temporal dependences of each component then take the forms Ex (z, t) = E0x ei(𝜔t−kz+𝛿x )

(15.2)

Ey (z, t) = E0y ei(𝜔t−kz+𝛿y )

(15.3)

where E0x and E0y are the real amplitudes of the x and y components, respectively, of the electric field vector, 𝜔 is the wave frequency, and k is the wave number. The absolute phases of the x and y components are denoted as 𝛿 x and 𝛿 y , respectively. Consequently, for a fixed point on the z-axis (where the observation plane is located), the time dependence of the electric field vector E(t) can be imagined as a superposition of the orthogonal electric vibrations Ex (t) and Ey (t) mutually shifted in phase by (𝛿 y –𝛿 x ) and with amplitudes E0x and E0y . It can be further shown [8] that in the observation plane, Eqs. (15.2) and (15.3) can be reorganized as the equation of an ellipse, ( ) ) ( Ey 2 cos(𝛿y − 𝛿x ) Ex 2 + −2 Ex Ey = sin2 (𝛿y − 𝛿x ) (15.4) E0x E0y E0x E0y hence proving our assumption of elliptical polarization being the most general polarization state. Linear (𝛿 y − 𝛿 x = 0, ± 𝜋, ± 2𝜋, …) and circular (E0x = E0y and 𝛿y − 𝛿x = ± 𝜋2 , ± 3𝜋 , …) polarizations are its special cases. 2 In ellipsometry, only the shape and orientation (with respect to the system of coordinates) of the ellipse are important. The light intensity I ∼ (E0x 2 + E0y 2 ) , which is related to the size of the ellipse, is usually normalized to unity. Therefore, only two parameters are sufficient to uniquely identify the polarization state: the relative phase shift of the components and the ratio of their amplitudes E0x and E0y . The two parameters Ψ and Δ are then defined as E (15.5) tan Ψ = 0x E0y Δ = 𝛿y − 𝛿 x

(15.6)

437

438

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

(ψ,Δ) = (45°,±90°)

Figure 15.2 Linear (pink), elliptical (blue), and circular (green) polarizations together with their corresponding parameters.

Ex

1

(ψ,Δ) = (45°,±45°) 0.5

(ψ,Δ) = (45°,0°)

Ey

0 –1

–0.5

0

0.5

1

–0.5

–1

with their geometrical meaning as indicated in Figure 15.1b,c. It is noteworthy that the angle Ψ should not be confused with the angle between the ellipse main semi-axis and the x-axis, which is also known as the azimuth. Selected cases of light polarization with the corresponding Ψ and Δ parameters are shown in Figure 15.2. Among these the linear and circular polarizations are special cases of the general elliptical polarization state. Note that the sign of the Δ parameter determines the handedness (for example, left-handed or right-handed circular polarization). In fact, it is not only the spatial dependence of E but also its temporal dependence in terms of its handedness that play important roles in light interaction with a medium. It is evident that the values of Ψ and Δ are dependent on the orientation of the base vectors ix and iy . To emphasize this dependence, we formally, wherever it is necessary, mark pairs of polarization parameters as (Ψ, Δ)xy . As all useful information on the polarization state lies in its component amplitudes and phases, it is convenient to combine ̂ = E0 ei𝛿 . In our case, we obtain for them into a single quantity: the complex amplitude E the two components Êx = E0x ei𝛿x

(15.7)

Êy = E0y ei𝛿y

(15.8)

Any change in the polarization is then solely expressed by these complex amplitudes of the wave.

15.3 Measureable Quantities Ellipsometry measures the polarization change in light induced by its non-normal reflection from a sample surface. This change is by definition expressed by the ratio of the reflection coefficients. For oblique incidence, we distinguish between the reflection coefficients r s and r p for s- and p- waves, respectively. The s-wave is linearly polarized with the vector of the electric field E perpendicular to the plane of incidence, whereas the p- wave is linearly polarized with the vector E parallel to the plane of incidence. When the s- and p- waves are reflected from an optically isotropic sample, they remain

15.3 Measureable Quantities

E is = E i0seiδisei(ωt–ki r)

E rs = E r0seiδrsei(ωt–kr r)

Eˆ i0s

ki

E i0s

kr

E r0s

θi

s

Eˆ r0s

s p p

E ip = E i0peiδipei(ωt–ki r)

E rp = E r0peiδrpei(ωt–kr r)

Eˆ i0p

ki

kr

Eˆ r0p

θi

s

E i0p p

s

E r0p p

Figure 15.3 S- and p- polarized waves reflected from an optically isotropic sample.

s- and p-polarized, respectively, but their amplitude and phase are changed due to light interaction with the sample (cf. Figure 15.3). These changes are expressed by the reflection coefficients, which are defined by the complex amplitudes of the reflected and incident waves at the same point on the sample surface: r ̂r Er𝟎p ei𝜹p E p = (15.9) rp = i i𝜹ip ̂i E e E 𝟎p p rs =

r ̂r Er ei𝜹s E s = 𝟎s i Ei𝟎s ei𝜹s Êi

(15.10)

s

As mentioned above, ellipsometry measures the ratio of the reflection coefficients r p and r s . This ratio can be written in terms of the angles Ψ and Δ as rp = tan ΨeiΔ (15.11) rs The relation between the ellipsometric angles Ψ and Δ (defined by Eq. (15.11)) and the parameters (Ψ, Δ)xy of the polarization ellipse introduced in Section 15.2 by Eqs. (15.5) and (15.6) is worth discussing. Consider an incident light with equal components of s- and p- waves that are in phase. In other words, the incident light is linearly polarized with the vector of the electric field E oscillating in the direction ±45∘ out of the incidence plane (cf. Figure 15.4). When light gets reflected from the sample surface, the s- and p- components undergo changes in their amplitudes and phases, and as a result, the reflected wave becomes elliptically polarized, in general. In this case, as

439

440

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

ki

kr

θi

E r0s Δ

E r0p

s s p p

ki

kr

θi s s p

ψ

p

Figure 15.4 Definition of the ellipsometric parameters Ψ and Δ.

shown below, the polarization parameters (Ψ, Δ)xy of the reflected wave are identical with the ellipsometric parameters Ψ and Δ, provided that the x- and y- axes coincide with the s- and p- directions, respectively. Using Eqs. (15.9) and (15.10), we get r

Er𝟎p ei𝜹p rp rs

i

=

Ei𝟎p ei𝜹p i𝜹rs

Er𝟎s e i Ei𝟎s ei𝜹s

=

Er𝟎p Er𝟎s

r

r

ei(𝜹p −𝜹s ) = tan ΨeiΔ

(15.12)

Accordingly, Eq. (15.12) determines the ellipsometric relative changes in amplitudes and phases of the s- and p- waves in their reflection from the sample surface. It is noteworthy that the phase change of the reflected p- or s- wave is in general difficult to measure independently, but its relative change 𝜹p − 𝜹s is accessible when polarization measurements are considered in ellipsometry. Therefore, ellipsometry belongs to the group of phase-sensitive techniques, which is rather sensitive to surface properties. On the other hand, photometric measurements such as, for example, reflectance measurement (which is usually easy to perform), access only the absolute value of the reflection coefficient as Rs,p = |r s,p |𝟐

(15.13)

but one cannot measure its phase. Reflection coefficients can be theoretically calculated for various samples, including, for example, bulks, single and multilayers, and linear gratings or photonic crystals. It is evident that both the sample inner structure to the

15.4 Instrumentation

depth that light reaches by its penetration under the surface and the optical properties of materials that are involved in the interaction influence the reflection coefficients. Therefore, experimentally determined ellipsometric parameters can provide information on sample geometry and its optical constants in the spectral range under consideration. In fact, this is the general goal of ellipsometry characterization. So far we have discussed optically isotropic samples, where the s- and p- polarizations are proper modes and do not mutually interchange when interacting with a sample. However, the situation can be more complex when optically anisotropic samples are to be characterized. In this case, not only the r s and r p reflection coefficients but also r sp and r ps (that represent mixing of s- and p- waves) have to be considered when describing the change in polarization. At the same time, more sophisticated calculations and selected experimental configurations are required. This approach is called general ellipsometry and is described, for example, in the following references [1, 9, 10].

15.4 Instrumentation Various aspects of instrumentation in ellipsometry are discussed in detail in monographs and handbooks (see, e.g. [11–13]). Historically, the first developed ellipsometers were null ellipsometers, where light first propagates through a polarizer and compensator and then gets reflected from the sample toward the analyzer and detector as schematically shown in Figure 15.5. In this experimental configuration, the orientation of the polarizer and compensator is adjusted in such a way that light reflected from the sample is linearly polarized. The analyzer is then rotated to the position where the light intensity on the detector gets extinguished or “nulled.” The schematic experimental setup of null ellipsometry is shown in Figure 15.5, where the polarizer and analyzer can be rotated. The compensator is a quarter-wave plate with fast and slow axes oriented ±45∘ with respect to the plane of incidence. It can be shown that when the light wave transmitted through the polarizer and compensator is incident on the sample surface, it has both s- and p- components of equal amplitudes but mutually shifted in phase by an angle 2P (P being the angle of polarizer, indicated in red in Figure 15.5). Hence, when the phase shift of the s- and p- components of the wave incident on the sample is compensated by the reflection, then the reflected wave is linearly polarized and can be extinguished by the analyzer (A being the angle of analyzer). From the positions of the polarizer and the analyzer, the ellipsometric angles Ψ and Δ can be calculated by [14] Ψ=A

(15.14)

Δ = 2(P − 45∘ )

(15.15)

This configuration in Figure 15.5, if performed in four equivalent quadrants, is very accurate, due to the compensation of systematic errors. However, even when automated, this approach is relatively slow, and measurements are very time consuming. In order to speed up the measurements, ellipsometers with rotating analyzer, polarizer, or compensator were developed. In these systems, either the analyzer, polarizer, or compensator is continuously rotated at a constant angular velocity (typically about 10–100 Hz) in order to modulate the light polarization state. Alternatively, one can also

441

442

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

sample compensator

ki

s s P

45°

p

kr s

θi A p

p analyzer polarizer

Figure 15.5 Schematic experimental setup of null ellipsometry.

use a photoelastic modulator, which generates a periodically changing light intensity in the detector. Fourier analysis of the corresponding detector signal then provides the ellipsometric values Ψ and Δ. Such systems can ensure high-speed and accurate measurements. Nevertheless, it is noteworthy that each of the above-mentioned ellipsometers has both strengths and weaknesses depending on its specific configuration. Further significant improvements in ellipsometers have extended the concepts developed for measurements at a single wavelength to measurements at multiple wavelengths. Spectroscopic ellipsometers capable of measuring the ellipsometric angles Ψ and Δ as a function of the wavelength have added another dimension to the analysis, permitting more reliable determination of material and structural parameters. Presently, a broad spectral range spanning the terahertz (THz), far infrared (FIR), mid infrared (MIR), near infrared (NIR), visible (VIS), ultraviolet (UV), DUV, and VUV can be by parts covered by various commercial ellipsometers. Furthermore, fast in-situ ellipsometers with charged coupled device (CCD) detectors and imaging ellipsometers with a lateral resolution of about 1 μm are also commercially available.

15.5 Single Interface Here, we use the term single interface for a sharp and plane interface between two homogeneous media with different refractive indices. When an optical experiment is carried out, usually the ambient medium (with incident and reflected beams) is air, and the other medium is the sample to be studied. The sample surface is then identical with the interface itself, as shown in Figure 15.6. Under real conditions, such a sample surface is never perfectly sharp and plane, but often it can be approximated by a single interface. The limitation of this approximation will be discussed later in this section. Reflection coefficients for a single interface are given by the well-known Fresnel relations [8]. They can be derived by application of Maxwell’s equations, where in the first step a solution to the wave equation is searched in the ambient and sample material independently (as if they were infinite media). Subsequently, by applying the boundary conditions of continuity of the tangential components of the electric and magnetic fields at the interface, the required solutions are obtained. For s- and p- polarizations, the corresponding reflection coefficients rs and rp obtained in the form of Fresnel relations are thus, respectively,

15.5 Single Interface

Figure 15.6 Cartesian coordinate systems, defined by base vectors s and p, for the incident and reflected waves. Indices of refraction of the ambient and sample are denoted by Ni and Nt , respectively.

p

p

s

s θi

ki

kr

Ni Nt = nt – iKt

given as rs =

N i cos 𝜽i − N t cos 𝜽t , N i cos 𝜽i + N t cos 𝜽t

(15.16)

rp =

N t cos 𝜽i − N i cos 𝜽t , N t cos 𝜽i + N i cos 𝜽t

(15.17)

where N i and N t are complex refractive indices of the ambient and sample, respectively. The incidence and refraction angles, which are related by Snell’s law, are denoted by 𝜃 i and 𝜃 t , respectively. For further use, we provide here also the normal incidence Fresnel formula for the transmission coefficient (the ratio of the complex amplitudes of the transmitted and incident waves): t=

𝟐N i Êt = Ni + Nt Êi

(15.18)

In the derivation of the Fresnel Eqs. (15.16) and (15.17), care should be exercised with respect to the coordinate system that one selects. In fact, this selection may influence the sign of the r p coefficient [12]. Here, we consider two coordinate systems assigned to the incident and reflected waves as indicated in Figure 15.6. For normal incidence, the s base vectors of the incident and reflected waves coincide with each other, whereas the p base vectors are opposite in direction. It is useful to discuss briefly the incidence angle dependence of the Fresnel reflection coefficients and ellipsometric angles. For this reason, it is convenient to express Eqs. (15.16) and (15.17) in the form of their absolute values and their phases: r p = |r p |ei𝜙p

(15.19)

r s = |r s |ei𝜙s

(15.20)

The definition of the ellipsometric angles Ψ, Δ by Eq. (15.11) can then be rewritten as r p | r p | i(𝜙 −𝜙 ) = | | e p s = tan ΨeiΔ , (15.21) r s || r s || and hence

| rp | Ψ = atan || || | rs |

(15.22)

443

444

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

Δ = 𝜙p − 𝜙s

(15.23)

The dependences of the reflection coefficients rs and rp in Eqs. (15.16) and (15.17) on the angle of incidence are presented in Figure 15.7a,b, where we have considered that the ambient is air (as is often the case in a real experiment; N i = 1). Samples with refractive index nt = 2.0 and four different extinction coefficients: K t = 0.0 (transparent), 0.2, 0.8, and 1.5 are used in the modeling. The angle of incidence where the rp coefficient becomes zero is called the Brewster angle and is denoted by 𝜽B . This occurs only in the case of a transparent sample. The reason for the disappearance of the reflected beam with p-polarization lies in the radiation diagram of the electric dipole [12]. It can be shown that the Brewster angle is given by 𝜽B = atan

Nt Ni

(15.24)

Using Eq. (15.24) we obtain in our case 𝜽B = 63.4 ∘ . For optically absorbing samples, the absolute value of the rp coefficient, plotted in Figure 15.7a in red, does not reach a zero value. Instead, its angle dependence gives a nonzero minimum that shifts toward higher incidence angles with respect to the Brewster angle. The position of this minimum is called the principal angle. However, the absolute value of the rs coefficient, plotted in blue in Figure 15.7a, monotonously increases from normal incidence (where it is equal to |rp |) up to an incidence angle of 90∘ , where it reaches unity (together with |rp |). The phase of the p-reflection coefficient is shown in Figure 15.7b in red for an abrupt change (from 360∘ to 180∘ ) in a transparent sample when crossing the Brewster angle. On the other hand, the phase of the s-reflection coefficient (in blue in Figure 15.7b) is the incident angle independent with a constant value of 180∘ . The phase dependence on sample absorption is also presented in Figure 15.7b for the same four different extinction coefficients. The ellipsometry parameters Ψ and Δ of the single interface under consideration are calculated from Eqs. (15.22) and (15.23) and plotted as a function of the incident angle in Figure 15.7c,d, respectively. The signature of the Brewster (principal) angle is manifested by a minimum in the Ψ parameter and by a step-like change in the Δ parameter. It is worth noting that from the ellipsometric parameters Ψ and Δ we can directly calculate the optical constants n (real part of refractive index) and K (extinction coefficient) of any sample by using √ ( √ ( )2 ) √ 1 − 𝜌 2 √ (15.25) n − iK = sin 𝜃i 1 + tan2 𝜃i 1+𝜌 where 𝜌=

rp rs

= tan ΨeiΔ

(15.26)

The relations in Eqs. (15.25) and (15.26) are derived from Eqs. (15.11), (15.16), and (15.17) [15] under the assumption that the ambient is air (N i = 1.0). Since the angle of incidence 𝜃 i can be fixed in an experiment, the measured parameters Ψ and Δ for a given wavelength can be used to deduce the complex parameters 𝜌 and n and K. The single interface is the only case where the inverse relation given in Eq. (15.25) can be

15.5 Single Interface

(a)

1.0 │rs│

│rs│ and│rp│

0.8 0.6 0.4 0.2

│rp│

0.0

ϕs and ϕp [deg]

(b) 350

ϕp

300 250 ϕs

200 150 50

(c)

Ψ [deg]

40 30 20 10 200 (d)

Δ [deg]

150 Kt = 0.0 100

Kt = 0.2 Kt = 0.8

50

Kt = 1.5 0 0

20

40 60 angle of incidence [deg]

80

Figure 15.7 (a) Absolute values and (b) phases of reflection coefficients together with (c and d) ellipsometric parameters as a function of the angle of incidence and extinction coefficient of the sample.

445

446

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

derived in this analytical form. For other structures such as thin films, multilayers, etc. we have to use numerical methods for determining the complex refractive index from the experimentally recorded ellipsometric angles. Even if the incidence angle 𝜃 i appears explicitly in Eq. (15.25), the value of the complex refractive index does not depend on its selection. However, most accurate results are obtained when 𝜃 i is set close to the principal or Brewster angle, where rp (∼0) has a significantly different value with respect to rs (see Figure 15.7a,b). Often, it is useful to express the optical properties of a sample by its relative complex dielectric permittivity 𝜀 = 𝜀1 − i𝜀2 . It can be shown that in the range of optical frequencies, where the value of the relative magnetic permeability 𝜇 can be considered close to unity, the complex refractive √ index relates to the relative electric permittivity of a material by the relation n-iK = 𝜀. Therefore, we get 𝜀1 = n2 − K 2

(15.27)

𝜀2 = 2nK

(15.28)

To give an example of the application of the inverse relation in Eq. (15.25), we present the determination of the refractive index of NiFe2 O4 bulk single crystal in the spectral range covering the visible (VIS) and ultraviolet (UV) regions [16]. Figure 15.8a shows the spectrally dependent Ψ and Δ recorded on a single-crystal (111) face for two incidence angles 40∘ and 60∘ , and Figure 15.8b shows the refractive index n and extinction coefficient K calculated using the inverse relation in Eq. (15.25). From Figure 15.8a, Ψ and Δ are different at the incidence angles of 40∘ and 60∘ , and hence these depend on the angle of incidence. However, the optical constants n and K shown in Figure 15.8b are identical, and hence these do not depend on the angle of incidence, as pointed out earlier in the text below Eq. (15.26). However, if the determined n and K are found to be incidence angle dependent, then the approximation of a single interface is no longer valid. It is often due to the fact that the sample surface exhibits a more complex structure. In this case, the inverse relation, Eq. (15.25), may still be used, but the determined n and K values are called pseudo-optical constants, rather than optical constants of a material. Therefore, precautions must be taken when approximating the surface of a bulk material with a single interface. A single interface is an ideal structure that presents an abrupt and sharp change in refractive index across a planar interface between two semi-infinitive media – in our case, the ambient and sample. In practice, it is rather difficult to prepare such surfaces. Mechanical polishing of a sample surface usually causes a so-called Beilby overlayer that accumulates various defects due to the mechanism of polishing [17, 18]. This damaged layer can be reduced in thickness if a proper polishing process is selected or replaced by electropolishing. The Beilby overlayer may present different optical properties than the bulk and, if not carefully considered, erroneous optical constants are determined. Alternatively, sample cutting or cleaving in vacuum or the deposition of an optically thick film may be used. Ellipsometry as a phase-sensitive method is extremely sensitive to the sample surface and to some extent also to what is beneath the surface level where the light can penetrate. The penetration depth dp is defined by the inverse of the absorption coefficient 𝛼 as dp =

𝟏 𝜶

(15.29)

15.5 Single Interface

(a) Δ_40° Δ_60° Ψ_40° Ψ [deg]

Δ [deg]

Ψ_60°

refractive index and extinction coeff

(b) n n_40° k_40° n_60° k_60°

K

Photon Energy [eV]

Figure 15.8 (a) Recorded ellipsometry parameters on NiFe2 O4 single crystal and (b) its optical constants determined by the inverse formula (Eq. (15.25)).

It is also worth noting the relation between the absorption and extinction coefficients of the material: 𝛂=

𝟒𝜋K 𝛌

(15.30)

where 𝛌 is the light wavelength. An accurate determination of the optical constants is one of the main goals of ellipsometry, and this goal is of both fundamental and practical importance. Optical constants are, in fact, the response function of a material to an electromagnetic field and contain useful information about material properties, directly or indirectly. The wider available the spectral range of optical constants, the wider the range of information on material properties and their potential application that is gained. For instance, Figure 15.9 schematically shows the spectral dependence of the imaginary part of

447

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

1 0.8

Inter-band transitions

Intra-band transitions

0.6 ε2

448

0.4

Excitons Multi-phonon absorption Defects dopants

0.2 0 IR

Short wavelength absorption edge

VIS Photon Energy (a.u.)

UV

Figure 15.9 Schematic spectral dependence of the imaginary part of the relative electric permittivity 𝜀2 (Eq. (15.28)).

the relative electric permittivity 𝜀2 (Eq. (15.28)), where various types of light–matter interactions can be identified. Just to mention a few, we can point out the spectral range around the onset of the short wavelength absorption edge. From this part, we can determine not only the bandgap energy Eg of a material but also the nature of the electronic transition (direct or indirect, allowed or forbidden), which is often rather important for optoelectronic applications. Furthermore, the possible presence of the Urbach tail extends absorption for photons with an energy below the bandgap (due to various kinds of structural defects). The Urbach tail can play a key role for photosensitizers by effectively broadening the spectral range of the harvested light in photovoltaic films. This effect can be further significantly amplified in nanostructured films due to intense internal light scattering [19]. In addition, as the optical constants depend, in general, also on temperature (due to lattice parameter dilatation and electron–phonon interaction), optical monitoring of the sample surface temperature can be developed (for c-Si wafers see, e.g. [20]). Furthermore, for example, photon absorption by free charge carriers (electrons or holes) provides the possibility of determining their concentration [21] in the material.

15.6 Single Layer A single layer can be regarded as a plane-parallel layer sandwiched between two semi-infinitive media, the ambient, and the substrate. Optical spectra recorded of a single layer often contain interference fringes due to multiple internal reflections. As an example, we present in Figure 15.10 an optical reflectance spectrum of a chalcogenide amorphous thin film. The interference fringes, which appear very clearly in the VIS range, disappear as the thin film starts to absorb, and completely vanish in the UV range because of high light absorption in the material. The derivation of the reflection coefficients of a single layer is presented in textbooks on optics (see, e.g. [2]). One possible way to derive these (solution of Maxwell’s equations with the appropriate boundary condition on the two interfaces) is an extension of that already mentioned for a single interface in Section 15.5. Alternatively, we can consider multiple internal reflections in a film and calculate the complex

15.6 Single Layer

Figure 15.10 Typical reflectance spectrum of an amorphous chalcogenide thin film.

0.5

Reflectance

0.4

semi-transp.

transparent

absorbing

0.3

0.2

0.1

0 1

2

3

4

Photon Energy (eV)

amplitude of the reflected beam as a coherent superposition (summation) of their infinite contributions. For s- and p- polarizations, respectively, we then get rs = rp =

r 𝟏s + r 𝟐s e− i𝟐𝝓 𝟏 + r 𝟏s r 𝟐s e− i𝟐𝝓 r 𝟏p + r 𝟐p e− i𝟐𝝓 𝟏 + r 𝟏p r 𝟐p e− i𝟐𝝓

(15.31) (15.32)

where r1 and r2 are the reflection coefficients for the upper (ambient-layer) and lower (layer-substrate) single interface, respectively (see Eqs. (15.16) and (15.17)). The angle 𝝓 is given by 𝟐𝜋d 𝟐 (15.33) (N 𝟏 − N 𝟐𝟎 sin𝟐 𝜽𝟎 )𝟏∕𝟐 𝝀 which represents the phase change of a wave propagating through the film with thickness d and refractive index N 1 . 𝜽0 is the angle of incidence, and N 0 is the refractive index of the ambient. Using Eqs. (15.31)–(15.33), we can easily calculate the reflectance for a single film as shown in Eq. (15.13) and the ellipsometry parameters from Eq. (15.11). However, it should be noted that relations in Eqs. (15.31)–(15.33) are valid only for sufficiently thin layers with thickness smaller than the coherence length of the interacting (probe) light. The reflection coefficients in Eqs. (15.31)–(15.33) are valid over the whole spectral range. Nevertheless, in a region with strong absorption (penetration depth being much smaller than the layer thickness), light senses only the layer surface (upper interface) and may not reach the substrate. In this case, the single-layer reflection coefficients reduce to the simpler Fresnel formulas given in Eqs. (15.16) and (15.17) derived for a single interface. In the absorbing spectral range, the reflected light does not carry any information on the sample thickness, and the absorbed light probes only the penetration depth. In general, the ellipsometric spectra provide useful information on the geometrical (film thickness) and optical (film complex refractive index) properties of a sample. However, the desired optical and geometrical parameters cannot be expressed analytically as explicit functions of the experimental Ψ and Δ values as was done previously 𝝓=

449

450

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

(in Eqs. (15.25) and (15.26)) in the case of a single interface (bulk samples). On the contrary, the solution of this inverse problem for a thin film requires the construction of a sample model that consists of the optical constants of the substrate, layer, and ambient together with the layer thickness. Selected parameters (refractive indices and/or layer thickness) are then adjusted by a numerical procedure where the differences between theoretically calculated and experimental Ψ and Δ spectra are iteratively minimized by nonlinear fitting procedures. For the calculation of fitting errors, various functions have been used and, for example, one of these is given by [22] √ ( √M ( )2 ( )2 ) √∑ Ψ (𝜆 ) − Ψ (𝜆 ) (𝜆 ) − Δ (𝜆 ) Δ exp j teo j exp j teo j √ 𝜒=√ + 𝛿Ψ(𝜆j ) 𝛿Δ(𝜆j ) M − P − 1 j=1 1

(15.34) where (𝛿Ψ, 𝛿Δ) show measurement errors in (Ψ, Δ). M and P give the number of measurements spectral points and adjusted parameters, respectively. Ellipsometry provides only two experimentally recorded angles, Ψ and Δ, for each wavelength. This limits the number of searched (adjusted) parameters of the sample model that can be independently and uniquely determined. Nevertheless, this number may be increased by following approaches: 1. Appropriate parameterization of optical constants by analytic formulas. 2. Requirement of Kramers–Kronig consistency between the real and imaginary parts of optical constants. For instance, for the complex relative electric permittivity 𝜀 = 𝜀1 − i𝜀2 we get 𝜔′ 𝜀2 (𝜔′ ) ′ d𝜔 𝜔′2 − 𝜔2 ∞ 𝜀1 (𝜔′ ) − 1 ′ 2𝜔 𝜀2 (𝜔) = − P d𝜔 𝜋 ∫0 𝜔′2 − 𝜔2

𝜀1 (𝜔) =

2 P 𝜋 ∫0



(15.35) (15.36)

3. Measurement of Ψ, Δ for different angles of incidence. 4. Application of appropriate strategies for the data treatment as, for example, separate determination of the layer thickness from the transparent part of the spectra. 5. Adding transmittance and reflectance spectra for simultaneous data treatment. 6. Multiple sample method. Considering some of the above suggestions enables one to increase number of adjusted parameters to about 12 and reduce their mutual correlations within the tolerable limits. As an example of ellipsometry data treatment, we present in Figure 15.11a–c the experimental ellipsometric, transmittance, and reflectance spectra compared with best-fitted theoretical ones. Spectra were recorded for chalcogenide As50 Se50 amorphous film deposited on the float glass surface. The backside of the substrate was grounded before performing the ellipsometry and reflectance measurements to suppress spurious reflections from the interface. The single-layer sample model structure (see Figure 15.11d) was then designed for a homogenous thin film sandwiched between

15.6 Single Layer

30 (a)

(d)

Ψ [deg]

25 20 chalcogenide film 15

d

Substrate - glass

10 5 10 (b) electric permittivity

60

Δ [deg]

40 20 0 –20

ε1

6 4

ε2

2

3000 (c)

0.06

T

0.8

(f)

2500

0.05

0.4

α [nm–1]

0.6 R

0.03 0.02

0.2

2000

0.04 dp film thickness

1500

α

1000 500

0.01 0 1

3 4 2 Photon Energy [eV]

5

0

dp (nm)

Reflectance and transmittance

–40

(e)

8

1

2 3 4 Photon Energy [eV]

5

0

Figure 15.11 Ellipsometry and spectrophotometry measurements of a thin film of As50 Se50 as a function of photon energy. The best-fit ellipsometry parameters Ψ and Δ are plotted in (a) and (b), respectively, spectrophotometry transmittance and reflectance in (c), the sample model in (d), the determined electric permittivity 𝜀1 and 𝜀2 of As50 Se50 in (e), and its absorption coefficient 𝛼 and penetration depth dp in (f ).

two semi-infinitive media: the ambient and substrate. The transmittance spectrum for normal incidence was calculated by T = n2 |t|2

(15.37)

where n2 is the refractive index of the substrate and t=

t1 t2 e−i𝜙 1 + r1 r2 e−i2𝜙

(15.38)

is a single-layer transmission coefficient. The phase change 𝜙 is defined by Eq. (15.33), r1 and r2 are reflection coefficients of the upper and lower interfaces, respectively (see Eqs. (15.16)–(15.17)) and, analogously, t 1 and t 2 are the transmission coefficients for the upper and lower interfaces, respectively (see Eq. (15.18)).

451

452

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

Table 15.1 Adjusted values of As50 Se50 film thickness and parameters of Tauc–Lorentz formula (Eq. (15.39)). d = 1085.5 ± 1.1 nm A = 155.16 ± 1.39 eV E0 = 4.55 ± 0.03 eV C = 6.09 ± 0.09 eV Eg = 2.351 ± 0.001 eV

The refractive index of float glass was determined formerly by spectroscopic ellipsometry carried out on a naked substrate considering it as a bulk material (see Section 15.5, Eqs. (15.25) and (15.26)). The optical constants (electric permittivity) of the chalcogenide film were parameterized by a Tauc–Lorentz formula (that is, a combination of the Lorentz oscillator and Tauc absorption edge models), which is the appropriate parameterization for amorphous semiconductors [23]. The imaginary part of the Tauc–Lorentz electric permittivity 𝜀2 at a photon energy E is then expressed as 𝜀2 =

AE0 C(E − Eg )2 (E2

𝜀2 = 0



E02 )2

+

CE2

.

1 E

E > Eg E < Eg

(15.39) (15.40)

and the real part 𝜀1 can be calculated with the help of the Kramers–Kronig relations given in Eqs. (15.35) and (15.36) or by an analytic solution [23]. Parameters of the Tauc–Lorentz formula (Eqs. (15.39) and (15.40)) together with the film thickness were free parameters adjusted by the fitting procedure. Figure 15.11e presents the determined spectral dependence of the electric permittivity of As50 Se50 . The adjusted values of the searched parameters are listed in Table 15.1. Finally, Figure 15.11f shows the absorption coefficient 𝛼 and penetration depth dp , and indicates the determined layer thickness for comparison. Ellipsometry is a precise and sensitive spectroscopic tool, but when operating solely in the reflection configuration, its sensitivity for material absorption determination with low values of the absorption coefficient is limited. This is because reflected light senses only the sample surface and its close vicinity. It is commonly agreed that ellipsometry can provide reliable results for absorption coefficients 𝛼 > 105 cm−1 [24]. If 𝛼 is in the range 103 –105 cm−1 , usually the transmittance spectrum is required to be treated simultaneously with ellipsometry data to ensure accurate determination of the absorption coefficient. In the region of low absorption, 𝛼 < 103 cm−1 , other techniques such as photocurrent measurement or photothermal deflection spectroscopy are more appropriate [25]. It is worth noting that thin film optical characterization can also be performed by analyzing the sole transmittance spectrum. This approach was developed by Swanepole [26] (see also Chapter 1). Roughly speaking, thickness of the film is determined by the density of the interference fringes, whereas its refractive index is calculated from the difference of their maximal and minimal values. The absorption coefficient in the semitransparent and absorbing region of spectra can also be estimated by its parameterization and the subsequent fitting of the transmittance spectrum. Nevertheless, ellipsometry, and especially when combined with spectrophotometric methods, gives more accurate

15.6 Single Layer

results because it is also phase sensitive and hence records two parameters instead of only one parameter for each wavelength. Most thin films usually have structural defects and non-idealities, for example, nonuniform thickness, surface roughness, internal non-homogeneity, refractive index gradient, etc. Ellipsometry is a sensitive method capable of detecting most of such defects. A relatively easy examination of possible presence of defects and non-idealities in thin films can be done by comparing its reflectivity spectrum with that of the bare substrate alone. Theoretical considerations show that for optically transparent and ideal thin film, the reflectance spectrum of the uncoated substrate coincides with the upper or lower (depending on the refractive index difference between the substrate and film) envelope of the film reflectance interference fringes. This situation is shown in Figure 15.11c, where the homogenous chalcogenide thin film reflectance spectrum is plotted together with the reflectance of its bare glass substrate (in blue). If such a coincidence is not found, then the film cannot be regarded as a homogeneous single layer, or its refractive index has changed in the vicinity of the film/substrate interface during the deposition. To demonstrate the latter case, we will consider a study of SiO2 thin film deposited by high-density reactive ion plating on a glass substrate [27]. Figure 15.12(left) compares the reflectance spectrum of the thin film (in black) with the reflectance spectrum of the uncoated substrate (in pink). As one can clearly see, the maxima of the interference fringes of the thin film reflectance spectrum do not coincide with the reflectance spectrum of the uncoated substrate. Therefore, to model it as a single homogeneous SiO2 layer on the glass substrate cannot be correct. Consequently, three different sample models consistent with the experimental data may be considered (see Figure 15.12 right and middle parts).

SiO2 film

6 5 4 3

200

400 600 800 Wavelength [nm]

TL

glass

SiO2

SiO2 layer TL

Liner index gradient in SiO2 film

linear profile Glass Air SiO2 layer

glass

7

Air

Modification of substrate refractive index Air

SiO2 layer

Glass

glass

8

Transition layer

Glass

SiO2

substrate

Parabolic profile

Sample models

SiO2

Reflectance [%]

9

Refractive index

signature of structural defects

Refractive index

10

Refractive index profiles

Refractive index

Reflectance of SiO2 film and uncoated substrate

Figure 15.12 Signature of structural defects in the SiO2 layer presented in the reflectance spectra (left), considered sample models for ellipsometry and reflectance spectra interpretation (right), and refractive index profile of each designed sample model (middle).

453

454

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

The first model, drawn on the top of the right-hand side of Figure 15.12, describes the possibility of formation of a thin transition layer (TL) between the SiO2 layer and the glass substrate with a parabolic refractive index profile. In this model structure, it is assumed that the refractive index of the part of the TL in contact with the substrate is the same as that of the glass substrate. The second model, in the middle, assumes that the SiO2 film is a single layer with a linear profile for its refractive index. In the third model, at the bottom, a change in the glass refractive index close to the SiO2 /glass boundary is assumed. It is worth mentioning here that the proposal of different sample models that are acceptable from the technological point of view is rather important to ensure correct interpretation of the recorded optical spectra. This helps in selecting the most reliable sample model by evaluating and assessing the fitting errors in simultaneous treatment of the ellipsometric, reflectance, and transmittance spectra together with results of complementary characterization methods. In case of the study of the SiO2 thin film, it was the third model (a change in the glass refractive index) that was recognized as the most reliable [27]. Theoretical approaches that have been developed to account for the thin film defects are presented in a few monographs focused on spectroscopic ellipsometry [1, 9, 12] and/or in selected review articles (see, e.g. [28]).

15.7 Multilayer Continuous demand for better-performing optical elements requires surface structuring or functionalization, which is often more complex than depositing a single layer. Alternatively, as a first step, a single layer may be considered to consist of multilayers. This concept has proved to have broad applications, e.g. as antireflection coatings, omnidirectional mirrors, and optical filters [29]. In fact, a single layer considered above in Section 15.6 can be assumed to have a complex inner structure due to specific modes of growth during deposition, which can often be satisfactorily approximated by a multilayered structure. Therefore, the understanding of optical interactions in a multilayer structure becomes very useful from both application and characterization points of views. The reflection and transmission coefficients for bilayer and trilayer structures have been derived in analytical form and can be found in a few textbooks or review papers (see, e.g. [1, 30, 31]). However, a more general approach of treating the light interaction within a stratified medium involves recursion or matrix calculations (see, e.g. [1, 2, 10]). The method of matrix calculation expands and generalizes the idea of the interaction of an electromagnetic wave with a single interface or single layer, which was discussed in Sections 15.5 and 15.6. Here again, one has to solve the wave equation in each layer independently, and then the solutions are subjected to the boundary conditions at the interfaces enforcing the requirement of continuity of tangential electric and magnetic field components. The light interaction with the whole structure is then obtained by subsequent matrix multiplication of 2 × 2 matrices for each layer and interface, which finally reveal the reflection and transmission coefficients or ellipsometry parameters of the multilayer. The aim of this section is to present a case study where the multilayer approach can be applied as an approximation for a nanocrystalline diamond (NCD) film’s inner

15.7 Multilayer

NCD film top view

growth mode van der Drift

Multilayered sample model

Surface rough.

Bulk layer Seed layer coalescence limit

c-Si ND particles

Figure 15.13 SEM top view of NCD film (left), schematic drawing of van der Drift growth mode (middle) and sample model structure analyzed from ellipsometry (right).

structure [32]. The objective is to emphasize the potentiality and complementarity of spectroscopic ellipsometry as a characterization tool of nanostructures. The structure of NCD films deposited by microwave-plasma-enhanced chemical vapor deposition (MWPECVD) on a substrate pre-seeded by nano-diamond (ND) particles is schematically shown in Figure 15.13. In the described technique, diamond growth starts on seeded ND particles. The small crystallites or grains increase in volume and then coalesce at a certain distance from the substrate. When the film is completely closed, the grains continue to increase in volume following van der Drift growth, where the larger grains suppress the growth of their smaller neighbors. This type of growth typically results in columnar film structure. The size of the grains increases with the film thickness (together with the surface roughness), and when its average diameter surpasses about 100 nm, the film becomes microcrystalline rather than nanocrystalline. To be able to analyze the recorded ellipsometric spectra, a sample model structure has to be designed in the first step. A good strategy is to start with a simple sample model (a single layer in our case) and then gradually refine it with respect to the expected growth mode of the layer and/or with respect to the complementary characterization tools if they are available. The refined sample model structure is presented in Figure 15.13(middle) and (right) and consists of a substrate that is a naturally oxidized crystalline Si wafer. The substrate is then covered by a seed layer (a layer grown from seeded ND particles, dots in red, up to a coalescence limit), bulk layer (dense columnar part in blue), and surface roughness layer on the top. Optical constants of NCD were parameterized by the Tauc–Lorentz formula [23] given in Eq. (15.39) and (15.40). The optical constants of seed and surface layers, where voids are expected, were parameterized by the effective medium approximation (EMA), which is treated in more detail later in the text. This multilayer system has been used to fit experimentally recorded ellipsometry spectra. The best-fit results are presented in Figure 15.14a and have revealed both the optical constants of NCD and the inner structure of NCD film of appropriate thicknesses of seed, bulk, and surface sublayers (see Figure 15.14c,d). It is often advisable especially when a large number of adjusted parameters is searched by the fitting procedure (9 in our case) to compare the determined sample structure with some complementary characterization techniques. These are usually scanning probe

455

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

70 60 50 40 30 20 10

0

1 2 3 4 5 Photon Energy [eV] (a)

Surface rough.

NCD film

Bulk layer Seed layer

c-Si (b)

300 250 200 150 100 50 0 –50 –100

Δ [deg]

Ψ [deg]

456

c-Si

6

tSR = 17

x

17 nm

tBL = 117 120 140 nm tSL = 17

x

x nm

SE SEM AFM (c)

(d)

Figure 15.14 (a) Best-fit spectra for a selected angle of incidence of 65∘ [32], (b) SEM picture of NCD film cross section, (c) sample model structure for ellipsometry data analysis, and (d) determined inner structure by spectroscopic ellipsometry (SE), scanning electron microscopy (SEM), and atomic force microscopy (AFM).

methods such as atomic force microscopy (AFM), scanning electron microscopy (SEM), or transmission electron microscopy (TEM). Figure 15.14b presents an SEM picture of the NCD film cross section, and Figure 15.14d compares values of the film surface roughness and bulk and seed layer thicknesses determined independently by spectroscopic ellipsometry (SE), SEM, and AFM. It is worth mentioning that SE provides most details of the NCD film inner structure and that the values of the searched parameters are found to be rather close with all the characterization techniques used. Besides determining the inner structure, we can further assess the “diamond quality” of NCD films. NCD consists mainly of sp3 hybridized carbon atoms arranged in grains with a diamond crystalline structure. Apart from this dominant phase, all possible hybridizations (sp3 , sp2 , and sp1 ) might be present at defect sites and are localized mainly in the grain joints and seed layer where the coalescence is reached. The optical constants of a material composed of a host medium filled with different types of inclusions can be approximated by the EMA [33] that takes into account the host and inclusion material optical constants and their corresponding filling factors. Therefore, if NCD optical properties are parameterized by EMA with the diamond phase as the host and non-diamond phases as the inclusions, then the “diamond quality” could be estimated from the spectro-ellipsometric analyses by assessing the filling factors ratios.

15.7 Multilayer

EMAs are based on the well-known Clausius–Mossotti relation given by [2]: 𝜀 − 1 N𝛼 = 𝜀 + 2 3𝜀0

(15.41)

which relates the macroscopic electric permittivity 𝜀 of a medium with its polarizability 𝛼 of N electric dipoles. When the medium consists of two components a and b with different polarizabilities 𝛼 a and 𝛼 b , we can rewrite Eq. (15.41) as 1 𝜀−1 (N 𝛼 + Nb 𝛼b ) = 𝜀 + 2 3𝜀0 a a

(15.42)

Combining Eqs. (15.41) and (15.42), we get the so-called Lorentz–Lorenz formula given by 𝜀 −1 𝜀 −1 𝜀−1 = fa a + (1 − fa ) b 𝜀+2 𝜀a + 2 𝜀b + 2

(15.43)

where f a and f b = (1 − f a ) are the filling factors of component a and b, respectively. This relation is valid for a medium consisting of spheres placed in a vacuum. If the vacuum is replaced by a host medium with electric permittivity 𝜀h we get 𝜀 − 𝜀h 𝜀 − 𝜀h 𝜀 − 𝜀h = fa a + (1 − fa ) b (15.44) 𝜀 + 2𝜀h 𝜀a + 2𝜀h 𝜀b + 2𝜀h Several variations of this relation can be found in the literature. One of them, also known as the Bruggeman approximation, is based on the assumption that 𝜀 = 𝜀h [34]. Then Eq. (15.44) becomes 𝜀 −𝜀 𝜀 −𝜀 + (1 − fa ) b (15.45) 0 = fa a 𝜀a + 2𝜀 𝜀b + 2𝜀 and it can be further extended for more phases. Bruggeman EMA is often used in ellipsometry for modeling the surface roughness formed by a mixture of air and top layer material with a filling factor, typically 50%. Further, Bruggeman EMA has also been found broader use in the determination of the effective optical constants of composite materials containing two or more phases and, in this form, it has also been applied for estimation of NCD film “diamond quality.” The diamond and non-diamond phases were represented by the optical constants of bulk diamond [35] and amorphous carbon [36], respectively. Numerical treatments of experimental ellipsometric spectra in the frame of EMA yield the filling factors of both the diamond and non-diamond phases. This analysis was applied for NCD films deposited under different deposition conditions (substrate deposition temperature and frequency of pulsed microwave plasma). The obtained results are summarized in the form of a graph shown in Figure 15.15(left), where we indicate the area of the deposition parameters that provides high-diamond-quality NCD films. A complementary technique to assess the diamond quality of the NCD films is Raman spectroscopy. Raman spectra, consistently with ellipsometry, clearly show the improvement of the diamond quality of the NCD films with increasing pulsed plasma frequency. This is demonstrated in Figure 15.15(right) for NCD films deposited with a different pulsed plasma frequency and at a fixed substrate temperature of 550 ∘ C. The presented spectra were normalized to the amplitude of the sharp diamond peak (1332 cm−1 ) to create direct evidence of the non-diamond dependence of the pulsed plasma frequency.

457

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

Diamond quality (SE)

Raman spectroscopy 1.2

16 14

9.0

12

7.5

High diamond quality

7.6

10 8

8% of a-C

6 Low diamond 12.3 quality 16.1 2 520

8.7

7.7

13.9

10.8

560

600

substrate temperature [°C]

normalized intensity

pulsed plasma frequency [kHz]

458

H M L

1.0

sp2

0.8 0.6 sp3

0.4 0.2 0.0 1200

1400

1600

frequency [cm–1]

Figure 15.15 Diamond quality of NCD films deposited with different pulsed microwave (MW) plasma frequency and substrate deposition temperature as determined by spectroscopic ellipsometry (SE). Indicated are (left) filling factors of non-diamond amorphous carbon phase and estimated line of constant a-C filling factor of 8% [32]. (Right) Raman spectra of NCD films deposited at 550 ∘ C for different frequencies of pulsed plasma (High 14.3 kHz, Middle 4.5 kHz, and Low 2.7 kHz).

The D (1340 cm−1 ) and G (1600 cm−1 ) bands related to disordered carbon and graphite, respectively, are clearly visible for all samples and were found to decrease in amplitude as the pulsed plasma frequency increased. Raman spectra deconvolution can be used for carbon sp3 /sp2 ratio estimation, but significantly different cross sections of diamond and non-diamond phases must be considered. The diamond quality of the NCD films determined from the Raman spectroscopy was found to be in the range 92–97%, and similar values of 84–93% were obtained by SE. It is worth noting that this reasonable agreement was reached in spite of the fact that the exact structure and ratio of sp1 , sp2 , and sp3 hybridized carbon in defect sites of NCD are not known; therefore, the approximation of the non-diamond phase by amorphous carbon is rather rough (for a more detailed discussion, see [32]).

15.8 Linear Grating As discussed in the preceding sections, ellipsometry is a characterization technique that is routinely used for characterization of bulks, single layers, and multilayers. In the last three decades, advanced theoretical approaches in the area of electromagnetic field interaction with nanostructured solids together with the enhanced power of personal computers have made it possible to apply ellipsometry for the characterization of laterally structured surfaces also. This includes, for example, linear gratings or others, even more complex nanostructures (such as photonic crystals, with periodicity in both lateral and perpendicular directions with respect to the sample surface) [37–41]. In this case, the generally accepted term is scatterometry rather than ellipsometry to distinguish its employment for the characterization of structured surfaces or photonic crystals. In recent years, scatterometry has gained increased attention also due to its potential application in on-line control of critical dimensions of nanostructures fabricated in the semiconductor industry.

15.8 Linear Grating

Calculation of the reflection coefficients of photonic crystals requires a more sophisticated mathematical approach compared to that presented previously for the case of a single interface and single or multilayers. Due to the lateral periodicity of a sample’s optical properties (or surface profile) in photonic crystals, it is convenient to expand the interacting electromagnetic field and the sample electric permittivity into Floquet and Fourier series, respectively. The wave equation with suitable boundary conditions is then solved for these expanded fields in the truncated series. This analytical calculation method, known as Fourier modal formalism, treats the Fraunhofer diffraction in planar multilayer anisotropic gratings (for more details, see, for example, [42–44] and the references therein). Special care has to be taken for the correct implementation of this method, including the control of the convergence of the solution with respect to the truncation of the series. Alternatively, numerical approaches are also available, for example, that proposed by Yee [45], which is based on the finite difference time domain (FDTD). Because of its computational demands, the current commercial ellipsometry software generally do not support this kind of calculations, although some stand-alone routines are available (see, e.g. [46]). Light reflected from gratings or photonic crystals presents diffraction. Experimentally, it is most convenient to measure ellipsometric parameters in the 0th diffracted order, because the angle of reflection for higher orders is spectrally dependent. Moreover, the change in light polarization even in the 0th order includes all the needed information on both the sample structure (for example, grating profile) and the optical properties of grating materials. The aim of this section is a brief presentation of the potential of scatterometry and its advantages and disadvantages with respect to the conventional characterization techniques of the selected case study: the characterization of a sine-like surface relief Ni grating [47]. A linear Ni grating was fabricated by holographic lithography. Its surface was covered by a native oxide. The grating cross section is schematically depicted in Figure 15.16 together with the top view of a set of gratings. Besides scatterometry, complementary measurements of the grating profile were performed by conventional techniques: SEM, AFM, and optical microscopy, as shown in Figure 15.17. All these tools provided the value of the grating period Λ. Nevertheless, the grating profile depth was determined only by AFM. The obtained results are summarized in Table 15.2. Scanning probe methods (SEM and AFM) including optical microscopy are relatively fast and flexible. SEM and AFM provide high resolution of the nanoscale features of the studied samples. Moreover, AFM also has access to the real surface topography. However, there are

< NiO

h Ni

1 cm

Figure 15.16 Grating cross-sectional (left, [47]) and top-view photo of a set of holographic gratings (right).

459

460

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

Axis, nm 250 200 150 100 50 0 0 1

Y Axis 8 6 2 3 4 5 6 7 8 9

4 2 0

X Axis, μm

10 μm

100 μm

Figure 15.17 Grating surface recorded by the (top) atomic force microscopy (AFM), (middle) scanning electron microscopy (SEM), and (bottom) optical microscopy.

some disadvantages as well in the use of these tools. For example, (1) SEM requires surface metallization and has only limited access to the surface profile, (2) AFM scans only a small surface area and therefore several scans are required to obtain statistically reliable results. In addition, the tip deconvolution artifacts often make precise surface topography evaluation difficult, and (3) optical microscopy suffers from the low resolution and with lack of access to the depth profile. The ellipsometric parameters Ψ and Δ were recorded for three angles of incidence: 20∘ , 30∘ , and 40∘ in the spectral range 300 to 1200 nm. The plane of incidence was set perpendicular to the grating lines. Only the zeroth order of diffraction (specular reflection) was measured and analyzed. Due to the non-ideal fabrication process, the Ni grating surface relief was not considered perfectly sinusoidal. Therefore, not only were the

15.8 Linear Grating

Table 15.2 Determined grating period and profile depth (× indicates unaccessible parameter). Methods Grating structural parameters

SEM

Optical microscopy

AFM

Spectroscopic ellipsometry

Grating period, nm

916.7

917.2

918.6

917.0

Profile depth, nm

×

170

×

205

period and depth of the grooves taken into account in the theoretical calculations, but an additional parameter describing the surface profile non-ideality was also considered. Moreover, for correct scatterometry spectra evaluation, we need to determine the thickness of the NiO surface overlayer. This was done by the standard ellipsometry measurement of a nonpatterned part of the sample (t NiO = 3.5 nm). The non-ideality in the profile shape was approximated by z = (h∕2){1 − cos[P(y)]}

(15.46)

P(y) = (2π∕Λ)[Ay + (2∕Λ)(1 − A)y2 ]

(15.47)

where

with the shape profile parameters A and y, and z are the Cartesian axes (A = 0 corresponds to the ideal sinusoidal case). Comparing the experimental ellipsometry spectra with the theoretical simulations, performed in the analytical framework of rigorous coupled wave analyses [48], and based on various depths and profile shapes, we have found the following: (1) grating period Λ = 917 nm, (2) profile depth h = 205 nm, and (3) shape profile parameter A = 0.52. The best-fit spectra of the Ψ parameter are presented in Figure 15.18, where we also compare the final grating profile determined by scatterometry with respect to that measured by AFM. It is worth noting that a high level 80

1 model 205 nm

Ψ [deg]

60 50 40 30 20

Scaled vertical axis [a.u.]

experiment

70

0.5

0

–0.5 AFM meas. profile SE - fitted profile

10 0 200

–1 400

600

800 1000 1200 1400

wavelength [nm]

–1

–0.5

0

0.5

Scaled horizontal axis [a.u.]

Figure 15.18 Best-fit scatterometry spectra and grating profile determined by (left) scatterometry compared with that measured by (right, [47]) AFM.

1

461

462

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

Table 15.3 Advantages and disadvantages of applied methods. Method

Advantage

Disadvantage

SEM

Fast and flexible, high resolution

Limited access to surface profile, surface metallization required

AFM

Real surface topography

Small surface area scan, tip deconvolution artifacts

Optical microscopy

Fast and flexible inspection

Low resolution, no access to depth profile

Spectroscopic ellipsometry

Sensitive on NiO overlayer and surface profile non-ideality, noninvasive

Difficult and time-consuming calculations

of agreement is achieved between both the profiles. The grating period and its depth thus obtained are added in Table 15.2. It is clear that a grating period of about 917 nm is obtained by all the applied methods. However, there is a discrepancy between the profile depths determined by AFM (170 nm) and SE (205 nm). The possible reasons of this discrepancy are AFM tip deconvolution artifacts or ellipsometry sensitivity on grating surface micro-roughness, which was not considered in the sample model. Moreover, the optical constants of Ni and NiO determined on the nonpatterned part of the sample may be slightly different from those of the grating. The sensitivity of scatterometry on an overlayer can be considered as an advantage or a disadvantage depending on how precise theoretical analyses are performed. An evident advantage of the scatterometry is its noninvasive nature and access to the real profile geometry of the grating surface. Nevertheless, the difficult and time-consuming calculations still limit the applications of scatterometry to a relatively small community of researchers. Despite the above disadvantage, scatterometry has proved to be a highly efficient tool for noninvasive characterization of laterally periodical patterned nanostructures. In the case of linear grating as presented here, this has been demonstrated by determination of the grating geometrical parameters (period, depth, and profile shape non-ideality) and also by identification of the NiO surface overlayer. Obviously, all the characterization techniques presented here show comparative advantages and disadvantages. The most evident ones are listed in Table 15.3.

15.9 Conclusions SE is a phase-sensitive optical tool that utilizes light polarization (more precisely, its change under light reflection) for surface characterization. It provides surface geometrical parameters such as, for example, the overlayer thickness, the inner structure of a film or surface profile of a grating and also the optical constants of materials that the light senses by its penetration beneath the surface. This technique is not direct, in the sense that ellipsometry data treatment usually requires design of a sample model and subsequent fitting of experimental ellipsometric spectra. Appropriate combination of SE with

References

complementary surface characterization tools guarantees precise and accurate results. The field of ellipsometry is growing continuously, and its potential is being applied for characterization of advance functionalized surfaces and nanomaterials.

Acknowledgments The author gratefully acknowledge support from the Ministry of Education, Youth, and Sports of the Czech Republic no. LM2015082.

References 1 Azzam, R.M.A. and Bashara, N.M. (2003). Ellipsometry and Polarized Light. Amster-

dam, Netherlands: Elsevier. 2 Born, M. and Wolf, E. (1999). Principles of Optics. Cambridge, UK: Cambridge

University Press. 3 Brosseau, C. (1998). Fundamentals of Polarized Light; a Statistical Optics Approach. 4 5 6 7 8 9 10 11

12 13 14 15

16 17 18 19 20 21

Wiley. Drude, P. (1889). Ann. Phys. 272: 865. Drude, P. (1889). Ann. Phys. 272: 532. Drude, P. (1890). Ann. Phys. 39: 481. Losurdo, M. and Hingerl, K. (eds.) (2013). Ellipsometry at the Nanoscale. Springer. Humlicek, J. (2005). Polarized Light and Ellipsometry. In: Handbook of Ellipsometry (eds. H.G. Tompkins and E.A. Irene), 3–91. Heidelberg, Germany: Springer. Tompkins, H.G. and Irene, E.A. (2005). Handbook of Ellipsometry. Heidelberg, Germany: Springer. Ohlidal, I. and Franta, D. (2000). Prog. Opt. 41: 181. Garcia-Caurel, E., Ossikovski, R., Foldyna, M. et al. (2013). Advanced Mueller Ellipsometry instrumentation and data analyses. In: Ellipsometry at the Nanoscale (eds. M. Losurdo and K. Hingerl), Ch. 2, 31–144. Springer. Fujiwara, H. (2007). Spectroscopic Ellipsometry, Principles and Applications. Chichester, UK: Wiley. Tompkins, H.G. and Irene, E.A. (eds.) (2005). Instrumentation. In: Handbook of Ellipsometry, Part II, 297–564. Heidelberg, Germany: Springer. Mansuripur, M. (2009). Classical Optics and its Applications, 2e. Cambridge University Press. Aspnes, D.E. (1976). Spectroscopic ellipsometry of solids. In: Optical Properties of Solids: New Developments (ed. B.O. Seraphin), Chapter 15, 801–846. North-Holland, Amsterdam: Elsevier Science Publishing Co. Inc. Mistrik, J., Visnovsky, S., Grondilova, J. et al. (2001). J. Magn. Soc. Jpn. 25: 267–270. Finch, G.I. and Quarrell, A.G. (1936). Nature 137: 516–519. Ward, L. (1988). The Optical Constants of Bulk Materials and Films. Bristol and Philadelphia: Adam Hilger. Zazpe, R., Sopha, H., Prikryl, J. et al. (2018). Nanoscale 10 (35): 16601. Postava, K., Aoyana, M., Mistrik, J. et al. (2007). Appl. Surf. Sci. 254 (1): 416. Taylor, A., Fekete, L., Hubik, P. et al. (2014). Diam. Relat. Mater. 47: 27.

463

464

15 Optical Characterization of Materials by Spectroscopic Ellipsometry

22 Herzinger, C.M., Johs, B., McGahan, W.A. et al. (1998). J. Appl. Phys. 83: 3323. 23 Jellison, G.E. Jr., and Modine, F.A. (1996). Appl. Phys. Lett. 69: 371, Erratum (1996).

Appl. Phys. Lett. 69, 2137. 24 Collins, R.W. and Vedam, K. (1995). Optical properties of solids. In: Encyclopedia of

Applied Physics, vol. 12, 285–336. Wiley-VCH. 25 Jackson, W.B., Amer, N.M., Boccara, A.C., and Fournier, D. (1981). Appl. Opt. 20:

1333. 26 Swanepoel, R. (1984). J. Phys. E 17: 896. 27 Mistrik, J., Ohlidal, I., Antos, R. et al. (2005). Appl. Surf. Sci. 244: 51. 28 Ohlidal, I., Cermak, M., and Vohanka, J. (2018). Optical Characterization of Thin

Solid Films, Springer Series of Surface Science, vol. 64, 271. Springer. 29 Heavens, O.S. (1965/1991). Optical Properties of Thin Solid Films. New York: Dover 30 31 32 33

34 35 36 37

38 39 40 41 42 43 44 45 46 47 48

Publications. Kildemo, M., Hunderi, O., and Drevillon, B. (1997). J. Opt. Soc. Am. A 14 (4): 931. Vasicek, A. (1960). Optics of Thin Films. Interscience Publishers. Mistrik, J., Janicek, P., Taylor, A. et al. (2014). Thin Solid Films 571: 230. For a review, see for example (J. Humlicek (2013). Data analyses for Nanomaterials: effective medium approximation, its limits and implementations. In: Ellipsometry at the Nanoscale (eds. M. Losurdo and K. Hingler) Ch. 3. Springer, and D.E. Aspnes (1982). Thin Solid Films, 89: 249. Bruggeman, D.A.G. (1935). Ann. Phys. 24: 636. Edwards, D.F. and Philipp, H.R. (1985). Handbook of Optical Constants of Solids (ed. E.D. Palik), 665. Orlando: Academic Press. S. A. Sopra Database of optical constants of solids, (http://sspectra.com/sopra.html) See for example: Bergmair, M., Hingerl, K., and Zeppenfeld, P. (2013). Spectroscopic Ellipsometry on metallic gratings. In: Ellipsometry at the Nanoscale (eds. M. Losurdo and K. Hingerl). Ch. 7. Springer. Veis, M. and Antos, R. (2013). J. Nanomater. 2013: 621531. Huang, H.T. and Terry, F.L. Jr., (2004). Thin Solid Films 455-456: 828. Novikova, T., De Martino, A., Bulkin, P. et al. (2007). Opt. Express 15: 2033. Chen, X., Liu, S., Zhang, C. et al. (2014). Opt. Express 22: 15165. Rokushima, K. and Yamakita, J. (1983). J. Opt. Soc. Am. 73: 901. Rokushima, K., Antos, R., Mistrik, J. et al. (2006). Czechoslov. J. Phys. 56: 665. and references therein. Visnovsky, S. (2006). Optics in Magnetic Multilayers and Nanostructures. CRC Press. Yee, K. (1966). IEEE Antennas Propag. Mag. 14: 302. J. Hugonin, P. Lalanne (2005). Reticolo software for grating analyses. Technical Report, Institut d’Optique, Orsay, France. Mistrik, J., Karlovec, M., Palka, K., and Antos, R. (2017). 2017 IEEE International Conference on Computational Electromagnetics, 298–299. Antos, R., Pistora, J., Mistrik, J. et al. (2006). J. Appl. Phys. 100: 054906.

465

16 Excitonic Processes in Quantum Wells Jai Singh and I.-K. Oh College of Engineering, IT and Environment, B-Purple 12, Charles Darwin University, Darwin, NT 0909, Australia

CHAPTER MENU Introduction, 465 Exciton–Phonon Interaction, 466 Exciton Formation in QWs Assisted by Phonons, 467 Nonradiative Relaxation of Free Excitons, 474 Quasi-2D Free-Exciton Linewidth, 485 Localization of Free Excitons, 491 Conclusions, 499 References, 500

16.1 Introduction During the last few decades, there has been astonishing progress in the field of semiconductor nanostructures due to the development of techniques in the epitaxial crystal growth of materials such as molecular-beam epitaxy and metal–organic chemical vapor deposition. These techniques have made it possible to manipulate and design semiconductor devices with high precision in the atomic scale and to produce high-quality low-dimensional systems such as quantum wells (QWs), superlattices, quantum wires (QWRs), and quantum dots (QDs) that give rise to several exotic and interesting phenomena [1]. These low-dimensional semiconductor systems have important applications in optoelectronic devices, e.g. light detectors, LASERs, and LEDs. Many physical properties of optoelectronic devices fabricated from such low-dimensional semiconductor systems are determined from information on excitonic processes. One of the most important properties of these semiconductor nanostructures is the exciton–phonon interaction, which plays a very significant role in the dynamics of excitons [2] associated with the formation [3, 4], relaxation [5, 6], dephasing [7], and localization [8] of excitons leading to the study of photoluminescence (PL) kinetics [9], phonon-induced luminescence [10], degenerate four-wave mixing [11], etc. In this chapter, we shall discuss the excitonic processes related to charge carrier–phonon interaction in QWs.

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

466

16 Excitonic Processes in Quantum Wells

16.2 Exciton–Phonon Interaction Excitonic processes such as the formation, relaxation, dephasing, and localization of excitons are influenced by the interaction with phonons. Since an exciton consists of an electron in the conduction band bound with a hole in the valence band through their Coulomb interaction, the exciton–phonon interaction is the sum of the electron–phonon and hole–phonon interactions. In noncentrosymmetric and polar crystals of III–V and II–VI semiconductors, charge carriers as well as excitons interact with acoustic phonons through the deformation potential (DP) and piezoelectric (PE) couplings and with optical phonons through the polar optical (PO) coupling [5]. However, it is to be noted that a strong electron–phonon or hole–phonon interaction does not mean a strong exciton–phonon interaction [3], which will be explained later. In general, as the conduction band near the minimum is isotropic, electron–acoustic phonon interaction obtained from the DP has a contribution from only longitudinal acoustic (LA) phonons. However, as the valance band is anisotropic, both holes–LA and holes–transverse acoustic (TA) phonons interactions are nonzero [5]. On the other hand, through the PE coupling, interactions of both the charge carriers, electrons in the conduction band and holes in the valence band, are nonzero with both LA and TA phonons. As regards the interaction of charge carriers with the optical phonons due to the polar coupling, the contribution of transverse optical (TO) phonons can usually be ignored because the TO phonons produce negligible electric fields, whereas the longitudinal optical (LO) phonons produce sizable electric fields in the direction of phonon propagation [12]. Accordingly, the exciton–phonon interaction Hamiltonians due to the DP HID , piezoelectric HIP , and polar LO phonon HIO couplings, can be written, respectively, as [5] HID (r || , R|| , ze , zh )𝜆,q = CD𝜆 [ieiq|| .R|| (Ξ𝜆c (q)e

i𝛼hq|| .r|| iqz ze

e

− Ξ𝜆v (q)e−i𝛼e q|| .r|| eiqz zh )̂ b𝜆q + c.c.]

(16.1)

b𝜆q + c.c.] HIP (r|| , R|| , ze , zh )𝜆,q = Cp𝜆 [ieiq|| .R|| (eiqz ze − e−i𝛼e qII . r|| eiqz zh )̂

(16.2)

HIO (r|| , R|| , ze , zh )LO,q = COLO [ieiq|| .R|| (eiqz ze − e−i𝛼e qII . r|| eiqz zh )̂ bLOq + c.c.]

(16.3)

and

Here, c.c. denotes the complex conjugate of the first term, and the center-of-mass (COM) coordinate and the relative coordinate for an exciton in the plane of QWs are given, respectively, as R|| = 𝛼e re|| + 𝛼h rh|| , 𝛼e =

m∗e|| M||∗

, 𝛼h =

m∗h|| M||∗

(16.4)

and (16.5)

r|| = re|| − rh|| where

m∗e|| ,

m∗h|| ,

and

M||∗

are the effective masses of the electron, hole, and exciton, respectively, in the plane of QWs. ̂ b𝜆q is a 𝜆-mode (𝜆 = LA, TA1, TA2) phonon annihilation operator with the wave vector q = (q|| , qz), where q|| is in the plane of QWs and qz is perpendicular to the QW walls, and ze and zh are the z-components of the electron

16.3 Exciton Formation in QWs Assisted by Phonons

and hole coordinates, respectively. Ξ𝜆c (q) and Ξ𝜆v (q) are the effective DPs [5] for the conduction and valence bands, respectively. CD𝜆 , Cp𝜆 , and COLO are respectively given by √ ℏq 𝜆 CD = (16.6) 2𝜌m v𝜆 V ( 𝜆 )√ ehq ℏ 𝜆 (16.7) Cp = i 𝜅𝜀0 2𝜌m v𝜆 qV √ ) ( ℏ𝜔LO 1 1 1 LO − (16.8) CO = ie 2𝜀0 V 𝜅∞ 𝜅0 q where rm denotes the material density, V volume, v𝜆 sound velocity, 𝜅 relative dielectric constant, 𝜀0 electric permittivity in the vacuum, h𝜆q effective PE constant for a 𝜆− mode phonon, 𝜅 ∞ and 𝜅 0 high-frequency and static relative dielectric constants, respectively, and 𝜔LO is the LO phonon frequency.

16.3 Exciton Formation in QWs Assisted by Phonons When the energy of an incident photon on a nanostructure semiconductor is above its bandgap energy, the photon is absorbed, and a free electron–hole pair is excited. Such photogenerated electron–hole pairs may relax nonradiatively and form excitons [9] by emitting phonons. At excitation energies larger than the bandgap energy, a photoexcited electron–hole pair can form an exciton first with a large total wavevector K|| , corresponding to its COM motion [13, 14]. The exciton then relaxes nonradiatively down to the K|| ∼ 0 state by emitting phonons, and finally it recombines radiatively from the K|| ∼ 0 excitonic state by emitting a photon to conserve the energy and momentum. Therefore, in exciton luminescence experiments, information on the formation time of an exciton as a function of the exciton wavevector is crucial to the study of luminescence rise time 𝜏 R . At low excitation densities, photoexcited electrons and heavy holes occupy only their first sub-band and so do heavy-hole excitons. In this case, an exciton is considered to be formed from a free quasi-2D electron and hole pair due to phonon emission or absorption in a 𝜆-mode (LA, TA1, TA2, and LO) via DP, PE, and PO couplings. Then the interaction Hamiltonian in the second quantized form involving such formation of excitons can be written as (see Refs. [3–5]) ∑ J ̂ ̂† ̂ ̂ J = √1 ̂ CJ𝜆 [F𝜆− (q|| , qz , k|| )B b d a H K|| 𝜆q h,𝛼h (K||−q|| )−k|| c,𝛼h (K||−q|| )+k|| I,𝜆 A0 q|| ,qz ,K|| ,k|| J ̂ ̂† ̂ ̂ + [F𝜆+ (q|| , qz , k|| )B ] b† ̂ b d a K 𝜆q 𝜆q h,𝛼h (K||−q|| )−k|| c,𝛼e (K||+q|| )+k|| ||

(16.9)

̂ where A0 is the 2D area of QWs, and d ac,𝛼h (K||−q|| )+k|| are the hole and h,𝛼h (K||−q|| )−k|| and ̂ ̂† electron annihilation operators in the valence and conduction bands, respectively. B K||

is the creation operator of an exciton. J = D, P, and O represent DP, PE, and PO couplings, 0 D P , F𝜆∓ [4] and FLO∓ [3] are, respectively, obtained as respectively. The factors F𝜆∓ D F𝜆∓ = ±i[Ξ𝜆c (q)Fe (±qz )G1 (k|| ± 𝛼h q|| )

− Ξ𝜆v (q) Fh (±qz )G1 (k|| ∓ 𝛼e q|| )]

(16.10)

467

468

16 Excitonic Processes in Quantum Wells P F𝜆∓ (q|| , qz , k|| ) = ±ih𝜆q [Fe (±qz )G1 (k|| ± 𝛼h q|| ) − Fh (±qz )G1 (k|| ∓ 𝛼e q|| )]

(16.11)

0 FLO∓ = [Fe (±qz )G1 (k|| ± 𝛼h q|| ) − Fh (±qz )G1 (k|| ∓ 𝛼e q|| )]

(16.12)

The form factors Fj ( =je,h) and G1 are given as [3]: Fj (qz ) =



dzj |𝜙j (zj )|2 eiqz zj , j = e, h

(16.13)

and √ G1 (k|| + 𝛼q|| ) =



dr|| 𝜙∗x (r|| )ei(k|| +𝛼q|| )

=

8𝜋 𝛽2

[(

∣ k|| + 𝛼q|| ∣ 𝛽

]−3∕2

)2 +1

(16.14) where 𝜙j is the electron (j = e) and hole (j = h) sub-bands, and 𝜙x is the 1s exciton wave function, which is chosen to be a trial wave function given by [15] √ 2𝛽 2 𝛽r|| 𝜙x (r|| ) = (16.15) e 𝜋 The first term of Eq. (16.9) corresponds to the formation of an exciton due to phonon absorption, and the second term corresponds to that due to phonon emission. It is obvious from Eqs. (16.11) and (16.12) that the PE and LO phonon contributions to the interaction Hamiltonian in QWs are mainly influenced by the form factors F j and G1 regardless of the individual strength of the electron– and hole–phonon interactions. For instance, when the effective mass of an electron is equal to that of a hole and the band offset of the conduction band is equal to that of the valence band, i.e. m∗ez = m∗hz , m∗e|| = m∗e|| , and ΔEc = ΔEv , the exciton–phonon interaction will become zero. As for the DP coupling (see Eq. (16.10)), the signs and magnitudes of each effective DP of both the conduction and the valence band play a very important role too. Using Fermi’s golden rule and the interaction Hamiltonian in Eq. (16.9), the rate of formation of an exciton with wave vector K|| from free electron–hole pairs due to a phonon emission in 𝜆-mode via J (J = D, P, O) coupling is obtained as [3, 4, 16] 1 2𝜋 ∑ J J W𝜆+ = |CJ𝜆 F𝜆+ (qz , q|| , k|| )|2 f𝛼h (K +q )−k f𝛼e (K +q )+k h || || || e || || || A0 ℏ q ,k ,q z

||

||,

× (fKex|| + 1)(n𝜆q + 1)𝛿(Ex − Ee−h + ℏ𝜔q𝜆 )

(16.16)

where f h , f ex , f e , and n𝜆q are the occupation numbers of holes, excitons, electrons, and phonons, respectively [3, 4]. Here, the exciton energy Ex and electron–hole-pair energy Ee–h are, respectively, given by Ex =

ℏ2 |K|| |2

Ee−h =

− Eb 2M||∗ ℏ2 |K|| + q|| |2 2M||∗

(16.17) +

ℏ2 |k|| |2 2𝜇||∗

(16.18)

where Eb is the exciton binding energy, and 𝜇||∗ = (1∕m∗e|| + 1∕m∗h|| )−1 is the reduced mass of the electron–hole pair. For simplicity, we assume that there are no excitons in the

16.3 Exciton Formation in QWs Assisted by Phonons

Exciton formation rate [s–1]

system in the initial state; i.e. fKex = 0. The densities of the photogenerated electrons and || holes can be considered to equal (N e = N h = N e–h ) in an intrinsic QW. We also assume that electrons and holes are at the same temperature; i.e. T e = T h = T e–h . Let us first discuss the rate of formation of an exciton due to the emission of an acoustic phonon in [001] GaAs/Al0.3 Ga0.7 As QWs of well width Lz = 80 Å at a lattice temperature T = 4.2 K. For comparing the individual rates due to the emission of an LA or TA phonon, we have calculated them from Eq. (16.16) for 𝜆 = LA and TA (TA1 + TA2) separately, and plotted them as a function of K|| at T e–h = 20 K for charge carrier densities N e–h = 1 ×1010 cm−2 and N e–h = 5 ×1010 cm−2 in Figures 16.1a,b, respectively. It is clear from Figures 16.1a,b that the rate of formation due to an LA phonon emission is comparable to that due to a TA phonon emission, though the rate of formation of an exciton via an LA phonon is slightly higher than that via a TA phonon for relatively small K|| , and the rate due to a TA phonon is slightly higher than that due to an LA phonon for a relatively large K|| . Also, the maximum value of the rate via TA phonon emission is relatively larger than that via LA phonon emission. For GaAs QWs, when considering only the DP coupling, holes in the valance band interact with both TA and LA phonons, whereas electrons in the conduction band interact with only LA phonons. Consequently, an exciton, which is a bound state of an electron and a hole, interacts with both LA and TA phonons due to DPs. In this case, it is clear from Eq. (16.10) that the sign of the DP of the charge carriers is very important in determining the strength of exciton–phonon interaction. A very strong electron–phonon interaction or hole–phonon interaction does not automatically mean a very strong exciton–phonon interaction. It also depends on the signs of the two coupling coefficients. For instance, if the effective DPs of electrons and holes are of comparable magnitude for any phonon mode but have opposite signs, then the strength

1012

1012

1010

1010

108

108

106

106

104

104 0

1

2

3

4

0

1

2

3

KII [108 m–1]

KII [108 m–1]

(a)

(b)

4

Figure 16.1 Formation rate of an exciton via LA (solid curve) and TA (dash-dotted curve) phonon emissions in GaAs quantum wells plotted as a function of the center-of-mass wavevector K|| for Lz = 80 Å, T = 4.2 K, and T e–h = 20 K: (a) Ne–h = 1 ×1010 cm−2 and (b) Ne–h = 5 ×1010 cm−2 [4]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 221. Copyright (2001) Elsevier.

469

16 Excitonic Processes in Quantum Wells

of the corresponding exciton–phonon interaction is enhanced because the coupling coefficients get added, but if they have the same sign, then the exciton–phonon coupling will be very weak as the electron–phonon coupling will cancel out the hole–phonon coupling. Our results show that, 𝜏 f , the time of formation of an exciton defined by 1/𝜏f = ∑ J J,𝜆 W𝜆+ (J = D, P, and 𝜆 = LA, TA1 , TA2 ), at the maximum rate of formation varies from 145 to 714 ps for N e–h = 1 ×1010 cm−2 , and it is 6.2–28 ps for N e–h = 5 ×1010 cm−2 at the carrier temperature T e–h ranging from 20 to 80 K. Accordingly, the rate of formation of excitons is very sensitive to both carrier temperatures and densities. Figures 16.2a,b illustrate the dependence of the rate of exciton formation on the COM wave vector K|| of the exciton, charge carrier temperature T e–h , and density N e–h . We have taken into account both the TA and LA phonons due to DP and PE couplings for AC phonon processes. As can be seen from Figures 16.2a,b, the rate of formation is very sensitive to K|| , T e–h , and N e–h . We find from Figure 16.2a that the rate of formation decreases for K|| ≤2.24×108 m−1 whereas it increases for K|| ≥2.72× 108 m−1 with increasing T e–h at the charge carrier density N e–h = 1 ×1010 cm−2 . Figure 16.2a also illustrates that the maximum rate of formation occurs at a nonzero value of K|| in the range 2.16–2.24 ×108 m−1 , which corresponds to a kinetic energy of COM motion of about 20–22 meV. Figure 16.2b shows the rate of formation due to AC phonon emission as a function of T e–h and N e–h at K|| = 2.16 ×108 m−1 and indicates that the rate of formation increases very fast in the low-carrier-temperature region T e–h , with increasing photoexcited charge carrier density N e–h in comparison with that in the high-temperature region. At higher excitation densities, Auger processes may also contribute to the formation of excitons. In this case, an excited electron–hole pair may form an exciton by giving the excess energy to another excited charge carrier but not to phonons. However, the reverse process to dissociate excitons will also have an equal probability of taking place at a high excitation density, and will deplete the number of excitons thus formed. Therefore, the 8

2.0

(a)

T e–h

–h

3

Ne

[10 10

2

K] [10

0.0

9 8 7 6 5 4 3 2

m –1]

0.5

1

3

KII [1 8 0

4

2

1

0

0

9 8 7 6 5 4 3 2

2

0.1

5

4

1.5

4

W [1011 s–1]

6 W [109 s–1]

470

T e–h

m –2]

[10

K]

(b)

Figure 16.2 Formation rate of an exciton due to acoustic phonon emission in GaAs QWs (Lz = 80 Å) (a) as a function of center-of-mass wave vector K|| and charge-carrier temperature T e–h at a charge carrier density of Ne–h = 1 × 1010 cm−2 and (b) as a function of charge-carrier density Ne–h and temperature T e–h at K|| = 2.16 × 108 m−1 [15]. Source: I.-K. Oh and J. Singh (2001). Int. J. Mod. Phys. B, 15, 3660. Reproduced with permission from World Scientific.

16.3 Exciton Formation in QWs Assisted by Phonons

Auger process of exciton formation may not be able to establish the excitonic density as effectively as the phonon emission processes at low temperatures. In Figure 16.3 are plotted the calculated values for the formation rate of an exciton as a function of the exciton wavevector KII for different values of charge carrier temperatures in QWs with a well width of Lz = 80 Å and a charge carrier density of N e–h = 1 × 1010 cm−2 . At charge carrier temperatures T e–h ≤30 K, the formation of an exciton occurs dominantly at KII = 0, but for T e–h ≥30 K it occurs at nonzero KII ≅ 0.0082 × 1010 m−1 . Our results indicate that, first, the formation rate of an exciton decreases with increasing K_, and then it increases to a peak value at about KII ≅ 0.0082 × 1010 m−1 , after which it decreases continuously as the exciton wavevector KII increases (see Figure 16.3). The results in Figure 16.3 are in agreement with those of Damen et al. [13] that excitons are formed dominantly at K_ ≠ 0, but our results also suggest that the formation process depends on T e–h and N e–h . For instance, according to Figure 16.3, an exciton can be dominantly formed at KII = 0 at low carrier temperatures T e–h ≤30 K for a carrier density of N e–h = 1 × 1010 cm−2 . However, at higher carrier temperatures, the situation can be different. For example, for a fixed carrier density of N e–h = 1 × 1010 cm−2 , by raising the temperature to T e–h = 40 K, we find that a formation time of 𝜏 f = 133 ps at KII = 0, which is slightly slower than 𝜏 f = 118 ps found at KII = 0.0082 × 1010 m−1 . Likewise, if the temperature is raised further to T e–h = 50 K, it is found again that 𝜏 f = 40 ps at KII = 0, higher than 𝜏 f = 30 ps at K|| = 0.0074 × 1010 m−1 . This trend seems to continue at even higher temperatures as well (see Figure 16.3). Figure 16.4 shows the formation rate of an exciton as a function of the exciton wavevector K|| at T e–h = 60 K for three different well widths, 80, 150, and 250 Å, and two different charge carrier densities. For all three QWs, the maximum formation rate occurs at nonzero K|| . This is obvious from Figure 16.4, which shows that for the QW with widths of 80 and 150 Å, the maximum formation rate

Exciton formation rate [s–1]

1012

1011

1010

109

108 0.000

0.010

0.020

0.030

0.040

KII [1010 m–1]

Figure 16.3 Formation rate of an exciton in GaAs quantum wells as a function of the center-of-mass wavevector KII for Lz = 80 Å, Ne−h = 1 ×1010 cm−2 , and different temperatures, i.e. T e−h = 30 K (–), T e−h = 40( . . . .), T e−h = 50 K (- - -), T e−h = 60 K (-•-), Te-h = 70 K (–· · ·–· · ·), and T e−h = 80 K (- -) [3]. Source: Reprinted with permission from I.-K. Oh et al., Phys. Rev. B 62, 2045 (2000) with permission of the American Physical Society.

471

16 Excitonic Processes in Quantum Wells

1013 Exciton formation rate [s–1]

472

1012

1011

1010

109 0.000

0.010

0.020 KII

[1010

0.030

0.040

m–1]

Figure 16.4 Formation rate of an exciton in GaAs QWs as a function of the center-of-mass wavevector K|| at T e−h = 60 K for different well widths and charge carrier densities, Lz = 80 Å, Ne−h = 5 ×1010 cm−2 (—), Lz = 150 Å, Ne−h = 5 ×1010 cm−2 (…), Lz = 250 Å, Ne−h = 5 ×1010 cm−2 (– – –), Lz = 80 Å, Ne−h = 1 ×1010 cm−2 (– • – •), Lz = 150 Å, Ne−h = 1 ×1010 cm−2 ( — • • •), Lz = 250 Å, Ne−h = 5 ×1010 cm−2 (– –) [3]. Source: Reprinted with permission from I.-K. Oh et al., Phys. Rev. B, 62, 2045 (2000) with permission of the American Physical Society.

occurs at K|| = 0.0074 ×1010 m−1 and 0.0016 ×1010 m−1 , at the carrier density of N e–h = 1 ×1010 cm−2 , which changes to 0.0081 ×1010 m−1 and 0.0025 ×1010 m−1 , respectively, at the carrier density of 5 ×1010 cm−2 . However, for the well Lz = 250 Å, the maximum rate occurs at K|| = 0.0041 ×1010 m−1 for both the carrier densities. For the QWs of widths Lz = 80 Å and Lz = 250 Å, the formation rates first decrease slightly, and then increase to a maximum, and then decrease again continuously as K|| increases. However, the formation rate for Lz = 150 Å QWs first increases to its maximum and then decreases continuously with increasing K|| . The formation time corresponding to the maximum formation rate in the QW with Lz = 150 Å is obtained as 𝜏 f = 17 ps and 0.7 ps at carrier densities of N e–h = 1 × 1010 cm−2 and N e–h = 5 × 1010 cm−2 , respectively. This trend – that the formation time corresponding to the maximum rate becomes shorter with an increase in the carrier density – is obtained in all the three well widths. In Figure 16.5, we have plotted the rate of formation of an exciton as a function of the charge carrier density N e–h , by varying it from 1 × 108 to 5 × 1010 cm−2 at T e–h = 60 K for the three well widths. From the calculated results of the formation rate, we have carried x out curve fittings using the relation W (N e–h ) = bNe−h , where b and x are fitting parameters. Our results give x ≈ 2, as shown in Figure 16.5. In other words, our theory shows a square-law dependence of the formation rate of an exciton on N e–h . This provides a theoretical confirmation of the observed experimental square-law dependence of the photoluminescence on excitation density obtained by Strobel et al. [17]. For a comparison between acoustic and LO phonon processes, we have plotted in Figure 16.6a,b the rate of formation of an exciton calculated as a function of K|| at T e–h = 20, 50, and 80 K for two different charge carrier densities N e–h = 1 × 1010 cm−2 and N e–h = 5 × 1010 cm−2 , respectively, and at a lattice temperature of 4.2 K. The —, ......, and – ⋅ –⋅ curves represent the formation rates due to acoustic phonon emission, and

Exciton formation rate [s–1]

16.3 Exciton Formation in QWs Assisted by Phonons

1.5 × 1012

1.0 × 1012

x = 2.024

+ + + x = 1.991

x = 2.000

x = 1.993

x = 2.059

x = 1.986

x = 1.993 5.0 × 1011

x = 2.002 x = 1.946

10

20 Ne–h

30 [109

40

50

cm–2]

Figure 16.5 Formation rate of an exciton in GaAs QWs as a function of the carrier density Ne−h at T e−h = 60 K for different well widths and center-of-mass wavevectors, Lz = 80 Å (–•–•), Lz = 150 Å (—Δ—), Lz = 250 Å (++++) at K|| = 0, Lz = 80 Å (–––), Lz = 150 Å (– •••), Lz = 250 Å (—) at K|| = 1010 cm−1 , and Lz = 80 Å (— ◊ —), Lz = 150 Å (– –) and Lz = 250 Å (•••••) at K|| = 0.0075 × 1010 cm−1 . The numbers x in the figure are obtained from the curve fitting W (Ne−h ) = b x Ne−h [3]. Source: Reprinted with permission from I.-K. Oh et al., Phys. Rev. B, 62, 2045 (2000) with permission of the American Physical Society.

the – ⋅⋅⋅, – – –, and — — curves are the formation rates due to LO phonon emission at T e–h = 20, 50, and 80 K, respectively. The results of Figure 16.6 show that the rate of formation of excitons emitting acoustic phonons first increases with increasing K|| to a peak value and then decreases with increasing K|| . It is also found that the charge carrier temperature (T e–h ) dependence of the rate of formation changes before and after the peak value. The rate of formation, including its peak position, decreases with increasing T e–h , but at a certain K|| after the peak value, the rate increases with increasing T e–h (see Figure 16.6). This dependence of the formation rate of excitons on the carrier temperature via acoustic phonons is different from that via LO phonons: in the latter case, the rate of exciton formation increases with increasing T e–h for all K|| [3]. As shown in Figure 16.6, the maximum rate of formation due to acoustic phonon emission occurs at relatively larger value of K|| ≈ 2.24 × 1010 m−1 in comparison with that due to LO phonon emission at K ≈ 0.82 × 108 m−1 for N e–h = 5 × 1010 cm−2 and T e–h = 50 K. In other words, our results suggest that the rate of formation of hot excitons (with large K|| values) due to acoustic phonon emission is much faster at T e–h ≤40 K, but then as T e–h approaches ∼50 K, it becomes comparable with the rate of formation due to LO phonons. At T e–h ≥50 K, the formation of an exciton via LO phonon emission becomes more efficient in comparison with that via acoustic phonon emission, which agrees reasonably well with Piermarocchi et al.’s result [18, 19]. Our results also show that the rate of formation due to acoustic phonon emission is relatively more sensitive to the COM wavevector K|| of excitons compared with that due to LO phonon emission. In particular, we have found that excitons are formed mainly due to LO phonon interaction at relatively small values of K|| , but at relatively large K|| the acoustic phonon interaction becomes dominant. For instance, the LO phonon process dominates for K|| ≤2.0 × 108 m−1 and the acoustic phonon process

473

16 Excitonic Processes in Quantum Wells

Exciton formation rate [s–1]

474

1012

1012

1010

1010

108

108

106

106

104

104 0

1

2

3

4

0

1

2

3

KII [108 m–1]

KII [108 m–1]

(a)

(b)

4

Figure 16.6 Formation rate of an exciton in GaAs QWs is plotted as a function of the center-of- mass wavevector K|| for Lz = 80 Å, T = 4.2 K with (a) Ne–h = 1 × 1010 cm−2 and (b) Ne–h = 5 × 1010 cm−2 . The —, ⋅⋅⋅⋅⋅, and –⋅ –⋅ curves correspond to the formation rate due to acoustic phonon emission at T e–h = 20, 50, and 80 K, respectively. The – ⋅⋅⋅, – – –, and – – curves correspond to the formation rates due to LO phonon emission at T e−h = 20, 50, and 80 K, respectively. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 221. Copyright (2001) Elsevier.

for K|| ≥2.0 × 108 m−1 at T e–h = 40 K and T = 4.2 K. The maximum rate of formation due to LO phonon emission is at K|| ∼ (0.74–0.82) × 108 m−1 (see Figure 16.6), which corresponds to a much smaller kinetic energy of about 2–3 meV in comparison with 20–22 meV obtained for the acoustic phonon process. This can be explained as follows: when the energy of photoexcited free electron–hole pairs is high enough, excitons formed at relatively small values of K|| are mainly due to the LO phonon process, whereas those formed at relatively large values of K|| are due to the acoustic phonon process because the energy involved in the formation of an exciton due to LO phonon emission is much larger than that due to AC phonon emission. In other words, as an excited electron–hole pair in a given state loses more energy in an LO phonon emission than that in an acoustic phonon emission, the exciton formed due to LO phonon emission has less kinetic energy than that formed due to acoustic phonon emission.

16.4 Nonradiative Relaxation of Free Excitons As discussed in the previous section, excitons formed from free electron–hole pairs have nonzero K|| exciton states. These excitons are “hot,” so they will relax, first, nonradiatively by emitting phonons, and then recombine radiatively by emitting photons. Therefore, the process of relaxation plays a significant role in the properties of photoexcited semiconductor systems. There are two important processes in the relaxation in QWs, intraband and interband. Here, we first consider intraband transitions of relaxation and then interband transitions.

16.4 Nonradiative Relaxation of Free Excitons

16.4.1

Intraband Processes

The interaction Hamiltonian associated with the intraband transition of a 1s exciton by interaction with a l-mode phonon via J (J = D, P, O) couplings can be written in the second quantized form as [5, 20] ̂J H (q , q ) = ex−ph,𝜆 || z



J J ̂† ̂† ̂ ̂ ̂ ̂ CJ𝜆 [G𝜆− (q|| , qz )B (q|| , qz )B b + G𝜆+ b† ] B B K +q K|| 𝜆q K −q K|| 𝜆q

K||

||

||

||

||

(16.19) where the subscripts + and − correspond to phonon emission and absorption proJ J∗ D cesses, respectively. It should also be noted that G𝜆+ (q|| , qz ) = G𝜆− (q|| , qz ).1 G𝜆− (q|| , qz ), O P G𝜆− (q|| , qz ), and GLO− (q|| , qz ) are obtained as [5] D G𝜆− (q|| , qz ) = i[Ξ𝜆c (q)Fe (qz )G(𝛼h q|| ) − Ξ𝜆v (q)Fh (qz )G(−𝛼e q|| )]

(16.20)

P G𝜆− (q|| , qz ) = i[Fe (qz )G(𝛼h q|| ) − Fh (qz )G(−𝛼e q|| )]

(16.21)

O GLO− (q|| , qz ) = Fe (qz )G(𝛼h q|| ) − Fh (qz )G(−𝛼e q|| )]

(16.22)

Here the form G(𝛼q|| ) is obtained as [5] G(𝛼q|| ) =

dr|| ei𝛼q|| .r|| |𝜙x (r|| )|2



(16.23)

Using the variational wave function of Eq. (16.15), we get an analytical expression for the form factor G as [5] [( ]−3∕2 ) 𝛼q|| 2 +1 (16.24) G(𝛼q|| ) = 2𝛽 Then, from Eq. (16.19) and Fermi’s golden rule, the relaxation rate of an exciton with wavevector K|| via J (J = D, P, O) couplings due to emission (absorption) of a phonon is given by [5] ) ( ∑ J 1 1 𝛿(E(K|| ∓ q|| ) W±J,𝜆 (K|| ) = W𝜆 (q|| , qz )(fKex|| ∓q|| + 1)fKex|| × n𝜆q + ∓ 2 2 q ,q ||

z

− E(K|| ) ± ℏ𝜔q , 𝛌 )

(16.25)

where the upper (lower) sign corresponds to the emission (absorption) of a phonon and W𝜆J (q|| , qz ) is given by [5] W𝜆J (q|| , qz ) =

2𝜋 𝜆 J 2𝜋 𝜆 J | CJ G𝜆+ (q|| , qz )|2 = | CJ G𝜆− (q|| , qz )|2 ℏ ℏ

(16.26)

Here, we study the total phonon emission from all excitonic relaxation processes as functions of the phonon wavevector q, exciton density, and exciton temperature. Taking J J 1 In References [5, 20], we have used F𝜆∓ for G𝜆∓ .

475

16 Excitonic Processes in Quantum Wells

into account the reverse excitonic process involving phonon emission, the net phonon emission rate of a 𝜆-mode phonon with wavevector q can be written as 𝜕n𝜆q 𝜕t

=

∑ J,K||

W𝜆J (q|| , qz ){[(fKex|| −q|| + 1)fKex|| (n𝜆q + 1) − (fKex|| + 1)fKex|| −q|| n𝜆q ]

× 𝛿(E(K|| − q|| ) − E(K|| ) + ℏ𝜔q,𝛌 ) − [( fKex|| +q|| + 1)fKex|| n𝜆q − (fKex|| + 1) fKex|| +q|| (n𝜆q + 1)]𝛿(E(K|| + q|| ) − E(K|| ) + ℏ𝜔q,𝛌 )}

(16.27)

The phonon emission rate obtained in Eq. (16.27) is the same as that derived by Vass [21] for fKex ≪ 1 and using the same coupling constants. Furthermore, if we ignore the || excitonic transitions K|| → K|| − q|| and K|| ← K|| − q|| and use the same coupling constants, then the rate in Eq. (16.27) also agrees with that obtained by Takagahara [22] provided that fKex ≪ 1. Thus, the present result can be regarded to be more general and || applicable to calculating the rates of all three types of phonon emissions. In Figure 16.7a we show the rate of an LA phonon emission due to the DP interaction as a function of the direction of phonon emission in spherical polar coordinates (angles 𝜃 and ∅). In this case, the rate is not very sensitive to the angle ∅, but it is quite sensitive to the 𝜃 angle. The rate has a maximum value of about 2.1 × 109 s−1 at 𝜃 ∼ 7∘ , but it is nearly independent of ∅. In Figure 16.7b, we show the rate of LA phonon emission using the PE interaction. Unlike Figure 16.7a with DP, the rate of scattering depends on both the q and f angles. The maximum rate of phonon emission is about 1.1 × 107 s−1 at 𝜃 ∼ 61∘ and ∅ ∼ 45∘ . Likewise we have also plotted the rate of emission of TA phonons in Figure 16.8a with DP and in Figure 16.8b with PE coupling. Although the maximum rate of scattering

5

LA Phonon (DP)

Phonon emission rate [107 s–1]

Phonon emission rate [109 s–1]

476

4 3 2 1

LA Phonon (PE)

3 2 1

0

40 θ [°]

20 (a)

60

80

ϕ

ϕ

0 40 30 20 10 0

[ °]

[ °]

0 40 30 20 10 0

4

0

40 θ [°]

20

60

80

(b)

Figure 16.7 LA phonon-emission rates due to (a) deformation potential (DP) and (b) piezoelectric (PE) couplings, as a function of phonon-emission angles for GaAs/Al0.3 Ga0.7 As QWs with well width Lz = 80 Å, phonon energy E ph = 0.2 meV, exciton density Nex = 5 × 1010 cm−2 , exciton temperature T ex = 20 K, and lattice temperature T = 4.2 K [5]. Source: Reprinted from I.-K. Oh and J. Singh, J. Lumin., 85, 233. Copyright (2000) Elsevier.

6

TA Phonon (DP)

Phonon emission rate [107 s–1]

Phonon emission rate [109 s–1]

16.4 Nonradiative Relaxation of Free Excitons

4

2

3 2 1

0

40 θ [°]

20 (a)

60

80

ϕ

ϕ

0 40 30 20 10 0

[ °]

[ °]

0 40 30 20 10 0

TA Phonon (PE) 4

0

60 40 θ [°]

20

80

(b)

Figure 16.8 TA phonon emission rates due to (a) deformation potential (DP) and piezoelectric (PE) couplings, as a function of phonon-emission angles for GaAs/Al0.3 Ga0.7 As QWs with well width Lz = 80 Å, phonon energy E ph = 0.2 meV, exciton density Nex = 5 × 1010 cm−2 , exciton temperature T ex = 20 K, and lattice temperature T = 4.2 K [5]. Source: Reprinted from I.-K. Oh and J. Singh, J. Lumin., 85, 233. Copyright (2000) Elsevier.

with the emission of TA phonons due to DP is higher (≈ 5.4 × 109 s−1 at 𝜃 ∼ 4∘ ), the directional dependence is similar to that of LA phonons shown in Figure 16.7a. However, TA phonon emission (Figure 16.8b) due to PE coupling is quite different from LA phonon emission (Figure 16.7b). In Figure 16.8b, one can see three peaks of emission of TA phonons. The first maximum at 𝜃 ∼ 37∘ and ∅ ∼ 45∘ , the third maximum at 𝜃 = 90∘ and ∅ = 45∘ corresponds to TA1 phonons, and the second maximum at 𝜃 ∼ 37∘ and ∅ ∼ 0∘ is from TA2 phonons. From our calculations, we find that LA and TA phonons are equally effective in excitonic processes of GaAs QWs. Furthermore, the emission rate of TA phonons due to DP is slightly higher than that of LA phonons. This result is quite different from those obtained earlier [21–23], because the anisotropy effect has not been taken into account previously. As a result, only the contribution of exciton–LA phonon interaction has been taken into account in earlier works [21–23]. In Figure 16.9, we have plotted the rate of LO phonon emission as a function of the angle 𝜃 and the phonon wavevector q, which gives a peak rate of about 1.3 × 109 s−1 at q = 3 × 108 m−1 and 𝜃 ∼ 90∘ , which is the plane of the QW. As the exciton–LO phonon interaction is isotropic in the x–y plane, the rate of scattering is independent of the angle f. The results shown in Figures 16.7–16.9 illustrate the point that exciton scattering is dominated by the emission of acoustic phonons due to DP at low 𝜃 angles, whereas LO phonon processes become dominant at high 𝜃 angles. The exciton scattering is dominated by acoustic phonons at small phonon wavevectors q ≪ 3 × 108 m−1 , but LO phonons dominate near q = 3 × 108 m−1 . It is to be remembered that the phonon wavevector is obtained from the momentum conservation equal to the difference between exciton wavevectors in the initial and final states (|K|| − K|| ′ | = q). Therefore, for LO phonon emission, we have used a fixed energy (e.g. ℏ𝜔LO = 36.2 meV for GaAs), and hence the rate has a peak value at a specific q value obtained from the value of momentum conservation. However, in the case of acoustic phonons, the phonon energy

477

16 Excitonic Processes in Quantum Wells

Phonon emission rate [109 s–1]

Figure 16.9 LO phonon-emission rate as a function of 𝜃 and q for GaAs/Al0.3 Ga0.8 As QWs with Lz = 80 Å, Nex = 8 × 1010 cm−2 , T ex = 80 K, and T = 4.2 K [5]. Source: Reprinted from I.-K. Oh and J. Singh, J. Lumin., 85, 233. Copyright (2000) Elsevier.

LO Phonon

2.6

1.5

1.0

0.5

An gle

θ[

°]

0.0 80 60 40 20 0 0.02

0.04

0.06

0.08

0.10

Wavevector q [1010 m–1]

depends linearly on the phonon wavevector, and therefore we get the phonon emission rate as a decreasing function of q. The calculated results of the dependence of the phonon emission rate on the exciton temperature suggest that the rate of LO phonon emission increases as the exciton temperature increases from 5 to 200 K. However, the rate of acoustic phonon emission first increases rapidly for exciton temperature 5 < T ex < 17 K at exciton densities in the range 5 × 109 to 8 × 1011 cm−2 , and then it decreases for 17 ≈ 19 < T ex < 200 K. It is also to be noted that our results suggest that the LO phonon emission rate is dominant at higher exciton temperatures, whereas the acoustic phonon emission rate is dominant at lower temperatures. We have plotted the rates of phonon emission in Figure 16.10 as a function of T ex at the exciton density N e–h = 5 × 1010 cm−2 . Accordingly, LA and 1.0 Phonon emission rate [1010 s–1]

478

0.8

LA Phonon TA Phonon

0.6

LO Phonon

0.4 0.2 0.0 20

40

60

80 Tex [k]

100

120

140

Figure 16.10 Phonon-emission rate as a function of T ex for GaAs/Al0.3 Ga0.7 As QWs with Lz = 80 Å, Nex = 5 × 1010 cm−2 , and T = 4.2 K [5]. Source: Reprinted from I.-K. Oh and J. Singh, J. Lumin., 85, 233. Copyright (2000) Elsevier.

16.4 Nonradiative Relaxation of Free Excitons

TA phonon processes are relatively more efficient at temperatures T ex ≤ 87 K and T ex ≤ 95 K, respectively, in comparison with LO phonon processes. However, as the temperature increases, LO phonon processes dominate both LA and TA phonon processes. As the exciton temperature is directly proportional to the exciton kinetic energy at a given exciton density, this suggests that if the exciton scattering takes place at higher exciton wavevector, the LO phonon emission will be dominant. At lower exciton wavevectors, acoustic phonon emission will be dominant. This agrees very well with the experimental results reported for GaAs QWs [24], where a faster exciton formation time is observed due to the involvement of LO phonons at higher exciton wavevectors. It is important to discuss the dependence of the phonon emission rate on the exciton density N ex . Our results suggest that at a given temperature of the exciton and lattice, the rate of phonon emission first increases with N ex and then shows a kind of saturation at higher exciton densities. For example, for T ex = 20 K and a lattice temperature of 4.2 K, the rate of emission of LA phonons with energy 0.2 meV becomes saturated at an exciton density N ex ∼ 2.5 × 1012 cm−2 . Likewise the rate of LO phonon emission becomes saturated at N ex ∼ 3.0 × 1012 cm−2 for T ex = 50 K and a lattice temperature of 4.2 K. 16.4.2

Interband Processes

In semiconductor QWs, depending on the well width and structure of the confinement potential, there exist several electron and hole sub-bands through which the exciton formation and relaxation may occur. For example, in GaAs/Alx Ga1−x As QWs (x = 0.3), there is only one electron sub-band and only one heavy-hole sub-band for a well width of Lz ≈41 Å.2 In this case, the relaxation by inter-sub-band transition is not possible. In QWs of width Lz ≈ 42–47 Å, there are two electron sub-bands (ne = 1, 2) and one heavy-hole sub-band (nh = 1). Then there are two electron sub-bands (ne = 1, 2) and two heavy-hole sub-bands (nh = 1, 2) in QWs of well width Lz ≈ 48–82 Å. QWs of even larger widths have increasingly more sub-bands. The number of sub-bands increases with increasing well width, but the energy separation between adjacent sub-bands decreases. This is an important point to consider in conserving the energy and momentum in inter-sub-band transitions. In the study of dynamics of excitons in intrinsic QWs, both processes, exciton formation and relaxation, are important. When charge carriers in QWs are created by optical excitations, the transitions occurring at Δn = ne − nh = 0 are called dipole allowed, and for Δn ≠ 0 dipole forbidden [25]. Among the transitions with Δn ≠ 0, those with Δn = even are parity-allowed and those with Δn = odd are parity-forbidden [26]. Thus, Δn = even (≠ 0) transitions are called parity-allowed forbidden transitions and Δn = odd are called parity-forbidden forbidden transitions. The parity-allowed forbidden transitions with Δn = even have been observed in parabolic QWs [27], and the parity-forbidden forbidden transitions with Δn = odd have been observed in square-well QWs at very high excitation intensities [26]. This illustrates quite clearly that an exciton can be excited through several possible combinations of electron and hole sub-bands. Each excitonic state corresponding to an optically allowed 2 For the calculations of electron and hole sub-band structures, the band discontinuities ΔEc = 1.1x eV and ΔEv = 0.17x eV (x = 0.3) for the conduction and valence bands, respectively, have been used as given in Reference [25].

479

480

16 Excitonic Processes in Quantum Wells

2e

ΔEc

1e

(2e, 2h)

1s(2e, 2h)

(2e,1h)

1s(2e,1h)

(1e, 2h)

1s(1e, 2h)

Eg

(1e,1h)

1s(1e,1h) 1h

ΔEv

Eg

2h lg> (a)

(b)

Figure 16.11 Schematic illustration of (a) the structure of electron and heavy-hole sub-bands and (b) energy levels of the 1s exciton states and continuum states of free electron–hole pairs (shaded) formed from possible combinations of electron and hole sub-bands for a GaAs QW Lz = 48–82 Å Ref. [6]. Source: Reprinted with permission from I.-K. Oh, J. Singh, and A.S. Vengurlekar, J. Appl. Phys., 91, 5796. Copyright (2002) American Institute of Physics.

transition will have its own discrete Rydberg series. Only the exciton state formed from the lowest electron (ne = 1) and hole (nh = 1) sub-band is well separated from the continuum states of free electron–hole pairs (see Figure 16.11). In other words, an exciton formed from higher electron or/and hole sub-bands, e.g. ne = 2 and nh = 1 or ne = 2 and nh = 2 sub-bands, is embedded in the continuum states of free electron–hole pairs formed from lower electron and hole sub-bands [28] (see Figure 16.11b). It is known that a resonant Coulomb interaction between the discrete exciton states and the continuum states of free electron–hole pairs results in Fano interference [29], which gives rise to an asymmetric absorption line shape called the Fano profile. As a result, the study of the dynamics of excitons or free electron–hole pairs in QWs involving inter-sub-band transitions due to LO phonon interaction becomes quite complicated. The relaxation of excitons by emitting photons (photoluminescence) also follows the same selection rules as in excitation. However, the excitonic relaxation by phonon emission is different. The relaxation of a higher-order exciton 1s (ne e, nh h) to any lower-order free electron–hole pair (n′ e e, n′ h h) by emitting phonons can be a multi-channel process. Here, 1s (ne e, nh h) denotes the 1s excitonic state created by exciting an electron in the ne -th and a hole in the nh -th sub-bands, and (n′ e e, n′ h h) denotes a free electron–hole pair with an electron in the n′ e -th and a hole in the n′ h -th sub-bands. For example, when a 1s (2e, 2h) exciton formed in ne = nh = 2 sub-bands relaxes to a (1e, 1h) free electron–hole pair in ne = nh = 1 sub-bands, it can do so in one step as (ne = nh = 2 → ne = nh = 1) or in two steps with two possibilities (ne = nh = 2 → ne = 1, nh = 2 → ne = nh = 1) or (ne = nh = 2 → ne = 2, nh = 1 → ne = nh = 1). However, as explained later, the orthogonality of the sub-band states does not allow a direct excitonic relaxation from 1s (2e, 2h) to (1e, 1h) by an LO phonon emission. Here, we consider only LO phonon processes, but the same selection rules apply to acoustic (AC) phonons as well. In the two-step process, energy conservation may exclude one of the steps. This happens when the energy separation between the adjacent sub-bands is less than the LO phonon energy.

16.4 Nonradiative Relaxation of Free Excitons

The interaction Hamiltonian for the relaxation of an exciton into a free electron–hole pair associated with interband transitions due to LO phonon coupling can be written in the second quantized form as [6] ∑ nm † ̂m ̂ni mi ̂ ex→eh = √1 ̂ H COLO [Fnfi mfi − (q|| , qz , k|| )d bq|| B a†n 𝛼 (K −q )+k ̂ I K|| f ,𝛼h(K|| −q|| )−k|| f e || || || A0 K|| ,k|| ,q|| ,qz nm † ̂m ̂ni mi ] ̂ + Fnfi mfi + (q|| , qz , k|| )d b†q|| B a†nf 𝛼 (K +q )+k ̂ K f ,𝛼h(K +q )−k ||

||

||

e

||

||

||

(16.28)

||

̂nm is the annihilation operator of an exciton formed from the n-th electron where B K|| and m-th hole sub-bands with COM momentum wavevector K|| in the plane QW, and ̂† ̂ a†n𝛼 K + k and d denote the electron and hole creation operators in the n-th elecm𝛼 K − k e

||

||

h

||

||

nm

tron and m-th hole sub-bands, respectively. Fnfi mfi ∓ is given by [6]: nn

nm

mm

nm

nm

Fnfi mfi ∓ (q|| , qz , k|| ) = [Fe i f (±qz )G1 i i (k|| ± 𝛼h q|| )𝛿mi ,mf − Fh i f (±qz )G1 i i (k|| ∓ 𝛼e q|| )𝛿ni ,nf ] (16.29) Here, Fjnm

(j = e, h) are the form factors associated with the inter-sub-band transitions and they are obtained as [6] Fjnm (qz ) =

j∗



j

dzj ∅m (zj )∅n (zj )eiqz zj ,

(16.30)

j = e, h

−𝛽nm r|| √ and G1nm is obtained using the variational wave function3 ∅nm of the x (r|| ) = 𝛽nm 2∕𝜋 e 1s exciton formed from the n-th electron and m-th hole sub-bands as [6] [( ]−3∕2 √ ) ∣ K|| + 𝛼q|| ∣ 2 8𝜋 nm nm i(K|| +𝛼q|| ).r|| dr|| ∅x (r|| )e = +1 G1 (K|| + 𝛼q|| ) = 2 ∫ 𝛽nm 𝛽nm

(16.31) The first term of Eq. (16.28) corresponds to the inter-sub-band transition of an exciton due to an LO phonon absorption and the second term corresponds to that due to an LO phonon emission. However, in a relaxation process at low crystal temperatures as considered here, only the second term of the emission of an LO phonon needs to be considered. It is to be noted here that, according to Eq. (16.29), the inter-sub-band relaxation Hamiltonian is nonzero only if one of the charge carriers (electron or hole) remains in its sub-band during the transition and the other goes through an inter-sub-band transition. In Eq. (16.29), the first term is zero if mi ≠ mf and the second term is zero if ni ≠ nf . In other words, a direct inter-sub-band relaxation from a 1s (2e, 2h) (ni = mi = 2) exciton to a (1e, 1h) (nf = mf = 1) free electron–hole pair is not allowed. This is due to the orthogonality of the sub-band states. Then, the total rate of relaxation of an exciton with wavevector K|| is obtained as [6] 1 2𝜋 ∑ nm nm |C LO F i i (q , q , k )|2 [1 − fmhi (𝛼h (K|| + q|| ) − k|| )] Wnfimfi + (K|| ) = A0 ℏ q q k o nf mf + || z || || z ||

nm

nm

f f − Eexi i + ℏ𝜔LO ) × [1 − fnei (𝛼e (K|| + q|| ) + k|| )]fnexi mi (K|| )[n(q) + 1]𝛿(Ee−h (16.32)

3 In Reference [6], we have used 𝛽 for 𝛽 nm .

481

482

16 Excitonic Processes in Quantum Wells ex where fne , fmh , fnm and n(q) are the occupation numbers of electrons in the n-th sub-band, holes in the m-th sub-band, excitons formed from them, and phonons, respectively, and nf mf

Ee−h =

ℏ2 |K|| + q|| |2 2M||∗

+

ℏ2 k|| 2 2𝜇||∗

h + Ene f + Em f

(16.33)

and n mi

Eexi

nm

=

ℏ2 |K|| |2 2M||∗

n mi

− Eb i

h + Ene i + Em f

(16.34) j

where Eb i i is the exciton binding energy, and En (j = e, h) the energy of n-th sub-band. nm The total rate Wnfimfi+ (K|| ) thus obtained gives the inverse of the time of relaxation, 1/𝜏 0 ni mi (K|| ) = Wnf mf + (K|| ), of an exciton with wavevector K|| [6]. Using the conduction- and valence-bands discontinuities (see Figure 16.11a) as ΔEc = 1.1x eV and ΔEv = 0.17x eV [25], respectively, we have calculated the structure of electron and heavy-hole sub-bands, binding energy, and wave functions of a 1s exciton formed from the possible combinations of electron and hole sub-bands. There are two electron sub-bands and one hole sub-band at a well width in the range Lz = 42–47 Å, two electron and two hole sub-bands in the range Lz = 48–82 Å, and four electron and three hole sub-bands at Lz = 130 Å. Therefore, for Lz = 42–47 Å QWs, there is only one possible inter-sub-band relaxation of an exciton [1s (2e, 1h)] formed initially with an excited electron in ne = 2 and a hole in nh = 1 sub-bands to a free electron–hole pair [(1e, 1h)], with the electron moving down to the ne = 1 sub-band by emitting an LO phonon. For Lz = 48–82 Å QWs, there are in principle four possible 1s excitonic states and thus four channels of relaxation of an exciton formed initially in a higher sub-band (see Figure 16.11). If the exciton is formed initially in the exciton state [1s (2e, 2h)] with an electron in the ne = 2 and a hole in the nh = 2 sub-bands, then there are in principle two possible channels of relaxations: (1) the exciton relaxes to a free electron–hole pair [(2e, 1h)], where the electron remains in the ne = 2 sub-band but the hole relaxes from the nh = 2 to the nh = 1 sub-band and (2) the exciton relaxes to a free electron–hole pair [(1e, 2h)], where the hole remains in the nh = 2 sub-band but the electron relaxes from the ne = 2 to the ne = 1 sub-band. The third possible channel is when the exciton is initially created in the state [1s (2e, 1h)] by exciting an electron in the ne = 2 and a hole in the nh = 1 sub-bands; it can then relax to a free electron–hole pair [(1e, 1h)] such that the hole remains in the nh = 1 sub-band but the electron relaxes from the ne = 2 to the ne = 1 sub-band. The fourth possible channel of relaxation is when an exciton excited initially in the state [1s (1e, 2h)] with an electron in the ne = 1 and a hole in the nh = 2 sub-bands relaxes to a free electron–hole pair [(1e, 1h)], where the electron remains in the ne = 1 sub-band but the hole relaxes from the nh = 2 to the nh = 1 sub-band. However, since the energy separation between the first and second hole sub-bands for such QWs is in the range 24–34 meV, the inter-sub-band relaxation of excitons by LO phonon emission involving hole sub-bands is not possible. Thus, out of the four channels, only the second and third channels are possible in practice and satisfy the condition of energy conservation for the relaxation of excitons by LO phonon emission. For comparison with recent experimental results [30], we have also considered QWs of well width Lz = 130 Å where, as stated above, there are four electron and three hole sub-bands. In this case, we have only considered the cases where excitons are excited to

16.4 Nonradiative Relaxation of Free Excitons

the sub-bands ne ≤ 2 and nh ≤ 2 for comparing our results with the experimental ones. The energy separation between the ne = 1 and ne = 2 sub-bands is about 65.1 meV and that between the nh = 1 and nh = 2 sub-bands is about 12.7 meV. The binding energies of 1s(1e, 1h), 1s(1e, 2h), 1s(2e, 1h), and 1s(2e, 2h) excitons are 7.0, 6.2, 6.2, and 6.1 meV, respectively. Since the energy separation between nh = 1 and nh = 2 is only 12.7 meV, the inter-sub-band relaxations from 1s(2e, 2h) to (2e, 1h) and from 1s(1e, 2h) to (1e, 1h) by an LO phonon emission are not possible. The only possible inter-sub-band relaxations by LO phonon emission are from 1s(2e, 2h) to (1e, 2h) and 1s(2e, 1h) to (1e, 1h). Accordingly, as stated above, a direct relaxation from 1s(2e, 2h) to (1e, 1h) by emitting a single LO phonon is not possible; it appears to be a two-phonon two-step Lo phonon

AC phonon

AC phonon

process: 1s(2e, 2h) −−−−−−−−→ (1e, 2h) −−−−−−−−→ (1e, 1h) or 1s(2e, 2h) −−−−−−−−→ (2e, LO phonon

1h) −−−−−−−−→ (1e, 1h). In Figure 16.12, we have plotted the rate of relaxation of 1s(2e, 2h) excitons into (1e, 2h) free electron–hole pairs by emitting an LO phonon as a function of the COM wavevector K|| for three different exciton densities and at a crystal temperature of T = 8 K and exciton temperature T ex = 20 K in QWs of width 130 Å. Similar results are obtained for the relaxation of a 1s(2e, 1h) exciton state to a (1e, 1h) electron–hole pair state by LO phonon emission. As illustrated in Figure 16.12, the rate decreases with increasing COM wavevector K|| and decreasing exciton density N ex . The maximum rate of relaxation is found to occur at K|| = 0. From the calculation, we have thus obtained values for the relaxation time of 305 fs at an exciton density of N ex = 1 × 1010 cm−2 , 616 fs at N ex = 5 × 109 cm−2 , and 3.11 ps at N ex = 1 × 109 cm−2 . The corresponding times for relaxation from 1s(2e, 1h) to (1e, 1h) are obtained as 304 fs, 614 fs, and 3.10 ps. Although these relaxation times for exciton densities in the range N ex = 5–10 × 109 cm−2 are in good agreement with those obtained by Pal and Vengurlekar [30], these are not times of relaxation from 1s(2e, 2h) to (1e, 1h). Our calculations also reveal that the relaxation time is very sensitive to the density of excitons, which

Exciton relaxation rate [s–1]

4 × 1012

3 × 1012 Nex = 1 × 109 cm–2 Nex = 5 × 109 cm–2

2 × 1012

Nex = 1 × 1010 cm–2

1 × 1012

0 0.00

0.01

0.02

0.03

0.04

KII [1010 m–1]

Figure 16.12 Relaxation rate of a 1s (2e, 2h) exciton to a (1e, 2h) free electron–hole pair by LO phonon emission in GaAs QWs as a function of the center-of-mass wavevector K|| for Lz = 130 Å, T ex = 20 K, T = 8 K, and three different exciton densities, Nex = 1 × 109 cm−2 (—), Nex = 5 × 109 cm−2 (⋅⋅⋅⋅⋅), and Nex = 1 × 1010 cm−2 (– ⋅ – ⋅) [6]. Source: Reprinted with permission from I.-K. Oh, J. Singh, and A.S. Vengurlekar, J. Appl. Phys., 91, 5796. Copyright (2002) American Institute of Physics.

483

484

16 Excitonic Processes in Quantum Wells

1013

1012

Nex = 1 × 109 cm–2

1011

Nex = 5 × 109 cm–2 1010 10

Nex = 1 × 1010 cm–2 20

30

40

50

Tex [k]

Figure 16.13 Relaxation rate of a 1s(2e, 2h) exciton to a (1e, 2h) free electron–hole pair by LO phonon emission in GaAs QWs as a function of the exciton temperature T ex (K) for Lz = 130 Å, K|| = 0, T = 8 K, and three different exciton densities, Nex = 1 × 109 cm−2 (—), Nex = 5 × 109 cm−2 (⋅⋅⋅⋅⋅), Nex = 1 × 1010 cm−2 (– ⋅ – ⋅) [6]. Source: Reprinted with permission from I.-K. Oh, J. Singh, and A.S. Vengurlekar, J. Appl. Phys., 91, 5796. Copyright (2002) American Institute of Physics.

has not been studied experimentally. However, as the chemical potential depends on the exciton density, the relaxation time depends quite sensitively on the exciton density [6]. In Figure 16.13, we have plotted the dependence of the relaxation rate for the same process as in Figure 16.12 [1s(2e, 2h) to (1e, 2h)] on the exciton temperature in the range T ex = 10–50 K at T = 8 K, K|| = 0, and at three different exciton densities, N ex = 1 × 109 cm−2 , 5 × 109 cm−2 , and 1 × 1010 cm−2 for Lz = 130 Å QWs. The results show that the rate at K|| = 0 decreases as the temperature of excitons T ex increases at a given exciton density. This is because at lower T ex the probability of excitons occupying states at K|| = 0 is higher compared with that at higher T ex . As shown in Figure 16.13, the rate of relaxation from a 1s(2e, 2h) exciton to a (1e, 2h) e–h pair at an exciton density of N ex = 5 × 109 cm−2 decreases from 3.28 × 1012 to 6.45 × 1011 s−1 ; as a result, the corresponding relaxation time decreases from 305 fs to 1.55 ps when T ex increases from 10 to 50 K. In Figure 16.14a,b, we have plotted the rates of relaxation of 1s(2e, 1h) and 1s(2e, 2h) excitons, respectively, at T ex = 20 K and T = 4.2 K, K|| = 0 and the above three different exciton densities as a function of Lz from 42 to 82 Å. As there is only one hole sub-band in QWs of width Lz = 42–47 Å, no relaxation of 1s(2e, 2h) excitons is possible in these QWs, as shown in Figure 16.14b. According to Figure 16.14, the rate of relaxation of both 1s(2e, 1h) and 1s(2e, 2h) excitons increases with the increasing width of QWs. It is found that the relaxation time from 1s(2e, 1h) to (1e, 1h), e.g. at N ex = 5 × 109 cm−2 , decreases from 26.2 to 1.23 ps and that from 1s(2e, 2h) to (1e, 2h) decreases from 3.57 to 1.24 ps when Lz increases from 48 to 82 Å. Such an increase is also found in the process of inter-sub-band scattering of 2D electrons in QWs because the form factor of the inter-sub-band transitions increases with well width [31]. However, this is quite different from the excitonic relaxation through intra-sub-band transitions by emission of an acoustic phonon. In that case, the rate of relaxation decreases with increasing Lz [32]. These two distinct features may be used to identify whether the mechanism of relaxation in QWs is through inter-sub-band or intra-sub-band transitions.

16.5 Quasi-2D Free-Exciton Linewidth

1013

Exciton relaxation rate [s–1]

1013

1012

1011

1010

1012

1011

1010 50

60

70

80

50

60

Lz [Å]

Lz [Å]

(a)

(b)

70

80

Figure 16.14 Relaxation rate of (a) 1s(2e, 1h) exciton to (1e, 1h) free electron–hole pair and (b) 1s(2e, 2h) exciton to (1e, 2h) free electron–hole pair by LO phonon emission in GaAs QWs as a function of the well width (Lz ) of QWs at T ex = 20 K, K|| = 0, T = 4.2 K, and three different exciton densities, Nex = 1 × 109 cm−2 (–), Nex = 5 × 109 cm−2 (⋅⋅⋅⋅⋅), Nex = 1 × 1010 cm−2 (– ⋅ – ⋅) [6]. Source: Reprinted with permission from I.-K. Oh, J. Singh, and A.S. Vengurlekar, J. Appl. Phys., 91, 5796. Copyright (2002) American Institute of Physics.

In narrow GaAs QWs, the relaxation of an exciton through an inter-sub-band transition by emission of an LO phonon is possible only via the inter-sub-band transitions of electrons. This is because the energy separation between different hole sub-bands is less than the LO phonon energy of ℏ𝜔LO = 36.2 meV. However, for QWs of wider well widths, e.g. 500 Å, the relaxation of excitons through higher inter-sub-band transition of holes involving acoustic phonon emission may be possible. In this case, the energy separation between nh = 1 and nh = 2 hole sub-bands is 1.2 meV and that between nh = 1 and nh = 3 sub-bands is 3.2 meV, and then the excitation of electrons from the nh = 2 or nh = 3 sub-band is expected. It is found that the inter-sub-band relaxation of excitons occurs only when one of the charge carriers (electron or hole) remains in the same sub-band during the transition and the other goes through an inter-sub-band transition. A direct transition from 1s(2e, 2h) exciton to (1e, 1h) free electron–hole pair is not allowed through inter-sub-band transitions. Such a selection rule cannot possibly be identified experimentally. The rate of relaxation is found to depend on the excitation density very sensitively, which has not yet been investigated experimentally. We have also found that the rate of inter-sub-band relaxation increases with increasing well width of GaAs QWs from 42 to 82 Å.

16.5 Quasi-2D Free-Exciton Linewidth In optical spectroscopic experiments such as transmission spectroscopy [33], photoluminescence [34, 35], and four-wave mixing [36, 37], ideal noninteracting excitons

485

486

16 Excitonic Processes in Quantum Wells

show very sharp optical peaks in their line shapes. However, the exciton line in low-dimensional semiconductors is broadened due to exciton scattering by acoustic [37], LO [34] phonons, interface, and alloy disorders [35], excitons [38, 39], free charge carriers [40], etc. In particular, at low exciton densities, the exciton line shape is determined mainly by exciton–phonon interactions, which are responsible for homogeneous line broadening [34]. In this section, we present a theory to calculate the free exciton homogeneous linewidth and dephasing rate at low excitation densities due to exciton–acoustic phonon interaction as a function of the lattice temperature, exciton temperature, well width, and exciton density in QWs. ph The total rate of dephasing of a free exciton Wh (K|| ) due to acoustic phonon interaction (J = D, P) is obtained by calculating the rate of transition from its initial state with K|| to other exciton states by emitting and absorbing phonons as [7] ∑ ph W𝜆J (q|| , qz )[(fKex|| +q|| + 1)fKex|| n𝜆q|| 𝛿(E(K|| + q|| ) − E(K|| ) − ℏ𝜔q𝜆 )] Wh (K|| ) = J,𝜆,q||, qz

+ [(fKex|| −q|| + 1)fKex|| (n𝜆q|| + 1)𝛿(E(K|| − q|| ) − E(K|| ) + ℏ𝜔q𝜆 )]

(16.35)

where E(K|| ) = ℏ2 K||2 ∕2M||∗ is the kinetic energy of a free exciton associated with its COM, and 𝜔q𝜆 is the frequency of a 𝜆-mode acoustic phonon with a wavevector q. From the rate of transition obtained in Eq. (16.35), the homogeneous linewidth due to acoustic phonons can be expressed as [33] ph

ph

Γh (K|| ) = ℏWh (K|| )

(16.36)

To interpret the experimental results [33, 39, 41], usually this linewidth is assumed to ph be linearly dependent on the lattice temperature T as Γh = 𝛾AC T [40], and 𝛾 AC is called the acoustic phonon scattering parameter [38, 39]. Assuming that excitons and phonons are in thermal equilibrium before and after transitions, n𝜆q|| and fKex can be replaced by || the average values obtained from the Bose–Einstein distribution as [5] [ ℏ𝜔q𝜆 ]−1 𝜆 kB T nq|| = e −1 (16.37) and f𝜆

[ ]−1 K|| = e[E(K|| )−𝜇]∕kB Tex −1

(16.38)

where m is the chemical potential of 2D excitons, and T ex is the exciton temperature. As the acoustic phonon energies are small, the exponential in the average phonon occupation in Eq. (16.37) can be expanded. Retaining terms only up to the first order, one gets n𝜆q|| ∝ T. This implies that the linear relation may be valid only for ℏ𝜔q𝜆 ≪ k B T. The chemical potential 𝜇 is a function of the exciton temperature T ex and 2D exciton density ∗ 2 N ex [41] as 𝜇 = kB Tex ln[1 − e−2𝜋Nex ℏ ∕(gM|| kB Tex ) ). Therefore, the dephasing rate as derived here is also obtained as a function of the exciton temperature T ex and density N ex , as well as the exciton energy E(K|| ) (or COM momentum ℏK|| ) and lattice temperature T. In this regard, the present result of dephasing rate is more general compared with that derived by Takagahara [32, 40]. It may be noted that the excitation density considered here is relatively low, which enables us to treat excitons through Bose–Einstein distribution. Here, we have calculated the dephasing rate and linewidth for GaAs/Al0.3 Ga0.7 As QWs of widths Lz = 28 and 80 Å. The QW width of 28 Å is chosen for comparing the present

16.5 Quasi-2D Free-Exciton Linewidth

Dephasing rate [s–1]

1011 1010 109 108 107 106 0.0

0.5

1.0 KII

[108

1.5

2.0

m–1]

Figure 16.15 Dephasing rate as a function of K|| for Lz = 28 Å QWs with Nex = 5 × 109 cm−2 at various lattice and exciton temperatures (–), T = 5 and T ex = 10 K; (⋅⋅⋅⋅⋅), T = 5 and T ex = 40 K; (– – – ), T = 80 and T ex = 10 K; (– ⋅ – ⋅), T = 80 and T ex = 40 K; (– ⋅⋅⋅), T = 150 and T ex = 10 K; (– –), T = 150 and T ex = 40 K [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

results with experimental ones where the same well width has been used [37], and the other width of 80 Å is used because, recently, several papers have appeared on excitonic process in GaAs QWs having this width. In Figures 16.15 and 16.16, we have plotted the dependence of the dephasing rate on the COM momentum at the exciton density N ex = 5 × 109 cm−2 for QWs of widths Lz = 28 and 80 Å, respectively, at three different lattice temperatures (5, 80, and 150 K) and two different exciton temperatures (10 and

Dephasing rate [s–1]

1011 1010 109 108 107 106 0.0

0.5

1.0

1.5

2.0

KII [108 m–1]

Figure 16.16 Dephasing rate as a function of K|| for Lz = 80 Å QWs with Nex = 5 × 109 cm−2 at various lattice and exciton temperatures (—), T = 5 and T ex = 10 K; (⋅⋅⋅⋅⋅), T = 5 and T ex = 40 K; (– – –), T = 80 and T ex = 10 K; (– . – .), T = 80 and T ex = 40 K; (– ⋅⋅⋅), T = 150 and T ex = 10 K; (– –), T = 150 and T ex = 40 K [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

487

16 Excitonic Processes in Quantum Wells

40 K). Both Figures 16.15 and 16.16 illustrate that the dephasing rate of free excitons at a relatively low exciton temperature is more sensitive to the exciton momentum (or kinetic energy) than that at a relatively high exciton temperature. At a given lattice temperature, the dephasing rate of excitons with lower momentum (or kinetic energy) is dominant at relatively low exciton temperatures, whereas that with higher momentum (or kinetic energy) is dominant at relatively high exciton temperatures. As expected, this trend implies that the average exciton kinetic energy is proportional to k B T ex , where k B is Boltzmann’s constant. In particular, in the case at low lattice and high exciton temperatures, e.g. T = 5 and T ex = 40 K, the maximum value of the dephasing rate occurs at a high exciton momentum (or kinetic energy), whereas at high lattice and low exciton temperatures it occurs at K|| ∼ 0, which means near the exciton band minimum. Figures 16.15 and 16.16 suggest that the dephasing rate for Lz = 28 Å QWs is higher than that for Lz = 80 Å QWs for the same N ex , T, and T ex. This agrees with the theoretical prediction for an infinite barrier potential by Borri et al. [11]. However, their experimental results for Inx Ga1−x As/GaAs QWs are available only for narrow QWs of widths 10–40 Å, in which range the penetration of electronic wave functions may play a very significant role. Also, the roughness of QW walls due to alloying will become important in such narrow QWs. The dependence of the dephasing rate on lattice and exciton temperatures obtained from Eq. (16.35) for Lz = 80 Å and N ex = 1 × 1010 cm−2 is shown in Figures 16.17 and 16.18 at two different exciton momenta K|| = 0.0004 × 108 m−1 and K|| = 1 × 108 m−1 , respectively. Here, K|| = 0.0004 × 108 m−1 corresponds to the kinetic energy E(K|| ) = 3.46 × 10−7 meV, and K|| = 1 × 108 m−1 corresponds to E(K|| ) = 2.16 meV. At a given exciton temperature and K|| , the dephasing rate of an exciton increases linearly as the lattice temperature increases, which agrees very well with the experimental result [36]. However, the dependence of the dephasing rate on the exciton temperature at a given T is different. As can be seen from Figure 16.17 60 Dephasing rate [109 s–1]

50 40 30 20 10 0

4 01

0 10

Tex [K ]

80

60

40

0 20

488

20

40

12 00 01 ] 8 K 60 T[

Figure 16.17 Dephasing rate as a function of T and T ex at K|| = 0.0004 × 108 m−1 for Lz = 80 Å QWs with Nex = 1 × 1010 cm−2 [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

16.5 Quasi-2D Free-Exciton Linewidth

10 Dephasing rate [109 s–1]

Figure 16.18 Dephasing rate as a function of T and T ex at K || = 1 × 108 m−1 for Lz = 80 Å QWs with Nex = 1 × 108 cm−2 [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

8 6 4 2

0

0 10

Tex [K ]

80

60

40

20

0

20

40

20 01 0 ] 1 80 T [K 60

14

at K|| = 0.0004 × 108 m−1 , which is near the exciton band minimum, the dephasing rate decreases with increasing exciton temperature. However, in the high-momentum region, shown in Figure 16.18, the dephasing rate increases first to a peak value and then decreases as the exciton temperature increases. For Lz = 28 Å QWs, one also finds a similar trend, although it is not shown here. Experimentally, the temperature dependence of the homogeneous exciton linewidth is determined by fitting the experimental data to the following equation [33, 42]: Γh (T) = Γh (0) + 𝛾AC T + 𝛾LO

1

(16.39) −−1 e where the first term is the contribution to the linewidth due to impurities, and the second and third terms are due to acoustic and LO phonon scatterings, respectively. The observed linewidth due to acoustic phonon interaction 𝛾 AC T from various experiments shows large variations as listed below: 𝛾 AC T = 3μeV K−1 for Lz = 45 Å [43], 1.7 μeV K−1 for Lz = 150 Å, 3 μeV K−1 for L𝛾 AC T = 325 Å [44], 2.5 μeV K−1 for Lz = 135 Å, 5 μeV K−1 for Lz = 277 Å [45], 3.8 μeV K−1 for Lz = 130 Å, 3.8 μeV K−1 for Lz = 170 Å, 5.2 μeV K−1 for Lz = 250 Å, and 4.4 μeV K−1 for Lz = 340 Å4 QWs, and 14 μeV K−1 [46] and 17 μeV K−1 [47]. In contrast to the prediction of Borri et al. [11] and present theory, these results do not indicate any systematic relation between 𝛾 AC and the QW width. It may, however, be noted that Eq. (16.39) does not take into account any dependence of Γh (T) on the temperature and density of excitons, which is found to exist quite sensitively from the present theory. This implies that all experimental results are not obtained at the same temperature and density of excitons, and therefore it is not meaningful to expect any systematic relation between Γh (T) and Lz , as obtained from the present theory. Recently, Fan et al.5 have measured Γh (T) ∼ (14–19) μeV at T = 10 K and N ex ∼ (1–10) × 109 cm−2 in Lz = 28 Å GaAs QWs. For the same well width of Lz = 28 Å and T = 10 K, our calculation gives Γh (T) = (1.1–6.4) μeV at N ex = 5 × 109 cm−2 and (2.4–16) μeV at N ex = 10 × 109 cm−2 with decreasing T ex from 5 to 1 K. These results ℏ𝜔LO kB T

4 It should be noted that the linewidths in Ref. [33] are given in terms of a half-width at half-maximum. 5 The values of linewidths are estimated from the experimental data given in Figure 2(b) in Reference [36].

489

16 Excitonic Processes in Quantum Wells

Dephasing rate [109 s–1]

6 5 4 3 2 1 0 0.0 K

0.5 .0 [10 8 1 m –1 1.5 ] 2.0

II

10

8

6

4 9 m–2 ] n ex [10

2

Figure 16.19 Dephasing rate as a function of Nex and K|| at T = 10 and T ex = 5 K for Lz = 28 Å QWs [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

agree reasonably well with Fan et al.’s experimental results although our results are slightly lower. Such a discrepancy between theoretical and experimental results is also found in bulk GaAs, where 𝛾 AC = 0.64 μeV K−1 [48] is obtained from theory, and 14 μeV K−1 [46] and 17 μeV K−1 [47] are obtained from experiments. In Figures 16.19 and 16.20, we have plotted the exciton dephasing rate at T ex = 5 K and 40 K, respectively, as a function of the exciton density and momentum at a lattice temperature of 10 K for a QW of width 28 Å as used by Fan et al. [36]. At a given 1.0 Dephasing rate [109 s–1]

490

0.8 0.6 0.4 0.2 0.0 0.0 K

II

0.5 .0 [10 8 1 m –1 1.5 ] 2.0

6

2

8

10

4 9 m–2 ] n ex [10

Figure 16.20 Dephasing rate as a function of Nex and K|| at T = 10 and T ex = 40 K for Lz = 28 Å QWs [7]. Source: Reprinted from I.-K. Oh and J. Singh, Superlattices Microstruct., 30, 287. Copyright (2001) Elsevier.

16.6 Localization of Free Excitons

exciton momentum, the rate of dephasing (homogeneous linewidth) increases linearly with increasing exciton density, which agrees very well with the experimental results of Fan et al. [36]. Such a linear relation has also been observed in other QW structures such as Inx Ga1− xAs/GaAs [11], ZnSe/Zn0.94 Mg0.06 Se [38], and CdTe/Cd1−x−y Mgx Zny Te [37]. The slope of the linear relation, as shown in Figures 16.19 and 16.20, depends on the temperature and momentum of excitons. However, this dependence cannot be compared with any experimental results, as no experiments have been done in connection with these parameters. As a rough estimate, we have calculated the slope of Γh with respect to N ex . Our result at T = 10 K and T ex = 2 K near the exciton band minimum gives a slope of 0.76 × 10−9 μeV cm2 , which agrees with the value of 0.75 × 10−9 μeV cm2 estimated from Fan et al.’s result.6 We have also found that the slope is very sensitive to the exciton temperature. For example, at a lattice temperature of 10 K, we have found that the slope increases from 0.3 × 10−9 to 1.9 × 10−9 μeV cm2 when T ex decreases from 5 to 1 K. However, more experimental results are required to verify this dependence on the exciton temperature.

16.6 Localization of Free Excitons The interfaces between two different semiconductors in QWs can dramatically alter physical properties such as confinements of charge carriers [49] and phonons [50]. In addition, there are inevitable structural disorders of the atomic lattice at the interfaces, which give rise to roughness and modify the physical properties of QWs even further. The interface roughness influences the optical properties and plays a very important role in the dynamics of excitons in QWs. The fine structures and splitting in the photoluminescence (PL) signals of excitons are some examples of the effects of interface roughness causing localization in the COM motion of an exciton. The aim of this section is to study the localization of free excitons due to interface roughness through the nonradiative processes in QWs as a transition from a free exciton state to a localized exciton state. When a sample of QWs is subjected to photons of energy greater than the bandgap of the well material, several excitonic processes may occur: (1) a free exciton with nonzero COM wavevector K|| can be formed, which then relaxes nonradiatively to another free-exciton state with K|| ∼ 0 and finally decays radiatively to the ground state, (2) a free exciton is formed and then becomes localized nonradiatively within the interface roughness potential by emitting phonons, and (3) the localized exciton formed due to the interface roughness in (2) is transferred nonradiatively to another localized state. There may be more complicated relaxation and radiative decay processes; however, in this section, we will focus on process (2), i.e. free excitons initially created by photons of energy higher than the bandgap becoming localized in the localized energy states caused by the roughness potential at the interface by emitting phonons. Here, we assume that free excitons are present in the QWs. If the interface of the QWs is perfectly smooth, these excitons will relax to exciton states with the COM wavevector K|| ∼ 0 and then recombine radiatively. In this case, the PL signal should be sharp and smooth, and its peak should also be well defined although it has a characteristic line broadening associated with the interaction of excitons with other excitations such as 6 The values of linewidths are estimated from the experimental data given in Figure 2 (b) in Reference [36].

491

492

16 Excitonic Processes in Quantum Wells

phonons and other excitons. In practice, however, there is always some roughness or disorder at the interfaces of the QWs. The potential energy fluctuations due to such interface roughness or alloy disorders in semiconductor heterostructures can hinder the smooth motion and cause localization of excitons, giving rise to fine structures and a red shift in PL [8]. Here, we consider the localization of a free exciton in the QWs due to the interface roughness involving acoustic phonons. This is based on the assumption that initially excitons are created with an energy higher than the bandgap, as stated earlier, so they are not localized. Thus, the initial wave function can be approximated by a plane wave. We assume that the density of free excitons is so low that the exciton–exciton interaction can be ignored. For the perpendicular motion of electron and hole in the QWs, we consider only the lowest sub-band. We also consider only 1s excitons for the internal motion of an exciton in the QW plane. Using the COM R|| and relative r|| coordinates for an exciton in the † describing the creation of n exciton with its plane of the QWs, the field operator 𝜓ex COM wave function 𝜓 ex (𝜉, R|| ) can be written as [50] ∑ † ∗ ̂† 𝜓ex = 𝜓ex (𝜁 , R|| )𝜙∗x (r)𝜙∗e (xe3 )𝜙∗h (xh3 )B (16.40) 𝜁 𝜁

where 𝜁 represents the set of quantum numbers for the COM of an exciton. As a convenient notation, we use 𝜁 = K|| for a free-exciton state with the COM wavevector K|| , and 𝜁 = Ra for a localized exciton state at a site Ra . Then the COM wave function 𝜓 ex (𝜉, R|| ) can be written as [50] { 𝜓ex (𝜉, R|| ) =

𝟏 √ eiK|| .R|| A𝟎

for

ζ = K||

𝜓loc (R|| − Ra ) for

ζ = Ra

(16.41)

In Eq. (16.40), 𝜙x , 𝜙e , and 𝜙h are wave functions of the exciton in the relative coordinate, electron, and hole for the motion perpendicular to the well of the QW, respectively, ̂† is the creation operator of an exciton in state 𝜁 . Likewise, using the annihilation and B 𝜁 ̂𝜁 of an exciton, the field operator describing the annihilation of an exciton operator B can be written as ∑ ̂𝜁 𝜓 ̂ex = 𝜓ex (𝜉, R|| )𝜙x (r|| )𝜙e (xe3 )𝜙h (xh3 )B (16.42) 𝜁

In principle, the free-exciton wave function is influenced by the presence of potential fluctuations due to interface roughness in the QW plane causing an exciton to be localized as it is considered here in the final state. Therefore, in the initial state, we have assumed the free-exciton wave function to be a plane wave as given in Eq. (16.41). We also assume that the energy fluctuations due to interfacial roughness or material compositions are not strong enough to influence the exciton wave function associated with its relative motion and charge carrier sub-band wave functions associated with their motion perpendicular to the QWs. That means 𝜙x , 𝜙e , and 𝜙h are assumed to be the same for both free and localized excitons. In accordance with Citrin [51], we consider a model potential V d due to a well-width fluctuation of lateral size b and thickness a/2 where a is the bulk GaAs lattice constant.

16.6 Localization of Free Excitons

Then V d in relative and COM coordinates can be written as ) ( |R|| + 𝜶 h r|| |2 e h Vd (R|| , r|| , x3 , x3 ) = −Ve exp − 2b2 [Θ(xe3 − Lz ∕2 + a∕4) − Θ(xe3 − Lz ∕2 − a∕4)] ) ( |R|| + 𝜶 e r|| |2 − Ve exp − 2b2 [Θ(xh3 − Lz ∕2 + a∕4) − Θ(xh3 − Lz ∕2 − a∕4)] with

{

1 0

Θ(x3 ) =

x3 ≥ 0 x3 < 0

for for

(16.43)

(16.44)

where V e and V h are the band offsets in the conduction and valence bands, respectively. It is to be noted that the model fluctuations considered here are one molecular layer (ML) thick from the QW wall, both inside and outside. Then the effective fluctuation eff potential of an exciton for the COM motion Vd (R|| ) can be expressed as [50] eff

Vd (R|| ) =



dr||



dxe3



xh3 Vd (R|| , r|| , xe3 , xh3 )|𝜙x (r|| ) 𝜙e (xe3 ) 𝜙h (xh3 )|2

(16.45)

The energy eigenvalue Eloc and wave function 𝜓 loc (R|| ) of the localized exciton for its COM motion can be determined from the Schrödinger equation [50]: ( 2 2 ) ℏ ∇|| eff + Vd (R|| ) 𝜓loc (R|| ) = Eloc 𝜓loc (R|| ) (16.46) 2M||∗ Here, we consider only the lowest bound state of a localized exciton due to the interface roughness. As a trial wave function for 𝜓 loc , we use the Gaussian type of COM envelope function for the localized state given by 𝟐∕𝟐 𝟐 𝟏 𝜓loc (R|| − Ra ) = √ e−|R|| −Ra | 𝜻 𝝅 𝜻

(16.47)

where Ra is the site center of localization, and x is the characteristic localization length [22] used as a variational parameter [52, 53]. Then the exciton–phonon interaction Hamiltonian in the second quantized form becomes [50] ∑ † ̂I = dR|| dr|| dxe3 dxh3 𝜓 ̂ex HIJ 𝜓 ̂ex H ∫ ∫ ∫ ∫ J ∑ J ̂ ̂ ̂†′ B = CJ𝜆 [F𝜆− (q|| , qz )M− (q|| , 𝜁 ′ , 𝜁 )B b 𝜁 𝜁 𝜆q|| J𝜆𝜁 𝜁 ′ q|| ,qz

J ̂†′ B ̂ ̂ + F𝜆+ (q|| , qz )M+ (q|| , 𝜁 ′ , 𝜁 )B b† ] 𝜁 𝜁 𝜆q ||

(16.48)

with M± (q|| , 𝜁 ′ , 𝜁 ) =



∗ dR|| 𝜓 ̂ex (𝜁 , R|| )̂ 𝜓ex (𝜁 , R|| )e∓iq|| . R||

(16.49)

̂ I represents an interaction operator for a transition The interaction Hamiltonian H from one exciton state to another through the interaction with acoustic phonons. Thus,

493

494

16 Excitonic Processes in Quantum Wells

when 𝜁 = K|| and 𝜁 = K′ || , i.e. two different free-exciton states, are considered, the Hamiltonian in Eq. (16.48) represents a transition between two free excitons involv′ ing acoustic phonons as given in Eq. (47) of Ref. [5]. When 𝜁 = Ra and 𝜁 = R′ a , i.e. two ′ localized-exciton states at two different sites Ra and R a , are considered, the interaction Hamiltonian represents the exciton transfer from site Ra to R′ a through the acoustic phonon interaction, which has been studied in detail by Takagahara [22]. The case of ′ 𝜁 = Ra and 𝜁 = K|| , one localized-exciton state at site Ra and the other a free-exciton state with K|| , has not been considered before. The interaction Hamiltonian in this case is associated with the localization of a free exciton or delocalization of a localized exciton through the acoustic phonon interaction. Here, we consider the process of transition from a free exciton to a localized exciton by emission of an acoustic phonon. For convenience, we write the localized wave function in Eq. (16.47) in K|| -space as ′ 1 ∑ 𝜓loc (R|| − Ra ) = √ g(K′|| , 𝝃, Ra )eiK|| . R|| (16.50) A0 K′|| ′

with g(K′|| , 𝜁 , Ra ) =

(

1 2𝜋

)2

dR|| 𝜓loc (R|| − Ra )e−iK|| . R|| = ′



′ 𝝃 2 ′ e−iK|| . R|| −𝜉 K|| 𝟐∕𝟐 3∕2 (2𝜋) (16.51)

Using Eqs. (16.41) and (16.49)–(16.51), we get M± (q|| , Ra , K|| ) = g(K|| ∓ q|| , 𝝃, Ra )

(16.52)

Using the Fermi golden rule and second term in Eq. (16.48), the rate of localization of a free exciton due to emission of an acoustic phonon can be obtained as 2𝜋 ∑ ̂ I ∣ i⟩|2 𝛿(Ef − Ei ) W (Ei ) = |⟨f ∣ H (16.53) ℏ f with Ei =

ℏ2 K||2 2M||∗

− Eb

(16.54)

and Ef = Eloc (𝜉) − Eb + ℏ𝜔𝜆q

(16.55)

where Eb is the binding energy of the exciton for the relative motion and 𝜔𝜆q the frequency of the 𝜆-mode phonon. Eloc (𝜉) is obtained from Eq. (16.46) with the help of Eq. (16.47). The initial state |i⟩ and final state |f ⟩ in Eq. (16.53) are characterized by the occupation numbers of the free exciton, localized exciton, and phonon states involved in the transition as given in Reference [54]. We consider the rate of localization of free excitons in [001]-oriented GaAs/Al0.3 Ga0.7 As QWs. As we consider the case that the energy of incident photons is greater than the energy gap of QWs, we assume that initially only 1s free excitons are resonantly excited and their density in the QW plane is N ex . The problem considered here is a transition from a free-exciton state to a localized-exciton state caused by the roughness at the QW interfaces. The free excitons thus created experience the potential fluctuations in their

16.6 Localization of Free Excitons

8 COM binding energy [meV]

Figure 16.21 COM binding energy as a function of Lz in the plane QWs due to interface roughness for four different values of b [50]. Source: Reprinted from I.-K. Oh and J. Singh, J. Appl. Phys., 95, 4883. Copyright (2004) American Institute of Physics.

b = 200 Å

6

b = 150 Å b = 100 Å b = 50 Å

4

2

0

60

80

100

120

140

Lz [Å]

COM motion and become localized by emitting acoustic phonons. In this process, we ignore the direct creation of localized excitons, which may occur dominantly when the energy of the exciting photons is less than the excitation energy of 1s free excitons. We also ignore the interaction between free excitons. For simplification and without the loss of generality, we consider that the localization occurs on a site at Ra = 0. Starting from the trial wave function in Eq. (16.47) and using the variational method and effective potential in Eq. (16.45), we have calculated the energy eigenvalue and the corresponding localized eigen function from Eq. (16.46) using different values of b and assuming that the well-width fluctuation in thickness is 1 ML (a/2 = 2.83 Å). This variational calculation is repeated for every well width in the range 50–150 Å, starting from 50 Å and increasing in steps of 5 Å. The COM binding energy of a localized exciton, which is thus obtained from the energy eigenvalues, is plotted in Figure 16.21 as a function of the well width for Lz = 50–150 Å and four different lateral sizes b = 50, 100, 150, and 200 Å. According to the results shown in Figure 16.21, the binding energy in relatively narrow QWs is larger than that in wider QWs at a given lateral size b. This feature can be explained as follows: the penetration of the sub-band wave functions 𝜙e and 𝜙h into the barrier becomes large for narrow QWs compared with that for wider QWs. As a result, the effective interaction of charge carriers with potential fluctuations at the interface of narrow QWs becomes larger compared with that for wider QWs. In other words, a larger overlap of the sub-band wave functions with the potential fluctuations at the interface enhances the effective potential of an exciton in the COM motion. Our results reveal that the binding-energy difference between different lateral sizes for relatively narrow QWs becomes large enough to observe the splitting of localized excitons even in conventional PL spectroscopy in comparison with that for relatively wide QWs. This feature agrees very well with the experimental results [54]. As can be seen clearly from Figure 16.21, the splitting or fine structures of excitons in wide QWs, e.g. Lz > 100 Å, is so small that it may only be observed through micro-PL spectroscopy. Our results also imply that the localization of the 1s free exciton may occur only due to acoustic phonon emission for QWs with well width Lz > 50 Å and b < 200 Å because the COM binding energy is much less than the LO phonon energy of 36.2 meV. This is the reason for considering the localization of 1s free excitons only through the acoustic phonon interaction here.

495

16 Excitonic Processes in Quantum Wells

1012 Exciton localization rate [s–1]

496

1011

LA TA

1010 109 108 107 106 105 0.000

Lz = 50 Å

0.005

0.010 KII

[1010

0.015

0.020

m–1]

Figure 16.22 Localization rates as a function of |K_| due to LA and TA phonon emission, respectively, at Nex = 5 × 109 cm−2 , b = 100 Å, T ex = 15 K, and T = 4.2 K for Lz = 50 Å QWs [50]. Source: Reprinted from I.-K. Oh and J. Singh, J. Appl. Phys., 95, 4883. Copyright (2004) American Institute of Physics.

Taking into account both DP and PE couplings, we have plotted in Figure 16.22 the rates of localization of a 1s free exciton by emission of LA and TA phonons, respectively, as a function of the COM wavevector |K|| | at a crystal temperature T = 4.2 K and exciton temperature T ex = 15 K in QWs of width Lz = 50 Å and lateral size b = 100 Å. The results show that the rate of localization by emission of acoustic phonons changes by about 103 and 102 for LA and TA phonon processes, respectively, when |K|| | increases from 0 to 0.007 × 1010 m−1 . This corresponds to a change in the COM kinetic energies by only 1.06 meV. At low COM kinetic energies (small |K|| |), the rate of localization is dominated by LA phonon emission. This is due to the combined effect of the anisotropic coupling constants, phonon dispersion relation, form factors, and energy–momentum conservation. For a free exciton near K|| = 0, the ( acoustic phonons ) in the localization process are dominantly those near √ involved 𝟐

q = q|| ∼ 𝟎, qz ∼ (ℏvE )𝟐 − K||𝟐 , [E = ℏ2 K|| 2 ∕(2M||∗ ) − Eloc ]. It is found from the phonon dispersion relation that the velocity of acoustic phonons vLA > vTA , which gives smaller qz for LA phonons than that for TA phonons. Furthermore, near q|| = 0, the localization due to the PE coupling can be ignored because of the PE coupling constants h𝜆q ∼ 0 for both LA and TA phonons (see Reference [5]). For the localization process due to D (q|| ∼ 0, qz )|2 is much larger the deformation coupling at q|| ∼ 0, we can show that |FLA+ D 2 7 than |FTA+ (q|| ∼ 0, qz )| . Therefore, for a free exciton near K|| = 0, the rate of the localization due to LA phonons is greater than that due to TA phonons. It is to be noted that in this case, although the dominant contribution of phonons to the localization is due to those near q|| = 0, the calculation of the rate has to be carried out by integration 𝝀

D 7 From Reference [5], we can show that FLA+ (q|| ∼ 0, qz ) ∼ −i[Ξc Fe (qz ) − mFh (qz )] and √ D D FLA1+ (q|| ∼ 0, qz ) = FLA2+ (q|| ∼ 0, qz ) ∼ −i[(n∕ 3)Fh (qz )]. Since |F j (qz )| (j = e, h) decreases as qz increases and qz for LA phonons is smaller than for TA phonons as explained in the text, |F j (qz )| for LA phonons is greater than that for TA phonons. For GaAs, the deformation potentials are given by Ξc = −8 meV, D D m = 2.86 eV, and n = −7.88 eV. Therefore, ∣ FLA+ (q|| ∼ 0, qz ) ∣ is larger than ∣ FTA+ (q|| ∼ 0, qz ) ∣.

16.6 Localization of Free Excitons

over all q = (q|| , qz ). The structured features appearing in the graphs of rates shown in Figure 16.22 are related to the anisotropy in both DP and PE coupling constants and the energy–momentum conservation. In the process of localization of a 1s free exciton with K|| , by emitting acoustic phonons, the contribution comes mainly from √ 𝟐 phonons with specific wavevectors near |q|| | = |K|| | and qz ∼ (ℏvE )𝟐 − K||𝟐 , because of the energy–momentum conservation. In other words, the energy–momentum conservation and anisotropy in coupling constants give rise to the structures in the rate of localization as shown in Figure 16.22. The total rate of localization decreases from 1.07 × 1011 to 2.37 × 109 s−1 and its corresponding time of localization increases from 9.3 to 423 ps when the COM kinetic energy increases from 0 to 1.12 meV (|K|| | = 0.0072 × 1010 m−1 ). In Figure 16.23, we have plotted the rate of localization of a 1s free exciton due to emission of both LA and TA phonons as a function of |K|| | at N ex = 5 × 109 cm−2 , T ex = 15 K, and T = 4.2 K for four different widths Lz = 50, 60, 85, and 130 Å of QWs with the same value of b = 100 Å. The results show that the dependence of the rates of localization on |K|| | for QWs with Lz = 60, 85 and 130 Å is more sensitive than that for QWs with Lz = 50 Å. This can be explained as follows. The COM binding energy of a localized exciton decreases with increasing width of the QWs as shown in Figure 16.21. This means that the characteristic localization length 𝜉 increases when Lz increases. Therefore, the Fourier transform of the localized wave function, g(K|| , 𝜉, 0) in Eq. (16.51), becomes more sensitive to |K|| | in wider QWs (large Lz ) compared with that in narrow QWs (smaller Lz ). As the rate of localization is proportional to |g(K|| , 𝜉, 0)|2, the resulting rate also becomes more sensitive to |K|| | in wider QWs than in narrow QWs. The results in Figure 16.23 also indicate that the plotted rates exhibit more fine structure at larger values of |K|| | for all four well widths, as also found in Figure 16.22. As explained above, these structures are also associated with the anisotropy and energy–momentum conservation.

Exciton localization rate [s–1]

𝝀

1012

1010

108

Lz = 50 Å Lz = 60 Å

106 0.000

Lz = 85 Å Lz = 130 Å 0.005

0.010

0.015

0.020

KII [1010 m–1]

Figure 16.23 Localization rate due to acoustic phonon emission as a function of |K_| at Nex = 5 × 109 cm−2 , b = 100 Å, T ex = 15 K, and T = 4.2 K for four different widths of QWs [50]. Source: Reprinted from I.-K. Oh and J. Singh, J. Appl. Phys., 95, 4883. Copyright (2004) American Institute of Physics.

497

16 Excitonic Processes in Quantum Wells

3.0 × 1012 Exciton localization rate [s–1]

498

2.5 × 1012 2.0 × 1012 1.5 × 1012 1.0 × 1012 × 50

5.0 × 1011

60

80

100

120

140

Lz [Å]

Figure 16.24 Localization rate as a function of Lz for Nex = 5 × 109 cm−2 , b = 100 Å, T ex = 15 K, and T = 4.2 K. The (—) curve is at |K|| | = 0, (⋅⋅⋅⋅⋅) curve at |K|| | = 0.52 × 108 m−1 , and (– ⋅ – ⋅) curve at |K|| | = 1 × 108 m−1 [50]. Source: Reprinted from I.-K. Oh and J. Singh, J. Appl. Phys., 95, 4883. Copyright (2004) American Institute of Physics.

In Figure 16.24, we have shown the dependence of the localization rate of a 1s free exciton on the well width in the range Lz = 50–150 Å for b = 100 Å at N ex = 5 × 109 cm−2 , T ex = 15 K, and T = 4.2 K. The (—), (⋅⋅⋅⋅⋅), and (– ⋅ – ⋅) curves correspond to the rates of localization at |K|| | = 0, 0.52 × 108 , and 1 × 108 m−1 , respectively, and the corresponding COM kinetic energies are 0, 0.58, and 2.16 meV. The rates of localization show a nonmonotonic behavior with respect to Lz . As explained above, the rates depend on |g|2 , which in turn depends on 𝜉, which is a monotonically increasing function of Lz . 2 2 According to Eq. (16.51), g depends on 𝜉 nonmonotonically as 𝜉e−𝜉 K|| ∕2 , and therefore 2 2 the rates also depend nonmonotonically on Lz as 𝜉e−𝜉 K|| ∕2 as shown in Figure 16.24. The results of Figure 16.24 can be summarized as follows: At |K|| | = 0, the rate of localization has a maximum of 2.34 × 1012 s−1 at Lz = 85 Å and the corresponding time of the localization is about 427 fs. At |K|| | = 0.52 × 108 m−1 , the maximum rate of 1.54 × 1012 s−1 occurs at Lz = 110 Å, and the corresponding time of localization is about 651 fs; and at |K|| | = 1 × 108 m−1 , the maximum rate is 2.40 × 1010 s−1 at Lz = 95 Å, and the corresponding time of localization is 41.6 ps. It is to be noted that, at Lz = 50 Å, the rates of localization are 1.07 × 1011 , 1.45 × 1010 , and 5.12 × 109 s−1 at |K|| | = 0, 0.52 × 108 , and 1 × 108 m−1 , respectively, and their corresponding times of localization are 9.3, 69, and 195 ps. Figure 16.25 shows the dependence of the rate of localization on |K|| | in Lz = 50 Å QWs for four different lateral sizes of the well-width fluctuations. The values of N ex , T ex , and T are the same as those used in Figure 16.24. The results show that, for a small lateral size of b = 50 Å, the rate of localization is less sensitive to |K|| | in the region of relatively small values of |K|| |, but as |K|| | increases it becomes more sensitive to |K|| |. In this case, the maximum rate of 2.85 × 1012 s−1 occurs at |K|| | = 0.002 × 1010 m−1 , and the corresponding time of localization is about 350 fs. In the small |K|| | region, when b changes from 50 to 100 Å, the rate of localization decreases by more than an order of magnitude

Exciton localization rate [s–1]

16.7 Conclusions

1012

1010

108

b = 50 Å b = 100 Å b = 150 Å

106 0.000

b = 200 Å 0.005

0.010

0.015

0.020

KII [1010 m–1]

Figure 16.25 Localization rate as a function of |K|| | for four different lateral sizes b = 50, 100, 150, and 200 Å, at Nex = 5 × 109 cm−2 , Lz = 50 Å, T ex = 15 K, and T = 4.2 K [50]. Source: Reprinted from I.-K. Oh and J. Singh, J. Appl. Phys., 95, 4883. Copyright (2004) American Institute of Physics.

at |K|| | ∼ 0. However, when b changes from 150 to 200 Å, such a drastic change in the rates is not found near |K|| | = 0. For b = 100 Å, the maximum rate and its corresponding time of localization are 1.07 × 1011 s−1 and 9.30 ps, respectively, at |K|| | = 0. On the other hand, for larger lateral sizes b = 150 and 200 Å, the rates of localization are less sensitive to |K|| |, and their maximum rates are 6.74 × 109 s−1 at |K|| | = 0.0048 and 7.34 × 109 s−1 at |K|| | = 0, respectively. The corresponding times of localization are 148 and 136 ps. Therefore, from Figures 16.23 and 16.25, we can conclude that the rate of localization is high and the time of localization is relatively faster for lateral sizes b = 100(50) Å and well width Lz = 85(50) Å at |K|| | ∼ 0.

16.7 Conclusions In summary, we have presented in this chapter a comprehensive theory of the formation processes of an exciton due to acoustic and LO phonon emission via DP, PE, and PO couplings as a function of the COM wavevector K|| , charge carrier temperature T e–h, and charge carrier density N e–h at a low lattice temperature of T = 4.2 K. The results in [001]GaAs/Al0.3 Ga0.7 As QWs show that the rate of formation of excitons is very sensitive to K|| , T e–h , and N e–h . We have found from this study that the rate of formation due to LO phonon emission is dominant over that due to acoustic phonon emission at all K|| for T e − h ≥50 K, but for T e − h ≤50 K there is a crossover at an exciton wavevector K0|| , where the formation via acoustic phonon emission becomes dominant for K|| > K0|| . It is also found that the rates of exciton formation by LA and TA phonons are comparable. A theory is presented for calculating the rate of dephasing of free excitons as a function of temperature and density of excitons, lattice temperature, COM momentum of excitons, and QW width. Our results for GaAs/Al0.3 Ga0.7 As QWs show that the dephasing rate and linewidth of free excitons are very sensitive to the temperature and density of excitons. We have also found from our calculations that the rate of dephasing increases

499

500

16 Excitonic Processes in Quantum Wells

linearly with increasing exciton density as well as lattice temperature, which agrees very well with experimental results. We have also developed a theory to calculate the rate of localization of 1s free excitons as a function of the well width, lateral size of well-width fluctuations, COM kinetic energy, exciton temperature, and exciton density due to acoustic phonon emission in QWs and have applied it to GaAs/Al0.3 GA0.7 As QWs. We have found that the time of localization is very sensitive to the COM kinetic energy.

References 1 Weisbuch, C., Benisty, H., and Houdré, R. (2000). J. Lumin. 85: 271. 2 Singh, J. (1994). Excitation Energy Transfer Processes in Condensed Matter. New

York: Plenum. 3 Oh, I.-K., Singh, J., Thilagam, A., and Vengurlekar, A.S. (2000). Phys. Rev. B 62: 2045.

and references therein. 4 Oh, I.-K. and Singh, J. (2001). Superlattices Microstruct. 30: 221. and references

therein. 5 Oh, I.-K. and Singh, J. (2000). J. Lumin. 85: 233. and references therein. 6 Oh, I.-K., Singh, J., and Vengurlekar, A.S. (2002). J. Appl. Phys. 91: 5796. and refer-

ence therein. 7 Oh, I.-K. and Singh, J. (2001). Superlattices Microstruct. 30: 287. and references

therein. 8 Runge, E. and Zimmermann, R. (1998). Ann. Phys. (Leipzig) 7: 417. and references

therein. 9 Butov, L.V., Imamoglu, A., Mintsev, A.V. et al. (1999). Phys. Rev. B 59: 1625. 10 Akimov, A.V., Moskalenko, E.S., Challis, L.J., and Kaplyanskii, A.A. (1996). Physica B

219/220: 9. 11 Borri, P., Langbein, W., Hvam, J.M., and Martelli, F. (1999). Phys. Rev. B 59: 2215. 12 Mahan, G.D. Polarons in heavily doped semiconductors. In: Polarons in Ionic Crys13 14 15 16 17 18 19 20 21 22 23 24 25

tals and Polar Semiconductors (ed. J.T. Devreese), 1972. Amsterdam: North-Holland. Damen, T.C., Shah, J., Oberli, D.Y. et al. (1990). Phys. Rev. B 42: 7434. Gurioli, M., Borri, P., Colocci, M. et al. (1998). Phys. Rev. B 58: 13403. Chuang, S.L. (1995). Physics of Optoelectronic Devices. New York: Wiley. Oh, I.-K. and Singh, J. (2001). Int. J. Mod. Phys. B 15: 3660. Strobel, R., Eccleston, R., Kuhl, J., and Köhler, K. (1991). Phys. Rev. B 43: 12564. Piermarocchi, C., Tassone, F., Savona, V. et al. (1997). Phys. Rev. B 55: 1333. Piermorocchi, C., Savona, V., Quattropani, A. et al. (1997). Phys. Status Solidi A 164: 221. Oh, I.-K. and Singh, J. (2000). J. Lumin. 87–89: 219. Vass, E. (1993). Z. Phys. B 90: 401. Takagahara, T. (1985). Phys. Rev. B 31: 6552. Basu, P.K. and Ray, P. (1992). Phys. Rev. B 45: 1907. Kumar, R., Vengurlekar, A.S., Prabhu, S.S. et al. (1996). Phys. Rev. B 54: 4891. Adachi, S. (1994). GaAs and Related Materials: Bulk Semiconducting and Superlattice Properties. Singapore: World Scientific.

References

26 Miller, R.C., Kleinman, D.A., Munteanu, O., and Tsang, W.T. (1981). Appl. Phys. Lett.

39: 1. 27 (a) Miller, R.C., Gossard, A.C., Kleinman, D.A., and Munteanu, O. (1984). Phys. Rev.

28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54

B 29: 3740; (b) Miller, R.C., Kleinman, D.A., and Gossard, A.C. (1984). Phys. Rev. B 29: 7085. Arlt, S., Siegner, U., Morier-Genoud, F., and Keller, U. (1998). Phys. Rev. B 58: 13073. Glutsch, S., Chemla, D.S., and Bechstedt, F. (1995). Phys. Rev. B 51: 16885. Pal, B. and Vengurlekar, A.S. (2001). Appl. Phys. Lett. 79: 72. Ridley, B.K. (1997). Electrons and Phonons in Semiconductor Multilayers. Cambridge: Cambridge University Press. Takagahara, T. (1989). J. Lumin. 44: 347. Gammon, D., Rudin, S., Reinecke, T.L. et al. (1995). Phys. Rev. B 51: 16785. Gopal, A.V., Kumar, R., Vengurlekar, A.S. et al. (2000). J. Appl. Phys. 87: 1858. Patané, A., Polimeni, A., Capizzi, M., and Matelli, F. (1995). Phys. Rev. B 52: 2784. Fan, X., Takagahara, T., Cunningham, J.E., and Wang, H. (1998). Solid State Commun. 108: 857. Brinkmann, D., Kudrna, J., Gilliot, P. et al. (1999). Phys. Rev. B 60: 4474. Wagner, H.P., Schätz, A., Maier, R. et al. (1998). Phys. Rev. B 57: 1791. Kim, D.-S., Shah, J., Cunningham, J.E. et al. (1992). Phys. Rev. Lett. 68: 2838. Takagahara, T. (1985). Phys. Rev. B 32: 7013. Segall, B. (1968). Proceedings of the IXth International Conference on the Physics of Semiconductors, Moscow, 1968 (ed. S.M. Ryvkin), 425. Leningrad: Nauka. Ivanov, A.L., Littlewood, P.B., and Haug, H. (1999). Phys. Rev. B 59: 5032. Ruf, T., Spitzer, J., Sapega, V.F. et al. (1994). Phys. Rev. B 50: 1792. Srinivas, V., Hryniewicz, J., Chen, Y.J., and Wood, C.E.C. (1992). Phys. Rev. B 46: 10193. Schultheis, L., Honold, A., Kuhl, J. et al. (1986). Phys. Rev. B 34: 9027. Tredicucci, A., Chen, Y., Bassani, F. et al. (1993). Phys. Rev. B 47: 10348. Schultheis, L. and Ploog, K. (1984). Phys. Rev. B 29: 7058. Rudin, S., Reinecke, T.L., and Segall, B. (1990). Phys. Rev. B 42: 11218. García-Moliner, F. and Velasco, V.R. (1992). Theory of Single and Multiple Interfaces. Singapore: World Scientific. Oh, I.-K. and Singh, J. (2004). J. Appl. Phys. 95: 4883. Citrin, D.S. (1993). Phys. Rev. B 47: 3832. Bastard, G., Delalande, C., Meynadier, M.H. et al. (1984). Phys. Rev. B 29: 7042. Madelung, O. (1978). Introduction to Solid-State Theory. Berlin: Springer. Luo, C.P., Chin, M.K., Yuan, Z.L., and Xu, Z.Y. (1998). Superlattices Microstruct. 24: 163.

501

503

17 Optoelectronic Properties and Applications of Quantum Dots Jørn M. Hvam Department of Photonics Engineering, Technical University of Denmark, DK-2800, Kgs. Lyngby, Denmark

CHAPTER MENU Introduction, 503 Epitaxial Growth and Structure of Quantum Dots, 504 Excitons in Quantum Dots, 508 Optical Properties, 513 Quantum Dot Applications, 520 Conclusions, 533 References, 534

17.1 Introduction In semiconductor quantum dots (QDs), the mobile carriers – electrons and holes – are confined in all three directions to a region that is so small that their energies are fully quantized into discrete levels. QDs are therefore also sometimes called artificial atoms, sharing some of the optical and optoelectronic properties with atoms. In contrast to natural atoms, however, QDs typically contain millions of atoms, and their energy levels are determined not only by the atomic constituents, but also by the size and shape of the QDs. This opens up a new way of controlling and engineering the optical and optoelectronic properties of matter. Historically, the first semiconductor QDs were nanocrystals of typically II–VI semiconductors embedded in a glass matrix. They were used already in the 1930s in color and edge filters ranging through the visible spectral region. It was known – and applied – that the color shifted with the temperature and duration of an annealing process, but it was only much later realized that this was coupled to the size of the nanocrystals through the quantization of the electron energy levels in the valence and conduction bands [1]. Later, these nanocrystals were produced from colloidal synthesis, resulting in highly monodisperse solutions of, e.g. CdSe/ZnS core/shell QDs, i.e. a core of a lower-bandgap semiconductor (CdSe) encapsulated and passivated by a shell of a higher-bandgap semiconductor (ZnS) [2]. The real breakthrough for typically III–V semiconductor QDs came with the new epitaxial growth techniques developed in the 1970s such as molecular beam epitaxy (MBE) [3] and metal organic vapor phase epitaxy (MOVPE), which allowed for the growth of two-dimensional semiconductor quantum wells (QWs) [4], one-dimensional quantum Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

504

17 Optoelectronic Properties and Applications of Quantum Dots

wires (QWRs) [5, 6], and zero-dimensional QDs [7]. In the literature, one will also find references to QDs where one type of carriers, typically electrons, that are already confined in a two-dimensional quantum well are further laterally confined electrostatically on a nanoscale by an arrangement of external electrodes [8, 9]. In the following, however, we will focus our attention on the epitaxially grown QDs and only marginally refer to the applications of colloidal QDs.

17.2 Epitaxial Growth and Structure of Quantum Dots In this section, we shall briefly discuss some different techniques for the epitaxial growth of semiconductor QDs of high optical quality. 17.2.1

Self-Assembled Quantum Dots

The most commonly applied technique for epitaxial growth of semiconductor QDs is via self-assembly in the so-called Stranski-Krastanov (SK) growth mode [7, 10]. If, e.g. InAs is epitaxially grown on top of GaAs, the lattice mismatch of 7% will cause competition between internal strain and interface energies, which after the growth of a few monolayers (MLs) of InAs, will result in the self-assembly of InAs pyramidally shaped QDs resting on top of a thin so-called wetting layer of InAs (Figure 17.1a,b). A capping layer of GaAs is finally added as sketched in Figure 17.1c, and the fully confined InAs QDs take shapes that can be approximated by truncated pyramids as seen in Figure 17.1d. Indium intermixing during the growth of the capping layer may influence the final shape and strain distribution, and thereby also the shape and location of the electron and hole wave functions inside the QDs. This is of importance for certain aspects of the optical properties of QDs. The size, shape, uniformity, and density of the self-assembled QDs can only to some extent be controlled by the detailed growth conditions such as substrate

(a)

(b)

5 nm 25 nm 5 nm

(c)

(d)

Figure 17.1 SK-growth of InAs QDs on GaAs. Uncapped QDs: (a) artist’s view and (b) a 0.5 μm × 0.5 μm scanning AFM picture. Capped QD: (c) artist’s view and(d) transmission electron microscope picture.

17.2 Epitaxial Growth and Structure of Quantum Dots

25 nm (a)

25 nm (b)

Figure 17.2 (a) Plan-view and (b) cross-sectional TEM images of arrays and stacks of InGaAs quantum dots in GaAs.

temperature, growth interrupts, etc. during the entire growth procedure. Typical sizes of self-assembled QDs are 2–10 nm in the growth direction and 5–30 nm in the lateral direction (see Figure 17.1d). They may be more or less disk shaped depending on the In–Ga intermixing during the growth of the capping layer. It is an inherent problem with self-assembled QDs that their locations are random and that there is a significant spread in the size distribution (see Figure 17.1b). The former is a problem if one wants to address individual QDs, and the latter will result in inhomogeneous broadening in the optical properties of ensembles of QDs. For certain applications, e.g. in lasers, a high density of QDs may be desired. For a very high layer density of QDs, they may self-arrange in a two-dimensional lattice as seen in Figure 17.2a. One may also stack several layers of QDs on top of each other separated by barrier layers thick enough to prevent electronic coupling between the layers, but thin enough so that the strain from the QDs in one layer serves a nucleation centers for QDs in the following layers (see Figure 17.2b). Thus, one may achieve dense three-dimensional lattices of QDs as active layers in QD lasers [11]. 17.2.2

Site-Controlled Growth on Patterned Substrates

It is a major challenge to grow semiconductor QDs at predestined locations. One possibility is to use the above-mentioned tendency of self-assembled QDs to nucleate at sites with a strain singularity. Thus, one may form a prepatterned substrate by lithography and etching techniques on which subsequently a pattern of QDs may be grown. The optical quality of the QDs grown directly on the patterned substrate is usually not very good due to impurities and defects from the lithographic processes. If, however, this layer is covered by a thin capping layer, a new layer of QDs may be grown following the strain pattern from the first layer and now without the impurities and defects. Following this scheme, two-dimensional patterns of good quality QDs are produced successfully [12]. A few groups have also managed to grow high-quality QDs directly on prepatterned substrates, the so-called pyramidal QDs [13, 14]. In these works, an etch mask is lithographically formed on the substrate followed by a wet etch leaving an array of etch pits in the form of inverted pyramids. If now, after proper cleaning, a standard quantum-well structure is grown on this substrate, a combination of QWs, QWRs, and QDs will form in the pyramidal etch pits due to different growth rates on the exposed crystallographic planes (see Figure 17.3). The QDs will form naturally in the bottom of the inverted

505

506

17 Optoelectronic Properties and Applications of Quantum Dots

100 μm 450 nm 270 nm

QWR

(111) A

QD

(111) B buffer layer

2.4 mm

Figure 17.3 Design of the substrate pattern before growth from Reference 15. The pitch of the triangular array of etched pyramids is 450 nm, and the side of the pyramids is 270 nm. An SEM image of the patterned substrate is depicted (center). A schematic top view of the pyramidal recess after growth is shown to the right. A QD is formed at the apex of the pyramid and QWRs at the wedges.

pyramid and with a self-regulated size that is only controlled by the width of the grown quantum well. In this scheme, well-defined arrays of relatively uniform QDs have been produced [15]. 17.2.3

Natural or Interface Quantum Dots

QDs may also form naturally from the lattice-matched growth of GaAs/AlGaAs hetero structures. If the layer-by-layer growth of AlGaAs is interrupted at half a monolayer (ML) before switching to GaAs and again when switching back to AlGaAs, the well width fluctuations of a thin GaAs quantum well will be sufficient to laterally confine both electrons and holes to form natural QDs or localized two-dimensional excitons (see Figure 17.4) [16]. These QDs are to a high degree free from strain and impurities and can exhibit quite variable sizes. They have been instrumental in the fundamental studies of the optical properties of semiconductor QDs [17]. A variant of interface QDs, and an alternative to SK-grown QDs are the so-called sub-monolayer (SML) QDs. SML InGaAs/GaAs QDs can be formed by alternating growth of 0.5 ML InAs and 2.5 ML GaAs repeated a number of times (e.g. 10) as sketched in Figure 17.5. Due to strain, the InAs islands will align more or less on top of each other and create a pillar of high indium content. Ultimately, In–Ga intermixing will result in the formation of In-rich QDs embedded in a QW of lower indium content, a so-called quantum-dot-quantum-well (QDQW) structure [18]. One of the advantages

AlGaAs GaAs

QD

AlGaAs (a)

(b)

Figure 17.4 (a) Monolayer fluctuations at both interfaces of thin GaAs quantum wells can localize quasi-two-dimensional excitons to form natural quantum dots. (b) Artist’s view of the potential landscape for the excitons in a thin quantum well with well width fluctuations.

17.2 Epitaxial Growth and Structure of Quantum Dots

InAs

GaAs

[110] [110]

20 nm SQD

LQD (a)

(b)

Figure 17.5 (a) Schematics of sub-monolayer (SML) growth of smaller (SQD) or larger (LQD) InAs/GaAs quantum dots. (b) Plan-view TEM picture of three layers of SML QDs.

of SML growth of QDs is a high density of relatively uniform QDs well suited for QD lasers. 17.2.4

Quantum Dots in Nanowires

The art of growing nanowires by the vapor-liquid-solid (VLS) scheme [19] has recently developed tremendously. In this technique, metal (Au) nanoparticles deposited on a substrate are heated and melted, and the resulting solid–liquid interfaces then serve as nucleation centers for the epitaxial growth from the precursor (III–V) materials in the vapor phase. This way, one can grow micrometer-long nanowires with diameters from tens to hundreds of nanometers. By varying the composition of the precursor materials during the growth, one can form p-n junctions as well as QDs along the wire directions (Figure 17.6a) as well as in the axial direction. Since this type of nanowire growth can be performed strain free on almost any substrate, irrespective of lattice mismatch, devices based on nanowire QDs have many potential applications. The growth of III–V nanowires may result in either zinc blende or wurtzite crystal structures having slightly different band structures. This is a problem if the crystal phase varies in an uncontrolled manner during the growth. If, however, the crystal phase can be controlled during the

(a)

(b)

Figure 17.6 (a) Schematic VLS-grown GaAs (light gray) nanowire with InAs (dark gray) section forming a QD. (b) Etched GaAs micropillar with few SK-grown QDs embedded.

507

508

17 Optoelectronic Properties and Applications of Quantum Dots

growth, it opens up the possibility of growing nanowire QDs with monolayer control over their position and size [20]. This is important if one wants to address and apply individual QDs or a well-controlled linear array of QDs. An alternative way to form and address individual or few QDs is to grow a layer of low-density InAs QDs centrally in a GaAs layer and then subsequently apply e-beam lithography and etching techniques to produce micro- or nanopillars with a single or few QDs embedded (Figure 17.6b). In both cases, Bragg reflectors below and above the QD layer can be formed by alternating growth of GaAs and AlAs and thereby placing the QD in a high-Q, small-volume cavity. These techniques have been successfully applied to produce small nanolasers and effective single-photon sources, as we shall see later.

17.3 Excitons in Quantum Dots The predominantly covalently bonded III–V semiconductors considered here have a direct band gap at the center of the Brillouin zone with a single conduction band of s-type symmetry (angular momentum 𝓁 = 0) and three valence bands of p-type symmetry (𝓁 = 1), allowing for direct optical transitions. The valence bands are inherently degenerate at k = 0. However, spin-orbit coupling and electronic confinement completely lift this degeneracy so that a simple two-band model composed of the s-like conduction band and the p-like heavy-hole valence band describes quite well, in an effective-mass approximation (parabolic bands), the electronic structure of the QDs [21]. The elementary electronic excitation, the exciton, in semiconductor QDs is an electron–hole pair excited across the electronic bandgap within the QD as defined by the bulk energy gap and the quantization energies within the conduction and valence bands, respectively, and modified by the Coulomb interaction between electron and hole. One can write an effective-mass Schrödinger equation for the QD exciton: H EM (r e , r h )X(r e , r h ) = Ex X(r e , r h )

(17.1a)

where X(r e , r h ) is the exciton wave function, Ex is the exciton energy, and ℏ2 ∇2e ℏ2 ∇2h e2 − + Ve (r e ) + Vh (r h ) − H EM (r e , r h ) = − 2me 2mh 4𝜋𝜀0 𝜀r |r|

(17.1b)

is the effective-mass Hamiltonian. r e(h) is the electron (hole) coordinate and ∇e(h) operates on the corresponding coordinates, me(h) is the effective electron (hole) mass, V e(h) (r e(h) ) is the confining electron (hole) potential, e is the elementary charge, 𝜀0(r) is the vacuum (relative) permittivity, and r = r e − r h . This equation can in general only be solved by numerical methods, but it is customary to distinguish between three different cases of (1) weak, (2) intermediate, and (3) strong confinement [21, 22]. In the limiting case of no confinement (V e (r e ) = V h (r h ) = 0), Eq. (17.1a) can be solved analytically with Ex,n (K) = Eg −

Exb ℏ2 |K |2 + n2 2M

(17.2a)

17.3 Excitons in Quantum Dots

and Xn (r e , r h ) = eiK ⋅R ue (r e )uh (r h )fn (r)

(17.2b)

where n = 1, 2, 3,… is a quantum number, Eg is the bulk bandgap, M = me + mh is the exciton mass, K = k e + k h is the exciton wave vector, and here we have introduced the exciton binding energy Exb =

𝜇e4 32ℏ2 𝜀20 𝜀2r

(17.2c)

with the reduced effective mass 𝜇 = m a0 𝜀r 𝜇0 ,

me mh . me +mh

The corresponding exciton Bohr radius is

where a0 is the hydrogen Bohr radius, and m0 is the free electron mass. ax = In Eq. (17.2b), we have further introduced the exciton center-of-mass coordinate m r + mh r h R= e e . ue(h) (r e(h) ) is the electron (hole) Bloch function, and fn (r) is an me + mh exciton envelope function containing a normalization factor, spherical harmonics, and Laguerre polynomials as known from the hydrogen solution. 17.3.1

Quantum-Dot Bandgap

Including the electron and hole confining potentials in the QD Hamiltonian (Eq. (17.1b)), the exciton energies, and thereby the QD bandgap, will depend more or less strongly on the size of the QD as described below: (1) In the weak confinement limit, the electron and hole quantization energies ΔEe , ΔEh ≪ Exb corresponding to ax ≪ L, where L is a linear extension of the QD. In this case, the relative motion of the electron and hole is essentially unaffected, and only the translational energy of the exciton is quantized by the confinement, resulting in a bandgap energy 3ℏ2 𝜋 2 ℏ2 𝜋 2 QD b or E = E − E + (17.3) g g x 2ML2 2MR2 for a cubic QD with side L and a spherical QD with radius R, respectively, and assuming infinitely high barriers. For natural or interface QDs, the localized excitons may be modified by the strong confinement in the growth direction, as in a quantum well, and the much weaker confinement in the lateral direction, thus approaching the situation of two-dimensional (2D) excitons with an increased overlap between electron and hole wave functions as well as an increased exciton binding energy. For b = 4Exb and a2D = 0.5 ax [21]. ideal two-dimensional (2D) excitons, we have E2D (2) In the case of intermediate confinement, the exciton and the quantization energies are comparable in size and have to be included on an equal footing. This can only be done with knowledge of the detailed structure and shape of the QD and requires complex numerical calculations. If in this regime a spherical QD has a radius between the electron and hole Bohr radii, and neglecting the Coulomb energy between electron and hole, one can in the literature [21] find an approximate expression for the QD bandgap energy as EgQD = Eg − Exb +

EgQD ≅ Eg + 0.67Exb

a2x 𝜋 2 R2

(17.4)

509

510

17 Optoelectronic Properties and Applications of Quantum Dots

(3) In the strong confinement limit of small QDs (L ≪ ax ), the electron and hole quantization energies are large compared to the exciton binding energy (ΔEe , ΔEh ≫ Exb ), which can then be incorporated as a perturbation. For a spherical QD with radius R, one approximate expression is [21] EgQD ≅ Eg +

ℏ2 𝜋 2 1.78e2 − + 0.752Exb 2𝜇R2 𝜀0 𝜀r R

(17.5)

Hence, the electronic structure and bandgap of the QDs are largely determined by the individual quantization energies of electrons and holes and only the finer details (splittings) are influenced by excitonic Coulomb and exchange interactions [23]. If we completely neglect the Coulomb interaction, the solution to Eq. (17.1) can be factorized X(r e , r h ) = Ψe (r e )Ψh (r h )

(17.6a)

with the electron (hole) wave function |Ψe(h) ⟩ = |Fe(h) ⟩|ue(h) ⟩|se(h) ⟩

(17.6b)

where |F e(h) ⟩, |ue(h) ⟩, and |se(h) ⟩ are the envelope function, the Bloch function at the center of the Brillouin zone, and the spin state for the electron (hole), respectively. If we further assume infinite barriers and isotropic effective masses, a cubic QD with a side length L has electron and hole quantization energies as ΔEe =

3ℏ2 𝜋 2 3ℏ2 𝜋 2 ; ΔE = h 2me L2 2mh L2

(17.7)

resulting in a QD bandgap energy EgQD = Eg + ΔEe + ΔEh = Eg +

3ℏ2 𝜋 2 2𝜇L2

(17.8)

To estimate the conditions for weak or strong confinement, we notice that ax ≅ 12 nm in GaAs and ax ≅ 37 nm in InAs. In this chapter, we will predominantly be concerned with InAs or InGaAs QDs embedded in a GaAs or AlGaAs matrix, which can be reasonably well described in the strong confinement picture. In all the above cases, it is clear that the QD bandgap increases when the size of the QD decreases and predominantly as one over the square of the linear dimension of the QD, as expressed in Eqs. (17.3)–(17.5) and (17.8). This dependence of the bandgap energy on the size of QDs has very useful applications in optoelectronic devices, like solar cells. The incorporation of certain size QDs can enhance absorption of solar radiation in a desired wavelength region.

17.3.2

Optical Transitions

The optical transition rate W if in a two-level system can in the dipole approximation be described by the Fermi golden rule as [24] Wif =

2𝜋 ̂ int |Ψh 𝜙i ⟩|2 𝜌(E) |⟨Ψe 𝜙f |H ℏ

(17.9)

17.3 Excitons in Quantum Dots

where |𝜙i(f ) ⟩ describes the initial (final) photon wave function, and 𝜌(E) is the density of final states with energy equal to the initial state and √ e e ℏ ̂ ̂ ̂ ⋅ Aq,𝜆 ≅ ̂ ⋅̂ p H int = a+q + ̂ aq } (17.10) p eq,𝜆 {̂ m0 m0 2𝜀0 𝜀r V 𝜔 ̂ q,𝜆 is the vector potential operator of the ̂ is the electron momentum operator, A where p single radiation mode with wave vector q, polarization state 𝜆 (unit vector ̂ eq,𝜆 ), and ̂ angular frequency 𝜔. Aq,𝜆 is further expressed in terms of the photon creation and annihilation operators ̂ a+q and ̂ aq , respectively. Since the envelope function in Eq. (17.6b) is slowly varying compared to the periodic Bloch function, the momentum operator only affects the latter when substituted in Eq. (17.9). Similarly, the photon operators only operate on the photon wave function. From Eqs. (17.6b), (17.9), and (17.10) we can now obtain an expression for the transition rate at frequency 𝜔 as Wif (𝜔) ≅

𝜋e2 fosc (𝜔t ) f (1 − ff )fq,𝜆 𝜌(𝜔)𝛿(𝜔 − 𝜔t ) 6m0 𝜀0 𝜀r i 3∕2

(17.11)

where 𝜌(𝜔) = ℏ𝜌(E) = 𝜀𝜋r 2 c𝜔3 is the density of photon states per unit volume in a homoV geneous medium (c is the velocity of light in vacuum); ℏ𝜔t = Eg + ΔEe + ΔEh is the electronic transition energy, i.e. the new bandgap, in the QD; f i(f ) is the probability that the initial (final) electronic state is occupied; f q, 𝜆 = nq, 𝜆 + 1 when a photon is emitted; and f q, 𝜆 = nq, 𝜆 when a photon is absorbed, nq, 𝜆 being the number of photons in mode (q, 𝜆). In Eq. (17.11), we have further introduced the oscillator strength, which in the strong confinement case takes the form [22, 25, 26] fosc (𝜔t ) =

2

Ep 6 ̂ ⋅̂ |⟨Fe ∣ Fh ⟩|2 |⟨ue ∣ p eq,𝜆 ∣ uh ⟩|2 = |⟨F ∣ F ⟩|2 m0 ℏ𝜔t ℏ𝜔t e h

(17.12)

̂ ⋅̂ where Ep = m6 |⟨ue ∣ p eq,𝜆 ∣ uh ⟩|2 is the well-known Kane energy [27]. Thus, the tran0 sition probability depends on the overlap integral of the electron and hole envelope functions, the symmetry of the conduction and valence band Bloch functions, and transitions are only allowed if no spin flip is taking place during the transition (Δj = ± 1; j = 𝓁 + s). If the QD is initially unoccupied (f i = 1, f f = 0), we then obtain the absorption rate as Γabs (𝜔) =

𝜋e2 fosc (𝜔t ) n(𝜔)𝛿(𝜔 − 𝜔t ) 6m0 𝜀0 𝜀r

(17.13)

where n(𝜔) is the density of photons at frequency 𝜔. Similarly, we obtain, if the QD is initially occupied by one electron–hole pair (exciton), the spontaneous radiative recombination rate √ e2 𝜀r 𝜔2 fosc (𝜔t ) 𝛿(𝜔 − 𝜔t ) (17.14) Γrad (𝜔) = 6𝜋m0 𝜀0 c3 To estimate the magnitude of f osc from Eq. (17.12) for strongly confined Inx Ga1-x As QDs, we note that Ep (x) = (28.8 − 7.3x) eV [27]. Thus, for pure InAs QDs with Ep = 21.5 eV and ℏ𝜔t ≅ 1.20 eV, an upper bound for the oscillator strength is f osc ≤ 18. In practice, it is very difficult to determine the oscillator strength of QDs from absorption experiments [28]. However, spontaneous emission can be conveniently studied in

511

512

17 Optoelectronic Properties and Applications of Quantum Dots

XX

Γ

Xb,x Xd

ΔEdb

γdb

Γd,nrad Γb,nrad

G

Γxy,rad

xx,rad

Γx,rad

Xb,y

ΔEfss

Γy,rad

G

Figure 17.7 QD energy level scheme showing ground state (G), dark excitons (Xd ), fine structure split (ΔEfss ), bright excitons (Xb,x and Xb,y ), and biexcitons (XX). The solid arrows indicate linearly polarized (x and y) radiative transitions, the dashed arrows show nonradiative transitions, and the dotted arrows indicate spin flip transitions between bright and dark excitons.

micro-photoluminescence (μ-PL) experiments with high spatial and spectral resolution [29]. Until now we have not considered the spin states in Eq. (17.6b). There are four different combinations of electron and hole spins in an exciton: |↑e ↓h ⟩, |↓e ↑h ⟩, |↑e ↑h ⟩, |↓e ↓h ⟩. Since a hole with spin-down corresponds to the absence of an electron in the valence band with spin-up, excitons with |↑e ↓h ⟩ and |↓e ↑h ⟩ are optically allowed (bright excitons, Xb ) and excitons with |↑e ↑h ⟩ and |↓e ↓h ⟩ are optically forbidden (dark excitons, Xd ). The dark excitons usually have lower energy (ΔEdb ) than the bright excitons and may only recombine non-radiatively. Due to a slight asymmetry of the confinement potential in the x-y plane for QDs grown on (001) surfaces, the exciton doublets (Xb,x , Xb,y ) exhibit a fine structure splitting (ΔEfss ) due to exchange interactions causing the optical transitions of the bright excitons to be linearly polarized in the respective x and y directions (see Figure 17.7). A QD can contain two electrons (of opposite spins) and two holes in their respective lowest quantized state, allowing for the formation of biexcitons (XX in Figure 17.7) as well as trions that are negatively charged excitons (X− ) containing two electrons and one hole or positively charged excitons (X+ ) containing one electron and two holes (not shown in Figure 17.7). Linearly polarized optical transitions may take place between the biexciton and exciton states (Γxx,rad and Γxy,rad in Figure 17.7). Negatively (positively) charged trions may recombine emitting right (left) circularly polarized light and leaving an electron (hole) behind. These excitonic transitions can be identified in a μ-PL experiment with a high spatial and spectral resolution [29]. Biexciton transitions are identified by their superlinear excitation density dependence and only appear when the average

17.4 Optical Properties

population of electron–hole pairs per QD exceeds one. Trion transitions are identified by their polarization (circular) and require the presence of excess carriers (doping).

17.4 Optical Properties Since quantitative absorption experiments are difficult to carry out on epitaxially grown QDs [28], the optical and optoelectronic properties of QDs are most conveniently studied in photoluminescence (PL) experiments where the QDs are excited via absorption in the barrier material and subsequent relaxation and capture of electrons and holes into their lowest quantized states in the QDs. The injection of electrons and holes into the QDs is particularly efficient for SK-grown QDs due to the wetting layer connecting them. In a simple PL experiment, a very large number of QDs is usually excited, resulting in a luminescence spectrum that is strongly inhomogeneously broadened, reflecting the size distribution (and the corresponding quantization energy distribution) of the contributing QDs. Figure 17.8a shows standard PL spectra from SK-grown Inx Aly Ga1-x-y As QDs with three different compositions. In all three cases, the spectral full width at half maximum (FWHM) exceeds 50 meV. Also shown is the response curve of a typical detector based on silicon charge-coupled devices (CCDs), indicating that if one wants to investigate the luminescence from a few QDs with good sensitivity, it is an advantage to add gallium and even some aluminum to the QDs. Figure 17.8b shows spatially and spectrally resolved PL of GaAs interface QDs. They have a somewhat narrower size distribution and emit in a wavelength range (photon energy above the GaAs bulk bandgap) that is convenient for efficient detection of single QD emission. Figure 17.9 shows examples of μ-PL spectra at low temperature from a low density of SK-grown QDs as viewed through a 500-nm metal aperture [30]. Lorentzian lines from individual QDs are resolved, showing a distribution of linewidths (FWHM) around

Ga

PL intensity

0.04

As

0.46

40 μm

In 0.5Ga0.5As In 0.5Al

CCD sensitivity InAs

GaAs Al0.08Ga0.92As

0 1.1

1.2

1.3

1.4

1.5

1.6

1.7 1.622

1.624

1.626

1.628

Photon energy (eV)

Photon energy (eV)

(a)

(b)

1.630

1.632

Figure 17.8 (a) PL spectra of SK-grown Inx Aly Ga1-x-y As QDs with three different compositions (x, y) = (1, 0), (x, y) = (0.5, 0), and (x, y) = (0.5, 0.04). The GaAs peak is from the substrate, and the bandgap of the AlGaAs barrier is indicated by the arrow. The sensitivity of a typical silicon CCD detector is shown by the dashed curve. (b) Spatially and spectrally resolved PL intensity (color coded) from natural QDs of GaAs formed by interface fluctuations in thin GaAs/AlGaAs quantum wells as in Figure 17.4.

513

17 Optoelectronic Properties and Applications of Quantum Dots

Number of lines

15

PL intensity

514

1.36

1.387

1.38 1.40 1.42 Photon energy (eV)

1.388

1.44

1.389

10

5

0 20

40 60 80 100 Linewidth (μeV)

1.390 1.391 Photon energy (eV)

1.392

1.393

Figure 17.9 High-resolution PL spectrum at 10 K as detected through a 500 nm aperture. The solid line is a Lorentzian fit to the data. The left inset shows all the lines from the aperture, and the right inset shows the linewidth statistics.

40 μeV (when corrected for instrumental resolution ∼20 μeV) for In0.5 Ga0.46 Al0.04 As QDs, and around 16 μeV for GaAs QDs (as in Figure 17.8b). It is a question whether these are the true homogeneous linewidths of the QD transitions and to what extent pure dephasing mechanisms contribute beyond lifetime broadening. This can be resolved by speckle analyses [31] and transient four-wave mixing (TFWM) experiments [30], as we shall see later.

17.4.1 Radiative Lifetime, Oscillator Strength, and Internal Quantum Efficiency First, we look at time-resolved μ-PL from the samples in Figure 17.8 to find the radiative recombination rate and thereby possibly determine the oscillator strengths of the QDs (Eq. (17.14)). Figure 17.10 shows PL decays of ensembles of QDs. There is some variation in the decay time within the size distribution of QDs [30], but self-assembled In0.5 Ga0.46 Al0.04 As QDs typically have a lifetime of T 1 =750 ps, while for the GaAs interface QDs T 1 = 60 ps. If we assume QDs to have an internal quantum efficiency of one, this would correspond to f osc = 21 for self-assembled QDs and f osc = 268 for interface QDs. The inferred oscillator strength for the self-assembled QDs is consistent with the above-found upper limit for strongly confined QDs. The much larger value found for the interface QDs suggests that they are only weakly confined with a (giant) oscillator strength such as the ones found for bulk excitons fosc (𝜔) =

Ep Vcoh Ep VQD ≤ 3 ℏ𝜔 𝜋ax ℏ𝜔 𝜋a3x

(17.15)

where V coh is the coherence volume of the exciton, and V QD is the volume of the QD. The interface QDs may be viewed as quasi-two-dimensional with hard walls (infinite barriers) in the growth direction and a weak parabolic potential in the lateral directions

17.4 Optical Properties

T=5K PL Intensity (a.u.)

Figure 17.10 Low-temperature PL decays of ensembles of GaAs interface QDs (solid circles) and In0.5 Ga0.46 Al0.04 As SK-grown QDs (open circles).

10 SK QDs T1 = 750 ps Interface QDs T1 = 60 ps 1

0

10

20

30

40

50

Time (ps)

for which one can obtain [22] ( ) 2Ep L 2 fosc (𝜔) = ℏ𝜔 ax

(17.16)

where L is the lateral diameter of the QD defined as four standard deviations of the Gaussian center-of-mass wave function, and ax ≅ 8 nm is the radius of a quasi-2D exciton in GaAs. The giant oscillator strength found above would then suggest a QD diameter L ≥ 20 nm, which is very reasonable. The assumption of a high internal quantum efficiency of excitons in low-dimensional structures at cryogenic temperatures is commonly used and is often valid for interface QDs. However, in self-assembled QDs nonradiative recombinations may contribute to the lifetime of the bright excitons and thereby the PL decay (see Figure 17.7). This can be tested by noting that radiative and nonradiative recombinations depend differently on the local environment of the QD. Specifically, the radiative decay rate will depend on the local density of optical states (LDOS) 𝜌l (r, 𝜔), which may vary significantly with the position of the QD in a nanostructured environment such as a photonic crystal [32] or close to a semiconductor–air (or metal) interface [25]. Thus, the total decay rate of bright excitons may be expressed as Γb,total (r, 𝜔) = Γb,nrad + Γrad (𝜔)

𝜌l (r, 𝜔) 𝜌(𝜔)

(17.17)

For a QD positioned at a distance z from a semiconductor–air interface, 𝜌l (z, 𝜔) can be calculated exactly and is shown in Figure 17.11 for a dipole oriented parallel or perpendicular to a GaAs–air interface. In Reference [25], spontaneous PL decays from self-assembled InAs QDs were recorded as a function of the distance z from the sample surface (GaAs–air interface). With the QDs polarized parallel to the interface and fitting with the LDOS in Figure 17.11, the radiative and nonradiative recombination rates could be independently determined and thereby also the oscillator strength (Eq. (17.14)) as well as the internal quantum efficiency: 𝜂int =

Γrad Γrad = Γb,total Γrad + Γb,nrad

(17.18)

515

17 Optoelectronic Properties and Applications of Quantum Dots

Figure 17.11 Normalized LDOS 𝜌l (z, 𝜔)/𝜌(𝜔) as function of distance z from a semiconductor–air interface calculated for dipole oriented parallel (full curve) and perpendicular (dashed curve) to the interface.

Normalized LDOS

1.2 1.0 0.8 0.6 0.4 0.2 0.0 0

50

100 150 200 Distance, z (nm)

250

300

17

1.00

16

0.95

15

0.90

14

0.85

13

0.80

12

0.75

Quantum efficiency

Oscillator strength

516

11 1.16

1.18

1.20

1.22

1.24

1.26

0.70 1.28

Photon energy (eV)

Figure 17.12 Oscillator strength (triangles) and internal quantum efficiency (squares) of SK-grown InAs QDs as a function of photon energy within the inhomogeneously broadened PL spectrum.

These measurements were performed for different photon energies within the inhomogeneously broadened PL spectrum, and the corresponding values for the oscillator strength and the internal quantum efficiency are shown in Figure 17.12. Since the photon energy increases with decreasing size of the QDs, one may conclude that the larger SK-grown QDs have a better overlap between the electron and hole wave functions, yielding a higher oscillator strength and internal quantum efficiency, but without reaching the level of interface QDs. In any case, the observed QD exciton lifetimes do not fully account for the linewidths observed in single QD PL spectra. Therefore, there must be significant contributions from pure dephasing mechanisms. 17.4.2

Linewidth, Coherence, and Dephasing

For a homogeneously broadened two-level system, the relation between the linewidth 𝛿 (FWHM) and the total dephasing rate 𝛾 is 𝛿 = 2ℏ𝛾, where 1 1 1 = + (17.19) 𝛾≡ T2 2T1 T2∗ and where T1 , T2∗ , and T2 are the population lifetime, pure dephasing time, and total coherence or dephasing time, respectively. T 1 and T 2 are also called longitudinal and transverse lifetimes, respectively.

17.4 Optical Properties

A direct way of observing the coherence and dephasing of QD emitters is by analyzing the speckle pattern in the secondary emission from a QD sample resonantly excited by a short laser pulse [31]. The secondary emission intensity I(t) in non-specular directions is a mixture of coherent resonantly enhanced Raleigh scattering and incoherent spontaneous emission: (17.20)

I(t) = Icoh (t) + Iincoh (t)

The coherence c(t) of the secondary emission can be defined and subsequently calculated as √ √ √ Icoh (t) √ (I(t) − I(t))2 = √N (17.21) c(t) ≡ 2 I(t) I(t) where the bars denote an average over emission directions, and c is calculated from the intensity variance over a large number of speckles (N > 100). For an inhomogeneously broadened ensemble of QDs, with energy variance ℏ𝜎, the intensity of the coherent emission is given by 2 2

Icoh (t) = Icoh (0)e−2𝛾t (1 − e−𝜎 t ) ≈ Icoh (0)e

− T2t

where the approximation is valid for t > 𝜎 , and we have used 𝛾 = coherence of the secondary emission is given by −1

c(t) =

Icoh (0)e I(0)e

− T2t

− Tt 1

2

= I0 e

− T2t∗ 2

(17.22)

2

1 . T2

In this limit, the

(17.23)

Thus, the coherence calculated as a function of time from the observed speckle pattern (Eq. (17.21)) will decay exponentially with a rate determined solely by the pure dephasing time as in Eq. (17.23). The speckle patterns are conveniently recorded by a streak camera, where one axis pictures the emission angle and the other axis the time as shown in Figure 17.13, where the coherence decay of an ensemble of interface GaAs QDs is also displayed, showing a pure dephasing time of T2∗ = 270 ps. This is significantly longer than the corresponding population lifetime T 1 = 60 ps (Figure 17.10). Hence, the total dephasing time (see 2T T ∗ Eq. (17.19)) T2 = 2T 1+T2 ∗ = 83 ps is strongly influenced by the short lifetime in these 1 2 interface QDs. It is remarkable that this technique allows for the determination of the homogeneous linewidth in strongly inhomogeneously broadened systems without involving nonlinear spectroscopies such as transient four-wave mixing (TFWM) to be described in the next section. 17.4.3

Transient Four-Wave Mixing

In a degenerate TFWM experiment, two input pulses of frequency 𝜔 are incident on the sample in the directions k 1 and k 2 . Via a third-order nonlinearity (𝜒 (3) ), the input pulses induce a third-order nonlinear polarization that subsequently emits a four-wave mixing (FWM) signal at the same frequency 𝜔 in the direction 2k 2 − k 1 (Figure 17.14a). This signal is then detected time-integrated as a function of the delay τ between the input pulses (Figure 17.14b).

517

17 Optoelectronic Properties and Applications of Quantum Dots

3.6

3 2 1

Log (Icoh/I)

Angle (degrees)

0 –1

T2* = 270 ps

3.4

3.2

–2 –3

3.0 0

10

20 30 Time (ps)

40

50

0

60

20

40 60 Time (ps)

(a)

80

(b)

Figure 17.13 Speckle analysis of resonantly excited secondary emission from GaAs interface QDs. (a) Time- and angle-resolved speckles recorded by a streak camera. (b) The coherence decay as calculated from the speckles in (a) revealing the pure dephasing time T2∗ . –2.5

2k2-k1 k2 k1

τ

Log (FWM intensity)

518

–3.0 T2 = 372 ps

–3.5

(a) –4.0

0

20

40

60

80

Delay τ (ps) (b)

Figure 17.14 (a) Transient FWM experiment on an ensemble of SK-grown QDs. (b) Decay of the time-integrated FWM signal in the direction 2k2 -k1 as a function of the delay τ between the input pulses from which the dephasing time T2 is determined by Eq. (17.25).

For a homogeneously broadened two-level system, the intensity of the FWM signal, I FWM (τ), which reflects the time-integrated polarization decay after the arrival of the second pulse, is given by IFWM (τ) = I0 e

− T2τ

2

;

τ≥0

(17.24)

from which the total dephasing time T 2 can be determined. If, however, the two-level system is inhomogeneously broadened, as from an ensemble of QDs, the FWM signal appears as a photon echo, at t = τ, with the result that the time-integrated echo decays as IFWM (τ) = I0 e

− T4τ

2

;

τ≥0

(17.25)

17.4 Optical Properties

Table 17.1 List of lifetimes (T 1 ), dephasing times (T 2 and T2∗ ), homogeneous linewidths (𝛿), and μPL linewidths for interface and SK-grown QDs. Emitters

T2∗ (ps)

T 2 (ps)

𝜹 (𝛍eV)

T 1 (ps)

Interface QDs

80

270

60

SK-grown QDs

372

500

750

15

𝛍-PL (𝛍eV)

10–50

3.5

10–50

Thus, the homogeneous dephasing time, and thereby the homogeneous linewidth, can still be determined from a strongly inhomogeneously broadened system. Figure 17.14b shows TFWM results from an ensemble of SK-grown QDs at low temperature probed at the low-energy part of the spectrum, where contributions from the ground state transitions of the QDs dominate [30]. Since the QD transitions are strongly inhomogeneously broadened, the observed TFWM decay reflects a dephasing time T 2 = 370 ps. Taking into account the lifetime T 1 = 750 ps found above, this implies a pure dephasing time T2∗ = 500 ps, which is somewhat longer than that found for the interface QDs above. The low-temperature (5 K) lifetimes, dephasing times, and homogeneous linewidths 𝛿 = 2ℏ T2 for the two QD systems considered here are summarized in Table 17.1 and compared with the observed μ-PL linewidths (right column) from the time-integrated spectra of individual QDs, showing that additional broadening mechanisms are present, particularly for SK-grown QDs. A very likely contribution is from spectral wandering caused by the Stark shift from fluctuating charges close to the QDs. The dephasing rates and linewidths recorded here are temperature dependent (see Figure 17.15) and found to increase linearly with temperature T at low temperatures due to scattering with acoustic phonons and exponentially at higher temperatures, where longitudinal optical (LO)-phonon scattering dominates. The temperature dependence of the linewidth 𝛿(T) is given by 𝛿(T) = 𝛿0 + akT +

b

(17.26) e −1 Here, a and b are constants, k is Boltzmann’s constant, and ℏ𝜔LO is the LO-phonon energy. Figure 17.15 shows linewidths from InAlGaAs QDs as a function of temperature ℏ𝜔LO kT

70 60 Linewidth δ (μeV)

Figure 17.15 Linewidth (FWHM) of SK-grown InAlGaAs QDs as a function of temperature measured directly from μ-PL spectra (open circles) and as calculated from TFWM experiments (full circles). The curves serve as a guide to the eye, indicating the low-temperature slopes.

50 μ-PL

40

0.50 μeV/K 30 20 FWM

10

0.45 μeV/K 0

0

10

20 30 Temperature (K)

40

519

520

17 Optoelectronic Properties and Applications of Quantum Dots

as measured directly from μ-PL spectra and as calculated from TFWM experiments, of which the latter represents the true homogeneous linewidth. The phonon scattering coefficient a ≅ 0.5 μev K−1 found from Figure 17.15 is three to six times smaller than for excitons in InGaAs QWs. The strong increase in the homogeneous linewidth at higher temperatures will be important for the application of QDs in devices operating at room temperature. By extrapolating the FWM to zero temperature, we find 𝛿 0 ≅ 2 μeV in good agreement with the expected lifetime broadening when the lifetime T 1 ≅ 750 ps as in Figure 17.10.

17.5 Quantum Dot Applications The intense research on semiconductor QDs has opened up a number of optoelectronic applications within information and communication technologies (ICTs). Some are already on a commercial level such as the application of colloidal QDs in high-definition TV displays, light emitting devices, solar cells, and the application of self-assembled QDs in semiconductor lasers and semiconductor optical amplifiers (SOAs), others are still at the development stage. Here, we shall focus on two different applications of epitaxially grown semiconductor QDs, namely, in semiconductor diode lasers (DLs) and SOAs, and as efficient single-photon emitters for quantum communication and quantum computing purposes. The first one involves dense ensembles of QDs operating at room temperature, implying a short dephasing time and a strong homogeneous broadening. The second application is based on individual QDs operating at low temperature to maintain the coherence time as long as possible to ensure the coherence and indistinguishability of the emitted photons. 17.5.1

Quantum Dot Lasers and Optical Amplifiers

QDs, or artificial atoms, were early on considered as an ideal medium for achieving optical gain and lasing. This was based on the similarities with the optoelectronic properties of atoms supplemented by the technology and ease of integration shared with conventional semiconductor devices. The anticipated advantages include wavelength selectivity and tunability, low threshold (transparency) currents, large material and differential gains (oscillator strength), low temperature sensitivity, low chirp, and high modulation bandwidths. Many of these early predictions [33, 34] were based on idealized assumptions of a dense ensemble of QDs with equal size and shape formed by lattice-matched heterostructures with infinite potential barriers and single confined electron and hole levels, even combined with conflicting assumptions of bimolecular e-h recombination and equilibrium carrier distributions. In reality, self-assembled quantum-dot devices consist of strained heterostructures with finite potential barriers and several electron and hole levels featuring strong inhomogeneous broadening due to a significant variation in size and shape. Recombination is by nature monomolecular (excitonic) and often from a non-equilibrium carrier distribution. One of the fundamental differences between ideal atoms and QDs is the unavoidable interaction with the solid-state host lattice, giving a finite width to the ideal discrete (transition) energy levels. Still, quantum-dot devices are potentially superior to quantum-well devices, especially with respect to transparency current [35], slope

Threshold current density (A/cm2)

17.5 Quantum Dot Applications

105

pn

Room temperature

104 DHS 10

QW

3

QD 102 101 1960

1970

1980

1990

2000

2010

Year

Figure 17.16 Development over time in diode laser threshold current density from simple GaAs p-n junctions over GaAs/AlGaAs double hetero structures (DHSs) and quantum wells (QWs) to quantum dots (QDs).

efficiency, and chirp [36], provided the growth technology is well controlled. Moreover, the technology is expected to be further advanced in analogy with the developments in the quantum-well technology experienced in the past. The high material gain (oscillator strength) within the QDs further enhance the possibilities for lasing in microcavities or nanocavities, and one can even speculate on the minimum number of QDs required to obtain lasing. The progress in semiconductor DL technology is illustrated in Figure 17.16, showing the development over time of the laser threshold current density for bulk, quantum-well, and quantum-dot devices, respectively [37]. The QD size is a crucial parameter for device applications. If the dots are too small, there will be no (populated) discrete states, and if the dots are too large, the inter-level spacing will be too small to have only the ground state populated. In the InAs/GaAs system, this means that the dot size should be in the range 4–20 nm. A high QD density is required to increase the interaction volume of the active QD region. The fine-tuning of the size is, together with the composition, an additional means of selecting the wavelength of the ground state transition in the QDs. In order to reach communication wavelengths in the range 1.3–1.55 μm, special care needs to be taken by either controlling the size or the alloying (or both) of the QDs. In order to reach these wavelengths, it has proved beneficial to overgrow the InAs dots with a thin layer of InGaAs, which reduces the strain and increases the size of the QDs at the same time [38]. Because of the high material gain coefficient g mat that can be achieved within a QD, the transparency current of a QD laser is inherently low. It is in principle reached when the QDs on average are occupied by one electron–hole pair (exciton). This is, however, counteracted by the much lower modal gain g = Γg mat , where Γ is the confinement factor, i.e. the ratio between the volume of gain material and the mode volume of the laser cavity. In SK-grown QDs, material gain coefficients as high as g mat = 105 cm−1 can be achieved. For a typical dot areal density of 4 × 1010 cm−2 and a QD volume V QD ≅1000 nm3 , the confinement factor for a single dot layer is estimated to be 10−4 . Hence, a material gain of 105 cm−1 results in a modal gain of 10 cm−1 for a laser structure with a single layer of QDs. This is sufficient for lasing in long cavities (L > 1 mm), whereas shorter cavity lasers will require higher dot volume density, e.g. more dot layers in the active region.

521

522

17 Optoelectronic Properties and Applications of Quantum Dots

Figure 17.17 Sketch of p-i-n laser or optical amplifier structure with QDs as the active medium.

I L Ridge Metal p-doped Barrier QD WL

H

n-doped Metal W

Figure 17.17 shows a sketch of an edge-emitting p-i-n laser or optical amplifier structure where the intrinsic active region, typically 120 nm thick (H), consists of stacked InAs/InGaAs QD layers with GaAs barriers. The AlGaAs p-doped and n-doped surrounding layers provide the injection of carriers, holes, and electrons respectively, into the QDs as well as optical confinement (double hetero structure). Guiding is provided by a ridge waveguide of width W = 6–8 μm and length L = 0.3–3.0 mm. The end facets may be uncoated or high-reflection coated for laser operation. For operation as an optical amplifier, they are tilted and possibly anti-reflection coated. 17.5.1.1

Gain Dynamics

Some of the important issues for the operation of QD lasers and optical amplifiers relate to their dynamic behavior, which is controlled and limited by carrier capture and relaxation into the QDs. This has been investigated in an electrically pumped QD amplifier by femtosecond time-resolved differential transmission experiments. We will here show examples from an MOVPE-grown p-i-n structure such as the one in Figure 17.17, where the active region consists of three stacked layers of InAs/InGaAs QDs with a luminescence spectrum peaking around 1.1 μm [39, 40]. E The transmission coefficient, or device gain G = Eout = egL , where Ein (Eout ) is the pulse in energy at the input (output) of the device and g is the modal gain coefficient, is shown in Figure 17.18 as a function of the input pulse energy for 150 fs pulses centered at 1.08 μm and for different forward bias currents of the device. For small input energies, an absorption of −3.35 dB (10 × log G) is found at zero bias (0 mA), transparency occurs at 4 mA, and a maximum gain of 1.85 dB is reached at 10 mA. From the corresponding modal gain g (see inset of Figure 17.18), a differential gain of 4.2 cm−1 mA−1 is deduced before saturation. This saturation is caused by strong filling of the QD ground state and occupation of excited states. With increasing input pulse energy, a bleaching of the absorption and a depletion of the gain occurs due to stimulated transitions induced by the intense pulse. For high input energies (>1 pJ), two-photon absorption (TPA) causes an additional reduction in the output pulse energy at all bias currents. The gain and index dynamics were investigated in pump-probe experiments for pump input energies still in the small signal regime (indicated by the vertical dotted line in Figure 17.18), where TPA plays a minor role [39, 40]. In Figure 17.19, the gain change in dB deduced from the probe transmission change is shown at forward bias currents 0, 4, and 20 mA as a function of the pump-probe delay. A striking difference between the absorption and the gain recovery dynamics appears. At zero bias a pump-induced

17.5 Quantum Dot Applications

1.85 dB

20 mA 4 mA

–2.27 dB

0 mA

–3.35 dB Gain coefficient (cm–1)

Device gain G = egL

1.0

1.14 dB

0.1

0.01

10 5 0 –5 dg/dI= 4.2±0.2 cm–1mA–1

–10 –15 –20

0

0.1

5 10 15 20 Bias current (mA) 1 Input energy Ein (pJ)

10

100

Figure 17.18 Gain of a QD semiconductor amplifier at room temperature as a function of input pulse energy and for different forward bias currents (0–20 mA). Inset shows the gain coefficient as a function of bias current, revealing a differential gain of 4.2 cm−1 mA−1 . Figure 17.19 Change of device gain as a function of delay after a 150 fs pump pulse of energy 0.27 pJ for three different forward bias currents: 0, 4, and 20 mA.

1.5 Epump = 0.27pJ

ΔG (dB)

1.0 0 mA

0.5

4 mA

0

20 mA

–0.5 0

2

4 6 Delay (ps)

8

10

bleaching of the absorption occurs that reaches a maximum value of 1.4 dB and then recovers over several picoseconds. At 4 mA forward bias, corresponding to (average) transparency, the transmission changes are small, and at 20 mA forward bias, a gain compression up to −0.7 dB is observed that recovers in less than 0.3 ps. These data show that the pump-induced bleaching of the absorption builds up with the time integral of the pulse, according to a recovery time much longer than the pulse duration, while the gain recovers on a time comparable to, or shorter than, the pulse duration. This is also reflected in the higher maximum absorption bleaching compared to the gain compression.

523

524

17 Optoelectronic Properties and Applications of Quantum Dots

At zero bias, the dots are initially empty. When the pump arrives, excitons are created in the dot ground state, leading to a bleaching of the absorption by spectral hole-burning (SHB). This recovers by spontaneous recombination of excitons or by carrier escape from the dot ground state. Typical spontaneous exciton lifetimes in QDs are several hundreds of picoseconds, as measured by time-resolved photoluminescence, much longer than the measured absorption bleaching recovery, which is then caused by carrier escape from the dots by phonon absorption, i.e. carrier heating. The absorption recovery can be fitted by a bi-exponential response function with two equally weighted time constants of 1.25 and 5.9 ps attributed to the escape times of holes and electrons, respectively [41]. At transparency, the SHB effects are small, and the remaining contribution to the gain changes is mainly due to TPA. At stronger bias with positive gain, the dots are initially occupied, and pump-stimulated emission leads to a reduction of the exciton population (gain compression) that recovers by relaxation of carriers into the dot ground state. The measured recovery time is, therefore, a direct measure of the carrier capture and relaxation into the dot ground state in a QD amplifier under working conditions [42]. The used bias current of 20 mA corresponds to a modal gain that is just enough for ground state lasing of the investigated structure, and is not an unrealistic working point for QD lasers. From a fit to the gain recovery at 20 mA, taking into account TPA, we estimate a time constant of 115 fs ± 10 fs [40], slightly shorter than the pulse duration. The measured ultrafast SHB recovery under gain thus indicates ultrafast carrier relaxation in electrically pumped QD amplifiers, which is very promising for high-speed applications. The carrier relaxation occurs probably via Auger scattering from the excited states, partly occupied at 20 mA, into the dot ground state, as theoretically proposed [43]. Note that the occurrence of such ultrafast scattering processes implies fast dephasing times, i.e. large homogeneous broadening (10–20 meV), which is confirmed by FWM experiments as discussed below. 17.5.1.2

Homogeneous Broadening and Dephasing

Some of the predicted superior performances of quantum-dot lasers rely on a delta-like density of states. This is, however, affected by the homogeneous and inhomogeneous broadenings. The inhomogeneous broadening due to size and/or strain fluctuations in real QD structures usually exceeds the homogeneous broadening. PL linewidths below 20 meV in ensembles of InAs/InGaAs QDs have been achieved [44], and the room-temperature dephasing time T 2 has been determined in femtosecond time-resolved FWM (TR-FWM) experiments [45]. It was found weakly dependent on the optical excitation intensity and attributed to phonon interaction, and the resulting ≅ 5 meV [46]. However, the intrinsic homogeneous homogeneous linewidth was 𝛿 = 2ℏ T2 linewidth in a real QD laser at working conditions has to be considered [47]. Figure 17.20 shows the time-integrated FWM (TI-FWM) signal (field amplitude) as a function of the delay between the two input pulses for different forward bias currents in a QD SOA. For low bias currents (100 nm) and low quantum efficiency (∼ 0.1%). In 2012, the first 3D perovskite emitters, CH3 NH3 PbBr3 nanoparticles in porous alumina [87], were reported, and the possibility of perovskites as emitters was effectively rediscovered. As a result, the research on PeLEDs has grown dramatically and focused on achieving high efficiency, achieving very high color purity (∼20 nm) and an external quantum efficiency (EQE) exceeding 20% so far [88], as represented in Figure 18.8. In comparison, organic emitters exhibit broad spectra (full width at half maximum [FWHM] ≥ 40 nm). Inorganic QD emitters exhibit relatively narrow (FWHM ∼30 nm) but size-sensitive spectra in limited dimension (≤10 nm). The lead halide PeLEDs also have a very wide color gamut (400 nm ≤ 𝜆 ≤ 780 nm) in the CIE diagram. This highlights the promise of perovskites as electroluminescent emitters, especially in display applications. In summary, comparing traditional quantum dots and organic semiconductors, perovskites exhibit great advantages including wide color gamut, low cost, high color purity, and color tunability. Since the active emission layers of PeLEDs are thin films composed of low-dimensional perovskite grains or structures, this section

549

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

FWHM ~ 30 nm

FWHM > 40 nm

D

400 500 600 700 Wavelength (nm)

LE

D

ic

LE

400 500 600 700 Wavelength (nm)

Q

D

O

10%

an

15%

400 500 600 700 Wavelength (nm)

rg

20%

EL Intensity (a.u.)

30% 25%

FWHM < 20 nm

EL Intensity (a.u.)

Max EQE

EL Intensity (a.u.)

550

5% Year 1990

1995

2000

2005

2010

2015

2020

Figure 18.8 (Top) Photos of high-efficiency perovskite LED devices in room temperature. (Bottom) Record external quantum efficiency (EQE) of perovskite, inorganic QD, and organic emitters based LEDs. Source: Adapted from Lin et al., 2018 [88] and Kim et al., 2016 [89].

reviews the current progress of PeLEDs using two approaches toward achieving high performance devices: layered structure and low-dimensional 3D structure. In 2015, Song et al. demonstrated multicolor PeLEDs (blue, green, orange) with CsPbX3 QDs, where X = Cl, Br, I [90]. These devices exhibited narrow linewidths (FWHM ninter 10

nmax, cm

-dE/dx, eV/nm

100

Strong interaction Weak interaction nmax < ninter

1019

1

Mean free path, nm

100 60 10 40

20

1

0,01

0,1 1 10 100 Electron energy E, KeV

Mean energy per scattering, eV

1018

0 1000

Figure 18.23 (Top frame, left scale) Linear energy deposition [dE/dx (eV/nm)] of an electron in NaI versus electron energy from 6 eV (the band gap) up to 1 MeV. (Top frame, right scale) Conversion to electron–hole pair density (eh/cm3 ) assuming a 3 nm effective radius of the electron track and that an average of 2.8Egap is invested per electron–hole pair. (Bottom frame, left scale) Mean free path for electron–electron scattering in NaI vs. energy. Source: From A.N. Vasil’ev in [140].

(top) is obtained [140]. It indicates that a 600 keV photoelectron produces roughly 1018 e-h cm−3 at the start of the track, increasing to about 2 × 1020 e-h cm−3 when the electron energy has been degraded to a few hundred eV. It is worth pointing out the high electron–hole pair densities near the track end, because such densities typically lead to nonlinear quenching (nonradiative recombination) processes that can cause lower detection yield and part of the nonproportional response mentioned earlier, degrading the resolution. The photoelectric cross section depends strongly on the atomic number, Z, of an element. For comparing the photon interaction cross sections and attenuation coefficients of compounds or mixtures of elements, it has been found useful to define an approximate effective atomic number, Zeff . The calculation method for Zeff that we will employ in Table 18.1 is the one due to Rodnyi [143]. For a compound Ax By Cz , the weight fraction

567

568

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

of each element is first defined as WA = xMA ∕(xMA + yMB + zMC )

(18.1)

where MA , MB , and MC, are the atomic masses of A, B, and C elements, respectively. In the approximation of the dominance of the photoelectric effect, the expression for the effective atomic number is given as [143] Zeff = (WA ZA 4 + WB ZB 4 + WC ZC 4 )1∕4

(18.2)

Ejection of a photoelectron into the material leaves behind a hole in the state it initially occupied, most often in a core level, and therefore it could be kilovolts deep. That binding energy of the electron initial state is subtracted from the kinetic energy of the ejected photoelectron, but that kinetic energy deficit can be and often is degraded finally to valence electron–hole pairs through the processes of Auger cascade and/or x-ray emission and reabsorption. In such cases, the full energy of the photoelectric absorption process is usefully converted to valence electron–hole pairs to be counted by the semiconducting or scintillating detector in the “full-energy peak” (see Figure 18.21). In the Auger cascade, an electron from the same or adjacent atom bearing the core hole fills it in the presence of a spectator electron or hole (Auger process) that carries away the energy to be degraded to low-energy e-h pairs by valence electron scattering (see the previous paragraph) or by creation of secondary shallower core holes on which the Auger cascade continues sequentially. While we are discussing Auger processes, it is worth mentioning that the phenomenon plays an additional role in spectroscopic radiation detection. Rather than contributing to the production of valence electron–hole pairs for sensing in the detector, as discussed just above for Auger cascade relaxation of core holes, valence Auger recombination at high electron densities found near track ends discussed two paragraphs above act to quench or destroy electron–hole pairs. While all such three-particle nonradiative recombination events can technically be considered Auger processes, we prefer, for clarity, to regard dipole–dipole quenching as distinct from free-carrier Auger recombination. In dipole–dipole quenching, an electron and hole exist on an excited atom or ion, and the spectator electron (or hole) on a nearby excited atom or ion carries away (in an ionizing higher excited state) the energy of the original recombining e-h pair. The sum of the rates involving each of the available spectator electrons or holes gives the Auger rate for quenching of the two closely located excited atoms. The same physical process can also be described as dipole–dipole energy transfer from one excited atom transitioning to its ground state with near-field excitation of the neighboring excited atom to an even higher excited state. The latter description allows writing the kinetic rate as second order in concentration of the excited atom or ion species. Free-carrier Auger recombination rate is third order in the concentration of electrons and holes. These nonlinear quenching or nonproductive recombination processes are particularly important at the high excitation densities encountered near track ends and/or low particle energy, or with heavy and multiply charged particles. Both the dipole–dipole and free-carrier quenching processes reduce the available signal to be counted and introduce nonlinearity (related to nonproportionality and loss of resolution) in the spectroscopic signal [144, 145]. The length of a photoelectron track starting with 600 keV energy is very roughly 100 μm in the solid materials chosen for radiation detection. We say “roughly,” because

18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects

60 simulated electron tracks in Nal

50

electron initial energy: 150keV

distance (μm)

40 30 20 10 0 10 20 30

0

10

20

30

40

50

60

distance (μm)

Figure 18.24 Simulated electron energy-loss trajectories projected in two spatial dimensions for 150 keV electrons in NaI, using GEANT4 Monte Carlo toolkit.

the trajectory of a track is highly random. Monte Carlo toolkits such as GEANT4 [146] can be used to simulate the shape and length of typical tracks, along with the corresponding linear energy deposition history contained in – dE/dx. Examples of an initial photoelectron energy of 150 keV in NaI are shown in Figure 18.24. The stopping power of a material interacting with an energetic electron (directly incident or produced within the material by gamma scattering, photoelectric events, electron–positron pair production, or charged particle scattering) refers to the rate of energy loss with distance, essentially the linear energy deposition rate (−dE/dx) discussed above. Electron stopping occurs by many electromagnetic scattering processes spaced very closely, from e-h cluster depositions tens of nanometers apart at megavolt electron energy to e-h generation at 1 nm or less separation for 1 kV electron energy. The track simulations in Figure 18.24 illustrate such electron stopping. In contrast, gamma ray stopping occurs in a few widely spaced events if it is by Compton scattering, or just one local event followed by electron stopping if it is primary photoelectric absorption, or by a mixture if a Compton-scattered secondary gamma ray experiences photoelectric absorption rather than a continued Compton cascade. Compton scattering describes the interaction of a gamma ray with a free or weakly bound electron. As pictured conceptually on the left side of Figure 18.21, the gamma ray’s direction and energy changes upon scattering from the electron, which carries away the necessary energy and momentum. As a fast electron in the detector material, this electron is stopped in the way discussed above, depositing electron–hole pairs to be counted. But the light or current sum no longer corresponds to the incident gamma ray energy for spectroscopic purposes unless the Compton-scattered secondary gamma ray and any subsequent scatters all occur in the material before any gamma escapes the boundary. This aspect is also illustrated in Figure 18.21. The sum of all partially stopped Compton-scattered gamma events contributes to the signal emission below

569

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

the “Compton edge” labeled in Figure 18.21. Compton scattering can contribute to the full-energy peak located above the Compton edge if all Compton scatter events occur within the detector. As already remarked, the full-energy peak is useful for spectroscopy to distinguish the energy of an incident gamma ray, whereas the Compton edge is not. The relative probability of Compton scattering increases with the ratio of the valence electron density to the core electron density, so small-Z elements will emphasize Compton over photoelectric interactions (see Figure 18.22), which is undesirable for spectroscopy and dosimetric applications. For a choice of a given effective Z and stopping power, it is obvious that a small detector will have a bigger fraction of events in the Compton edge than in the full-energy peak, compared to a bigger detector. Detector size and the proportion of valence electrons as well as density become factors to consider when exploring a relatively new class of materials containing organics, such as the HOIP. X-ray escape peaks can also be discussed with reference to Figure 18.21 as an analogy and Figure 18.25 as an actual example of K escape peaks in CsPbBr3 . Photoelectric absorption produces core holes as discussed above. Instead of always decaying entirely by Auger cascade, the radiative emission of characteristic x-rays is also a core-hole relaxation mechanism. If the x-ray is absorbed again within the detector, it still contributes 57Co

source

–150 V

–900 V

1000

Backscatter peak Compton edge

400 Pb escape peak

122 keV

Count (a.u.)

Count (a.u.)

570

Compton edge Cs escape peak

500

Pb escape peak

200

662 keV 137Cs

136 keV 0

source

0 0

40

80 120 Energy (keV)

160

0

200

400 600 Energy (keV)

(a)

4 x 4 x 3 mm3

160

(b)

5 x 5 x 3 mm3

6 x 6 x 3 mm3

Figure 18.25 Energy-resolved spectra of (a) 3 × 3 × 0.9 mm3 CsPbBr3 room-temperature semiconductor detector with 57 Co gamma source, and (b) 4 × 2 × 1.24 mm3 CsPbBr3 detector with 137 Cs source. The bottom photograph shows an as-grown single-crystal ingot with diameter 11 mm and cut crystals at sizes used in the measurements. Source: He et al., 2018 [147]. https://www.nature .com/articles/s41467-018-04073-3. Licensed under CCBY 4.0.

18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects

to the total electron–hole pair count in the full energy peak. But if the x-ray escapes the detector volume similar to the scattered gamma ray pictured in Figure 18.21, it represents a deficit of energy that is sharply defined for a characteristic x-ray unlike a scattered gamma ray. Therefore, a small replica of the full-energy peak shifted down by the escaped characteristic x-ray energy occurs. Lead halide perovskite materials have been tried for fabricating both photoconducting and scintillating radiation detectors. The photoconducting (semiconductor type) perovskite detectors have met with the best success so far, and so we begin our survey of progress with examples of lead halide perovskite semiconductor detectors. The most advanced detector of that group presently is the melt-grown CsPbBr3 single-crystal detector announced in 2018 by the groups of Mercouri Kanatzidis and Duck Young Chung in Northwestern University and Argonne National Laboratory [147]. Its energy resolution (R, as defined above) was reported in 2018 as 3.9% for 122 keV gamma rays and 3.8% for 662 keV. The pulse height spectra and photographs of an ingot and cut crystals are reproduced in Figure 18.25. To put this in perspective, the resolution of commercialized cadmium zinc telluride (CZT) tested at the same 122 keV energy and simple planar measurement setup as the CsPbBr3 was 4.1% [147]. With special contact design, pixilation, and pulse discrimination analysis, CZT detection has been highly developed to yield the best performance for selected ingot sections reported as about 1% at 662 keV [148]. CZT is presently the most widely used room-temperature semiconductor detector and is the second in resolution of solid-state detectors only to high-purity germanium (HPGe). HPGe has an excellent energy resolution, e.g. 0.2% at 1 MeV, but it is very expensive and requires cryogenic operation, so there are many applications for which it is not practical. An intense quest continues (with lead halide perovskites among the new entries) to find a cost-effective room-temperature spectroscopic detector with a resolution targeted at 3% or better. CZT itself is rather expensive because of high-temperature growth and the need to select sections from grown crystals to try to avoid defects that otherwise ruin the performance. CsPbBr3 grows at a lower temperature and already seems much less impacted by defects of both cluster size and atomic size. In fact, it appears to be an excellent example of defect tolerance, conveying the remarkably long hole lifetime of 25 μs at room temperature [147]. The μ𝜏 product for holes is 1.34 × 10−3 cm2 V−1 . Melt-grown CsPbBr3 is neither fully optimized nor commercialized as a radiation detector as of this writing. It is a relatively “young” material in the specific sense of trials for radiation detection, so the above comparison to the performance of CZT under the same measurement conditions suggests optimism for the melt-grown all-inorganic single crystals. Hybrid organic–inorganic perovskites discussed below are even younger with respect to trials of ionizing radiation detection. Both perovskite material groups considered for semiconducting radiation detection stand to benefit from the defect tolerance property associated notably with the perovskite semiconductors, especially from their employment as solar cells. Evidence for and concepts of defect tolerance in the perovskites have been reviewed and discussed in References [70, 84, 149], for example, as well as being discussed in the photovoltaic section of this chapter. The important point is that more often in the perovskites than in most other materials, the energy levels of electrons and/or holes trapped on lattice defects lie within the conduction band or valence band, respectively, and so are unstable; i.e. trapping does not occur. The reason we phrased it in this awkward way

571

572

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

is that the reported phenomena include theoretical results wherein the empty trap level may indeed lie within the bandgap, but the occupied trap state moves into degeneracy with the band and ionizes. Semiconducting gamma detectors made using HOIP materials have so far exhibited significantly poorer energy resolution than the all-inorganic melt-grown perovskite discussed above. Selected reports in the literature up to 2018 are summarized in Table 18.2. The performance of the melt-grown CsPbBr3 [147] was already discussed above. It and the representative entries for CZT will serve as the standards for performance comparison at this point of solution-grown inorganic and hybrid perovskite semiconducting detectors. Looking first at the energy resolution column, one notices that there are just five reports of resolution at this point among the solution-grown perovskite detectors we found, and three of those five are reported only for low-energy 59.6 keV gamma rays, and it is in the range of 35%. The single study reporting the resolution for higher energy [154, 155] reports 6.5% (best) up to 12% resolution at 662 keV. Clearly, these are not yet at competition with the melt-grown CsPbBr3 paradigm of the perovskite class, nor the commercial benchmark CdZnTe (CZT). As has been discussed and analyzed in some of the reports, part of the shortcoming at this point in the testing of hybrid perovskite detectors is the small size of the detectors tested relative to the effective stopping range of gamma rays, which of course increases with the gamma ray energy. The escape of gamma rays or scattered gamma rays is aggravated to some degree by the relatively lower density (Table 18.2) and lower effective Z of the hybrid perovskites relative to the Cs perovskites. However, the hybrid perovskite semiconductors often have an excellent carrier lifetime and compete fairly well in the μ𝜏 product and background resistivity relative to the CZT standard. The transport properties are good as a material class; however, the combination of modestly good density and effective Z, which is a fundamental physical characteristic of organic hybrid perovskites, together with the currently available and used sample size, is allowing many gamma events to escape from contributing to a full-energy peak. This degrades the energy resolution, often negating it at high gamma energies such as 662 keV, where Compton escape becomes serious. An answer would seem to be increasing the detector size. However, as several of the groups have pointed out, that remedy works in a semiconducting radiation detector only to the point that one can still collect charge reliably (proportionally to the gamma ray energy) over the dimensions of the larger detector. The effective collection range (μ𝜏E), distortion of the electric field, and other inhomogeneities have to be solved along with a scale-up in size [148]. If we approach the review of scintillating lead halide perovskite detectors in the same material order of presentation as we did for the semiconducting detectors, the lead-off material in our introduction of perovskite radiation detectors, i.e. melt-grown single-crystal CsPbBr3 [147], is as much a failure as a scintillator as it was a success for photoconduction. Bulk crystalline CsPbBr3 is a dark material at room temperature, and still not a good light emitter even at low temperature. The optical properties of bulk CsPbBr3 and Cs4 PbBr6 single crystals and melted or annealed film perovskites are known from a series of studies from 1996 onward [130, 164–168]. In addition to determination of bandgaps and exciton luminescence energies of the two compounds in the early work, Nitsch et al. [164] have characterized the light yield of ∼ 2.35 eV green luminescence from bulk crystalline CsPbBr3 as weak even at low temperature.

Table 18.2 Material, processing, gamma energy resolution (or x-ray dosimetric sensitivity as appropriate), carrier mobility-lifetime product, lifetime, bulk resistivity, mass density, effective atomic number calculated by the method of Rodnyi [143] and physical thickness of the tested sample along the incidence direction, for selected semiconductor spectroscopic radiation detectors using lead halide perovskite materials, reported through 2018. Energy resolutions were measured at room temperature unless otherwise indicated. Proc.

Resolution @E

𝛍𝝉 (cm2 V−1 )

𝝉

𝝆 (𝛀-cm)

Dens (g cm−3 )

Z eff

References

CsPbBr3

melt

3.8% @662 keV 3.9% @122 keV

1.34 × 10−3 (h) 8.77 × 10−4 (e)

25 μs

3.4 × 1011

4.83

65.9

[147]

CsPbBr3

sol.

27% @59.6 keV for T = 220K

2 × 10−4

2 × 109

FAPbI3

sol.

35% @59.6 keV

1–1.8 × 10−2

109 –1011

MAPbI3

sol.

9.2% @122 keV

8.1 × 10−4 (h)

17 μs (h)

6.8% for T = 275K

7.4 × 10−4 (e)

11 μs (e)

Material

MaPbBr3

∼1 × 10−3 (h)

sol.

−2

MAPbBr3-x Clx

sol.

6.5% @662 keV

1.8 × 10

Cx F1-x PbI3-y By

sol.

no res @662 keV 50% @59.6 keV

0.04–0.12

Cd0.9 Zn0.1 Te (CZT)

melt

4.1% @122 keV 1% @ 662 keV

5 × 10−5 (h) 2.5 × 10−3 (e)

x-ray dosimetric sensitivity (𝛍CGy−1 cm−2 )

5 μs

65.9

[150] [151]

108 –109

4.15

66.8

7.8 × 1010

3.83

67.1

3.6 × 109

∼ 3.8

[152] [153] [154, 155] [156]

108 –109

5.86

50

[147, 157]

7 × 108

4.15

66.8

[158]

4.15

66.8

[159]

66.8

[160]

Pulse response time

MAPbI3

sol.

2.5 × 104

2 × 10−7

MAPbI3

sol.

3.8 × 103

1 × 10−4

MAPbI3

sol.

2.1 × 104 @60 keV

MAPbI3

sint.

2.53 × 103

2 × 10−4

Cs2 AgBiBr6

sol.

105

6.3 × 10−3

350 ps

3.76 109 –1011

66.8

[161]

59.8

[162, 163]

574

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

The weakness was attributed to nonradiative energy transfer to trap states, and the luminescence was especially weak at room temperature in view of thermal quenching attributed to bimolecular recombination of free or trapped carriers [164]. Therefore, bulk CsPbBr3 has not been found suitable as a scintillator, but the flexibility afforded by HOIP and all-inorganic perovskites in confined dimensions (e.g. quantum dots or wires) show some promise or potential. We begin with Table 18.3, which summarizes published findings on scintillating radiation detectors using perovskite materials up through 2018. This is meant as a parallel summary to Table 18.2 for the semiconducting perovskite radiation detectors. However, the column for energy resolution in Table 18.3 is mostly devoid of entries with one exception, because the scintillating lead halide perovskites generally have not exhibited full-energy peaks so far in their pulse height spectra. Phenethylamine lead bromide (Phe-PbBr4) in row 6 of Table 18.3 is an exception to this generalization, with full-energy peak resolutions reported [172]. The most promising practical successes of lead halide perovskite scintillators so far have been in the form of thin or moderately thick (100 μm) films composed of nanoparticles, employed for dosimetric x-ray imaging. If the film is sufficiently thin, escape of light from the interior of the scattering and absorbing collection of nanoparticles does not seriously attenuate the high efficiency of light emission from individual nanoparticles. In that sense, the nanoparticle films that are thick enough to stop x-rays in the target energy range can function with respect to light extraction similarly to the LED devices discussed in the previous section, with similar bright operation. The x-ray scintillation of such lead halide perovskite imagers was shown to be bright by Chen et al. [169]. Such films are useful only in dosimetric and imaging mode operations, not in pulse height resolution for gamma ray energy analysis. The latter application requires stopping a sufficient number of the gamma rays for a full-energy peak. That requires much thicker films, which in turn raises the problem of getting the light out from the interior. Many versions of lead halide perovskite scintillators suffer from self-absorption of their own scintillation light due to small Stokes shift of the exciton luminescence below the absorption edge. We will discuss one example of trying to deal with this later. For the scintillators in Table 18.3 that are primarily x-ray dosimetric, the best figure of merit is the “x-ray detection limit (μGyair s-1 ),” abbreviated in the column as “x-ray lim.” For the few that have yielded a gamma pulse height spectrum with a resolvable full-energy peak, a column for the gamma light yield, abbreviated “𝛾 Lt Yield,” lists values. As we conceded in the beginning, Table 18.3 for scintillating perovskite detectors has fewer data entries than Table 18.2 for semiconducting detectors. Only one of the entries reports the energy resolution, and that is 35% at 662 keV and 49% at 122 keV. For comparison, resolutions of 6% and 2% in NaI:Tl and LaBr3:Ce:Sr at 662 keV represent a standard low-cost detector and the best-resolution detector among current scintillators. Note that the thickness of the PhePbBr3 scintillator crystal used by Kawano et al. was 4 mm [172]. Assuming that transparency issues are solved, the useful size of scintillator crystals can in general be increased more readily to capture full-energy peaks at high gamma energies than can the size of semiconducting crystals. Possibly perovskite scintillators in radiation detectors could become the tortoise that has a chance to overtake the current semiconducting perovskite hares. For concreteness in discussing some of the issues involved in realizing a good scintillating detector from lead halide perovskites, we will now discuss investigations in the

Table 18.3 Material, processing method (melt or solution), measurement temperature, photoluminescence quantum yield, gamma light yield, gamma resolution (if any) at specified gamma energy, x-ray sensitivity limit as an imager, main decay time of scintillation light, and literature reference for scintillating radiation detectors using lead halide perovskite materials reported up through 2018. 𝛄 Lt Yield (phot/MeV)

Temp. (∘ K)

Photolum. Q Yield

both

300

< 1% (RT)

sol.

300

MAPbBr3

sol.

10 300

>150 000 ∼ 100

2D (EDBE)PbCl4

sol.

130 300

>120 000 9000

NR

PhePbBr4

sol.

300 300

14 000

35% @622 k 49% @122 k

Material

Proc.

bulk CsPbBr3 NPfilm CsPbBr3

Resolution @E

x-ray lim. (𝛍Gyair s−1 )

𝝉 decay

0.013

Reference

[169] [170]

NPfilm CsPbBr3

QD CsPbBr3

sol.

300

90%

CsPbBr3 /Cs4 PbBr6 nanocomposite

sol.

300

70%

NaI:Tl

melt

300

38 000

LaBr3 :Ce:Sr

melt

300

73 000

[171] [171] 11 ns

[172] [121]

1900

345 ps

[121]

6% @662 k

250 ns

[173]

2% @662 k

15 ns

[173, 174]

576

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

DMSO

CsB

PbBr2

Molar ratio > 1 Step 1

Acetone vapor < 24 hrs Cs_Pb-Br crystal

Cs-Pb-Br precursor 0.2 M

Step 2

Step 3

Step 4

Figure 18.26 Depiction of steps in growth of Cs4 PbBr6 crystals containing nano-inclusions of CsPbBr3 .

authors’ laboratory of light emission and scintillation from all-inorganic cesium lead bromide perovskite nanocomposite materials. This direction of scintillation investigations was frankly motivated by the remarkably high PLQY from the similar nanocomposites used for the LEDs discussed in the previous section of this chapter. Optimism was suggested by the high density and effective Z of cesium lead bromide compounds, as well as by our studies on photoluminescence kinetics, electroluminescence, and the composite nanostructure of CsPbBr3 active islands in Cs4 PbBr6 host [121]. The time resolution seemed promising given the 345 ps photoluminescence decay time at 70% QY that we measured in the Cs4 PbBr6 / CsPbBr3 nanocomposite [121]. In the previous section on nanocomposite perovskite LEDs, films were spun down from the CsBr/PbBr2 precursor solution in a ratio chosen to form grains of Cs4 PbBr6 containing nano-inclusions of CsPbBr3 , which are the radiative recombination centers. For a chance of gamma ray scintillation, we need a much thicker nanocomposite, ideally a single crystal of the host Cs4 PbBr6 with embedded nanocrystals. The solution crystal growth method depicted in Figure 18.26 was used. Crystals of up to 7 mm dimension on the edge and about 1 mm thickness were obtained after cycling successive growth stages with the renewed precursor solution. Figure 18.27 shows them under white light

Figure 18.27 Crystals of the Cs4 PbBr6 /CsPbBr3 nanocomposite (left) under white light, (middle) under ultraviolet illumination emitting green light that saturates the camera response to appear white, and (right) UV-illuminated crystal.

18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects 500k

1 A1 = 720.6 (29.8%); τ1 = 0.107 ns Intensity (a.u.)

Photons/nm/s

400k 300k 200k

A2 = 16.4 (4.8%); τ2 = 0.764 ns

0.1

A3 = 11.0 (65.4%); τ3 = 15.5 ns

0.01

100k 0

1E–3 300

400 500 Wavelength (nm)

600

0

5

10 15 Time (ns)

20

25

30

Figure 18.28 (a) X-ray excited luminescence spectrum and (b) decay in the CsPbBr3 /Cs4 PbBr6 nanocomposite macroscopic crystals from Wake Forest University, measured at Lawrence Berkeley National Laboratory. The x-ray energy was ≤40 keV.

(appearing yellow) and under ultraviolet illumination (emitting green, which saturates the camera response to white). The x-ray excited luminescence spectrum of the Cs4 PbBr6 /CsPbBr3 nanocomposite is shown in Figure 18.28a. The intense luminescence at 530 nm is from the CsPbBr3 nano-inclusions, and the weaker band near 330 nm might be spectrally shifted from the 375 nm intrinsic emission of the pure Cs4 PbBr6 host. The time response of the nanocomposite excited by picosecond x-ray pulses with photon energies ≤40 keV is shown in Figure 18.28b. The two main decay components are observed with decay times of 107 ps and 15.5 ns. In the following, we report the optical response and kinetics of pure CsPbBr3 and the CsPbBr3 /Cs4 PbBr6 nano-inclusion composite under high-intensity photoexcitation, and preliminary testing with x-ray and gamma excitation. Taking a cue from the earlier findings of Nikl et al. [164, 167] and Kondo et al. [165, 166] on the nano-precipitates of CsPbBr3 in CsBr, multiple studies have now shown that the PLQY of CsPbBr3 nanocrystals, including CsPbBr3 nano-inclusions in Cs4 PbBr6 , under steady blue or UV illumination can be much higher (a factor of up to 1000) than that of bulk CsPbBr3 [103, 121]. However, under a high-intensity femtosecond laser excitation (60 μJ cm−2 in 300 fs at 420 nm), the streak camera images in Figure 18.29a,b show that the pure CsPbBr3 micro-grain powder emits a roughly equal peak intensity and a slightly larger time-integrated luminescence yield than the nanocomposite CsPbBr3 in Cs4 PbBr6 powder which had been approximately 1000 times brighter in the steady state [121]. The phenomena at work to produce this observation should include the competition among carrier trapping, bimolecular radiative recombination, and Auger decay such as discussed recently by Xing et al. for organic/inorganic hybrid lead iodide perovskite electroluminescence [131]. Applying their discussion to the inorganic perovskite light emitters of present interest, we first note that the low exciton binding energy reported as 19 meV [175] to 62 meV [176] in pure CsPbBr3 means that at room temperature the recombination kinetics are expected to be mainly those of free and trapped carriers, not excitons. Radiative recombination should thus obey bimolecular kinetics, and so the Xing et al. rate equation [131] applies. Qualitatively in this picture, the very low light yield of pure CsPbBr3 at micro-grain and larger size for the low carrier densities

577

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

Wavelength (nm) 425 450 475 500 525 550 575 600

Wavelength (nm) 425 450 475 500 525 550 575 600 0

0 Excitation

1

2

2

3

3

4 5

Time (ns)

Time (ns)

1

CsPbBr3 micro-crystal

6

4 5 6

7

7

8

8

9

9

10

10

(a)

Excitation

CsPbBr3 nano-crystal in Cs4PbBr6

(b)

1 Intensity (a.u.)

1 Intensity (a.u.)

578

0.1 1:1 1:1.2 1:1.5 1:2 CsPbBr3 Cs4PbBr6

0.01

0

1000

0.1 0.01 1E–3

CsPbBr3 CsPbBr3 fitting Cs4PbBr6 Cs4PbBr6 fitting

1E–4 1E–5 2000 3000 Time (ps) (c)

4000

0

1000

2000 Time (ps) (d)

3000

4000

Figure 18.29 Subnanosecond luminescence under 300 fs, 60 μJ cm−2 , 420 nm excitation. Streak camera images of (a) CsPbBr3 microcrystals and (b) CsPbBr3 nanocrystals in Cs4 PbBr6 perovskite at room temperature. (c) PL lifetime decay curves of Cs-Pb-Br thin films with different molar ratios, CsPbBr3 microcrystals, and CsPbBr3 nano-inclusions in Cs4 PbBr6 crystals. (d) Fitting analysis on exciton dynamics of CsPbBr3 microcrystals and CsPbBr3 nano-inclusions in Cs4 PbBr6 crystals.

typical of steady-state optical excitation or LED operation with a homogeneous active recombination layer is a result of the bimolecular recombination rate being smaller than the trapping rate at defects. Using the bimolecular rate constant of k2 = 7 × 10−10 cm3 s−1 tabulated and reported by Xing et al. [131] as typical for hybrid lead iodide perovskites, and LED injected carrier densities of n = 1015 cm−3 also typical of hybrid perovskite LEDs with homogeneous active layers, the initial rate of descent of the bimolecular luminescence excited by a pulse of moderate intensity can be characterized by an effective initial decay rate of (𝜏 PL 1/e )−1 = k2 n = 7 × 105 s−1 . This is significantly slower than the carrier trapping rate achieved without taking unusual steps to exclude defects, and thus corresponds (in the present case) to a very low light yield in pure extended CsPbBr3 excited at ordinary intensity. However, in our short-pulse laser experiment, the 60 μJ cm−2 of 420 nm laser light arrives as a 300 fs pulse. Taking into account the measured absorption coefficient of 3.3 × 104 cm−1 and the reflection loss measured in the course of PLQY determination,

18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects

such a pulse produces approximately 1.4 × 1018 e-h pairs per cm3 in CsPbBr3 . Putting this initial carrier density into the bimolecular rate equation with the rate constant k2 = 7 × 10−10 cm3 s−1 , we obtain a model prediction of the effective initial time constant 𝜏 PL 1/e = (k2 n)−1 = 1 ns for the conditions of our experiment, whose results are shown in Figure 18.29. The observed effective initial lifetime seen in Figures 18.29a,c, and e for the pure CsPbBr3 powder is 0.97 ns. This agreement with the prediction based on excitation density and estimated rate constant suggests that the simple rate equation model properly describes pure CsPbBr3 powder of micro-grain size such as those used in our streak camera experiment. The bimolecular radiative decay time becomes extremely fast (∼1 ns) for intense pumping and thus competes very well against the trapping rate, resulting in a dramatic increase of PLQY along with the fast decay time. Remember that this is bulk (micro-grain) CsPbBr3 , which exhibits a very low light yield under continuous or moderate-intensity pulsed excitation. The complete second-order decay curve soon bends toward the slower rate of decay as time progresses, and the carrier density falls. By fitting the streak camera data for pure CsPbBr3 powder to second-order kinetics, we obtain the fitted second-order curve as shown in Figure 18.29d with the time constant for decay to half value, t1/2 = 0.69 ns. We want to emphasize that in fitting Figure 18.29d for pure CsPbBr3 powder, the initial excitation density was fixed at the measured value 1.4 × 1018 e-h pairs per cm3 , so that there was only one variable fitting parameter, the second-order rate constant k2 . The fitting specifies k2 = 4.6 × 10−10 cm3 s−1 , to be compared to 7 × 10−10 cm3 s−1 , reported by Xing et al. [131] for extended (3D) hybrid lead iodide perovskites. Note that the embedded nanocrystal composite powder labeled as Cs4 PbBr6 in the streak camera data did not fare as relatively well in gaining luminescence intensity versus the pumping density as the pure CsPbBr3 powder. This is partly a statement to the effect that the composite started out quite bright at a low (average) intensity pumping achieved by steady illumination or LED injection, and then remained high in PLQY, possibly decreasing slightly due to dipole–dipole or Auger recombination, as the excitation density was increased. There was simply not headroom or opportunity for such a dramatic improvement of PLQY upon increasing the excitation density in the nano-inclusion composite because it already started out with a high PLQY at a low excitation density. The point is that quantum confinement effects in the composite already overcame the slow recombination that bimolecular kinetics would have predicted at a low injection or excitation density, in favor of confined electron–hole pairs. The nanocrystal boundaries in the Cs4 PbBr6 -encapsulated inclusions of CsPbBr3 enforce effective “excitons” despite the low binding energy of the actual excitons in CsPbBr3 , and consequently we found approximate first-order decay kinetics even at a low excitation density in a material whose excitons are thermally ionized in the bulk at room temperature. The data for the nanocomposite powder in Figure 18.29c,d could not be well fitted with bimolecular kinetics alone, but instead the main part (longer component) could be fitted with first-order kinetics (exponential decay time 𝜏 = 345 ps). The faster initial decay could be fitted with the addition of second-order kinetics having a rate constant k2 = 1 × 10−9 cm3 s−1 , corresponding to 𝜏 PL 1/e = (k2 n0 )−1 = 710 ps. The sum of the first- and second-order rates gives the steepening, which can be seen at the earliest time in Figure 18.29d. We suggest that the embedded nanocrystals exhibit first-order decay of the “excitons” (geminate e-h pairs) enforced at room temperature by nanocrystal confinement. The initial second-order decay at the highest

579

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

5

3

4

2 Energy (eV)

Energy (eV)

580

3 2 1

0 –1 –2

0 –1

1

Γ

A

H

K (a)

Γ

M

L

–3

Γ

X

S

Y (b)

Γ

Z

U

Figure 18.30 Calculated band structures of (a) Cs4 PbBr6 and (b) CsPbBr3 using PBE functional including spin-orbit (SO) interaction. The zero of energy is set at the top of the valence band. Source: From Kang and Biswas [177].

density is attributed to dipole–dipole Förster quenching of multiple geminate pairs per nanocrystal, often considered a kind of Auger quenching. As commented in the earlier discussion of the LED voltage threshold in this chapter, it appears that carrier injection may occur into the host Cs4 PbBr6 bands. If so, the CsPbBr3 nano-inclusions with a smaller bandgap represent heterostructure wells that could collect and concentrate the density of both carrier types in the radiative recombination regions, thus boosting the bimolecular recombination rate, and with it, the QY. Whether the band offsets of CsPbBr3 in Cs4 PbBr6 indeed form a Type I heterostructure with the advantageous carrier collection and concentration properties outlined has been examined in electronic band structure calculations by Kang and Biswas [177]. Their band structures of CsPbBr3 and Cs4 PbBr6 are shown in Figure 18.30 by DFT with the semilocal PBE functional, which underestimates the fundamental gap of these compounds. However, the band characteristics and spin-orbit splitting remain consistent. The dispersive band edges of CsPbBr3 do not support self-trapped carriers, which agrees with reports of weak exciton binding energy and high photocurrent. The decoupled octahedra in Cs4 PbBr6 create narrow, nondispersive valence and conduction bands, and further calculations reveal polaronic and lattice-coupled excitonic features in the 0-D material. Information regarding the relative band edge positions in Cs4 PbBr6 /CsPbBr3 heterostructures and composite crystals were predicted from the calculated ionization potential (IP) and electron affinity (EA) [177]. These were estimated from separate bulk and slab structure calculations of the respective compounds. An appropriately thick slab with an adequate vacuum was used to estimate the difference between the average electrostatic potential of the slab’s bulk-like center and the vacuum level. This potential step enabled Kang and Biswas to align the bulk valence band maximum (VBM, obtained from a separate host calculation) with the vacuum level and obtain the IP. The IP values could then be used for band edge alignment between Cs4 PbBr6 and CsPbBr3 . The calculated IPs are sensitive to the orientation and termination of the slab surfaces, which may introduce some ambiguity in the results. Still, a comparison of the values of the IP and estimates of EA (obtained using reported experimental bandgaps) provides important insight regarding band alignment. According to Figure 18.31, it is of type I character; i.e. the VBM and conduction band minimum (CBM) of CsPbBr3 are contained within the gap of Cs4 PbBr6 .

18.4 Ionizing Radiation Detectors Using Lead Halide Perovskite Materials: Basics, Progress, and Prospects

vacuum

EA = 3.22 eV

Eg = 3.96 eV IP = 7.18 eV

Cs4PbBr6

EA = 4.08 eV

Eg = 2.25 eV

IP = 6.33 eV

CsPbBr3

Figure 18.31 Calculated ionization potential (IP) and electron affinity (EA) relative to the vacuum level (dashed line). Experiment bandgaps [178] are used to align CBMs of Cs4 PbBr6 and CsPbBr3 . Source: From Kang and Biswas [177].

Prediction of a Type I heterostructure of CsPbBr3 in Cs4 PbBr6 is itself very promising for prospects of a scintillator with Cs4 PbBr6 serving as a host in which gamma rays are stopped, and the CsPbBr3 nanocrystals serving as quantum dot “activators” for radiative recombination. However, one still must contend with whether the band structure of the Cs4 PbBr6 host supports sufficient carrier mobility for such collection to the activator islands. The calculations of Kang and Biswas confirm theoretically what was known about the “zero-dimensional” perovskite Cs4 PbBr6 . Figure 18.30 confirms that its bands are very flat, and furthermore Kang and Biswas found that both carriers are self-trapped. They reported that the signature of a localized electron-polaron level becomes evident from a singly occupied state appearing below the host CBM. Figure 18.32 shows the atomic structure and charge density plot of the electron polaron localized around a Pb (6p-orbital) in a 132-atom supercell. In order to accommodate the localized electron, the lattice responds by significantly extending the six Pb—Br bonds from its optimized value of 3.04 to 3.29 Å. Similar to the electron polarons, a valence hole is also subject to self-trapping in Cs4 PbBr6 when it “floats up” to a localized state above the VBM of the host. The localized hole of Cs4 PbBr6 shown in Figure 18.32 is composed of antibonding Pb 6s* and Br 4p (𝜎*). All six Pb—Br bonds in the [PbBr6 ] octahedron decrease in length from 3.04 Å in the perfect host to about 2.88 Å. Kang and Biswas showed that the electron and hole polarons may create bound excitonic structures. The self-trapped exciton comprising a bound state of these lattice-coupled carriers are likely responsible for the observed ultraviolet emission around ∼375 nm reported in bulk Cs4 PbBr6 and Cs4 PbBr6 /CsPbBr3 nanocomposites. It appears that the electronic structure of CsPbBr3 nano-islands in the Cs4 PbBr6 host offers good news and bad news for prospects of effective scintillation. There are two of each. The two pieces of good news are that (1) quantum confinement makes the emission of light fast and efficient (mostly first-order) once the electron–hole pairs enter the nano-islands, and (2) Type I band offsets in the heterostructure of the island-in-host

581

582

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

(a)

(b)

Figure 18.32 Calculated atomic structure and charge density isosurface of (a) electron and (b) hole polaron localized in a single octahedron of a 132-atom supercell. Outward and inward relaxation of the Pb—Br bonds are indicated by arrows. Source: From Kang and Biswas [177].

are expected and would have the very beneficial effect of collecting and concentrating electrons and holes in the islands if there is sufficient mobility in the host to transport them to the island boundaries. The first piece of offsetting bad news is that theory (see above) and experiment (see below) so far present a strong case for the fact that there is not sufficient mobility in the host for good scintillation. The second piece of bad news is tied to good news #1. Quantum confinement confers fast and efficient radiative recombination partly because it enhances oscillator strength. But oscillator strength works for upward as well as downward electronic transitions. Quantum confinement makes the nano-islands very good absorbers [121]. Unlike light emission from LED active films only a hundred nanometers or so thick, the light from a scintillator must make its way outward from the gamma stopping depth, typically centimeters. Because the exciton light emitted from CsPbBr3 islands has a very small Stokes shift equal to the exciton binding energy of a few tens of millivolts, it must contend with reabsorption by the quantum dots on its way out of the scintillator. We have surmised that this must be a serious problem with a Cs4 PbBr6 /CsPbBr3 nanocomposite used as a scintillator because materials with very high PLQY are observed to display very low light yield when excited by gamma rays. Self-absorption has been encountered in other scintillators and has sometimes been overcome by strategies like adding a wavelength shifter. We have tried some of those measures with Cs4 PbBr6 /CsPbBr3 nanocomposite micro-grains potted in a dye medium, but without significant success yet. It is a tantalizing quandary to have two such promising pieces of technical good news counterbalanced or perhaps even nullified by two other pieces of technical bad news. It remains tantalizing in this particular system, and the search should surely go on.

18.5 Conclusions Perovskites as a class of materials have been known a long time, are reasonably well characterized, and have been considered generally understood. But the recent introduction of organic–inorganic perovskite hybrids and the realization of record photovoltaic

References

efficiencies from solution-processed perovskites have led to a re-evaluation of such materials in “high-tech” applications. Indeed, the broad range of accessible properties found in 2D and 3D classes of perovskites have pointed to significant potential in electronic, optical, and energy-related technologies. The synergisms and tunability that make these new perovskites so interesting can be seen as generally emergent from their chemical and structural flexibility. Photovoltaics based on hybrid lead halide perovskites benefit from long carrier lifetimes attributed partly to phenomena lumped as defect tolerance, and partly to weak exciton binding energy in the 3D perovskites. LEDs constructed with the lead halide perovskites benefit from very high photo- and electroluminescent quantum yield, in some cases attributable to quantum confinement in nanoparticle films or nanocomposites. It also appears fortuitous that LEDs are thin film devices, so that self-absorption of the typically weakly Stokes-shifted exciton light is not a significant problem. Lead halide perovskite-based detectors of high-energy radiation can benefit from the high density and effective atomic number of these compounds along with the aforementioned long carrier lifetime for semiconducting gamma detection, and high PLQY along with very fast radiative lifetime, especially in materials with quantum confinement. Semiconducting radiation detectors from, e.g. melt-grown CsPbBr3 seem particularly promising, but energy-resolving scintillation detectors seem hampered at the moment by poor full-energy stopping in the available crystal sizes, and more fundamentally by problems in extracting weakly Stokes-shifted light in materials that may have seemed very promising on the basis of the PLQY from films or powders. There are opportunities for further research and development.

Acknowledgments JX, SG, and RTW acknowledge support of Department of Homeland Security – CWMD-Academic Research Initiative grant DHS-2014-DN-077-ARI077-05 and DOE National Nuclear Security Administration subcontract through Lawrence Berkeley National Laboratory LB15-ML-GammaDetMater-PD2Jf for work on ionizing radiation detectors. KB acknowledges DHS-2014-DN-077-ARI075. DLC acknowledges NSF Grant 1610641. We thank Arnold Burger and Liviu Matei of Fisk University for gamma pulse height measurements on cesium lead halide nanocomposites.

References 1 2 3 4 5 6 7

Goldschmidt, V.M. (1926). Naturwissenschaften 21: 477–485. Megaw, H. (1945). Nature 155 (3938): 484–485. Mitzi, D.B. (1999). Prog. Inorg. Chem. 48: 1–121. Mitzi, D.B., Chondroudis, K., and Kagan, C.R. (2001). IBM J. Res. Dev. 45: 29–45. Cheng, Z. and Lin, J. (2010). CrystEngComm 12: 2646–2662. Mercier, N., Louvain, N., and Bi, W. (2009). CrystEngComm 11: 720–734. S. Roth and D. Carroll (2015). One-Dimensional Metals: Conjugated Polymers, Organic Crystals, Carbon Nanotubes and Graphene 3rd Edition (Chapter 3), Wiley-VCH, NY. NY. ISBN-13: 978–3527335572.

583

584

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

8 Kieslich, G., Sun, S., and Cheetham, A.K. (2014). Chem. Sci. 5: 4712–4715. 9 Shannon, R. (1976). Acta Crystallogr., Sect. A: Cryst. Phys., Diffr., Theor. Gen. Crys10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43

tallogr. 32: 751–767. Kieslich, G., Sun, S., and Cheetham, A.K. (2015). Chem. Sci. 6: 3430–3433. Saparov, B. and Mitzi, D.B. (2016). Chem. Rev. 116: 4558–4596. Mitzi, D.B. and Chem, J. (2001). Soc., Dalton Trans.: 1–12. Mitzi, D.B., Feild, C.A., Harrison, W.T.A., and Guloy, A.M. (1994). Nature 369: 467–469. Needham, G.G., Willett, R.D., and Franzen, H.F. (1984). J. Phys. Chem. 88: 674–680. Barman, S., Venkataraman, N.V., Vasudevan, S., and Seshadri, R. (2003). J. Phys. Chem. B 107: 1875–1883. Naik, V.V. and Vasudevan, S. (2010). J. Phys. Chem. C 114: 4536–4543. Tang, Z., Guan, J., and Guloy, A.M. (2001). J. Mater. Chem. 11: 479–482. Weber, O.J., Marshall, K.L., Dyson, L.M., and Weller, M.T. (2015). Acta Crystallogr., Sect. B: Struct. Sci. Cryst. Eng. Mater. 71: 668–678. Knutson, J.L., Martin, J.D., and Mitzi, D.B. (2005). Inorg. Chem. 44: 4699–4705. Li, Y., Lin, C., Zheng, G., and Lin, J. (2007). J. Solid State Chem. 180: 173–179. Xu, Z., Mitzi, D.B., and Medeiros, D.R. (2003). Inorg. Chem. 42: 1400–1402. Mitzi, D.B., Dimitrakopoulos, C.D., and Kosbar, L.L. (2001). Chem. Mater. 13: 3728–3740. Takahashi, Y., Obara, R., Nakagawa, K. et al. (2007). Chem. Mater. 19: 6312–6316. Chung, I., Song, J.-H., Im, J. et al. (2012). J. Am. Chem. Soc. 134: 8579–8587. Stoumpos, C.C., Malliakas, C.D., and Kanatzidis, M.G. (2013). Inorg. Chem. 52: 9019–9038. Sourisseau, S., Louvain, N., Bi, W. et al. (2007). Chem. Mater. 19: 600–607. Mercier, N., Poiroux, S., Riou, A., and Batail, P. (2004). Inorg. Chem. 43: 8361–8366. Sourisseau, S., Louvain, N., Bi, W. et al. (2007). Inorg. Chem. 46: 6148–6154. Pearson, R.G. (1983). J. Mol. Struct. THEOCHEM 103: 25–34. Lemmerer, A. and Billing, D.G. (2010). CrystEngComm 12: 1290–1301. Castro-Castro, L.M. and Guloy, A.M. (2003). Angew. Chem. Int. Ed. 42: 2771–2774. Svensson, P.H., Rosdahl, J., and Kloo, L. (1999). Chem. Eur. J. 5: 305–311. Riggs, S.C., Shapiro, M.C., Corredor, F. et al. (2012). J. Cryst. Growth 355: 13–16. Daub, M. and Hillebrecht, H. (2015). Angew. Chem. Int. Ed. 54: 11016–11017. Halder, A., Chulliyil, R., Subbiah, A.S. et al. (2015). Chem. Lett. 6: 3483–3489. Van Aken, B.B., Palstra, T.T.M., Filippetti, A., and Spaldin, N.A. (2004). Nat. Mater. 3: 164–170. Seo, D.K., Gupta, N., Whangbo, M.H. et al. (1998). Inorg. Chem. 37: 407–410. Swainson, L., Chi, L., Her, J.-H. et al. (2010). Acta Crystallogr., Sect. B: Struct. Sci. 66: 422–429. Donaldson, J.D., Silver, J., Hadjiminolis, S., and Ross, S.D. (2010). J. Chem. Soc., Dalton Trans. 1975: 1500–1506. Halasyamani, P.S. (2010). Chem. Mater. 2004 (16): 3586–3592. Alonso, J.A., Martínez-Lope, M.J., Casais, M.T., and Fernández-Díaz, M.T. (2000). Inorg. Chem. 39: 917–923. Mizokawa, T. (2013). J. Phys. Conf. Ser. 428 : 012020. Lufaso, M.W. and Woodward, P.M. (2004). Acta Crystallogr. Sect. B: Struct. Sci. 60: 10–20.

References

44 Paul, A.K., Reehuis, M., Ksenofontov, V. et al. (2013). Phys. Rev. Lett. 111 : 167205. 45 Yan, B., Paul, A.K., Kanungo, S. et al. (2014). Phys. Rev. Lett. 112 (147202). 46 Shanmuga Priya, S., Rao, A., Thirunavukkarasu, I., and Nayak, V. (2008). Prog. Pho47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85

tovolt. Res. Appl. 16: 235. Wang, W., Winkler, M.T., Gunawan, O. et al. (2014). Adv. Energy Mater. 4: 1. Che, X., Li, Y., Qu, Y., and Forrest, S.R. (2018). Nat. Energy 3: 422. Yang, W.S., Park, B.W., Jung, E.H. et al. (2017). Science 356: 1376. Akihiro Kojima, T.M., Teshima, K., and Shirai, Y. (2009). J. Am. Chem. Soc. 131: 6050. De Wolf, S., Holovsky, J., Moon, S.J. et al. (2014). J. Phys. Chem. Lett. 5: 1035. D’Innocenzo, V., Grancini, G., Alcocer, M.J.P. et al. (2014). Nat. Commun. 5: 3586. Manser, J.S. and Kamat, P.V. (2014). Nat. Photonics 8: 737. Nie, W., Tsai, H., Asadpour, R. et al. (2015). Science 347: 522. Stranks, S.D., Stranks, S.D., Eperon, G.E. et al. (2013). Science 342: 341. Xing, G., Mathews, N., Sun, S. et al. (2013). Science 342: 344. Burschka, J., Pellet, N., Moon, S.J. et al. (2013). Nature 499: 316. Xiao, M., Huang, F., Huang, W. et al. (2014). Angew. Chem. Int. Ed. 53: 9898. Jung, H.S. and Park, N.G. (2015). Small 11: 10. Jeon, N.J., Noh, J.H., Kim, Y.C. et al. (2014). Nat. Mater. 13: 897. Zhou, H., Chen, Q., Li, G. et al. (2014). Science 345: 542. Liu, M., Johnston, M.B., and Snaith, H.J. (2013). Nature 501: 395. Petrus, M.L., Schlipf, J., Li, C. et al. (2018). Adv. Energy Mater. 8: 1700264. Ishihara, T. (1994). Optical properties of PbI-based perovskite structure. J. Lumin 269: 60–61. Mitzi, D.B. (1996). Chem. Mater. 8: 791. Mercier, N. and Riou, A. (2004). Chem. Commun. 33: 844. Tsai, H., Nie, W., Blancon, J.C. et al. (2016). Nature 536: 312. Schmidt, L.C., Pertega, A., Gonza, S. et al. (2014). J. Am. Chem. Soc. 136: 850. Martiradonna, L. (2018). Nat. Mater. 17: 377. Huang, H., Bodnarchuk, M.I., Kershaw, S.V. et al. (2017). ACS Energy Lett. 2: 2071. Walsh, A. and Zunger, A. (2017). Nat. Mater. 16: 964. Yin, W.-J., Shi, T., and Yan, Y. (2014). Appl. Phys. Lett. 104: 063903. Kim, J., Lee, S.-H., Lee, J.H., and Hong, K.-H. (2014). J. Phys. Chem. Lett. 5: 1312. Walsh, A., Scanlon, D.O., Chen, S. et al. (2015). Angew. Chem. Int. Ed. 54: 1791. Du, M.-H. (2015). J. Phys. Chem. Lett. 6: 1461. Zhang, S.B., Wei, S.-H., and Zunger, A. (1998). Phys. Rev. B 57: 9642. Zakutayev, A., Caskey, C.M., Fioretti, A.N. et al. (2014). J. Phys. Chem. Lett. 5: 1117. Brandt, R.E., Stevanovi´c, V., Ginley, D.S., and Buonassisi, T. (2015). MRS Commun. 5: 265. Buin, A., Pietsch, P., Xu, J. et al. (2014). Nano Lett. 14: 6281. Oga, H., Saeki, A., Ogomi, Y. et al. (2014). J. Am. Chem. Soc. 136: 13818. Du, M.H. (2014). J. Mater. Chem. A 2: 9091. Juarez-Perez, E.J., Sanchez, R.S., Badia, L. et al. (2014). J. Phys. Chem. Lett. 5: 2390. Kang, B. and Biswas, K. (2017). Phys. Chem. Chem. Phys. 19: 27184. Brandt, R.E., Poindexter, J.R., Gorai, P. et al. (2017). Chem. Mater. 29: 4667. Era, M., Morimoto, S., Tsutsui, T., and Saito, S. (1994). Appl. Phys. Lett. 65: 676.

585

586

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128

Chondroudis, K. and Mitzi, D.B. (1999). Chem. Mater. 11: 3028. Kojima, A., Ikegami, M., Teshima, K., and Miyasaka, T. (2012). Chem. Lett. 41: 397. Lin, K., Xing, J., Quan, L.N. et al. (2018). Nature 562: 245. Kim, Y.-H., Cho, H., and Lee, T.-W. (2016). Proc. Natl. Acad. Sci. USA 113: 11694. Song, J., Li, J., Li, X. et al. (2015). Adv. Mater. 27: 7162. Protesescu, L., Yakunin, S., Bodnarchuk, M.I. et al. (2015). Nano Lett. 15: 3692. Sichert, J.A., Tong, Y., Mutz, N. et al. (2015). Nano Lett. 15: 6521. Huang, H., Susha, A.S., Kershaw, S.V. et al. (2015). Adv. Sci. 2: 1500194. Tyagi, P., Arveson, S.M., and Tisdale, W.A. (2015). J. Phys. Chem. Lett. 6: 1911. Dou, L., Wong, A.B., Yu, Y. et al. (2015). Science 349: 1518. Bekenstein, Y., Koscher, B., Eaton, S.W. et al. (2015). J. Am. Chem. Soc. 137: 16008. Wu, X., Trinh, M.T., Niesner, D. et al. (2015). J. Am. Chem. Soc. 137: 2089. Hong, X., Ishihara, T., and Nurmikko, A.V. (1992). Phys. Rev. B 45: 6961. Yuan, Z., Shu, Y., Xin, Y., and Ma, B. (2016). Chem. Commun. 52: 16008. Xiao, Z., Kerner, R.A., Zhao, L. et al. (2017). Nat. Photonics 11: 108. Zhao, L., Yeh, Y.W., Tran, N.L. et al. (2017). ACS Nano 11: 3957. Byun, J., Cho, H., Wolf, C. et al. (2016). Adv. Mater. 28: 7515. Quan, L.N., Zhao, Y., Garc, F.P. et al. (2017). Nano Lett. 17: 3701. Yuan, M., Quan, L.N., Comin, R. et al. (2016). Nat. Nanotechnol. 11: 872. Xiao, Z., Zhao, L., Tran, N.L. et al. (2017). Nano Lett. 17: 6863. La-Placa, M.G., Longo, G., Babaei, A. et al. (2017). Chem. Commun. 53: 8707. Wang, Q., Ren, J., Peng, X.F. et al. (2017). ACS Appl. Mater. Interfaces 9: 29901. Lee, S., Park, J.H., Nam, Y.S. et al. (2018). ACS Nano 12: 3417. Zhang, S., Yi, C., Wang, N. et al. (2017). Adv. Mater. 29: 1606600. Wang, N., Cheng, L., Ge, R. et al. (2016). Nat. Photonics 10: 699. Chen, Z., Zhang, C., Jiang, X.F. et al. (2017). Adv. Mater. 29: 1603157. Yang, X., Zhang, X., Deng, J. et al. (2018). Nat. Commun. 9: 570. Wu, C., Zou, Y., Wu, T. et al. (2017). Adv. Funct. Mater. 27: 1700338. Ji, X., Peng, X., Lei, Y. et al. (2017). Org. Electron. Physics, Mater. Appl. 43: 167. Bade, S.G.R., Shan, X., Hoang, P.T. et al. (2017). Adv. Mater. 29 : 1607053. Ling, Y., Tian, Y., Wang, X. et al. (2016). Adv. Mater. 28: 8983. Bade, S.G.R., Li, J., Shan, X. et al. (2016). ACS Nano 10: 1795. Song, L., Guo, X., Hu, Y. et al. (2017). J. Phys. Chem. Lett. 8: 4148. Yantara, N., Bhaumik, S., Yan, F. et al. (2015). J. Phys. Chem. Lett. 6: 4360. Cho, H., Wolf, C., Kim, J.S. et al. (2017). Adv. Mater. 29: 1700579. Xu, J., Huang, W., Li, P. et al. (2017). Adv. Mater. 29: 1703703. Saidaminov, M.I., Almutlaq, J., Sarmah, S.P. et al. (2016). ACS Energy Lett. 1 (4): 840–845. Makarov, N.S., Guo, S., Isaienko, O. et al. (2016). Nano Lett. 16: 2349–2362. Pietryga, J.M., Park, Y.S., Lim, J. et al. (2016). Chem. Rev. 116: 10513–10622. Dennis, A.M., Mangum, B.D., Piryatinski, A. et al. (2012). Nano Lett. 12: 5545–5551. Hiroshige, N., Ihara, T., Saruyama, M. et al. (2017). J. Phys. Chem. Lett. 8: 1961–1966. Lim, J., Jeong, B.G., Park, M. et al. (2014). Adv. Mater. 26: 8034–8040. Stranks, S.D., Burlakov, V.M., Leijtens, T. et al. (2014). Phys. Rev. Appl. 2: 034007.

References

129 Quan, L.N., Quintero-Bermudez, R., Voznyy, O. et al. (2017). Highly emissive green

perovskite nanocrystals in a solid state crystalline matrix. Adv. Mater. 29: 1605945. 130 Kondo, S., Amaya, K., and Saito, T. (2002). Localized optical absorption in 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167

Cs4 PbBr6 . J. Phys. Condens. Matter 14: 2093–2099. Xing, G., Wu, B., Wu, X. et al. (2017). Nat. Commun. 8: 14558. Gridin, S., Onken, D.R., Williams, R.T. et al. (2018). J. Appl. Phys. 124: 154504. Jones, T. and Townsend, D. (2017). J. Med. Imaging 4: 011013. Lecoq, P. (2017). Trans. Rad. Plasma Med. Sci. 1: 473. Lecoq, P. (2012). IEEE Trans. Nucl. Science 59: 2313. Derenzo, S.E., Choong, W.-S., and Moses, W.W. (2014). Phys. Med. Biol. 59: 32613286. Knoll, G.F. (2010). Radiation Detection and Measurement, 4e. Wiley. NIST (2010). Standard Reference Database 8 (XCOM Photon Cross Sections) [http://www.nist.gov/pml/xcom-photon-cross-sections-database]. NIST (2004). Standard Reference Database 126 (X-Ray Mass Attenuation Coefficients) [http://www.nist.gov/pml/x-ray-mass-attenuation-coefficients]. Bizarri, G., Moses, W.W., Singh, J. et al. (2009). J. Appl. Phys. 105: 044507. Dorenbos, P. (2010). IEEE Trans. Nucl. Sci. 57: 1162–1167. Alig, R.C. and Bloom, S. (1978). J. Appl. Phys. 49: 3476–3480. Rodnyi, P.A. (1997). Physical Processes in Inorganic Scintillators. Boca Raton: CRC Press. Xinfu, L., Qi, L., Bizarri, G.A. et al. (2015). Phys. Rev. B 92: 115207. Lu, X., Gridin, S., Williams, R.T. et al. (2017). Phys. Rev. Appl. 7 : 014007. (a) Agostinelli, S. et al. (2003). Nucl. Instrum. Methods A 506: 250. (b) Allison, J. et al. (2006). IEEE Trans. Nucl. Sci. 53: 270. He, Y., Matei, L., Jung, H.J. et al. (2018). Nat. Commun. 9: 1609. Zhang, F., He, Z., and Seiffert, C.E. (2007). IEEE Trans. Nucl. Sci. 54: 843–848. Kang, J. and Wang, L.–.W. (2017). J. Phys. Chem. Lett. 8: 489–493. Dirin, D.N., Cherniukh, I., Yakunin, S. et al. (2016). Chem. Mater. 28: 8470–8474. Yakunin, S., Dirin, D.N., Shynkarenko, Y. et al. (2016). Nat. Photonics 10: 585–589. He, Y., Ke, W., Alexander, G.C.B. et al. (2018). ACS Photonics 5: 4132–4138. Xu, Q., Wei, H., Wei, W. et al. (2017). Nucl. Inst. and Meth. In Phys. Res. A 848: 106–108. Wei, H., DeSantis, D., Wei, W. et al. (2017). Nat. Mater. 16: 826–833. Wei, H. et al. (2016). Nat. Photon. 10: 333–339. Nazarenko, O., Yakunin, S., Morad, V. et al. (2017). NPG Asia Mater. 9 (373). Cho, H.Y., Lee, J.H., Kwon, Y.K. et al. (2011). JINST 6: C01025. Yakunin, S., Sytnyk, M., Kriegner, D. et al. (2015). Nat. Photonics 9: 444–449. Kim, Y.C., Kim, K.H., Jeong, D.N. et al. (2017). Nature 550: 87. Wang, X., Wu, Y., Li, G. et al. (2018). Adv. Electron. Mater. 4: 1800237. Shrestha, S., Fischer, R., Matt, G.J. et al. (2017). Nat. Photonics 11: 436–440. Pan, W. et al. (2017). Nat. Photon. 11: 726–732. Zhuge, F., Luo, P., and Zhai, T. (2017). Sci. Bull. 62: 1491–1493. Nitsch, K., Hamplova, V., Nikl, M. et al. (1996). Chem. Phys. Lett. 258: 518. Kondo, S., Kakuchi, M., Masaki, A., and Saito, T. (2003). J. Phys. Soc. Jpn. 72: 1789. Kondo, S., Nakagawa, H., Saito, T., and Asada, H. (2004). Curr. Appl. Phys. 4: 439. Nikl, M., Nitsch, K., Mihodova, E. et al. (1999). Radiat. Eff. Defects Solids 150: 341.

587

588

18 Perovskites – Revisiting the Venerable ABX3 Family with Organic Flexibility and New Applications

168 169 170 171 172 173 174 175 176 177 178

Nikl, M., Mihokova, E., Nitsch, K. et al. (1999). Chem. Phys. Lett. 306: 280. Chen, Q., Wu, J., Ou, X. et al. (2018). Nature 561: 88–93. Heo, J.H., Shin, D.H., Park, J.K. et al. (2018). Adv. Mater. 30: 1801743. Birowosuto, M.D., Cortecchia, D., Drozdowski, W. et al. (2016). Sci. Rep. 6: 37254. Kawano, N., Koshimizu, M., Okada, G. et al. (2017). Sci. Rep. 7: 14754. Saint-Gobain www.crystals.saint-gobain.com/products/nai. Alekhin, M.S., de Haas, J.T.M., Khodyuk, I.V. et al. (2013). Appl. Phys. Lett. 102: 161915. Yettapu, G.R., Talukdar, D., Sarkar, S. et al. (2016). Nano Lett. 16: 4838. Li, J., Yuan, X., Jing, P. et al. (2016). RSC Adv. 6: 78311. Kang, B. and Biswas, K. (2018). J. Phys. Chem. Lett. 9: 830. Stoumpos, C.C., Malliakas, C.D., Peters, J.A. et al. (2013). Cryst. Growth Det. 13: 2722–2727.

589

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures Akihiro Murayama 1 and Yasuo Oka 2 1 Graduate School of Information Science and Technology, Hokkaido University, Kita 14, Nishi 9, Kita-ku, Sapporo 060-0814, Japan 2 Institute of Multidisciplinary Research for Advanced Materials, Tohoku University, Katahira 2-1-1, Aoba-ku, Sendai 980-8577, Japan

CHAPTER MENU Introduction, 589 Quantum Wells, 591 Fabrication of Nanostructures by Electron-Beam Lithography, 596 Self-Assembled Quantum Dots, 599 Hybrid Nanostructures with Ferromagnetic Materials, 604 Conclusions, 607 References, 609

19.1 Introduction Diluted magnetic semiconductors (DMSs) are compound semiconductors of magnetic ions such as Mn, Fe, and Co. They show remarkable spin-dependent properties owing to the s and p-d exchange interactions of band-electrons (-holes) with the magnetic ions [1, 2]. These characteristic properties are very attractive for applications to magneto-optical devices. Much research interest has recently been focused on both III–V and II–VI DMS materials, the aim being to develop spin-related optical and electronic devices [3]. A significant optical property of DMSs is the giant Zeeman splitting of excitonic states, by which the energy levels of the spin states of the electron and hole in an exciton produce very large splittings in the presence of magnetic fields. The spin-splitting energy of an exciton is up to 100 meV. Therefore, the exciton relaxes rapidly to the lowest energy state, which has a specific spin direction, and the excess energy is given to phonons. As a result, the exciton spins can be fully polarized in a DMS, which is a potential advantage in applying these materials for semiconductor devices, where the direction of the electron spin can be controlled by the direction of the applied external magnetic field. Another important property of DMSs is the giant Faraday rotation, especially for optical applications at room temperature. For example, an optical isolator using a Cd1-x Mnx Te-based crystal has already been developed and employed in optical communication [4]. In addition to these two properties, Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

590

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

several fundamental phenomena relating to the spin dynamics of excitons are still under investigation. One of the very interesting properties is excitonic magnetic polaron formation, where the exciton interacts with many local magnetic moments of the doped magnetic ions through exchange interactions [5–7]. The exciton spin is significantly affected by the dynamical behavior of the local spins. This cooperative phenomenon of many spins is one of the most interesting properties of spin dynamics in semiconductors. However, one still requires a basic understanding of the quantum confinement effects, the exchange interactions between excitons and magnetic ions, and the exciton spin dynamics in DMS nanostructures. Here, we describe our studies on spin-dependent optical processes and the dynamics in several types of nanostructures based on II–VI DMS [8, 9]. These nanostructures show many attractive spin-related functionalities such as spin injection and ultrafast spin polarization. Moreover, the coupled nanostructures of the DMS, including self-assembled quantum dots (QDs), provide opportunities for spin manipulation of carriers based on quantum tunneling. Therefore, the coupled DMS nanostructures are worth investigating for future developments of spintronic devices and quantum computing. Hybrid nanostructures consisting of DMSs and ferromagnetic materials can be used in a new technology, since the small dimensions of the ferromagnetic materials can generate very strong local magnetic fields, which can align the exciton spins in the DMS nanostructures. Spin injection and transport of excitons or carriers have been reported, where a DMS is used as the source of spin-polarized excitons or carriers [10–13]. The details of the spin-injection process, however, have not been completely explored. Therefore, double quantum well (DQW) systems with proper tunneling barriers have been investigated by transient circularly polarized photoluminescence (PL) spectroscopy [14–17] to understand the dynamical processes of spin injection and relaxation. The transient optical phenomena of electron-hole (e-h) plasma, excitons, and magnetic ions in DMS-QWs involve basic physics related to the dynamics of their spin states [18–20]. To clarify these ultrafast spin dynamics, transient pump-probe spectroscopy has been performed under magnetic fields. These measurements reveal the rich physics of the dynamics of the creation and relaxation of e-h plasma, excitons, and the subsequent spin alignment of the excitons [21–23]. Nanoscale dot and wire structures of DMS are expected to show efficient magnetooptical properties owing to their low-dimensional confinement effects and high quantum yields of exciton luminescence [24–28]. We have designed and fabricated such DMS nanostructures by electron-beam lithography [29–32]. Exciton dynamics and the giant magneto-optical properties are studied in DMS nanostructures having lateral dimensions ranging from 20 to 100 nm. Further, in a coupled QD system composed of the DMS and nonmagnetic QDs, interdot injection of the spin-polarized excitons is realized. A new type of coupled system with self-assembled QDs that show strong quantum confinement effects is also examined. Hybrid nanostructures composed of DMS and ferromagnetic metals exhibit various properties that are applicable to spin-related electronics [33]. We have fabricated a nanoscale hybrid structure of DMS wires sandwiched between Co wires [34–36] and DMS nanodisks embedded in Co/Pt multilayered thin films [37]. The magneto-optical properties are affected by the local magnetic field generated by the neighboring Co wires. This magnetic field induces giant Zeeman effects on excitons induced by the Mn-spin alignment.

19.2 Quantum Wells

In this chapter, in Section 19.2, the optical properties of DMS quantum wells are described, whereas in Section 19.3, the nanostructures fabricated by electron-beam lithography are presented. In Section 19.4, we have covered the properties of selfassembled QDs and, finally, Section 19.5 presents the hybrid nanostructures involving ferromagnetic materials.

19.2 Quantum Wells 19.2.1

Spin Injection

The spin injection of excitons or carriers has been studied in a coupled DQW system with a magnetic quantum well (MW) and a nonmagnetic well (NW) [13]. However, the dynamical processes in spin injection have not been sufficiently elucidated, although knowledge of such detailed dynamics is crucial for the development of ultrafast spin-related devices. For instance, the spin-injection efficiency is determined by the injection rate of spin-polarized carriers from the MW to the NW relative to the spin relaxation rate in both the wells. The spin-injection efficiency decreases if carrier injection into the NW occurs faster than the spin polarization in the MW. The insertion of a proper tunneling barrier between the MW and the NW can increase the spin-injection efficiency, since the rate of carrier injection into the NW can be controlled by suitably designing the thickness and/or the potential height of the barrier. Therefore, the dynamics of spin injection has been studied in magnetic DQWs for various thicknesses of the tunneling barrier [15–17]. The energy diagram of the DQW studied is shown in Figure 19.1, which consists of an NW of Zn0.76 Cd0.24 Se and an MW of Zn0.96 Mn0.04 Se/CdSe. The NW and MW are separated by a ZnSe barrier. The well widths of NW and MW were fixed at 7 and 40 nm, respectively. The barrier width (LB ) was varied from 4 to 8 nm to study the variation in

ZnSSe

Zn0.73Cd0.27Se (NW) ZnSe

Zn0.96Mn0.04Se/CdSe (MW) ZnSSe ZnSe

Photoexcitation 2.77 eV

2.87 eV 2.80 eV

2.53 eV

7 nm LB = 4, 8 nm

2.72 eV

40 nm

Figure 19.1 Energy diagram of the magnetic DQW with a magnetic well (MW) of Zn0.96 Mn0.04 Se/CdSe and a nonmagnetic well (NW) of Zn0.73 Cd0.27 Se.

591

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

the spin-injection dynamics that depended on the carrier tunneling rate. The transient and circularly polarized PL from the DQW was measured at 2 K using a spectrometer and a streak camera. Magnetic fields of up to 6 T were applied in the Faraday configuration. The PL spectrum of the DQW (LB = 8 nm) is shown in Figure 19.2. The DQW was excited by a linearly polarized light of energy 2.770 eV. The exciton emissions from the NW and MW appear at 2.530 and 2.720 eV, respectively, at B = 0 T. The exciton PL in the MW is very weak, although an MW of width 40 nm is wide enough to give rise to a PL intensity. This suggests that tunneling of excitons or carriers occurs from the MW to the NW in the present DQW. By increasing the magnetic field to 6 T, the PL peak of the MW exciton with left-circular (𝜎 + ) polarization at 2.677 eV shows a shift toward a lower energy (red shift) of 43 meV, which is the typical giant Zeeman shift. Here, the 𝜎 + polarization is defined as the exciton recombination for the down-spin state with a lower energy (magnetic quantum number: mj = −1/2 and + 3/2 for electrons and heavy holes [hh], respectively). In the NW, the exciton PL peak energy remains unchanged over the whole range of magnetic fields considered. For optical excitation under magnetic fields, the carriers generated in the MW occupy mostly the down-spin states, with the spin alignment being parallel to the direction of magnetic field. This is due to the fast spin relaxation; the exciton spin is therefore highly polarized. These polarized down-spin excitons or carriers then tunnel into the NW. As a result, circularly polarized exciton PL is observed from the NW if the spin relaxation is not significant in the NW. As can be seen in Figure 19.2, the circular polarization property appears in the NW with a polarization degree P of 0.20 at the magnetic field B = 3 T (shown as a thick dotted line). Here, the circular polarization degree is defined as P = (I𝜎+ − I𝜎− )/(I𝜎+ + I𝜎− ), where I𝜎+ and I𝜎− are the circularly polarized PL intensities with 𝜎 + and 𝜎 − components. The observation of P from the NW is experimental evidence for spin injection through the tunneling process. For studying the spin-injection dynamics, we have measured the time-resolved PL of the DQW by exciting it with a linearly polarized light pulse of energy 2.774 eV at B = 3 T, as 1.2

σ+

0.8

0.8 0.6

σ–

0.6

MW 0.4

0.4

σ

+

0.2 0.2 x 50 0.0 2.45

2.50

2.55

2.60

2.65

2.70

2.75

Circular Polarization Degree

1.0

NW 1.0 Intensity (a.u.)

592

0.0 2.80

Photon Energy (eV)

Figure 19.2 Excitonic PL spectra with circular polarizations (𝜎 + : solid lines; 𝜎 − : a broken line) for the magnetic DQW (LB = 8 nm) at an applied magnetic field of 3 T. The circular polarization degree is also plotted for the exciton PL of the NW (a thick dotted line with an arrow). The excitation was achieved by using linearly polarized light.

19.2 Quantum Wells 0.8 0.6 0.4 0.2

0.1

MW σ+ NW σ– 0

0.0

–0.2 100 200 300 400 500 600 700 800

1

Polarization

NW σ+

0.6 0.4

NW σ–

0.2

0.1

0.0

MW σ+ 0

Circular Polarization Degree

Polarization NW σ+

0.8 Circular Polarization Degree Intensity (a.u.)

Intensity (a.u.)

1

–0.2 100 200 300 400 500 600 700 800

Time (ps)

Time (ps)

(a)

(b)

Figure 19.3 Circularly polarized PL intensities (𝜎 + : solid lines; 𝜎 − : a broken line) and the circular polarization degree (a dotted line) as a function of time, for DQWs with (a) LB = 4 nm and (b) 8 nm at 3 T.

shown in Figure 19.3. In the DQW sample with LB = 4 nm, shown in Figure 19.3a, the PL lifetime for the NW was 115 ps, and the P value was 0.20 just after the excitation. On the other hand, in the sample with LB = 8 nm, shown in Figure 19.3b, the PL lifetime for the NW was 138 ps, and P increased from 0.03 to 0.40 with a rise time of 400 ps. The increase in P of the NW with a time constant of 400 ps can be considered to be a result of the exciton tunneling from the MW to the NW with a long time constant. However, in the present case, the PL lifetime of excitons in the MW is much shorter (30 ps), and a model of slow exciton tunneling cannot be applied to these short-lifetime excitons. Therefore, we have proposed a model of individual tunneling of electrons and holes to quantitatively interpret the experimental results. Using this model, we have calculated the injection and recombination rates of the electrons and holes in DQWs with the aid of rate equations. The calculated results of the time developments of the MW and NW exciton PL intensities show good agreement with the experimental time evolution of P, where the electron-spin tunneling time is 30 ps, while the hh spin tunneling time is 1 ns. Therefore, when the PL lifetime of the MW is short and P builds up gradually in the NW due to slower hh tunneling, the individual charge carrier tunneling model reproduces the observed results quantitatively. Experimental evidence of carrier-spin tunneling has been reported in DMS DQWs, consisting of a DMS-QW (MW) of Zn0.77 Cd0.15 Mn0.08 Se and an NW of Zn0.82 Cd0.18 Se, by means of spin-resolved photoluminescence excitation (PLE) spectroscopy [38]. The spin-dependent PLE signal clearly shows efficient spin-conserving tunneling of excitons, which is sustained by the significant suppression of a spin-reversing process during the tunneling in magnetic fields. The spin-reversing tunneling is suppressed by two orders of magnitude in high magnetic fields. In this tunneling process, phonon emissions can induce energy relaxation just after the tunneling and therefore play an important role in the whole spin-transfer dynamics. The time constant of the exciton spin transfer, including the carrier tunneling and subsequent energy relaxation due to the phonon emissions, decreases significantly when the exciton energy difference between both wells is equal to the energy of one longitudinal optical (LO) phonon [39]. Therefore, this experimental observation indicates that the spin transfer can be accelerated with the conservation of spin polarization by the energy relaxation with a resonant LO-phonon emission immediately after the tunneling. This finding is important for developing ultrafast spin functional optical devices utilizing spin tunneling and transfer processes.

593

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

19.2.2

Study of Spin Dynamics by Pump-Probe Spectroscopy

The transient optical processes in DMS-QWs involve various interesting spin dynamics of electron-hole (e-h) plasma, excitons, and local spins of the magnetic ions [19, 20]. These phenomena are the basic physical processes for the potential of spin-related optical applications. To distinguish each spin dynamics that appears in an ultrashort time range, the technique of transient pump-probe absorption spectroscopy performed under magnetic fields has been developed [21–23]. The technique has been applied to study Cd0.95 Mn0.05 Te/Cd0.80 Mg0.20 Te multiple quantum wells (MQWs) grown by molecular beam epitaxy on ZnTe (100) substrates. Figure 19.4 shows the time evolution of the pump-probe absorption spectra in the magnetic MQW at a zero magnetic field, with the pump power density being 1.5 mJ cm−2 . Δ𝛼 d is the transient differential absorbance with and without pumping. An induced absorption (defined as a positive value of Δ𝛼 d), which rises and decays within 3 ps, is observed at 1.735 eV after the pump-pulse excitation. This induced absorption directly reflects the bandgap renormalization (BGR) due to the exchange correlation in the dense e-h plasma excited by the intense pump pulse. In the meantime, a bleached absorption (negative Δ𝛼 d) appears at the exciton energy position of 1.795 eV. This bleached transient absorption results from the elimination of screening for the Coulomb interactions in excitons due to their high-density effects, after the e-h plasma decays. The renormalized energy bandgap after the optical pumping can be extracted from the spectra of the transient absorption Δ𝛼 d calculated by taking into account the effects of the Coulomb screening, BGR, and carrier heating [18]. The rise and decay of this BGR energy in the absence of a magnetic field, i.e. at 0 T, at a temperature of 1.8 K can be fitted by a function with two exponential components. The time constants for the increase and decrease of this bandgap shift, ΔEg , are 0.2 and 1.0 ps, respectively. ΔEg BGR

Exciton

0 +0.05 1 2

0 Δad

Time (ps)

594

3 4 5

–0.08

6 1.70

1.75 1.80 Photon Energy ( eV )

1.85

Figure 19.4 Transient differential absorption spectra of a magnetic Cd0.95 Mn0.05 Te/Cd0.80 Mg0.20 Te MQW, measured by the pump-probe technique, with a zero magnetic field (0 T).

19.2 Quantum Wells

Ex (–1/2, +3/2)

1.0

Ex (+1/2, –3/2)

absorption PL

0.5

1 0

Time (ps)

0 50 100

Δαd –0.10

probe : σ–

Time (ps)

150 0

probe : σ+

0.15

Ex (–1/2, +3/2)

0.10 –Δαd

αd

1.5

PL Intensity (a.u.)

reaches its maximum value at 77 meV in about 0.5 ps after the pumping, which yields the maximum plasma density of 3 × 1012 cm−2 . In DMS materials, the strong s and p-d exchange interactions between the carriers and the localized spins of the magnetic ions result in giant Zeeman splitting. The BGR effect was clearly resolved for each spin-split Landau subband when the measurement was carried out under the application of magnetic fields. Following the decay of the dense e-h plasma, the bleached peak of the exciton absorption at 1.795 eV attains its maximum intensity 4 ps after the pumping. The maximum intensity indicates the maximum population of the excitons generated by the pumping. The excitonic bleached peaks measured at 1.8 K without any magnetic field show an evident red shift within the exciton lifetime. At a magnetic field of 3 T, the bleached peak splits into two branches of 𝜎 + and 𝜎 − polarizations, which is caused by the giant Zeeman splitting of the excitons. Subsequently, we discuss the time dependence of the bleached peaks under the action of magnetic fields. The transient differential absorption spectra are shown in Figure 19.5a in the longer time range at a magnetic field of 5 T for excitation by a 𝜎 − -circularly polarized pump pulse of energy 1.860 eV and a lower pump power density of 10 μJ cm−2 . Two intense bleached peaks were observed with an energy splitting of up to 80 meV corresponding to the spin-polarized hh-exciton levels due to the giant Zeeman effect. The field dependence of the energy splitting can be expressed by a Brillouin function. The intensity of the bleached peak at 1.820 eV corresponding to the up-spin exciton state decays faster than that of the down-spin state. This reflects the

0.05

Ex (+1/2, –3/2)

0.00 0

50

20

40 60 Time (ps)

80

100

(b)

100 0

150 1.70

1.75 1.80 1.85 Photon Energy (eV)

1.90

(a)

Figure 19.5 (a) Transient differential absorption spectra measured with the probes of 𝜎 + and 𝜎 − polarizations, for the Cd0.95 Mn0.05 Te MQW in an applied magnetic field of 5 T and excited by a 𝜎 − -polarized pump-pulse of 1.860 eV. The linear absorption spectrum and PL spectrum are also shown above the differential absorption spectra. (b) Differential absorbances as a function of time for the higher- (open circles) and lower-energy (closed circles) spin states of excitons in an applied magnetic field of 5 T. The solid lines are the best-fitted curves calculated using the rate equations.

595

596

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

spin relaxation occurring through the spin-flip processes observed between those levels that is accompanied by energy relaxation. However, a PL peak is observed only from the down-spin hh-exciton state. The disappearance of the up-spin PL spectrum can be attributed to the fast relaxation of the hh spin in the exciton state. The faster decay of the hh spin results in the formation of a dark exciton state. Therefore, the electron cannot recombine efficiently at the up-spin exciton branch. After the relaxation of the electron from the up- to the down-spin state, the electron and the hh can recombine radiatively to exhibit PL. The differential absorbances with 𝜎 + and 𝜎 − polarizations are plotted as functions of time in Figure 19.5b. The observed time dependences are fitted with those calculated using the rate equations. From the best-fitted results, we have successfully determined the spin relaxation time of an electron to be 29 ps at a magnetic field of 5 T. The saturated absorbance for the down-spin state increases in accordance with its decrease for the up-spin state. The decay in the saturated absorbance for the up-spin state directly reveals the dynamics of the spin-flip process of the electron. The spin relaxation time of the hh is found to be lower than 0.2 ps, which is our instrumental resolution limit. This fast decay of the hh spin may be attributed to the participation of multiple LO phonons in the relaxation process in addition to the band mixing of the hole states, because the calculated energy splitting of the hh state is found to be 64 meV at a magnetic field of 5 T and the LO-phonon energy is known to be 21 meV for CdTe. On the other hand, the relaxation time of the electron spin is obtained as 17 ps at a zero magnetic field, which is shorter than the 46 ps observed in the case of CdTe MQW. Such a short relaxation time can be attributed to the spin relaxation occurring via the s-d exchange mechanism with many Mn spins. At the applied magnetic field of 5 T, the spin relaxation time of 29 ps is found to be longer than the value of 17 ps obtained at 0 T. At such a high magnetic field, the Mn spins are strongly pinned down by the external field, which may eliminate the relaxation pathways of the s-d exchange mechanism. As the energy splitting of the electron spin state is calculated to be 16 meV at the applied magnetic field of 5 T, which is still lower than the LO-phonon energy of 21 meV, the LO-phonon cannot participate in the electron-spin relaxation process. Therefore, the electron spin relaxation caused by the exchange mechanism is experimentally elucidated as being suppressed by the magnetic fields.

19.3 Fabrication of Nanostructures by Electron-Beam Lithography The fabrication of intentionally designed nanostructures of DMS is important for developing spin electronic devices [24, 25]. We have fabricated nanostructures based on the QDs and wires of Zn1-x-y Cdx Mny Se [29–32]. Single QWs of Zn1-x-y Cdx Mny Se with thicknesses ranging from 5 to 10 nm were grown on GaAs (100) substrates by molecular beam epitaxy. DMS nanostructures with lateral dimensions of 20–100 nm were successfully fabricated from the QW by electron-beam lithography followed by chemical wet etching. Figure 19.6 shows (a) a schematic illustration of the dot structure fabricated from a DMS-QW of Zn1-x-y Cdx Mny Se by using electron-beam lithography, (b) scanning electron microscope (SEM) image of the dot array, (c) SEM image of the dots with an average diameter of 20 nm, and (d) SEM image of dots of diameter 80 nm.

19.3 Fabrication of Nanostructures by Electron-Beam Lithography

(b) Zn1–x–yCdxMnySe QW ZnSe cap ZnSe buffer

500 nm (a)

(c)

GaAs (100)

100 nm

100 nm

(d)

Figure 19.6 (a) Schematic drawing of the dot structure fabricated from a DMS-QW of Zn1-x-y Cdx Mny Se by using electron-beam lithography. (b) SEM image of the dot array. (c) SEM image of the dots with an average diameter of 20 nm and (d) 80 nm.

ZnSe barrier

PL Intensity (a.u.)

Figure 19.7 Excitonic PL spectra from a DMS-dot of Zn0.80 Cd0.15 Mn0.05 Se with a lateral diameter of 30 nm (a solid line) and from the QW before dot fabrication (dotted line). The PL peaks due to the ZnSe barrier and the defect-related ZnSe y-line are also shown.

QW

QD (x10)

ZnSe Y-line

2.55

2.60

2.65

2.70 2.75 Energy [eV]

2.80

2.85

The excitonic PL spectrum obtained from a dot sample of diameter 30 nm is shown in Figure 19.7 (solid line); also shown for comparison is the PL spectrum obtained from the QW before the fabrication of the dot (dotted curve). The PL peak of the exciton of these QDs appears at 2.690 eV, in comparison to its appearance at 2.680 eV in the case of the QW (Figure 19.7). A blue shift of 10 meV is observed in the PL peak of the dot, which can

597

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

Figure 19.8 Exciton PL peak energies as a function of time for DMS-QDs with a lateral diameter of 30 nm (closed circles) and the QW (open circles) at an applied magnetic field of 7 T at 4.2 K. The solid and broken lines were plotted by fitting to single exponential functions.

0 –1 Energy Shift (meV)

598

QW

–2 –3 –4 –5 –6 –7 –8 –9

QD (30 nm)

7T 4.2 K 0

100

200

300

400

Time (ps)

be attributed to the lateral confinement effects of the carriers, while the exciton binding energy tends to be lowered in the confined structure. The relaxation of the lattice strain during dot formation, which causes a shift toward lower exciton energy, is negligible, since the lattice mismatch between the QW and the buffer or barrier layer is relatively small (1.4%). In addition, the exciton dynamics are found to be different between the QDs and the QW. The transient PL spectrum was measured in a high magnetic field, as shown in Figure 19.8. The exciton PL peak corresponding to the QDs of diameter 30 nm shows a lower energy shift of 2.5 meV with a time constant of 102 ps, while that of the QW shows a larger shift of 6 meV with a shorter time constant of 60 ps. The application of high magnetic fields prevents the transient formation of excitonic magnetic polarons [7]. Therefore, the smaller spectral diffusion with a longer time constant for the QD excitons indicates a significant suppression of the localization of excitons, possibly due to the limitation of the diffusion length of the excitons in the dots. In the Zn1-x-y Cdx Mny Se QW, the potential fluctuations due to the alloying can cause exciton localization and hence spectral diffusion. The exciton diffusion length is calculated to be 40 nm, which is larger than the dot diameter of 30 nm. Therefore, the exciton diffusion is significantly limited in the dots. The giant Zeeman shift of the excitons in the QDs appears, with the shift in energy being 45 meV at 7 T, which is larger than that observed in the QW before the fabrication of the QDs. By applying a magnetic field, a significant increase in the luminescence intensity is also observed, which originates from the suppression of Auger energy transfer from the excitons to the Mn ions [40]. A double quantum dot (DQD) system, one DMS and the other nonmagnetic, was fabricated from a DQW to study spin manipulations such as spin injection. Using the lithography technique, the DQD, with a lateral diameter of 30 nm, was successfully fabricated from a DQW composed of Zn0.70 Cd0.22 Mn0.08 Se-MW (5 nm thick) and Zn0.76 Cd0.24 Se-NW (10 nm thick). These quantum wells were separated by a 5-nm-thick ZnSe barrier. The structure of the DQD is schematically illustrated in Figure 19.9a. The exciton energy of the DMS-QD was designed to be 100 meV higher than that of the nonmagnetic QD, by adjusting the Cd content and well thickness. In the DQD system, the measured exciton lifetime in the DMS-QD was as short as 20 ps, suggesting that the

19.4 Self-Assembled Quantum Dots

Zn0.76Cd0.24Se QD “Spin detector” Zn0.76Cd0.24Se QD “Spin detector”

ZnSe barrier ZnSe buffer GaAs (100)

σ+

PL Intensity (a.u.)

ZnSe cap



σ

ZnSe y-line

Zn0.70Cd0.22Mn0.08Se DMS QD “Spin aligner” (a) 2.45

2.50

2.55

2.60

2.65

Energy (eV) (b)

Figure 19.9 (a) A double quantum dot (DQD), one magnetic and the other nonmagnetic, structure for the study of spin injection. (b) Excitonic PL spectra with 𝜎 + (a solid line) and 𝜎 − (a broken line) polarizations from the Zn0.76 Cd0.24 Se nonmagnetic QD in the DQD, at a magnetic field of 3 T. A PL peak due to the defect-related ZnSe y-line is also shown. The DQD was excited by linearly polarized light.

excitons tunnel from the DMS-QD to the nonmagnetic QD. Furthermore, the intensity of the exciton PL from the DMS-QD decreased to 1/100 when the barrier thickness decreased from 15 to 5 nm. Owing to the applied magnetic fields, the photoexcited carriers, and hence, the excitons in the DMS-QD, were highly spin-polarized by the Mn spins, which was directly confirmed by the high value of the circular polarization degree P = 0.80 for the exciton PL. Therefore, the spin-polarized excitons could be efficiently injected into the nonmagnetic QD by quantum tunneling through the potential barrier. P values up to 0.20 were observed for the exciton PL from the nonmagnetic QD, as shown in Figure 19.9b. This is experimental evidence of the spin injection occurring in this DQD system. This P value is higher than 0.11, which is the value corresponding to the initial QW before the fabrication of the dot. The enhanced P value can be attributed to the increase in the number of spin-polarized carriers injected from the DMS-QDs into the nonmagnetic QDs. As the DMS-QD is next to the ZnSe buffer layer, a large number of excitons in the ZnSe buffer layer can migrate into the DMS-QD and then get spin-polarized. Therefore, the number of spin-polarized excitons is significantly increased in this QD structure, since the area of the dot is markedly smaller than that of the exciton diffusion, as discussed before. Consequently, a larger number of spin-polarized carriers are injected into the nonmagnetic QD, resulting in a higher P for the exciton PL from the nonmagnetic QD.

19.4 Self-Assembled Quantum Dots Self-assembled quantum dots (SAQDs) are also very attractive for optical applications, because of the discrete density of states of the photoexcited carriers, due to strong

599

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

Figure 19.10 Micro-PL spectra of self-assembled quantum dots (SAQDs) of CdSe in the nanopillar sample (a thick solid line) and in the sample of the plane QD layer before nanopillar fabrication (a thin solid line). The sharp PL peaks of the SAQDs originate from excitons in individual dots.

PL intensity (a.u.)

4.2 K

ZnSe Y-line

2.60

2.65 2.70 Photon Energy ( eV )

quantum confinement effects. In addition, the spin relaxation time of an electron has been reported to be as long as several nanoseconds in SAQDs [41, 42]. Such a long time constant of spin relaxation is an advantage for spin manipulation in semiconductor materials. Figure 19.10 shows the micro-PL spectra of the SAQDs of CdSe, where the plane QD layer was fabricated into a nanopillar shape using electron-beam lithography. The SAQD layer was grown by molecular beam epitaxy with proper growth conditions to realize the Stranski–Krastanov (SK) growth mode [26, 27]. As shown in Figure 19.10, sharp PL peaks appear, with a spectral width of 0.4 meV, which is the instrumental width, indicating excitonic PL from individual SAQDs. The dot diameter and density were determined as 3.5 ± 0.2 nm and 5000 μm−2 , respectively, from the PL energy and the number of PL peaks in the pillar sample with a specific diameter. Subsequently, a coupled SAQD system with a DMS layer was studied [43]. The magnetic field dependences of the Zeeman shift of the exciton are shown in Figure 19.11, for SAQDs of CdSe coupled with a DMS-cap layer of Zn0.80 Mn0.20 Se. Typical giant Zeeman shifts of excitons are observed, which can be well fitted by a Brillouin function. The highest Zeeman shift in the SAQDs reaches 20 meV at a magnetic field of 5 T. The exciton lifetime is also affected by the Mn ions, due to the magnetic-field-induced suppression of the energy transfer process from the exciton to the internal d-d transitions via Auger Figure 19.11 Giant Zeeman shifts of the excitons in self-assembled CdSe QDs (closed circles) and in the magnetic cap layer of Zn0.80 Mn0.20 Se (open circles). The solid lines were fitted by calculations by using Brillouin functions.

30 Zn0.8Mn0.2Se Cap 25 Zeeman shift (meV)

600

ZnSe Buffer

20

CdSe QDs

GaAs (100)

15 10 CdSe QDs

5

Zn0.8Mn0.2Se Cap 0

0

1

2 3 4 Magnetic field (T)

5

19.4 Self-Assembled Quantum Dots

processes [40]. The lifetime of the excitons is 40 ps, which is markedly shorter than the 230 ps for the CdSe dots without a DMS layer. These properties can be explained by the spatial overlapping of the wave functions of the excitons in the QDs with those of the Mn ions in the DMS layer, and also by some diffusion of the Mn ions into the QDs. The coupled SAQDs are studied to demonstrate the spin manipulation. Spin injection into nonmagnetic SAQDs is an important subject to investigate, because it opens the way for realizing semiconductor spintronic devices and quantum computing based on such nanostructures. The direction of the spin can be controlled by the direction of the external magnetic field in the DMS, due to the giant Zeeman effects of the carriers. Therefore, the carrier spins are polarized in SAQDs if the carriers are injected from spin-conserving states. Thus, spins can be manipulated in QDs by using a long relaxation time. A coupled SAQD structure of CdSe with a DMS-QW is schematically illustrated in Figure 19.12a. In this case, a barrier layer of ZnSe is inserted to separate the SAQDs and DMS-QW. The circularly polarized PL spectra from such nanostructures, for a magnetic field of 5 T, are shown in Figure 19.12b. The broad emission band with 𝜎 − polarization corresponds to the excitonic PL from the QD ensemble, which is caused by the dot-size distribution. The peak energy of the QD emission band can be controlled by changing the average dot size. The resultant energy difference between the excitonic PL peaks for the QD band and DMS-QW is 50 meV at a zero magnetic field. The P value of the exciton PL of the CdSe SAQDs without the DMS-QW is about −0.10 in magnetic fields; the negative P value indicates that the intensity of the 𝜎 − -polarized PL is stronger than that of the 𝜎 + polarization. The negative P value originates from the negative g-values of excitons in the SAQDs of CdSe. The exciton PL peak for the DMS-QW is indicated by an arrow of EMW in Figure 19.12b. The magnetic field dependence of the EMW peak shows a giant Zeeman shift, and therefore, the spectrum appears only in the 𝜎 + -polarized component, in nonzero magnetic fields. Moreover, as shown in Figure 19.12b, an additional PL peak, indicated by EA , appears for 𝜎 + polarization. The energy of the additional PL peak depends on the magnetic field, and agrees with

E MW

ZnSe-cap Zn0.68Cd0.22Mn0.10Se-QW

CdSe-QDs ZnSe-barrier

ZnSe-buffer

PL Intensity (a. u.)

EA

2.60 (a)

E CdSe–QD

σ+ σ–

2.65

2.70 2.75 Energy (eV)

ZnSe

2.80

(b)

Figure 19.12 (a) A coupled SAQD structure of CdSe with a DMS-QW of Zn0.68 Cd0.22 Mn0.10 Se (MW) and a 5 nm-thick tunneling barrier of ZnSe. (b) Circularly polarized excitonic PL spectra of the coupled SAQDs at the magnetic field of 5 T. The 𝜎 + -polarized PL spectrum (a solid line) is composed of three components (see text), while the 𝜎 − -polarized spectrum (a broken line) shows the exciton PL from the CdSe QD ensemble only.

601

602

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

that calculated from a Brillouin function with the same parameters as those for the DMS well. The energy of the additional PL peak is 25 meV, which is lower than that of the exciton energy EMW in the DMS-QW. This energy difference is almost equal to the LO-phonon energy of 27 meV for a single CdSe QD layer. To interpret this additional PL peak, the electron and hh levels were calculated for the coupled structure in nonzero magnetic fields. As a result, type-II band alignment can be assumed for a down-spin electron in the QDs and a down-spin hh in the DMS-QW under high magnetic fields. The time-resolved circularly polarized PL shows that the lifetime of such an additional PL is as long as 3.5 ns. The exciton PL lifetime in the DMS well and CdSe QDs are 60 and 240 ps, respectively, which are nearly the same as those obtained from their corresponding single layers. Therefore, we conclude that the additional PL of the coupled structure originates from the type-II transition between electrons injected from the DMS well and hhs remaining in the DMS part. The photoexcited down-spin electrons in the DMS-QW can efficiently penetrate SAQDs via LO-phonon scattering, although the down-spin hhs remain inside the DMS because of the type-II band alignment. The down-spin hh level in the DMS-QW is 6 meV lower than that in the dot at a magnetic field of 5 T, where the electron energy in this dot is assumed to be 1-LO-phonon energy lower than that in the DMS-QW. This scenario is appropriate for realizing electron spin injection into SAQDs. The electron spin injection takes place efficiently into the SAQDs via LO-phonon-assisted resonant tunneling, since the energy level in the QD is completely discrete. The magnetic field dependence of the above spin injection as well as the related type-II transition between electrons in the same SAQDs of CdSe and hhs have also been observed in the DMS-QW [44]. The type-II emission energy depends on a magnetic field, since the electron energy in the DMS-QW is a strong function of the field strength. The magnetic field dependence of the intensity of the type-II emission suggests a level-crossing field of the hh states between the SAQD and the DMS-QW. The type-II transition is induced by an electron injected from the DMS-QW into the SAQD and a spin-polarized hh in the DMS-QW, indicating that the electron tunneling is resonantly assisted by the energy relaxation involving the abovementioned LO-phonon scattering. The spin-injection dynamics has been further studied in a CdSe-SAQD layer stacked with a 100-nm-thick DMS layer of Zn0.80 Mn0.20 Se, where both layers were separated by a 10-nm-thick spacer layer of ZnSe [45]. In this layered system, one can avoid the complicated band lineup previously studied between the SAQDs and DMS-QW, where excitons in the DMS are off-resonantly situated energetically above the exciton system of the SAQDs. P values of excitonic PL at 5 T in the coupled SAQDs exhibit a rapid increase with increasing delay time, up to 0.3 at 25 ps after pulse excitation of the DMS by a linearly polarized light, as shown in Figure 19.13. This transient PL is observed at the high-energy side of the PL emission band of the SAQDs due to the size distribution. The development of a positive P value directly reflects the spin-injection dynamics from the DMS, since the intrinsic polarization of the SAQD excitons due to Zeeman splitting is P = −0.1 when only the SAQDs are selectively excited. The P value tends to gradually decay with time after reaching its maximum, as a result of the exciton spin relaxation with a time constant of 800 ps in the SAQDs. A time constant of 10 ps was deduced to be responsible for the spin injection aided by the rate equation fitting. A time-integrated P value of 0.2 is obtained for the higher-energy region of the inhomogeneous SAQD

19.4 Self-Assembled Quantum Dots

σ+

0.8

0.4

σ–

0.0 0.4 0

100 Time (ps)

0.0

Circular polarization degree

PL intensity (arb. units)

0.8

E = 2.585 – 2.564 eV 0.0

0.2

0.4

0.6

Time (ns)

2.45

0.2

σ–

σ+ 0.0

2.50

2.55

Circular polarization degree

Figure 19.14 Circularly polarized exciton PL spectra integrated over a time window of 0–200 ps (thick solid lines) and the corresponding P values as a function of photon energy (a thin solid line) for the SAQDs, obtained at 2 K and 5 T. The solid circles show the calculated P values based on the simplified spin-injection model taking the spin-injection dynamics and the subsequent spin relaxation inside the individual SAQDs into account.

PL intensity (arb. units)

Figure 19.13 Circularly polarized exciton PL intensities (the open circles) and the corresponding P values (the closed circles) as a function of time for the spectral window 2.564–2.585 eV, measured at 5 T in the coupled SAQDs. The solid lines are the calculations using the rate equations for the PL intensities and P, taking the spin-injection dynamics and the subsequent spin relaxation in the SAQDs into account. The inset shows a close-up over the shorter time period.

2.60

Photon energy (eV)

emission band, with an exciton transfer time of 40 ps toward lower-energy SAQDs with larger QD sizes via interdot tunneling. The time-integrated P value decreases down to −0.04 with decreasing exciton energy, where the spin relaxation within the SAQD plays a crucial role in the P value observed in the spin-injection dynamics. Figure 19.14 shows circularly polarized PL spectra and the P value variation in the PL emission band of these SAQDs, integrated over the time range 0–200 ps. In the high-energy region above 2.55 eV, the P values calculated by the rate equations, taking the spin-injection dynamics and subsequent spin relaxation inside individual SAQDs into account, agree well with the experimental values. This is consistent with the fact that the rate equation analysis can reproduce the experimental decay curves for the high-energy region of the QD emission band, as shown in the above Figure 19.13. At lower energies, a systematic decrease in the P value from +0.25 to −0.04 is observed with decreasing exciton energy.

603

604

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

The discrepancies between the experimental P values and the calculated ones are also apparent in the middle- and low-energy regions. This observation can at least partly be explained by the interdot exciton transfer occurring in the studied SAQD structure due to the high dot density. It should be noted that the simplified model of the rate equation analysis includes only the exciton transfer from the dot under consideration to other dots, but not the other way around. Therefore, it is only appropriate for QDs with the highest exciton energy. The observed discrepancies in the P value thus suggest the existence of multiple paths for exciton transfer in this dense dot system, which should be considered when analyzing the results in the middle- and lower-energy regions of the QD emission band. A PLE study on the exciton spin injection efficiency was also reported in this coupled SAQD system with the DMS-QW [46]. The injected spin polarization during the injection process, including crossing the heterointerfaces and energy relaxation within the SAQDs, is determined to be 32% at 5 T, decreasing from 100% in the DMS-QW before the injection. The layered structure of the SAQDs stacked with the DMS provides the off-resonant band lineup with marked energy-level differences between them. Therefore, a remarkable spin loss can be induced. Further studies are required to identify the exact physical mechanisms for the observed spin loss, with the aim of improving the spin-injection efficiency in the SAQDs.

19.5 Hybrid Nanostructures with Ferromagnetic Materials The characteristic optical properties such as the giant Zeeman effects of excitons, and the giant Faraday rotation in II–VI DMSs, are attractive for applications in magneto-optical devices operating in the visible light region; an optical isolator operating at room temperature has already been developed [4]. In addition, spin-polarized carriers can be generated in DMSs due to the giant Zeeman splitting of their electronic states, which is also used as a spin aligner in spintronic devices. In these studies, applications of external fields to the II–VI DMSs are necessary for the alignment of paramagnetic spins in the DMSs. For applying efficient magnetic fields to DMS nanostructures, it is important to prepare hybrid structures of DMSs with ferromagnetic materials. This offers a new technology for developing microscopic magneto-optical and spin-related devices without significant energy consumption. Therefore, several types of hybrid structures with DMS-QWs have been proposed [33, 34]. We fabricated hybrid nanostructures of DMS-QWs with ferromagnetic cobalt (Co) wires [34–36]. The DMS-QW was made into wires of width as low as 100 nm that were sandwiched between Co wires. Resonant spin-flip light scattering was observed at room temperature in the hybrid nanostructure, indicating a Zeeman shift of paramagnetic spins of Mn ions and hence spin alignment of Mn ions in DMS. Moreover, we have studied the giant Zeeman effects of the excitons at 4.2 K in the same hybrid nanostructure, where the Co magnetization in the wires was aligned normal to the wire surface by applying an external magnetic field. The advantage of this hybrid nanostructure is that nearly perpendicular magnetic fields can be produced between the Co wires, originating from the overlapping magnetic fields generated by neighboring Co wires. Therefore, the whole DMS-QW part in the hybrid structure can experience the perpendicular fields applied from the Co wires, which provide a Faraday configuration for the optical

19.5 Hybrid Nanostructures with Ferromagnetic Materials

Co DMS BCo-flux

DMS-QW

M

Co Bext

1 μm (a)

(b)

Figure 19.15 (a) A schematic diagram of the hybrid nanostructure of the DMS-QW with ferromagnetic Co wires. The magnetic field distribution generated by Co magnetization is also illustrated, where the Co magnetization was aligned by applying an external magnetic field. (b) An AFM image of the surface of the hybrid nanostructure.

measurements. As a result, the magneto-optical properties, such as the giant Zeeman effects of excitons, are expected to be affected by the magnetic flux from the Co ions. The hybrid nanostructure fabricated for this study is schematically illustrated in Figure 19.15a. Wires of a Zn0.69 Cd0.23 Mn0.08 Se-QW of 200 nm width are sandwiched between Co wires of 400 nm width and 50 nm thickness. A calculation of the magnetic field distribution around the gap of the Co wire system shows that a field of intensity up to 0.5 T, which is nearly perpendicular to the well plane, can be applied at the position of the DMS-QW when the Co magnetization is perfectly aligned normal to the wire plane. The intensity of the lateral component of the magnetic field is less than 10% of that of the perpendicular component. To realize this situation, external magnetic fields higher than an in-plane magnetic shape-anisotropy field of Co (4𝜋Ms = 1.8 T) are applied normal to the wire surface. When the Co magnetization orients perpendicular to the film surface, a magnetic flux BCo-flux from the Co magnetization can be applied in the opposite (anti-parallel) direction of the external field. Therefore, an effective perpendicular magnetic field B⟂ eff = Bext − B⟂ Co-flux is finally applied to the DMS-QW. The perpendicular magnetic field B⟂ eff induces giant Zeeman effects of excitons in the DMS-QW. The effect of the B⟂ Co-flux on the giant Zeeman shift of excitonic PL can be distinguished from the PL spectral shape, since the strength of the B⟂ Co-flux is inhomogeneous in the DMS-QW part, and thus can cause a broadening in the PL spectrum, whereas that of the Bext was made perfectly uniform in our experimental setup using a micro-PL system. An atomic force microscopy (AFM) image of the surface of the hybrid nanostructure is shown in the Figure 19.15b. The full width at half maximum (FWHM) of the exciton PL spectrum for the hybrid nanostructure increases significantly with an increase in the external field, as shown in Figure 19.16a. It is attributed to the giant Zeeman effects with various energy shifts due to the inhomogeneous magnetic flux applied from the Co wires. At zero external magnetic field, the Co magnetization lies completely in-plane, and no magnetic flux

605

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

0.04

0.03

0.02 DMS-QW

0.01

4πMs = 1.8 T 0.00

0

1

2 3 4 Magnetic field (T) (a)

PL data Mn d-d Calculation Fitting

Normalized PL intensity

Hybrid structure FWHM (eV)

606

2T 5

–0.05

0.00 Energy (E – Epeak) (eV)

0.05

(b)

Figure 19.16 (a) FWHM of the exciton PL spectrum as a function of the external field in the hybrid nanostructure (closed squares) and in the plane QW before fabrication of the hybrid nanostructure (closed circles). The in-plane shape-anisotropy field of the Co wire is 1.8 T, at which the Co magnetization aligns perpendicular to the wire surface. (b) PL spectrum of the hybrid nanostructure at 2 T (open circles). The calculated PL spectral shape (a broken line) and the broad spectrum tail due to d-d internal transitions of Mn ions in the DMS-QW (a dotted line). The superposition of both the spectra (a solid line) is compared with the PL spectrum. On the horizontal axis, the E-E peak corresponds to the photon energy difference from the PL peak energy E peak .

in the Faraday configuration is produced from the Co wire system to the DMS-QW. As the external field increases, the Co magnetization rotates gradually toward the perpendicular direction of the film surface, and the B⟂ Co-flux starts to act on the excitons in the DMS-QW, as in the Faraday configuration. Therefore, the FWHM of the exciton PL spectrum in the hybrid nanostructure gradually increases with an increase in the external magnetic field, and saturates at 1.9 T, which is almost equal to the in-plane shape-anisotropy field (1.8 T) of the Co film. To confirm the giant Zeeman effects due to the application of the B⟂ Co-flux , a quantitative analysis of the spectral shape was made on the basis of the field distribution of the B⟂ Co-flux in the position of the DMS-QW. The result was compared with the observed PL spectrum at a magnetic field of 2 T, as shown in the Figure 19.16b. The calculated result for the spectral shape shows a good agreement with the spectrum observed, except for the tail on the lower-energy side. A further study was attempted by using ferromagnetic thin films with a perpendicular residual magnetization at a zero external magnetic field [37]. Nanoscale disks of DMS-QW with a diameter of 80 nm were embedded in Co/Pt multilayered films with interfacial uniaxial perpendicular magnetic anisotropy, as shown in Figure 19.17, generating magnetic fields perpendicular to the DMS-QW plane in the nanodisk. An advantage of this disk shape of the DMS-QW lies in the field strength from the magnetic film, which is higher than that in the case of a wire shape with the same lateral dimension. As a result, exciton PL with circular polarization properties arises in a zero external field owing to the giant Zeeman effects, as shown in Figure 19.18. Before the PL measurements, external fields of 2 T were applied in opposite directions perpendicular to the Co/Pt film surface. Therefore, the magnetization direction in the Co film was changed. Then, the magnetic fields were completely switched off, and the PL measurements were performed with no external magnetic field. As can be

19.6 Conclusions

Figure 19.17 SEM image for the surface of the fabricated hybrid nanostructure, where the disk shape of DMS-QW is embedded in the Co/Pt multilayer.

DMS-disk

100 nm σ+

σ–

σ–

PL intensity (arb. units)

σ+

2.50

2.55

2.60

2.65 2.70 2.50 2.55 Photon energy (eV)

2.60

2.65

2.70

Figure 19.18 Circularly polarized excitonic PL spectra in the hybrid DMS nanostructure with the Co/Pt film, in a zero external magnetic field, where the excitation was made by using a linearly polarized light. The external magnetic fields were applied prior to the PL measurements in opposite directions (left and right figures) perpendicular to the Co/Pt film plane, and were then removed before the measurement.

seen in Figure 19.18, the intensity difference between the PL spectra with different circular polarization probes direct experimental evidence for the giant Zeeman effects of excitons in the DMS-QW is in the present hybrid nanostructure. The typical value of circular polarization degree P is 10%. The sign of P reverses when the direction of the external field applied before the PL measurement is reversed. This also is experimental evidence that the perpendicular magnetic fields from the Co/Pt film align with the exciton spins in the DMS-disk.

19.6 Conclusions Dynamical spin-injection processes of excitons have been studied in a DQW composed of an MW and NW. In the magnetic Zn0.96 Mn0.04 Se/Zn0.76 Cd0.24 Se DQWs with 4- and

607

608

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

8 nm-thick barriers, spin injection takes place by the fast and slow injection processes with time constants of 5 and 1 ns, respectively. Excitons in the MW recombine with a faster decay time of 30 ps than that in the slow spin-injection process. Therefore, the slow spin injection from the MW is quantitatively explained by individual tunneling of electrons and hhs. The dynamical processes of optically created e-h plasma and spin-polarized excitons in the DMS-MQW of Cd0.95 Mn0.05 Te have been studied by using transient pump-probe spectroscopy. The BGR induced by the pump-created e-h plasma is observed in the time region 0–3 ps. After the e-h plasma decays, the bleached peak of exciton absorption increases in intensity and shows spin-dependent transient behavior in the presence of magnetic fields. The electron spin relaxation time is determined as a function of the magnetic field. The electron spin relaxation time is found to be affected by the magnetic field, through the s-d exchange mechanism with the local magnetic spins of Mn ions. Nanoscale dots and wires of a Zn1-x-y Cdx Mny Se-based DMS have been fabricated using electron-beam lithography. A blue shift of the excitonic PL is observed with a decrease in the dot diameters, which indicates lateral confinement for the exciton. The exciton diffusion is significantly suppressed in these nanostructures, which also results in enhancement of the giant Zeeman effects of excitons. Coupled QDs composed of the DMS-MW and NW have been successfully fabricated. The excitons tunnel from the MW into the NW, and a circular polarization degree of 20% is obtained for the exciton PL of the NW due to spin injection from the MW to the NW. In addition, SAQDs of CdSe coupled with a DMS-QW were examined, where the electron spin levels of the SAQD align quasi-resonantly. Spin injection via one LO-phonon-assisted resonant tunneling has been observed, with high efficiency of the spin-conserving injection. This is expected to launch a new technology for spin manipulation in SAQDs. The spin-injection dynamics and the role of LO-phonon scattering have also been studied in the case of off-resonant band alignment between the SAQDs and DMS. Ultrafast spin injection is demonstrated; however, a spin loss during the injection processes has been observed, where the spin relaxation due to the interdot migration of excitons among high-density SAQDs as well as the energy relaxation between the DMS and SAQDs are found to be responsible for the spin loss. Hybrid nanostructures of DMSs with ferromagnetic materials have been fabricated, such as DMS-Co hybrid wires and DMS nanodisks embedded in Co/Pt multilayered films, with uniaxial magnetic anisotropy. The magneto-optical properties of the hybrid nanostructures have been studied by excitonic PL in the hybrid structure. The circular polarization properties of PL have been observed under a zero magnetic field in the latter DMS-hybrid nanodisks with Co/Pt, owing to the giant Zeeman effects induced by the perpendicular magnetic flux due to the perpendicular magnetization alignment in the Co/Pt.

Acknowledgments The authors are indebted to K. Nishibayashi, Z.H. Chen, K. Hyomi, I. Souma, K. Kayanuma, S. Shirotori, H. Sakurai, K. Seo, H. Ikada, A. Uetake, T. Tomita, T. Asahina, and M. Sakuma for their helpful collaborations. The authors are also grateful to Professors Weimin Chen and Irina Buyanova in Linkoping University for their

References

fruitful collaboration. This work was supported by the Nanotechnology Project of NEDO and also by the Ministry of Education, Science, and Culture, Japan.

References 1 Furdyna, J.K. and Kossut, J. (eds.) (1988). Diluted Magnetic Semiconductors, Semicon-

ductors and Semimetals, vol. 25. New York: Academic Press. 2 Awschalom, D.D. and Samarth, N. (1999). J. Magn. Magn. Mater. 200: 130. 3 Awschalom, D.D., Loss, D., and Samarth, N. (eds.) (2002). Semiconductor Spintronics 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37

and Quantum Computation. Berlin: Springer. Onodera, K., Masumoto, T., and Kimura, M. (1994). Electron. Lett. 30: 1954. Harris, J.H. and Nurmikko, A.V. (1983). Phys. Rev. Lett. 51: 1472. Yakovlev, D.R., Mackh, G., Kuhn-Heinrich, B. et al. (1995). Phys. Rev. B 52: 12033. (a) Nogaku, M., Shen, J.X., Pittini, R. et al. (2001). Phys. Rev. B 63: 153314. (b) Nogaku, M., Pittni, R., Sato, T. et al. (2001). J. Appl. Phys. 89: 7287. Oka, Y., Kayanuma, K., Shirotori, S. et al. (2002). J. Lumin. 100: 175. Oka, Y., Kayanuma, K., Nakayama, E. et al. (2002). Nonlinear Opt. 29 (7–9): 491–499. Fiederling, R., Keim, M., Reuscher, G. et al. (1999). Nature 402: 787. Oestreich, M., Huebner, J., Haegele, D. et al. (1999). Appl. Phys. Lett. 74: 1251. Ohno, Y., Young, D.K., Benschoten, B. et al. (1999). Nature 402: 790. Chen, W.M., Buyanova, I.A., Rudko, G.Y. et al. (2003). Phys. Rev. B 67: 125313. Kayanuma, K., Shirado, E., Debnath, M.C. et al. (2001). J. Appl. Phys. 89: 7278. Kayamuma, K., Shirado, E., Debnath, M.C. et al. (2001). Physica E 10: 295. Kayanuma, K., Shirotori, S., Chen, Z.H. et al. (2003). Physica B 340-342: 882. Chen, W.M., Buyanova, I.A., Kayanuma, K. et al. (2005). Phys. Rev. B 72: 073206. Trankle, G., Leier, H., Forchel, A. et al. (1987). Phys. Rev. Lett. 58: 419. Baumberg, J.J., Awschalom, D.D., Samarth, N. et al. (1994). Phys. Rev. Lett. 72: 717. Crooker, S.A., Awschalom, D.D., Baumberg, J.J. et al. (1997). Phys. Rev. B 56: 7574. Chen, Z.H., Sakurai, H., Seo, K. et al. (2003). Physica B 340–342: 890. Sakurai, H., Seo, K., Chen, Z.H. et al. (2004). Phys. Status Solidi (c) 1: 981. Chen, Z.H., Sakurai, H., Tomita, T. et al. (2004). Physica E 21: 1022. Illing, M., Bacher, G., Kummell, T. et al. (1995). J. Vac. Sci. Technol. B 13: 2792. Ray, O., Sirenko, A.A., Berry, J.J. et al. (2000). Appl. Phys. Lett. 28: 1167. Shibata, K., Nakayama, E., Souma, I. et al. (2002). Phys. Status Solidi (b) 229: 473. Kuroda, T., Hasegawa, N., Minami, F. et al. (1999). J. Lumin. 83/84: 321. Bhattacharjee, A.K. and Benoit a la Guillaume, C. (1997). Phys. Rev. B 55: 10613. Takahashi, N., Takabayashi, K., Souma, I. et al. (2000). J. Appl. Phys. 87: 6469. Chen, Z.H., Debnath, M.C., Shibata, K. et al. (2001). J. Appl. Phys. 89: 6701. Ikada, H., Saito, T., Takahashi, N. et al. (2001). Physica E 10: 373. Uetake, A., Ikada, H., Asahina, T. et al. (2004). Phys. Status Solidi (c) 1: 941. Kossut, J., Yamakawa, I., Nakamura, A. et al. (2001). Appl. Phys. Lett. 79: 1789. Sakuma, M., Hyomi, K., Souma, I. et al. (2003). J. Appl. Phys. 94: 6423. Sakuma, M., Hyomi, K., Souma, I. et al. (2004). Phys. Status Solidi (b) 241: 664. Sakuma, M., Hyomi, K., Souma, I. et al. (2004). Appl. Phys. Lett. 85: 6203. Murayama, A. and Sakuma, M. (2006). Appl. Phys. Lett. 88: 122504.

609

610

19 Optical Properties and Spin Dynamics of Diluted Magnetic Semiconductor Nanostructures

38 Park, J., Murayama, A., Souma, I. et al. (2008). Jpn. J. Appl. Phys. 47: 3533–3536. 39 Park, J.H., Saito, K., Souma, I. et al. (2007). Jpn. J. Appl. Phys. 46: 7290–7293. 40 Nawrocki, M., Rubo, Y.G., Lascaray, J.P., and Coquillat, D. (1995). Phys. Rev. B 52: 41 42 43 44 45 46

R2241. Paillard, M., Marie, X., Renucci, P. et al. (2001). Phys. Rev. Lett. 86: 1634. Mackowski, S., Nguyen, T.A., Jackson, H.E. et al. (2003). Appl. Phys. Lett. 83: 5524. Souma, I., Kayanuma, K., Hyomi, K. et al. (2005). J. Supercond. 18: 219. Murayama, A., Asahina, T., Nishibayashi, K. et al. (2006). J. Appl. Phys. 100: 084327. Murayama, A., Furuta, T., Hyomi, K. et al. (2007). Phys. Rev. B 75: 195308. Dagnelund, D., Buyanova, I.A., Chen, W.M. et al. (2008). Phys. Rev. B 77: 035437.

611

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors Ruben Jeronimo Freitas 1 and Koichi Shimakawa 2 1 2

Department of Electrical and Electronic Engineering, National University of Timor Lorosae, Dili, East Timor Department of Electrical and Electronic Engineering, Gifu University, Gifu, 501-3122, Japan

CHAPTER MENU Introduction, 611 A Review of PPC in III-V Semiconductors, 613 Key Physical Terms Related to PPC, 615 Kinetics of PPC in III-V Semiconductors, 617 Conclusions, 623 On the Reaction Rate Under the Uniform Distribution, 623 References, 625

20.1 Introduction Semiconducting materials when illuminated by light of energy equal or greater than the bandgap energy create electron–hole pairs. These pairs are subsequently separated, contributing to the electrical transport in such materials and making them photoconductors. Crystalline III-V semiconductors, including particularly Gallium nitride (GaN), are important materials for applications in optoelectronic and microelectronics devices [1–3]. The 2014 Nobel Prize in Physics was awarded to Profs. I. Akasaki, H. Amano and S. Nakamura for their invention of the light emitting diode (LED) using GaN. Photoconductivity spectroscopy is known as a powerful technique to understand the charge carrier generation, trapping, and recombination processes in semiconducting materials [4]. Persistent photoconductivity (PPC), i.e. persistence of photocurrent after the termination of photoexcitation, particularly at low temperatures, is found in III-V semiconductors [1–3, 5–9]. This striking feature is also found in disordered semiconductors such as hydrogenated amorphous silicon (a-Si:H) [10] and amorphous chalcogenides [11, 12]. Figure 20.1 shows a typical example of PPC observed in GaN [6]. The buildup and decay behavior depend highly on temperature. We will find a very slow carrier relaxation at low temperatures. The relative change in conductance becomes significantly smaller at higher temperatures, and the time constant decreases. Note also that these characteristics depend on the photoexcitation energy, which is shown in Figure 20.2. We notice that at low excitation energy the relaxation time becomes shorter than that Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

5

4.0 light off T = 77 K 3.9

4

light off

3.8

T = 302 K

3

Conductance (mS)

Conductance (mS)

612

3.7

light on 2 0.0

0.5

1.0

1.5

2.0

2.5

4

Time (10 s)

Figure 20.1 Buildup and decay transient of the conductance at 77 and 302 K. Source: Reprinted from M.T. Hirsch et al., Appl. Phys. Lett. 71, 1098 (1997) by American Institute of Physics.

at higher energies, and no significant PPC is observed [1]. It is important to understand why such a slow photo-response appears in these semiconductors. As will be discussed in Sections 20.3 and 20.4 the time evolution of changes in the number of carriers or current, noted by y in Eq. (20.1), during illumination (buildup) is empirically represented as [ { ( )𝛽 ] t (20.1) y = C 1 − exp − 𝜏 where C is a constant, 𝜏 is the effective response time, and 𝛽 (0 < 𝛽 < 1) is the dispersion parameter. The decay relaxation is proportional to exp[−(t/𝜏)β ], which is called the stretched exponential function (SEF) [12]. It is known that the SEF appears frequently when we discuss kinetic behaviors for any physical response involving mechanical phenomena in disordered matter [13–17]. This means disorder should be related to the origin of the SEF. However, since the first discovery of SEF by Kohlrausch [18], the origin of the SEF has remained one of nature’s best-kept secrets even though there have been many attempts to derive the form of the SEF as a universal law [16]. Trachenko and Dove [17], for example, discussed the SEF in terms of local events for relaxation in glasses, in which the activation thermal barrier itself changes with the number of local events. An alternative approach to deriving the SEF is to assume a certain distribution of the relaxation (or reaction) time [12, 14]. In this chapter, the origin of the SEF observed for the PPC in III-V semiconductors is discussed in detail. Of particular interest is the exploration of why the PPC dynamics follows the SEF. It is found that the dispersive nature in defect-related kinetics is deeply related to the origin of the SEF.

20.2 A Review of PPC in III-V Semiconductors

Figure 20.2 Typical buildup and decay behavior of PPC in n-type GaN, which depends on the exciting light energy. Source: Reprinted from C.V. Reddy et al., Appl. Phys. Lett. 73, 244 (1998) by American Institute of Physics.

Room Temperature (297 K)

2.4

White Light

1.8 1.2 0.6 0 0.4

1.96 eV

Current (nA)

0.2

0 0.15

1.77 eV

0.1 0.05 0 1.38 eV

0.04

0.02

0 0

1000

2000 3000 4000 Time (sec)

5000

6000

20.2 A Review of PPC in III-V Semiconductors Before proceeding with the discussion, we will first review the previous works on PPC in III-V semiconductors. Figure 20.3 is a well-known result for the decay of free-electron concentration (deduced from the Hall measurement) after stopping illumination in n-type Ga1-x Alx As [13]. As the conductivity is proportional to the number of free charge carriers, the data here are equivalent to the decay of photoconductivity. It was found that the high electron concentration persists, indicating that the capture cross section for the photoexcited carriers is less than 10−30 cm2 , which is six orders of magnitude smaller than the normal cross section for ionized or neutral impurity scattering [13]. The capture process is thermally activated with a barrier height of 0.18 eV, and the photoionization energy is 1.1 eV for the sample of x = 0.36, indicating a large Stokes shift due to a large lattice relaxation. It was then suggested that the donor atom associated with an unknown defect (X) should be related to the PPC, which was called the DX center. The term “DX” stands for the donor atom “D” associated with an unknown defect “X” [15].

613

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

Free-Electron Concentration (1016cm–3)

Ga1-xalx As x = 0.36 Te DOPED

T = 63 °K

10 T = 76 °K

T = 82 °K

1 0

100

200 Time (s)

Figure 20.3 Decay of the free-electron concentration after stopping illumination at different temperatures in n-type Ga1-x Alx As. Source: Modified from R.J. Nelson, Appl. Phys. Lett. 31, 351 (1977).

Electronic + Defect Coordinate Energy (eV)

614

3.0

D

C

1.1 eV Eec = 0.2 eV Eee = 0.3 eV

2.0 ED = 0.1 eV

1.0

V

0.0 0

QR

Defect Configuration Coordinate, Q (arb. units)

Figure 20.4 Configuration coordinate diagram proposed for n-type Alx Ga1-x As. Source: Modified from D.V. Lang and R.A. Logan, Phys. Rev. Lett. 39, 635 (1977).

Figure 20.4 shows the configuration coordinate diagram (CCD) for a DX center proposed for Alx Ga1-x As [19]. Curves C and V correspond to the system energy of the conduction and valence bands, respectively, of an unoccupied defect with a delocalized electron. Curve D corresponds to the shifted conduction band due to vibrations of an occupied defect. The thermal barrier to electron capture and to electron emission

20.3 Key Physical Terms Related to PPC

Figure 20.5 Microscopic model for the DX center in Si-doped Alx Ga1-x As. “0” and “−” denote the neutral and negatively charged states (ground state), respectively. Source: Modified from A. Alkauskas et al., J. Appl. Phys. 119, 181101 (2016).

0 As

Si

– Eabs As

Si

are indicated as Eec and Eee , respectively. The DX center is an example of a defect that is formed when shallow donors undergo a large lattice relaxation and become deep donors. The microscopic model of a DX center in Si-doped GaAs is shown in Figure 20.5 [20]. The fourfold-coordinated Si (sp3 bonding) forms a shallow-donor level (denoted by “0”), whereas the threefold-coordinated Si (p3 bonding) representing the distorted atom forms a deeper level (denoted by “−”: charged negatively with a negative-U character). The DX center itself is a donor impurity that relaxes away from its substitutional site (“0”) to the charged site (“−”) (ground state). It is therefore bistable in nature, i.e. one relaxed and other unrelaxed states. As will be discussed in Section 20.4, the DX centers lie at the shallow-donor state (“0”). To revert to the DX ground state (“−”), the defect must capture an electron and surmount a thermal barrier as shown by the arrow. If the thermal barrier height is much greater than the thermal energy, the excited state may last for hours or days, which is the origin of PPC. Note here that Eabs is its optical ionization energy. As the PPC is also found in p-type III-V semiconductors, it is expected that the acceptor should also show the same behavior, and hence we use a similar CCD for the acceptor (AX center) [21].

20.3 Key Physical Terms Related to PPC Before discussing PPC, the following key physical terms are introduced in this section to make the understanding of PPC easier in photoconducting materials. 20.3.1

Dispersive Reaction

What is a dispersive reaction? The definition of a dispersive reaction is that in any reaction, for example, in physics and chemistry, the reaction rate depends on time: usually the reaction rate decreases with time, but it does not remain constant. It is important to note that the dispersive reaction is a general feature of random systems such as amorphous semiconductors. In many cases of reactions, the time-dependent reaction rate K(t) (s−1 ) is given by [22] K(t) = Bt 𝛽−1 ,

(20.2)

where B is a constant, and 𝛽 (0 < β < 1) is called the dispersion parameter. The main objective of this chapter is to clearly understand how and why dispersive reactions occur

615

616

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

in some systems. Therefore, first we will discuss the origin of PPC in crystalline III-V semiconductors. It may be noted that PPC is always found in disordered (amorphous) semiconductors [10–12]. The term order means lattice regularity from crystallography. However, we suppose that the local structure of localized states, including defect states, has disorder in ordered materials. 20.3.2

SEF and Power Law

The SEF and the power law have been commonly observed in any relaxation (electronic or mechanical) [5, 10, 12, 23]. The SEF itself was first introduced by Kohlrausch, who applied it in studying the discharge of a capacitor [18]. From the mathematical point of view, the SEF and the power law can be derived, respectively, from the following differential equations: dn∕dt = −K(t)n, (monomolecular type)

(20.3)

dn∕dt = −K(t)n , (bimolecular type)

(20.4)

2

where n is the charge carrier concentration that takes part in the reaction. The respective solutions of Eqs. (20.3) and (20.4) are ( ) t𝛽 n = Cexp −B , (20.5) 𝛽 𝛽 (20.6) n = t −𝛽 , B where C is a constant. It should be noted that if 𝛽 = 1.0. Eq. (20.5) gives simple exponential decay with a constant reaction rate K(t) = B. In this case, the simplest rate equation for the change in the photoexcited free carrier density n is expressed as [24] dn n =G− , (20.7) dt 𝜏 where G is the photocarrier generation rate, and 𝜏 = 1/B is the recombination time. The solution of Eq. (20.7) is obtained as n(t) = G𝜏[1 − exp(−t∕𝜏)],

(20.8)

and with G = 0 in Eq. (20.7), the solution becomes n(t) = n(0) exp(−t∕𝜏),

(20.9)

which is just the decay equation, with n(0) being the initial concentration. According to Eq. (20.8), in the steady state as t → ∞, we get n = G𝜏. These simple rise and decay characteristics are obtained from a time-independent reaction rate K(t) = B = 1/𝜏, and are usually not observed in amorphous semiconductors, because of the involvement of electron and hole traps that are distributed randomly in energy and space. When the traps are present with a random distribution, the carrier recombination process becomes complex, and the recombination rate is perturbed by the trapping and de-trapping processes: The response time becomes longer and longer due to the filling and emptying of the traps during the rise and decay. In this case, n(t) can be described as the SEF or the power law.

20.4 Kinetics of PPC in III-V Semiconductors

20.3.3

Waiting Time Distribution

What is the waiting time for an event that occurs randomly? This can be answered through probability theory. The Poisson distribution is normally used to govern the occurrence of random events in space or time. The Poisson waiting-time probability distribution for a single occurrence of an event is given by [25] Ψ(t) = 𝜈e−𝜈t ,

(20.10)

where 𝜈 (s−1 ) is the rate of occurrence of a random event; the inverse of time. Using Eq. (20.10), the average waiting time for an event is given by ∞

t=

t𝜓(t)dt

∫0 ∞

=

∫0

𝜈te−𝜈t dt =

1 , v

(20.11)

which is the inverse of the constant rate 𝜈. As described above, the constant reaction rate is not applicable when 𝜈 does not take a single value and is given by a distribution function P(𝜈), which gives a modified Poisson distribution for the waiting-time distribution as ∞

Ψ(t) =

∫0

P(ν)νe−𝜈t dν.

(20.12)

where the rate 𝜈 is a function of another physical parameter U that is statistically distributed (i.e. P(U)). Then the probability density P(𝜈) is given by [25] P(U) . (20.13) | dν | | dU | | | The above equation is very useful for connecting the relation between P(U) and P(𝜈). P(ν) =

20.4 Kinetics of PPC in III-V Semiconductors As discussed above, PPC is commonly observed in crystalline III-V semiconductors [1–3, 5–9]. In one of the III-V materials, Alx Ga1-x As (x > 0.22), PPC is believed to be caused by the DX centers. The PPC observed in GaN [1, 2], is of interest here because PPC is related to the so-called yellow luminescence (YL; peak around 2.2 eV) in n-type GaN [2]. In this section, PPC observed in GaN is discussed as a case example for the III-V semiconductors. Figure 20.6 shows an example of PPC observed in an n-type GaN film at 179 K [6] and at room temperature [2]. It should be noted that in Figure 20.6 the photocurrent is normalized by an initial photocurrent just after stopping the illumination; i.e. I p (t)/I p (0). The solid circles and triangles show the experimental results at 179 K and room temperature, respectively, while the solid lines show the fitting to SEF given by Ip (t) Ip (0)

= exp[−(t∕𝜏)𝛽 ],

(20.14)

where 𝜏 is the effective response time, and 𝛽 (0 < 𝛽 < 1) the dispersion parameter. Here, 𝛽 (= 0.35) is the same at both the temperatures, but 𝜏, on the other hand, depends on

617

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

1.0

0.8 Ip (t)/Ip (0)

618

0.6

0.4

0.2

0.0 0

5000

10000 Time (s)

15000

20000

Figure 20.6 Comparison of PPC decays with the stretched exponential function (SEF) at 179 K (solid circle) [6] and at room temperature (solid triangle) [2] in n-type GaN. Source: Reprinted from R.J. Freitas and K. Shimakawa, Philos. Mag. Lett. 97, 257 (2017).

the temperature being 12 000 s at 179 K and 800 s at room temperature. In amorphous chalcogenides, I p (t) is known to be given by the SEF in Eq. (20.14) [11, 12]. As stated above, the DX centers represent, in principle, donor-related localized states and hence are expected to exist in n-type semiconductors [20]. Let us discuss the correlation between YL and PPC observed in GaN. First, YL decreases with increasing Mg (p-type dopant) to a maximum, and then it disappears at further higher doping in p-type GaN, while PPC is still observed [2, 3, 26]. This implies that the centers responsible for PPC are not caused by the extrinsic dopants. This may indicate that PPC is related to the intrinsic structural defects [2]. However, the DX behavior can be expected to occur due to extrinsic doping also for acceptor dopants, which are called the AX centers [21]. There are many structural models proposed on the DX and AX centers in III-V semiconductors [3]. The microscopic origin of the DX center has already been discussed (see Figure. 20.5), and the main objective of the present chapter is to make it clear why PPC occurs if DX-like centers are present in III-V semiconductors. Hence, we will discuss the origin of the dispersive decay kinetics originating from the DX-like centers. The particular interest is why such a slow dispersive reaction dominates the dynamics of photoexcited carriers when DX-like centers are present. In this regard, let us discuss the role of a DX center in contributing to the photocurrent in III-V semiconductors. A CCD shown in Figure 20.7, which is similar to that shown in Figure 20.5, is useful to understand the dynamics of PPC [2, 3, 6]. Here, A and B represent the ground and the excited states, respectively; V is the height of the potential barrier that needs to be surmounted (∼0.2 eV) when the transition from B to A occurs; and U is the energy difference between the transit state O and the ground state A. As already stated in Section 20.2, a DX (AX) center acts as a shallow-donor (acceptor) impurity that relaxes away from its substitutional site (B) to a deep-level defect (A) [20]. It is well known that the charge carriers generated in state B can move to

20.4 Kinetics of PPC in III-V Semiconductors

Figure 20.7 Configuration coordinate diagram for a DX-like center. O

V U B A

the DX ground state (A) by surmounting the barrier V . In this process, the reaction time due to surmounting the barrier V becomes longer (hours) at low temperature, leading to the occurrence of PPC [2, 6]. However, such a long relaxation time cannot be explained quantitatively with the above model, and why PPC is given by the SEF is not clearly understood. As a large lattice relaxation is involved in the transition between B to A [3, 6, 27, 28], a multiphonon transition should be taken into consideration [8, 29, 30]. We will now discuss the kinetics of PPC related with the relaxation of DX (AX)-like centers, in which multiphonon nonradiative transitions are involved. For the transition reaction (B → A) by surmounting the potential barrier V, the reaction rate 𝜈 can be given as [31] ( ) ) ( U V exp − , (20.15) ν = 𝜈00 exp − kT ℏ𝜔p where 𝜈 00 ≈ 1 × 1012 s−1 is the characteristic frequency. The second exponential term represents multiphonon emission, where ℏ𝜔p is a single phonon energy (∼40 meV), which shows that the rate falls off exponentially with the energy U taken up by the lattice (the energy gap law) [21]. Let us, first, assume that the relaxation energy U for DX (AX) center is distributed in an exponential manner, as shown in Figure 20.8: ( ) U − Umin 1 exp − (20.16) , (U − Umin ≥ 0), P(U) = U0 U0 P (U)

U0 Umin U

Figure 20.8 Exponential distribution P(U) of the potential barrier U involved in the transition reaction.

619

620

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

where U 0 is the extent of broadening of the potential barrier height, and U min is the minimum energy taking up by the lattice. According to Eq. (20.15), the rate v should be distributed in the manner shown in Figure 20.8 [25, 31, 32]: ) ( U − Umin P(U) ℏ𝜔p 1 ℏ𝜔p P(ν) = exp − P(U) = = | d𝜈 | 𝜈 𝜈 U0 U0 | dU | | | ( ) ⎞ ⎛ V ν ( ) + ln ℏ𝜔 Umin ⎜ p kT ν00 ⎟ 1 ℏ𝜔p = exp exp ⎜ ⎟ 𝜈 U0 U0 U0 ⎟ ⎜ ⎠ ⎝ ) ( ( ( )) Umin 𝛽 V ν = exp exp 𝛽( + ln 𝜈 U kT ν00 ( 0) ( ) Umin 𝛽V = 𝛽 exp exp 𝜈00 −𝛽 𝜈 𝛽−1 U0 kT = C 𝜈 𝛽−1 , (20.17) where 𝛽 and C are given by 𝛽=

ℏ𝜔p U0

,

and

(20.18) (

C = 𝛽 exp

Umin U0

)

( exp

𝛽V kT

) 𝜈00 −𝛽 .

(20.19)

We now introduce the waiting-time distribution for the transition reaction shown in Figure 20.7. It should be emphasized again that the main objective of the present work is to understand how and why the time-dispersive reaction is involved in the PPC mechanisms. Using Eq. (20.17) in Eq. (20.12), the modified Poisson waiting-time probability distribution is obtained as ∞

Ψ(t) =

∫0



P(v)ve−𝜈t dv = C

∫0

νβ e−vt dv = C ′ t −β−1 ,

(20.20)

where 𝜈 is the average reaction rate (s−1 ) for the above transition, and C ′ is a constant (≠C). Using Eq. (20.20), the average waiting time for the reaction is obtained as t

⟨t⟩ =

∫0

t𝛹 (t)dt = C ′

t 1−𝛽 . 1−𝛽

(20.21)

The average reaction rate can be given as ⟨ν⟩ =

1 − 𝛽 𝛽−1 1 t . = ⟨t⟩ C′

(20.22)

20.4 Kinetics of PPC in III-V Semiconductors

Thus, under the assumption of an exponential distribution for the lattice relaxation energy for the DX-like centers given in Eq. (20.17), the average relaxation rate ⟨𝜈⟩ is given by the time-dependent power law as obtained in Eq. (20.22). In the DX (AX) model, the number of photocarriers is proportional to the number of charge carriers N B in the state B (shallow-donor state) [20]. Using Eq. (20.22), the rate equation for the time-dependent N B can be given as dNB 1−𝛽 = −⟨ν⟩ NB = − ′ t β−1 NB . dt C N B is then obtained by the solution of Eq. (20.23) as }𝛽 { [ ] ⎤ ⎡ 1−𝛽 𝛽 (1 − 𝛽)1∕𝛽 t ⎥ ⎢ NB = NB (0)exp − ′ t = NB (0)exp − 1 ⎥ ⎢ C𝛽 ′ 𝛽 (C ) ⎦ ⎣ [ ( )𝛽 ] t , ≡ NB (0)exp − 𝜏

(20.23)

(20.24)

where N B (0) is the initial value of N B just before stopping the illumination, and 𝜏 is called the effective response time and is given by ( 𝜏=

C′𝛽 1−𝛽

)1 𝛽

.

(20.25)

As the photocurrent I p (t) is proportional to N B (t), Eq. (20.24) gives the empirical Eq. (20.1). According to Eq. (20.25), we find that the reaction time can be written as ) 1∕𝛽 ( ( ) 𝛽V 1 V , = exp 𝜏 ∝ (C ′ ) ∕𝛽 ∝ (exp) kT kT

(20.26)

Accordingly, 𝜏 increases exponentially with the barrier height V at a temperature T. Using Eq. (20.26) for the experimental results shown in Figure 20.6, we get V = 0.11 eV. According to the present model, 𝛽 = ℏ𝜔p /U 0 , which is temperature independent. Using ℏ𝜔p = 40 meV, the extent of broadening of the potential barrier height U 0 is then estimated to be 0.11 eV. C is given by Eq. (20.19), where assuming 𝜈 00 = 1 × 1012 s−1 , we obtain C = 30 at 297 K. We then get U min = 1.4 eV, which may be a reasonable value for the DX center in GaN [20, 27]. The temperature-independent 𝛽 obtained above agrees with the experimentally reported temperature-independent 𝛽 = 0.35. Finally, we should discuss briefly the alternative possibility of an energy distribution P(U) for the DX (AX) center obtained from the Gaussian distribution as ( ) (U− < U >)2 1 P(U) = √ exp − , (20.27) 2U𝜎 2 2𝜋U𝜎 where U 𝜎 is the standard deviation (eV), and < U > is the average value.

621

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

103

102 p(𝜈) (arb. unit)

622

101

100

10–1 10–6

10–5

10–4

10–3

𝜈

10–2

10–1

100

(S–1)

Figure 20.9 Numerical calculation of P(𝜈) vs. 𝜈 as a function of temperature. Solid triangles and circles represents T = 200 and 300 K, respectively. Source: Reprinted from R.J. Freitas and K. Shimakawa, Philos. Mag. Lett. 97, 257 (2017).

The distribution function P(𝜈) is then given by [ ] P (U) ℏ𝜔p (U− < U >)2 1 ℏ𝜔p exp − P(𝜈) = U PU (U) = √ = | d𝜈 | 𝜈 ν 2𝜋U 2U𝜎 2 | dU | 𝜎 | | 2 ) ( ⎡ ⎡ ⎧ ⎫ ⎤ ⎤ V exp − 𝜈 ⎢ ⎢ ⎪ 00 ⎥ ⎥ kT ⎪ − < U >⎥ ⎥ ⎢ ⎢ℏ𝜔p ln ⎨ ⎬ 𝜈 ⎢ ⎢ ⎪ ⎪ ⎥ ⎥ ⎦ ⎥ ⎩ ⎭ ⎢ ⎣ 1 ℏ𝜔p exp ⎢− = √ (20.28) ⎥, 2 𝜈 2𝜋U 2U𝜎 ⎥ ⎢ 𝜎 ⎥ ⎢ ⎥ ⎢ ⎥ ⎢ ⎦ ⎣ ) ( ) ( 𝜈 V . where U = ℏ𝜔p ln 𝜈0 , and 𝜈0 = 𝜈00 exp − kT The 𝜈 dependence of P(𝜈) is not clear from the analytical form, but a numerical calculation of P(𝜈) can be done by using, for example, ℏ𝜔p = 0.05 eV, = 1.0 eV, U 𝜎 = 0.3 eV, V = 0.1 eV, and 𝜈 00 = 1 × 1012 s−1 . Using these in Eq. (20.28), we get P(v) as shown in Figure 20.9. For plotting Figure 20.9, the 𝜈 values selected for the calculation are in the range 10−6 ∼ 100 (s−1 ), which covers very well the measured time range. According to Figure 20.9, the power law is obtained as P(𝜈) ∼ 𝜈 β−1 ,

(20.29)

20.A On the Reaction Rate Under the Uniform Distribution

with 𝛽 = 0.24 at 200 K and 0.30 at 300 K. Thus, 𝛽 depends on the temperature, which is not consistent with the experimental results stated above. Therefore, it may be concluded that the Gaussian distribution is not applicable for U, and the exponential distribution gives better results.

20.5 Conclusions We have discussed the origin and kinetics of the PPC observed in GaN as a case example for III-V semiconductors. The principal results are as follows: (1) The DX(AX) center plays the principal role in th PPC. (2) A large lattice relaxation with the multiphonon emissions from the photoexcited DX-like state to the ground state dominates the reaction rate. (3) The energy U should be given by a distribution function due to, e.g. accumulation of lattice strains during crystalline growth in preparation. (4) An exponential distribution P(U) with a minimum U min produces the SEF parameters 𝛽 and relaxation time 𝜏 in agreement with the experimental results.

Acknowledgments The authors wish to thank Mr. A. Takahashi and Ms. M. Sakaki (Japan International Cooperation Agency: JICA) for encouragement and financial support, which enabled the authors to prepare the present chapter.

20.A On the Reaction Rate Under the Uniform Distribution The uniform distribution (the box approximation) has not been discussed in the chapter. The uniform distribution can be an extreme case of the exponential and Gaussian distributions. It is of interest to discuss what happens to the reaction rate under the uniform distribution. An exponential distribution function as given by Eq. (20.16) is ( ) U − Umin 1 exp − . P(U) = U0 U0 If U 0 ≫ U (𝛽 → 0) (see Figure 20.A.1), then 1 P(U) = . U0

(20.A.1)

This is called a uniform distribution, since P(U) is a constant (independent of U). For the Gaussian distribution shown already in Eq. (20.27), P(U) is given as } { (U − Um ) 2 1 P(U) = √ (20.A.2) exp − 2𝜎 2 2𝜋𝜎 √ As 2𝜋𝜎 corresponds to U 0 , when U 0 ≫ U (𝛽 → 0) (see Figure 20.A.2), then we get 1 PU (U) = (uniform distribution). U0

623

624

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

P (U)

U0 ≫ U U

Figure 20.A.1 Schematic drawing of the extreme case of the exponential distribution. P (U) U0 ≫ U

Figure 20.A.2 Schematic drawing of the extreme case of the Gaussian distribution.

U P (U)

Figure 20.A.3 Schematic drawing of a uniform distribution. 1

0 Umin

Umax U

The above two extreme cases can be equivalent to the box approximation (uniform distribution) as shown in Figure 20.A.3. It looks like a BOX. This is given by 1 1 ≡ , for Umin < U < Umax P(U) = Umax − Umin U0 =0 otherwise (20.A.3) where we denote here U 0 = U max − U min . In conclusion, If U 0 ≫ U in the exponential and Gaussian distributions (i.e. broad distribution), these can be approximated by the uniform distribution. Let us discuss how the reaction rate K(t) is modified under the box approximation. P(𝜈) under the box approximation (20.A.3) is given as [25] P(ν) =

P(U) ℏ𝜔p −1 ν . = | d𝜈 | U0 | dU | | |

(20.A.4)

Then, we get ∞

Ψ(t) = ⟨t⟩ =

∫0

tmax

∫tmin

ℏ𝜔p



ℏ𝜔p

t −1 , U0 ℏ𝜔p ℏ𝜔p tΨ(t)dt = (tmax − tmin ) ≅ t , U0 U0 max

P(v)ve−𝜈t dv =

U0 ∫0

e−vt dv =

where t max (min) is given by (see Eq. (20.15)) ( ) ( ) Umax(min) V −1 exp . tmax(min) = ν00 exp kT ℏ𝜔p

(20.A.5) (20.A.6)

(20.A.7)

References

⟨𝜈⟩ is then given as ⟨ν⟩ =

U 1 = 0 t , ⟨t⟩ ℏ𝜔p max

(20.A.8)

Thus, according to Eq. (20.A.8), the reaction rate < v > takes a constant value (not time-dependent) in the BOX approximation, while the reaction rate distributes broadly. Thus, for the monomolecular case, I p (t) can be given simply by an exponential function. When the time range we are concerned with is shorter than t max (t < t max ), the upper limit of the integral in Eq. (20.A.6) should be t. In this case, ⟨t⟩ is given by t

⟨t⟩ =

∫tmin

tΨ(t)dt =

ℏ𝜔p U0

(t − tmin ) ≅

ℏ𝜔p U0

t.

(20.A.9)

< 𝜈 > is therefore proportional to t −1 (time-dependent), which leads to different kinetic behavior, even under the box approximation.

References 1 Reddy, C.V., Balakrishna, K., Okumura, H., and Yoshida, S. (1998). Appl. Phys. Lett.

73: 244. 2 Johnson, C., Lin, J.Y., Jiang, H.X. et al. (1996). Appl. Phys. Lett. 68: 1808. 3 Chen, H.M., Chen, Y.F., Lee, M.C., and Feng, M.S. (1997). J. Appl. Phys. 82: 899. 4 Bube, R.H. (1992). Photoelectronic Properties of Semiconductors. Cambridge: Cam-

bridge University Press. 5 Beadie, G., Rabinovich, W.S., Wickenden, A.E. et al. (1997). Appl. Phys. Lett. 71:

1092. 6 Hirsch, M.T., Wolk, J.A., Walukiewicz, W., and Haller, E.E. (1997). Appl. Phys. Lett. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

71: 1098. Bonfigilio, A., Traetta, G., Lomascolo, M. et al. (2001). J. Appl. Phys. 89: 5782. Salis, M., Anedda, A., Quarati, F. et al. (2005). J. Appl. Phys. 97: 033709. Hou, Q., Wang, X., Xiao, H. et al. (2011). Appl. Phys. Lett. 98: 102104. Shimakawa, K. and Yano, Y. (1984). Appl. Phys. Lett. 45: 8. Shimakawa, K. (1985). J. Non-Cryst. Solids 77 & 78: 1253. Shimakawa, K. (1986). Phys. Rev. B 34: 8703. Nelson, R.J. (1977). Appl. Phys. Lett. 31: 351. Palmar, R.G., Stein, D.L., Abrahams, E., and Anderson, P.W. (1984). Phys. Rev. Lett. 53: 958. Langouche, G. (1989). Hyperfine Interact. 47: 85. Phillips, J.C. (1996). Rep. Prog. Phys. 59. Trachenko, K. and Dove, M. (2004). Phys. Rev. B 70: 132202. Kohlrausch, R. (1854). Pogg. Ann. Phys. Chem. 91: 56/179. Lang, D.V. and Logan, R.A. (1977). Phys. Rev. Lett. 39: 635. Alkauskas, A., McCluskey, M.D., and Van de Walle, C.G. (2016). J. Appl. Phys. 119: 181101. Itoh, N. and Stoneham, A.M. (2001). Material Modification by Electronic Excitation. Cambridge: Cambridge University Press. Vardeny, Z., O’Connor, P., Ray, S., and Tauc, J. (1980). Phys. Rev. Lett. 44: 1267.

625

626

20 Kinetics of the Persistent Photoconductivity in Crystalline III-V Semiconductors

23 Nakagawa, N., Shimakawa, K., Itoh, T., and Ikeda, Y. (2010). Phys. Status Solidi C 7

(3–4): 857. 24 Singh, J. and Shimakawa, K. (2003). Advances in Amorphous Semiconductors. London

and New York: Taylor & Francis. 25 Papaulis, A. (1965). Probability, Random Variables and Stochastic Processes. Tokyo:

McGraw- Hill Kogakusha Ltd. 26 Li, Z., Lin, J.Y., Jiang, H.X. et al. (1996). Appl. Phys. Lett. 69: 1474. 27 Mattila, T., Seitsonen, A.P., and Nieminen, R.M. (1996). Phys. Rev. B 54: 1474. 28 Tanaka, K. and Shimakawa, K. (2011). Amorphous Chalcogenide Semiconductors and

Related Materials. New York: Springer. 29 Rose, A. (1951). RCA Rev. 12: 362. 30 Andreev, A.A., Zherzdev, A.V., Kosarev, A.I. et al. (1984). Solid State Commun. 52:

589. 31 Freitas, R.J. and Shimakawa, K. (2017). Philos. Mag. Lett. 97: 257. 32 Freitas, R.J., Shimakawa, K., and Wagner, T. (2014). J. Appl. Phys. 115: 013704.

627

Index a a-As2 S3 optical absorption spectrum 69 photodarkening 204 Abbe number 88–89 absorption band-to-band (fundamental) 45–48, 68 cross-section 48, 103 by 3d metal ions 99–102 due to free carriers 43 nonlinear 272–280 p-type GaAs 45 types of 45 vibronic potential-energy curves 302 absorption coefficient calculation of 68 deviations from Tauc’s relation 75 as function of photon energy 38 Tauc’s relation for 71 theory 74–79 ABX3 structure 539 acceptors 160 acoustic phonon scattering 43 acousto-optic modulator (AOM) 178 activators 160 add/drop multiplexing devices 263–264 Ag20 As32 Se48 , linear and nonlinear refractivity 279 AgCl, dispersion relations parameters 9 a-Ge:H effective mass of electrons in extended and tail states 77 QFRS spectra 178 a-Gex Si1-x :H, optical gap 80 AlAs, optical and structural properties 39

Al2 O3 , Wemple–DiDomenico dispersion relationship parameter 15 Al-PET foil 350 AlP, optical and structural properties 39 amorphous chalcogenides 69–72 chemically ordered bond network (CON) type 80 compositional variation of the optical bandgap 79–80 indefinite network type 80 photodarkening 204 photoinduced changes in 225–232 random bond network (RBN) type 80 types of 80 volume expansion (VE) 201 amorphous semiconductors absorption coefficient calculation 68 defects in 67 excitons in 135–139 photoluminescence lifetime spectroscopy by QFRS 175–192 spectral dependence of optical absorption coefficient 79 Tauc’s relation for absorption coefficient 75 Wannier–Mott excitons 135 amorphous solids effective mass of charge carriers in 76 excited charge carriers 206–207 excitonic absorption in 137–139 rates of spontaneous emission in 211–216 singlet excitonic states in 137 triplet excitonic states in 137 Urbach’s tail 79

Optical Properties of Materials and Their Applications, Second Edition. Edited by Jai Singh. © 2020 John Wiley & Sons Ltd. Published 2020 by John Wiley & Sons Ltd.

628

Index

angular-dependent color stability angular radiance 395 CIE coordinates 396–400 color rendering index 397 current efficiency 396–397, 399 normalized EL spectra 395–396 perfect Lambertian emitter 395 power efficiency 396–397, 399 resonant wavelength 398–399 anodes modification, for enhanced OLED performance compositional analysis 336–339 electrical properties 331–333 electroluminescence performance of J–V characteristics 340–341 experimental methods 328–329 low-temperature high-performance ITO 327–328 morphological properties 329–331 optical properties 333–336 anthracene 317 a-Se 71 optical absorption spectrum before and after annealing 69 transient photodarkening 204 a-Si:H broad PL lifetime distribution 178–189 double-and triple-peak structures in PL 203 monochromatized PL QFRS spectra 182 optical and structural properties 39 photoluminescence 175 PL decay 189–192 QFRS spectra 175 QFRS spectra at various PL excitation energy 183 radiative lifetime, by TRS and QFRS 203 spontaneous emission and radiative lifetime 221 SWE in 204 a-Si1-x Cx :H, optical bandgap obtained from Tauc’s, Sokolov and Cody’s relations 79 As2 S3 bandgap illumination 282

bandgap reduction 231 linear and nonlinear refractivity 279 photodarkening 284 photo-induced changes 282 spectral dependence of the one-and two-photon absorption coefficients 276 two-photon absorption spectra 279 As2 S(Se)3 binary systems 71 As2 Te3 , bandgap reduction 231 athermal processes 281 atomic force microscopy (AFM) 456, 459–460 Auger processes 568

b BaFBr 121 bandgap compositional variation in amorphous chalcogenides 79–80 and temperature dependence 11 bandgap energy 38, 39 bandgap excitations, of one-photon and resonant two-photon 285 bandgap narrowing/reduction 47, 231 bandgap renormalization (BGR) 594, 595 band-pass filters (BDFs) 165 bandtailing 47 band-to-band (fundamental) absorption crystalline solids 45–48 of photons 45 barium strontium titanate (BST) 257 Beilby overlayer 446 beta-factor 264 bipolarons 232, 299 birefringence 91 Bloch function 510 Bolings’ relation 270 bond arrangements, for polythiophene 305 bottom-emitting OLED 348 Bragg gratings 107, 252 Brewster angle 88, 444 Brillouin function 595 Bruggeman approximation 457 Burstein–Moss shift 47

Index

butylammonium (BA) 553 BX6 octahedron 538

c cadmium zinc telluride (CZT) 571–572 carrier-spin tunneling 593 Cauchy’s dispersion parameters 8 Cauchy’s dispersion relation 7–8 Cauchy’s formula 7 cavity enhancement factor 264 cavity quantum electro dynamics (c-QED) 530–531 center-of-mass (COM) 492–493 chalcogenide glasses (ChGs) 283–285 films of 241 nonlinear excitation-induced structural changes 283–285 photo induced volume changes (see photo induced volume changes) properties 241 chalcogenides, amorphous 69–72 channel waveguides 261 charge carriers, in amorphous solids 76 charge carriers, pairing of like excited 229–232 chemically ordered bond network (CON) type, amorphous chalcogenides 80 chemical perturbations 94–95 chromogenic coatings 413 Clausius–Mossotti relation 94, 457 complex refractive index (n*) 3, 40 complex relative permittivity 40 Compton scattering 569–570 condensed matter, optical properties of disordered 67–80 configuration coordinate diagram (CCD) 614 configuration coordinate model 101 confined Stark effect 55 constant-fraction discriminator (CFD) 171 continuous wave (CW) 526 copper indium gallium selenide (CIGS) 544

copper zinc tin sulfide (CZTS) 544 Coulomb interaction 508, 510 coupled resonant optical waveguide (CROW) structures 261 Co wire system 605 critical angle 88 crown glass, Verdet constant 57 crystalline III-V semiconductors 613–615 CCD 614 distribution function 622–623 DX center 613–615, 618–619 energy distribution 621 exponential distribution 619, 620 multiphonon emission 619 n-type GaN film 617, 618 p-type GaN 618 time-dependent N B 621 waiting-time distribution 620 crystalline semiconductors, refractive index 10–11 crystalline solids band-to-band or fundamental absorption 45–48 excitonic absorption in 133–135 excitons in 130–135 crystals, ligand field 97 CsCl, Wemple–DiDomenico dispersion relationship parameter examples 15 cubic energy dependence 72 Curie–Weiss law 92 current density–voltage (J–V) characteristics 385 EL spectra 378, 380 emission characteristics 383 energy levels 379 hole-electron current balance 377 HTL/ETL layer 383–384, luminance function 379–380 MoO3 HIL 377–378 power efficiency 377–380 recombination zone 382–383 TCTA 380–384 TmPyPB interface 381–384 triplet-triplet energy transfer 379

629

630

Index

d DAP recombination 160 PL lifetime 161 in semiconductors, schematic optical transition models 160 Davydov splitting 317 dc/rf magnetron sputtering 327 defects, in amorphous semiconductors 67 defect tolerance 548–549 degenerate indirect bandgap semiconductors, absorption coefficient 47 delayed fluorescence 235 dense wavelength-division multiplexing (DWDM) systems 262 density functional theory (DFT) 424, 562 density-of-states (DOS) 421, 424 amorphous chalcogenides 68 for extended states 68 in Euclid space 71 in fractal space 71 of fractals 72 diamagnetic glasses 94 diamond Cauchy’s dispersion parameters 8 optical and structural properties 39 Sellmeier coefficients 9 dichromatic WOLED. see transparent white organic light-emitting diodes (WOLEDs) diffused reflection 27 diluted magnetic semiconductors (DMSs) electron-beam lithography 596–599 hybrid nanostructures with ferromagnetic materials 604–607 nanoscale dot and wire structures 590 nanostructures 590 properties 589–590 quantum wells pump-probe spectroscopy 594–596 spin injection 591–593 SAQDs Brillouin function 602 micro-PL spectra 600 PLE study 604 P value 604 spin injection 601–603

type-II transition 602 Zeeman shift 600–601 spin-splitting energy 589 dimethyl sulfoxide (DMSO) 554 dipole–dipole energy transfer 568 direct-gap semiconductor 274 direct transitions 45, 159 disordered condensed matter, optical properties 67–80 dispersion 2, 88–90 in insulators 4 refractive index and 7–16 dispersion wavelength, zero-material 15 dispersive devices 262 dispersive reaction 615–616 distant DAP recombination 160 donors 160 double perovskite structures 540, 541 double quantum dot (DQD) system 598, 599 double quantum well (DQW) systems pump-probe spectroscopy 594–596 spin injection 591–593 Drude theory for free charge carriers 335 dual-phase double lock-in (DPDL) QFRS 176–178 block diagram 176–178 experimental set-up 177 DX (AX) model 621 dynamic Franz–Keldysh effect (DFKE) 55

e E0 (optical gap) 218 effective mass of charge carriers in amorphous solids 76 for electrons and holes 77 of electrons in extended and tail states of a-Si:H and a-Ge:H 77 effective medium approximation (EMA) 73, 456–457 elastic-ether theory, of the refractive index 7 electric dipole oscillator strength, for absorption or emission 102 electric permittivity 457

Index

electric susceptibilities, roles in frequency scale upon excitation 273 electro-absorption 55–56, 312 electroabsorption modulator (EAM) 55–56 electro-absorption spectra three-level model 313 three-level model with asymmetric inhomogeneous broadening and vibronic structure 313 two-level model 313 electroluminescence 157, 234 electroluminescence performance, OLEDs 340–345 electroluminescence (EL) spectra angular-dependent color stability 395–396 J–V characteristics 378, 380 L–J characteristics 391, 393, 396 optimization 376–377 electron affinity (EA) 580, 581 electron-beam lithography 596–599 electron–electron interaction, 𝜋-conjugated polymers 309–314 electron-hole current balance 384–385, 386 electronic configurations, for 3d metal ions 99 electron injection layer (EIL) 375 electron–phonon coupling 101 electron transporting layer (ETL) 383–384, electro-optic modulator (EOM) 176 ellipsometery 2 energy level diagram of low-lying 4fN states of trivalent ions doped in LaCl3 50 for Pr3+ ions 98 energy levels, for d7 configuration 100 epitaxial films 421–422 epitaxial growth techniques in nanowires 507–508 natural/interface quantum dots 506–507 self-assembled QDs 504–505 site-controlled growth 505–506

even functions (states) 51 excitation, of single electron–hole pair 228–229 excited charge carriers, pairing of like 229–232 exciton absorption peaks 38 exciton binding energy 132, 301 excitonic absorption in amorphous solids 137–139 in crystalline solids 133–135 excitonic bands 131 excitonic Bohr radius 137 exciton radius, dependence on bandgap of semiconductors 132 excitons 129–153 in amorphous semiconductors 135–139 concept of 223–225 in crystalline solids 130–135 in organic semiconductors exciton dissociation 148–153 exciton up-conversion 147–148 photoexcitation and formation 140–146 radiative lifetime 225 excitons, quantum dots (QDs) bandgap 509–510 Coulomb interaction 508 optical transitions 510–513 Schrödinger equation 508 external fields electro-absorption and Franz–Keldysh effect 55–56 electro-optic effects 54–55 Faraday effect 56–57 external quantum efficiency (EQE) 393, 394 extinction coefficient K of AlSb 42 experimental techniques for measuring 2 and Kramers–Kronig relations 5–6 spectral dependence for typical glasses 105 values of quantities required for 42 extrinsic photoluminescence 160–162

631

632

Index

f fac-tris(2-phenlypyridine)iridium [Ir(ppy)3 ] 159 Faraday effect 56–57, 92–94 Faraday, Michael 56, 92 ferroelectric transparent glass ceramics 121 ferromagnetic materials 604–607 fiber optics 104–106 films, uniform-thickness 16–21 finite difference time domain (FDTD) 459 flexible OLEDs on plastic foils 347–353 on ultra-thin glass substrates 348–349 flint glass, Verdet constant 57 fluorescence 158 from Pr3+ ions in ZBLA glass 103 in rare-earth-doped glass 102–104 fluorescent-line narrowing (FLN) 102 fluorophenylmethylammonium (FPMA) 553 fluorozirconate glass, Verdet constants for undoped and rare-earth-doped 93 Forouhi–Bloomer (FB) coefficients 10 Fourier transform infrared (FTIR) spectrometry 166 Fourier transform photoluminescence (FTPL) 166 Franck–Condon factor 314 Franz–Keldysh effect 55–56 free-carrier absorption 42–44 due to holes in p-Ge 44 in n-GaP, p-PbTe and n-ZnO 44 free charge carriers, Drude theory 335 free excitons 131 free exciton/Wannier excitons, photoluminescence from 159 Frenkel excitons 129, 131 frequency-resolved spectroscopy (FRS) 172–173 Fresnel reflection coefficients 405 f -sum rules 6 full-energy absorption peak 564 full width at half maximum (FWHM) 605–606 fundamental absorption crystalline solids 45–48

experimental 69–74 of photons 38

g GaAs exciton binding energy 131 Forouhi–Bloomer (FB) coefficients 12 GaAs multiple quantum wells (MQW), saturation intensity 134 optical and structural properties 39 quantum wells low-temperature absorption 134 photoluminescence spectra 134 Wemple–DiDomenico dispersion relationship parameter examples 15 GaAs0.88 Sb0.12 , optical and structural properties 39 gallium nitride (GaN) 611 GaN, optical and structural properties 39 GaP Forouhi-Bloomer (FB) coefficients 12 optical and structural properties 39 PL spectrum 161 GaSb 39 Ge Cauchy’s dispersion parameters 8 Forouhi–Bloomer (FB) coefficients 8, 12 free-carrier absorption in p-type 44 heavy-hole band 44 light-hole band 44 optical and structural properties 39 spin-off band 43 valence band of 44 geminate pairs 205 type I 205 type II 205 geminate recombination 175 general ellipsometry 441 GeS2 bandgap reduction 231 linear and nonlinear refractivity 271 Gladstone–Dale coefficient 12–13 Gladstone–Dale formula, and oxide glasses 12–13 glass ceramics, transparent 114–124

Index

ferro electric transparent glass ceramics 121 photonic applications 115 rare-earth-doped transparent glass ceramics for active photonics 120–121 theoretical basis 116–120 X-ray storage phosphors 121–124 glass color 95–102 coloration by colloidal metals and semiconductors 95–96 rare-earth-doped glass 96–99 3D metal ions, absorption by 99–102 glasses optical properties 83–124 Verdet constants 57 glass interfaces 86–88 reflection coefficient at normal incidence vs wavelength for air/silica glass interface 86 GRadient INdex (GRIN) fibres 107 grating monochromator 164 group index 15–16

h Hagen–Rubens relationship 43 Hervé–Vandamme relationship 11 highest occupied molecular orbital (HOMO) 244 highest occupied molecular orbital (HOMO) level 375 high-purity germanium (HPGe) 571 hole injection layer (HIL) 375 hole-transporting layer (HTL) 383–384 Holstein’sapproach 226 Hong–Ou–Mandel (HOM) experiment 529, 530 Hopper theory 117 hot-injection method 550 Huang–Rhys parameters 303, 304 hybrid organic–inorganic perovskites (HOIP) 541 hydrogenated amorphous silicon (a-Si:H) distribution of electronic energy states 68 distributions of conduction and valence-bands electronic states 77

effective mass of electrons in extended and tail states 77 optical absorption coefficient 73 hydrogenated nanocrystalline silicon (nc-Si:H) 72–74 hydrogen-collision model 204

i illumination, effects of 244–245 impurity absorption 48–54 impurity absorption band 48–54 impurity scattering 43 InAs FB coeficients 12 Forouhi–Bloomer (FB) coefficients 12 optical and structural properties 39 indefinite network type, amorphous chalcogenides 80 indirect transition 159 indium tin oxide (ITO) films 325, 554 anodes 339–340 compositional analysis 336–339 conductivity of oxygen-deficient 343 electrical properties 331–333 experimental methods 328–329 function of hydrogen partial pressure 332 grown on glass substrates, AFM images 330 morphological properties 329–331 optical properties 333–336 oxygen plasma-induced low-conductivity layer thickness and ΔR 343 resistivity, hydrogen partial pressure 331 SIMS depth profile 337, 338 square of absorption coefficient 335 thickness 328 wavelength-dependent transmittance 334 infrared reflection 40–42 infrared refractive index (n) 7 of AlSb 41 values of quantities required for 42 infrared (IR) transmittance 414–415 InGaAs on InP, optical and structural properties 39

633

634

Index

InGaAsP/InP photocathodes 166 InP Forouhi–Bloomer (FB) coefficients 12 optical and structural properties 39 InSb Forouhi–Bloomer (FB) coefficients 12 optical and structural properties 39 reflectance of doped 42 insulators, dispersion in 3 interband processes conduction-and valence-bands discontinuities 482 discrete Rydberg series 479–480 excitonic relaxation 480 Fano profile 480 interaction Hamiltonian 481 parity-allowed forbidden transitions 479 parity-forbidden forbidden transitions 479 relaxation times 483–485 sub-bands 479, 481–482 transitions 485 interchain interaction, 𝜋-conjugated polymers 314–320 intermolecular interaction 317 internal quantum efficiency 514–516 internal quantum efficiency (IQE) 234 International Commission on Illumination (CIE) coordinates angular-dependent color stability 394–400 L–J characteristics 390–393, 396 intraband processes LA phonon emission 476 l-mode phonon 475 LO phonon emission 477–479 TA phonon emission 476–477 total phonon emission 475–476 variational wave function 475 intrinsic photoluminescence 159–160 ionic crystals, dispersion relations 9 ionization potential (IP) 580, 581 ionizing radiation detectors atomic structure and charge density plot 581, 582 Auger processes 568

band structures 580 bimolecular rate equation 579 bulk crystalline 572, 574 Compton scattering 565, 566, 569–570 crystal growth method 576 Cs4 PbBr6 /CsPbBr3 nanocomposite 577 CZT 571–572 defect tolerance 562–563, 571 electronic structure 581–582 embedded nanocrystal composite 579–580 full-energy absorption peak 564 functionalities 563 high-energy radiation detection 565, 566 histogram 564 ionization potential and electron affinity 580, 581 linear energy deposition 565, 567 linear energy deposition curves 565, 567 nonproportional response 564 optical response and kinetics 577 particle-type discrimination 565 photoconducting 571 photoconductive detectors 563 photoelectric effect 567, 568 photoelectron energy 568–569 photoluminescence 563 photon counting (dosimetry) function 563 photon energy down-converters 563 scintillating radiation detectors 571, 574, 575 scintillation 563 self-absorption 582 semiconducting detectors 572, 573 spectroscopic detection 564 steady-state optical excitation 577–578 TOF-PET 565 Type I heterostructure 581 x-ray 570–571 IR bands, frequency of 105

j Jahn–Teller distortions 541 Judd–Ofelt analysis 51

Index

Judd–Ofelt parameters 103 for rare-earth ions in heavy-metal fluoride glasses 103

k Kasha’s rule 311 Kerr coefficient 54 Kerr effect 54 Kramers–Kronig relations 38 extinction coefficient K and 5–6 refractive index n and 6 relations between real and imaginary parts 6

l large-radii orbital excitons 129 lasers 264 35La2 S3 .65Ga2 S3 , linear and nonlinear refractivity 278 lattice (Reststrahlen) absorption 40–42 leading-edge discriminator (LED) 171 lead lanthanum zirconate titanate (PLZT) 256 ligand field parameters 97 light-emitting diodes (LEDs) 264 butylammonium 552–553 CsPbBr3 nano-inclusions, Cs4 PbBr6 absorption strength 554 Auger recombination 559 biexcitons and charged excitons 559 Cs-Pb-Br thin films 554–556 current density, current efficiency, and external quantum efficiency 559, 560 current-versus-voltage characteristics 554 device structure 559 DFT 562 DMSO 555, 557, 561 indium tin oxide 554 luminance–voltage characteristics 561 morphology 557, 558 quantum yield 555, 557 TEM images 559, 560 XRD 554–555

dimension-dependent properties 551 hot-injection method 550 nanometer-sized 3D crystallites/quasi-2D phases 551–552 PeLED 549–550 perovskite films preparation 553 photoluminescence spectra 550–551 PLQY 551 polymer-perovskite nanocomposite films 553 3D and low-dimensional lead halide 551, 552 light-induced defects 232 light-induced electron-spin resonance (LESR) intensity 175 light-induced metastable defects (LIMD) 204 LiNbO3 , Wemple–DiDomenico dispersion relationship parameter examples 15 linear grating advantages 462 characterization 458 diffraction 459 disadvantages 462 discrepancy 462 ellipsometric parameters 460–462 holographic lithography 459–460 reflection coefficients 459 lithium-fluoride (LiF) 9, 342 lock-in amplifier 176, 178 longitudinal acoustic (LA) phonon emission 476 longitudinal optical (LO) phonon emission 473–474, 477–479, 593 long-pass filter (LPF) 164 long-wavelength lasers 526–527 Lorentz–Lorenz formula 457 Lorentz–Lorenz relationship 94 lowest unoccupied molecular orbital (LUMO) 244, 375 low-temperature absorption, GaAs quantum wells 134 𝜆–PADMR spectra, of disordered and ordered polythiophene films 307

635

636

Index

luminance–current density (L–J) characteristics CIE coordinates 391–393, 396 emission spectra 391, 393, 396 EQE 393, 394 power efficiency 392–393 resonant mode 390 symmetrical and bidirectional emission 391 weak microcavity 390 luminescence 157 Lyddane–Sachs–Teller relation 40

m Magnetic field dependence 92–94 magnetic quantum well (MW) 591, 592 matrix methods 406–407 Maxwell–Boltzmann distribution function 214 Maxwell’s formula for refractive index 2 metal–ligand complexes (MLCs) 159 metals, free-electron reflectance 42 MgO, dispersion relations parameters 9 micromachining technique 108 microscopic model 615 mid-gap absorption 285 Mie scattering theory 95, 115 mixed-valent perovskites 540 molar refractivity 94 molecular excitons 129, 139 molecular layer (ML) 493 molecular structure, of 𝜋-conjugated polymers 296–298 monochromatized PL, QFRS spectra of a-Si:H 182 monochromator (MNC) 164 Moss–Burstein effect 335 multi-channel analyser (MCA) 165 multiple quantum wells 56 multiple quantum wells (MQWs) 594

n NaCl Verdet constant 57 Wemple–DiDomenico dispersion relationship parameter examples 15

nanocrystalline diamond (NCD) film 454–458 nano-diamond (ND) particles 455 nano lasers 527, 528 nanostructures fabrication of 596–599 with ferromagnetic materials 604–607 15Na2 O⋅85GeO2 , linear and nonlinear refractivity 278 National Renewable Energy Laboratory (NREL) 544, 545 nc-Si:H 72–74 near-infrared (NIR) transmittance 414–415 negative-U effect 226 neutral density filter (NDF) 164 noncrystalline condensed matter 203–236 nonemissive excited states, 𝜋-conjugated polymers 306–309 nongeminate e–h recombination 175 nongeminate recombination 175 nonlinear absorption 272–280 nonlinear optical properties, of photonic glasses 269–288 nonlinear refractivity 278–280 nonmagnetic well (NW) 591, 592 non-metal-insulator transition (MIT) 417 nonmetallic crystalline solids, bandgap reduction 226 nonradiative relaxation 474 interband processes conduction-and valence-bands discontinuities 482 discrete Rydberg series 479–480 excitonic relaxation 480 Fano profile 480 interaction Hamiltonian 481 parity-allowed forbidden transitions 479 parity-forbidden forbidden transitions 479 relaxation times 483–485 sub-bands 479, 481–482 transitions 485 intraband processes LA phonon emission 476 l-mode phonon 475

Index

LO phonon emission 477–479 TA phonon emission 476–477 total phonon emission 475–476 variational wave function 475

o odd functions (states) 51 one-photon absorption 286 one-photon excitation 285 optical absorption fundamental 69–74 in rare-earth-doped glass 96–99 of trivalent rare earth ions 48–54 optical absorption coefficients 69 a-Si:H 73 c-Si 73 nc-Si:H 73 spectral dependence in amorphous semiconductors 79 optical absorption cross-section 103 optical absorption spectrum of Co2+ in ZBLA glass 101 of obliquely deposited a-As2 S3 70 of obliquely deposited a-Se 71 of Pr3+ ions in ZBLAN glass 99 rare-earth ions 99 of a spin-coated film of polythiophene 301 optical bandgap of a-Si1-x Cx :H, from Tauc’s, Sokolov and Cody’s relations 79 compositional variation in amorphous chalcogenides 79 optical band-pass filter (BPF) 165 optical constants 1–7 extinction coefficient 2–5 Kramers-Kronig relations 5–7 refractive index 2–5 optical excitations, nonlinear 280 optical fibres 274 optical gap, of a-Gex Si1-x :H 80 optical glasses, Abbe diagram for 89 optical phonon scattering 40 optical properties of disordered condensed matter 67–80 of glasses 83–124

optical transitions 510–513 organic light-emitting devices (OLED) 234–236, 375–377 anode modification 339–340 architecture and operation principle 325–326 bottom-emitting OLED 348 electroluminescence performance 341–345 energy-level diagram 325 flexible displays on ultra-thin glass substrates 348–349 flexible OLEDS ultrathin glass substrate 346–347 flexible top-emitting OLEDs on plastic foils 347–353 with hybrid MoO3 -PEDOT:PSS hole injection layer (HIL) morphological properties of 361–363 performance of 353–361 surface electronic properties of 363–368 luminance–voltage characteristics 325 top-emitting OLEDs 348–353 organic materials, radiative recombination processes 159 organic semiconductors and applications 295 in excitons exciton dissociation 148–153 exciton up-conversion 147–148 process integration 325–327 radiative recombination of excitons 232–236 technical challenges 325–327 photoexcitation and formation singlet excitons, exciton–photon interaction 141–142 triplet excitons 142–146 oscillator strength 51, 103, 514–516 oxide glasses 281–283 Gladstone–Dale formula and 12 linear and nonlinear refractivity 278 oxyfluorides, photonics application studies 116 oxygen plasma treatment 339, 340

637

638

Index

p paramagnetic glasses 94 parity 48 parity-allowed forbidden transitions 479 parity-forbidden forbidden transitions 479 parity selection rule 48 partial dispersion 89 particle-type discrimination 565 PbO-SiO2 277 electronic structure 277 linear and nonlinear refractivity 278 PbTe, free-carrier absorption coefficient in p-type 43 𝜋-conjugated polymers absorption spectrum 300–304 backbone structures 296 charged excited states in 307 with degenerate ground states 299 disordered films of 305 electron–electron interaction 309–314 interchain interaction 314–320 molecular structure of 296–298 with nondegenerate ground states 299 nonemissive excited states 306–309 photoluminescence 304–306 theoretical models 298–300 PC refractive index tuning 256–257 free carriers in semiconductor-based PCs 257 of infiltrated PCs 257–258 using applied electric field 256–257 using light 256 PEDOT (poly(3,4-ethylenedioxythiophene)) 326 perovskites lead halide perovskite materials (see ionizing radiation detectors) multiplicity of hybrids 539–540 photovoltaics ammonium bromide 547 defect tolerance 548–549 mesoscopic and planar structures 544, 546 NREL 544, 545 power conversion 544

Ruddlesden–Popper layered perovskites 547 solution-processed deposition methods 544, 546 2D perovskites 547 simple cubic frameworks 538–539 solution-processed lead halide perovskites (see light-emitting diodes (LEDs)) structural variation AMnO3 structure 541 anionic inorganic substructure 541–542 antiferrimagnetic compounds 543 A-site organics 540–541 carbon nanotubes, graphene 2D chalcogenides, 543–544 double perovskite structures 543 immense diversity 543 3D ABX3 structure 541 “tuning knobs,” 537–538 persistent photoconductivity (PPC) crystalline III-V semiconductors CCD 614 distribution function 622–623 DX center 613–615, 618–619 energy distribution 621 exponential distribution 619, 620 multiphonon emission 619 n-type GaN film 617, 618 p-type GaN 618 time-dependent N B 621 waiting-time distribution 620 dispersive reaction 615–616 GaN 611 photoexcitation energy 611, 613 power law 616 SEF 612, 616 waiting time distribution 617 p-Ge, free-carrier absorption due to holes in 43 phase-modulated fluorometry (PMF) 173 phenethylammonium (PEA) 553 phenyl alkoxyphenyl PPV copolymer-based OLEDs 346 phonon absorption 46 phonon emission 316

Index

phonon interaction 226–228 phosphate glass 93 phosphorescence 158 phosphors 160 photocarrier density 168 photocarrier lifetime 168 photocatalytic microbiocidal effects 420 photochromic glasses 115 photoconductive detectors 563 photoconductor 166 photodarkening 226 metastable 232 transient 232 photodiode 166 photoelectric effect 567, 568 photoellipsometry ambient medium 408 iterative procedure 412 Jones/Stokes–Mueller formalisms 409 pseudointerfaces 410–411 quantities 411 semitransparent substrates 412 series developments 411–412 signal measurement 409 total reflection and transmission amplitudes 409–410 photoexcitation 226–228 photoinduced absorption (PA) measurements 307 PA spectra, of disordered and ordered polythiophene films 307–308 photoinduced changes in amorphous chalcogenides 225–232 and photoluminescence in noncrystalline condensed matter 203–236 photoinduced phenomena 270, 282 photo induced volume changes in amorphous chalcogenides 242 atomic-scale computer simulations 243–244 effect of illumination 244–245 As2 Se3 246–248 photoinduced volume expansion (PVE) 225 photolithography 107

photoluminescence (PL) 157–197, 205–225, 374–375, 513–514, 550–551 detection 173 experimental aspects frequency-resolved spectroscopy (FRS) 172–173 PLE and PLAS 167–168 quadrature frequency resolved spectroscopy (QFRS) 173–175 time-correlated single photon counting (TCSPC) 168–171 time resolved spectroscopy (TRS) 168–169 extrinsic 160–162 from free Wannier excitons 159 fundamental aspects 158–164 GaAs quantum wells 134 intrinsic 159–160 𝜋-conjugated polymers 304–306 and photoinduced changes in noncrystalline condensed matter 203–236 PL decay, for a-Si:H 189–192 PL lifetime Spectroscopy of a-Si:H, QFRS technique dual-phase double lock-in (DPDL) technique 176–178 electric and magnetic fields 185–189 excitation intensity, excitation, and emission energies 179–183 residual decay 189–192 temperature dependence 184–185 PL signal 491 QFRS on UCPL, of RE-doped materials 192–197 second-harmonic generation (SHG) 163 static PL spectroscopy 164–167 Stokes shift 158 temperature dependence 222–223 two photon absorption (TPA) 163 up-conversion photoluminescence 162–163 vibronic potential-energy curves 302 photoluminescence absorption spectroscopy (PLAS) 167–168

639

640

Index

photoluminescence excitation (PLE) 167–168, 593, 604 photoluminescence lifetime spectroscopy, of amorphous semiconductors by QFRS technique 175–192 photoluminescence quantum yield (PLQY) 551 photoluminescence recombination, nonlinear effects in 171 photomultiplier tube (PMT) 166 analogue and digital modes 166 photon-counting method 166 photon energy absorption coefficient as function of 38 absorption in crystalline solids 78 relationship between absorption coefficient and 38, 45 photon energy down-converters 563 photonic bandgap (PBG) 251 complete 253 incomplete 253 in 1D PC 252 2D waveguide low-loss 90∘ bending of light 253 photonic crystals (PCs) applications 251–252 applications for LEDs and lasers 263 1D 252 introduction 252–253 materials selection for 255 nano-engineering of PC architectures 253–255 overview 252–253 photonic devices 251 properties and applications of 251–265 selected applications 260–265 3D 252 2D 252 photonic crystal waveguide 531–533 photonic glasses, nonlinear optical properties 269–288 nonlinear absorption fundamentals 271–275 nonlinear refractivity 278–280 two-photon absorption 275–277 nonlinear excitation-induced structural changes

chalcogenides 283–285 fundamentals 280–281 oxides 281–283 photovoltaics (PV) ammonium bromide 547 defect tolerance 548–549 mesoscopic and planar structures 544, 546 NREL 544, 545 power conversion 544 Ruddlesden–Popper layered perovskites 547 solution-processed deposition methods 544, 546 2D perovskites 547 physical vapor deposition (PVD) 242 planar lightwave circuits (PLCs) 251 planar photonic crystals (PCs) 261 plasma resonance 96 plasmons, surface 264 PLC (planar lightwave circuits) 251 PL excitation spectra (PLE) for 3 QFRS QFRS (quadrature frequency-resolved spectroscopy) components 183 for three QFRS components 183 Pockels coefficient 55 Pockels effect 54 Poisson distribution 617 polarons 226, 307 energy structure 308 poly(3,4-ethylenedioxythiophene) (PEDOT) 326 poly(p-phenylenevinylene) (PPV) 304, 326 poly(styrenesulfonate) (PSS) 326 polyacetylene 297 polycrystalline film 421–422 polydiacetylene 311, 312 polyfluorene (PFO) 298, 300 polymer light-emitting devices 324 polythiophene 298 absorption spectrum of spin-coated film of 301 bond arrangements in the ground state 305 bond arrangements in the self-trapped excited state 305

Index

𝜋-orbital 297 power conversion efficiency (PCE) 544 power law 616 principal angle 444 Pr3+ ions energy level diagram for 98 in ZBLA glass, fluorescence of 103 prompt fluorescence 233 pseudo-optical constants 446 p-type GaAs, absorption behavior 45 pump-probe spectroscopy 594–596 Purcell effect 530 Purcell factor 264 pyramidal quantum dots 505

q quadrature frequency-resolved pectroscopy (QFRS) 173–175 photoluminescence lifetime spectroscopy of amorphous semiconductors by 175–192 PL spectra of a-Ge:H 91 179 spectra of monochromatized PL for a-Si:H 182 spectrum of Rhodamine 6G 178 wideband 178–179 quantum-dot-quantum-well (QDQW) structure 506 quantum dots (QDs) epitaxial growth techniques 504–508 excitons in bandgap 509–510 Coulomb interaction 508 optical transitions 510–513 Schrödinger equation 508 lasers and optical amplifiers 522 advantages 520 development 521 gain dynamics 522–524 homogeneous broadening and dephasing 524–526 long-wavelength lasers 526–527 material gain coefficient 521 nano lasers 527, 528 optical properties internal quantum efficiency 514–516

linewidth, coherence, and dephasing 516–517 oscillator strength 514–516 PL spectra 513–514 radiative lifetime 514–516 TFWM 517–520 single-photon emitters efficiency 530 HBT experiment 529, 530 HOM experiment 529 micropillars and nanowires 530–531 photonic crystal waveguide 531–533 qubits possessing 528 requirements 528 quantum efficiency 159 quantum wells (QWs) 55 exciton formation acoustic phonon 469–470, 474 Auger processes 470–471 charge carrier density 472 charge carrier temperature 473 electron–hole pairs 468–469 excitation energies 467 factors 467–468 formation rate 470–473 interaction Hamiltonian 467 LO phonons 473–474 square-law dependence 472 trial wave function 468 exciton–phonon interaction 466–467 localization acoustic phonon 494–495 annihilation operator 492 COM 492–493 effective potential 495 free excitons 491–492 interaction Hamiltonian 493–494 interface roughness 491–492 localized exciton 491–492 molecular layer 493 rates of 496–499 variational method 495 nonradiative relaxation (see nonradiative relaxation) quasi-2D free-exciton linewidth 485–491 quartz (ne ), Sellmeier coefficients 9

641

642

Index

quartz (no ), Sellmeier coefficients 9 quartz, Verdet constant 57 quasi-2D free-exciton linewidth Bose–Einstein distribution 486 dephasing rate 486–488 exciton density and momentum 490–491 experimental results 489–490 experiments 485–486 lattice and exciton temperatures 488–489

r radiative lifetime 514–516 radiative lifetimes 221, 234 radiative recombination operator 206–210 radiative recombination processes, in organic materials 159 radiative tunneling transition 175 radio frequency (RF) magnetron sputtering 327, 414 Raman scattering 163–164 Raman spectra 457–458 random bond network (RBN) type, amorphous chalcogenides 80 rare-earth-doped glass fluorescence in 102–104 optical absorption in 96–99 rare-earth-doped transparent glass ceramics 120–121 rare-earth elements, occupation of outer electronic shells 49 rare-earth ions absorption spectrum 99 optical absorption of trivalent 48–54 rate of fluorescence 233 rate of phosphorescence 233–234 Rayleigh–Gans (or Rayleigh–Debye) theory 117 RC low-pass filter (LPF) 169 reflectance, of partially transparent plate 25, 27 reflectance R 4 of AlSb 40 of doped InSb 40 reflection coefficient 3, 85–86

vs angle of incidence in degrees 87 vs relative refractive index 86 vs wavelength for an air/silica glass interface 86 reflection coefficients 438–439 refractive index 2–5, 38, 84–86 chemical perturbations 94–95 of crystalline semiconductors 10–11 and dispersion 7–16 elastic-ether theory 7 engineering 106–109 experimental techniques for measuring 2 of glasses 11–14 and Kramers–Kronig relations 5–6 and K versus normalized frequency from single electronic dipole oscillator model 4 magnetic field dependence 92–94 sensitivity of 90–95 of silica–germania glass 8 stress dependence 91 temperature dependence 90–91 vs wavelength for silica glass SiO2 85 refractivity 278–280 renewable-energy technologies electrochromic thin films 413 solar-absorbing surfaces 412–413 thermochromism phenomena 414–417 resonant two-photon absorption 286 Reststrahlen (lattice) absorption 40–42 rod lens 107 root mean square error (RMSE) 21 ruby glass 95, 96 Ruddlesden–Popper layered perovskites 547

s 𝜎 and 𝜋 bonds, in a single chain of polyacetylene 297 sapphire, Sellmeier coefficients 9 saturation intensity, GaAs multiple quantum wells (MQW) 134 saturation parameter 134 S, bandgap reduction 231 Sb2 S3 , bandgap reduction 231

Index

scanning electron microscopy (SEM) 456, 459–460 scatterometry. see linear grating Schuster–Kubelka–Munk theory 27–31 scintillator 563 Se bandgap reduction 231 spectral dependence of one-and two-photon absorption coefficients 276 Se amorphous thin film construction of envelopes 18 regenerated transmission spectrum 21 transmission spectrum, determination of refractive index 20 transmission spectrum with interference fringes 17 selenide glasses, linear and nonlinear refractivity 278 selenium atoms 242 self-assembled quantum dots 504–505 self-assembled quantum dots (SAQDs) Brillouin function 602 micro-PL spectra 600 PLE study 604 P value 604 spin injection 601–603 type-II transition 602 Zeeman shift 600–601 Sellmeier coefficients 8, 9 Sellmeier equation 8–10 Sellmeier formula 90 Sellmeier relationship 8–10 semiconductors DAP recombination schematic optical transition models 160 optical and structural properties 39 refractive index of 10–11 temperature coefficient of refractive index (TCRI) 11 semiconductors, refractive index of bandgap and temperature dependence 11 crystalline semiconductors 10–11 sheet resistance 340 Si Cauchy’s dispersion parameters 8

FB coefficients 8 Forouhi–Bloomer (FB) coefficients 12 fundamental absorption at two temperatures 46 hydrogenated nanocrystalline (nc-Si:H) 72–73 optical and structural properties 39 Si (crystal) complex relative permittivity as function of photon energy 6 optical properties versus photon energy 6 Wemple–DiDomenico dispersion relationship parameter 15 SiC, values of quantities required for calculating n and K 42 Si-doped GaAs 615 silica–germania glass, refractive index of 8 single electron–hole pair, excitation of 228–229 single electronic dipole oscillator model n and K versus normalized frequency 4 reflectance versus normalized frequency 4 single interface absorption and extinction coefficients 447 Beilby overlayer 446 Brewster angle 444 Cartesian coordinate systems 442–443 ellipsometric angles 443–444 ellipsometric parameters 444–446 Fresnel relations 442–443 penetration depth 446 principal angle 444 pseudo-optical constants 446 refractive index 446–447 relative electric permittivity 446 relative magnetic permeability 446 spectral dependence 447–448 transmission coefficient 443 single-photon emitters efficiency 530 HBT experiment 529, 530 HOM experiment 529 micropillars and nanowires 530–531 photonic crystalwaveguide 531–533

643

644

Index

single-photon emitters (contd.) qubits possessing 528 requirements 528 singlet exciton binding energy 136 singlet excitonic states, in amorphous solids 137 SiO2 286 absorption coefficient vs wavelength 85 atomic structure 271 DOS of the valence and conduction bands 276 linear and nonlinear refractivity 278 refractive index n and the group index as a function of wavelength 16 refractive index vs wavelength 85 Sellmeier coefficients 8 spectral dependence of the one-and two-photon absorption coefficients 276 structural responses to bandgap and mid-gap excitations 281 two-photon absorption spectra 278 Wemple–DiDomenico dispersion relationship parameter 15 site-controlled growth 505–506 site-selective fluorescence measurement 306 Snell’s law 263, 405–406 solitons 307 solution-processed deposition methods 544, 546 spectral hole-burning (SHB) 524 spectral selective radiating materials (SSRMs) ambient temperature 418 atmospheric window 417–418 radiative cooling power 419–420 surface temperature 418 thermochromic materials 418 spectroscopic ellipsometry (SE) 421 amplitudes and phases 440 ellipsometric angles 439 instrumentation 441–442 light polarization 436–438 linear grating 458–462 multilayer applications 454

EMA 456–457 matrix calculations 454 NCD film 454–458 ND particles 455 optically isotropic samples 441 parameters 439–440 photometric measurements 440 reflection coefficients 438–439 single interface (see single interface) single layer absorption coefficient 452 approaches 450 defects and non-idealities 453 optical reflectance spectrum 448 phase change 451 reflection coefficients 448–449 SiO2 thin film 453–454 spectrophotometry measurements 449–450–451 Tauc–Lorentz formula 452 transmittance spectrum 451–453 sp2 -hybridized orbitals, and pz orbital 297 spin injection 591–593, 602 spin multiplicity 51 spontaneous emission determining E0 215–216 at nonthermal equilibrium 212–215 possibilities 217–218 and radiative lifetime 216–222 rates of 211–216 rates under two-level approximation 211–212 results 211–222 thermal equilibrium 214–215 two-level approximation results 211 Staebler–Wronski Effect (SWE) 204 Stark effect, confined 51 Stark shift 312 static photoluminescence spectroscopy 164–165, 167 Stokes shift 218, 236, 319, 320 stop band, PBG parameters 254 Stranski–Krastanov (SK) growth mode 504, 600 streak camera 172 stretched exponential function (SEF) 612

Index

structural changes, nonlinear excitation-induced 280–285 sub-monolayer (SML) 506 sulfide glasses, linear and nonlinear refractivity 278 superprism effect 262 Swanepoel technique 16–25

t tail states 67 Tauc–Lorentz formula 452 Tauc plot 69 Tauc’s coefficient 74 Tauc’s relation 68, 71, 138, 213 Tb3 –Ga5 garnet (633 nm), Verdet constant 57 Tb–Ga garnet (1064 nm), Verdet constant 162 temperature coefficient of refractive index (TCRI), of semiconductors 11 temperature dependence, bandgap and 11 terbium 94 thickness variation 22 thin films anisotropic films 407–408 energy management and renewable-energy technologies (see renewable-energy technologies) environmentally friendly technologies, TiO2 420–424 isotropic film 404–406 multi-layered system 406–407 phase transitions incidence angles and film thicknesses 426–427 interference effects 425 optical properties 424 phase-change materials 424 phase shifts 428 polarization effects 425 polarization modulation 427–428 polarization rotation 428–429 transition metal oxides 424 VO2 films 425–426, 428 SSRMs 417–420 transmission-mode photoellipsometry 408–412

thin flims absorbing system of film with wedge-like cross-section 22 with nonuniform thickness 22–25 sketch of typical behavior of light passing through 17 three 3d metal ions, absorption by 99 tight-binding molecular dynamics, simulation of a-Se 243 time amplitude converter (TAC) 171 time-correlated single-photon counting (TCSPC) 168–171 block diagram of conventional system 171, 172 time-of-flight positron emission tomography (TOF-PET) 565 time resolved spectroscopy (TRS) 168–171 TlCl, Wemple–DiDomenico dispersion relationship parameter 15 top-emitting OLED (TOLED) 324 on Al-PET, cross-sectional view 352 cross-sectional view 352 on glass and Al-PET 354 luminous efficiency–voltage characteristics 354 photograph of a flexible 352 on TFT, cross-sectional view of active-matrix displays with 352 transient four-wave mixing (TFWM) 517–520 transient photodarkening 204, 225 Transition layer (TL) 454 transition matrix element 74–75, 206–210 transitions extended-to-extended states 210 extended to tail states 209 transmission coefficients vs angle of incidence in degrees 87 vs relative refractive index 86 transmission electron microscopy (TEM) 116, 456 transmittance, of partially transparent plate 25, 27 transparent conducting oxide (TCO) thin films 327

645

646

Index

transparent glass ceramics 114–124 empirical criteria 116 ferroelectric 121 photonic applications 115 rare-earth-doped, for active photonics 120–121 for X-ray storage phosphors 121–124 transparent white organic light-emitting diodes (WOLEDs) angular-dependent color stability 395–400 electron-hole current balance 384–385, J–V characteristics (see current density-voltage (J–V) characteristics) L–J characteristics CIE coordinates 390–393, 396 emission spectra 390, 393, 396 EQE 393, 394 power efficiency 392–394 resonant mode 390 symmetrical and bidirectional emission 390 weak microcavity 390 optimization blue phosphorescent emitter bis 374 device architecture 375 EL spectra 376–377 energy diagram 375 OLEDs 375–376 orange phosphorescent emitter bis 374 PL emission 374–375 TCTA 375–377 TPBi 375 visible-light transparency 386–389, 395 transverse acoustic (TA) phonon emission 476–477 trapped-exciton model 225 triplet exciton binding energy 136 triplet excitonic states, in amorphous solids 137 triplet-triplet annihilation (TTA) 380–381 1,3,5-tri[(3-pyridyl)-phen-3-yl]benzene (TmPyPB) interface J–V characteristics 381–384

2,2′ ,2′′ -(1,3,5-benzinetriyl)-tris(1-phenyl-1H-benzimidazole) (TPBi) optimization 375 triplet-triplet energy transfer 379 tris(8-hydroxyquinoline)aluminum (Alq3 ) 326 4,4′ ,4′′ -tris(carbazol-9-yl)-triphenylamine (TCTA) J–V characteristics 380–383 optimization 375–377 trivalent rare earth ions, optical absorption of 120 tunable photonic crystals (PCs) 256–257 tuning PC response by altering physical structure of PC MEMS actuation 260 piezoelectric effects 259–260 strain 258–259 temperature 258 refractive index of constituent materials applied electric field 256–257 electric field temperature, in semiconductor-based PCs 257 infiltrated PC 257 light 256 using magnetism 258 using micro-electro-mechanical systems (MEMS) actuation 260 using strain 258–259 using temperature 258 tunneling theory 344 two-level approximation 211 two-photon absorption 274, 286 two-photon absorption technique 311 two-photon excitation 285 two-photon excitation measurement 311 two-step absorptions 273 Twyman–Green interferometer 242 type I germinate pair 205 type I heterostructure 581 type II germinate pair 205

u ultraviolet (UV) 420 uniform-thickness films 16–21 up-conversion photoluminescence 162–163

Index

Urbach rule 48, 79 Urbach tail 78, 105 Urbach width 48 UV–ozone-treatment 340 UV range, free-electron reflectance of metals 42

v valence change glasses 111–115 vanadium dioxide (VO2 ) thin films chemical substitutions 414 metallic substrates 428 MIT 417 optical-switching behavior 417 properties 426 refractive indices 425 RF magnetron sputtering 414 spectral transmittance 415–416 spin-coating method 417 transition temperature 416 water-cooled metal targets 414–415 van Hove singularities 47 vapor-liquid-solid (VLS) scheme 507 variable-frequency fluorometry 173 Verdet constants 92 of some glasses 56 for undoped and rare-earth-doped fluorozirconate glass 94 vertical-cavity surface emitting lasers (VCSELs) 526 vibronic potential-energy curves absorption 302 photoluminescence 302 vibronic structures, calculated for Huang–Rhys parameters 303 vibronic wave function 303 visible-light transparency 385 Ag thicknesses 389–390 calculated transparency 387–388 design 386 EL emission 388–389 index matching layer 387 ITO 386–387 wavelength function 388–389

w waiting time distribution 617 Wannier excitons 159 Wannier–Mott excitons 129 in amorphous semiconductors 135 Coulombic interaction between hole and electron 130 energy of 130 waveguide devices 261–262 waveguide dispersion, superprism effect 262 Wemple–DiDomenico dispersion relation 14–15 wet cleaning process 340 wideband QFRS 178–189

x X-ray diffraction (XRD) 554–555 X-ray storage phosphors, transparent glass ceramics for 121–124

y YAG laser

162

z Zeeman effect 595, 604–605 Zeeman shift 600, 601 ZerodurR 115 zero-material dispersion wavelength 15 ZnO free-carrier absorption coefficient in n-type 43 Wemple–DiDomenico dispersion relationship parameter examples 15 ZnSe optical and structural properties 39 Wemple–DiDomenico dispersion relationship 15 ZnSe buffer layer 599 ZnS, Verdet constant 57 ZnTe, optical and structural properties 39

647