Surface and Interface Science, Volumes 9 and 10: Volume 9 - Applications I; Volume 10 - Applications II (Wandelt Hdbk Surface and Interface Science V1 - V6) [9-10, 1 ed.] 3527413812, 9783527413812

Table of contents :
Cover
Surface and Interface Science,
Volume 9: Applications of Surface Science I
Contents
Surface and Interface Science,
Volume 10: Applications of Surface Science II
Contents

Citation preview

Edited by Klaus Wandelt Surface and Interface Science

Surface and Interface Science Edited by Klaus Wandelt Volume 1: Concepts and Methods Volume 2: Properties of Elemental Surfaces Print ISBN 978-3-527-41156-6 oBook ISBN 978-3-527-68053-5 (Volume 1) oBook ISBN 978-3-527-68054-2 (Volume 2) Volume 3: Properties of Composite Surfaces: Alloys, Compounds, Semiconductors Volume 4: Solid-Solid Interfaces and Thin Films Print ISBN 978-3-527-41157-3 oBook ISBN 978-3-527-68055-9 (Volume 3) oBook ISBN 978-3-527-68056-6 (Volume 4) Volume 5: Solid-Gas Interfaces I Volume 6: Solid-Gas Interfaces II Print ISBN 978-3-527-41158-0 oBook ISBN 978-3-527-68057-3 (Volume 5) oBook ISBN 978-3-527-68058-0 (Volume 6) Volume 7: Liquid and Biological Interfaces Volume 8: Interfacial Electrochemistry Print ISBN 978-3-527-41159-7 oBook ISBN 978-3-527-68059-7 (Volume 7) oBook ISBN 978-3-527-68060-3 (Volume 8) Volume 9: Applications of Surface Science I Volume 10: Applications of Surface Science II Print ISBN 978-3-527-41381-2 oBook ISBN 978-3-527-82249-2 (Volume 9) oBook ISBN 978-3-527-82250-8 (Volume 10)

Edited by Klaus Wandelt

Surface and Interface Science Volume 9: Applications of Surface Science I

The Editor Prof. Dr. Klaus Wandelt University of Bonn Institute of Physical and Theoretical Chemistry Germany and University of Wroclaw Institute of Experimental Physics Poland Cover: Designed by Klaus Wandelt Cover Pictures: Left: Agresti et al., RSC Adv., 2014, 4, 12366–12375, Reproduced by permission of The Royal Society of Chemistry. Middle: © IBM Research. Right: Schulz Grafik-Design, Fußgönheim, Germany. Cover Design: Klaus Wandelt and Grafik-Design Schulz

All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and publisher do not warrant the information contained in these books, including this book, to be free of errors. Readers are advised to keep in mind that statements, data, illustrations, procedural details or other items may inadvertently be inaccurate. Library of Congress Card No.: applied for British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at . © 2020 Wiley-VCH Verlag GmbH & Co. KGaA, Boschstr. 12, 69469 Weinheim, Germany All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, microfilm, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law. Print ISBN: 978-3-527-41381-2 Typesetting SPi Global, Chennai, India Printing and Binding

Printed on acid-free paper 10 9 8 7 6 5 4 3 2 1

V

Contents Volume 9 About the Editor XIII Preface XV List of Abbreviations XIX

61 61.1 61.1.1 61.1.2 61.1.2.1 61.1.2.2 61.1.3 61.1.3.1 61.2 61.2.1 61.3 61.3.1 61.3.2 61.3.3 61.4 61.4.1 61.4.1.1 61.4.1.2 61.4.1.3 61.4.1.4 61.4.1.5 61.4.1.6 61.4.1.7 61.5

Thin Films: Sputtering, PVD Methods, and Applications 1 John Colligon and Vladimir Vishnyakov Introduction 1 Outline of the Scope of the Chapter 1 Sputtering 3 History and Development of Sputtering 3 Fundamentals of Sputtering 3 Sputtering Methods 10 Ion Beam Systems 10 Thin Film Coatings 19 Advantages and Some Disadvantages of Energy-Assistance 19 Surface Roughening, Smoothing, and Patterning 21 Surface Topography 21 Cluster Beams 27 Surface Topography: The Consummation 29 Ion-Induced Property Changes with Potential Applications 30 Changes in Material Properties by Ion-Based Processing 30 Phase Changes 30 Bio Changes 31 Improved Adhesion of Coatings 31 Hardness, Stress, and Crystal Properties 31 Friction and Wear 35 Electrical Properties 36 Antiferromagnetic Coatings for Spintronics 36 Current Commercial Applications of Ion-Surface Processes 37

VI

Contents

61.6 61.7

Quo Vadimus 2018 44 Final Words 47 Acknowledgments 48 References 48

62

Wafer Bonding 57 Bernhard Rebhan, Viorel Dragoi, Thomas Plach, Christoph Flötgen, Kurt Hingerl, Markus Wimplinger, and Nasser Razek The Importance of Wafer Bonding 57 Physics of Wafer Bonding 61 Gibbs Free Energy 61 Relation Between Binding Enthalpies and Surface Energy 62 Intermediate State 65 Wafer Bonding Characterization Techniques 68 Characterization Techniques 68 C-Mode Scanning Acoustic Microscopy (C-SAM) 69 Maszara Method 70 Electron Backscatter Diffraction (EBSD) 71 Sample Preparation Techniques 73 AES and SIMS Specimen Preparation 74 X-SEM Specimen Preparation with FIB 74 X-TEM Specimen Preparation 74 EBSD Specimen Preparation 77 Wafer Bonding Processes 77 Low-Temperature Direct Wafer Bonding 77 Introduction 77 Process Flow 78 Surface Characterization 79 Interface Characterization 86 Summary 88 Room Temperature Covalent Wafer Bonding 89 Introduction 90 Process Flow 92 Surface Characterization 94 Interface Characterization 99 Summary 103 Application Example: GaAs–Si Covalent Wafer Bonding 103 Metal Thermo-Compression Wafer Bonding 106 Introduction 107 Process Flow 108 Surface Characterization 109 Interface Characterization 120 Summary 126 Example: Al–Al Thermo-compression Wafer Bonding 127 Example: Hybrid Wafer Bonding 128

62.1 62.2 62.2.1 62.2.2 62.2.3 62.3 62.3.1 62.3.1.1 62.3.1.2 62.3.1.3 62.3.2 62.3.2.1 62.3.2.2 62.3.2.3 62.3.2.4 62.4 62.4.1 62.4.1.1 62.4.1.2 62.4.1.3 62.4.1.4 62.4.1.5 62.4.2 62.4.2.1 62.4.2.2 62.4.2.3 62.4.2.4 62.4.2.5 62.4.2.6 62.4.3 62.4.3.1 62.4.3.2 62.4.3.3 62.4.3.4 62.4.3.5 62.4.3.6 62.4.3.7

Contents

62.5

Summary 130 References 132

63

Superconformal Deposition 139 Daniel Josell and Thomas P. Moffat Introduction 139 The Short List of Superconformal Deposition Mechanisms 141 A Caveat: Geometric Leveling 143 Superconformal Filling 145 The Underlying Equations 145 Superconformal Filling by Leveling 148 Measurements to Quantify Leveling 149 Modeling of Leveling 150 Superconformal Deposition by Curvature Enhanced Adsorbate Coverage Mechanism 153 Measurements to Quantify Superconformal Deposition by Area Change 155 Superconformal Feature Filling 159 Damascene CEAC Feature Filling Beyond Copper 159 Superconformal Chemical Vapor Deposition 163 Overshoot Phenomenon and Its Control 164 Surface Diffusion in CEAC Superfill 167 Surface Smoothing Through the CEAC Superconformal Deposition 168 Transient Effects 169 Nonlinear Effects Associated with Suppression Breakdown: Extreme Bottom-Up Filling 170 Measurements to Quantify Superconformal Deposition by S-NDR 172 S-NDR Modeling of Superconformal Deposition 174 Other Systems 180 Conclusion and Outlook 182 References 183

63.1 63.1.1 63.2 63.3 63.3.1 63.4 63.4.1 63.4.2 63.5 63.5.1 63.5.2 63.5.3 63.5.4 63.5.5 63.5.6 63.5.7 63.5.8 63.6 63.6.1 63.6.2 63.6.3 63.7

64 64.1 64.2 64.3 64.3.1 64.3.2 64.3.3 64.3.4 64.3.5 64.4

Spintronics: Surface and Interface Aspects Claus M. Schneider Introduction 187 Replacing Charge by Spin? 188 Basic Issues in Spintronics 190 Passive vs. Active Concepts 191 Electronic Aspects 192 Transport Aspects 194 Geometrical Aspects 196 Material Aspects 197 Giant Magnetoresistance (GMR) 199

187

VII

VIII

Contents

64.4.1 64.4.2 64.4.3 64.4.4 64.4.4.1 64.4.4.2 64.4.4.3 64.4.4.4 64.5 64.5.1 64.5.2 64.5.3 64.6 64.6.1 64.6.2 64.7 64.7.1 64.7.2 64.7.3

Interlayer Exchange Coupling 199 GMR in Multilayers 202 GMR in Trilayer Structures 208 The Role of the Interface 211 Exchange Biasing 212 Interfacial Dusting 214 Role of Oxygen 215 Temperature-Induced Degradation 217 Tunneling Magnetoresistance 219 Spin-Dependent Tunneling in the Jullière Picture 219 MTJs with AlOx Barriers 220 Spin-Dependent Tunneling Beyond Jullière’s Model 222 Spin Transfer Torque 226 Microscopic Picture 226 Experimental Realization 228 Technological Perspectives 230 Magnetic Mass Storage 231 Magnetic Sensors 231 Nonvolatile Core Memory 232 Acknowledgment 235 References 235

65

Device Efficiency of Organic Light-Emitting Diodes 243 Wolfgang Brütting Introduction 243 OLED Operation 244 OLED Architecture and Stack Layout 244 Working Principles of OLEDs 245 OLED Materials 247 White OLEDs 248 Electroluminescence Quantum Efficiency 249 Factors Determining the EQE 249 Fundamentals of Light Outcoupling in OLEDs 251 Optical Loss Channels 251 Optical Modeling of OLEDs 253 Simulation-Based Optimization of OLED Layer Stacks 258 Influence of the Emitter Quantum Efficiency 260 Comprehensive Efficiency Analysis of OLEDs 261 Approaches to Improved Light Outcoupling 265 Overview of Different Techniques 265 Reduction of Surface Plasmon Losses 268 Basic Properties of SPPs 268 Scattering Approaches 268 Index Coupling 270 Emitter Orientation 272

65.1 65.2 65.2.1 65.2.2 65.2.3 65.2.4 65.3 65.3.1 65.4 65.4.1 65.4.2 65.4.3 65.4.4 65.4.5 65.5 65.5.1 65.5.2 65.5.2.1 65.5.2.2 65.5.2.3 65.5.2.4

Contents

65.6

Summary and Outlook 278 Acknowledgement 279 References 280

66

Dye-Sensitized Solar Cell 287 Sara Pescetelli, Antonio Agresti, Angelo Lembo, and Aldo Di Carlo Introduction 287 DSCs: Material Interfaces, Energy Matching, and Layer Morphology 289 Transparent Conducting Substrate 289 Semiconductor Materials 291 Dye Sensitizer 292 Electrolyte 294 Counter Electrode 297 DSC: Working Principle 298 Photovoltaic Characterization of DSCs 300 DSC: Interface and Stability 304 Interfaces 305 TCO/TiO2 305 TiO2 /Dye/Electrolyte 306 From Cell to the Module: Module Structure Optimization 313 DSC: Stability 315 Photoelectrode Stability 316 Counter Electrode Stability 320 Stability of Dye-Sensitized Module 322 Thermal Stability 323 Outdoor Stability 324 Stability and Degradation Under Reverse Bias 324 References 327

66.1 66.2 66.2.1 66.2.2 66.2.3 66.2.4 66.2.5 66.3 66.3.1 66.4 66.4.1 66.4.1.1 66.4.1.2 66.5 66.6 66.6.1 66.6.2 66.6.3 66.6.4 66.6.5 66.6.6

67 67.1 67.2 67.2.1 67.2.2 67.3 67.3.1 67.3.2 67.3.3 67.3.4 67.3.5 67.4 67.4.1 67.4.2

Electronic Nose: Current Status and Future Trends 335 Anna Staerz, Frank Roeck, Udo Weimar, and Nicolae Barsan Introduction to the Device 335 Sample Handling 336 Sampling 336 Filtration and Analyte Separation 337 Sensors 338 Gravimetric Sensors 339 Chemiresistive Sensors 339 Chemically Sensitive Field Effect Transistors (CHEMFETS) 339 Amperometric Gas Sensors 340 Optical Sensors 340 Classic Analytical Methods 341 Infrared Spectroscopy 341 Mass Spectrometer 342

IX

X

Contents

67.4.3 67.5 67.5.1 67.5.2 67.5.3 67.5.4 67.6 67.7 67.8

Ion Mobility Spectrometer 343 Application Areas 343 Food Analysis 343 Environmental Comfort and Safety Monitoring Applications 346 Indoor and In-Cabin Air Monitoring 351 Medical Diagnostics 354 Data Evaluation 358 Commercially Available E-Nose Systems 359 Conclusion and Outlook 359 References 368

68

Surface Science in Batteries 381 Yuping Wu and Rudolf Holze Batteries: An Overview 381 Experimental Methods 392 Overpotentials 408 Electrode Materials 410 Examples 413 Aqueous Systems 413 Lithium-Ion Batteries 414 Alkaline Primary Batteries 418 Supercapacitors 418 Acknowledgments 422 References 422

68.1 68.2 68.3 68.4 68.5 68.5.1 68.5.2 68.5.3 68.5.4

69

69.1 69.1.1 69.1.2 69.1.3 69.1.4 69.1.5 69.1.5.1 69.1.5.2 69.2 69.2.1 69.2.1.1 69.2.1.2 69.2.1.3 69.2.1.4

Surface and Interface Science in Fuel Cell Research 429 Indro Biswas, Kaspar Andreas Friedrich, Pawel Gazdzicki, Jens Mitzel, and Mathias Schulze Introduction to Fuel Cells 429 Principles of Fuel Cell Technology 429 Types of Fuel Cells 432 Electrocatalytic Reactions 435 Applications 437 Frequently Used Analytic Tools 438 Physical Characterization 438 Electrochemical Characterization 445 Surface Analysis of PEMFC Components and Degradation Issues 449 Catalyst Degradation 451 Electrochemical Analysis of Catalyst Degradation by CV 452 Morphological Analysis of Catalyst Degradation by Electron Microscopy 453 Physical Analysis of Catalyst Degradation by XPS 454 Physical Analysis of Catalyst Degradation by TPR Measurement 455

Contents

69.2.2 69.2.3 69.2.4 69.2.5 69.3 69.3.1 69.3.2 69.3.3 69.4

Carbon Support Degradation 456 Membrane and Ionomer Degradation 457 Aging of Hydrophobic Agents 461 Bipolar Plate Corrosion 464 Examples of Technology Implementation and Demonstration 465 Transport Application 465 Stationary Applications 467 Portable Applications 467 Summary 468 References 469

Index 927

XI

XV

About the Editor

Klaus Wandelt is currently Professor Emeritus at the University of Bonn, Germany, where he was also Director of the Institute of Physical and Theoretical Chemistry until 2010. He received his PhD on electron spectroscopy of alloy surfaces in 1975 in München; spent a postdoctoral period at the IBM Research Laboratory in San Jose, California, in 1976/1977; and qualified as a professor in 1981 in München. Since then his research focuses on fundamental aspects of the physics and chemistry of metal surfaces under ultrahigh vacuum conditions and in electrolytes, on the atomic structure of amorphous materials, and more recently on processes at surfaces of plants. Professor Wandelt was visiting scientist at the University of Caracas, Venezuela; the University of Hefei, China; the University of Newcastle, Australia; and the University of California, Berkeley, and he was guest professor at the University of Messina, the University of Padua, and the University of Rome Tor Vergata, Italy; the University of Linz and the Technical University of Vienna, Austria; and the University of Wroclaw, Poland. He chaired the surface physics divisions of the German and European Physical Society as well as of the International Union of Vacuum Science Techniques and Applications, has organized numerous workshops and conferences, and was editor of journals, conference proceedings, and books.

XVII

Preface Surfaces and Interfaces: A “Divine Gift”

For decades books, book chapters, theses of generations of PhD students, and, more recently, also presentations on the Internet about subjects of surface and interface science, i.e. the research of physical and chemical properties and processes at solid surfaces, often start with the quotation God made the bulk, surfaces were invented by the devil attributed to Wolfgang Pauli, Nobel Prize Laureate in Physics 1945 [1]. Of course, quotes like this are to be understood from the respective era; a systematic experimental “surface science” did not exist at that time. A description of the field ion microscope (FIM), which for the first time enabled the visualization of individual surface atoms, was published only a few years later by Erwin W. Müller [2]. Now, nearly seventy years later, our profound scientific understanding of the fascinating peculiarities of solid surfaces presented in Volumes 1–8 of this series of books and their fundamental importance for so many vital technological areas emphasized below, and in part addressed in Volumes 9 and 10, make the “invention of surfaces” truly a gift from God. Surfaces and interfaces enrich our world in a double sense. On the one hand, they structure our world and make it so diverse and beautiful. On the other hand, surfaces and interfaces are locations of gradients, which drive spontaneous and mancontrolled processes. These processes change our world and, therefore, our all living conditions in a fundamental way. On the one hand, heterogeneous catalysis of chemical reactions at solid surfaces has enabled the large-scale production of (i) fertilizers and pesticides for agriculture, (ii) a vast variety of plastic commodities, and (iii) pharmaceuticals for medicine and the “health industry.” These products (i) have contributed to a better food supply of the world population and thereby its rapid growth, (ii) appear no longer indispensible in our daily life, and (iii) help to fight diseases and save lives, if produced and applied responsibly and sustainably. On the other hand, besides the growing world population itself, the profit-driven excess production of these products and the accompanying ruthless exploitation of our natural resources are

XVIII

Preface

an increasing thread for humanity’s survival. The excess production and thoughtless use and uncontrolled disposal of these products disturbs natural equilibria and leads to an increasing contamination of soil and groundwater, pollution of the atmosphere and oceans, and a weakening or failure of the natural immune systems. Insufficient or neglected air pollution control is most likely a reason for the obvious “global warming.” The concomitant rise of the sea level will cause dramatic erosion processes at ocean shores and dikes, the largest-scale solid/liquid interfaces. The consequent shrinkage of man’s living space will, at best, cause a process of mass migration of people. The physics of interfaces and low-dimensional systems has opened the door to modern electronic devices that are revolutionizing the collection, processing, and availability of information, which not only changes our own communication behavior but has also created the vision of the “Internet of things (IoT)” in which people and mobile and immobile physical objects including buildings communicate within a single and common network with each other, which in the opinion of some people will change the world for the better, while others fear that man may lose control. Biological processes function via processes at and through interfaces of membranes, which in turn can be influenced by traces of drugs. It is, thus, not only a great scientific challenge to investigate the properties and processes at surfaces and interfaces, but also of vital importance for mankind’s future, provided we make wise use of this knowledge. Although theoretical predictions about properties of surfaces as well as intuitive models of surface processes existed much earlier, modern experimental surface science started by now about 50 years ago with the commercial availability of ultrahigh vacuum (UHV) technology. Under UHV conditions, it became possible to prepare clean surfaces and to develop and apply a growing number of “surface-sensitive” methods based on particle beams. Unlike photon beams, for instance, used in X-ray crystallography, electron, ion, and atom beams interact only with the outermost layers of a solid and therefore provide information pertaining only to the surface. While in the beginning, practical surface investigations were concentrated on the changes of surface properties due to exposure to gases or vapors, it soon turned out that the properties of the bare surfaces themselves pose a lot of scientific surprises. Now 50 years later, the so-called reductionist “surface science approach,” that is, the use of well-defined, clean single-crystal surfaces under UHV conditions, enables a microscopic and spectroscopic characterization of these bare surfaces atom by atom. The overwhelming achievements of this research may ultimately be summarized by the general statement: Surfaces are a different state of matter! Moreover, nowadays, it is possible not only to study the interaction of individual atoms and molecules with a surface but also to manipulate them on the surface according to our will. The present series of books aims not only at giving a broad overview of the present state of understanding of the basic physics and chemistry at surface and interfaces but also at highlighting a number of technological applications that rely on the established knowledge about surfaces, like thin film and nanotechnology, highly integrated electronics, heterogeneous catalysis in gaseous and liquid phases,

Preface

electrochemical energy conversion and storage, and bio-functionalization of inorganic materials, to name a few. The intention of this series of books is, thus, not only to give an introduction for those who enter the field of surface research but also to provide an overview for those whose work needs conceptual and analytical input from surface science. According to the original concept, this book series should comprise six volumes. The first volume was planned to describe “bare surfaces and methods,” that is, all the physical properties of clean surfaces of elemental and composite solids as well as the most relevant surface analytical methods. However, it turned out immediately that an adequate treatment of just these topics exceeded by far the reasonable size of a single volume and instead filled three volumes, extending the number of intended volumes to eight. But also the material for Volumes 7 and 8 went beyond the limits of one book each, so, after all, the series comprises 10 volumes now: Volume 1: Concepts and Methods Volume 2: Properties of Elemental Surfaces Volume 3: Properties of Composite Surfaces: Alloys, Compounds, Semiconductors Volume 4: Solid/Solid Interfaces and Thin Films Volume 5: Solid/Gas Interfaces I Volume 6: Solid/Gas Interfaces II Volume 7: Liquid and Biological Interfaces Volume 8: Interfacial Electrochemistry Volume 9: Applications of Surface Science I Volume 10: Applications of Surface Science II. The first eight volumes emphasize the basic insights into the physics and chemistry at surfaces and interfaces and the most important experimental and theoretical methods, which led to these results. The methods are grouped according to the applied probe, namely, electrons, ions, photons, and proximity probes, and are described to an extent to give the reader enough confidence in “what surface scientists are able to do nowadays”; more detailed descriptions of these methods can be found in the existing specialized literature. The last two volumes present a selection of some daily phenomena and technological applications, which depend on and arise from surface-specific properties and processes. The vast material is laid out in 80 chapters and is structured according to increasing complexity of the subject in question. Each chapter is written by experts in the respective field and is supposed to start with an introduction of the basic phenomenon, to develop the problem from simple to more specific examples, and to end, if possible, with the identification of open questions and challenges for future research. This intended strategy “from simple to complex” is graphically expressed by the veil rising from left to right on all book covers. One person alone could hardly ever have written such an extensive and divers oeuvre. I am extraordinarily thankful to all authors who have contributed to this series of books. I am also very grateful to the publisher, namely, Ulrike Werner, Nina Stadthaus, Dr. Frank Weinreich, and Dr. Martin Preuss at Wiley, for their continuous

XIX

XX

Preface

support and their understanding and flexibility to adapt the original concept of the whole project to “new circumstances” and to agree with the expansion from 6 to 10 volumes. Altogether it took 12 years to realize this project, and obviously a great deal of patience and persistence was necessary to complete it, patience of the authors and the publisher with the editor, but also persistence of the editor and his patience with some authors. The result of this joint effort of all three parties is now in the hands of the critical readers. After all, surfaces and interfaces are a “divine gift” and as such by no means fully fathomed. Bonn, Wroclaw January 11, 2019

References 1. Quoted in:Jamtveit, B. and Meakin, P.

(eds.) (1999). Growth, Dissolution and Pattern Formation in Geosystems, 291. Kluwer Academic Publishers. 2. Müller, E.W. (1951). Z. Phys. 131: 136.

Klaus Wandelt

XIX

List of Abbreviations 1D 2D IC 3D AEM AES AF AFC AFM AM ASIC ATR BAW BB-MZI BC BEOL BET BIPV bipy BN BPP BSI CIS ca CB CCD CCM CD CDC CDCA CE CEAC CFUBMS CHEMFET CHP

one-dimensional two-dimensional integrated circuit three-dimensional anion exchange membrane Auger electron spectroscopy antiferromagnetic alkaline fuel cell atomic force microscopy air mass application-specific integrated circuit attenuated total reflection bulk acoustic wave broadband Mach–Zehnder interferometer bladder cancer back-end-of-line Brunauer, Emmett, Teller building integrated photovoltaics 2,2′ -bipyridyl butyronitrile bipolar plate backside-illuminated CMOS image sensor citric acid conduction band charge coupled device catalyst-coated membrane Crohn’s disease carbide-derived carbon chenodeoxycholic acid counter electrode curvature enhanced accelerator coverage closed-field unbalanced magnetron system chemically sensitive field effect transistor combined heat and power

XX

List of Abbreviations

CKD CL CMOS CMP CNT COPD CRC C-SAM CSL Cu(dmp)2 Cu(I)-hfac CV CVD CWA DCA DCMS DEMS DL dmby DMC DMFC DMII dmp DMPI DVS EAD EASA EBSD ECD EDLC EDX EDXS EiPS EIS EXAFS FCEV FDA FENO FEOL FF FIB FPA FTIR FTO GBL

chronic kidney failure catalyst layer complementary metal-oxide-semiconductor chemical mechanical polishing carbon nanotube chronic obstructive pulmonary disease colorectal cancer c-mode scanning acoustic microscopy coincidence site lattice bis(2,9-dimethyl-1,10-phenanthroline)copper(I/II) Cu(I)-hexafluoroacetylacetonate cyclic voltammetry chemical vapor deposition chemical warfare agent deoxycholic acid direct current magnetron sputtering differential electrochemical mass spectrometry diffusion length bis(2,9-dimethyl-1,10-phenanthroline) dimethyl carbonate direct methanol fuel cell 1,3-dimethylimidazolium iodide 4,4′ ,6,6′ -tetramethyl-2,2′ -bipyridine 1,2-dimethyl-3-propylimidazolium ion dynamic vapor sorption energy-assisted deposition electrochemical active surface area electron backscatter diffraction electrochemically deposited electrochemical double-layer capacitor energy dispersive X-ray emission energy-dispersive X-ray spectroscopy ethyl isopropyl sulfone electrical impedance spectroscopy extended X-ray absorption fine structure/spectroscopy fuel cell electric vehicle food and drug administration fraction exhaled nitric oxide front-end-of-line fill factor focused ion beam focal plane array Fourier transform infrared fluorine-doped tin oxide 𝛾-butyrolactone

List of Abbreviations

GC GCIB GCIS GDL GdmSCN GIS GIXAFS GNP HAADF-STEM HF HfO2 -MGC HIPIMS HNC HOMO HOR HREELS HR-TEM HSAC HSE HT HV IBS IMPS IMS INDEx IPCE IPD IPF IR ISF ISS ITO JPL KC LC LDA LEED LMIS LSV LT LUMO LVQ MBE MCFC

gastric cancer gas cluster ion beams gas cluster ion source gas diffusion layer guanidinium thiocyanate gas injection system grazing incidence X-ray absorption fine structure/spectroscopy graphene nanoplatelet high-angle annular dark field scanning transmission electron microscopy hydrofluoric mesoporous–graphitic–carbon-supported Hafnium(IV) oxide high power impulse magnetron sputtering head and neck cancer higher occupied molecular orbital hydrogen oxidation reaction high-resolution electron energy loss spectroscopy high-resolution-transmission electron microscopy high surface area carbon high stability electrolyte high-temperature high vacuum irritable bowel syndrome intensity-modulated photocurrent spectroscopy ion mobility spectrometry inside-needle dynamic extraction incident photon to current convertion efficiency idiopathic Parkinson’s disease inverse pole figure infrared instrumental spreading function International Space Station indium tin oxide jet propulsion laboratory kidney cancer lung cancer linear discriminant analysis low energy electron diffraction liquid metal ion source linear sweep voltammetry low-temperature lowest unoccupied molecular orbital linear vector quantization molecular beam epitaxy molten carbonate fuel cell

XXI

XXII

List of Abbreviations

MEAs MEMS MLCT MOR MOSFET MPL MPN MPN MPP MPPMS MS MS MT NAP-XPS NBB ND NDIR NEGS NEXAFS NMBI NPD OA OC OLC ORR PAFC PAH PBI PC PC PCA PCE PDMS PE PEDOT PEDOT-PSS PEG PEM PEMFC PET PET PEUT PFSA PID

membrane electrode assemblies microelectromechanical systems metal-to-ligand charge transfer methanol oxidation reaction metal oxide field-effect transistor microporous layer methoxypropionitrile most probable number maximum power point modulated pulse power magnetron sputtering mass spectrometry multiple sclerosis mid-temperature near-ambient pressure X-ray photoelectron spectroscopy N-butyl-1H-benzimidazole normal (to the surface) direction non-dispersive infrared non-evaporable getters near-edge X-ray absorption fine structure structure/spectroscopy N-methylbenzimidazole national purchase diary optical axis ovarian cancer onion-like carbon oxygen reduction reaction phosphoric acid fuel cell pulmonary arterial hypertension polybenzimidazole propylene carbonate prostate cancer principal component analysis power conversion efficiency polydimethylsiloxane photoelectrode poly(3,4-ethylenedioxythiophene) poly(3,4-ethylenedioxythiophene)–poly(styrenesulfonic acid) polyethylene glycol polymer electrolyte membrane polymer electrolyte membrane fuel cell polyethylene terephthalate positron emission tomography polyether urethane perfluorosulfonic acid photoionization detector

List of Abbreviations

PIII PIPS PMII PN PTFE PtNPs/MWCNT PV PVA PVD QMB RB RD RDE RDS RFB RFID RH RMS RRS RSF RTIL SAW SBF SBSE SE SEI SEM SFC SIMS SiNC SMOX S-NDR SOFC SOI SPME STEM TBP TCO TD TDS TEM TGA TLC tmby TMS

plasma immersion ion implantation precision ion polishing system 1-methyl-3-propylimidazolium iodide propionitrile polytetrafluoroethylene Pt nanoparticles in multi-walled carbon nanotube photovoltaic pivalic acid physical vapor deposition quartz microbalance reverse bias rolling direction rotating disk electrode rate-determining step redox flow battery radio-frequency identification relative humidity root mean square resonance Raman spectroscopy relative sensitivity factor room-temperature ionic liquid surface acoustic wave simulated body fluid stir bar sorptive extraction spectroscopic ellipsometry solid electrolyte interface scanning electron microscopy smart fuel cell secondary ion mass spectrometry silicon nanocrystal semiconducting metal oxide S-shaped negative differential resistance solid oxide fuel cell silicon on insulator solid-phase microextraction scanning transmission electron microscopy 4-tert-butylpyridine transparent conductive oxide transverse direction thermal desorption spectroscopy transmission electron microscopy thermogravimetric analysis thin layer cell 6,6′ -dimethyl-2,2′ -bipyridine trimethylamine

XXIII

XXIV

List of Abbreviations

ToF ToF-SIMS TPD TPO TSV TVBN UC UHV UPS VOC WE WF WLI XANES XPS XRD X-SEM X-TEM YSZ ZnC

time-of-flight time-of-flight secondary ion mass spectrometry temperature-programmed desorption temperature-programmed oxidation through silicon vias total volatile basic nitrogen ulcerative colitis ultrahigh vacuum uninterrupted power supply volatile organic compound working electrode working function white light interferometry X-ray absorption near-edge structure/spectroscopy X-ray photo-electron spectroscopy X-ray diffraction cross-sectional scanning electron microscopy cross-sectional transmission electron microscopy yttria-stabilized zirconia zinc-carbon cell

1

61 Thin Films: Sputtering, PVD Methods, and Applications John Colligon and Vladimir Vishnyakov

61.1 Introduction 61.1.1 Outline of the Scope of the Chapter

This chapter is intended to provide an overview of the background, development and application of thin film, and surface modification processes at a level suitable for readers who are new to the field or for those who simply need an update of latest developments and applications. Many references are included for more comprehensive details of work in specific areas of application for readers with specialist interests. Surface coating has been used for many centuries for decorative purposes with the earliest gold films documented in an excellent review of this earlier work by Greene [1]. This chapter is concerned with thin films created by the ejection of atoms, molecules, and other fragments from a target material bombarded by energetic particles known as “sputtering.” This process is now widely used to provide a source of material for coating various substrates, forming multicomponent layers [2], depth profiling [3], preparing thin samples for atomically resolved imaging [4], and forming periodic structures [5]. The ability of the ion sputtering technique to coat various substrates permits a design engineer great flexibility to select a material for an application for its best bulk properties while, at the same time, optimizing the surface properties by coating it with a suitable different material. As discussed in more detail in Section 61.4, there are many potential applications of surface coatings and, in Section 61.5, examples of many engineering components already in use, which are coated by this method. These have special features such as hard wear-resistant coatings on cutting tools, low friction surfaces on moving parts in engines, and diffusion barriers. In addition there is a wide range of optical coatings offering special features on window glass and spectacles such as chromatic control to reduce transmission in bright light and increase in hydrophobicity to promote self-cleaning action on windows. The sputtering process lends itself to large-scale coating of components and to the production Surface and Interface Science: Applications of Surface Science I, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

2

61 Thin Films: Sputtering, PVD Methods, and Applications

Table 61.1 Some fields of application of surface coatings. Property

Typical application

Antireflection Scratch resistant Ceramic Window glass Low friction Diffusion barriers Bio-resistant High temperature Decorative

Optical Optical Tools Energy saving Bearings, tools Food packaging Medical applications Tools, engine components House fittings

of thin films comprising several materials so that new multi-element films can be created that may exhibit nanostructural features and can have exceptional hardness or durability features in extreme environments. Table 61.1 lists just some of the many current applications of coatings that contribute to the rapidly growing $12 billion per annum coating market. This market total includes sales of coating equipment, coating services, and coated components. The components are often used as vital parts in much larger systems, so they underpin an even larger economic return. In terms of their tribology properties, coatings on moving parts offer a further economic and environmental benefit by reducing friction and wear to reduce energy consumption and by avoiding coolants when machining components. The JOST Report to the UK government in 1966 put a value on this saving, which, scaled to 2015 values, is 20 billion pounds per annum [6]. The key to all this successful application of coatings is the ion-surface interaction process, which leads to the sputtering of atoms that form the coatings. We start this chapter therefore with a short overview of the sputtering process. Sputtering methods for forming these layers and latest developments in this area of surface modification and applications will be described in Sections 61.1.3 and 61.2. Section 61.3 looks at surface topography and its minimization when using sputtering to etch a surface, of importance when analyzing the concentration depth profiles of materials using X-ray photoelectron spectroscopy (XPS), secondary ion mass spectrometry (SIMS), or similar techniques. Here the development of roughness of the surface can seriously limit the depth resolution of the data although, in other applications such as catalysis, a rough surface can be an advantage and the ability of ion sputtering to form periodic surface structures opens various other possibilities. New and recently developed applications of surface treatment by sputter-deposition methods are reviewed in Section 61.4, and current commercial applications are described in Section 61.5. Section 61.6, quo vadimus, looks ahead

61.1 Introduction

at some of the latest developments where new surface modification procedures and improvements are reported and some challenges in this field remain. Final remarks and acknowledgments are made in Section 61.7. 61.1.2 Sputtering 61.1.2.1 History and Development of Sputtering

“Sputtering” is the name given to a vapor deposition process in which atoms, molecules, and other fragments of a material are removed as a result of bombardment by energetic particles of atomic or molecular size; these usually are in the form of charged positive or negative ions. The process itself has some similarities with sandblasting and even rain erosion, but, whereas these occur at atmospheric pressure, sputtering occurs at a reduced atmosphere allowing the ejected particles to flow through the rarefied gas and deposit onto another surface. The first recorded observation of sputtering was by W.R. Grove in 1852 [7]. Grove was, in fact, interested in the study of glow discharge plasmas that, at that time, had the attribute of being a new state of matter (adding to the solid, liquid, gas forms). He noted in his report that there was a deposit on the inside of the glass vessel containing the discharge but did not assign the name sputtering to this event. Much of the subsequent early work on sputtering was described by Fruth in 1932 [8], but the data was inconsistent because vacuum conditions were poor. Measurement of the sputtering coefficient, defined as a ratio Y = number of atoms sputtered∕number of ions arriving at the sample being eroded depends critically on the pressure in the system. At higher pressure the mean free path for sputtered particles is lower, which results in fewer sputtered atoms reaching the collector and an apparent low value of Y . Similarly, if measurements were by weight loss of the sample being sputtered in a high pressure system, particles can collide with gas atoms and return to the sputtered surface, again producing a low measured value of Y . A review in 1968 of work in this field by Carter and Colligon [9] shows the main trends in the understanding of the sputtering process and measurement of sputtering coefficient drawn from experiments performed in improved vacuum conditions in the post-1932 years. This and later data will be referred to in the subsequent sections to give the reader a quick appreciation of the current state-of-the-art understanding of the sputtering process. 61.1.2.2 Fundamentals of Sputtering

Figure 61.1 illustrates a typical sequence of events that can occur when an energetic charged particle strikes the surface of a material. Assuming there is no energy loss in the process in the form of heat or radiation, then conservation of momentum rules

3

4

61 Thin Films: Sputtering, PVD Methods, and Applications

Ion or neutral atom

Figure 61.1 Schematic diagram of a collision sequence induced by ion bombardment, leading to sputtering of an atom. Incident ion (red); displaced atoms (blue); sputtered atom (yellow); channeling ion (green).

applies in all atom–atom collisions. Thus, if the ion has mass m1 and approaches a target atom and mass m2 with velocity v1 , m1 𝑣1 = m1 𝑣1 ′ + m2 𝑣2 ′

(61.1)

where v1 ′ and v2 ′ are the velocities of ion and struck atom, respectively, after collision (assuming the initial velocity of m2 is zero). Depending on the collision geometry, whether the original ion path is directed through the center of the struck atom (headon collision with so-named “zero impact parameter”) or through a point a distance p to the side of the center (glancing incidence with impact parameter p), the momentum transfer to the struck atom and subsequent particle trajectories vary. For an ion with initial energy Eo and head-on collision, the energy transfer E is E = [4m1 ⋅ m2 ∕(m1 + m2 )2 ] ⋅ Eo

(61.2)

For m1 = m2 all the energy is transferred from the ion to the struck atom in a headon collision. For non-head-on collisions, the ion retains some energy, and both ion and struck atom continue to make further collisions forming what is known as a collision cascade of moving atoms. Energy is required to displace atoms from their normal sites in a material, so the extent of this cascade is limited. However there is still a reasonably high probability for an ion to bounce back toward a surface atom and to transfer energy to it greater than the surface binding energy (Eb ) in a direction perpendicular to the surface. This atom is then ejected as a sputtered atom. It should be noted that, especially for targets bombarded at temperatures below 0.2–0.3 of their melting point (in K), cooling rates for the cascade atoms can be exceptionally high (of order 1013 K/s) [10]. This means that stress and defects become frozen into the post-bombarded material with potentially useful effects on material properties such as hardness (see Section 61.4).

61.1 Introduction

With the above, rather elementary, description of the sputtering process, it is possible already to predict some trends:

Normalized sputtering yield (a.u.)

1. The bombarding ions must have energy greater than the surface atom binding energy, Eb ; in fact the energy must be quite a bit larger as the escaping atom normally has to receive this energy after several collisions. The minimum energy for atom ejection is called the sputtering threshold energy. 2. If the sputter coefficient Y is high, then many sputtered atoms will have energies above this threshold value [11, 12]. It was found that the energy distribution of sputtered atoms had the form shown in Figure 61.2, where the most probable energy is of order Eb and the high energy tail follows a 1/E2 dependence. Some sputtered atoms can have energies up to about 50 eV, far higher than the energies of thermally evaporated atoms, which are of the order 0.07 eV. 3. As the energy of the ions becomes higher, first the number of moving atoms in the surface region and the sputtering coefficient will increase. However, at higher ion energy, more ions will penetrate deeper into the target material, and the probability of energy transfer back to the surface atoms is low so that the sputtering coefficient will tend to a constant value and, at higher incident energy, will even decrease. This is also the case if the incoming ion is aligned with an open row between atoms in a crystalline target in a process known as “channeling”; for example, ions striking the target in Figure 61.1 from directly above the surface can enter the open channels and penetrate to deeper levels, leading to a lower value of Y . 4. Momentum transfer will be maximized if the ion mass is equal to the target atom mass. There is however a further factor related to the electronic structure

Eb

4

3

2

1

0

10

20

30

40

Sputtered atom energy (eV) Figure 61.2 Typical energy distribution of sputtered ions. E b is approximately half the surface binding energy. (Source: Based on early work by Anderson and Wehner [13].)

5

6

61 Thin Films: Sputtering, PVD Methods, and Applications

of the ion species, which influences the ion range; noble gases with filled electronic outer shells have higher sputtering ratios than other nearby elements in the periodic table owing to increased stopping power and therefore more energy deposition near the target surface. Sigmund [14] encapsulated most of the above factors in a theoretical model of the sputtering process, which allowed sputtering coefficients to be predicted reasonably well. The basic equation for the sputtering coefficient contains parameters that include kinematic factors (taking into account the mass of ion and target atoms), stopping power (dictating the range of the ion in the target), and a term related to the reciprocal atomic binding energy of the target material (i.e. energy required for the surface atom to escape). These equations provided a good analytical prediction method for values of sputtering coefficients, but, in later years, molecular dynamics simulations have been developed in which many atom-target collisions are followed, each set with randomly varied impact parameters. An average of ejected atoms per ion for each of the many sequences followed gives the sputtering coefficient. Probably the most well used of these methods is TRIM [15, 16], which was later developed as SRIM [17]. Much of the sputtering data is reported in the form Y divided by (1 + 𝛾). The term 𝛾 is the secondary electron emission coefficient that is the number of electrons ejected from a surface when it is struck by a positive ion. The ammeter measuring the ion current will record the current from negatively charged electrons, leaving the surface as if they were positive ions arriving. The measured current I m is thus the sum of these two currents: Im = It + 𝛾It that means It = Im ∕(1 + 𝛾) where I t is the true ion current. Many sputtering measurements have been made without special current detectors and therefore measure 1 + 𝛾 times the true ion current. Any Y value obtained using I m must therefore be divided by 1 + 𝛾 to obtain the true atoms/ion sputtering coefficient. Measurement of Y is usually done by recording the weight loss m of a target for a given total charge Q = (I t × t) or by collecting sputtered material on a quartz crystal oscillator where the frequency change is a known function of the increase in weight of the sputtered film. The number of atoms is then mN A /M where N A is Avogadro’s number (6.022 × 1023 ) of the sputtered material having atomic or molecular weight M. The number of ions is Q divided by the electronic charge “e,” so that Y = mN A ∕M∕(Q∕e) Figure 61.3 shows a typical Y vs. E curve for Ar+ ion sputtering copper. There is a threshold energy at about 20 eV, a linear increase in Y up to about 600 eV ion energy as the near-surface collision cascade develops and then evidence of a turn toward a maximum level as the ion range increases to form deeper collisions, which are too far from the surface to contribute to additional sputtering. For energy above 40 keV, the

Sputtering yield (atoms/ion)

61.1 Introduction

6

4

2

0 0

10k

20k Ion energy (eV)

30k

Figure 61.3 Sputtering coefficient Y vs. energy for Ar+ ions on copper using data from Ref. [9].

values of Y start to decrease. This is the energy where the elastic collision probability is reduced, and also the argon ions entering the material with their orbiting electrons have velocities similar to those of electrons orbiting the copper atoms. In this situation the “inelastic” or electronic energy loss predominates. This inelastic energy loss per unit transit distance is lower than elastic losses were, thereby allowing ions to penetrate the material to even deeper levels before slowing down to a deeper elastic collision regime. Hence near-surface collision events leading to sputtering are reduced. There is a general guideline that, for an ion of atomic number A (i.e. 40 for argon), the value of Y begins to fall at energies above about A keV (i.e. 40 keV for argon). If you are designing a system to sputter at a high rate, an argon ion energy of up to 40 keV is therefore the best to aim for. Note that the value of Y already exceeds 5 atoms/ion for argon bombardment of copper at 1 keV, which is a reasonably high value. It is possible, in principle, to increase this sputter rate further for a target that remains flat by changing the impact direction of the ion beam from normal to the surface to an angle of 𝜃 = 40∘ to normal. This brings moving atoms in the cascade between the surface and maximum range of the ions (Rp ) to a lower depth below the target surface of Rp cos 𝜃 giving a higher energy deposition nearer the target surface and thus a higher chance of liberating a surface atom. A typical curve for the change of Y with 𝜃 is shown in Figure 61.4. However, whereas the normal bombardment of most targets produces a cosine distribution of sputtered atoms, the non-normal bombardment will produce a different distribution, often with a peak at the specular direction. Substrates would need to be moved in an appropriate way to average out the sputtered flux at the sample and thereby produce a coating of uniform thickness. Another uniformity problem arises when sputtering single crystal and even polycrystalline materials. Where incident ions are aligned to enter between atom rows

7

61 Thin Films: Sputtering, PVD Methods, and Applications

6 Sputtering yield (atoms/ion)

8

5 4 3 2 1

0

10

50 20 30 40 60 Angle of incidence (°)

70

Figure 61.4 Typical sputtering coefficient Y vs. angle of incidence Θ of ion beam to the normal to the target surface.

going into the crystal from the surface (so-called channeling), e.g. the green ion in Figure 61.1, it is possible that it penetrates more deeply. The deeper cascade leads to some atoms striking the lower atoms in an atomic row in an upward direction. Momentum transfer along such a row of equal mass atoms is highly efficient, and the collisions tend to focus to transfer energy into the row direction. The final surface atom of all these sequences is ejected into space in the row direction. The sputtered atom distribution is now far from cosine indeed; as shown by Anderson and Wehner, by mounting a hemispherical collector with a hole to allow the ions through, dense deposits of sputtered material will form a specific spot pattern showing the row directions from which the crystal face of the target can be identified [13]. Having chosen the optimum energy for highest sputter rate, a further option is to choose a different species of ion. Almén and Bruce [18] showed that the changes in Y for copper, silver, and tantalum targets follow a series of periodic rises and falls with the ion species where the noble gas ions (Ne, Ar, Kr, Xe) always have the highest Y value in the period. The results are again related to the penetration depths of the different ions in copper and the ion-copper mass ratio. In practice however the availability and cost of non-noble gases becomes a factor, and the choice of ions is usually limited to argon, krypton, or xenon. When sputtering an alloy target comprising elements A and B, one of the species usually sputters more quickly than the other. If B is more quickly sputtered, the nearsurface composition of the alloy becomes depleted in B, and subsequent emission of B will be reduced, while the surface will be enriched in A. This should be kept in mind especially in surface sensitive analysis such as X-ray photoelectron and Auger spectroscopy. Unless there are chemical reactions, such as oxidation of one of the elements, an equilibrium situation will eventually be reached, and the correct stoichiometric sputter rates for the alloy will be achieved thereafter. Exposed to air almost all materials form surface oxides, which do not form an initial coating,

9

61.1 Introduction

(b) Ion ranges Ion range = 977 Å Straggle = 369 Å

Skewness = –0.1645 Kurtosis = 2.6505

800

Implanted profile after sputtering to 200 nm depth

600

400 Implanted profile without sputtering

200

8 × 104 104

4 × 104 2×

104

0 0

50

100 150 Depth (nm)

200

250

(c) Concentration (a.u.)

6×

Layer 1

(Atoms/cm3)/(Atoms/cm2)

10 × 104

Concentration (a.u.)

(a)

800 600 Implanted atom concentration on the surface

400

200

0A

-Target depth-

0 2500 A

0 0

Figure 61.5 Accumulation of implanted atoms in the target. Model system – implantation of 30 keV C+ into Si. (a) SRIM ion depth distribution [17], (b) implant depth

20 40 60 80 100 120 140 160 180 Sputtered depth (nm)

profile without and with sputtering, and (c) implant accumulation on the target surface as the implantation goes on and more and more material has been sputtered.

representing that of a pure material. The use of a shutter to prevent sputter deposition during this initial period is essential if the composition of the first layers of a coating on a substrate is important. This shutter is also important when more than one magnetron, each with different target materials, is used to form an alloy coating. If we look at Figure 61.5 and consider stages of the initial bombardment then, as ion bombardment starts, there is a certain probability that any ion will reach a certain depth up to a maximum range 𝜆; this is an approximate Gaussian type of probability distribution as shown in Figure 61.5a. During sputtering the surface recedes, thus exposing already implanted atoms and allowing a deeper implantation (as referred to the original surface). The implanted atom profile is modified by this process, and surface atom concentration grows and reaches the saturation concentration, given that the implantation is continued long enough. Of course the argument is simplified because the probability curve will change as the number of trapped implant atoms increases, but this should be a small effect as the total number of implant atoms is only of the order of 4 at.%. A similar argument can be used for buildup of displaced atoms (damage), which can also influence the value of Y . A fuller mathematical description of this process is given by Carter et al. [19]. Early experiments by Colligon et al. [20] for N2 ions bombarding a radioactive gold target show that Y , measured using a sensitive radioactive count to measure the sputtered gold, can change by more than 5% over this initial period of bombardment. Similar periods for stabilization of Y have also been found by Anderson and Bay [21] in situations where either ion species or ion energy is changed.

10

61 Thin Films: Sputtering, PVD Methods, and Applications

Another factor that can change the number of atoms per unit area per second arriving at a surface being coated is the development of surface roughness. This will be considered in Section 61.3 where the application of sputtering to depth profile analysis of materials is discussed. 61.1.3 Sputtering Methods 61.1.3.1 Ion Beam Systems

Sputtering of a material, using a beam of ionized particles from an ion source to bombard the target material to form the coatings, is a well-tried film-deposition method. The technique offers good control of the process parameters and allows these to be changed in a controlled way. Thus fundamental data on the effects of changing these parameters on the nucleation and growth of a coating can be obtained. In this type of system, the ion species, energy, current, and vacuum conditions can be monitored. The method does not, however, lend itself to the production of larger area coatings. An example of such a system is shown schematically in Figure 61.6. This type of system is particularly useful for production of a wide range of compositions of a multi-element coating so that its properties can be optimized for a particular application. A typical application of this type of study was a study of retention of nickel in nickel–chromium alloy used for dental implants [22]. Many patients are allergic to nickel so it is important to find the safe range of compositions of the alloy for which Ni does not leech out into the mouth. Samples with different concentrations of Ni in a Ni–Cr sputtered alloy were tested in phosphate-buffered saline solution (serving as an artificial saliva solution) and indicated that the nickel content in the alloy must remain below 90 at.% (see Figure 61.7). Another example of the versatility of the ion beam method is demonstrated by Schalk et al. [23] when exploring properties of TiAlON formed by sputtering in (a)

Ar+ Assisting ion gun 200−600 eV 0−10 mA

(b)

Sputtering ion gun 1.25 keV 50 mA +

+

+

+ + +

+ + +

+ + + +

N2+

Co Ti

Deposition shutter

mp

+

Heated up to 800 °C substrate

+

os ite an targ d Si et

Partial reactive gas pressure

Figure 61.6 (a) Schematic diagram of dual ion beam coating system. (b) Photo of the authors’ system in operation.

Released nickel (ng/cm2)

61.1 Introduction

3 × 105

High nickel release at high nickel content above 90 at.%

2 × 105

1 × 105

0 60

70

80

90

Nickel content (at.%) Figure 61.7 Ion beam-sputtered Nix -Cr samples with increasing Ni content to find maximum safety level for which Ni release in artificial saliva does not occur. (Source: Data from Ref. [22].)

different gas partial pressure ratios. This material has many potential applications owing to its good mechanical properties, tunable optical properties, and chemical stability. A broad scan of deposition parameters could be obtained by using magnetron sputtering having a fixed oxygen partial pressure and increasing the nitrogen partial pressure. The highest nano-hardness of 23 GPa was found for an (Al + Ti)/(O + N) ratio of 0.78, and, with increasing oxygen content, a decreasing refractive index and extinction coefficient were found using spectroscopic ellipsometry. Other work on properties and applications of these coatings for cutting tools has been reported elsewhere [24–26]. A broad ion beam source used by Harper and Kaufman [27] and developed initially as an ion thruster to move satellites in orbit is ideal for this type of fundamental study and has been used for many coating studies since 1980. Such a source, shown schematically in Figure 61.8, can produce 500–1500 eV ion beams of several milliamps/cm2 with good uniformity over an area of diameter of 75 cm. Electrons are emitted from a hot filament into a cylindrical chamber with an inner cylindrical shield and outer cylindrical magnets. A well-aligned two- or three-grid lens system produces an extraction field, which penetrates into the plasma and extracts ions to a crossover focus between the grids. The ion beam is thus a set of small diverging beamlets that merge into each other to produce a 75 mm diameter spatially uniform ion beam at the sputter target. Two other ion beam systems, liquid metal ion sources (LMIS) (also known as a focused ion beam [FIB]) and a cluster ion beam, will be discussed in Sections 61.3 and 61.5 as these have other specific applications for micro-etching and depth profiling, respectively.

11

12

61 Thin Films: Sputtering, PVD Methods, and Applications Electron emitting filament

External magnet

Gas

Plasma

Ion beam

Accelerator grid

Figure 61.8 Schematic diagram of a Kaufman source.

The ability of ion beam sputtering to control deposition parameters, to allow an easy method to vary coating composition, and to find optimum properties of multicomponent coatings makes this method extremely valuable. However, the size of samples that can be coated and the relatively low film growth rates (order 1 μm/h), although important for fundamental work, limit the use of this coating method to such studies of new coating systems and for thin film coating of relatively small area samples (about 1–5 cm dia). The development of ion plating systems, where an electron beam is steered onto the sputter target causing evaporation of material into a glow discharge region, followed an idea formulated in a UK patent by Berghaus in 1938 [28, 29]. The practical application of this technique, illustrated schematically in Figure 61.9, was developed by Mattox in the 1960s [30]. In this system the coating material is evaporated using an electron beam and enters the discharge region. Although electron beam evaporation produces low Figure 61.9 Schematic diagram of an ion plating system.

Biased susbstrate Growing film – –

–

Evaporated material

Water cooled crucible

Electron source

61.1 Introduction

energy evaporated material, many of the emitted particles in this system are ionized in the discharge and gain additional energy from the applied electric field before reaching the negative substrate. The additional arrival energy generates a dense coating with good adhesion. The next developments, which are fully described in a review article by Anders [31], were arc ion plating [32], steered arc ion plating [33, 34], and arc [35] deposition systems. The first two of these methods produced coatings at a high rate but with added condensed droplets of material, which became embedded in the coatings and caused local stress and material weakness. The steered arc arrangement simply has an added magnet under the target that is moved and encourages the arc to move also, thereby promoting more uniform use of the target material. In the filtered arc, a magnetic coil arrangement is used to draw out the ionized material to form a deposit away from the region with the droplets. This provides good quality coatings but increases the system complexity and reduces the material flux at the substrate and hence the coating rate. For many applications in industry, magnetron sputtering has offered the best solution. The principal of operation is illustrated in Figure 61.10. In operation, electrons start to move away from the negative cathode, but the magnetic field provides a trap, and the electrons are constrained to spiral around the magnetic field lines as they move above the negative target. Electron collisions with gas atoms result in ionization with the positive ions, then accelerating to the target to cause sputtering and (importantly) secondary electron emission that provides a source of electrons to replace losses and maintain the discharge. These original magnetron systems provide high rate deposition of coatings but limited ion bombardment of the growing film. Window and Savvides [36, 37] realized that, by using north and south poles of the magnets of different strength, the field lines could move away from the cathode and pass through the substrate. This, so-named “unbalanced magnetron,” allows positive ions to be steered toward the substrate that, if electrically floating in the plasma, is usually at a voltage of about −30 V. Thus considerable low energy ion bombardment accompanies the film deposition, and, as we shall discuss in Section 61.2, this densifies and can modify the coating microstructure. Figure 61.10 Magnetron sputtering system.

Coated substrate

High density plasma

N

S

N

13

14

61 Thin Films: Sputtering, PVD Methods, and Applications

N

B

N Electron “Race track” (electron confinement in front of target)

S

Target Position of magnets (behind target) N

B

E N S

S N Cross section

N S

Magnet polarities

N Front elevation Figure 61.11 Racetrack on magnetron-sputtered target. (Source: courtesy Hauzer Techno Coating.)

In all magnetron systems the most intense region of ion bombardment is about halfway between the center and outer magnet, i.e. below the region where the magnetic flux is parallel to the sputter target. This means that the target is eroded by sputtering at a higher rate at this point than elsewhere, leading to a deepening channel, known as the “racetrack.” This results in poor utilization of target material and also causes changes in the spatial distribution of sputtered material, which further change for different target–substrate separations because the flat target surface becomes a valley section (see Figure 61.11). Mahieu et al. [38] have developed a model for a specified racetrack profile for a 5 cm diameter circular magnetron sputtering onto a 40 cm diameter substrate. This predicts well the flux (and hence thickness profile) of arriving sputtered particles for pressures between 0.3 and 1 Pa and target–substrate distances in the range 7–16 cm when compared with experimental data. It also allows the energy and direction of atoms reaching the target, which can influence film microstructure, to be calculated. Takayuki Iseki [39] has shown that better target utilization can be attained by introducing a system to move the magnets located below the target material. For some coatings, comprising several different materials or their oxides, special sputter targets are required, which can prove to be expensive. By the use of powder targets, the composition of the coating can be changed to produce optimum coating properties, and, once the specific composition of these is established, the target material utilization is improved. This offers a good method for identifying the required target composition, but, in general, the use of powder targets can offer some difficulty. Minami et al. [40] have used DC planar magnetron sputtering of powder targets to form transparent conducting AZO films using a target comprising dopant powder of

61.1 Introduction

Closed megnetic field line

N

S

N

Ar

S

S

N

N

S

S

Sample holder

N

S

N

Magnetron

Figure 61.12 Closed-field unbalanced magnetron system. (Source: Courtesy of Teer Coatings Ltd., MIBA Group.)

Cr or Co added to a mixture of ZnO and Al2 O3 powder. They showed that the chemical stability of transparent conducting AZO films could be improved by co-doping Cr or Co without significantly altering the original electrical and optical properties. Kelly et al. [41] have used the flexibility of powder target composition to study in more detail the formation of AlO/Zn2 O3 transparent conductive oxides at different doping levels using a pulsed DC magnetron system. Their results show that, after annealing in a controlled atmosphere, well-bonded defect-free films with suitable electrical and optical properties can be formed. For improved coating uniformity it is necessary to provide suitable sample movement with respect to the sputtered flux of coating material. A good solution, which is ideal for coating machine tools, is the “closed-field unbalanced magnetron system” (CFUBMS) developed by Teer et al. [42] and shown schematically in Figure 61.12. In this system there is a circular array of four equispaced magnetrons with adjacent magnetrons having magnets with polarity such that the north of one is next to the south of the neighbor. There is thus a completely closed magnetic field around the system, which traps the charged particles in the system, intensifies the discharge, and increases the ionization of sputtered material to provide additional ion bombardment of the growing coating. Samples are mounted on a carousel at the center, each component rotating about its own axis as the whole set rotates about the central axis of the carousel. Although magnetron sputtering is already used commercially for production of conducting coatings, problems arise when forming insulating films such as oxides. For example, when forming TiO2 coatings, the oxygen flow into the system begins to interact with the titanium target to form an insulating Ti oxide layer, initially building at the edges of the Ti target where the ion bombardment is weakest and then spreading across the whole target. The insulated sections charge up and lead to arcing and

15

61 Thin Films: Sputtering, PVD Methods, and Applications

Oxygen partial pressure

16

4

3

1

2 Oxygen flow rate

Figure 61.13 Illustration of target oxidation hysteresis problems (see loop 1-2-3-4-1) for sputtering a metal target.

a reduced TiO2 sputtering yield, which slows down the coating growth rate. This is shown schematically in Figure 61.13. Figure 61.13 shows that the oxygen partial pressure suddenly rises from point 2 to 4 in Figure 61.13 (a region where the TiO2 deposition rate falls) and then continues to rise at a lower rate beyond point 4. When the oxygen flow rate is reduced, the partial pressure reduces via point 4 to point 1. Sproul et al. [43] have reviewed this problem and have identified a few solutions. One solution is to monitor the partial pressure of oxygen in the system and use this value to set the flow controller feeding in this gas. After some preliminary calibration a working flow level corresponding to a point near point 3 in Figure 61.13 can be set, which not only maintains a high Ti sputtering rate but also provides enough oxygen to form the metal oxide. Another method, proposed by Schiller [44], is to use an optical emission monitor tuned to a characteristic Ti emission line in the plasma and take the signal from this to control the gas inlet valve. An alternative solution to the charging problem has been the addition of a negative voltage pulse to the sputter target. This process is known as “pulsed DC magnetron sputtering” and is described by Schiller et al. [45]. During the negative pulse the target is bombarded by a higher energy ion flux, which removes the insulating layer. The system works well also for sputtering two components to form an alloy where each material is sputtered from one of two magnetrons. Here a single pulse unit can be shared; i.e., the negative cleaning pulse can be switched between the two cathodes. In recent years commercial sputtering systems with rotating cylindrical cathodes have been introduced [46]. In these systems, as the surface of the vertical cylinder rotates, any charged material moves out of the ion beam and loses this charge before returning briefly into the sputtering ion beam. Arcing is thereby minimized but, of course, specially shaped, and expensive target cathodes are required.

61.1 Introduction

1800 1600 Discharge voltage Discharge current

–500

1400 1200 1000

–1000

800 600

–1500

400

Discharge current (A)

Discharge voltage (V)

0

200 –2000 –50

0

50 100 Time (μs)

150

0 200

Figure 61.14 Typical HIPIMS pulse with associated current waveform. (Source: Courtesy of Hauzer Techno Coating.)

A more recently developed sputtering method is high power impulse magnetron sputtering (HIPIMS), originally described by Kouznetsov and others [47] and further reviewed by Anders [31], Helmersson [48], Konstantinidis [49], and Alami [50]. In this the system geometry is much the same as for conventional magnetron sputtering, but the power supply now delivers a short (of order 50 μs) pulse of negative voltage (of order 0.6–1 kV) to the sputter target (Figure 61.14). The intense ion current that subsequently bombards the target material heats the surface and generates a much higher flux of ejected material than in conventional magnetron sputtering. A substantial fraction of the ejected material becomes ionized. This has an advantage, again as we shall see in Section 61.2, of providing additional bombardment of the growing film mounted on the anode surface. However, some of the ejected atoms that are positively charged are pulled back to the negative cathode, and the deposition rate at the substrate is reduced. The intense discharge can lead to frequent arcing, but the latest power supplies now have sensors that react to increases in current and reduce power to prevent these arcs from forming. Whereas in conventional magnetron sputtering most of the species are argon atoms, in HIPIMS, Ehiasarian et al. [51] has shown that target species are also ionized, sometimes with two or even three charges. Other versions of HIPIMS-type pulsed systems have been used to minimize the arcing problem by using high-frequency pulses, in particular the modulated pulse power magnetron sputtering (MPPMS) of Lin, Moore et al. [52] applied to two facing unbalanced magnetrons in the closed-field arrangement. Sputtering of Cr films in an Ar/N2 gas indicated a much higher percentage of metal ions compared with sputtering using pulsed closed-field and conventional closed-field systems. The MPPMS method also produced ions with a peak ion energy of 4 eV in a narrow energy band of 0–30 eV that, as we shall see in the next section, is a good energy range for ion-assisted deposition because it provides added energy without significant re-sputtering of coating material. Typical MPPMS pulses are illustrated in Figure 61.15.

17

18

61 Thin Films: Sputtering, PVD Methods, and Applications

MPPTM Technology utilizes: Modulated Pulse Power Multi-step voltage pulse ▪ First step – ignition of low power discharge ▪ Second step – low power discharge ▪ Third step – transient stage from low power discharge to high power discharge ▪ Fourth step – high power discharge

V

t0

t1

t2

t3

t4 t 5

0 1

2

3

4

Time

VLP VHP

(1)

VLP / (t1 – t0)

(3)

(VHP – VLP) / (t3 – t2)

Figure 61.15 Typical pulses in a magnetron pulsed power system. (Source: Courtesy of Hauzer Techno Coating.)

A comparison of ionized fractions of species between conventional DC sputtering and maximum power point (MPPMS) shows significant differences [53]. Whereas, for DC sputtering of Cr by Ar ions at 5 kV power, the ionized percentage of Cr+ in the Ar–Cr discharge is about 25%, the ratio for the MPPMS system becomes 50 : 50 at 1.2 kW power and the Cr percentage even higher at powers above this. In addition, the MPPMS system shows a small fraction of double-charged Ar and Cr ions [54]. As we noted in Section 61.1.2, bombarding with the ions of the same mass as the target species leads to more efficient momentum transfer and increased sputtering rates. From a commercial coating point of view, these plasma methods with suitable substrate movement provide uniform layers of a wide range of materials at deposition rates of order a few microns thickness per hour. The ability to form ions of the depositing species in HIPIMS and MPPMS systems provides more control of added energy to the growing film. However, these ions must traverse the plasma from sputter target to film substrate, which means there will be collisions and energy loss. This can occur and differ as a result of changes in working pressure or indeed changes in gas composition as shown by Bowes and Bradley [55] who found over 100% increase in deposition rate when using a Kr/O gas mixture compared with Ne/O. Another factor, often overlooked, is a change in the secondary electron emission coefficient of the sputter target for different ion species or when the nitrogen pressure or composition of a multi-element target changes during sputtering, as shown by Haijuan Mei et al. when sputtering a Mo–Cu–V target by nitrogen in an HIPIMS system [56]. In this work both the composition and mechanical performance of the sputtered Mo–Cu–V–N changed. Schmidt et al. have compared the properties of DC magnetron-sputtered and HIPIMS CNx coatings and found substantial differences using inert gases, especially for

61.2 Thin Film Coatings

N2 /Ne flow ratios below 20% [57], which they attributed to changes in the plasma with HIPIMS having a higher ratio of ion sputtered to gas atom ratios. As found by Bradley [57] film properties were different not only with changes in working gas but also with the direct current magnetron sputtering (DCMS) or HIPIMS method used. Other plasma-based systems include the hollow cathode magnetron, described by Klawuhn et al. [58] and adapted for use on an ion implanter by Tonegawa et al. [59], and the plasma immersion ion implantation (PIII) system reported by Ensinger [60]. The hollow cathode system is often used in the semiconductor industry because it provides a relatively narrow beam that is useful when using a metal species for implantation of electronic devices. Tonegawa et al. [59] have also shown that the source has a better lifetime than the hot filament ion sources. Lanthanum boride cathodes last up to 150 hours before they need replacement and can produce ion currents for up to 150 hours, but other cathodes such as titanium carbide have lower lifetimes of 20 hours. Current densities are of order 10 mA/cm2 for an extraction voltage of 10 kV. The PIII system, described by Ensinger [61], is used for ion implantation and also for surface modification. An electron cyclotron resonance source generates the plasma inside the chamber containing the sample to be analyzed that is electrically connected to a supply that applies negative voltage pulses of order −2 kV. Ions from the plasma then cross the cathode dark space and bombard the sample. A recent application of this technique to improve the biocompatibility of surgical implants has been reported by Mengqi Cheng et al. [62]. Zr and N ions were ion implanted into AZ91 Mg alloy, which is used for bio-implant prostheses. It was shown that the Mg corrosion resistance was enhanced both in vitro and in vivo tests and its antibacterial properties in vitro were also improved. A biocompatible metal nitride and metal oxide layer forms near the surface that offers a good medium for cell adhesion and growth, thereby promoting osseointegration. The plasma methods are thus excellent for production of coatings on a commercial scale once plasma conditions for a new coating have been optimized. Ion beam methods are still very important for searching for optimum parameters for new types of coating, but, as mentioned earlier, coating area and rate of growth is limited. In the above section we have covered the main present-day physical vapor deposition (PVD) thin film coating methods. We now look in Section 61.2 at some other specific systems that have been used to provide and study the effects of added energy to the depositing atoms, so-named “energy-assisted deposition.”

61.2 Thin Film Coatings 61.2.1 Advantages and Some Disadvantages of Energy-Assistance

The concept of bombarding a growing coating with ions has made possible the creation of many new multi-element and uniquely structured coatings and has

19

20

61 Thin Films: Sputtering, PVD Methods, and Applications

dramatically expanded the range of commercially useful surface treatments. As shown in Figure 61.12, the energy of sputtered atoms ranges from about 5 to 50 eV with the number vs. energy curve showing a 1/Eo 2 dependence. Their energies on arrival at the substrate are far larger than those of thermally evaporated atoms that allows arriving atoms the possibility of surface diffusion to better bonding sites and, thereby, to produce better quality coatings. However further added energy can make a significant improvement to the film density and microstructure. We should remember that this energy is delivered to the substrate by a conservation of momentum collision process (as indicated in Eq. (61.2), Section 61.1.2.1) in an energy regime where elastic collisions take place and there is negligible electronic energy transfer and other losses. For a 100 eV Ar ion bombarding a titanium substrate atom, the maximum energy transfer is thus 90 eV for a head-on (zero impact parameter) collision, but there will of course be many more non-head-on collisions with somewhat lower energy transfer. Atoms and molecules of coating material that have not found sites with high binding energy can be displaced and effectively given further opportunities to move along the surface to better sites. Atoms of coating material sited above a hollow region can be recoiled into the bulk to densify the coating. This energy-assisted deposition (EAD) process is predicted in molecular dynamic models by Ferrón et al. [63]. Many of the deposition methods described in Section 61.1.3 already have a degree of ion bombardment of the growing coating, but more detailed studies using ion beams were first made by Weissmantel in 1976 [64] and 1981 [65]. The earlier work looked at production of coatings using EAD with a crossed dual ion beam system of the type already shown schematically as Figure 61.6 in Section 61.1.3. Weissmantel [64, 65] used the dual ion beam system and single substrate system in various ways. For example, for Si3 N4 deposition argon or nitrogen ion bombardment from one source could be used for the sputtering, and nitrogen from the other to provide the EAD. Other coatings studied included chromium and indium nitrides, and silicon carbide. Weissmantel also noted the possibility of improvement of adhesion by using the system to form a complicated structure to form gradient coatings that are discussed in Section 61.4. An early example of an experiment where the added energy was controlled was reported by Martin et al. [66] and showed the effect of EAD on the refractive index of zirconium oxide that increases to the value for a fully dense ZrO2 coating as the oxygen ion current density (i.e. added energy) increases. Densification causes important and usually advantageous changes in many coating properties, in particular in coating hardness, corrosion resistance, and electrical conductivity. This will be discussed further in Section 61.4 on coating applications. Harper et al. [67] gathered together a set of such changes already identified in 1984 where improvements in film properties had been found and where the data was sufficient to calculate the average added energy (Ea ) per deposited atom (N d ). Most of the experimental data showing the added energy per atom Ea /N d that caused a specific change in coating property was found to lie in the range Ea /N d from 1 to 100 eV/atom. More recent results on other coating materials were shown

61.3 Surface Roughening, Smoothing, and Patterning

also to fall within this range of added energy per atom by Martinu et al. [68] for coatings formed by plasma-enhanced chemical vapor deposition (CVD). However, Petrov et al. [69] realized that there may also be a difference in film properties if the ion energy is changed; for example, 50 eV/atom can be added using a single 50 eV ion/atom or by using five 10 eV ions/atom. Petrov’s results indicate that film properties are, indeed, changed in different ways for the same energy per atom but different ion energies. This is not unexpected because, for lower energy ions, momentum transfer collisions are more localized and the added energy occurs nearer the surface rather than at a deeper level where the coating structure has already formed. For viable comparisons of added energy effects, we really need to be sure that the whole surface of the sample has a uniform ion/atom experience, i.e. a spatially uniform ion beam and a spatially uniform atom arrival rate. Sidelev et al. [70] have shown that heat radiation from a sputter target provides a significant increase in added energy at the substrate (from 0.06 to 0.43 W/cm2 ) that changes the properties of sputtered chromium coatings. A further problem relating to the spatial distribution of sputtered atoms can arise when changing sputter ion energy. Lautenschläger et al. [71] have shown that the directional distribution comprises an isotropic and anisotropic component for Xe and Ar sputtering a Ti target in an oxygen atmosphere for TiO2 . The isotropic part increases with increase in ion energy, decreasing the incidence angle or changing from Xe to Ar ions. Most commercial coating units use complex rotation systems to ensure that an average ion/atom ratio seen by all areas of the samples being coated is constant. The situation for planar stationary samples is quite complicated as the sputtered atoms forming the coating would not always have a uniform spatial distribution so that, if the energy-assisted beam is uniformly spread across the sample, different spatial regions on the coatings will experience varied levels of added energy. This may, of course, lead to some interesting mixed phase coatings and is a topic that needs further study; in fact many of the experimental coatings formed so far may not have been produced with the optimum uniform deposition conditions. Another potential problem of ion-assisted coating is that, under certain circumstances, the ion bombardment can trigger the development of surface roughness. This can generate some useful features, but for depth profiling of implanted species, the depth resolution can be poor. Section 61.3 now looks at these effects.

61.3 Surface Roughening, Smoothing, and Patterning 61.3.1 Surface Topography

We have established that particle bombardment of surfaces leads to material sputtering, material ion mixing, and ion implantation. It should be noted that a particle

21

22

61 Thin Films: Sputtering, PVD Methods, and Applications

beam can contain charged ions, ionized molecules, or ionized atomic clusters. During ion bombardment three other parameters can be varied: ion beam energy, angle of incidence, and material temperature. In general terms, the ion beam can be neutralized or even have negatively charged particles. The first is relatively easily done with electron sources, as simple as hot electron-emitting filaments. This process is widely used for sputtering less well-conducting and well-insulating materials. Negatively charged beam production is a much more technically complicated process and only used rather rarely for fundamental science purposes or special applications. As noted earlier, the angle of incidence affects the sputtering coefficient and increases from normal incidence to approximately 70∘ to the normal, and at the higher angles the sputtering coefficient falls rather sharply. One needs to bear in mind that this dependence is, for macroscopic setting, only strictly true for atomically flat surfaces. The sputtering coefficient also depends on the sputtered material itself, the atomic species, chemical bonding, crystallographic orientation, and crystallinity and if this crystallinity can be maintained under the ion bombardment conditions. However, one needs to remember that all this can only easily be taken into account for relatively small fluences. At any significant fluence, another, well-pronounced, phenomena will be established: surface roughening, which is due to sputtering and other rather difficult to account for processes. Figures 61.16 and 61.17 demonstrate two surfaces of polycrystalline aluminum and monocrystalline silicon after high fluence bombardment. These results are for extreme fluence cases as both materials have been used as sputtering targets. Nevertheless in some cases almost similar fluences are used for depth profiling when establishment of any significant surface roughness leads to reduction of depth resolution. An enormous amount of research has gone into understanding and elimination of surface roughening. This effort was skillfully supported and summarized by Hofmann [72].

20 μm Figure 61.16 Scanning electron microscopy image of polycrystalline aluminum bombarded by 1.25 keV Ar+ at 35∘ to normal. Approximately 420 K, ion fluence 2 × 1019 ions/cm2 .

61.3 Surface Roughening, Smoothing, and Patterning

40 μm Figure 61.17 Scanning electron microscopy image of single crystal bombarded by 1.25 keV Ar+ at 35∘ to normal. Approximately 420 K, ion fluence at around 2 × 1019 ions/cm2 .

For the highest sputtering yield, we need to use energies at around 10 keV. Going to higher energies produces slightly higher yields but is associated with technical (essentially cost) penalties. The link between sputtering and surface topography development due to various phenomena was described a long time ago. Pioneering work of Cunningham et al. [73] has established that etch hillocks develop under ion bombardment and in the case of metals, (aluminum in this case, which remains crystalline under ion bombardment), dislocations can also be revealed. The latter is a very important point; one gets much less roughening in the case of pure, perfect targets. Even miniscule surface contamination, which is very difficult to completely avoid and to monitor in many cases, plays a role in surface roughness development. Difference is also observed for crystalline and amorphous samples, with crystalline materials amorphizing under ion bombardment [73]. Metal surface structure can be understood from the basic stochastic material removal and presence of crystallites. For the other class of materials, e.g. ceramics, glasses, and many semiconductors, the surface, after prolonged bombardment and when the process depends on many parameters, adopts a ripple structure. Ripple development on glass was first reported by Navez et al. [74]. Much work related to amorphous materials has followed this earlier study combined with early theories, for example, see the papers on equilibrium topography by Nobes and Carter et al. [75–77]. Ample efforts went at the time to establish and understand the mechanisms of surface roughening. Several theories have been proposed to explain ripple formation or more general surface roughening. Some models employ the sputtering process alone, while others combine it with surface diffusion, whereas still others attribute the topography changes to the relaxation of stress built in by the primary ion implantation. The first, widely acknowledged, theory was developed by Bradley and Harper [78] and used an approximation of dominance of incoming ion penetration depth over surface roughening and the importance of local curvature. The theory followed more general theoretical work on sputtering by Sigmund [79].

23

24

61 Thin Films: Sputtering, PVD Methods, and Applications

However, the curvature-mediated stochastic sputtering was unable to describe all observed phenomena, such as the absence in some cases of dependence of roughening on the sample temperature, and a more complicated theory, taking into the account the directed flux of atoms parallel to the surface and produced by ion bombardment, was developed [80, 81]. It seems that the material under bombardment can be modeled in many ways, for instance, as a viscous liquid in the model by Castro and Cuerno [82] or in a model almost purely based on near-surface elastic collisions by Numazawa and Smith [83] where it was shown that ripple structure can be established even without sputtering. The models of ripple formation are developed all the time, and one of the later ones by Smith [84] returns again to the consideration of relatively simple processes when only nonlinear effects due to dependence of the sputtering yield on the angle of incidence are taken into account and are enough to describe the structure formation. Full description of surface roughness phenomena is very complex, and one inevitably needs to take into account implantation that is occurring during the ion bombardment. This implantation phenomenon has been considered by Hofsass in his latest work [85]. Historically and rather macroscopically the surface roughening can be observed by the naked eye. One can notice that aluminum, for instance, after prolonged ion bombardment, loses its typical shine and becomes rather dull. On the other hand, for small ion fluences and corresponding low roughening, it is quite challenging to image low amplitude surface variations by scanning electron microscopy (SEM), which was an instrument of choice not so long ago. This is probably the main reason that low ion fluence effects only have been extensively studied for the last 20 years, essentially after the atomic force microscopy (AFM) technique was invented [86] and has become ubiquitous in the 1990s. It is also worth mentioning that this last 20 years coincided with the transition of approach to the roughening. Earlier, the roughening was viewed as an undesirable phenomena, and the effort was to understand it with the goal to eliminate structure development. In the last 20 years, we have used it to create nanostructures. AFM allows access to broad dimensional scales and to precisely analyze surface topography. (A detailed description can be found in Chapter 3.5 in volume 1.) Most importantly, the AFM data then permit us to analyze surface roughening at the very early stages and distinguish the roughening from the native, as polished, surface structures. We have to stress that nowadays flatness of polishing allows atomic level perfection to be achieved. It can be seen (Figure 61.18) for routinely polished silicon (this rather more than 20 years ago) with peak-to-trough roughness less than 1.5 nm development of amplified roughness after ion bombardment (Figure 61.19). An increase of ion energy in this case leads to much faster development of correlated periodic structure. The structure develops in the latter case at much lower fluence (Figure 61.20). AFM not only provides pleasing images in this case but also allows accumulation of digital data as shown on Figure 61.21 for the further analysis and theory development. Lately, roughening, especially one- and two-dimensional (2D) periodic structures, is studied with the goal to create periodic patterns for various

61.3 Surface Roughening, Smoothing, and Patterning

1.7 nm

0.0 nm

y: 2

.0 μ

m

x:

2.0

μm

Figure 61.18 An AFM tapping mode of native (100) polished silicon [87].

27 nm

2 nm

y: 5.0

μm

5 x:

μ .0

m

Figure 61.19 An AFM tapping mode of silicon surface bombarded to a fluence 1.9 × 1019 ions/cm2 with 3 keV Xe+ . Angle of incidence 45∘ , room temperature [87].

50 nm

0 nm

y:

10 .0

μm

.0 10 x:

μm

Figure 61.20 An AFM tapping mode of silicon surface bombarded to a fluence 4 × 1018 ions/cm2 with 20 keV Xe+ . Angle of incidence 45∘ , room temperature [87].

25

61 Thin Films: Sputtering, PVD Methods, and Applications

10

2D Power spectral density (a.u.)

26

3 × 1019 ions/cm2

1

1 × 1019 ions/cm2

0.1

4 × 1018 ions/cm2 0.01

1E–3

3.4 × 1017 ions/cm2 1 × 1015 ions/cm2

1E–4

Unimplanted 0.1

1

10

Wave length (μm) Figure 61.21 2D power spectral densities of Si samples bombarded to different fluences by 20 keV Xe+ ions. Incidence angle 45∘ , room temperature [87].

nanotechnology-related applications. The study effectively started with the work of Fackso et al. [88], which described creation of periodic 2D surface structure on GaSb with the characteristic size of 35 nm. A study of topography using real-time spectroscopic ellipsometry and grazing incidence X-ray diffraction during RF sputtering of hydrogenated silicon to follow the development of surface topography has been carried out by Adhikari et al. [89]. The topography changes were seen to depend on hydrogen content in the discharge, total pressure, and RF power level. The hydrogen content in the coating was seen to decrease when the film changes from an amorphous to crystalline phase. Probably the most interesting discovery was made by Gago et al. [90] when, under 1.2 keV Ar+ bombardment at normal incidence, a silicon surface was seen to develop self-ordered hexagonal structures. The near-surface material remained crystalline. Approximately at the same time, periodic structures on metals by low energy ion beams were created by Valbusa’s group [91]. Another way to create 2D periodic structures was demonstrated in the work of Smith and Ranjan [92], where the one-dimensional periodic structure on silica is created by an ion beam and then silver self-organizing clusters decorate the rippled surface. This method creates a 2D periodic structure. The most important fact about periodic structures created by the ion beams is the ability to create structure over large areas. Applications of nanostructured surfaces and multistep preparation routes are growing at an astonishing rate (see, for instance, Pellegrino et al. [93], García et al. [94] and Araf et al. [95]). As a reflection of the fast developing field, a workshop on “Nanoscale Pattern Formation at Surfaces” has been established in 2000. The ninth workshop was in 2017 in Helsinki.

61.3 Surface Roughening, Smoothing, and Patterning

61.3.2 Cluster Beams

Low ion beam energies provide numerous advantages including very low damage and low surface roughening or even smoothing. Ions with incoming energies at few tens of eV can only penetrate a few atomic layers of the material and provide on impact mostly lateral movements of the surface atoms. Sputtering yield in this case is very low and exhibits a threshold that depends on many parameters and lies in the region above 25–30 eV for pure metallic targets [94]. Due to difficulties associated with experimental setups at those energies, data for sputtering are scarce as the main interest in the past was mostly related to spacecraft ion thrusters. One of the main problems can be easily understood from the existence of the Child–Langmuir law based on space-charge-limited current. The law essentially states that the maximum current density (number of ions/cm2 ) J max that can be extracted from an ion source is limited to Jmax ∼ Ea 3∕2 ∕m1∕2 where Ea is charged particle (ion) energy and m is ion mass. If we make a particle (ion, cluster) out of N atoms and then accelerate it to energy NEa , then our atom flux rises proportionally to ∼N 2 . At the same time energy per atom remains very low. Work in this area was conducted by many around the globe for countless years, but the vision and persistence by Isao Yamada in gas cluster ion sources (GCIS) in the 1990s [96] should be acclaimed to bring the gas cluster ion beams (GCIB) technology into the research and associated technology focus. The basic set of GCIS elements can be seen in Figure 61.22. The application of the de Laval supersonic nozzle leads to creation of supersaturation and condensation of working gas (argon is shown) into big gas clusters containing thousands of atoms or molecules. Skimmer and aperture allow pressure to be balanced for differential pumping and maintain workable pressure in the electron ionization chamber. Clusters with velocity along and close to the source axis are ionized in the ionization chamber and accelerated to the desired energy. The next important element is filtering, which will define a relatively narrow final energy spread per atom. In the example shown this is performed by a Wien filter and bend plates. The illustrated example is Ar GCIS schematics 1. de Laval Nozzle 2. Skimmer 3. Aperture 4. E.l. source 5. Condenser lens 6. Alignment plates 7. Wien filter 8. Beam monitor 9. Bend plates 10. Quadrupole 11. Objective lens

P0

P1 P2

P3/PSEC

PSAC

Arn

Ar

1

2

3

Arn+

4

5

6

7

8

9

10 11

Figure 61.22 Schematic diagram of gas cluster ion source. (Source: Courtesy of Kratos Analytical Ltd.)

27

61 Thin Films: Sputtering, PVD Methods, and Applications

intended for focused cluster beam sputtering, and the rest of the components allow the cluster beam to be focused and scanned. In the broad beam GCIB systems, the cluster beam is sometimes charge neutralized after mass filtering if the intention is to direct it onto an insulating surface. One of the big challenges with GCIS is associated with balancing performance and handling differential pressure Pn . Cluster beams can be used to smooth the surfaces, for surface etching, surface cleaning, surface sputtering, deposition, shallow doping, and changing surface biocompatibility. More information about the cluster beam sources and application areas can be found in the review papers by Yamada [97–99]. The nonlinear effects during cluster–surface interaction lead to low near-surface damage and potentially high sputtering yield, especially for sputtering of heavy molecules. For SIMS the utilization of GCIBs offers less complex molecule fragmentation. The use of other than noble gases clusters (CO2 , for instance) enhances ionization efficiency and also improves pumping of the system [100]. While total energy supplied to the surface during GCIB bombardment is relatively high, the energy per atom is rather low and is, essentially, beam energy divided by the number of atoms. This needs to be taken into account when choosing the application. For instance, ion sputtering by GCIB strongly depends on cluster energy (see Figure 61.23).

0.9

1

0

5

10

15

20

0.8 Ion sputtering yield (ion/cluster)

28

0.7

0.1

0.6 0.5

0.01

0.4 0.3 0.2 0.1 0 0

5

10

15

20

Cluster energy (keV) Figure 61.23 Ion production yield during GCIB profiling of 50 nm Ta2 O5 thin film. The inset represents modeling results. (Source: Courtesy of Kratos Analytical Ltd.)

Relative atomic %

61.3 Surface Roughening, Smoothing, and Patterning

40

40

30

30

20

20 Ti 2p

10

N 1s

Cr 2p

10

0 0 (a)

100 Depth (nm)

200

Cr 2p

0 0 (b)

Ti 2p

N 1s

500 Depth (nm)

Figure 61.24 XPS depth profiling of TiN (7 nm thickness)/CrN (3 nm thickness) multilayer coating: (a) depth profiling by 500 eV Ar+ ion beam and (b) depth profiling by GCIB 20 keV Ar2000 + . (Source: Data courtesy of Kratos Analytical Ltd.)

An additional benefit of GCIB profiling is low near-surface mixing. This allows successful ultrathin multilayer profiling without losing the resolution with depth. High-resolution depth profiling is shown in Figure 61.24. It should be noted here that the surface roughening during a depth profiling analysis is a protracted standing issue and the long-standing solution, so-named Zalar rotation, was proposed [101] and implemented by many authors quite a while ago. One needs to look very carefully into the pros and cons of both the methods described above if very high depth resolution is required. Cluster beam sputtering and surface smoothing provides many technological opportunities where ultimate surface smoothness is a must for the ultimate performance, for instance, in high-gradient RF technology [102]. Argon cluster beams have also been used to sputter InP where preferential removal of phosphorous at the start of bombardment results in the formation of discrete self-organized plasmonic particles of indium. The resulting devices can be of use for surface-enhanced Raman scattering of visible light and could be the basis of optical switches.

™

61.3.3 Surface Topography: The Consummation

In the above overview it has been shown that the ion sputtering is a fairly complex process. Indeed, just very small compositional variations on the levels of impurities can lead to huge sputtering yield changes. Sputtering yield amplification described by Berg et al. [103] just by itself can lead to surface structuring, especially at the low fluences, if the impurities have an inhomogeneous distribution. One needs to bear in mind that in a polycrystalline system, many low concentration constituents tend to segregate to the grain boundaries. This segregation, even at the miniscule overall concentration, can lead to either desired or unwanted surface structuring by itself. It is also difficult to predict and account for the temperature effects. It is worth remembering that the surface under ion bombardment

29

30

61 Thin Films: Sputtering, PVD Methods, and Applications

is a system that is very far from the thermodynamic equilibrium condition. Still the overall sample temperature plays an important role, and it has been shown that silicon bombardment with silicon self-ion produces surface ripples at 100 K, while the surface remains smooth at a temperature of 300 K. Nevertheless, over the years we have developed the ability to suppress the unwanted topography development and smooth the surface by ion beams. We also can create nanostructured, periodic patterns. This is a hugely important development. We can thus use ion bombardment for many ends; it is much more unclear if we can confidently define the basic set of processes responsible for the observed phenomena. However the general concept of applying ion-surface interactions to industrially useful surface modification processes is well established as outlined in Sections 61.4 and 61.5.

61.4 Ion-Induced Property Changes with Potential Applications

There is a wide range of coating applications based on ion-surface interactions. Some of the major areas of application of such coatings are given in Table 61.1. The reader who is interested in the full range is referred to a thorough 95-page review with 694 references by Jain and Agarwal [104] in which there is a discussion of the many effects of ion-surface interactions that have been studied up to 2011. A review of advances in vacuum science and applications by Mozetic et al. [105] is also of interest. The market value for surface coatings is estimated to be of order USD 12 billion per annum, but this is only part of the ion-surface market as ion implantation, used in the process to form electronic devices, and the use of ion beams for surface analysis (SIMS, Rutherford backscattering, elastic recoil detection) at least doubles this estimate. Some examples of material properties that can be controlled using ionbased methods will be described in this section, and examples of products already on the market will be given in Section 61.5. 61.4.1 Changes in Material Properties by Ion-Based Processing 61.4.1.1 Phase Changes

The additional energy that can be provided to a growing coating by ion bombardment can often influence the phase of the material. This has been seen by Park et al. [106] for formation of boron nitride when depositing boron during simultaneous bombardment by a mixture of nitrogen and argon ions. The Raman spectra they obtained show a clear transition from non-bombarded hexagonal Boron Nitride (BN) to cubic BN as the ion current (i.e. added energy/atom) increases.

61.4 Ion-Induced Property Changes with Potential Applications

61.4.1.2 Bio Changes

Puligundla et al. have reported that a three minute oxygen and air plasma jet treatment of rape seeds grown to form sprouts significantly reduced the bio-contaminant level of these seeds [107]. At the same time the seed germination rate also improved. Similar work on cabbage seeds by Ono et al. [108] but with a stirring device to move seeds during treatment was shown to improve the bio-contaminant reduction. Pedroni et al. [109] have shown that an atmospheric pressure plasma jet applied for 60 seconds significantly reduced the level of Escherichia coli bacteria on the surface of agar plates and attributed this to the level of reactive species such as oxygen in the plasma. 61.4.1.3 Improved Adhesion of Coatings

Cathodic arcs were used by Bergman [110] first to ion-etch a substrate surface at 1 kV bias using Ti ions and then to add nitrogen ions at a much lower voltage to build up a well-bonded TiN coating. Later an “arc-bonding” process using a steered arc for the metal etch followed by unbalanced magnetron sputtering (ABS ) was developed by Münz et al. [111]. The result of these predeposition treatments was to create a gradual composition transition from substrate to coating material over a depth of about a few microns, leading to greatly improved coating adhesion properties. Similar adhesion improvement can be produced using ion-assisted deposition by changing the arriving species. The deposition starts by depositing substrate atoms, then a material X that bonds well with the substrate material, and then the film material that bonds well with X. This is a graded interface and the process is often called “gradient” deposition. For examples of these graded coatings, which can be deposited onto ̇ sintered tool materials, see Dobrza´nski and Zukowska [112], and, for a surface layer of diamond-like carbon (DLC), see the work of Voevodin et al. [113]. Voevodin et al. used a hybrid method of sputtering and pulsed laser deposition on 440C stainless steel substrates and showed that gradual replacements from hcp α-Ti to fcc TiC and a two-phase region consisting of crystalline TiC and amorphous carbon (a-C) in transitions from Ti to TiC and from TiC to DLC could be formed. In this way a superhard (60–70 GPa) self-lubricating DLC layer with a low friction surface could be bonded to the α-Ti.

™

61.4.1.4 Hardness, Stress, and Crystal Properties

These three properties of coatings are interrelated, and the resulting hardness of a material and its dependence on deposition temperature is basically a result of how the coating deals with induced stress. Limiting dislocation movement helps to retain stress and therefore hardness, which is created by generation of high-density defects, formation of grains, phases, and columnar or multilayer boundaries. Ionassisted deposition creates an ideal situation to promote these processes by providing additional atomic movement during deposition and sudden (order 1013 K/s) cooling rates. For magnetron-sputtered TiCx N1−x (x = 0, 0.15, 0.45), with changes

31

61 Thin Films: Sputtering, PVD Methods, and Applications

in composition obtained by control of the relative flow of CH4 and N, Karlsson et al. [114] showed that in coatings for different negative bias on the substrate V s from 20 to 820 V (i.e. increasing ion energy Es ), the intrinsic stress was shown to be compressive and to decrease as V s increased. A linear relation between hardness and stress magnitude was reported with H increasing with x. Nanocrystalline structure can be created in some alloy materials that contain immiscible elements, which form small size nanocrystals surrounded by a tissue material lattice. In these materials strong grain boundaries are formed in which grain boundary sliding or dislocation movement is minimized. These materials have been shown to have hardness values in excess of 40 GPa [115]. An extensive discussion of these and multilayer superlattice systems has been given by Mayrhofer et al. [116]. A typical example of such a system is the TiN-SiNx system for which, with 9 at.% content of Si, the hardness is seen to increase with ion-assisted energy [117]. This hardness, approaching 40 GPa, did not significantly decrease even after heating samples in a vacuum up to 1000 ∘ C, indicating there is minimum stress release. A high-resolution transmission electron microscopy (TEM) picture showed that the microstructure has crystalline grains of order 4–5 nm in size giving enhanced hardness as expected from the Hall–Petch relation [118, 119], which predicts that, as crystalline size d is reduced to a smaller value, the hardness of a material increases up to a maximum at d ∼ 3 nm and then decreases again for smaller nanocrystal sizes. Deposition onto the substrate at higher and higher temperature reveals more complex hardness behavior; see Figure 61.25. It is not clear at the moment what atomic processes are responsible for two-region behavior. The high stability of hardness with temperature occurs because there is little dislocation activity inside the TiN grains, and plastic deformation can only occur by grain boundary sliding. 45

Region B Region A

40 Hardness (GPa)

32

35 30 25 20 200

300

400

500 600 700 800 900 1000 1100 Substrate temperature (K)

Figure 61.25 Changes of TiN/Si3 N4 film hardness depending on the substrate temperature during deposition. Author’s unpublished data.

61.4 Ion-Induced Property Changes with Potential Applications

Dislocation activity can also be reduced by sputtering alternate layers of two materials with different lattice parameters and different Young’s moduli to prevent movement of dislocations (and hence stress reduction) across the layer boundaries. This has been shown by Wu et al. [120] for CNx /ZrN coatings formed using a dual cathode unbalanced magnetron to sputter multilayers of order 2 nm thickness. Hardness values of 50 GPa and moduli of 410 GPa were obtained when deposition conditions were chosen to form (111) ZrN texture. Riedl et al. [121] studied multilayered Ti-Al-N/Mo-Si-B coatings by varying the bilayer period from 7 to 237 nm to find optimum hardness, growth morphology, thermal stability, and oxidation for a bilayer period of 37 nm comprising 6 nm thin Mo-Si-B and 31 nm thick Ti-Al-N layers. With optimum growth rates these were stable at temperatures up to 90 ∘ C. Musil et al. [122, 123] have reported work on the evolution of microstructure and micro-stress resulting from increasing added energy to the deposition process when using magnetron sputtering of TiAlN alloy targets. They showed that the total added energy (from ions and fast neutrals) dictated the mechanical, structural, and physical properties, the latter including resistance to cracking, which required the highest added energy. Other papers by Sidelev [70] looking at Cr sputtering by heated and cooled targets and Bondarenko et al. [124] using a physicochemical model plus an experimental study for sputtering of tantalum strongly support the Musil’s concept. Another set of layered structure materials, first reported some years ago by Nowotny et al. [125–127], is in the form of bulk ternary carbides and nitrides. These are known as MAX phases because they comprise a transition Metal, a material from the A section of the periodic table of elements, and another material X, which is carbon or nitrogen. There is a specific formula for these materials: My+1 AXy where y mostly has values 1, 2, or 3. Great interest in these materials was generated by the work of Barsoum et al. [128–130] who showed that the materials have many useful properties, for instance, some retaining hardness and oxidation resistance at temperatures above 1000 ∘ C (see Table 61.2). Formation of bulk MAX phases and thin coatings requires high temperatures [131], but recent developments using PVD techniques and ion assistance have Table 61.2 Properties of Ti3 SiC2 and some other max phase materials.

Stable at temperatures above 2000 ∘ C Thermal shock resistant Oxidation resistant in excess of 1500 ∘ C High elastic stiffness (Young’s modulus in the region of 300 GPa) Superplasticity at low loads High electrical conductivity (up to ∼1.5 × 105 S/cm) High radiation damage tolerance Low friction Easy to machine

33

34

61 Thin Films: Sputtering, PVD Methods, and Applications

(a)

(b)

(c)

0002 a b _ 1100 a

a b

312

40 nm

c

312

3 nm

5Å

Figure 61.26 (a) Cross-sectional transmission electron microscopy image, (b) lattice image, and (c) corresponding selected diffraction area electron diffraction pattern from the Ti3 SiC2 sample deposited at 650 ∘ C. (Source: Courtesy of Elsevier [113].)

shown that some MAX phase layers can be produced at temperatures below 700 ∘ C [132, 133], which means they can be formed on industrially useful substrates such as tool steel without significant effect on substrate properties. To reduce the required deposition temperature for the Ti–Si–C MAX phase coating, a special heated and rotating substrate holder was designed and had three magnetrons, each with either Ti, Si, or C targets. Magnetron powers were set on the three magnetrons to deposit the Ti, Si, and C layers in the 3 : 1 : 2 proportions for every rotation. Using this method characteristic X-ray diffraction (XRD) spectra for MAX phase Ti3 SiC2 were obtained when using a deposition temperature of 650 ∘ C. A high-resolution TEM cross-sectional view of such a coatings is shown in Figure 61.26. Cr2 AlC MAX phase coatings were also produced by Vishnyakov et al. [133] on a substrate at 400 ∘ C using the dual ion beam system (shown in Figure 61.6, Section 61.1.3.1) and an elemental combined Cr + Al + C sputter target. It was found that, even by depositing the components at room temperature and subsequently annealing in vacuum, good MAX phases were formed. Annealing in air also produced MAX phase in the bulk of the coating but with a thin oxide layer that may be acceptable and even have an advantage as a barrier for some applications of this coating. Both these subsequent annealing methods lend themselves to development for production of this MAX phase on a larger area industrial scale. Extending the application of layer-by-layer deposition, thin films were deposited onto suspended graphene (directly onto a TEM grid) at 500 and 600 ∘ C [115]. Two MAX phase layers of material were deposited. The analysis is presented on Figure 61.27. It should be noted that some of the Cr:Al ratios measured in the previous experiments are closer to 1 : 1 than to 2 : 1, as shown in the table (Figure 61.27f ). This ratio is different from the reported ratio range for the formation of the M2 AX structure. Therefore, it is not clear at this stage whether the structure shown in Figure 61.26 has actually formed in the samples. Although the analysis of the diffraction pattern seems to fit the proposed structure, more chemical analysis is needed to further confirm the formation of Cr2 AlC.

61.4 Ion-Induced Property Changes with Potential Applications

(a)

(b)

(c)

(e)

(f)

(d)

EI

AN

Cr AI O C

24 13 8 6

Series norm. C (wt. %) K K K K

51.19 25.26 12.33 10.48

Atom. C (at. %) 27.42 26.08 21.46 24.30

Figure 61.27 (a) High-angle annular dark(d) Combination of Cr, Al, and C maps. field (HAADF) image of Cr2 AlC cluster: energy (e) Combination of HAADF image and Cr, Al, dispersive X-ray emission (EDX) maps from and C maps. (f ) EDX results [134]. the selected region. (b) Cr map. (c) Al map.

61.4.1.5 Friction and Wear

The majority of inserts in cemented carbide tools for wear limitation in metal cutting are coated using either CVD or PVD methods. To improve coating properties multicomponent materials such as Ti-Al-N were produced by Münz et al. [135]. Wang et al. [136] have studied the wear and corrosion properties of TiN, TiAlN, a-C:H, and CrN sputtered coatings on a WC base. An 8 mm dia Al2 O3 ball with 2N load was used in a pin-on-disc instrument in simulated body fluid (SBF) to assess the tribological performance of interacting surfaces of implant prostheses. Standard electrochemical tests were used to show that the CrN coating had the best corrosion and wear resistance with a friction coefficient about 0.23 and complete wear out after 1000 m sliding distance. This work shows again the convenience of ion beam sputtering for an initial assessment of the use of different materials, in this case of importance in biomedical studies. Stoyanov et al. [137] have devised a neat experiment to show how the mechanical properties and microstructure appear to be linked by sputtering vanadium– carbide/amorphous carbon multilayers. Increase in the amorphous (carbon) content of the coating was achieved by reducing the V–C proportion. The increase resulted in reduction in both indentation hardness H and Young’s moduli E but lower wear and friction. Ion bombardment of materials by nitrogen ions to reduce wear was first studied by Hartley and Dearnaley at the UK Atomic Energy Establishment in Harwell in the early 1970s, and work was reviewed by Hartley in 1979 [138]. The implantation of steel by nitrogen ions was shown to reduce the coefficient of friction by factors between 15% and 60%, and the improvements continued long after the original depth of implant was removed. This was because a fortuitous combination of radiation damage, formation of an oxide layer, and subsequent mobility of ions into the material maintained the improved performance of the material. A general description of “Friction: Friend or Foe” can be found in Chapter 76 in Volume 10.

35

36

61 Thin Films: Sputtering, PVD Methods, and Applications

61.4.1.6 Electrical Properties

The mass spectrometer was developed in the early 1920s by Dempster [139], which is often referred to as “the first mass spectrometer,” and Aston [140, 141]. It was important initially for the separation of isotopes, in particular U-238 and U-235. The ability to use a magnetic mass analyzer to separate specific ion species was ideal for ion implantation, which was well established in the 1950s and allowed precise doping of silicon wafers to produce semiconductor devices. The manufacture of devices on silicon continues to be a multi-billion-dollar-per-annum industry. Ion implanters comprise an ion source, usually at a voltage of order +200 kV, and a 120∘ sector magnet with adjustable edge field systems to focus the selected ions onto an earthed target comprising a rotating carousel containing the silicon wafers. Control to parts per million of dopant levels can be obtained. A review, published in 2017, of the field of ion implantation in silicon has been made by Michael Current, where he shows the continuing importance of ion and possibly neutralized ion implantation, which has been essential in meeting the challenges of smaller devices. Now the transition to nm-scale and quantum-controlled IC devices will ensure that ion implantation will continue to play key roles in many areas [142]. Another application of ion bombardment is the formation of ohmic contacts on electronic devices. This can be achieved by a process known as ion beam mixing to produce metal silicides and has been studied by Lin et al. [143]. In this process a thin layer of the metal required to form the silicide is sputtered onto a silicon substrate and subsequently bombarded by high energy ions. In this study 80 keV Xe+ and Ar+ ions bombarded molybdenum films with initial thickness chosen so that the maximum number of atomic displacements occurred at the Mo–Si interface. Figure 61.28 shows the resulting Mo and Si depth profile, obtained using Rutherford backscattering, as deposited and after an ion dose of 2 × 1016 ions/cm2 , and a target temperature of 300 ∘ C. The data shows that a region with 2Si and 1Mo (i.e. MoSi2 ) can be formed. This silicide is a good ohmic contact for very large-scale integrated electronic devices. The fundamental principles of ion beam mixing, where a thin layer of metal is sputtered onto the silicon device and then bombarded by several keV Argon ions, have been discussed in a review by Averback [144]. Martinu et al. [68] have also shown that, by varying the target bias voltage (V B ) during ion-assisted deposition, other electrical parameters, such as resistivity, also change during EAD, for example, resistivity (𝜌), dielectric loss, microhardness, and film density. Patel et al. [145] have used 140 keV nitrogen ions to recrystallize a layer of silicon above a silicon nitride layer. This demonstrates the capability of recovering amorphicity in a sample. It is a very similar technique to that used earlier by Reeson et al. [146] who used swift oxygen ions to form a buried oxide layer in silicon (SIMOX technology). 61.4.1.7 Antiferromagnetic Coatings for Spintronics

The field of spintronics, or spin electronics, involves active control and manipulation of spin degrees of freedom in solid-state systems. Exchange-coupled bilayers containing antiferromagnetic (AF) Cr2 O3 have potential applications in novel

61.5 Current Commercial Applications of Ion-Surface Processes

100 Silicon

Concentration (at. %)

Molybdenum

As deposited

50

0 0

5

10

15

20

25

30

35

40

45

50

55

60

Concentration (at. %)

100 Silicon

Molybdenum

After 80 keV Ar2+ 4.5 × 1016 ions/cm2 @ 250°C

50

0 0

5

10

15

20

25

30 35 40 Depth (nm)

45

50

55

60

Figure 61.28 Formation of MoSi2 silicide by ion beam mixing. (Source: data from Ref. [143].)

spintronic devices with magnetoelectric properties. Chang et al. [147] have studied the microstructures and magnetic properties of ion beam-sputtered NiFe/Cr2 O3 bilayers on single-crystal SiTrO3 (STO) (001) using different Cr2 O3 thicknesses on single-crystalline SiTrO3 (STO) (001) and on amorphous SiO2 substrates and have found changes in microstructure and in the preferred orientation of Cr2 O3 deposited on different substrates. A thickness dependence of the microstructure and magnetic properties of NiFe/Cr2 O3 bilayers prepared on STO substrates was found. The data has relevance to the development of spintronic devices. (See also Chapter 64 in this volume.) In this section we have looked mainly at new techniques and applications that are still under development, or less used, but have potential industrial applications. The following section looks at already established commercial applications of thin film coatings.

61.5 Current Commercial Applications of Ion-Surface Processes

This is a field of research and development that is developing at such a rapid rate that latest work listed at the time of writing will quickly become part of the history.

37

38

61 Thin Films: Sputtering, PVD Methods, and Applications

For current advances the reader is referred to many major scientific conferences where the titles of presentations illustrate the wide range of new techniques for production of materials with improved properties. These conferences include, but are not limited to, “Ion Beam Modification of Materials,” “Society of Vacuum Coaters Annual meeting,” some of the topical seminars in the “Materials Research Society” (and “European Materials Research Society”) spring and fall meetings, the International Union for Vacuum Science, Technique and Applications “International Conference on Thin Films,” and The American Vacuum Society “International Conference on Metallurgical Coatings and Thin Films.” In addition there are conferences devoted to a better understanding of present processes and the development of novel ion-surface methods such as “International Conference on Atomic Collisions in Solids” and “Ion Surface Interactions” (Moscow). In addition, a wide range of surface treatment processes appear in plasma conferences such as “International Symposium on Applied Plasma Science” and “International Conference on Plasma Science.” More specifically on tribology and wear, there are the “International Conference on Wear of Materials,” “International Conference on Tribology Technology,” “World Tribology Conference,” “International Conference on Tribology and Interface Engineering,” and IET UK annual conference on “New Challenges in Tribology.” With such a large community of scientists attending and presenting new data at these and other meetings, a visit to the conference websites and a study of technical programs are the best way to find latest developments. Only some of the more significant applications are discussed here. Ion beams have provide an ultimate tool for surface patterning if some kind of mask-like conventional photolithography or scanning electron lithography is used. This process has been used for a long time for creation of active and passive optical structures (see, for instance, Garvin et al. [148]). Later, the material structuring work has been extended and has become known as ion beam structuring. The technique is mostly used for final subtractive high-value optical element, such as lenses or mirrors, correction [149–151]. The LMIS produces an intense finely FIB and is now often referred to as an FIB. It is used routinely in the preparation of thin samples for TEM studies. A full history of the background to the development of this device starting with studies of the effect of electric fields on liquids is given by Bischoff et al. [152]. A model of its operation, which depends on the rapid expansion of a liquid metal from a heated cylinder through a fine nozzle into a high electric field region, has been presented by Forbes [153], and an improved source producing high intensity currents was described in 1975 by Clampitt et al. [154] who discovered that a blunt needle carrying the feed liquid metal gallium produced a much higher current of order 200 μA. Present-day ion sources have various needle and capillary liquid feeds and use various liquid metals including liquid metal alloys, but the most stable/popular system is based on liquid gallium. The first major technological interest in the use of the LMIS was reported by Krohn [155] for use as a possible space vehicle thruster.

61.5 Current Commercial Applications of Ion-Surface Processes

These LMIS systems have many other important applications than sample preparation for TEM. These include maskless implantation, lithography, mask repair, and nano-fabrication. An example of nano-fabrication is the production of superconducting nanowires by Sadki et al., using a gallium FIB and a precursor gas of tungsten carboxyl (W [CO6 ]) to form a W–C–Ga wire with composition 40, 40, and 20 at.%, respectively [156]. Another example is the three-dimensional (3D) components formed by Matsui et al. [157], an example of which is a micro-bellows formed by entraining an aromatic carbon-containing precursor onto a substrate at a point coinciding with the area bombarded by the FIB. The ion beam cracks the precursor leaving a carbon deposit and gaseous products, which are pumped away. By rotating the substrate and slowly withdrawing the focus point of the beam, a 3D micro-bellows with 0.8 μ pitch, 0.1 μ wall thickness, 2.75 μ external dia, and 6.1 μ height could be formed. The FIB has become a widely used technique for TEM sample (lamellae) preparation. Many examples of combined FIB and SEM instruments are available from various instrument producers. The latest technique is to use the ion beam for surface imaging instead of electrons. In this case it is possible to focus the beam down to 0.25 nm and image the top outmost material layers. The only instrument on the market currently available is produced by Zeiss SMT (see Figure 61.29).

Figure 61.29 Helium scanning ion microscope – Orion NanoFab system. (Source: Courtesy of Zeiss.)

39

40

61 Thin Films: Sputtering, PVD Methods, and Applications

Che-Wei Yang et al. [158] have used a focused He ion beam in single-spot mode to drill nanoholes in SiO2 /Si(111), with smallest hole diameter of 8 nm. A low He+ dose can then trigger the nucleation and then vertical growth of InAs nanowires. One of the many successful industrial applications of coatings is in the production by magnetron sputtering of self-cleaning and photochromic layers onto window glass. Self-cleaning windows have a coating of the anatase phase of TiO2 , which is the most photocatalytic form of this material. In sunlight, a process of photocatalysis causes the coating to break down chemically organic dirt particles, which are adsorbed onto the window. When rain wets the glass, the hydrophilic nature of the titania reduces contact angles to very low values, causing the water droplets to form a thin layer, which washes the dirt away. Pilkington Glass Company (now part of the Nippon Sheet Glass Company) created their Pilkington Activ process for coating large area glass sheets in 2001. (See also Chapter 80 in Volume 10.) An equally important product is the production of coatings for ophthalmic lenses to form antireflection and tinted layers on spectacles. A merger between the Carl Zeiss Ophthalmic Lens Division and SOLA International Inc. was made in 2004 to form Carl Zeiss Vision with coatings contributing to a company turnover of order 800 million Euro per annum. An early example of one of the coating methods is given in Ref. [159], which uses magnetron sputtering to form multilayers of coatings including TiO2 and SiO2 on sheets of glass as they are fed through an online processing system. Another important industrial user of surface coating is the engineering tool industry. A drill bit that lasts for a longer time minimizes the downtime required during replacement. This also has an environmental impact in reducing the number of drill bits thrown away and, in some cases, allows processing with reduced or zero levels of coolant. Finer dimensional tolerances of drilled holes can be attained. Teer Coatings (MIBA Group) have shown that a special TiN coating with their trademark MOST coating working with reduced coolant flow increases the drill-bit lifetime from 6 to 600 holes. In another application a Cr plus MOST coating on a die used by the Canadian Mint to stamp coins increased the number that could be stamped with quality images from the usual 100 000 to over 800 000. DLC coatings have exceptional properties similar to, but not quite as good as, diamond. They can be produced by a CVD process using hydrocarbon gases such as C2 H2 , which break up to C and H on interaction with a hot wire filament or in a plasma discharge and subsequently form a hard carbon layer containing H on the substrate. However the process requires exceptionally high temperatures of order 900 ∘ C, which are too high for DLC deposition on most materials. Aisenberg and Chabot [160] described an ion beam deposition system, which was able to deposit DLC coatings onto substrates at room temperature. This is an ion energy-assisted process with the ions providing additional energy near the substrate surface in a so-called “subplantation” process described by Robertson [161]. In this subplantation process the momentum transfer and near-surface collisions promote the formation of the DLC coating. For commercial applications, complex objects and large area substrates need to be coated. This led to the development of the system now known as the closed field unbalanced magnetron (CFUBM)

™

™

61.5 Current Commercial Applications of Ion-Surface Processes

sputtering system by Teer Coatings (previously part of the Miba Group and is described earlier in Section 61.4.1.2) but, for DLC coatings, now runs using a hydrocarbon gas in a so-named “Plasmag” system. The C2 H2 (acetylene) used decomposes on the Ti sputter target where the C and H components separate and join the flow of sputtered material. The C2 H2 gas also decomposes at the substrate. By lowering the acetylene flow for short periods, multilayers of TiC and DLC could be formed. The resulting coatings have extremely low friction and wear rates. A full description of the process is given by Monaghan, Teer et al. [162]. There are now several forms of diamond-like coatings available including the tetrahedral DLC (ta-C) with friction coefficients μ of order 0.02–0.1, amorphous (a-C) with μ of order 0.05–0.1, and Me-DLC (WC-C:H) with μ 0.1–0.2. The adhesion of the coatings when deposited directly on a substrate can have problems, but, by using a gradient coating as discussed and described by Voevodin et al. [113] earlier in Section 61.4.1.2, this problem can be overcome. Solar cells form another established area of application for multilayers coatings; some formed by sputtering, and others by the use of plasma-assisted CVD. A review of the field in 2009 has been given by Aberle [163]. Stable energy conversion efficiencies in 2009 were usually of the order 10% with a few claims of levels of up to 13%. Research and development of high-efficiency solar cells, using beam splitters, is estimated to save over 750 M Australian dollars in electricity generation for the country over the next decade in addition to triggering production and sales of energy-saving products, which will have a huge worldwide impact [164]. Another potential improvement in performance of silicon crystal solar cells has been proposed by Tang et al. [165]. It involves preservation of the cells from deterioration in performance as a result of sodium ion penetration from glass through the protection encapsulation, which uses ethyl vinyl. It has been shown that treatment of the silicon using a 2 MHz driven atmospheric plasma using argon and oxygen produces O and OH radicals. These promote formation of an oxide layer on the silicon crystal and reduce the influx of Na and the falloff in solar efficiency. Non-evaporable getters (NEGS) are used in particle accelerator systems such as synchrotrons. In these systems it is important to maintain pressures below 10−12 mbar to avoid loss of the beam from scattering by residual gas atoms. The beamline is however rather long and narrow providing a limit to pumping speed. Benvenuti and colleagues [166, 167] found that a new Ti-Zr-V coating in a vacuum system could be activated in vacuum at a temperature of about 180 ∘ C. In operation, when the beamline has been pumped to its lowest pressure using turbo-molecular pumps, further heating causes residual gases on the getter surface to be absorbed. It was shown that the use of such NEGS allowed the base pressure in the system to be reduced from 10−12 to 10−13 mbar. For subsequent activations and pumping cycles, increased temperatures up to 400 ∘ C were required. The NEG pumps residual gases that include H, CO, and CO2 . More recently Gupta et al. [168] have been able to demonstrate that oxygen on the surface of these NEGS moves into the Ti-Zr-V coating during the pumping process.

™

41

42

61 Thin Films: Sputtering, PVD Methods, and Applications

NEG coatings have been marketed by SAES Getters for many years, and their use not only extends synchrotron beam lifetimes, but also when NEGS are distributed along the inside of the beamline tube, the use of NEGS can reduce the number of turbo-molecular pumps required. Some special forms of NEGS that do not require activating are the COMBOGETTER used to maintain vacuum in vacuum insulating panels used by the building construction industry. These pump O2 , H2 O, CO, CO2 , and N2 by a chemical reaction under vacuum. NEGS contribute with other company products to the 1.66 M Euro income reported by SAES Getters in 2015. Studies by Valizadeh, Malyshev et al. at the ASTeC group at STFC Daresbury Laboratories, UK [169], have shown that the use of Ti-Zr-V alloy sputter wire cathodes, rather than the twisted wire cathodes used earlier, allows NEG coatings to be formed, which have a lower activation temperature than 180 ∘ C. In addition, by increasing the pressure during sputtering from a low to a higher value, these authors showed that a dense layer of coating could be formed adjacent to the inside cylinder wall but a more open coating structure at the pumping surface. The dense layer minimizes outgassing from the chamber wall, whereas the open-structured coating at the vacuum surface provides a much larger surface area for pumping the residual gases. An illustration of the coating system is given in Figure 61.30, and a curve showing the reduced activation temperature for pumping CO is given in Figure 61.31.

Ceramic Vacuum pump

Kr injection

Target: Ti-Zr-V Twisted and alloy wires

Solenoid

Test pieces Ceramic

Pulsed dc power supply

–

+ HV

Figure 61.30 Layout for DC magnetron deposition installation and the targets used: twisted wires and an alloy wire. (Source: From Ref. [169] courtesy of AIP.)

61.5 Current Commercial Applications of Ion-Surface Processes

CO sticking probability

1E–00

1E–01

TiZrV (twisted wires) TiZrV (alloy wire) TiZrV (alloy wire)

1E–02 140

160

180

200

220

240

260

280

300

320

Activation temperature (°C) Figure 61.31 Initial CO sticking probability as a function of NEG coating activation temperature. (Source: From Ref. [169] courtesy of AIP.)

For smaller glass tubes (17.6 mm dia, 40 mm long), Kaune et al. have shown that copper films can be formed on the inside of the glass using a similar method but with a set of magnetic coils to produce the magnetic field pattern for uniform coating thickness [170]. With the increase in requirements for temperature sensors, both for domestic and industrial applications, thin film devices using Pt are already in use [171] and marketed by Heraeus and other companies. However thin film Pt sensors can degrade in high-temperature situations above 600 ∘ C [172], and there is recent interest in the use of indium tin oxide (ITO) films as sensors that are stable up to 1100 ∘ C or even higher in some environments [173]. In a search for suitable hydrogen storage materials, Zhijie Cao et al. [174] have studied Zr-Fe-V-based alloys have found that (Zr0.7Ti0.3)1.04Fe1.8V0.2 shows the best overall properties with a reversible hydrogen capacity of 1.51 wt% and a hydrogen desorption pressure of 11.2 atm at 0 ∘ C. It appears to be viable after many load/unload cycles and may form a suitable storage system for fuel cell vehicles. In recent years ion accelerators have been used by medical researchers to assess their value for analysis and treatment of tumors and other conditions in humans. Already many such accelerators are used to produce positron-emitting radionuclides for positron emission tomography (PET) in which the radionuclide is attached to a bioactive molecule and introduced into the body. Ion beam therapy is also already used because the radiation is confined to a small region allowing precise treatment without damaging surrounding areas of the body. For a review of this field, see Denker et al. [175].

43

44

61 Thin Films: Sputtering, PVD Methods, and Applications

61.6 Quo Vadimus 2018

There are many reports of new experimental data forming the basis for new applications, and we can only give a selection of these here. An interesting new application has been reported by Yunlong Zi et al. [176] who have studied surface modification to form nanosized energy generators by creating electric power from movement of a person. The relatively new areas of development are in the biological field where coatings are needed to reduce friction for moving parts of implant prostheses within the body. As reported in Section 61.4.1.5, Qianzhi et al. [136] have made a thorough comparison of erosion and abrasion of TiN, CrN, TiAlN, and diamond-like a-C:H coatings in SBF. The TiAlN coatings had the worst tribological and electrochemical properties because of abundant products formed from hydration. The a-C:H (diamond-like) coating showed outstanding corrosion resistance, but premature delamination occurred in an aqueous environment, which might limit its application. The CrN coating exhibited superior characteristics of tribology and electrochemistry because Cr2 O3 is formed, and, from this study, this appears to be the best coating to use in the SBF environment. We noted earlier in Section 61.5 the important developments of the use of ion beams for surgical analysis. Ion implantation has many more potential applications for improving the biocompatibility of materials for surgical implants. A review by Jagielski et al. in 2006 [177] has outlined some of the achievements and remaining challenges. To reduce friction in hip joints, the surface of Ti alloy femoral heads can be modified by carbon and nitrogen implantation to give an order of magnitude reduction in friction and to encourage retention of synovial fluid [178]. To alleviate problems that arise when using Co–Cr alloy, which has good wear resistance but damages the mating polymer surface, ion implantation of the alloy has also been shown to increase synovial fluid retention [179]. Another application of ion implantation is the treatment of the Ti femoral stem by bombarding with Ca and P ions to form a compatible interface with the surrounding bone, thereby encouraging formation of a strong biological bond [179]. Ali Ghanbarzadeh et al. [180] have developed a tribomechanical model based on the thermodynamics of interfaces and the kinetics of tribochemical reactions with an aim to implement tribochemistry into deterministic modeling of boundary lubrication to point the way to reduce friction and wear. In optoelectronics Saxena et al. [182] have shown that, by sputtering a fused silica target with Si(100) pieces attached onto a Si(100) substrate and subsequently irradiating this coating by 160 MeV Ni11 , intense red luminescent and narrow size distributed silicon nanocrystals (SiNCs) having intense red emission centered at 675 nm were obtained. This could have importance in the future use of SiNC’s in optoelectronic devices. Future developments in electronics and photonics require development of high-dose Sn− ion implantation of germanium as an n-type dopant. Unfortunately,

61.6 Quo Vadimus 2018

(a)

(b)

(c)

Figure 61.32 (a) Bronze 10 kopek coin. (b) The coin with totally deleted image. (c) The same coin after oblique ion bombardment by 7 keV Kr+ ions at 70∘ to the normal. (Source: From Ref. [183] courtesy of Elsevier to be published.)

associated with such high doses, the germanium becomes porous. However, Tran et al. [182] have recently shown that, by adding a nanometer-scale silicon dioxide capping layer and implanting at liquid nitrogen temperature, the development of this porosity is significantly reduced. This opens the way for the possible production of devices such as metal oxide field-effect transistors (MOSFETs) and other devices based on germanium. In the field of archeology, a novel method for recovering details on old metal artifacts is described by Tolpin and Yurasova [183]. The data on such items was imprinted centuries ago by using a patterned stamp. The protrusions on the stamp compressed the metal to a greater extent than the hollow parts of the pattern image. This event induces different levels of stress, surface binding energy, and film density below the various features on the stamp, effects which change the local sputtering coefficient and remain after the original pattern has been worn away. Figure 61.32 shows a Russian 10 Kopeek coin that has had the image removed by chemical etching and then brought back by sputtering. This is a novel way of dating the coins. Coin collectors will need to decide if they are prepared to risk damaging pristine specimens if they want to establish dates and other details of their coins using this method. We have reported many areas where ion-surface modification and thin film coatings are already established and used by industry and many more new results that indicate possible new commercial advantages. There is definitely a strong future in this field with many new applications, and these will be highlighted in the programs of the various future conferences listed at the start of Section 61.5. The major challenges expected to be overcome appear to be in the areas listed in Table 61.3. It is difficult to rank these in importance. The first two are important for improved medical care, while the next four relate to energy-saving and energy sources, e.g. high-temperature coatings allowing engines to operate with higher fuel efficiency and lower toxic emissions. Voevodin et al. [184] have offered a partial solution to wear reduction by creating a self-lubricating coating. In this a broken paving-type structure of hard yttrium–strontium–zirconium nanocrystals forms reservoirs in the gaps that contain MoS2 . The latter is slowly drawn out during the sliding movement between mating surfaces and acts as a lubricant. A diagram of this long lifetime system is given in Figure 61.33 (Figure 61.34).

45

61 Thin Films: Sputtering, PVD Methods, and Applications

Table 61.3 Future challenges for ion-surface and thin film coating solutions.

(a)

2–5 nm MoS2 amorphous or nano-crystalline inclusions for vacuum and dry lubrication

Optmum structure for toughness improvement

3–10 nm

Surgical implant materials to improve accommodation by the body and promote strong bonds Sensors for health monitors (diabetes) and ion-based medical analysis systems Friction and wear reduction Improved solar cell efficiencies Coatings withstanding temperatures above 1000 ∘ C High-efficiency thermoelectric devices Self-monitoring coatings to warn of imminent failure

Amorphous YSZ/Au + DLC for Iubrication in humid air (DLC) and At high temperature (Au)

46

Hard YSZ nanocrystals for wear resistance

Adaptive transfer film (‘‘triboskin’’) on contact surfaces

Wear debris ‘‘Chameleon’’ coating with lubricant reservoirs

Gradient interface Substrate (b) (b) cross section of a coating at the friction Figure 61.33 Schematic of a conceptual design of the YSZ/Au/MoS2/DLC tribological chameleon contact. (Source: ref [184], courcoating with chameleon-like surface adaptive tesy of AIP.) behavior. (a) Composite layer structure and

61.7 Final Words

0.5

Friction coefficient

0.4 Lab air, 500 °C against Si3N4

0.3

Air 40% RH, 25 °C against 440 C

0.2

0.1 N2 (GA + GB ) + E A , a spontaneous reaction can be triggered. EA is the activation energy with respect to the initial state.

62.2 Physics of Wafer Bonding

that (GA + GB ) + EA is exceeded. A simple, but not always applicable, approach (see Section 62.1) would be an increase of the (bonding) temperature. However, another, more purposeful, approach foresees to generate a temporary state with increased Gibbs free energy (GA* + GB* ). Technically, this is done by enhancing the energy state of the surfaces and/or topmost layers, such that afterward the bonding process starts spontaneously, as the maximum of the activation barrier (and the final state) is lower than the temporary state with (GA* + GB* ). From the Arrhenius behavior E

k∝e

− k AT B

(62.11)

that describes the chemical reaction rate of a system, it is obvious that (for an intermediate state) processes with activation energies less than or around k B T still have a reasonable reaction rate, while this is definitely not the case for activation energies greater than 3 k B T. In Eq. (62.11) k is the reaction rate, and the product of the Boltzmann constant k B and the temperature T represents the thermal energy of the system. As a typical example for the intermediate enhancement of G, a two-component adhesive can be used where each component is applied to one surface. As long as the surfaces are in ambient air, no reaction starts. As soon as the two components come close (reasonable overlap of the electron wavefunctions), the chemicals intermix, and the reaction starts and lowers the Gibbs free energy considerably, until a stable state is reached. In the following, modifications are described, where either the activation energy EA is reduced or G is increased prior bonding in order to allow for reduction of G during the subsequent wafer bonding process (for details see Section 62.4):

• For SiO2 –SiO2 wafer bonding: H2 O deposition followed by the oxidation of Si. • For covalent Si–Si wafer bonding: sputtering results in surface damage (increased number of point defects) and amorphization. Subsequently these two surfaces are formed, followed by a crystallization reaction. • For Cu–Cu: using initially small grains that increase in size during the bonding process. • For all examples: lower surface roughness reduces the (interface) void surface area and hence decreases 𝛾dA (compared with surfaces with high roughness) when samples are bonded. If the roughness is low enough, even spontaneous bonding (as a result of low activation energy) of the wafers is possible. So far the time scale of the (intermediate) state (GA* + GB* ) was not further discussed; however, it is very important that the lifetime in ambient conditions and at room temperature (or slightly enhanced temperatures) is at least in the order of minutes and therefore sufficiently large to get the wafer surfaces bonded before the system relaxes from its intermediate state and thereby releases the required energy. Besides the increase of the Gibbs free energy, the additional condition of long lifetime for successful wafer bonding is fulfilled for all aforementioned examples. This implies that the reaction path diagram, as shown in Figure 62.4, has a time scale of greater than a minute and not as ordinarily a fraction of a second.

67

68

62 Wafer Bonding

62.3 Wafer Bonding Characterization Techniques

This section aims to introduce the main characterization and preparation techniques used in the experimental section. As some of these techniques are well established in surface science and thus well known, this section presents only selected characterization techniques in detail. 62.3.1 Characterization Techniques

The characterization techniques used to study wafer bonding processes can be split into two categories: (1) Surface characterization of individual incoming substrates (prior bonding). (2) Bonded interface characterization (after bonding). The techniques used for surface characterization are atomic force microscopy (AFM), spectroscopic ellipsometry (SE), white light interferometry (WLI), scanning electron microscopy (SEM), transmission electron microscopy (TEM), energydispersive X-ray spectroscopy (EDXS), Auger electron spectroscopy (AES), X-ray photoelectron spectroscopy (XPS), time-of-flight (ToF) secondary ion mass spectrometry (SIMS), and electron backscatter diffraction (EBSD). Bonded interfaces are characterized using c-mode scanning acoustic microscopy (C-SAM), SEM, TEM, EDXS, AES, and SIMS. (A general description of surface sensitive analytical methods can be found in Volume 1 of this series of books.) The bonding energy for direct bonding was quantified using the Maszara test, as well as qualitatively verified using a simple blade test in order to inspect substrate separation, while the electrical performance was determined based on the I–V characteristics. Table 62.2 presents an overview of the characterization techniques used in this chapter with respect to the properties investigated. Table 62.2 Overview of the characterization methods used in this chapter. Characterization type

Property

Method

Surface characterization Dimensions Elemental composition Crystallography Interface characterization Elemental composition Quality Bonding energy (direct bonding) Bond strength (metal bonding)

Roughness SE, WLI, SEM, TEM SE, EDXS, AES, XPS, SIMS EBSD Dimensions EDXS, AES, SIMS C-SAM, SEM, TEM Maszara test Substrates separation test (qualitative) I–V characteristics

AFM

Electrical conductivity

SEM, TEM

62.3 Wafer Bonding Characterization Techniques

Within this section the characterization techniques specific to wafer bonding processes such as EBSD, C-SAM, and the Maszara method will be described briefly. 62.3.1.1 C-Mode Scanning Acoustic Microscopy (C-SAM)

C-SAM is a scanning method employing acoustic waves in mega hertz range used for the detection of unbonded areas, often referred to as bonding voids. Therefore an acoustic beam with a frequency of tens/hundreds of mega hertz is generated by a piezoelectric transducer and focused perpendicular to the bonded interface of the investigated sample [31] using an acoustic lens and water as coupling media. As the transducer is working in dual mode (emission/detection), the waves reflected by the different surfaces/interfaces from the sample are received by the transducer. Such a diagram, which is shown in Figure 62.5, is referred to as A-scan. Further, by laterally scanning the wafer, the signal reflected after a constant ToF – related to the interface depth – is processed in order to obtain a contrast map, also referred to as C-scan. While the amplitude of the reflected wave provides information about the acoustic impedance difference Ar = Ai

Z2 − Z1 , Z2 + Z1

(62.12)

the time scale t refers to the depth z of the inhomogeneity: (62.13)

z = v1 t

Reflected amplitude Ar

Ai H2O Trapped air

Ar2

Ar2

Si

Ar1

Si

Ar1

Ar3

(b)

Ar3

(a)

Time of flight

Chuck

Figure 62.5 In (a) the schematic crosssectional view of an acoustic wave propagating through a bonded wafer pair is shown. In this case, the unbonded area (shown as trapped air) produces a strong reflection of the incident acoustic beam, which is shown in the A-scan (b). By scanning the bonded wafer pair in x- and y-directions and recording the waves reflected at a certain depth

(c) in the sample corresponding to the bonded interface (focus depth), a C-scan image is generated, corresponding to the defects chart of the bonded wafer pair. Picture (c) shows a C-scan measurement of a bonded wafer pair with a particle-induced bonding defect (unbonded area) close to the edge, in 6 o’clock position.

69

70

62 Wafer Bonding

here Ar and Ai are the amplitudes of the reflected and incident acoustic waves, respectively. Z1 and Z 2 are the acoustic impedances of the two materials at the interface. The acoustic impedance is defined as Z = 𝜌v,

(62.14)

where 𝜌 and v are the materials density and the corresponding speed of sound, respectively. Due to the very large acoustic impedance mismatch of a gas/solid or a gas/liquid interface, almost 100% is reflected. Thus, even ∼10 nm thin, laterally expanded (greater than the lateral resolution) air voids can be clearly detected. Practically the detection limit ddet for unbonded areas (bonding voids) is about one third of the theoretically predicted resolution d: d (62.15) 3 For Si samples, voids with 10–15 μm diameter can be detected using current stateof-the-art acoustic microscope equipment. ddet ≈

62.3.1.2 Maszara Method

Several destructive testing methods are available for the evaluation of the bond strength. They are all based on measuring the force needed to separate the bonded wafers followed by the calculation of the energy that is needed to produce the bonded interface failure. The most commonly used method is the crack opening method. Often this method is also called “Maszara method” [3], “razor blade method,” or “double cantilever beam method.” The main benefits of using this method are the straightforward procedure without a need for specific sample preparation as well as the very basic experimental setup required to analyze a bonded substrates pair: an infrared (IR) inspection stage consisting of a lamp and an IR-sensitive camera (used also as sample holder), a blade with known thickness and a ruler (see Figure 62.6). The Maszara method is derived from the theory on crack propagation in solids and is based on the minimization of total energy of the system (bonded wafers with inserted blade). Initially the total energy is enhanced by opening a crack, which

Figure 62.6 IR image showing a typical result of the blade test. The crack length is measured with a ruler at the middle of the razor blade, adjusted perpendicular to the edge of the razor blade.

62.3 Wafer Bonding Characterization Techniques

increases the elastic energy of the system. This crack formation on the other hand increases the surface energy of the system and therefore leads to a larger total energy. When an equilibrium between these two driving forces is reached, the crack stops and the surface energy can be calculated, using the measured length of the crack and geometric dimensions, by using the following expression [3, 32]: 𝛾=

3E1 t13 E2 t23 h2 1 (𝛾1 + 𝛾2 ) = 2 16(E1 t13 + E2 t23 )L4

(62.16)

Here 𝛾 is the wafer interface energy, E1 and E2 are the Young moduli, and L is the measured crack length. The values of t 1 , t 2 , and h describe the thicknesses of the wafers and the blade, as can be seen from Figure 62.7. The accuracy of this method is dominated by the measurement error of the crack length, since the surface energy is ∝ L−4 . For a given measurement geometry (wafer and blade thicknesses), the error gets larger with increasing bonding energy, since the crack length decreases and the measurement error gets more significant. In cases of high bonding energy values measured with thin blades (which lead to short crack lengths), the relative error can reach values up to 60%. 62.3.1.3 Electron Backscatter Diffraction (EBSD)

With EBSD the absolute crystalline orientation and phases of single- or polycrystalline samples can be identified by using an SEM system. Therefore the stage is tilted to usually 70∘ , and the EBSD detector is inserted close to the specimen surface at an angle of about 90∘ to the primary electron beam [33]. Typically, the lateral resolution is in the range of 20–100 nm, and the information depth is about 5–50 nm. Shadowing effects have to be reduced, and so high surface roughness of the specimen has to be avoided. Further, contamination and defects have to be avoided, too. In compliance with these requirements, the scattering of the electrons with an energy close to the energy of the primary electrons produces a diffraction pattern, consisting of the so-called Kikuchi lines or bands at the phosphor screen of the EBSD detector. The phosphor screen converts the backscattered electrons into light, which is recorded by a charge-coupled device (CCD) camera. In Figure 62.8 the Kikuchi lines are shown schematically. In order to investigate one point in Figure 62.9a, first the corresponding Kikuchi pattern observed on the phosphor screen is visualized with the CCD camera, and then Kikuchi bands are detected and classified using a dedicated EBSD software (Figure 62.9b–d). Finally, the absolute crystal orientation (three Euler angles for L t1

E1

h1

t2

E2

h2

Figure 62.7 Schematic drawing that illustrates the geometry of the blade test. When the razor blade is inserted, a crack opens, whose length is measured for determining

h = h1 + h2

the bonding energy. The crack length L has to be measured starting from the edge of the razor blade, as shown in the figure.

71

72

62 Wafer Bonding

SEM column Electron beam

EBSD detector

ND

TD

Specimen ~70° RD Figure 62.8 Schematic view of the Kikuchi lines, projected onto the phosphor screen.

(a)

10 μm

(b) –11–1

–10–1

–12–1

–130 –21–1

–120

–20–1 –110 –30–1 –210 –310

–121

–100 –211 –111

(c)

(d)

–301

Figure 62.9 Sequence of classifying a single Kikuchi pattern: (a) SEM image recorded with forward scattered detector. (b) Kikuchi pattern observed on phosphor screen. (c) Band detection and (d) its Kikuchi line classification evaluated by the software.

62.3 Wafer Bonding Characterization Techniques

every grain) and phase information can be determined automatically after applying the Hough transformation and comparing it with the simulation. The physical mechanism for the electron interaction with matter to obtain Kikuchi lines is currently described by two contradicting theories: “diffuse scattering followed by diffraction” and “channeling in and channeling out.” However, the exact physical model is still unclear and under discussion [34]. By scanning an area, the determination of grains with one specific phase and orientation is possible. Different orientations (phases) are separated by grain (phase) boundaries, which mainly cannot be indexed due to two or more different orientations (phases). Scanning areas, which are much greater than the average grain size, deliver a multitude of results and statistical crystallographic data, such as phase mapping, orientation mapping, grain size distribution, orientation contribution, grain boundary misorientation angle distribution [35, 36], and intragranular misorientation for strain analysis. 62.3.2 Sample Preparation Techniques

This section is reviewing the different preparation techniques and sequences used for AES, cross-sectional scanning electron microscopy (X-SEM), cross-sectional transmission electron microscopy (X-TEM), and EBSD measurements (see also [37, 38]). Generally it has to be distinguished between preparation performed for surface analysis prior bonding or for bonded interface characterization. The analysis of bonded specimens first requires (except for Gatan method) that one of the bonded substrates is completely removed or at least thinned back to less than 5 μm to ensure that the subsequent Ar sputtering or focused ion beam (FIB) milling can reach the bonded interface. The removal of one substrate was performed by grinding using grinding pads with SiC particles and polishing with diamond particles. Such a pretreatment is shown in Figure 62.10 for bonded Cu–Cu wafers. This back-grinded

Si (bottom wafer)

Bonding interface with surrounding Cu layers

Si (top wafer) 2 mm

Figure 62.10 Cu–Cu bonded specimen after back-grinding and polishing for subsequent interface characterization.

73

74

62 Wafer Bonding

specimen, with simultaneously accessible thin Si (top wafer), the two intermediate Cu layers and the second Si (bottom wafer), can be prepared with a tiny finite wedge with relatively small efforts. This preparation is used for the further preparation steps of bonded specimens. For both types of characterization, the surface and the interface, the preparation and measurement is almost the same. 62.3.2.1 AES and SIMS Specimen Preparation

Depth profiling in an AES or SIMS of bonded layers should start in the thin top substrate (e.g. Si layer) and end in the opposite substrate to gain elemental information of the entire layer structure including the bonded interface (see Figure 62.10). Depth profile spectroscopy is a destructive method of alternating spectroscopic analysis and sputtering to gain depth-resolved elemental information. For sputtering most frequently electron impact ionization guns with a continuous argon gas inlet are used. Therefore gaseous Ar is first ionized and then accelerated on the specimen, where the specimen material is removed (sputtered) with typical rates in the range of nanometer per minute. Hence, the analysis typically starts at a position of the thinned RP , and thus Eq. (64.4) may yield values in excess of 100%.2) ⃗ 1 and M ⃗ 2 continuously rotate Since in the applied field the magnetization vectors M toward each other, sometimes this is also referred to as “scissoring motion” – the magnetic field dependence of the magnetoresistance exhibits a typical bell-shaped curve. The microscopic origin of the GMR is the spin-dependent scattering of the charge carriers, when they move from one ferromagnetic layer to the neighboring one. As mentioned above the nonmagnetic interlayer has to fulfill one important condition, namely, negligible spin scattering for the traversing electrons. This picture can be applied for the electrical current flowing perpendicular and also parallel to the layers. The original data from the Fert group on Fe/Cr samples is displayed in Figure 64.10 [16]. These single-crystalline samples have been grown by molecular beam epitaxy. 2) The normalization to RP is just a convention, which established itself in the field.

203

204

64 Spintronics: Surface and Interface Aspects

R(H)/R(H = 0) 1.0 (Fe 30 Å/Cr 18 Å)30

(Fe 30 Å/Cr 12 Å)35 (Fe 30 Å/Cr 9 Å)60 0.5 –40

–20

0

20

40

Magnetic field H (kOe) Figure 64.10 Giant magnetoresistance in Fe/Cr superlattices for different thicknesses of the Cr interlayer ranging from 9 to 18 Å. (Source: After Ref. [16].)

As discussed above the resistance exhibits a bell-shaped characteristics as a function of the external magnetic field. The magnetic ground state of these superlattices corresponds to antiparallel coupling; therefore the maximum resistance is found for zero external field. It continuously drops with the strength of the external field due to the above scissoring motion of the layer magnetizations. Above the saturation field HS , all layer magnetizations are oriented parallel to each other and to the external field, and the magnetoresistance assumes a constant value. The highest GMR value is obtained for the thinnest Cr interlayer. This situation corresponds to the first antiferromagnetic maximum of the IEC, which also needs the highest magnetic field of HS = 2 T to obtain magnetic saturation. With increasing interlayer thickness the antiferromagnetic coupling becomes weaker, and therefore the saturation fields are significantly reduced. An interlayer thickness of 18 Å brings the system close to the crossover into the ferromagnetic coupling state. In this region, also 90∘ coupling contributions may occur, which further reduce the magnetoresistance. In Figure 64.11, we show an example for the GMR in the Co/Cu system, which has probably the highest technological relevance. Here, the GMR is plotted as a function of the Cu interlayer thickness tCu and therefore directly displays the oscillatory variation of the GMR with the strength of the IEC. As a marked difference to the Fe/Cr case discussed above, these Co/Cu multilayers have been grown by magnetron sputtering, a technique that is extensively used for large-area deposition of thin films in hard disk industry [6]. The resulting multilayers are polycrystalline with a strong {111} texture [66]. We clearly see the reduction of the GMR magnitude from the first to the second and to the third antiferromagnetic maximum of the IEC. The data set also compares the magnetoresistance measured at cryogenic and room temperature. In the first antiferromagnetic coupling maximum, the GMR at 4 K reaches values around 110%, which are reduced to about 60% at room temperature. A similar temperature dependence is observed for the second and third coupling maxima. The reduction of GMR at room temperature is a consequence of the increase of the overall resistivity with temperature and thermally activated spin wave excitations.

64.4 Giant Magnetoresistance (GMR)

1. afm max. T = 4.2 K RT

GMR (%)

100

2. afm max. 50

3. afm max.

0 0

1

2 tCu (nm)

3

Figure 64.11 GMR in sputtered Co/Cu multilayers as function of the Cu interlayer thickness tCu . The GMR maxima are related to the first, second, and third antiferromagnetic

4

coupling maximum of the IEC, respectively. The maximum GMR is achieved at cryogenic temperatures (T = 4.2 K) and is a factor of 2 larger than at room temperature (T = 300 K).

The highest GMR values at room temperature for the Co/Cu system have been reported as ΔR∕R ≈ 80% [67]. The combination of Co and Cu exhibits the highest magnetoresistance values at moderate magnetic fields and therefore became a reference systems for technological applications of GMR. This raises the question, what makes this material combination so special. A part of the explanation is found in the matching of specific electronic bands and wave functions at the interface between Co and Cu. In order to see this, let us inspect the system in more detail. The first thing to realize is that Co films grown on a Cu template by MBE take an epitaxial relationship and start to grow in the face-centered cubic (fcc) phase, rather than the bulk hexagonal phase [68, 69]. The in-plane lattice mismatch between the fcc phase of Co and the Cu lattice is only 1.8%, resulting in a small vertical compression of the films. The difference in the electron count for Co and Cu is only two electrons. We therefore expect the electronic band structure of fcc-Co to show some similarities to the one of Cu. An inspection of the non-relativistic band structures, for example, along the (001) direction, reveals this similarity (Figure 64.12a). The majority spin bands of fcc-Co (red solid lines) are almost completely occupied

205

206

64 Spintronics: Surface and Interface Aspects 2

2 Cu(001)

Energy (eV)

1

Cu

1

0

0

–1

–1

–2

–2

–3

–3

–4

–4

Co majority

–5 (a)

Co(001)

0 0.2 0.4 0.6 0.8 1 1 0.8 0.6 0.4 0.2 0 k-value

k-value

Co minority

–5 (b)

Figure 64.12 (a) Electronic bands in Cu and face-centered cubic Co along the (100) direction neglecting the influence of spin–orbit coupling. The Co majority spin states (↑) are represented by solid lines, whereas the

broken lines mark the Co minority spin states (↓). (b) Fermi surfaces for Cu, Co majority spins, and Co minority spins. Note the similarities of the Cu and Co majority spin surfaces.

and show similar shapes and band symmetries as the respective copper bands. The minority spin bands of Co (green broken lines) are shifted above the Fermi level due to the exchange interaction (exchange splitting). Since the electrical transport in the diffusive limit is carried by all electrons at the Fermi level and not those in a particular crystalline direction, Figure 64.12b shows a comparison of the respective Fermi surfaces. Here, the similarity between the Fermi surface of Cu and the majority spin Fermi surface of fcc-Co is even more striking, whereas there is a large difference to the minority spin Fermi surface. This means that minority spin electrons from Co will be more strongly scattered when entering the Cu layer than the majority spin electrons. Therefore, the spin-dependent scattering upon both entering and leaving the Cu layer plays a role in determining the magnitude of the GMR. The above argument holds also for polycrystalline film stacks grown by magnetron sputtering. By using appropriate seed layers, one can usually achieve a highly textured morphology of the films. For example, Cu as a seed layer on a polycrystalline or amorphous template (e.g. native SiO2 on a Si wafer) takes a {111} textured orientation and therefore imprints this texture also in the Cu/Cu multilayer [67]. The Fermi surface argument sketched above can be also transferred to the textured multilayers, although the situation at the interface is more complex. In general, due to the texture and the grain boundaries, the interfacial roughness may be higher than in a multilayer grown by molecular beam epitaxy. In addition, the higher kinetic energy of the atoms during sputtering growth may increase the tendency to intermixing at the interfaces. As a consequence, small superparamagnetic Co clusters may form in the interfacial region influencing the temperature dependence and/or high magnetic field characteristics of the GMR signal [70]. Moreover, with increasing interlayer thickness and number of repetitions in the multilayer, there is an accumulation of interfacial roughness, leading to a gradient of the interfacial quality from

64.4 Giant Magnetoresistance (GMR)

the bottom to the top of the multilayer. This roughness can “smear out” the coupling conditions for the IEC, for example, a mixture of antiparallel and 90∘ couplings. This is probably one of the reasons for the weak third antiferromagnetic maximum seen in Figure 64.11. Finally, even an ideally sharp interface has a finite “magnetic width.” The reason for this is a magnetic proximity effect: the exchange interaction polarizes electrons in copper across the Co/Cu interface, inducing a magnetic moment on the Cu side of the interface [71]. This has some effect on the spin-dependent scattering at the interface, which we have neglected in our simplified picture. Let us look into the microscopic mechanisms of the spin transport in some more detail. For this we choose the CIP geometry, i.e. the electrical current flows parallel to the film planes. We have to keep in mind that the transport in GMR is diffusive. This means that although the charge carriers at large have a defined drift direction parallel to the layers, the trajectory of an individual electron has momentum components perpendicular to the layer plane. Therefore, even in CIP geometry the electrons pass through several neighboring layers, i.e. probing their magnetic state during their propagation. On a microscopic level we can distinguish two main processes contributing to the spin-dependent resistivity (Figure 64.13, left). The first one is related to the spindependent scattering of the electrons while passing through the interfaces. In the parallel orientation of the magnetic layers, the spin-up carriers undergo very little scattering at the Co/Cu interfaces, because of the similarity of the Cu and Co spinup Fermi surfaces. Quantitatively, the strength of the scattering is determined by spin scattering asymmetry 𝛼 introduced above. By contrast, the spin-down carriers undergo comparably stronger scattering because of the difference between Cu and the minority spin Co Fermi surface. The overall resistivity RP is therefore mainly determined by the spin-up channel and is thus low. In the antiferromagnetic coupling arrangement of the layers, the spin-up electrons starting from one Co layer Spin-dependent scattering

Spin-dependent reflectivity Scattering event

M

M

RP

M

M

RAP

Figure 64.13 Microscopic mechanisms determining the spin transport in a layered metallic system for a current flowing parallel to the layers. (Left) Spin-dependent

M

M

RP

M

M

RAP

scattering at the interfaces. (Right) Spindependent reflectivity. (Bottom) Qualitative contributions to the spin-dependent resistivities in the Mott two-current model.

207

208

64 Spintronics: Surface and Interface Aspects

become spin-down electrons in the neighboring Co layer and vice versa. Therefore, the scattering for both spin channels is comparable, resulting in a higher resistivity RAP . This interface-dominated mechanism is the main microscopic origin for the above observation of RP < RAP . The second microscopic mechanism entering the spin transport is the spindependent reflectivity introduced by the spin-dependent potential step at the interfaces in combination with spin-dependent scattering in the bulk of the layers. In the parallel configuration (P), the potential step is low, and the spin-up electrons can move through the entire layer stack where they basically undergo spin scattering in the bulk of the ferromagnetic layers. The spin-down electrons, however, experience a higher potential step. This means that once they have entered the Cu interlayer, there will be a higher probability for reflection at the interfaces and a preference of the spin-down electrons to propagate along the Cu interlayer. There they encounter little spin scattering, resulting in a low resistivity contribution in the two-current model. In the AP configuration only one side of the Cu interlayer has a higher reflectivity depending on the spin direction. The situation for spin-up and spin-down carriers becomes comparable and is dominated by the scattering in the bulk of the layers. It results in comparable contributions to the spin-up and spindown resistivities. Also in this case we find RP < RAP . The quantitative contribution of spin-dependent reflectivity to the overall magnetoresistance depends strongly on the structural and chemical quality of the interface. The issues discussed above are important for any kind of GMR structure and appear in multilayers as well as trilayer structures. 64.4.3 GMR in Trilayer Structures

The multilayer geometry was important for the discovery of the GMR and the understanding of the spin transport processes in metallic systems. Technologically more relevant, however, are structures comprising a magnetic trilayer as the main building block. We therefore introduce a prototypical device structure in spin-dependent transport, the spin valve [72], in some detail. By choosing the layer materials and thicknesses and tailoring the interfaces between them, a huge parameter space for tuning both magnetic and spin transport properties becomes available. A spin valve and its various derivatives can therefore be designed to the specific needs of the application, for instance, a magnetic sensor or a memory cell. In addition, spin valves and MTJs share a very similar geometry, and many experiences and developments made on spin valves can be directly transferred to MTJs. The simplest version of a spin valve consists of two ferromagnetic layers, sandwiching the separating nonmagnetic interlayer. In order for the spin valve to properly work, the two magnetic layers in the structure in Figure 64.14 have to switch at sufficiently different coercive fields HC1 and HC2 . This simple trilayer is also often called a “pseudo spin valve.” The relation between the coercive fields and the GMR signal is depicted in (Figure 64.14a, left) for an idealized switching behavior. Neglecting any coupling effects between the two ferromagnetic layers (i.e. the residual interlayer

64.4 Giant Magnetoresistance (GMR)

M

M

ΔR

H

H

H

HC1 HC1 < HC2

Figure 64.14 Increasing structural and magnetic complexity of spintronic devices for the example of a spin valve with weak interlayer exchange coupling. (a) Magnetic hysteresis loops of the individual magnetic layers (red, blue), the resulting hysteresis loop of the pseudo spin valve (green) and the

H HC1

HC2

(a)

ΔR

HC1 < HC2 (b)

HEB > 0

HC2

HEB

corresponding magnetoresistance characteristics ΔR(H). (b) Same for an exchange-biased spin valve (for explanation, see text). The arrows indicate the relative orientation of the magnetization vectors in the ferromagnetic layers.

coupling must be small), the variation of the magnetization in the external magnetic field (hysteresis loop) can be obtained by simply superposing the hysteresis loops of the individual magnetic films. Schematically, this leads to the complex stepped loop shape – which can be measured in a simple magnetometer – indicating the various parallel and antiparallel configurations of the two layers. Starting from high magnetic fields where both layers are aligned parallel with the external field, upon field reversal first the layer with the lower coercive field (“soft layer”) switches, leading to an antiparallel configuration. Increasing the external field then also switches the layer with the higher coercive field (“hard layer”), resulting in a parallel configuration in the opposite field direction. The hysteresis loops and thus the magnetic switching are symmetric with respect to H = 0. Assuming a high (low) resistance state for the relative antiparallel (parallel) magnetic orientation of the layers, we end up with the symmetric magnetoresistance pattern (Figure 64.14a, right), which is an abstraction of the “butterfly” curves typically found in experiments. An example for this butterfly curve may be found in Figure 64.17. Note that although a pseudo spin valve already provides the principal functionality of measuring the magnitude of a magnetic field strength, the symmetric GMR characteristics make it difficult to determine the spatial orientation of the magnetic field. This is a shortcoming for many technological applications, for example, in magnetic angle or rotation sensors. With respect to applications in data storage and processing, the drawback is rather the low rigidity of the hard layer against external fields, which results in a low stability of GMR-based memory cells in a magnetic random access memory (MRAM) arrangement. In order to overcome these limitations, expanding on this simple trilayer, more complex structures have been developed by adding further functional layers. A very important example for engineering the magnetic properties is the mechanism of exchange biasing [73]: one of the ferromagnetic (FM) layers is exchange-coupled

209

210

64 Spintronics: Surface and Interface Aspects

to an antiferromagnet (AF) (Figure 64.14b). An antiferromagnet is characterized by a well-defined spin alignment axis along which neighboring atomic magnetic moments are pointing opposite to each other. As a consequence, the net magnetization vanishes. Below a critical ordering temperature (Néel temperature), the antiferromagnetic alignment between neighboring spins can only be broken by very strong magnetic fields in the range of several 10 T. Being coupled to an antiferromagnet changes the magnetization curves of a ferromagnetic film markedly. First, the coercive field of the ferromagnet is increased, i.e. it becomes magnetically harder. Second, the hysteresis loop may no longer be centered around H = 0, but appears shifted along the field axis. This latter phenomenon is called exchange bias [74]. The microscopic origin of the exchange bias is the direct exchange coupling of the spins across the AF/FM interface. The structural, morphologic, and chemical properties of the interface therefore have a pronounced impact on the strength and characteristics of the exchange bias [75] (see also Section 64.4.4.1). The exchange bias stabilizes the magnetization direction of an adjacent ferromagnet against an external field in a particular direction in space and enables a control of the working point in devices, such as magnetic sensors, read heads, and memory cells and logical circuits. This is clearly seen by the change in the magnetic hysteresis loop of the second (hard) layer in contact with the antiferromagnet (Figure 64.14b, left): the loop becomes much wider and undergoes an unidirectional shift along the field axis. The shifted hysteresis loop is centered around a magnetic field value H = HEB , marking the strength of the exchange bias. As a consequence, the resistance characteristics are no longer symmetric with respect to the field-free case and have a wider regime for the antiparallel alignment (Figure 64.14b, right). In an exchange-biased spin valve, the hard layer acts as a real magnetic reference in space, which is needed for angle-sensitive sensors (for an example, see also Section 64.7). The direction of the exchange bias can be “imprinted” into the system in two ways: (i) the entire film system is grown in the presence of a magnetic field, or (ii) after the growth of the layers, the system is heated above the Néel temperature of the antiferromagnet and cooled down in a magnetic field that aligns the hard layer. When cooling through the Néel temperature, the interfacial exchange coupling between antiferromagnet and hard layer imprints the spin alignment into the antiferromagnet and thus establishes the exchange bias direction [73]. An instructive example for the magnetic and transport design of a realistic spin valve with exchange biasing is given in Figure 64.15. The GMR trilayer consists of Ni80 Fe20 /Cu/Ni80 Fe20 , and the antiferromagnet is the intermetallic compound FeMn. The resistance curves along the film plane (sheet resistance) as a function of the magnetic field exhibit the characteristic spin-valve signature with a sharp onset at zero field, a broad plateau of higher resistance indicative of the antiparallel alignment, and an extended hysteresis at higher fields corresponding to the reversal of the hard layer (see also Figure 64.14b). Note that the direction of the exchange bias depends on the imprinted alignment of the antiferromagnet. There is a clear maximum in the GMR signal around 2 nm Cu thickness, which is close to the second antiferromagnetic coupling maximum in Ni80 Fe20 /Cu multilayers [77]. Toward higher copper thicknesses the GMR signal falls off, which can be related to

Sheet resistance (Ω)

64.4 Giant Magnetoresistance (GMR)

1.3 nm

2.5 nm

9.5

18.5

16.0

18.0

15.5 –40 0 Magnetic field (kA/m)

4.7 nm

9.0

–40

0

–40

0

8

ΔR/Rsat (%)

6

4 2 0

1 2 5 Copper layer thickness (nm)

Figure 64.15 Magnetoresistance in Ni80 Fe20 /Cu/Ni80 Fe20 /Fe50 Mn50 spin valves. Data show the variation of the magnetoresistance curves and the MR ratio on the Cu

10

interlayer thickness. (Source: After Ref. [72]). The dashed line gives a fit to the data for the Cu thicknesses for which full antiparallel alignment is obtained [76].

two mechanisms: (i) a decrease of the sheet resistance R◽ of the layer stack, i.e. a shunting effect by the thicker Cu layer, and (ii) with the Cu layer thickness also the diffuse scattering probability in the Cu layer increases, i.e. the scattering probability of spin-polarized electrons with the second ferromagnetic layer and thus the GMR is reduced. The origin of the GMR decay at Cu thicknesses tCu < 2 nm is different. It is related to the residual ferromagnetic coupling between the two Ni80 Fe20 , which prohibits a full antiparallel arrangement of the two layer magnetizations needed for the full GMR signal. At very low thicknesses tCu ≃ 1 nm where one observes the first afm coupling maximum in the Ni80 Fe20 /Cu multilayers, the GMR disappears completely. This may be a consequence of strong ferromagnetic coupling conveyed through roughness or pinholes. Exchange-biased spin valves are nowadays the most prominent building blocks for technological applications in GMR. 64.4.4 The Role of the Interface

The GMR has been shown to be an interface-dominated effect; more precisely, it arises by the spin scattering of the electrons at the interface between the

211

212

64 Spintronics: Surface and Interface Aspects

nonmagnetic and ferromagnetic layers [78]. For the parallel configuration, the meaning of spin-down (↓) and spin-up states (↑) is the same in both layers; therefore also the scattering is the same. In the antiparallel configuration, however, (↓)-electrons leaving the first ferromagnetic layer become (↑)-electrons, when they enter the second ferromagnetic layer. Likewise (↑)-electrons have to find empty (↓)-states in the second layer to be able to pass through the interface. Because the density of available (↓)-states is small, the resistivity increases. This explains why the resistivity in the antiparallel state is often higher than in the parallel state. For specific material combinations showing, for example, opposite spin scattering at the two interfaces, i.e. 𝛼1 > 1 and 𝛼2 < 1, the resistivity in the antiparallel state may be lower, which leads to an inverse GMR [79]. From this brief review of the GMR phenomenon, it becomes clear that the interface and its chemical, structural, electronic, and magnetic properties will play a crucial role. In fact, it turns out that the magnetotransport is determined by a complex interplay of all these properties. For the sake of clarity, we will focus mainly on the Co/Cu layered system, which is the most prominent GMR system, also with respect to technological applications. In the following we will discuss some selected aspects that highlight the role of the interface and the complex interplay of magnetic properties and spin transport. 64.4.4.1 Exchange Biasing

The functionality of spin-valve structures depends strongly on the exchange bias phenomenon. Exchange bias occurs due to the coupling of the magnetoresistive layer stack to an antiferromagnet (AF). As we have already discussed above, exchange bias has two consequences (Figure 64.14b): (i) the coercivity HC of the ferromagnetic film (FM) in direct contact increases, and (ii) the magnetization curve exhibits a unidirectional shift HEB along the magnetic field axis. The latter behavior is also the reason why exchange bias is sometimes referred to as unidirectional magnetic anisotropy. The microscopic mechanism of exchange bias involves the exchange coupling of spins at and across the FM/AF interface. Although this coupling may be straightforward to understand on the level of a single spin or single magnetic moment, the situation becomes quite complex at extended interfaces. This is due to the fact that the local and global AF spin arrangement depends on the details of the structure, morphology, and crystalline orientation of the interface. As an example, let us look into the simple case of a layered antiferromagnet. In a layered AF there exists a crystallographic direction along which atomic planes with opposite spin orientation are stacked (Figure 64.16a). The individual atomic planes are magnetically uncompensated. Such a situation is found, for example, in Cr along the (100) direction. A flat (100) surface will therefore exhibit a defined magnetic state and orientation of the magnetic moments, whereas a surface with monoatomic steps will expose terraces with opposite spin orientation. This magnetic topology has been visualized in spin-polarized STM [80, 81]. Intuitively one might expect that the AF spins at a sharp interface to orient parallel to the FM. However, the exchange coupling depends on the details of the electronic structure of the materials on either side of

64.4 Giant Magnetoresistance (GMR)

Layered AF

(a)

Spin-flip coupling

(b)

Spin-flop coupling

(c)

Magnetic frustration

(d)

Figure 64.16 Exchange coupling across an FM/AF interface. (a) Layered AF with antiparallel coupling. (b) Compensated AF with spin-flip coupling. (c) Compensated AF with spin-flop coupling. (d) Magnetic frustration at rough interfaces.

the interface. In the Fe/Cr system, the exchange coupling between Fe and Cr favors an antiparallel orientation of the Fe and Cr spins. Element-selective investigations in the Fe(100)/Cr(100) system have indeed shown the magnetic moments in the first atomic layer of Cr to couple antiferromagnetically to the underlying Fe substrate [82]. The adjacent Cr layers take their antiferromagnetic stacking, leading to the structure depicted in Figure 64.16a. A different situation occurs in the case of compensated AF planes. In the simplest case, the magnetic moments in the AF have a long-range spin alignment axis, along which neighboring spins couple antiparallel (Figure 64.16b). Across the AF/FM interface, the AF spin alignment axis can align either parallel or orthogonal to the magnetization in the ferromagnet. The first case is known as spin-flip coupling; the second one is termed spin-flop coupling (Figure 64.16c). The nature of this coupling depends on details of the atomic configuration and bonding situation at the interface. For the same material system, it may vary with the crystalline orientation of the interface. This has been demonstrated, for example, in the oxidic system Fe3 O4 /NiO. The AF NiO exhibits a spin-flip coupling to the ferrimagnet Fe3 O4 along the (110) and (111) directions, whereas the (100)-oriented interface revealed a spin-flop coupling [83, 84]. The microscopic mechanisms behind this behavior involve aspects beyond exchange coupling, for example, the influence of magnetoelastic interactions. In contrast to the idealized cases discussed above, a realistic interface introduces a number of complications, which can significantly impair the properties of the exchange bias. First, roughness at the interface that extends over several atomic planes in the AF may introduce magnetic frustration on different length scales at the interface (Figure 64.16d). Magnetic frustration means that the magnetic moments are partially quenched and locally deviate from the spin orientation (FM) and alignment axes (AF) in both materials. Second, interdiffusion between FM and AF will also alter the magnetic moments and coupling strengths across the interface. Third, due to the magnetic and electronic interactions across the interface, the interfacenear regions in both materials may also exhibit proximity effects, i.e. a modification of the magnetic properties of the materials close to the interface. Finally, one also has to take into account the presence of magnetic domains and domain walls in the AF. In order to capture these complications and arrive at quantitative description of the exchange bias, several models have been successfully developed and applied [85].

213

214

64 Spintronics: Surface and Interface Aspects

64.4.4.2 Interfacial Dusting

The magnetic materials employed in GMR structures have a wide range of magnetic and spin transport properties. For certain sensor applications a combination of a low coercivity, i.e. soft magnetic response, and high GMR is required. As we have seen above, soft magnetic properties can be easily obtained with nickel-iron alloys, such as Ni80 Fe20 or permalloy. Unfortunately, extensive studies on Ni80 Fe20 based GMR structures showed GMR values significantly smaller than those obtained from comparable Co/Cu-based systems [86]. Typically the GMR in the first antiferromagnetic coupling maximum in Ni80 Fe20 /Cu does not exceed ΔR∕R0 ≈ 20% at room temperature. The reason is the smaller spin scattering asymmetry at the Ni80 Fe20 /Cu interface. The task is now to increase the spin scattering asymmetry while keeping the soft magnetic properties of the permalloy. How can the advantages of Ni80 Fe20 and Co be combined? Alloying Co into the Ni80 Fe20 will change both the magnetic and the magnetotransport properties. Recalling that the major contribution to GMR comes from interfacial scattering, the solution is to place the Co as a monolayer thick coverage only at the interface between Ni80 Fe20 and Cu. The strong exchange coupling between Ni80 Fe20 and Co ensures that this artificial “bilayer” behaves magnetically as one unit. As a result, the magnetic properties (coercive field) are dominated by the Ni80 Fe20 layer, which is usually a few nm thick, whereas the spin scattering is dominated by the effective Co/Cu interface. This technique is sometimes called “interfacial dusting” and has proven a significant enhancement of the GMR as compared with pure Ni80 Fe20 /Cu interfaces [87]. It is a particularly nice example for the interfacial engineering in spintronic structures. An example for the effect of interfacial dusting is given in Figure 64.17. The GMR data in the panels (a) and (b) refer to a “clean” multilayer with the stacking [Ni80 Fe20 (15 Å)/Cu(20 Å)]14 without Co dusting. The thickness of the Cu spacer layer is chosen for the second maximum of the IEC. At room temperature the GMR signal is very small (a) and reaches sizable values of about ΔR∕R0 ≈ 10% only at cryogenic temperatures (b). The GMR curves in panel (b) show a particular property of the Ni80 Fe20 /Cu system, namely, a negligible magnetic hysteresis even at low temperature. Both the GMR and hysteretic behaviors change significantly with the addition of small amounts of Co. In the samples with additional Co, the multilayer structure is set to [Ni80 Fe20 (10 Å)/Co(ti )/Cu(19 Å)/Co(ti )/Ni80 Fe20 (10 Å)/Co(ti )]19 to maintain the conditions for the second antiferromagnetic IEC maximum. With increasing Co thickness ti at the interfaces one notes first an increase in the coercivity (panel (c)). This coercivity increase becomes even more pronounced at cryogenic temperature (panel (d)). Between ti = 2.6 Åand ti = 3.5 Å, the GMR increases dramatically by almost an order of magnitude, with a factor of 2 between the room temperature and cryogenic data (panels (e) and (f )). A further increase of the Co interfacial layer to ti = 4.4 Å, however, leads only to small changes (panels (g) and (h)) indicating that the effect has already reached saturation. This example shows the power of interfacial magnetic and magnetotransport engineering in spintronics. The Co interfacial layer must be just thick enough to establish most of the spin-dependent scattering properties of Co at the relevant interfaces.

64.4 Giant Magnetoresistance (GMR)

295 K

0 10

4.2 K No Co

5 (a)

(c)

2.6 Å

0 20 (e)

3.5 Å

No Co

(b)

(d)

2.6 Å

(f)

3.5 Å

0 20 10

5

ΔR/R (%)

10

0 40

15

30

10

20

5

10

0 20 (g)

4.4 Å

(h)

4.4 Å

0 40

15

30

10

20

5

10

0

–400

0

400

–400

0

400

0

Field (Oe) Figure 64.17 Effect of cobalt “dusting” on the GMR signal in a sputtered permalloy–copper multilayer measured at room (T = 295 K) and cryogenic temperatures (T = 4.2 K). For explanation, see text. (Source: From Ref. [87].)

The experiments show that a layer of the order of ti = 3.5 Å, which can be interpreted as a full Co coverage of the interface, is sufficient to achieve this goal. 64.4.4.3 Role of Oxygen

A major challenge in spintronics technology is not only to obtain the highest possible magnetoresistance values for a given material system but also to obtain it in a reproducible manner. In order to achieve this goal, all growth parameters and conditions need to be carefully controlled. In particular when using magnetron sputtering techniques, which is the method of choice for industry-scale production, the control of the process gas and the background pressure of the residual gas may be quite important. In the early days of GMR technology, one observed significant fluctuations in the GMR signal of samples prepared under nominally identical conditions. It became soon clear that the residual gas in the sputtering system may play an important role in the growth of magnetic layers and spin valves, thereby affecting both the crystallinity and the interface structure and morphology. In order to discuss this in some detail, we turn again to the Co/Cu system. We have already discussed above that Co/Cu multilayers develop a {111}-oriented fiber texture. This growth orientation is actually quite interesting. From experiments with molecular beam epitaxy under UHV conditions, we know that Cu

215

216

64 Spintronics: Surface and Interface Aspects

grows quite well epitaxially on Co(111). However, the growth of Co on Cu(111) is much more complicated, because it proceeds via the formation of twinned islands, which cannot coalesce into a flat film [69]. As a consequence, the Co film grows in an island fashion. This situation can be changed by the use of surfactants, for example, the twin formation can be suppressed by using Pb or Sb as surfactant [88]. A magnetron sputtering system for Co/Cu multilayers uses a clean process gas, e.g. argon, at pressures around 10−6 mbar. Usually, these growth chambers are not designed for UHV conditions, but a base pressure of 10−9 mbar may be achieved after an extended period of operation. The observation in the experiment was that the quality of the Co/Cu multilayers, i.e. the magnitude of the GMR signal, varied with the background pressure of residual gas mostly containing water [89]. This finding suggested that a component of the residual gas may act as a surfactant. Several experiments with oxygen clearly showed that there is a partial pressure regime of oxygen of ∼ 4 × 10−9 mbar, which leads to the highest GMR values, in this case in Co/Cu-based spin valves [90]. The interpretation assumes oxygen to act as surfactant during the layer deposition, influencing both the crystallinity and the interface structure and morphology. The layer stacking in technologically relevant spin valves can be quite complex. Often the stack starts with a seed layer to obtain the desired crystallinity of the stack. Then there may be an exchange bias layer, which may be a natural antiferromagnet coupled to a synthetic antiferromagnet. The GMR section of the stack may also comprises in itself more than just a trilayer, considering the dusting described above. The entire film stack must then be protected from oxidation by an appropriate cap layer. Most or all of these layers will contribute to the overall resistance of the stack, and they may also participate in the spin-dependent scattering without contributing to the GMR. Therefore, one may develop the idea to confine the spin-dependent transport to the GMR layer and reduce spin scattering channels outside the GMR layer. This is achieved by oxygen, not only to improve the growth of the layers and therefore the reflectivity at the interfaces but also to oxidize selected interfaces in a well-defined manner. The formation of such “nanooxide layers” (NOL) enhances the specular scattering of charge carriers and spins, confines them to the GMR-active layer, and causes a significant increase of the magnetoresistance. An example of the influence of NOLs is given in Figure 64.18. The GMR-active layer stack comprised as basis a [CoFe(20 Å)Cu(20 Å)/CoFe(20 Å)] trilayer [91]. The NOLs on either side of the layer stack were generated by plasma oxidation. The exchange biasing was performed by an IrMn antiferromagnet grown on a Ni80 Fe20 seed layer. The GMR curves show the characteristics of an exchange-biased spin valve. Without the NOLs the maximum GMR reaches up to about ΔR∕R = 6%. With the NOLs the GMR reaches up to ΔR∕R = 13%. We also note some changes in the field dependence, which show that the NOLs also influence the exchange biasing and the coercivity of the hard and soft magnetic layers. The effect of the NOLs depends sensitively on the architecture of the layer stack and their placement within the stack [92].

64.4 Giant Magnetoresistance (GMR)

14

MR (%)

12 10

Ta CoFe Cu CoFe

With oxidation

8

NOL NOL

IrMn 6

NiFe Ta

4 2 0 –400 –300 –200 –100 0

100 200 300 400

H (Oe) Figure 64.18 Change of the giant magnetoresistance signal by oxidation at the interfaces increasing the specular scattering. The maximum GMR is increased by more than

a factor of 2 by insertion of the nanooxide layers. In addition, however, also the GMR characteristics and the switching fields are changed. (Source: After Ref. [91].)

64.4.4.4 Temperature-Induced Degradation

The quality of a surface or interface can often be optimized by a careful control of the parameters during growth of the thin film stack. In the case of magnetron sputtering, one of the technologically preferred deposition techniques, these parameters include not only growth rate and background pressure but often also the growth temperature. However, applications may require operation in a quite different temperature regime, which raises the question of thermal stability of the thin film stack and eventual degradation mechanisms impairing the interface quality particularly at elevated temperatures. As an example, let us take the Co/Cu GMR system. It is known from the bulk phase diagram that Co and Cu have a miscibility gap, i.e. they do not form an intermetallic phase. Co in a Cu matrix usually forms precipitates. On the other hand, it is well established that growth of Co thin films on Cu at elevated temperatures may result in Cu diffusion onto the surface. This process is driven by the surface free energy of Cu being lower than that of Co [93]. In a GMR multilayer we have a sequence of alternating Co/Cu and Cu/Co interfaces. How do they behave as a function of temperature? This is illustrated in Figure 64.19. The data in Figure 64.19a depict the magnetization and GMR characteristics for a Co/Cu GMR multilayer in the first antiferromagnetic coupling maximum, i.e. for ∼ 1 nm thickness of the Cu interlayers. The multilayer has been grown by magnetron sputtering, resulting in a polycrystalline growth with a {111} texture. The GMR signal reaches values of close to ΔR∕R ≈ 45% at moderate magnetic fields of 200 mT. The Sshaped magnetization curve is a result of the antiferromagnetic interlayer coupling. Both characteristics change markedly, once the system has been treated at elevated temperature (200 ∘ C). The GMR signal is reduced by about a factor of 2, and the

217

64 Spintronics: Surface and Interface Aspects 1. afm max. As deposited

40 30

GMR (%)

(R–Rs)/Rs (%)

50

200 °C

20 10 1.0

IKerr / Isat

218

0.5

As deposited

0.0 –0.5

0

100

Co/Cu 1. afm (tCu~1 nm)

Recrystallization

2. afm (tCu~2 nm)

0

100 200 300 400 500 600 700 T (°C)

(b)

200 °C

–1.0 –200 –100

(a)

50 45 40 35 30 25 20 15 10 5 0

200

Magnetic field (mT)

(c)

100 nm

Figure 64.19 Temperature stability of magnetron sputter-deposited Co/Cu GMR multilayers. (a) Magnetization and magnetoresistance characteristics for films in the first antiferromagnetic coupling maximum measured at room temperature in the “as-deposited” state and after treatment at elevated temperature (200 ∘ C). (b) Variation of the GMR with annealing temperature for films in the first (Cu thickness tCu ≈ 1 nm) and second

Co 20 nm Cu

10 nm

antiferromagnetic coupling maximum (tCu ≈ 2 nm). (c) Cross-sectional energy-filtered TEM micrographs of multilayer after the annealing procedure showing degradation of the layer periodicity and the formation of pinholes (arrow) through the Co layers. The chemical information in the two right-hand micrographs is encoded in a false color representation: Co (red) and Cu (green).

magnetization curves exhibit a clear hysteretic part with low coercivity. The latter finding suggests that part of the film has now adopted a ferromagnetic interlayer coupling and therefore can no longer contribute to the GMR. In Figure 64.19b we compare the GMR results for the first and second coupling maximum – the latter corresponding to a Cu interlayer thickness of about 2 nm. In agreement with the above discussion, the system with the thinner Cu interlayer exhibits a strong degradation of the GMR already at 200 ∘ C, whereas the thicker Cu interlayer is clearly more stable. In this system the GMR starts to degrade only beyond 400 ∘ C. In fact, below 400 ∘ C there is even an improvement of the GMR signal up to ΔR∕R ≈ 30%. Improvement is due to a recrystallization of the grains in the multilayer, resulting in an improved {111} texture [94]. Above 400 ∘ C there is another recrystallization step, which favors a {100} orientation. This is only part of the degradation mechanisms. As already discussed in the previous paragraph, the loss of GMR can be traced back to an increase of ferromagnetically coupled regions in the film. The microscopic degradation mechanisms are the loss of coherency in the multilayer stacking and the formation of local “pinholes” through the Co layer. This is illustrated by cross-sectional micrographs taken in energy-filtered transmission electron microscopy (TEM) of a sample after thermal treatment (Figure 64.19c).

64.5 Tunneling Magnetoresistance

The chemical information from the energy-filtered TEM images is translated into a color code (Co, red; Cu, green). In the bottom part of the image, we still see a confined region with the original layer stacking corresponding to 1 nm periodicity of the Cu interlayer. Outside this region the layered structure has changed by partly merging neighboring Co layers and therefore leaving behind also thicker Cu interlayers. The second TEM micrograph shows an example for a local pinhole, connecting two Cu interlayers, which is eventually the initial step in the formation of the thicker Cu layers. Already in the vicinity of this pinhole, the interlayer coupling will be strongly changed due to the higher effective Cu thickness. In summary, the loss of GMR after treating the sample at elevated temperatures is due to a combination of pinhole formation, recrystallization, and loss of the layer stacking sequence.

64.5 Tunneling Magnetoresistance

It is well known that there may be charge transport through a very thin insulating barrier due to quantum mechanical tunneling [10]. The tunneling conductivity occurs due to a finite overlap of the electronic wave functions of the metallic electrodes on either side of the insulating film. The magnitude of the conductivity is determined by the width d and the height Φ of the potential barrier in the insulator. The first spin-dependent tunneling effects were demonstrated by Jullière in 1975 [33]. In this case the tunneling junction was composed of an Fe and a Co film separated by a very thin oxidized germanium barrier. The experiments performed at cryogenic temperatures (4.2 K) gave first evidence for the existence of spin-polarized tunneling. Jullière found a 14% lower resistance for parallel orientation of the magnetization directions in the electrodes than for an antiparallel alignment. 64.5.1 Spin-Dependent Tunneling in the Jullière Picture

The work by Jullière also gave a first explanation of spin-dependent tunneling on the basis of the spin-split DOS in a ferromagnet. The idea is sketched in Figure 64.20. Assuming elastic tunneling only, the tunneling current density is generally determined by the number of occupied states in the left electrode NL , the number of unoccupied states at EF in the right electrode NR , and a tunneling matrix element. The current contributions have to be determined for both spin-up and spin-down channels independently, as long as we can neglect spin-flip processes. For illustration purposes let us assume that the majority spin DOS is almost completely occupied – this situation corresponds to the ferromagnets Co or Ni. For a parallel alignment of the magnetization directions in the two electrodes, we therefore find a strong spin-down tunneling current, which is due to the fact that we have many occupied NL↓ and empty spin-down NR↓ states available around the Fermi level. The tunneling probability and thus the tunneling current are proportional to NL↓ NR↓ . In the spin-up channel NL↑ and NR↑ are both small and thus result in a significantly

219

64 Spintronics: Surface and Interface Aspects

Antiparallel alignment

(a)

Barrier

Parallel alignment

Barrier

220

DOS

(b)

Figure 64.20 Principle of the tunneling magnetoresistance effect (TMR) in a simple density of states picture. Each ferromagnetic electrode is represented by a schematic density of states (including both s- and d-type

electrons), with the arrows denoting the majority (↑) and minority spin (↓) characters. Parallel alignment of the electrode magnetizations (a) vs. antiparallel alignment (b).

smaller spin-up current contribution. If we reverse the magnetization of the righthand electrode, the relative spin orientation is reversed as well, and the spin-up and spin-down DOS contributions in this electrode are interchanged. This means that the |↑⟩ tunneling current is now determined by ∝NL↑ NR↓ and the |↓⟩ current contribution is ∝NL↓ NR↑ . From this simple consideration we expect the parallel (antiparallel) alignment to lead to a lower (higher) resistivity, according to the expressions RP =

V V ∝ maj maj IP NL NR + NLmin NRmin

RAP =

V V ∝ maj maj min IAP NL NR + NLmin NR

(64.5) (64.6)

with the minority and majority spin densities NL,R in the left- and right-hand electrodes, the tunneling voltage V , and the tunneling currents IP and IAP . From these equations and the spin polarization values PL and PR of the electrodes, Jullière’s formula for the TMR is derived: 2PL PR R − RP ΔR = (64.7) = AP R RP 1 − PL PR With the typical band polarization of the 3d ferromagnets of the order of P ∼ 45–50%, the predictions for the maximum magnitude of the TMR are in the range of 70%. 64.5.2 MTJs with AlOx Barriers

Spin-polarized tunneling from ferromagnets into superconductors was extensively used to determine the magnetic moments [31, 95]. These experiments all involved cryogenic temperature and perfected the use of Al2 O3 tunneling barriers. Based on this knowledge first magnetic tunneling experiments at room temperature were

64.5 Tunneling Magnetoresistance

0 Co film

–0.25 –0.50 0.12

ΔR/R (%)

0.06

CoFe film

0 10.0 CoFe/Al2O3/Co junction

7.5 5.0 2.5 0 –600

–400

–200

0 H (Oe)

200

Figure 64.21 Tunneling magnetoresistance in a tunneling junction with an Al2 O3 barrier. The AMR signals from the individual Co (top) and CoFe films (center) exhibit the different magnetic switching of the two

400

600

ferromagnetic electrodes. The bottom panel shows the magnetoresistance signal from the magnetic tunneling junction. (Source: After Ref. [34].)

demonstrated in the mid-1990s. The example in Figure 64.21 reproduces the results from a Co/Al2 O3 /CoFe MTJ [34]. The TMR ΔR∕R reveals the characteristic “butterfly” curve as a function of the external magnetic field, which is a consequence of the different coercivities of the Co and CoFe electrodes. This magnetic situation is very similar to that of a GMR pseudo spin valve, in that the two electrodes need to have distinctly different magnetic switching fields to enable well-defined and addressable parallel and antiparallel configurations of the magnetization vectors. The magnetic tunneling resistance reaches values of about 10% at room temperature for an alumina barrier thickness of about 16 Å. One may raise the question, why the observed TMR value is significantly smaller than the predictions from Jullière’s model. First of all, at room temperature there are magnetic excitations (magnons), which tend to reduce the spin polarization of the tunneling current. However, the TMR of this system at cryogenic temperature increases only by a factor of 2 to about 22% [34]. As a second mechanism the quality of the tunneling barrier and the barrier/electrode interfaces may play a role. Defects in the tunneling barrier and at the interfaces can introduce additional scattering and reduce the TMR. As serious shortcomings of the Jullière’ model, the detailed nature and properties of the tunneling barrier are not considered. In technological applications of the TMR effect, for example, in magnetic sensors or memory cells, the magnetic stability is an issue. Therefore, also in MTJs exchange biasing to an antiferromagnet is used to define the switching characteristics. The example in Figure 64.22 displays the TMR signal of a tunneling junction coupled

221

64 Spintronics: Surface and Interface Aspects

50 40 TMR (%)

222

Annealed at 300 °C

30 CoFe 20 10

Al2O3 CoFe MnFe

0 –1000

–500

0

500

1000

H (Oe) Figure 64.22 Tunneling magnetoresistance of a MnFe/CoFe/Al2 O3 /CoFe tunnel junction measured at room temperature after annealing at 300 ∘ C for improving the barrier homogeneity and quality. (Source: After Ref. [96].)

to MnFe, which leads to the asymmetry of the switching curve and a broad plateau extending up to almost 50 mT. The TMR value achieved on the plateau approaches almost 50% [96]. It should be pointed out that these TMR values were only obtained after an annealing step up to 300 ∘ C, which more than doubled the magnetoresistance value measured on the untreated system. This improvement is believed to be a consequence of barrier homogenization and the reduction of defects in the Al2 O3 barrier. At cryogenic temperatures the influence of magnetic excitations is reduced, and the TMR values increase up to 69 %. Applying Jullière’s formula to derive the spin polarization of the Co75 Fe25 electrodes, one finds a value of P ∼ 50%, which is close to the band polarization to be expected in this material. 64.5.3 Spin-Dependent Tunneling Beyond Jullière’s Model

The predictions from the Jullière model suggest that a value of the TMR beyond 100% may be achieved, if only the spin polarization of the magnetic electrodes can be sufficiently increased. A plot of the TMR magnitude as a function of the electrode polarization is given in Figure 64.23. This perspective initiated a broad and still ongoing materials research activity, exploring the properties of nominal highly spin-polarized materials. Among these are half-metallic ferromagnets, e.g. half and full Heusler alloys [97], oxides such as CrO2 or Fe3 O4 , or perovskites and double perovskites. However, we should point out that the prediction of a half-metallic state is usually derived from a bulk band structure calculation. At an interface, the electronic structure may be significantly changed as compared to the bulk, even in the case of an atomically sharp interface. These changes are related to the symmetry breaking at the interface, which can lead to the formation of electronic interface states. For instance, a major problem in the use of Heusler alloys – in addition to lattice site

64.5 Tunneling Magnetoresistance

500 400 500 400 200 300 100 200

TMR effect (%)

300

100

n tio iza ) lar 2 (% po in ode Sp lectr e

80

60 40 20 0

20

40

80 60 tion lariza %) o p ( in p 1 S rode elect

Figure 64.23 Tunneling magnetoresistance in Jullière’s model as a function of the electrode spin polarization values.

defects – is the appearance of electronic states at the gap, resulting in a partial or total loss of the half-metallicity [29]. So far we have focused on the role of the magnetic electrodes only. One can raise the question about alternative barrier materials. Most of the early studies have focused on alumina barriers, because of the vast body of experience on how to grow nanometer thin film with a low density of defects, e.g. pinholes. Generally, growing a good quality and continuous oxide film on a metal is a major challenge. In order to establish a reliable transport by tunneling, the density of highly conductive defects, i.e. “pinholes,” must be low. [98]. In the case of Al2 O3 barriers, the procedure that has been established involves the deposition of metallic aluminum followed by an oxidation step. In this way, an accidental oxidation of the magnetic electrode underneath can be mostly avoided. This is particularly important in the case of 3d ferromagnets as their oxides are often antiferromagnetic. An antiferromagnetic or paramagnetic metal/barrier interface significantly impairs the spin-dependent tunneling process [99]. Another issue is formation of correlated roughness of the interface, which particularly for extremely thin barrier may lead to an effective magnetic coupling

223

64 Spintronics: Surface and Interface Aspects

(Néel or “orange peel” coupling) between the ferromagnetic layers [100, 101]. This impairs the magnetic switching ability of the tunneling junction. Other tunneling barriers, which were considered, were related to insulating films used successfully in semiconductor technology, for instance, SiO2 , HfO, or MgO. In fact, replacing the amorphous Al2 O3 barrier by an epitaxial (100)-oriented MgO barrier while keeping all other components of the tunneling junction, the same led to a significant increase of the TMR [102]. This finding is already a good indication that the actual electronic structure of the barrier and at the interfaces does play a role and must be taken into account. In 2001 Butler et al. then predicted a “giant” TMR with MgO barriers [103]. These predictions were made for a single-crystalline tunneling contact with bcc-Fe(100) electrodes. An experimental proof for this prediction was given about three years later independently by the Yuasa group and the Parkin group [104, 105]. The result by Yuasa et al. is displayed in Figure 64.24. The MTJ used an epitaxial Fe(100)/ MgO(100)/Fe(100) layer sequence and was exchange-biased by a top IrMn layer, which gives rise to the characteristic switching curve. As we can see the resistance–area product (RA) changes from about 7 to 23 kΩ μm2 at cryogenic temperature. This corresponds to a TMR ratio of ΔR∕R0 = 274%. The corresponding value at room temperature amounts to ΔR∕R0 = 180%. The Parkin group employed a polycrystalline FeCo/MgO/FeCo layer stack grown by sputtering. The relatively low TMR in the as-grown state was strongly enhanced by a high-temperature annealing step, which led to a recrystallization in the MgO layer forming a strong (100) texture. In this way, room temperature values of the TMR of about ΔR∕R0 = 220% were obtained [105]. The quantum mechanical mechanism leading to this large TMR signal is an interplay of spin- and symmetry-dependent tunneling processes. The first thing that has to be taken into account is that there are complex evanescent electronic states in the MgO tunneling barrier, which participate in the tunneling process. The Bloch states in the ferromagnetic electrode (Fe) can couple to these barrier states, depending on T = 20 K T = 293 K

25 RA (kΩ μm2)

224

20 15 10 5

–200

–100

0 H (Oe)

100

200

Figure 64.24 TMR characteristics in epitaxial Fe(100)/MgO/Fe(100) tunneling junction at cryogenic and room temperature. The resistance times junction area product (RA) is given as a function of magnetic field. (Source: Taken from Ref. [104].)

64.5 Tunneling Magnetoresistance

Δ1

Density of states

1 10–5 10–10

Δ5

Fe

MgO

Fe

10–15 10–20 2′

10–25

2

3

4

5

6

7 8 9 10 11 12 13 14 15 Layer number

Figure 64.25 Theory of the TMR in the Fe/MgO system. Calculated DOS for different band symmetries inside the tunneling barrier. The highest DOS is obtained for a

Δ1 majority spin state. It dominates the spin transport through the MgO tunneling barrier. (Source: After Ref. [103].)

the symmetry character of the wave function. As a result, the decay of the Bloch states in the barrier depends on their symmetry. As it turns out by the theoretical treatment [103], the conductivity in the spin-up tunneling channel close to the Fermi level is particularly high for states with Δ↑1 symmetry. This behavior is depicted in Figure 64.25, which plots the change of the DOS for a particular state symmetry as a function of the position in the tunneling junction. It can be clearly seen that the states of Δ1 symmetry experience the least decay within the barrier. Other states close to the Fermi level of Fe that can participate in the tunneling process have symmetry Δ5 or Δ2′ . However, their DOS within the barrier decays much more strongly by orders of magnitude. In the Fe electrodes the Δ1 states are exchangesplit. The Δ↑1 band crosses the Fermi level and can contribute to the tunneling current, whereas the Δ↓1 band is more than 1 eV above the Fermi level and cannot contribute to tunneling. It is this selection of the predominant symmetry of the tunneling electrons by the MgO barrier and their spin by the Fe electrodes that strongly enhances the spin tunneling and leads to this favorable behavior of the Fe/MgO system. Taking the above role of the interface for the matching of electronic wave functions, one would expect the TMR to depend sensitively on the preparation conditions. The calculation assumes a structurally and chemically perfect interface. However, we have to keep in mind that binary oxides like Al2 O3 or MgO are usually grown by depositing the metal in the presence of an oxygen atmosphere, forming the oxide directly on the surface. This reaction-based growth process is not trivial in the case of an Fe surface, which is also quite susceptible to oxidation. Indeed, careful investigations of the Fe/MgO interface revealed the formation of an FeO interfacial layer under certain growth conditions [106, 107]. This layer affects the interface magnetism of Fe and with it the properties of the electronic states participating in the tunneling process. Spin-polarized photoemission showed that the spin polarization of the electronic states at the Fermi level is significantly reduced in the case

225

226

64 Spintronics: Surface and Interface Aspects

of a partial oxidation of the Fe [108]. These studies also showed that the reduction of the spin polarization can be avoided by growing a minute amount of metallic Mg prior to the growth of MgO. The reactivity of Mg with O is higher than that of Fe, and the metallic Mg basically protects the Fe from oxidizing during the initial phase of the MgO growth. This is an example that a carefully designed and controlled growth process is a must in the successful manufacturing of high-level spintronic devices.

64.6 Spin Transfer Torque

So far we have only considered passive device concepts and spintronic processes. Passive in this context means that the magnetization state affects the spin scattering and in this way changes the device resistance. In a quantum mechanical picture, the spin has the properties of an angular momentum. One may therefore ask the question about angular momentum conservation in the scattering process. In other words, what happens to the scattered spin component? 64.6.1 Microscopic Picture

Let us consider the situation at an interface between the spacer and the ferromagnetic electrode. For this we adopt the description by Stiles and Zangwill [109] (Figure 64.26). ⃗ which is nonLet us assume a spin-polarized current with a polarization vector P, ⃗ collinear with the magnetization direction M of the receiving electrode. The angle ⃗ is denoted 𝜙. Since M ⃗ defines the reference quantization axis between P⃗ and M in the ferromagnet, upon entering the ferromagnetic electrode, the spin current has to adapt to this quantization axis. This happens over a distance of a few lattice constants [109, 110]. In order to analyze the behavior of a single spin, we have to resort to a quantum mechanical description, which also means we have to transform Motion of electron Nonmagnet

ψin =

cos(θ/2)

Ferromagnet Torque ∝ sin(θ)

θ

sin(θ/2) ψtrans=

0 ψref = sin(θ/2)

cos(θ/2) 0

M

Figure 64.26 Simplified principle of spin transfer torque (STT).

64.6 Spin Transfer Torque

between the polarization vector P⃗ in three-dimensional space and the spin state in the two-dimensional spin space. This transformation is done by means of the Pauli spin matrices 𝜎⃗ = (𝜎x , 𝜎y , 𝜎z ). We choose the magnetization and thus the quantization axis in the ferromagnet along the z-axis and the electron propagation direction (current direction) along the x-axis. In this reference system, the incident spin state |𝜓in ⟩ is not a pure spin state and becomes a mixture of spin-up and spin-down contributions3) : ( ) cos 𝜙2 |𝜓in ⟩ = (64.8) sin 𝜙2 which yields the following projections of the spin polarization vector: ⟨𝜓in |𝜎x |𝜓in ⟩ = sin 𝜙, ⟨𝜓in |𝜎y |𝜓in ⟩ = 0, ⟨𝜓in |𝜎z |𝜓in ⟩ = cos 𝜙

(64.9)

For simplicity, we assume an ideal (e.g. half-metallic) ferromagnet in which only the spin-up component can propagate. The transmitted spin state traveling in the ferromagnet |𝜓T ⟩ then takes the form ) ( cos 𝜙2 (64.10) |𝜓T ⟩ = 0 The respective expectation values for the spin polarization in real space become 𝜙 (64.11) 2 and exhibit solely a z-component as expected. There is also a spin component |𝜓R ⟩ scattered back from the interface, which takes the form ( ) 0 (64.12) |𝜓T ⟩ = sin 𝜙2 ⟨𝜓T |𝜎x |𝜓T ⟩ = 0, ⟨𝜓T |𝜎y |𝜓T ⟩ = 0, ⟨𝜓T |𝜎z |𝜓T ⟩ = cos2

The expectation values of this spin component in real space become 𝜙 (64.13) 2 which describes a negative spin polarization contribution traveling backward through the interlayer to the other ferromagnetic electrode. Angular momentum conservation requires that incoming, transmitted, and reflected components should add up as ⟨𝜓R |𝜎x |𝜓R ⟩ = 0, ⟨𝜓R |𝜎y |𝜓R ⟩ = 0, ⟨𝜓R |𝜎z |𝜓R ⟩ = −sin2

⎛sin 𝜙⎞ ⎜ ⎟ ⟨𝜓in |⃗ 𝜎 |𝜓in ⟩ − (⟨𝜓R |⃗ 𝜎 |𝜓R ⟩ + ⟨𝜓T |⃗ 𝜎 |𝜓T ⟩) = ⎜ 0 ⎟ ⎜ ⎟ ⎝ 0 ⎠

(64.14)

The remaining angular momentum component along the x-direction (∼ sin 𝜙) describes a spin transfer to the spin system of the ferromagnetic electrode. This spin 3) Although discussing a single spin, the effect is ultimately related to the spin-polarized current density impinging onto the interface.

227

64 Spintronics: Surface and Interface Aspects

⃗ Relaxing the transfer effectively generates a torque acting on the magnetization M. constraint of an ideal ferromagnet, but taking into account the average over the whole ensemble of the electron states on the Fermi surface that contribute to the incoming spin-polarized current, yields to a good approximation the above result: the transverse component of the incoming spin current (sin 𝜙, 0, 0) is at the interface transferred to the ferromagnet [109]. This spin transfer torque (STT) must lead to some reaction in the ferromagnet in order to be dissipated, for example, by the excitation of spin waves. Such a behavior has been predicted by Berger [111] and Slonczewski [112] for magnetic multilayers. In order to observe a sizable effect, however a sufficiently high density of spin-polarized current is needed to build up the required spin torque. 64.6.2 Experimental Realization

One of the proof-of-principle experiments on STT was performed by Myers et al. in 1999 [113]. The sample comprised a nanocontact of about 10 nm diameter structured into a CPP Co/Cu/Co trilayer. The two cobalt layers are very different in thickness, resulting in a spin-valve configuration. The inset in Figure 64.27a shows the CPP-GMR signal of this nanocontact as a function of the external magnetic field. The butterfly curve shows that the top (hard) and bottom Co layers (soft) switch at distinctly different coercive fields.

dV/d/ (Ω)

16

14.0 –0.5

t1 = 4 nm µ 0H = 0 T

14 (a)

R (V = 0)

µ0H(T)

14.2

–4

–2

0.0

0.5

0

2

4

11 dV/d/ (Ω)

228

10 9

t1 = 2 nm µ 0H = 0 T –2

(b)

0 Current (mA)

Figure 64.27 Spin transfer torque switching in Co/Cu/Co nanocontacts. The resistance vs. current characteristics are given for two different Cu interlayer thicknesses, 4 nm (a)

2

and 2 nm (b). The inset in top panel depicts the CPP-GMR response of the nanocontact as a function of magnetic field in the range of ±0.5 kOe. (Source: After Ref. [113].)

64.6 Spin Transfer Torque

The curve in Figure 64.27a displays the variation of the CPP-GMR as a function of the current passing through the nanocontact. It shows several peculiar characteristics. First of all, there is an almost linear variation of the resistance at higher current values. This can be explained as a result of temperature increase in the nanocontact due to Joule heating. The current of a few mA corresponds to a current density of 109 Acm2 , which is actually close to the damage threshold caused by electromigration. The second feature is a sequence of spikes between 1 and 1.4 mA. These spikes are caused by the onset of magnetization excitations, which relate to spin waves and precessional motion of the local magnetization in the nanocontact [114]. The third feature is a distinct hysteretic switching between two branches of the curve. In a second sample with about half the Co thickness for the soft layer (Figure 64.27b), this switching takes place at considerably smaller current values. This experiment was a first proof that the phenomenon of STT can be used to switch a magnetic system deterministically between stable states. The pathway of the current-induced switching may be quite complex, as the current first introduces a precessional motion of the magnetization and then gradually increases the amplitude of the precession until the system has gained enough energy to jump to another low energy state. This process is illustrated with an experiment on single-crystal nanocontacts comprising a Fe/Ag/Fe trilayer combination [115]. The system was epitaxially grown on a Ag(100) template and exhibits a fourfold in-plane anisotropy with two orthogonal easy axes of the magnetization. The Ag interlayer thickness was chosen such as to effectively decouple the hard (fixed) and soft (free) Fe layers. Therefore, one observes a two-step switching of the free layer as seen in Figure 64.28. The almost linear increase of the differential resistance with current density is again due to Joule heating of the contact. However, one observes two distinct switching events, corresponding to rotation of the soft layer by 90∘ at the critical current IC1 and a further 90∘ switching at the critical current IC2 . This behavior can be described by micromagnetic (macro-spin) simulations, and the respective trajectories of the magnetization vector are displayed in Figure 64.29. As we can see the switching at IC1 occurs by an increasing precession of the magnetization vector around the easy axis at 0∘ . Only after the amplitude of the precession exceeds a critical value, the magnetization is able to switch to the neighboring easy axis by 90∘ rotation (Figure 64.29a). As long as the current is maintained, the magnetization vector will continue to precess around this new direction. If the current is increased, however, the precession amplitude increases again, and the magnetization vector switches to the neighboring easy axis at 180∘ (Figure 64.29b). The trajectory of this switching process is rather complex and does not take the shortest pathway. We should also note that there is an additional symmetry-breaking contribution to the switching, which arises from the dipolar (Oersted) field generated by the current passing through the nanocontact. This effect is included in the macrospin simulations in Figure 64.29.

229

64 Spintronics: Surface and Interface Aspects

Current density j (107 A/cm2) –20

–10

0

10

20

Differential resistance dU/dl (a.u.)

230

Ic1 –8

–4

0 4 DC current I (mA)

Ic2 8

Figure 64.28 Spin transfer torque switching in a single-crystalline Fe/Ag/Fe pseudo spin valve. The thicknesses are 20 nm Fe for the bottom (hard or fixed) layer, 6 nm for the Ag interlayer, and 2 nm Fe for the top (soft or free) layer. (Source: After Ref. [115].) Mfree Mfree

0°

–1

90°

–1 –0.5

0.5

11

(a)

–1

180° –0.5

–1

0 0.5

0

90°

–0.5 –0.5

Mfixed

0 0

0.5

0.5 11

Mfixed

(b)

Figure 64.29 Trajectories of the magnetization vector in the nanocontact in Figure 64.28 derived from macro-spin simulations. The simulations also include the effect of the Oersted field. (Source: After Ref. [115].)

64.7 Technological Perspectives

The technological perspectives for spintronics are vast. There are multiple applications of spintronic concepts and devices in information technology, mechanical and electrical engineering, medicine, etc., and the field is still growing. We will limit ourselves to just three applications, which can be directly related to the topics discussed in the above chapters.

64.7 Technological Perspectives

64.7.1 Magnetic Mass Storage

The first technological application of spintronics was introduced at the end of the 1990s revolutionizing magnetic data storage. In a hard disk or magnetic tape drive, the information is written into the magnetic storage material as a sequence of “0” and “1” bits, which are represented by magnetic domains of opposite magnetization directions. The magnetic stray field of these domains is read by a magnetoresistive sensor, which is brought in close proximity to the storage material. The AMR-based read heads for magnetic hard disk drives being in use in the mid-1990s were replaced by more sensitive GMR-based spin-valve devices. Note that this technological advance happened only 10 years after the discovery of GMR, which is an extremely short time. However, it was facilitated by the fact that magnetic thin film technology had already been developed with a focus on AMR, so that GMR did not come as a disruptive technology. Switching to GMR-based read heads in CIP geometry [116] started a breathtaking growth of the storage densities, making hard disk drives much more powerful and compact over the last 20 years. The typical magnetic bit size shrunk from 10 μm in 1990 to a few 10 nm in 2010. Around 2006 the CIP-GMR heads were replaced by TMR-based systems again boosting the hard disk storage density toward the TB/inch2 regime. The TMR read heads involve a CPP geometry, and the size of the sensor can be more easily adapted to the size of the magnetic domains. Currently there is a discussion whether or not CIP-GMR heads may replace the TMR heads in the future [117]. 64.7.2 Magnetic Sensors

Many applications in engineering require not only a contact-less monitoring of positions and angles but also rotation speeds. A particularly illustrative example is automotive applications [118, 119]. In today’s cars we find several dozens of magnetic sensors not only at and around the engine but also in other areas of the car. At the engine these sensors monitor the crank shaft speed and position with directional information. The antilock brake systems rely on magnetic sensors that measure the wheel speed. Electrical power steering (EPS) systems involve electrical motors, where the position of the rotor needs to be precisely known. Another quantity that needs to be measured is the steering angle and the position of the steering wheel. Typically, the measured data from all of these sensors are also input to the stability program of the car and are used to avoid critical driving conditions. Let us focus only on one example, namely, the angle sensor. We have already discussed above that in a spin valve with weak interlayer coupling, the soft magnetic layer may be easily rotated by an external field against the hard magnetic layer (Figure 64.30a). In other words, the GMR signal will be a function of the angle 𝜃 ⃗ h of the soft and hard layer, respectively ⃗ s and M between the magnetization vectors M [120]: 1 (64.15) R(𝜃) = RP + (RAP − RP )(1 − cos 𝜃) 2

231

64 Spintronics: Surface and Interface Aspects

6 N S

M

ΔR/R0 (%)

232

4

2

0 (a)

(b)

0°

90°

180°

270°

360°

Angle

Figure 64.30 Rotation angle sensor using the spin-valve concept. (a) Principle of the rotation sensor with CIP-GMR geometry. (b) Angular variation of the GMR signal in a rotation angle sensor. Inset: packaged GMR angle sensor (for illustration purpose only).

This equation describes the ideal situation, i.e. the residual coupling between the two magnetic layers and any magnetic anisotropies are negligible, and the magnetic field is considerably smaller than the coercive field of the hard layer. In reality, R(𝜃) may deviate more or less strongly from Eq. (64.15). By means of simulations the deviations may be quantified and related to the microscopic mechanisms [121]. The example in (Figure 64.30a ) shows the angular dependence of the GMR signal from an exchange-biased spin-valve system, which comes quite close to the ideal variation R(𝜃). However, one notes a small hysteresis depending on the direction of angular variation. This is another issue that must be taken into account in the sensor design. Another important property of GMR sensors, in particular, in automotive applications is the temperature stability. Close to the engine temperatures of up to 150 ∘ C may be reached, and close to the brake peak temperatures may easily exceed 250 ∘ C. GMR sensors for these applications have to be therefore very robust, which requires particular design rules [122]. 64.7.3 Nonvolatile Core Memory

Another key application of spintronics was also proposed quite early: the magnetic random access memory. The principle of a magnetic core memory providing nonvolatility of data was already around before the now ubiquitous semiconductor memory. The ring core memory in the 1960s consisted of an array of ferrite rings [123]. Each ring supported a bit in the form of a permanent clockwise or anticlockwise magnetization structure, which was written and read inductively. This architecture was not scalable and was therefore replaced by semiconductor memory, however losing the advantage of data nonvolatility. Semiconductor core memory needs frequent refreshing cycles in the 100 MHz range to maintain the information stored. The energy consumption problem of modern computer systems brings up this question again (nonvolatility). The simplest architecture approach to

64.7 Technological Perspectives

an MRAM would be a crossbar architecture with a GMR spin valve or an MTJ at each crossing point. The information “0” and “1” can be encoded in the resistivity of the spin valve in parallel (RP ) and antiparallel (RAP ) configurations. However, for this simple architecture to work, the ON/OFF ratio of the magnetic memory cell, i.e. the ratio RAP ∕RP , should be typically of the order of 103 to avoid parasitic currents through neighboring cells. Unfortunately, the magnitude of the GMR is only in the order of 50% and so is the ON/OFF ratio. This requires additional semiconductor circuitry to select a contact for read and write. In the 1980s the idea of an MRAM using magnetoresistive elements was put forward, and 256 kbit samples were manufactured around 1990, for example, by Honeywell. These systems still used AMR as magnetoresistive effect [124]. The driving force for this development were aerospace applications, where radiation hardness is a major requirement. In this respect metallic systems have a significant advantage over semiconductor devices. Later versions of the Honeywell chips also used GMR technology [125]. Nowadays 64 M chips are available [126]. A serious limitation of the GMR-MRAM is the limited resistivity ratio RAP ∕RP . Therefore, MTJs have been considered in MRAM development, and significant progress has been made in several companies [127]. There are numerous technological challenges that need to be solved, particularly with respect to high-density MRAM. The magnetic and GMR characteristics of the individual MTJ must fall within narrow limits to ensure a sufficiently high yield in fabrication [128]. During the last 10 years, a variety of different MRAM concepts based on MTJs has evolved [129]. The conventional way of switching the state of an MTJ MRAM cell uses external magnetic fields. A sketch of the respective architecture is shown in Figure 64.31a. The MTJ is connected to a bit line on the top and to a selection transistor at the bottom. The MRAM consists of an array of these bit lines with MTJs at regular distances. During the write operation a high current passing through the bit line generates a magnetic field that switches the free layer of the MTJ. In high-density MRAMs this magnetic field acts also on the MTJs, which are connected to the same bit line. This may eventually cause an unwanted (partial) switching in the neighboring MTJs. This cross talk is thus an considerable problem in field-switched MRAM. As an improvement Motorola has introduced the “toggle switching” mechanism [130]. It uses an additional write line, which runs perpendicular to the bit line (Figure 64.31a). The magnetization in the soft layer, which is actually designed as a synthetic antiferromagnet, is switched by applying a sequence of three pulses to the bit and write lines. In this way, the probability of cross talk between neighboring MRAM cells is greatly reduced, and the reliability of the switching is greatly enhanced. Currently, MRAMs with toggle technology are available up to 16 Mb capacity [131]. Despite the success of the toggle switching approach, this is still not yet the approach to high-density MRAM. The main reason is a dilemma in nanomagnetism. In order to obtain high integration densities in MRAM, the individual MTJ must be reduced in size far below the 100 nm limit. In order to maintain the magnetic stability of the MTJ, materials with sufficiently large magnetic anisotropy

233

234

64 Spintronics: Surface and Interface Aspects

Toggle switching

Magnetic field

Spin transfer torque switching

Bit line Free layer Tunnel barrier Fixed layer

Flux concentrating cladding layer Inlaid copper interconnects

Isolation transistor OFF

Isolation transistor OFF

(a) Figure 64.31 MRAM architectures. (a) Toggle switching approach by means of a sequence of magnetic field pulses applied by the bit and write lines. Chip image for illustration purpose only. (b) Spin–torque

Free layer Tunnel barrier Fixed layer

(b) switching. Chip image for illustration purpose only. (Source: Information taken from https:// www.everspin.com/toggle-mram-technology [131].)

and coercivity must be chosen to withstand thermal excitations. As a consequence, however, the material becomes harder to switch, which requires higher magnetic fields, i.e. current densities in the case of field-switched MRAMs. This trend is sketched in Figure 64.32, which gives the result of a simulation depicting the field switching current I as a function of the size of the MTJ cell. It is apparent that above densities of 100 Mb/cm2 , the switching current strongly increases. As a consequence, the field switching mode in MRAM is approaching both power and thermal dissipation issues. A solution is provided by the STT mechanism discussed in Chapter 6. For the same MTJ cell sizes, the development of the current needed for STT is shown. Even at current densities JC = 5 × 106 A/cm2 , the currents needed for STT are significantly lower than those for the field switching mode and become even smaller with reduced feature size. The current density needed for an efficient STT switching depends strongly on the material system; therefore another critical current line for JC = 1 × 106 A/cm2 is included. From these simulations it becomes clear that STT is the switching mechanism of choice, if a higher storage density in MRAMs is to be achieved. Consequently, in recent years we have witnessed broad range of activities around the development of STT-MRAM concepts. The architecture for spin–torque switching is illustrated in Figure 64.31b. A selection transistor is still needed to address the MTJ cell reliably.

References

1 Gb/Cm2

10

100 Mb/Cm2

I (mA)

MRAM 5

STT-RAM JcD = 5 × 106 A/cm2 JcD = 1 × 106 0

0

50

150 100 Magnetic cell width (nm)

Figure 64.32 Development of the switching currents with the magnetic feature size, i.e. size of an MTJ. Compared are the currents for field switching (MRAM) and spin–torque

200

switching (STT-RAM) for a critical current density of JC = 5 ⋅ 106 A/cm2 . For comparison, also a current variation for JC = 1 ⋅ 106 A/cm2 is included.

On the market, one may find already 256 Mb STT-MRAM chips [131]. Respective Gb devices are announced for the near future. The perspective of MTJs goes well beyond nonvolatile memory. There have been pathways outlined to an MTJ-based architecture for logics, or even reconfigurable logics [132]. In principle, MTJs offer the potential to realize memory and logics using the same physical device principle, which can enable novel and more energy-efficient chip architectures. Acknowledgment

I am indebted to Daniel Bürgler for scientific discussions and critical reading of the manuscript. I would like to acknowledge the hospitality of the Lawrence Berkeley National Laboratory where parts of this manuscript were written.

References 1. Moore, G.E. (1975). Progress in digi-

tal integrated electronics, EDM Tech. Digest 11. 2. Waldrop, M.M. (2016). The chips are down for Moore’s law. Nature 530: 144. 3. Hoefflinger, B. (ed.) (2012). Chips 2020 - A Guide to the Future of Nanoelectronics. Berlin: Springer-Verlag. 4. Van Heddeghem, W., Lambert, S., Lannoo, B. et al. (2014). Trends in

worldwide ICT electricity consumption from 2007 to 2012. Comput. Commun. 50: 64. 5. Kim, J.H. and Lee, M.J. (eds) (2011). Green IT: Technologies and Applications. Berlin: Springer-Verlag. 6. Wang, S.X. and Taratorin, A.M. (1999). Magnetic Information Storage Technology. San Diego, CA: Academic Press.

235

236

64 Spintronics: Surface and Interface Aspects 7. Forrester, J.W. (1951). Digital informa-

8.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20.

tion storage in three dimensions using magnetic cores. J. Appl. Phys. 22 (1): 44. Thomson, W. (1856). On the electrodynamic qualities of metals:–effects of magnetization on the electric conductivity of nickel and of iron. Proc. R. Soc. London 8: 546. Coey, J. (2009). Magnetism and Magnetic Materials. Cambridge: Cambridge University Press. Ashcroft, N. and Mermin, N. (1976). Solid State Physics. Philadelphia, PA: Saunders College. McGuire, T. and Potter, R. (1975). Anisotropic magnetoresistance in ferromagnetic 3d alloys. IEEE Trans. Magn. 11: 1018. Mott, N.F. (1936). The electrical conductivity of transition metals. Proc. R. Soc. London, Ser. A 153 (880): 699–717. Fert, A. and Campbell, I.A. (1968). Two-current conduction in nickel. Phys. Rev. Lett. 21 (16): 1190. O’Handley, R.C. (2000). Modern Magnetic Materials: Principles and Applications. New York: Wiley. Binasch, G., Grunberg, P., Saurenbach, F., and Zinn, W. (1989). Enhanced magnetoresistance in layered magnetic structures with antiferromagnetic interlayer exchange. Phys. Rev. B 39 (7): 4828. Baibich, M.N., Broto, J.M., Fert, A. et al. (1988). Giant magnetoresistance of (001)Fe/(001)Cr magnetic superlattices. Phys. Rev. Lett. 61 (21): 2472. Kroemer, H. (2001). Nobel lecture: quasielectric fields and band offsets: teaching electrons new tricks. Rev. Mod. Phys. 73 (3): 783. Apalkov, D., Dieny, B., and Slaughter, J.M. (2016). Magnetoresistive random access memory. Proc. IEEE 104 (10): 1796. Bland, J. and Heinrich, B. (1994). Ultrathin Magnetic Structures, vol. I. Berlin: Springer-Verlag. Heinrich, B. and Bland, J. (1994). Ultrathin Magnetic Structures, vols II. Berlin: Springer-Verlag.

21. Bland, J. and Heinrich, B. (2005). Ultra-

22.

23.

24.

25.

26.

27. 28. 29.

30.

31.

32.

33.

34.

35.

36.

thin Magnetic Structures, vols III. Berlin: Springer-Verlag. Heinrich, B. and Bland, J. (2005). Ultrathin Magnetic Structures, vol. IV. Berlin: Springer-Verlag. Gomonay, E.V. and Loktev, V.M. (2014). Spintronics of antiferromagnetic systems. Low Temp. Phys. 40 (1): 17. Jungwirth, T., Marti, X., Wadley, P., and Wunderlich, J. (2016). Antiferromagnetic spintronics. Nat. Nanotechnol. 11: 231. Baltz, V., Manchon, A., Tsoi, M. et al. (2018). Antiferromagnetic spintronics. Rev. Mod. Phys. 90: 015005. Datta, S. and Das, B. (1990). Electronic analog of the electro-optic modulator. Appl. Phys. Lett. 50: 665. Rashba, E.I. (1960). Rashba effect. Sov. Phys. Solid State 2: 1109. Bychkov, Y.A. and Rashba, E.I. (1984). JETP Lett. 39: 78. Galanakis, J. (2002). Surface properties of the half- and full-Heusler alloys. J. Phys. Condens. Matter 14: 6329. Attema, J.J., de Wijs, G.A., and de Groot, R.A. (2006). The continuing drama of the half-metal/semiconductor interface. J. Phys. D: Appl. Phys. 39: 793. Meservey, R. and Tedrow, P.M. (1994). Spin-dependent tunneling. Phys. Rep. 238: 173. Fert, A. and Campbell, I.A. (1976). Electrical resistivity of ferromagnetic nickel and iron based alloys. J. Phys. F: Met. Phys. 6 (5): 849. Jullière, M. (1975). Tunneling between ferromagnetic films. Phys. Lett. 54A (3): 225. Moodera, J.S., Kinder, L.R., Wong, T.M., and Meservey, R. (1995). Large magnetoresistance at room temperature in ferromagnetic thin film tunnel junctions. Phys. Rev. Lett. 74 (16): 3273. Miao, G.-X., Münzenberg, M., and Moodera, J.S. (2011). Tunneling path toward spintronics. Rep. Prog. Phys. 74 (3): 036501. Ohno, H. (1998). Making nonmagnetic semiconductors ferromagnetic. Science 281: 951.

References 37. Dietl, T., Ohno, H., Matsukura, F. et al.

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

48.

(2000). Zener model description of ferromagnetism in zinc-blende magnetic semiconductors. Science 287: 1019. Meier, F. and Zakharchenya, B. (eds) (1984). Optical Orientation, Modern Problems in Condensed Matter Sciences. Amsterdam: North-Holland. Mikkelsen, M.H., Myers, R.C., Fuchs, G.D., and Awschalom, D.D. (2008). Single Spin Coherence in Semiconductors, 1. Amsterdam: Elsevier. Fiederling, R., Keim, M., Reuscher, G. et al. (1999). Injection and detection of a spin-polarized current in a lightemitting diode. Nature 402 (6763): 787. Ohno, Y., Young, D.K., Beschoten, B. et al. (1999). Electrical spin injection in a ferromagnetic semiconductor heterostructure. Nature (London) 402: 790. Schmidt, G., Ferrand, D., Molenkamp, L.W. et al. (2000). Fundamental obstacle for electrical spin injection from a ferromagnetic metal into a diffusive semiconductor. Phys. Rev. B 62 (8): R4790. Katsnelson, M.I., Irkhin, V.Y., Chioncel, L. et al. (2008). Half-metallic ferromagnets: from band structure to many-body effects. Rev. Mod. Phys. 80: 315. Rashba, E.I. (2000). Theory of electrical spin injection: tunnel contacts as a solution of the conductivity mismatch problem. Phys. Rev. B 62 (24): R16267. Fert, A. and Jaffrès, H. (2001). Conditions for efficient spin injection from a ferromagnetic metal into a semiconductor. Phys. Rev. B 64: 184420. Jiang, X., Wang, R., Shelby, R.M. et al. (2005). Highly spin-polarized room-temperature tunnel injector for semiconductor spintronics using MgO(100). Phys. Rev. Lett. 94: 056601. Bhat, S.G. and Kumar, P.S.A. (2014). Room temperature electrical spin injection into GaAS by an oxide spin injector. Sci. Rep. 4: 5588. Dash, S.P., Sharma, S., Patel, R.S. et al. (2009). Electrical creation of spin polarization in silicon at room temperature. Nature 462 (7272): 491.

49. Li, C.H., van ’t Erve, O.M.J., and

50.

51.

52.

53.

54.

55.

56.

57.

58.

59.

Jonker, B.T. (2011). Electrical injection and detection of spin accumulation in silicon at 500 K with magnetic metal/silicon dioxide contacts. Nat. Commun. 2: 245. https://doi.org/10 .1038/ncomms1256. Miao, G.-X., Müller, M., and Moodera, J.S. (2009). Magnetoresistance in double spin filter tunnel junctions with nonmagnetic electrodes and its unconventional bias dependence. Phys. Rev. Lett. 102 (7): 076601. Schmehl, A., Vaithyanathan, V., Herrnberger, A. et al. (2007). Epitaxial integration of the highly spin-polarized ferromagnetic semiconductor EuO with silicon and GaN. Nat. Mater. 6 (11): 882. Caspers, C., Gloskovskii, A., Gorgoi, M. et al. (2016). Interface engineering to create a strong spin filter contact to silicon. Sci. Rep. 6: 22912. Hasan, M.Z. and Kane, C.L. (2010). Colloquium: topological insulators. Rev. Mod. Phys. 82: 3045. Wang, K.L., Lang, M., and Kou, X. (2016). Spintronics in Topological Insulators, 431. Dordrecht: Springer Science. Grünberg, P., Schreiber, R., Pang, Y. et al. (1986). Layered magnetic structures: evidence for antiferromagnetic coupling of Fe layers across Cr interlayers. Phys. Rev. Lett. 57 (19): 2442. Bürgler, D.E., Grünberg, P., Demokritov, S., and Johnson, M. (2001). Interlayer Exchange Coupling in Layered Magnetic Structures, vol. 13, 3. Amsterdam: Elsevier Science Publishers. Carbone, C., Vescovo, E., Rader, O. et al. (1993). Exchange split quantum well states of a noble metal film on a magnetic substrate. Phys. Rev. Lett. 71 (17): 2805. Nordström, L., Lang, P., Zeller, R., and Dederichs, P. (1995). Ab initio calculations of quantum well states in Cu/Co(100). EPL 29 (5): 395. Qiu, Z. and Smith, N. (2002). Quantum well states and oscillatory magnetic interlayer coupling. J. Phys. Condens. Matter 14: R169.

237

238

64 Spintronics: Surface and Interface Aspects 60. Unguris, J., Celotta, R., and Pierce, D.

61.

62.

63.

64.

65.

66.

67.

68.

69.

70.

71.

(1991). Observation of two different oscillation periods in the exchange coupling of Fe/Cr/Fe(100). Phys. Rev. Lett. 67 (1): 140. Schmidt, C.M., Bürgler, D.E., Schaller, D.M. et al. (1999). Correlation of shortperiod oscillatory exchange coupling to nanometer-scale lateral interface structure in Fe/Cr/Fe(001). Phys. Rev. B 60 (6): 4158. Leng, Q., Cross, V., Schäfer, R. et al. (1993). Interlayer coupling through Au. J. Magn. Magn. Mater. 126: 367. Bruno, P. (1995). Theory of interlayer magnetic coupling. Phys. Rev. B 52 (1): 411. Stiles, M.D. (1999). Interlayer exchange coupling. J. Magn. Magn. Mater. 200: 322. Bloemen, P.J.H., van Kesteren, H.W., Swagten, H.J.M., and de Jonge, W.J.M. (1994). Oscillatory interlayer exchange coupling in Co/Ru multilayers and bilayers. Phys. Rev. B 50: 13505. Hecker, M., Tietjen, D., Prokert, F. et al. (2000). Investigation of Co/Cu/NiFemultilayers by X-ray reflectometry and diffraction. Mikrochim. Acta 133: 239. Parkin, S., Li, Z., and Smith, D. (1991). Giant magnetoresistance in antiferromagnetic Co/Cu multilayers. Appl. Phys. Lett. 58: 2710. De Miguel, J.J., Cebollada, A., Gallego, J.M. et al. (1991). Influence of the growth-conditions on the magneticproperties of fcc cobalt films - from monolayers to superlattices. J. Magn. Magn. Mater. 93: 1. Camarero, J., Spendeler, L., Schmidt, G. et al. (1994). Surfactant-induced suppression of twin formation during growth of fcc Co/Cu superlattices on Cu(111). Phys. Rev. Lett. 73: 2448. Elefant, D., Tietjen, D., Schaefer, R. et al. (2002). High-field magnetoresistance, giant magnetoresistance, and superparamagnetism in Co/Cu multilayers. J. Appl. Phys. 91 (10): 8590–8592. Samant, M., Stöhr, J., Parkin, S. et al. (1994). Induced spin polarization in Cu spacer layers in Co/Cu multilayers. Phys. Rev. Lett. 72 (7): 1112.

72. Coehoorn, R. (2003). Giant Magne-

73.

74.

75.

76.

77. 78.

79.

80.

81.

82.

83.

toresistance and Magnetic Interactions in Exchange-Biased Spin-Valves, vol. 15, 4. Amsterdam: Elsevier Science Publishers. Nogues, J. and Schuller, I.K. (1999). Exchange bias. J. Magn. Magn. Mater. 192: 203. Meiklejohn, W. and Bean, C. (1956). New magnetic anisotropy. Phys. Rev. 102: 1413. Berkowitz, A. and Takano, K. (1999). Exchange anisotropy - a review. J. Magn. Magn. Mater. 200: 552. Rijks, T.G.S.M., Coehoorn, R., Daemen, J.T.F., and de Jonge, W.J.M. (1994). Interplay between exchange biasing and interlayer exchange coupling in Ni80 Fe20 /Cu/Ni80 Fe20 /Fe50 Mn50 layered systems. J. Appl. Phys. 76 (2): 1092. Parkin, S. (1992). Py/Cu multilayers and GMR. Appl. Phys. Lett. 60: 512. Parkin, S. (1993). Origin of enhanced magnetoresistance from magnetic multilayers: spin-dependent scattering from magnetic interface states. Phys. Rev. Lett. 71: 1641. Renard, J., Bruno, P., Mégy, R. et al. (1996). Inverse giant magnetoresistance (invited). J. Appl. Phys. 79 (8): 5270. Wiesendanger, R., Güntherodt, H.J., Güntherodt, G. et al. (1990). Observation of vacuum tunneling of spin-polarized electrons with the scanning tunneling microscope. Phys. Rev. Lett. 65: 247. Kleiber, M., Bode, M., Ravli´c, R., and Wiesendanger, R. (2000). Topologyinduced spin frustrations at the Cr(001) surface studied by spin-polarized scanning tunneling spectroscopy. Phys. Rev. Lett. 85 (21): 4606. Schneider, C., Meinel, K., Kirschner, J. et al. (1996). Element specific imaging of magnetic domains in multicomponent thin film systems. J. Magn. Magn. Mater. 162: 7. Krug, I.P., Hillebrecht, F.U., Gomonaj, H. et al. (2008). Magnetic coupling in highly ordered NiO/Fe3 O4 (110): ultrasharp magnetic interfaces vs. longrange magnetoelastic interactions. EPL 81 (1): 17005.

References 84. Krug, I.P., Hillebrecht, F.U., Haverkort,

85.

86.

87.

88.

89.

90.

91.

92.

93.

M.W. et al. (2008). Impact of interface orientation on magnetic coupling in highly ordered systems: a case study of the low-indexed Fe[sub 3]O[sub 4]/NiO interfaces. Phys. Rev. B 78 (6): 064427. Radu, F. and Zabel, H. (2007). Exchange Bias Effect of Ferro-/Antiferromagnetic Heterostructures, Springer Tracts in Modern Physics, 97. Berlin: SpringerVerlag. Hütten, A., Mrozek, S., Heitmann, S. et al. (1999). Evolution of the GMReffect amplitude in copper/permalloymultilayered thin films, Acta Mater. 47 (15–16): 4245. Parkin, S. (1992). Dramatic enhancement of interlayer exchange coupling and giant magnetoresistance in Ni81 Fe19 /Cu by addition of thin co interface layers. Appl. Phys. Lett. 61 (11): 1358. Matthes, F., Rzhevskii, A., Tong, L.-N. et al. (2003). Growth and magnetic properties of epitaxial ultrathin Ni films on Cu(111) using Sb as a surfactant. J. Appl. Phys. 93: 8740. Egelhoff, W.F., Chen, P.J., Powell, C.J. et al. (1996). Optimizing the giant magnetoresistance of symmetric and bottom spin valves. J. Appl. Phys. 79 (8): 5277. Egelhoff, W.F., Chen, P.J., Powell, C.J. et al. (1997). Oxygen as a surfactant in the growth of giant magnetoresistance spin valves. J. Appl. Phys. 82 (12): 6142. Veloso, A., Freitas, P.P., Wei, P. et al. (2000). Magnetoresistance enhancement in specular, bottom-pinned, Mn83 Ir17 spin valves with nano-oxide layers. Appl. Phys. Lett. 77: 1020. Wang, L., Qiu, J.J., McMahon, W.J. et al. (2004). Nano-oxide-layer insertion and specular effects in spin valves: experiment and theory. Phys. Rev. B 69: 214402. Schmid, A.K., Atlan, D., Itoh, H. et al. (1993). Fast interdiffusion in thin films: scanning-tunneling-microscopy determination of surface diffusion through microscopic pinholes. Phys. Rev. B 48 (4): 2855.

94. Hecker, M., Thomas, J., Tietjen, D.

95.

96.

97.

98.

99.

100.

101.

102.

103.

104.

et al. (2003). Thermally induced modification of GMR in Co/Cu multilayers: correlation among structural, transport, and magnetic properties. J. Phys. D: Appl. Phys. 36: 564. Meservey, R., Paraskevopoulos, D., and Tedrow, P.M. (1976). Correlation between spin polarization of tunnel currents from 3d ferromagnets and their magnetic moments. Phys. Rev. Lett. 37 (13): 858–860. Han, X.-F., Daibou, T., Kamijo, M. et al. (2000). High-magnetoresistance tunnel junctions using Co75 Fe25 ferromagnetic electrodes. Jpn. J. Appl. Phys. 39: L439. Felser, C. and Fecher, G.H. (2013). Spintronics: From Materials to Devices. Dordrecht: Springer Science and Business Media. Han, X. and Yu, A. (2004). Patterned magnetic tunnel junctions with al conduction layers: fabrication and reduction of pinhole effect. J. Appl. Phys. 95 (2): 764. Jansen, R. and Moodera, J.S. (1998). Influence of barrier impurities on the magnetoresistance in ferromagnetic tunnel junctions. J. Appl. Phys. 83 (11): 6682. Schrag, B.D., Anguelouch, A., Ingvarsson, S. et al. (2000). Néel “orange-peel” coupling in magnetic tunneling junction devices. Appl. Phys. Lett. 77 (15): 2373. Tegen, S., Monch, I., Schumann, J. et al. (2001). Effect of neel coupling on magnetic tunnel junctions. J. Appl. Phys. 89 (12): 8169. Bowen, M., Cros, V., Petroff, F. et al. (2001). Large magnetoresistance in Fe/MgO/FeCo(001) epitaxial tunnel junctions on GaAs(001). Appl. Phys. Lett. 79 (11): 1655. Butler, W.H., Zhang, X.-G., Schulthess, T.C., and MacLaren, J.M. (2001). Spindependent tunneling conductance of Fe/MgO/Fe sandwiches. Phys. Rev. B 63: 054416. Yuasa, S., Nagahama, T., Fukushima, A. et al. (2004). Giant room-temperature magnetoresistance in single-crystal Fe/MgO/Fe magnetic tunnel junctions. Nat. Mater. 3: 868.

239

240

64 Spintronics: Surface and Interface Aspects 105. Parkin, S.S.P., Kaiser, C., Panchula, A.

106.

107.

108.

109.

110.

111.

112.

113.

114.

115.

116.

117.

et al. (2004). Giant tunnelling magnetoresistance at room temperature with MgO(100) tunnel barriers. Nat. Mater. 3: 862. Meyerheim, H.L., Popescu, R., Kirschner, J. et al. (2001). Geometrical and compositional structure at metaloxide interfaces: MgO on Fe(001). Phys. Rev. Lett. 87 (7): 076102. Meyerheim, H.L., Popescu, R., Jedrecy, N. et al. (2002). Surface X-ray diffraction analysis of the MgO/Fe(001) interface: evidence for an FeO layer. Phys. Rev. B 65 (14): 144433. Muller, M., Matthes, F., and Schneider, C.M. (2007). Spin polarization at ferromagnet-insulator interfaces: the important role of stoichiometry in MgO/Fe(001). EPL 80 (1): 6. Stiles, M.D. and Zangwill, A. (2002). Anatomy of spin-transfer torque. Phys. Rev. B 66: 014407. Zhang, S., Levy, P.M., and Fert, A. (2002). Mechanisms of spin-polarized current-driven magnetization switching. Phys. Rev. Lett. 88: 236601. Berger, L. (1996). Emission of spin waves by a magnetic multilayer traversed by a current. Phys. Rev. B 54 (13): 9353. Slonczewski, J. (1996). Current-driven excitation of magnetic multilayers. J. Magn. Magn. Mater. 150: L1. Myers, E.B., Ralph, D.C., Katine, J.A. et al. (1999). Current-induced switching of domains in magnetic multilayer devices. Science 285: 867. Tsoi, M., Jansen, A.G.M., Bass, J. et al. (1998). Excitation of a magnetic multilayer by an electric current. Phys. Rev. Lett. 80 (19): 4281. Lehndorff, R., Buchmeier, M., Burgler, D.E. et al. (2007). Asymmetric spintransfer torque in single-crystalline Fe/Ag/Fe nanopillars. Phys. Rev. B 76 (21): 6. Kools, J. (1996). Exchange-biased spinvalves for magnetic storage. IEEE Trans. Magn. 32: 3165. Maat, S. and Marley, A.C. (2016). Physics and Design of Hard Disk Drive Magnetic Recording Read Heads, vol. 2.

118.

119.

120.

121.

122.

123.

124.

125.

126.

127.

128.

Dordrecht: Springer Science+Business Media. Treutler, C.P.O. (2001). Magnetic sensors for automotive applications. Sens. Actuators, A 91: 2. Kapser, K., Weinberger, M., Granig, W., and Slama, P. (2013). GMR Sensors in Automotive Applications, Smart Sensors, Measurement and Instrumentation, Vol. 6, 133. Berlin: Springer-Verlag. Coehoorn, R. (1996). Giant Magnetoresistance in Exchange-Biased Spin-Valve Structures and its Application in Read Heads, 63. Berlin: Springer-Verlag. Tietjen, D., Elefant, D., and Schneider, C.M. (2002). Rotation angle sensors based on spin valve structures: a modeling approach. J. Phys. D: Appl. Phys. 91: 5951. Giebeler, C., Adelerhof, D.J., Kuiper, A.E.T. et al. (2001). Robust GMR sensors for angle detection and rotation speed sensing. Sens. Actuators, A 91: 16. technikum29 Living Museum (2018). Historic internal storage media. http:// www.technikum29.de/en/computer/ storage-media (accessed 30 June 2018). Pohm, A.V., Comstock, C.S., and Hurst, A.T. (1990). Quadrupled nondestructive outputs from magnetoresistive memory cells using reversed word fields. J. Appl. Phys. 67: 4881. Prinz, G.A. and Hathaway, K. (1995). Special issue: magnetoelectronics. Phys. Today 48: 58. Honeywell (2015). Radiation-hardened 64 Mb Nonvolatile MRAM from Honeywell. https://aerospace.honeywell .com/en//media/aerospace/files/ datasheet/hxnv06400-datasheet.pdf (accessed 30 May 2017). Gallagher, W.J. and Parkin, S.S.P. (2006). Development of the magnetic tunnel junction MRAM at IBM: from first junctions to a 16-Mb MRAM demonstrator chip. IBM J. Res. Dev. 50 (1): 5. Koh, G.H., Kim, H.J., Jeong, W.C. et al. (2004). Fabrication of high performance 64kb MRAM. J. Magn. Magn. Mater. 272–276: 1941.

References 129. Yoda, H. (2016). Handbook of Spin-

tronics, 1031. Netherlands, Dordrecht: Springer. 130. Engel, B.N., Akerman, J., Butcher, B. et al. (2005). A 4-Mb toggle MRAM based on a novel bit and switching method. IEEE Trans. Magn. 41 (1): 132–136. 131. Everspin Technologies Inc. (2018). Everspin’s toggle MRAM technology.

https://www.everspin.com/toggle-mramtechnology (accessed 30 December 2018). 132. Schneider, C.M. (2012). Spin-based logics: Principles and concepts. In R. Waser (Ed.), Nanoelectronics and Information Technology. Wiley-VCH: Weinheim.

241

243

65 Device Efficiency of Organic Light-Emitting Diodes1) Wolfgang Brütting

65.1 Introduction

The availability of artificial light has been a seminal cultural development of mankind. After open fires for thousands of years, the introduction of electricity together with the invention of the light bulb in the second half of the nineteenth century has revolutionized our daily life. Having dominated the lighting market for more than 100 years, however, the light bulb faces its decline due to the need for technologies that convert electricity more efficiently into visible light. Besides fluorescent lamps, which are already well established in the market, the availability and progress in white light-emitting diodes, both inorganic and organic, over the last decade has led to a new lighting technology called solid-state lighting [1–3]. Its working principle, namely, the radiative recombination of injected electron–hole pairs in a solid, a process termed electroluminescence, is fundamentally different from thus far existing techniques and holds the promise for highly efficient, long-lived, and environmentally friendly light sources. In contrast to their inorganic counterparts, organic light-emitting diodes (OLEDs) are flat and thin large area light sources that could rather lead to complementary luminaires than competitors to existing technologies. Historically speaking, electroluminescence in organic molecular crystals dates back to the early 1960s [4, 5]; however, the important step toward applicable devices was the demonstration of thin film OLEDs by researchers at Eastman Kodak in 1987 [6] using vacuum-deposited molecular materials and by a group at Cambridge University in 1990 [7], making use of a solution-processed conjugated polymer. Inspired by these publications, intense research and development has led to the introduction of first commercial products based on OLED displays in the late 1990s. Since 2000 the focus in many laboratories shifted toward OLEDs for lighting applications, i.e. to white OLEDs. After steady improvements in efficiency and lifetime over the years, OLEDs have already become part of our everyday life, 1) This chapter is an updated version of an article published in Physica Status Solidi A 210 (2013) 44–65. Surface and Interface Science: Applications of Surface Science I, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

244

65 Device Efficiency of Organic Light-Emitting Diodes

e.g. as displays for smartphones or, more recently, in high-end television screens. However, we are still waiting for their market entry on a larger scale in the general lighting sector [8]. The scope of this article is to present the working principles of OLEDs and to review ongoing efforts to improve their efficiency, in particular through enhanced light outcoupling. Therefore we will restrict ourselves to devices fabricated by vacuum deposition of low-molecular-weight materials (small molecules). Nevertheless, most of the following discussion is equally valid for polymer OLEDs where thin films are prepared by solution processing of large macromolecules. For an overview of various technical aspects of OLED fabrication and the current state of the art of white OLEDs for lighting applications, see Refs. [9–15].

65.2 OLED Operation 65.2.1 OLED Architecture and Stack Layout

In principle, an OLED can consist of only one organic layer sandwiched between two electrodes, i.e. the anode and the cathode [7]. However, a multilayer OLED made of several different functional organic layers is superior in terms of efficiency and lifetime because the properties of the individual layers can be tuned through the proper choice of materials. A typical stack layout of such an OLED is displayed in Figure 65.1. The positively biased anode is required in order to inject holes into Encapsulation Cathode Electron transport layer (ETL) Emission layer (EML) Voltage

Hole transport layer (HTL) Anode (ITO) Glass substrate

Light emission Figure 65.1 Simplified illustration of a three-layer OLED stack. ETL, EML, and HTL denote the electron transport layer, the emission layer(s), and the hole transport layer,

respectively. In order to obtain light emission, a positive voltage must be applied to the anode, which often consists of optically transparent indium tin oxide (ITO).

65.2 OLED Operation

the hole transport layer (HTL). Similarly, electrons are injected from the cathode into the electron transport layer (ETL). The emission layer (EML) comprises one or more dyes emitting light in different regions of the visible spectrum. Highefficiency OLEDs often contain several additional layers like separate blocking and injection layers. The most common substrate material is glass, but it is also possible to employ other materials, like polymer or metal foils or semiconductor substrates. One of the two metallic or metallic conducting electrodes must be semitransparent so that the produced light can leave the device either to the substrate side (bottom emission) or through the upper electrode (top emission). Frequently, the anode of an OLED consists of indium tin oxide (ITO), which is highly transparent and shows a good electrical conductivity. In general, anode materials need a high work function, which also makes metals like gold a suitable choice. For the cathode, low work function metals like calcium or magnesium are often used in combination with highly reflective and less reactive metals like aluminum or silver. Since many organic materials and low work function metals are not stable under ambient conditions [16], i.e. they react with oxygen and moisture, it is necessary to protect the materials with an encapsulation. The most common method utilizes a cover glass that is glued to the substrate, thus creating a cavity filled with inert gas or a desiccant acting as a getter to absorb small amounts of oxygen and water that penetrate through the encapsulation in the course of time. The active area of the OLED is defined as the overlap of the two electrodes. Hence, OLEDs can in principle have any shape and size, and they are therefore suited for a wide range of applications. Laboratory OLEDs have typical dimensions of a few square millimeters. In industrial prototypes the active area of the OLEDs has already been scaled up to sizes in the range 0.1–1 m2 . The thickness of the organic layers is typically in the range between 10 and a few hundred nanometers. 65.2.2 Working Principles of OLEDs

The basic processes in an OLED under operation are illustrated in Figure 65.2 for a device with three organic layers. An external voltage source of typically a few volts is applied to the device so that the two types of charge carriers are injected from the opposite electrodes, i.e. electrons from the cathode and holes from the anode, and drift toward each other. When the initially free electrons and holes meet, they form strongly bound electron–hole pairs (excitons) in the EML, which then may decay radiatively and emit photons. In detail, the whole process can be separated into four fundamental steps as denominated in Figure 65.2: (1) Injection of electrons and holes at the electrodes. (2) Transport of charge carriers through the organic layers. (3) Formation of bound electron–hole pairs (excitons). (4) Radiative exciton decay and emission of light.

245

246

65 Device Efficiency of Organic Light-Emitting Diodes

vel

um le

Vacu

ΦC (2)

(1)

O

LUM

ΦBI

(3)

ΦA

(4)

Exciton

e·V

(3)

O

(1)

Anode

HOM (2)

Hole injection Emission Electron injection and transport layer layer and transport layer

Figure 65.2 Schematic energy diagram of a three-layer OLED illustrating the basic processes of electroluminescence. Electrons are injected from the cathode into the lowest unoccupied molecular orbital (LUMO) of the ETL, and holes from the anode into the highest occupied molecular orbital (HOMO) of the HTL. Since the anode and cathode have different work functions ΦA and ΦC , respectively, a built-in voltage ΦBI exists in the device, which has to be overcome by an external voltage V before current can flow through the device. If materials are properly chosen, recombination of electrons and holes will be confined to the EML, and the flow of

Cathode

excess carriers not recombining will be minimized. It should be noted that owing to disorder the energy levels of all involved states are not discrete but distributed in energy, thus leading to important consequences for charge carrier transport and recombination. (For further details on this aspect, we refer to the literature, e.g. [17, 18].) In addition, the possibility of trap-assisted recombination (e.g. [19]) is not included in this diagram. Furthermore, vacuum level shifts and other effects occurring at interfaces between different materials (see, e.g. [20]) are not taken into account.

One distinct difference between OLEDs based on molecular materials and polymer OLEDs is the extent and location of the recombination zone inside the device. While in the latter case the emission zone is spread over a sizeable fraction of the light-emitting polymer layer (depending of course on the detailed material properties) [21], in small molecule OLEDs, the generation of light can be confined by a proper choice of materials with suitable energy levels and layer thicknesses to a relatively narrow zone sandwiched between HTL and ETL. Moreover, if doped hole and electron injection layers are used [22], this region can be placed at almost arbitrary distance away from the electrodes. This is favorable because recombination near the electrodes usually causes quenching and therefore a reduction in efficiency. Further on, as will be discussed later in this article, this additional degree of freedom is a handle to improve OLEDs with respect to light outcoupling by making use of optical interference effects.

65.2 OLED Operation

65.2.3 OLED Materials

The development of OLEDs has been enabled by and will continue to rely on the availability of tailor-made functional organic materials that can be applied to well-controlled thin films. Therefore the requirements to the materials are manifold: starting from processability and film formation via electrical transport to optical properties. The key factor is obviously the availability of efficient and stable light emitters in the full visible spectral range. In this respect one has to distinguish between fluorescent and phosphorescent materials. A seminal step was the introduction and further development of emitters based on heavy-metalcentered metal–organic complexes, as shown in Figure 65.3. In these compounds strong spin–orbit coupling mixes singlet and triplet states much more than in pure hydrocarbons, so that phosphorescence becomes an allowed transition [23]. In the

N

N

N

N

N

N

α-NPD

TPD

N

N N

S-TAD

O N

Al O

O N

N

N

N

BCP

N

BPhen

Alq3 N N

N N

N

Ir

N

O Ir O

N 2

Ir(ppy)3

CBP

Ir(MDQ)2 (acac)

O

N

n

O

N N

S

n

SO3H

BDASBi

Figure 65.3 Selection of organic materials used for OLEDs that are discussed in this article. TPD, α-NPD, and S-TAD are hole transport materials. Alq3 , BCP, and BPhen are electron transport materials, with Alq3 often being used as green fluorescent emitter material, too. CBP serves as matrix material for the green phosphorescent emitter

PEBA

PEDOT

PSS

Ir(ppy)3 or the two blue fluorescent emitters BDASBi and PEBA. The red phosphorescent emitter Ir(MDQ)2 (acac) is doped into an αNPD host layer. PEDOT:PSS is a conducting polymer that can be used as hole injection layer, because it effectively increases the work function of ITO and serves as a planarization layer.

247

77 nm

S-T AD

30 nm PEDOT :PSS

ITO

140 nm

Glass

3

10

2

10

1

10

0

10

−1

10

−2

10 −4

−2

−3

0

2

4

6

10

25

30

20

25 15

20 15

Ir(ppy)3 OLED

5

5 0

Voltage (V)

10

Alq3 OLED

10

3

4

5

6

7

Luminous efficacy (lm/W)

Ca Alq3

35

4

10

2 Luminance (cd/m )

60 nm

10 nm

Alq3 OLED

2

10 1 10 0 10 −1 10 −2 10 −3 10 −4 10 −5 10 −6 10 −7 10 −8 10

Current efficiency (cd/A)

Al

100 nm

2 Current density (mA/cm )

65 Device Efficiency of Organic Light-Emitting Diodes

0

Ca Alq3

10 nm 60 nm 10 nm

BCP

20 nm

Ir(ppy)3:CBP

TPD

60 nm 30 nm

PEDOT:PSS ITO

140 nm

Glass

Ir(ppy)3 OLED

2

10 1 10 0 10 −1 10 −2 10 −3 10 −4 10 −5 10 −6 10 −7 10 −8 10

4

10

3

10

2

10

1

10

0

10

−1

10

−2

10

−3

−4

−2

0

2

Voltage (V)

Figure 65.4 Layer structures of two prototypical OLEDs with Alq3 as fluorescent and Ir(ppy)3 as phosphorescent green emitters, respectively. Also shown are their

4

6

10

2 Luminance (cd/m )

Ag

180 nm

Photoluminescence intensity (normalized)

Voltage (V)

Current density (mA/cm2)

248

1.0 0.8

Alq3 OLED Ir(ppy)3 OLED

0.6 0.4 0.2 0.0 400

500 600 700 Wavelength (nm)

800

current–voltage–luminance characteristics as well as their electroluminescence emission spectra and their efficiency.

meantime, impressive efficiency data have been published for OLEDs based on these materials [13, 24, 25]; however, the bottleneck is still the limited availability and stability of deep-blue phosphorescent emitters. A selection of common organic materials that were used for fabricating OLEDs in our group is displayed in Figure 65.3. Figure 65.4 shows electroluminescence characteristics of two prototypical OLEDs, one with the fluorescent emitter Alq3 and the other one with the phosphorescent material Ir(ppy)3 . Both will serve as reference devices throughout this article. 65.2.4 White OLEDs

One of the peculiarities of organic emitters is their intrinsically broad luminescence spectrum. By the combination of several organic emitters with different emission color, it is possible to generate light spreading over the whole visible spectral range, thus having an excellent color rendering index (CRI), i.e. the ability to reproduce the color of illuminated objects. There are several methods in order to create such white OLEDs. The first report dates back to 1994, when a Japanese group combined red, green, and blue (RGB) laser dyes in a common matrix and achieved light emission over a broad spectral range [26]. Despite its simplicity in preparation, achieving white light emission with good and stable color quality in this way is not that straightforward, because charge recombination and energy transfer processes between the different dyes need to be well controlled. Thus it is nowadays more common to employ distinct EML for RGB that can either be directly stacked on top of each

65.3 Electroluminescence Quantum Efficiency

other in one OLED (see Figure 65.1) or in three separated OLEDs – one for each color – that are in turn vertically stacked by optically transparent electrical interconnecting units. In addition, white light from OLEDs may be produced similar to inorganic LEDs, i.e. by combining a blue OLED with either red and green or yellow color conversion layers [27]. Depending on the method and the choice of materials, it is thus possible to cover a large variety of colors and color temperatures within the CIE 1931 color space (Commission internationale de l’éclairage). Further details about different concepts of white OLEDs and the current state of the art can be found in Refs. [10–13].

65.3 Electroluminescence Quantum Efficiency 65.3.1 Factors Determining the EQE

Apart from its spectral characteristics, the most important parameter characterizing an OLED is the external electroluminescence quantum efficiency (EQE) that describes the ratio between the number of emitted photons and injected charge carriers [28]: 𝜂EQE = 𝛾 ⋅ 𝜂S∕T ⋅ qeff ⋅ 𝜂out

(65.1)

where 𝛾 is the charge carrier balance factor, describing whether or not equal amounts of electrons and holes are injected and what fraction of them recombines to form an exciton. The second factor 𝜂S∕T gives the fraction of excitons that is allowed to decay radiatively by spin statistics. The third factor qeff indicates how many of the spinallowed excitons actually do decay by emitting a photon (instead of dissipating the excitation energy non-radiatively to their environment). Finally, the last factor 𝜂out determines which fraction of the generated photons is in the end able to leave the device to the outside world. Hence the external quantum efficiency can be split into an internal quantum efficiency 𝜂IQE = 𝛾 ⋅ 𝜂S∕T ⋅ qeff times the outcoupling factor 𝜂out . In detail (see Figure 65.2), the charge carrier balance 𝛾 depends on the numbers of electrons and holes that are injected and the fraction of them, which is consumed by recombination: 𝛾=

jR jtot

(65.2)

with jtot = jh + je′ = je + jh′ being the total current density and jR = jh − jh′ = je − je′ the recombination current density. Therein jh,e are the injected hole and electron currents, respectively, and the primed quantities are the corresponding fractions of carriers, leaving the device at the opposite electrode without recombination. Under ideal conditions the latter will be zero and consequently 𝛾 = 1. However, if there is an imbalance of the numbers of injected electrons and holes or if recombination is not complete, then there is excess of charge carriers that does not contribute to the

249

250

65 Device Efficiency of Organic Light-Emitting Diodes

production of light, and the efficiency of the OLED is reduced. In small molecule multilayer OLEDs, the charge carrier balance can be brought close to unity by using doped transport layers and additional selective carrier and exciton blocking layers [22, 29]. The so-called singlet/triplet factor 𝜂S∕T describes the probability for the formation of an exciton that is allowed to decay radiatively according to the spin selection rules. Under electrical operation, singlet and triplet excitons are formed by recombination of electrons and holes, each of them carrying spin 12 . The probability of forming a triplet exciton with total spin S = 1 is statistically three times higher than the probability of forming a singlet exciton with S = 0, because the spin orientations of the injected electrons and holes are random and every triplet state is threefold degenerate. Therefore, 𝜂S∕T is believed to have a value of 25% for fluorescent emitters, which only make use of singlet excitons, while it can be 100% for phosphorescent emitters, where both singlets and triplets contribute to light emission [30, 31]. We note that even for singlet emitters, 𝜂S∕T might be enhanced beyond 25% by triplet–triplet annihilation (TTA) [32–35] as well as thermally activated delayed fluorescence (TADF) [36, 37] in certain material systems. In the context of conjugated polymer systems, there has been a debate whether 𝜂S∕T might intrinsically be larger than 25% (even in the absence of TTA or TADF) [38–40]; however, recent results indicate that this is probably not the case [41]. As will be discussed below, an unambiguous assignment requires a careful efficiency analysis including effects of non-isotropic orientation of the light-emitting chromophores. All in all, a significant efficiency enhancement can be obtained by using phosphorescent materials; hence 𝜂S∕T can be raised to 1 [23, 42]. The effective radiative quantum efficiency qeff is derived from the intrinsic radiative quantum efficiency q of the emitter material, which is obtained if the emitting species is surrounded by an unbounded medium in the limit of very low excitation densities. Usually the photoluminescence quantum efficiency measured by using a thin film of the emitter embedded in the same matrix as in the OLED comes close to this value. Per definition q is given as q=

Γr Γr + Γnr

(65.3)

where Γr is the radiative decay rate of the excited state and Γnr is the sum of all competing non-radiative decay rates. It is apparent that every non-radiative contribution to the exciton decay reduces the radiative quantum efficiency. In addition, in an OLED environment the presence of stratified media with different refractive indices and the vicinity to metallic electrodes lead to a modification of the radiative decay rate of an exciton via the Purcell effect: Γr → Γ∗r = F ⋅ Γr (with F being the Purcell factor) [43–45]. By contrast, the non-radiative decay rates, e.g. the dissipation of excitation energy into heat, are not affected by the cavity environment [46]. Thus, the effective radiative quantum efficiency qeff can be defined as qeff =

Γ∗r

Γ∗r F ⋅ Γr = + Γnr F ⋅ Γr + Γnr

(65.4)

65.4 Fundamentals of Light Outcoupling in OLEDs

Depending on the details of the layer stack, the cavity effect can either increase or reduce qeff with respect to the intrinsic value q. Consequently, the optimization of the OLED cavity is very important not only in terms of the light outcoupling efficiency, as will be discussed below, but also in terms of enhancing the radiative decay processes in cases where the emitter does not a priori have q = 1. And even if q is close to 1, one has to be aware that in an operating device the radiative quantum efficiency can be significantly reduced by bimolecular quenching of excitons or interactions with charge carriers at high current densities [47]. This is especially relevant for long-lived triplet excitons (triplet–triplet quenching, triplet–polaron quenching) as will be shown later on. With the abovesaid, the internal quantum efficiency of OLEDs can be brought up toward the theoretical limit of 100%, if charge carrier injection and recombination are well balanced, if phosphorescent emitters are used, and if non-radiative exciton quenching processes are suppressed [29]. Nevertheless, only a fraction of the light will in the end be able to leave the device to the outside world. The reason is that light is generated in a region of the OLED stack with higher refractive index than the glass substrate and, obviously, ambient air. In a simple model based on ray optics, the light outcoupling efficiency is given by [48] 1 (65.5) 2n2 where n denotes the (average) refractive index of the organic layer stack. With typical values of n = 1.6–1.8, one immediately finds that only between 15 and 20% of the optical power is actually extracted from an OLED. However, this should only be taken as a rough estimation; a more sophisticated analysis has to take the coupling of the excited molecules to the modes of the OLED cavity into account, as will be discussed in detail in the next section. 𝜂out =

65.4 Fundamentals of Light Outcoupling in OLEDs 65.4.1 Optical Loss Channels

As already mentioned above, a ray optical model can only give a rough estimate of the light extraction efficiency 𝜂out of an OLED. For a more accurate treatment, wave optical methods are required, which will be addressed in this section. An excited molecule can couple to different optical modes in such a thin film structure (see Figure 65.5). Viewed from the emitter position, the light escape cone has an opening angle of some 30∘ with respect to the surface normal, and the energy contained in it typically amounts to less than 20% of the total energy. This is followed by the contribution of substrate modes, where total internal reflection at the glass/air interface is the limiting process. This contribution is comparable in energy also at around 20%. For higher emission angles the light can not even reach the glass substrate, but is waveguided in the organic layers (including the transparent ITO electrode) and in

251

252

65 Device Efficiency of Organic Light-Emitting Diodes

Surface plasmons Cathode Emitter

Waveguide modes

Organic layers

ϕ

Anode Emission to substrate

Substrate Glass prism

Emission to air

Figure 65.5 Schematic illustration of an OLED showing different optical loss channels. Without outcoupling enhancement, only a small fraction of light is directly emitted to air as shown in the left part of the figure.

By applying a macro-extractor, e.g. an indexmatched glass hemisphere, also the substrate emission can be extracted (cf. the right half of the figure).

Surface plasmons: 43.7 % Waveguide modes: 11.0 %

Absorption: 5.7 % Emission to substrate: 24.3 % Figure 65.6 Amount of power coupled to different optical channels in the prototypical Alq3 OLED from Figure 65.4. The numbers are obtained by integrating each region in the power dissipation spectrum shown

Emission to air: 15.3 % in Figure 65.8a. In this diagram, a radiative quantum efficiency of q = 1 is assumed. The influence of q-values below unity is discussed in the text.

the end lost by residual absorption or edge emission. Finally, the emitter can couple to the evanescent field of surface plasmon polaritons (SPP) traveling at the interface between the metal electrode and the organic layers. Quantitative calculations, treating the emitting molecules as classical electrical dipoles (for details see below), reveal that in planar OLED stacks, typically around 50% of the light is trapped in waveguide and plasmon modes [44, 49, 50]. As an example, Figure 65.6 shows the contribution of different optical channels for the prototypical Alq3 OLED stack discussed before in Figure 65.4. Note that this chart is obtained with a radiative quantum efficiency q = 1. It thus describes the maximum light outcoupling of the stack under ideal conditions. The given numbers are valid regardless if the emitter is fluorescent (e.g. Alq3 ) or phosphorescent (e.g. Ir(ppy)3 ), as long as the layer stack has comparable thicknesses and refractive indices and the emission spectra are not too different. Given the low number of directly emitted light, it is therefore not surprising that developing new concepts for improving light extraction efficiency has been a major

65.4 Fundamentals of Light Outcoupling in OLEDs

issue over recent years (for an overview see, e.g. Ref. [51]). Some of these approaches (including our own results) will be presented in Section 65.5. 65.4.2 Optical Modeling of OLEDs

Fluorescence and phosphorescence of excited molecules are not intrinsic properties of a material, but are modified by the optical environment of the emitting species. Originally discovered for magnetic resonance, this so-called Purcell effect [52] has been found to be relevant also in experiments with fluorescent molecules near interfaces [53, 54]. After its experimental demonstration Chance, Prock, and Silbey (CPS) developed a theoretical framework by applying the classical theory of an oscillating electrical dipole near a dielectric interface to the problem of molecules fluorescing near a surface [55, 56]. Thereby they made use of the fact that the probability for the emission of a photon by an excited molecule via a dipole transition is equivalent to the power radiated by a classical dipole antenna. The formalism has later on been extended to structures where the emitter is embedded in a layer stack with multiple interfaces [57] and to microcavity structures [58, 59]. Here we follow the approach by Wasey and Barnes [60], where the problem of an incoherent ensemble of dipole emitters being embedded in a planar thin film structure is solved by a plane wave expansion of the electric field with appropriate consideration of the electromagnetic boundary conditions (see Figure 65.7). Formally the whole OLED stack is split into an upper and a lower half with respect to the emitter position, and the propagation of radiation in both directions is calculated by taking into account the Fresnel reflection and transmission coefficients at the +++---+++---+++---+++---+++---

θ

k

Excited state τ0 = (Γr + Γnr)−1

z

k|| +

d

Γnr

Γr

–

Ground state

Figure 65.7 Schematic illustration of an oscillating electrical dipole embedded in a dielectric layer, which is sandwiched between two interfaces. The upper material is assumed to be a metal, so that surface plasmon polaritons are supported at this interface. The other parameters are the layer thickness d, the distance z of the dipole to the metal, its relative orientation 𝜃 to ⃗ and the surface normal, the wave vector k, its in-plane component k⃗|| . The quantum

mechanical analogue is a two-level system consisting of the energy levels of the excited state of the molecule and the ground state with radiative and non-radiative decay rates between them, leading to a finite lifetime of the excited state. Note that if the dipole is embedded in a cavity, the radiative rate will be modified to become Γ∗r = F ⋅ Γr (with F being the Purcell factor, see text) and the excited state lifetime accordingly 𝜏 = (Γ∗r + Γnr )−1 .

253

254

65 Device Efficiency of Organic Light-Emitting Diodes

involved interfaces. Key parameters entering into the model are the position (z) of the emitter with respect to interfaces (in particular the highly reflective cathode in bottom-emitting structures), the emitter orientation (𝜃), the thicknesses (d) and optical constants (refractive index n, extinction coefficient 𝜅) of all involved layers, and the intrinsic radiative quantum efficiency (q) of the emitter (often embedded in a matrix material). The presence of the OLED cavity now has a twofold effect on the properties of light emission with respect to the case when the emitter resides in an infinite medium with the same index of refraction: it will modify the radiative decay rate of the emitter (thus leading to an effective radiative quantum efficiency qeff in Eq. (65.1)), and it will redistribute the optical power between different channels (thus determining the light outcoupling factor 𝜂out ). According to Ref. [60] the relative decay rate (Γ∕Γ0 ) of an ensemble of isotropically oriented and mutually incoherent dipoles emitting at a single wavelength 𝜆 takes the following form: Γ = (1 − q) + q ⋅ P̃ = ∫0 Γ0

∞

P(k|| , 𝜆, z)

dk||

(65.6)

where Γ = Γ∗r + Γnr is the decay rate of the emitter modified by the presence of the cavity, Γ0 = Γr + Γnr is the original decay rate without a cavity, q is the intrinsic radiative quantum efficiency (again for the emitter in a homogeneous unbounded medium), and P is the layer specific power dissipation function depending on the in-plane wave vector k|| , the wavelength 𝜆, and the emitter position within the layer stack denoted by z. For the sake of simplicity, we will always use a discrete emitter position, i.e. an infinitely sharp recombination zone inside the OLED. This assumption is not so far from reality for small molecule OLEDs, where the exciton formation zone is confined by using a thin dye-doped EML surrounded by selective carrier blocking layers. It is, however, quite straightforward to extend the formalism to an extended emission zone by introducing some distribution function in z and averaging over all emitter positions (see, e.g. [21, 61]). The same holds for the emitter orientation. The calculation is actually performed for the three fundamentally different dipole orientations: – P⊥,TM : dipoles oriented perpendicular to the substrate plane (vertical), which emit p-polarized (transverse magnetic (TM)) light. – P||,TM : dipoles oriented parallel to the substrate plane (horizontal), which emit p-polarized (transverse magnetic (TM)) light. – P||,TE : dipoles oriented parallel to the substrate plane (horizontal), which emit s-polarized (transverse electric (TE)) light. For symmetry reasons, the contributions of the two horizontal dipole components are equal, so that in the case of a random distribution of dipole orientations, P is given as P(iso) = 13 P⊥ + 23 P|| . If not otherwise stated, we will in the following assume isotropic emitter orientation. The effect of non-isotropic orientation on the light outcoupling efficiency will be discussed separately at the end of Section 65.5. Figure 65.8 shows the power dissipation spectrum corresponding to the prototypical OLED stack shown in Figure 65.4, where the emission is assumed to take

65.4 Fundamentals of Light Outcoupling in OLEDs Normalized in-plane wave vector 800

Intensity

750

1.0 0.8 0.6 0.4 0.2 0.0

700 650 600 550 500 1

2

3

4

1.5

2.0

2.5

3.0

0.6 0.4

0.0

1× 107 2× 107 3× 107 4 ×107 (a)

1.0

0.2

450 400

0.5

0.8 Intensity (a.u.)

Wavelength (nm)

0.0 1.0

0

−1

In-plane wave vector (m )

7

2× 10

7

3 ×10

7

4× 10

−1

(b)

Figure 65.8 (a) Simulation of the total dissipated optical power for the Alq3 OLED shown in Figure 65.4. The emitter position is assumed at the interface between Alq3 and S-TAD. The result is shown for an isotropic dipole orientation. Red and yellow areas indicate a high amount of dissipated power. The dashed white lines divide the graph into four

1 ×107

In-plane wave vector (m )

regions: (1) emission into air, (2) emission into substrate, (3) waveguide modes, and (4) coupling to surface plasmons. (b) Cross section of (a) at a wavelength of 510 nm, i.e. at the emission maximum. The upper axis shows the normalized in-plane wave vector with respect to the vacuum wave vector at a wavelength of 510 nm. (See also Ref. [44].)

place in Alq3 directly at the interface to the HTL. Furthermore we have used a hypothetical radiative quantum efficiency q = 1 and weighted the simulation results for different wavelengths between 400 and 800 nm with the normalized photoluminescence spectrum S(𝜆) of the emitter. Also shown in the figure is a cross section at a wavelength of 510 nm, where the different optical loss channels can be clearly identified. According to Eq. (65.6) this power dissipation spectrum can now be used to calculate the total relative decay rate by integrating over all relevant wavelengths (between 𝜆1 = 400 nm and 𝜆2 = 800 nm) and in-plane wave vectors: P̃ tot = (1 − q) + q ⋅

𝜆2

∫𝜆1

∞

S(𝜆)

∫0

P(k|| , 𝜆, z) dk|| d𝜆

≡ (1 − q) + q ⋅ F

(65.7)

The latter equivalence follows from the definition of the decay rates in free space Γ0 = Γr + Γnr and inside the cavity Γ = Γ∗r + Γnr = F ⋅ Γr + Γnr . Thus the integral of the spectrally weighted power dissipation directly yields the Purcell factor F. With that the effective radiative quantum efficiency of the emitter in the OLED cavity is given by Γ qeff F ⋅ (Γr + Γnr ) F =F⋅ 0 = = q F ⋅ Γr + Γnr Γ (1 − q) + q ⋅ F

(65.8)

Of course, this concept is equally valid for non-isotropic emitter orientation, if the calculation of the Purcell factor is performed separately for each of the three orthogonal orientations (x, y being parallel to the substrate plane and z being perpendicular to it) and by taking deviations from randomness into account via an orientation

255

256

65 Device Efficiency of Organic Light-Emitting Diodes

Θ=0

0.0

0.33

Horizontal

1.0

Isotropic

Vertical

Figure 65.9 Relation between the orientation parameter Θ and the orientation distribution of molecular transition dipole moments.

parameter Θ so that F in the previous expression is replaced by F(Θ) (for details see Ref. [62]): 1−Θ (65.9) ⋅ (Fx + Fy ) + Θ ⋅ Fz 2 where the orientation parameter Θ is defined as the fraction of power radiated by vertical dipoles, i.e. molecules with a component pz of their transition dipole moment vector p⃗ pointing along the surface normal of the film, in relation to the total optical power radiated by all dipoles. If 𝜃 is the angle of the dipole axis to the surface normal (see Figure 65.7), the orientation parameter Θ is given by [63] ∑ 2 pz = ⟨cos2 𝜃⟩ (65.10) Θ= ∑ p⃗ 2 F(Θ) ≡

Loosely speaking, Θ denotes the fraction of vertical dipoles: Θ = 1∕3 for isotropic orientation, Θ = 0 for completely horizontal (x–y-plane) orientation, and Θ = 1 for fully vertical (z-direction) emitter orientation, respectively, as shown schematically in Figure 65.9. We note that in this terminology the coupling of the excited molecules to modes of the cavity that do a priori not radiate into the far field, such as waveguide modes or surface plasmons, is a radiative process, since they contribute to F and thus change the radiative decay rate Γr . (Also note that in practice the integration over k|| is only carried out up to a finite cutoff value, where the contribution of surface plasmons has dropped to sufficiently small values (in most cases at around 4 × 107 m−1 ).) The simulated power dissipation spectrum can furthermore be used to calculate the fraction of power of the dipole that goes into different optical channels as discussed in the previous section: P̃ mode = q ⋅

𝜆2

∫𝜆1

ku

S(𝜆)

∫kl

P(k|| , 𝜆, z)

dk||

d𝜆,

(65.11)

with suitable lower (kl ) and upper (ku ) integration limits for each channel. For this purpose the spectrum shown in Figure 65.8 has to be subdivided into different regions according to the allowed range of the in-plane wave vector of the radiation to be able to couple to distinct optical modes. For example, for direct emission into air (P̃ air ), the in-plane wave vector must satisfy the condition 0 ≤ k|| ≤ (nair ⋅ k0 ) with k0 = 2𝜋∕𝜆 being the vacuum wave vector at a given wavelength 𝜆. In addition,

65.4 Fundamentals of Light Outcoupling in OLEDs

in this case also transmission losses are included in the calculation. Finally, by dividing the power contribution of the air mode by the total power dissipation, one arrives at P̃ (65.12) qeff ⋅ 𝜂out = air P̃ tot

(a)

Simulation p-pol. without prism

(b)

Measurement s-pol. without prism

(e)

Measurement p-pol. without prism

(f)

0° 15° 30° 45° 60° 75° 90° 0° 15° 30° 45° 60° 75° 90°

Angle

Angle

Figure 65.10 Simulation (upper row, a–d) and measurement (lower row, e–h) of the spectrally resolved angular-dependent emission of the Alq3 OLED in the case of sand p-polarized light. Left half is data for

Wavelength (nm)

800 750 700 650 600 550 500 450 400

Simulation s-pol. without prism

800 750 700 650 600 550 500 450 400

Wavelength (nm)

Wavelength (nm)

800 750 700 650 600 550 500 450 400

Wavelength (nm)

from which the outcoupling factor (for direct emission to air) is easily obtained, if qeff is known from Eq. (65.8). As an example, Figure 65.6 shows a pie chart of the different fractions of power to which the emitting molecules in a prototypical OLED dissipate their excitation energy. Obviously, the aforementioned 1∕(2n2 ) estimation based on ray optics is not that far from the calculated value of about 15% for the air mode. However, this is due to the fact that the distance of the emitter to the cathode was chosen not too far from the values favorable for light outcoupling (see also Figure 65.13). As will be shown below, this number can vary strongly if the spacing between the emitter and a metallic electrode is varied. It is further evident from this chart that the amount of light that is accessible directly and via substrate light extraction is limited to about 40% in this particular OLED stack. The biggest loss is the coupling to surface plasmons, amounting to more than 40% of the total power. In order to verify simulation results with experimental data, one can on the one hand use measured EQE values (without and with a macro-extractor attached to the glass substrate), which is an angle-integrated quantity. However, as expressed in Eq. (65.1) 𝜂EQE is a product of four factors so that additional assumptions are necessary. On the other hand one can also use angular- and polarization-dependent emission spectra containing much more information. Figures 65.10 and 65.11 show a comparison of measured and simulated spectra where all observed features (in particular the appearance of an s-polarized substrate mode) are well reproduced. Moreover, a detailed analysis of the angular emission spectra allows to recalibrate the layer thicknesses of the OLED stack (as shown in Figure 65.11) and to determine

800 750 700 650 600 550 500 450 400

Simulation s-pol. with prism

(c)

Simulation p-pol. with prism

(d)

Measurement s-pol. with prism

(g)

Measurement p-pol. with prism

(h)

Intensity 1.0 0.8 0.6 0.4 0.2 0.0

0° 15° 30° 45° 60° 75° 90° 0° 15° 30° 45° 60° 75° 90°

Angle

Angle

the planar OLED, and right half shows the results for an index-matched fused silica halfcylinder prism attached as a macro-extractor to the glass substrate. (See also Ref. [44].)

257

65 Device Efficiency of Organic Light-Emitting Diodes

0.8 0.6 0.4 0.2 0.0 0°

(a)

1.0 Intensity (a.u.)

1.0 Intensity (a.u.)

258

s-pol. measured s-pol. simulated p-pol. measured p-pol. simulated

15°

30°

45° Angle

0.8 0.6 0.4 0.2

60°

75°

0.0 0°

90° (b)

s-pol. measured s-pol. simulated p-pol. measured p-pol. simulated

15°

30°

45°

60°

75°

90°

Angle

Figure 65.11 Cross section at a wavelength (b) Emission with attached macro-extractor. of 510 nm of the s- and p-polarized angular- Best agreement is obtained with layer thicknesses of 60 nm for Alq3 and 77 nm for Sdependent measurement and simulation of Figure 65.10. (a) Direct emission to air. TAD as given in Figure 65.4.

the spatial extend of the emission profile as well as the orientational distribution of the transition dipole moments being particularly important in polymer OLEDs [21, 61, 64, 65]. 65.4.3 Simulation-Based Optimization of OLED Layer Stacks

The real benefit of optical simulations is to study and optimize OLED stacks without the need for elaborate and time-consuming experiments. In this section, the prototypical Alq3 OLED introduced in Section 65.2 will be investigated in terms of thickness variations of the ETL and HTL. This basically corresponds to a variation of the distance between the emitter position and the interfaces to the metallic electrode and the glass substrate. Thus, the cavity will be strongly modified, which results in changes of the outcoupling efficiency into air and into the substrate. We will assume a radiative quantum efficiency of q = 1 in this example. Thus, variations of the cavity environment will only affect the light outcoupling factor 𝜂out ; the effect on q will be discussed separately in the next section. Figure 65.12 shows how the fraction of power that is dissipated into different optical modes varies as a function of the ETL thickness in the prototypical Alq3 OLED. Oscillations are clearly observable for direct emission into air and substrate emission. As the distance increases, the strength of the oscillations decreases because the intensity of the dipole radiation field weakens with increasing distance from the dipole; thus the strength of the reflected field will also be decreased [59]. It is noteworthy that the optimum for extraction to air is not in the first cavity maximum at about 75 nm ETL thickness but in the second antinode at about 220 nm (in agreement with similar results by Lin et al. [66]). However, if the contribution of both the direct and substrate emissions are considered, i.e. if substrate light extraction enhancement tools are used, then the optimum ETL thickness will be at about 90 nm. Power dissipation to waveguide modes and surface plasmons shows a quite different progression. Coupling to waveguides is very low for small ETL thicknesses,

65.4 Fundamentals of Light Outcoupling in OLEDs

1.0

Absorption Plasmon

Fraction of power

0.8

Waveguides

0.6 0.4

Emission to substrate

0.2 0.0

Emission to air

50

100

150

200

250

300

350

400

ETL thickness (nm) Figure 65.12 Amount of power coupled to different optical channels for the Alq3 OLED as depicted in Figure 65.4 in dependence of the electron transport layer (Alq3 ) thickness

assuming a radiative quantum efficiency of q = 1. The simulation was performed polychromatically according to the emission spectrum of Alq3 . (See also Ref. [44].)

but it is strongly enhanced with increasing thickness. This can be explained by the fact that there is a lower limit for the thickness of an asymmetric waveguide before a waveguide mode can exist. Above this limit an increasing number of modes are supported for larger thicknesses. By contrast, power dissipation to SPPs is strongest for a small ETL thickness, i.e. if the emitter is positioned close to the metallic cathode. This is obvious, because SPPs are evanescent waves that are only excited if the dipole couples to the metal in the near field. This contribution is reduced with increasing distance of the dipole to the metal. It should be noted that the simulation does not differentiate between coupling to surface plasmons and non-radiative energy transfer to lossy surface waves, i.e. luminescence quenching at the metal, which especially dominates for very small distances below about 25 nm and significantly reduces the efficiency [45, 59, 67]. Further optimization of the OLED layer stack is possible if both the ETL and HTL are systematically varied in thickness as shown in Figure 65.13. For the sake of simplicity, these simulations have been carried out at a single wavelength of 510 nm only (in the emission maximum of Alq3 ). It is again found that changing the ETL thickness has the strongest impact on EQE as the distance to the cathode is the main determining factor for interference effects. Moreover, reducing the HTL thickness (or in practice the ITO thickness) will lead to a further improvement of the efficiency [68]. As a result, modifying the cavity environment of the emitting molecules allows for optimizing the light outcoupling of an OLED. Nevertheless, coupling to both surface plasmons and waveguide modes causes combined losses of around 50% of the total available power – almost independent of the chosen thickness. Therefore, it is inevitable either to reduce coupling to these unfavorable loss channels or at least to partially recover some of the dissipated energy in order to develop highly efficient OLEDs. For this reason, we will later on present several promising approaches.

259

65 Device Efficiency of Organic Light-Emitting Diodes Fraction of power 0.25 0.20 0.15 0.10 0.05 0.00

HTL thickness (nm)

350 300 250 200 150 100

+

400

50 0

Fraction of power 0.50 0.40 0.30 0.20 0.10 0.00

350 HTL thickness (nm)

400

300 250 200 150 100

+

50 0

50 100 150 200 250 300 350 400

(a)

50 100 150 200 250 300 350 400

(b)

ETL thickness (nm)

Figure 65.13 Optical power coupled directly into air (a) and into the glass substrate for an applied macro-extractor (b) in dependence of the electron and hole transport layer thicknesses. The stack consists of a glass substrate, 140 nm ITO, 30 nm PEDOT:PSS, varying thicknesses of S-TAD (HTL) and Alq3 (ETL), 10 nm Ca, and 100 nm Al. The position of the prototypical OLED

ETL thickness (nm)

stack used in this work (cf. Figure 65.4) is marked by a cross in the diagram. The simulations were performed at a single wavelength of 510 nm, i.e. close to the emission maximum of Alq3 . The ETL and HTL thicknesses were varied in steps of 5 and 10 nm, respectively. A radiative quantum efficiency of q = 1 was assumed.

65.4.4 Influence of the Emitter Quantum Efficiency

In the discussion so far, we have assumed that the intrinsic radiative quantum efficiency q of the emitter material is equal to one. In real OLEDs, however, and in particular in the considered prototypical Alq3 OLED, this quantity is often significantly less than unity. According to the abovesaid, the cavity environment will then not only modify the light outcoupling factor 𝜂out but also the overall decay rate and thus lead to an effective radiative quantum efficiency qeff (see Eq. (65.8)). Figure 65.14 0.50 0.45 0.40

qeff

260

0.35 0.30 0.25 0.20 0.15

0

100

200

300

400

ETL thickness (nm) be an infinitely thin sheet of Alq3 molecules Figure 65.14 Variation of qeff with the distance of the emitting dipoles to the with q = 0.2 residing immediately at the cathode in the prototypical OLED given in Alq3 /S-TAD interface. (See also Ref. [44].) Figure 65.4. The emission layer is assumed to

65.4 Fundamentals of Light Outcoupling in OLEDs

Fraction of power

1.0 0.9 0.5

Intrinsic loss: 71.2%

0.4

Intrinsic loss

0.3 Absorption Plasmon

0.2 Waveguides

0.1

Emission to substrate Emission to air

0.0 50

(a)

100

150

200

250

300

350

Emission to air: 4.4 % Emission to substrate: 7.0 % Waveguide modes: 3.2% Surface plasmons:12.6 % Absorption: 1.6 %

400

ETL thickness (nm)

Figure 65.15 Amount of power coupled to different optical channels as in Figure 65.12 in dependence of the electron transport layer (Alq3 ) thickness assuming a radiative

(b) quantum efficiency of q = 0.2. Also shown is a pie chart of the different fractions of power for an Alq3 thickness of 60 nm. (See also Ref. [50].)

shows the effect on qeff of the distance of Alq3 emitter molecules with q = 0.2 to the cathode in the prototypical OLED given in Figure 65.4 (cf. Ref. [44]). It is found that qeff oscillates around the value of q and reaches the true value only asymptotically for very large distances. For small distances (less than about 120 nm), however, qeff strongly increases due to constructive interference but most importantly due to coupling to SPPs at the metal–organic interface. We note that for distances less than about 25 nm, the radiative quantum efficiency should strongly decrease again due to non-radiative energy transfer to the metal and subsequent quenching of the excitation [56, 59, 67], which is not distinguished for SPPs in our analysis. The consequences for the efficiency of OLEDs with emitters having q ≪ 1 are manifest. Due to the enhancement of qeff for smaller distances to the metal cathode, the first cavity maximum will gain in power relative to the second one, where the cavity effect on the radiative rate is much weaker. This is directly observable in Figure 65.15, where the dissipation of power to different optical channels is shown for the same prototypical Alq3 OLED stack as before (Figure 65.12), but now with a realistic value of q = 0.2 [50]. As expected, the biggest loss is now due to intrinsic non-radiative exciton decay; however, for both direct emission to air and emission to substrate, the optimum thickness is at the first cavity maximum below 100 nm distance to the cathode. And it is also worth noting that the fraction of power that can be coupled out of the device is significantly more than what would be expected by taking the simulation for q = 1 and simply multiplying with the lower q-value of 0.2. Thus the correct value of qeff is essential for making reliable OLED efficiency predictions according to Eq. (65.1). 65.4.5 Comprehensive Efficiency Analysis of OLEDs

As outlined in the previous section, the knowledge of the intrinsic radiative quantum efficiency is of paramount importance for OLED optimization, in particular for designing the optimum layer stack. In many cases, however, this information

261

65 Device Efficiency of Organic Light-Emitting Diodes

is not available, or the value of q is taken from photoluminescence measurements performed on films fabricated under quite different conditions than an OLED. Furthermore, in an operating OLED the excitation profile is usually different than in an optical experiment, and excitons can be quenched at interfaces or due to interactions with other excited molecules or charge carriers residing in the EML. Thus, a method for the determination of q applicable directly in an OLED environment or even in an operating OLED will give the most direct information on the value of the radiative quantum efficiency [69]. We have recently shown that the above described variation of the distance between the emitter and a metal electrode, which changes the optical environment most significantly, can be performed in real OLEDs without changing the electrical properties, provided that a doped ETL is used [43, 70]. In order to prove the feasibility of this approach, a green phosphorescent OLED based on Ir(ppy)3 as emitter was chosen (see Figure 65.16; for details we refer to Ref. [70]). Photoluminescence lifetime measurements were performed on a variety of OLED stacks with different ETL thicknesses. As shown in Figure 65.17, the phosphorescence lifetime of Ir(ppy)3 is not constant but varies with the ETL thickness. This directly reflects the influence of the optical cavity on the radiative decay rate of the emitter. With the above presented optical modeling, a quantitative analysis is now possible. Recall that the measured PL lifetime 𝜏 in the presence of a cavity is given by 𝜏 = Γ−1 = (Γ∗r + Γnr )−1 and in the absence of the cavity = (Γr + Γnr )−1 . Thus, according to Eq. (65.7) the ratio between these by 𝜏0 = Γ−1 0 two values only depends on the cavity enhancement factor F and the intrinsic radiative quantum efficiency of the emitter q but is independent of the light outcoupling factor: Γ 𝜏 = 0 = [(1 − q) + q ⋅ F]−1 (65.13) 𝜏0 Γ 60 Al-Cathode (200 nm) ETL (varied) HBL (10 nm) Matrix:Ir(ppy)3 (20 nm) EBL (10 nm) HTL (94 nm) ITO (129 nm)

34 nm 65 nm 93 nm 127 nm 164 nm 191 nm 225 nm 253 nm 284 nm

50 Intensity (a.u.)

262

40 30 20 10

Substrate

0 450

Figure 65.16 Green phosphorescent OLED stack with Ir(ppy)3 as emitter, where the ETL thickness is systematically varied. Simulated emission spectra show both spectral as well

500

550 600 650 Wavelength (nm)

700

750

as intensity changes, indicating strong effects of the variation of the optical cavity, while the electrical characteristics (not shown) are not affected. (See also Ref. [70].)

65.4 Fundamentals of Light Outcoupling in OLEDs

Relative lifetime τ/τ

0

1.1 1.0 0.9 Simulation with:

0.8

q = 1.0 q = 0.8 q = 0.6 q = 0.4 q = 0.2

0.7

q= 0.9 q= 0.7 q= 0.5 q= 0.3 q= 0.1

Measurement

0.6 0

50

100

150

200

250

300

ETL thickness (nm) Figure 65.17 Experimentally determined phosphorescence lifetimes of Ir(ppy)3 in OLED stacks with different ETL thicknesses as shown in Figure 65.16. The lines are

simulations for the given stacks with different values of the intrinsic radiative quantum efficiency q. (See also Ref. [70].)

As F is obtained independently from simulation for each of the used OLED stacks with different ETL thicknesses, the two free parameters q and 𝜏0 are easily determined by comparing measured lifetime data with simulation results. As shown in Figure 65.17, the best agreement is obtained with q(PL) = 0.5 ± 0.1 and an intrinsic phosphorescence lifetime 𝜏0 ≈ 700 ns. Remarkably, the value of the quantum efficiency is significantly less than unity. Thus, care should be taken when published IQE values for the emitter Ir(ppy)3 [71] are generalized to other OLED stacks containing this emitter. q should rather be considered as an OLED stack specific quantity, depending on the type of matrix material for the emitting dye or the materials in the immediate vicinity of the EML. Other factors could be concentration quenching and/or the effect of dye aggregation. The same series of OLEDs with systematic variation of the ETL thickness can now be used for the determination of the radiative quantum efficiency q(EL) under electrical operation, which need not necessarily be identical to q(PL) as will be seen below. For this purpose the EQE of the complete series of OLEDs was measured in an integrating sphere, once without any substrate modification and once with an index-matched hemispherical lens attached to the glass substrate to extract all of the substrate modes. Both measurements were performed for various current densities to investigate possible effects of exciton quenching with increasing currents. Again, the experimental results can nicely be reproduced by optical simulation as shown in Figure 65.18 with q being the only free parameter. As demonstrated in Ref. [70], the charge balance factor 𝛾 can thereby be assumed to be close to unity. The analysis now allows extracting the radiative quantum efficiency q(EL) at each current density as shown in Figure 65.19. It is evident that only in the limit of very small currents, q(PL) and q(EL) take roughly the same value of 0.5; however, with increasing

263

65 Device Efficiency of Organic Light-Emitting Diodes Direct emission

External quantum efficiency

0.25

0.4

0.15

0.3

0.10

0.2

0.05

0.1

0.00

0

50

100

150

200

Direct + substrate emission

0.5

0.20

250

300

0.0

q =1.0 q =0.9 q =0.8 q =0.7 q =0.6 q =0.5 q =0.4 q =0.3 q =0.2 q =0.1 Current density (mA/cm2):

0.1 0.5 1.0 2.0 10 100 500 1000

0

50

ETL thickness (nm)

100

150

200

250

300

ETL thickness (nm)

Figure 65.18 Experimentally determined external quantum efficiencies of OLED stacks with different ETL thicknesses as shown in Figure 65.16 (a, direct emission to air; b,

values obtained with a macro-extractor). The lines are simulations for the given stacks with different values of the intrinsic radiative quantum efficiency q. (See also Ref. [70].)

60 Radiative quantum efficiency (%)

264

50 40 30 20 10 0

10−1

100

101

102

103

Current density (mA/cm2) Figure 65.19 Dependence of the extracted radiative quantum efficiency from Figure 65.18 as a function of the driving current of the OLEDs. The horizontal dashed line corresponds to a q-value of 0.5 determined

from photoluminescence measurements; the vertical shaded area indicates the typical drive conditions for OLEDs in lighting applications. (See also Ref. [70].)

current density, q(EL) rapidly decreases and may only be some 20% in the range where OLEDs are typically operated for lighting applications. Thus one has to conclude that for simulation-based optimization of OLEDs, i.e. the question whether the emitter should be placed in the first or second cavity maximum, not only has the radiative quantum efficiency q in the limit of low excitation densities to be known but also its value under realistic operating conditions.

65.5 Approaches to Improved Light Outcoupling

Finally, a comprehensive efficiency analysis of OLEDs has to include the investigation of the radiation pattern, i.e. the angular- and polarization-dependent emission spectra as mentioned before, to obtain information about the emitter orientation [64, 65] and the spatial extend of the emission zone [21, 61]. We have performed this analysis for the series of Ir(ppy)3 OLEDs and found no evidence for a non-isotropic emitter orientation of Ir(ppy)3 as well as good agreement with the assumption of a sharply localized emission zone in the middle of the only 10 nm wide EML [70]. We will later on come back to the issue of non-isotropic emitter orientation when rodlike dyes (e.g. BDASBi [72]) or asymmetrically substituted metal–organic complexes (e.g. Ir(MDQ)2 (acac) [62, 73]) are used as emitters. We also want to note that a careful efficiency analysis is indispensable for the emerging class of TADF emitters, because for these materials the radiative exciton fraction (i.e. the factor 𝜂S∕T ) is no longer fixed to 0.25 (fluorescent) or 1 (phosphorescent), but can take any value in between these limits [37]. In this case, radiation pattern analysis is extremely helpful to disentangle the contributions of the TADF mechanism and other efficiency boosting factors like horizontal emitter orientation [74, 75].

65.5 Approaches to Improved Light Outcoupling 65.5.1 Overview of Different Techniques

As shown above, in planar bottom-emitting OLEDs typically around 50% of the generated light is trapped in waveguide and surface plasmon modes, and only a fraction of about 20% of it is directly radiated into air. In the previous section, we have already made use of the fact that light captured in the glass substrate can be fully extracted if an index-matched macroscopic lens is used and if the active pixel area is not too large. However, if the unique form factor (thin and flat) is to be preserved, this method is not practicable in large area devices. Thus, the development of new concepts for improving light extraction efficiency has been a major issue over recent years (for reviews see, e.g. Refs. [14, 51, 76, 77]). The different approaches can roughly be distinguished into techniques staying with planar structures and others utilizing scattering methods [78, 79]. Among the former ones are approaches toward further optimization of the OLED cavity, i.e. modifications of the layer stack with respect to thicknesses, refractive indices, or reflectivity. A relatively simple method is to place the emission zone in the second antinode of the interference pattern relative to the cathode [66] and thus reduce the coupling to SPPs. As discussed before, however, this will only be beneficial if the radiative quantum efficiency of the emitter is very high; otherwise the emitter will experience a stronger enhancement of its radiative rate when placed in the first cavity maximum. Alternatively, the excitation of SPPs can be completely avoided if metalfree OLEDs are used [80], but there is usually no overall gain in light outcoupling

265

65 Device Efficiency of Organic Light-Emitting Diodes

since the reduction of SPP losses is mostly at the expense of enhanced waveguiding [50]. Another way of boosting the direct emission to air is the use of microcavity structures, i.e. OLEDs with two (partially) reflecting electrodes, where the radiation is directed more toward small angles with respect to the surface normal. This can be achieved by (aperiodic) dielectric Bragg reflectors placed underneath the ITO electrode [81] or in top-emitting devices with a highly reflective anode on glass and a semitransparent metallic cathode, often followed by a dielectric capping layer as antireflection coating [82–84]. Though very high EQE values close to 30% have recently been published for a red phosphorescent top-emitting OLED [85], these structures usually have a non-Lambertian emission characteristics with non-negligible spectral shifts as a function of the viewing angle. We also want to note that in the case of the particular red phosphorescent dye Ir(MDQ)2 (acac) used in Ref. [85], non-isotropic emitter orientation might at least partially contribute to the high EQE values as will be discussed at the end of this chapter [62, 73]. A third method where the planar layer structure is preserved lies in matching the refractive indices of the EML, where the light is generated, and the outside world, either by bringing the refractive index of the EML down to one or by using a highindex (HI) substrate. Simulations show that if the refractive index of the EML could be reduced to n = 1, a theoretical limit of almost 70% direct emission to air would be possible [86]. An experimental realization might be difficult simply because no such materials are available, but lowering nEML below the value of glass already would bring an enormous boost in EQE. The use of glass substrates with higher refractive index than the organic layers (including ITO) is possible [87, 88]. Simulations for the prototypical Alq3 OLED demonstrate (see Figure 65.20) that in this way all the light from waveguide modes 1.0 Absorption

0.8 Fraction of power

266

Plasmon

0.6 Emission to substrate

0.4 0.2 0.0

Emission to air

50

100

150 200 250 300 ETL thickness (nm)

Figure 65.20 Amount of power coupled to different optical channels in dependence of the electron transport layer thickness for the Alq3 OLED as depicted in Figure 65.4, but now with a high-index substrate (SF6

350

400

glass with n = 1.82). The simulation was performed polychromatically according to the emission spectrum of Alq3 , and a radiative quantum efficiency of q = 1 was assumed.

65.5 Approaches to Improved Light Outcoupling

Cathode Organic Glass/ITO

(a)

Microlens array

(b) Scattering particles

Figure 65.21 (a) Schematic illustration of substrate mode extraction by a microlens array. Light paths indicated in red correspond to radiation that would normally not be able to leave the device due to total internal reflection at the glass/air interface. Multiple reflections inside the stack reduce the efficiency due to absorption in

(c)

Grating

the organic layers and in ITO as well as due to reflection losses at the cathode. (b) Concept of light outcoupling by scattering particles in a film applied to the backside of the substrate. (c) Realization of an internal scattering structure by a periodic grating placed between the glass substrate and the thin film stack.

goes into the HI glass substrate from where it can be extracted more easily. Combining HI substrates with a macro-extractor (an index-matched lens), record EQEs in excess of 40% have been achieved in this way [24, 89]. Nevertheless, one has to be aware that HI glass substrates would increase the overall cost of OLEDs considerably. Thus for practicable devices thin film solutions have to be applied. A fundamentally different approach to extract trapped light in OLEDs is the use of scattering structures. Here one has to distinguish between periodic and nonperiodic (random) structures on the one hand and internal or external scattering on the other hand (see Figure 65.21). The latter realization is quite straightforward and can be any modification of the backside of the glass substrate that serves to scatter out light rays that would otherwise suffer from total internal reflection at the glass/air interface. Examples are ordered microlens arrays [90, 91], scattering particles [92], or mechanical roughening of the glass substrate [93]. In the meantime light scattering foils are commercially available, but in contrast to a macro-extractor these techniques typically extract only part of the light trapped in the substrate and the reflecting appearance of the OLED in the off-state changes to a milky, nonreflecting one. If one wants to get access to waveguide modes, an internal scattering structure, e.g. placed between the glass substrate and the ITO layer, has to be used. Therefore both periodic gratings (photonic crystals) and random scattering structures are possible. The effectiveness of this approach relies on the spatial overlap of the waveguide modes with such features; in other words, they have to be employed close to the emission zone of the OLED [94]. In general, photonic crystal structures with periodicities on a length scale comparable with optical wavelengths have the disadvantages that they require elaborate fabrication techniques and that they induce a wavelength- and angular-dependent scattering efficiency. Other interesting approaches are therefore the use of so-called low-index grids with periods on the micron scale [95] or random structures fabricated by morphological instabilities of thin film structures [96]. The abovesaid is also true for the extraction of surface plasmons by scattering structures, which consequently need to be in the vicinity of the metal cathode [97]. This boundary condition is even more challenging since many of the established patterning techniques are not compatible with OLED technology as they would damage or destroy the underlying organic films.

267

65 Device Efficiency of Organic Light-Emitting Diodes

65.5.2 Reduction of Surface Plasmon Losses 65.5.2.1 Basic Properties of SPPs

Before presenting our own results on the reduction of surface plasmon losses, some general properties of SPPs will be discussed (for details we refer to Refs. [98–100]). SPPs are longitudinal p-polarized waves traveling at the interface between a metal and a dielectric with evanescent fields decaying exponentially into both adjacent media. For semi-infinite layers their dispersion relation is given by √ √ 𝜖1 ⋅ 𝜖2 𝜖1 (𝜔) ⋅ 𝜖2 (𝜔) 𝜔 = (65.14) kSPP (𝜔) = k0 𝜖1 + 𝜖2 c 𝜖1 (𝜔) + 𝜖2 (𝜔) with 𝜖1,2 (𝜔) being the complex dielectric functions of the metal and the adjacent dielectric layer, respectively, and k0 the vacuum wave vector. This relation is shown in Figure 65.22 for three different metal surfaces adjacent to air as dielectric for the range of frequencies and in-plane wave vectors relevant in this context. It is obvious that the SPP dispersion curve and the light line in air do not intersect for finite frequencies; thus energy and momentum conservation cannot be fulfilled simultaneously, and as a consequence SPPs cannot couple to far-field radiation. The application to OLEDs, however, requires an important modification due to the fact that the involved layers are not thick enough to be treated as bulk material. Thus, the evanescent field of the SPP can extend through an adjacent thin organic layer and sense an effective refractive index, and, if the metal is thin enough, SPPs can exist on both sides of the metal layer and couple with each other [101]. 65.5.2.2 Scattering Approaches

There are well-established methods how SPPs can be excited by far-field radiation, most importantly by grating coupling or prism coupling [98, 99]. Making use 5 ×1015

400

4 ×1015

533

3 ×1015 Silver Aluminum Gold

2 ×1015 1 ×1015 0 0

1×107

800 1600

Wavelength (nm)

Angular frequency (rad/s)

268

2 ×107

In-plane wave vector (m−1) Figure 65.22 Dispersion relation of surface plasmon polaritons at metal/air interfaces for three different metals in the visible spectral range. The dashed line is the light line in air following the relation 𝜔 = c ⋅ k|| .

65.5 Approaches to Improved Light Outcoupling Angular frequency ω

Angular frequency ω Air light cone Glass light cone

Air light cone Glass light cone

Surface plasmon

Surface plasmon

−4kg−3kg −2kg −kg

In-plane wave vector kx Metal film

Surface plasmons ++ − −+ + − −

kg 2kg 3kg 4kg

λg Metal film

Organic emitter

Organic emitter

Glass substrate

Glass substrate

In-plane wave vector k||

+ + −− + −− +

Emission from surface plasmons

(a)

Direct emission

Figure 65.23 Simplified layer structure to demonstrate the principle of SPP grating coupling. The SPP dispersion relation, lying below the glass light-cone in planar layered structures (a), is scattered by multiples of

(b)

Direct emission

the grating wave vector to result in several branches being located within the glass light cone (b), so that light from SPPs can be coupled out. (See also Ref. [102].)

of reciprocity in optics, it is quite straightforward to transfer these techniques to light-emitting structures [97, 103]. The first approach uses a periodic grating with period 𝜆g to scatter SPPs so that they gain an extra momentum ′ = kSPP ± m ⋅ (2𝜋∕𝜆g ) (with an integer number m, see Figure 65.23). This kSPP scattering approach relies on the fact that there is sufficient overlap between the SPP mode and a periodic modulation of the refractive index, which implies that the grating has to be placed next to the metal layer (or there has to be a height modulation reaching through the complete layer stack including the cathode). The fabrication of OLEDs fulfilling this condition is still challenging [104], as not only the optical properties have to be tuned but also the electrical functioning has to be ensured. For that reason we have investigated this coupling approach only in simplified structures comprising a luminescent film adjacent to a metal layer [102, 105]. Figure 65.24 shows experimentally measured angular-dependent p-polarized emission spectra for photoluminescence excitation of a 30 nm thick Alq3 film adjacent to a silver layer, both deposited on a line grating that was nanoimprinted into a PMMA layer. One can clearly see that SPPs, which are excited by fluorescent Alq3 molecules via near-field coupling, are scattered back into the substrate up to the fifth diffraction order. There is excellent agreement between the measured angular dispersion and the simulated curves using the above described methods. One can also readily show that waveguide modes that will become relevant for thicker organic layers are equally well scattered out in this way. Furthermore, we have demonstrated that even structures containing some randomness, produced by an ordinary DVD stamp, can be used [105]. Nevertheless, one has to be aware that periodic structures will induce a pronounced angular dependence of the perceived

269

65 Device Efficiency of Organic Light-Emitting Diodes

750

m = −2

m = −4

650

m=3

600

−60°

(a)

+30° 0°

m=4

500 450 −90°

Laser excitation

833 nm period

m = −5

550

Glass/PMMA Alq3 30 nm Silver 150 nm

Intensity (a.u.) 1.0 0.8 0.6 0.4 0.2 0.0

m=2

m = −3 m = 1

700

Wavelength (nm)

270

−30°

0°

−90° Rotation stage

30°

Angle

(b)

Figure 65.24 (a) Experimentally measured angular-dependent p-polarized emission spectra after photoluminescence excitation of a 30 nm thick Alq3 film adjacent to a silver layer, both deposited on an 833 nm periodic line grating that was nanoimprinted into a PMMA layer. The dotted lines show simulated SPP dispersions of the corresponding planar structure, where the curves were shifted by integer multiples of the grating

wave vector. (b) Schematic sample layout used in the experiment; both excitation and detection are through the glass prism. Excitation is made by a 375 nm laser diode incident at a fixed angle of 45∘ ; detection is performed by a fiber optical spectrometer through a linear polarizer in front of which the sample is rotated on a motorized turn table. (See also Ref. [105].)

color of the OLED [106]. Thus, for white OLEDs with color coordinates independent of the viewing angle, random scattering structures have to be used [96]. 65.5.2.3 Index Coupling

Figure 65.25 shows the principle of index (or prism) coupling utilizing the so-called inverse Kretschmann configuration [107]. This technique makes use of the fact that the in-plane wave vector within the prism is stretched by a factor given by the refractive index of the prism, so that energy and momentum conservation are fulfilled. If the refractive indices are properly chosen, it is thus possible to extract SPPs that are evanescent modes at the metal–organic interface (the bottom side Laser excitation

Angular frequency

Glass Silver Alq3

Incident light k0

Air light cone Glass light cone

90°

Emission from surface plasmons

θ npk0 klight − − + + − −+ + kSPP (a)

Surface plasmon

Prism 0°

Metal Air (b)

(c)

In-plane wave vector

Rotation stage

Figure 65.25 Principle of SPP index coupling. (a) shows the excitation of SPPs by light from the far field in the so-called Kretschmann configuration. (b) Inverse Kretschmann configuration to extract SPPs

that are excited by near-field coupling from a luminescent organic film. (c) The dispersion relation of the SPP traveling at the interface to air overlaps with the light cone for the attached glass prism. (See also Ref. [102].)

65.5 Approaches to Improved Light Outcoupling

Air Alq3 --++ --++ --++ Ag

Air Medium with refractive index nSPP

0–195 nm 5 nm

- - + + - - + + - - + + SPP Ag

Alq3 + --++ --++ --++ nm + - - +Ag + -40 -++ --++ -

Emitter position SPPs

Glass (c)

1.6 1.4

1.0 (b)

nSPP

1.2

0

Alq3

1.8

Alq3

Effective SPP index

Effective SPP index

1.8

50 100 150 Alq3 thickness (nm)

1.6

(d)

Glass

1.4

nSPP1 nSPP2

1.2 1.0

200

Alq3 thickness (nm)

(a)

500 Intensity 450 p-pol 1.0 400 0.8 TM1 350 0.6 300 0.4 250 0.2 200 TM0 0.0 150 100 SPP 50 0 0° 15° 30° 45° 60° 75° 90° Angle (e) Alq3 thickness (nm)

5 nm 0–195 nm

in Figure 65.25) to become radiative modes at the opposite side of the metal layer (here the top side). To realize this concept, however, some prerequisites have to be fulfilled. First, the metal film obviously has to be semitransparent. Since there are two counteracting processes, namely, an enhanced coupling of the emitting dipoles to SPPs with an increasing metal thickness on the one hand and a concomitant reduced optical transmission on the other hand, there is typically an optimum thickness at around 50 nm [102, 108]. More importantly, the refractive index on the extraction side of the metal has to be larger than on the other side in order to match the dispersion relation of the plasmon (traveling at the bottom side) with the far-field light line in the prism at the top side. Due to the evanescent nature of SPPs, this condition depends not only on the bulk values of the organic material and the prism but also in particular on the thickness of the organic layer. Since the decay length of the SPP field amplitude perpendicular to the metal/dielectric interface is typically of the order of half a wavelength, the SPP field probes an extended vertical distance to the metal surface. In order to account for this feature, we have introduced an effective refractive index that has to be entered in Eq. (65.14) to calculate the correct SPP dispersion relation. This concept of the effective index of SPPs and its dependence on the thickness of the organic layer for a one-sided SPP (thick metal) and a two-sided SPP configuration (thin metal) is shown in Figure 65.26. Without going into the details (see, e.g. Ref. [109]), it is clear from the figure that only for thin Alq3 layers (up to about 50 nm), ordinary glass can be used to extract SPPs. For larger organic layer thickness, and in particular for the thicknesses typically used in OLEDs, the extraction of SPPs (in addition to waveguide modes that can also be extracted by this method) requires a medium on the top side of the metal that has a significantly higher refractive index.

0

50 100 150 Alq3 thickness (nm)

Figure 65.26 Concept of the effective refractive index of the adjacent dielectric medium that is probed by an SPP for a thick metal layer (a, b) and for a thin metal layer (c, d). Also shown is the simulated

200 (f)

500 Intensity 450 s-pol TE2 1.0 400 0.8 350 0.6 300 0.4 TE 1 250 0.2 200 0.0 150 100 TE0 50 0 0° 15° 30° 45° 60° 75° 90° Angle

angular dispersion of the modes that can be extracted through a fused silica prism as a function of the organic layer thickness: (e) p-polarized and (f ) s-polarized. (See also Refs. [102, 109].)

271

65 Device Efficiency of Organic Light-Emitting Diodes

(a)

500 450 400 350 300 250 200 150 100 50 0

Intensity 1.0 0.8 0.6 0.4 0.2 0.0

TM1

TM0

n = 1.82 p-pol

SPP 0°

15°

30°

45°

60°

75°

Angle

Alq3 thickness (nm)

Alq3 thickness (nm)

272

90°

(b)

500 450 400 350 300 250 200 150 100 50 0

Intensity 1.0 0.8 0.6 0.4 0.2 0.0

TM1

TM0

n = 2.28 p-pol

SPP 0°

15°

30°

45° 60° Angle

75°

90°

Figure 65.27 Simulated angular dispersion of the p-polarized modes extracted through a high-index SF6 glass prism (a) and a LiNbO3 prism (b) as a function of the organic layer thickness. (See also Ref. [109].)

Figure 65.27 shows exemplarily the calculated angular dispersion of p-polarized modes that can be extracted if an HI glass (SF6, n = 1.82) or a LiNbO3 (n = 2.28) prism is used. In the former case SPP modes can now be extracted up to an organic layer thickness of about 100 nm, while in the latter case the index is so high that there is no limit any more and all modes, regardless of the thickness of the organic film, can be extracted under an angle of not more than 60∘ . The application of this concept to a white bottom-emitting OLED with a thin semitransparent Ag top contact is presented in Figure 65.28. The photograph clearly shows that only the air mode is emitted to the bottom side, whereas all other modes (including substrate, waveguide, and SPP modes) are emitted in characteristic angular ranges to the top side. We note that the index of the used SF6 prism was not high enough to extract the SPP completely, so only the red part of the SPP branch is seen under large angles close to 90∘ . Furthermore, the used index matching fluid was strongly absorbing in the short wavelength range so that the substrate light has a brownish appearance. As the simulations in (c, d) show, the extraction of a major fraction of the energy contained in SPPs requires a medium with nHI > 2.0. Nevertheless, this is a clear demonstration that the concept of HI coupling is applicable not only to extract waveguide modes in bottom-emitting OLEDs, as was demonstrated before by different authors [24, 88], but also to get access to all trapped modes – including surface plasmons [109]. It should be noted, however, that these results can only serve as a proof of principle, because using an HI prism is not practicable in OLED applications. It would therefore be necessary to develop thin film solutions to get rid of the viewing angledependent color shift. In that context it should be noted that top-emitting OLEDs with HI capping layers have been known for many years [82, 83]; however, the connection to SPPs was not made. Only recently has it been shown that including a fluorescent dye in such a capping layer can allow to extract energy from SPPs in top-emitting OLEDs [111]. 65.5.2.4 Emitter Orientation

Instead of developing tools to extract energy from surface plasmons, one can also consider means to reduce their excitation in OLEDs. Apart from the already

65.5 Approaches to Improved Light Outcoupling

Substrate emission Prism

WGM

Substrate emission Waveguided modes

HI prism

SPP

Ag (30nm)

Plasmons

OLED ITO/glass substrate

Direct emission

Direct emission 50 45 40 35 30 25 20 15 10 5 0

DT ST WGM SPP

WGM Increasing nHI SPP nHI = 2.5

1.0

15°

0.8

SPP lost

0.6

SPP extracted WGM lost WGM extracted

0.4 Direct + subs. lost

Substrate

0.2 Direct

nHI =1.5

0°

(c)

OLED

(b)

Fraction of power

Intensity (a.u.)

(a)

30°

45°

60°

75°

0.0 1.5

90°

Angle

Figure 65.28 Sample layout (a) and photograph (b) of a white OLED with direct emission toward the bottom and extraction of all other modes to the top side using an SF6 prism with n = 1.82. The simulations in (c) show the angular dispersion of the modes extracted through the high-index glass prism and in (d) the angle-integrated

(d)

1.7

1.9

2.1

2.3

2.5

Refractive index nHI

contribution of the different modes as a function of the refractive index of the prism. Note that these simulations were performed at a single wavelength of 600 nm and a semi-infinite glass substrate was assumed to avoid back-reflected light from the glass/air interface appearing as top emission. (See also Refs. [109, 110].)

mentioned distance-dependent coupling between a radiating dipole and SPPs, there is also a characteristic dependence on the orientation of the dipole [55, 59]. Thus another way to avoid the excitation of surface plasmons, even if the emitter is rather close to the metal, is to control the orientation of the emitting molecules and thus of their transition dipole moments. Keeping the radiation pattern of a classical electrical dipole in mind (cf. Figure 65.7) and considering that surface plasmons are transverse magnetic modes, one readily concludes that perfectly horizontally oriented dipoles would only very weakly couple to SPPs [49]. This effect has been known for many years in polymeric OLEDs [112]; only very recently, however, we have been able to show that orientation effects also play a role in small molecule OLEDs fabricated by vacuum evaporation, where the fluorescent or phosphorescent dyes are embedded with only a few percent content in a matrix material [62, 72, 73].

273

Wavelength (nm)

65 Device Efficiency of Organic Light-Emitting Diodes 800 750 700 650 600 550 500 450 400

px-dipole

1

2 7

3 7

py-dipole

4

1

7

2 7

3

1

1 × 10 2 × 10 3 × 10 4 × 10

In-plane wave vector (m )

In-plane wave vector (m )

−1

(a)

7

pz-dipole

4

7

1 × 10 2 × 10 3 × 10 4 × 10

7

(b)

2

3

7

Intensity 1.0 0.8 0.6 0.4 0.2 0.0

4

1 × 10 2 × 10 3 × 107 4 × 107

7

−1

7

−1

In-plane wave vector (m )

(c)

Figure 65.29 Simulation of power dissipation for the prototypical Alq3 OLED (structure shown in Figure 65.4) for the three orthogonal dipole orientations separately (x, y in the substrate plane, z perpendicular to the

substrate plane). In all cases the radiative quantum efficiency was set to q = 1. The dashed lines separate the power dissipation spectra into different regions as denoted in Figure 65.8. (See also Ref. [50].)

To demonstrate the potential of controlling the emitter orientation for efficiency enhancement in OLEDs, Figure 65.29 shows simulated power dissipation spectra separately for each of the three orthogonal dipole orientations in the prototypical Alq3 OLED (structure shown in Figure 65.4). From this plot one can clearly see that horizontal dipoles (px and py ) couple to various optical channels, whereas vertical (pz ) dipoles dissipate their energy almost exclusively to SPPs, which makes it difficult to detect them in OLEDs, e.g. from angular-dependent emission spectra of OLEDs. As a workaround we have therefore developed a method that is based on photoluminescence excitation of layer stacks that contain the same EML as the corresponding OLED but no metal layer [113]. It is thus possible to quantitatively determine the orientational distribution from angular- and polarization-dependent photoluminescence spectra. It is furthermore instructive to look at the power dissipation of both dipole orientations into different optical channels as a function of the distance to the cathode, as displayed in Figure 65.30 (cf. Figure 65.12 for the isotropic case). It is obvious Horizontal orientation

1.0

Emission to substrate

0.4 0.2 0.0

(a)

Waveguides

0.6

Fraction of power

0.8

Emission to air

50

100

150

200

250

Figure 65.30 Simulation of power dissipation for exclusively horizontal (a) or vertical dipole orientation (b) in the prototypical Alq3 OLED (structure shown in Figure 65.4) as a

(b)

Absorption

0.8 Plasmon

0.6

Waveguides

0.4 Emission to substrate Emission to air

0.2 0.0

300

ETL thickness (nm)

Vertical orientation

1.0

Absorption Plasmon

Fraction of power

274

50

100

150

200

250

300

ETL thickness (nm)

function of the ETL thickness. In both cases, the radiative quantum efficiency was set to q = 1. (Compare with Figure 65.12 for the isotropic case.)

65.5 Approaches to Improved Light Outcoupling

from these simulations that the light outcoupling efficiency (be it only direct emission or including substrate emission) can be enhanced by about a factor of 1.5 with respect to the case of random emitter orientation, which is mostly due to the strongly reduced coupling to SPPs, if vertical dipoles are absent. Thus, from the point of view of efficiency, the vertical dipole orientation should be avoided in OLEDs. It is also interesting to note in this context that vertical dipoles exhibit markedly different distance-dependent couplings to the various optical channels as compared with horizontal ones, which is due to different interference conditions. First, the coupling of vertical dipoles to SPPs has a longer range than in the horizontal case (as predicted already in Ref. [114] for energy transfer of dyes to a metal surface). Second, for vertical dipoles the maximum outcoupling efficiency to air is achieved for an ETL thickness of about 135 nm, exactly where the outcoupling from horizontal ones has an interference node. This property can be used for the determination of the amount of vertical dipoles in an OLED with non-isotropic emitter orientation by intentionally fabricating a non-optimized layer stack [65]. To realize this concept one needs emitter molecules that show non-isotropic orientation in a small molecule OLED environment, where the emission layer is usually a thin dye-doped layer prepared by co-evaporation on top of other – usually amorphous – layers. Good candidates are rodlike chromophores with a large shape anisotropy, such as the blue fluorescent dye BDASBi shown in Figure 65.3. This material has already been known to exhibit non-isotropic orientation in neat evaporated films [115], but interestingly it preserves this feature even when it is doped into a CBP matrix [113]. As described in detail in Ref. [72], we have compared OLEDs incorporating two blue emitting fluorescent dyes (BDASBi and PEBA; see Figure 65.3) exhibiting different emitter orientations with respect to their external quantum efficiency. In the actual OLEDs investigated by us, the emitter orientation is not perfectly horizontal, and the emitter quantum efficiency is less than unity. Nevertheless, the experimentally observed increase in EQE with respect to the isotropic reference OLED can consistently be explained by the preferentially horizontal emitter orientation in the case of the rodlike BDASBi molecule as emitting dye. Meanwhile, other groups have also found evidence for a possible contribution of non-isotropic emitter orientation in fluorescent OLEDs with efficiencies beyond the spin-statistical limit [34, 35]. We have extended these studies toward highly efficient phosphorescent emitters. As expected, symmetrically substituted – i.e. homoleptic – metal–organic emitter complexes such as Ir(ppy)3 have random emitter orientation when doped in a CBP matrix (see Figure 65.31a). However, in the case of compounds with different ligands – so-called heteroleptic complexes, there is evidence for non-isotropic emitter orientation. For example, the well-known red phosphorescent emitter system Ir(MDQ)2 (acac) doped into an α-NPD matrix also exhibits predominantly horizontal orientation of the emitting dipoles (see Figure 65.31b and Ref. [73]). If this circumstance is ignored, efficiency analysis based on the assumption of isotropic emitter orientation will lead to an overestimation of the radiative quantum efficiency [116, 117], as has been worked out in detail in Ref. [62]. Figure 65.32 shows the OLED stack, on which we have performed a comprehensive efficiency analysis combining

275

65 Device Efficiency of Organic Light-Emitting Diodes

Ir(ppy)3: CBP (8%)

0.8 0.6 0.4 0.2

Ir(MDQ)2(acac): α-NPD (8%)

1.0 Intensity (a.u.)

Intensity (a.u.)

1.0

Measurement Simulation px : pz =1:1 px : pz =1: 0

0.8 0.6 0.4 0.2

Measurement Simulation px :pz = 1:1 px :pz = 1:0 px :pz = 1:0.6

0.0 0.0 0° 10° 20° 30° 40° 50° 60° 70° 80° 90° 0° 10° 20° 30° 40° 50° 60° 70° 80° 90° (a) Angle (b) Angle

Figure 65.31 Angular-dependent ppolarized photoluminescence intensity of simplified layer stacks on glass substrates with two different phosphorescent dyes: (a) Ir(ppy)3 embedded in a CBP matrix (wavelength 510 nm) and (b) Ir(MDQ)2 (acac) in an α-NPD matrix (wavelength 610 nm). (For details of the experiment, see Ref. [113]). The

Spectral density (a.u.)

10 nm 10 nm 10 nm 220 nm 111 nm

105

ETL HBL EML EBL HTL ITO

Intensity (a.u.)

Varied

comparison to simulations for exclusively horizontal and fully random emitter orientation reveals isotropic emitter orientation in case (a), while case (b) shows a predominantly horizontal orientation with a ratio of 2 : (0.6 ± 0.1) parallel (px + py ) vs. perpendicular (pz ) emitters, i.e. Θ = 0.23, in good agreement with Ref. [73].

Ag

200 nm

τ = 1.36 μs

104

103 PL intensity Monoexp. fit

Substrate 102 0.0

(a)

2.5

5.0

(b)

7.5

550

10.0

(c)

Time (μs)

600

650

700

750

Wavelength (nm)

0.50

1.2

0.45 1.1

0.40 0.35

1.0

0.30 EQE

Relative lifetime τ/τ0

276

0.9 0.8 0.7

0.25 0.20

Isotropic, q =0.8 Mainly parallel, q=0.7 PL measurement τ =1.37 μs

0.15 0.10 0.05

0.6

(d)

50

100

150

200

250

300

350

0.00

400

ETL thickness (mn)

Figure 65.32 (a) Structure of the red Phosphorescent OLED stack containing Ir(MDQ)2 (acac) in an α-NPD matrix as EML, (b) an example of a phosphorescence lifetime measurement for an ETL thickness of 249 nm, and (c) the electroluminescence spectrum used for simulation. (d) Measured phosphorescence lifetimes normalized to the intrinsic value 𝜏0 = 1.37 μs together with simulated curves for different emitter

(e)

50

100

150

200

250

300

350

400

ETL thickness (mn)

orientations: isotropic with q = 0.8 and mainly horizontally oriented with an orientation parameter Θ = 0.63∕2.63 = 0.24 and q = 0.7. (e) Measured EQE values without and with a macro-extractor at a current density of j = 1 mA∕cm2 are compared to optical simulations for both situations, clearly confirming the non-isotropic orientation of this emitter system. (For details see Ref. [62].)

65.5 Approaches to Improved Light Outcoupling

Quantum efficiency (q)

1.0

ηEQE 0.00

0.9

0.05 0.10 0.15

0.8

0.20 0.25

0.7

0.30 0.35

0.6 0.5 0.0

0.40

0.1

0.2

0.3

0.4

0.5

Orientation parameter (Θ) Figure 65.33 Relation between the orientation parameter Θ and the external quantum efficiency (𝜂EQE ) in a phosphorescent OLED. For perfectly horizontal emitter orientation (Θ = 0) and q = 1, the EQE can reach up to almost 40%. (For details see Ref. [63].)

phosphorescence lifetime and EQE measurements as function of the ETL thickness, as described before in Section 65.4.5 for an Ir(ppy)3 -based green OLED. Both data sets show that a consistent description can only be obtained, if non-isotropic emitter orientation is taken into account. Moreover, optical simulations clearly reveal that horizontally oriented phosphorescent emitters with high radiative quantum efficiencies (q → 1) have the potential to achieve EQE values close to 40% for direct emission to air [119], as shown in Figure 65.33. Meanwhile, several groups have demonstrated that preferably horizontal orientation of the transition dipole moments can be achieved in OLEDs including fluorescent, phosphorescent, and even TADF emitters (for a review see, e.g. Ref. [63]). In some cases, even complete horizontal emitter orientation could be achieved, and EQE values approaching 40% were consistently reported [120, 121]. This development was enabled by a better understanding of the driving forces for molecular orientation in small molecule guest–host systems [63]. In spite of the fact that OLEDs commonly comprise noncrystalline layers, it has become clear that orientational anisotropies play an important role in these devices. Among the major factors controlling emitter orientation are the shape of the dye molecules as well as the substrate temperature at which the film is grown by thermal evaporation or similar vapor deposition techniques [34, 122]. As an example, Figure 65.34 shows the case of a heteroleptic Ir complex, where the different nature of the ligands drives alignment of the C2 symmetry axis of the complex perpendicular to the surface plane, while the film is growing by molecular beam deposition in vacuum.

277

278

65 Device Efficiency of Organic Light-Emitting Diodes

O

N

O Ir O

O 2

Film

Substrate

Figure 65.34 Schematic orientation mechanism of a heteroleptic Ir complex on the surface of a vapor-deposited film. The less aromatic ligand (green) points to the surface of the growing film, i.e. to vacuum, while the aromatic ligands are directed to the bottom.

In this way, the optical transition dipole moments residing on the aromatic ligands are oriented preferentially in the film plane, which leads to improved light outcoupling. (For further details see Ref. [118].)

65.6 Summary and Outlook

In conclusion, we have shown that optical modelling is an indispensable tool not only for optimization of OLED light outcoupling but also for a comprehensive efficiency analysis. In the latter case, simulations have to be combined with systematic parameter variation of the OLED layer stack, e.g. the distance of the emission zone to the highly reflecting cathode. In this way it is possible to separate microcavity effects on the radiative rate, determining the effective radiative quantum efficiency qeff , from a redistribution of energy between different optical channels, affecting the light outcoupling term 𝜂out . We find a significant reduction of qeff under typical OLED operating conditions, which has to be considered in the stack design. Furthermore, we have presented approaches for enhancing 𝜂out , with special emphasis on the reduction of the surface plasmon loss channel. Beside scattering approaches and HI coupling, the control of the emitter orientation has been identified as a particularly powerful handle to obtain this goal. Remarkably, even

Acknowledgement

dye-doped small molecule phosphorescent emitter systems prepared by vacuum deposition can exhibit pronounced anisotropic orientation distributions of their optical transition dipole moments. Thus, after more than 20 years of research on OLEDs, the microscopic understanding of the underlying processes has progressed substantially. With respect to the expression for the external quantum efficiency (Eq. (65.1)) put forward by Tsutsui in 1997 [28], we now have a sound understanding of all four factors allowing for quantitative calculations of OLED efficiencies and predictions toward the potential in improvement of it by newly developed OLED materials, structures, and concepts. Moreover, these methods are the basis for a reliable experimental determination of the relevant factors for the OLED efficiency. Nevertheless, the development is ongoing, and important new concepts, such as the use of TADF [123, 124], to name just one hot topic, are still being elucidated and integrated into OLEDs. By now, white OLEDs are specified with luminous efficacies exceeding that of fluorescent tubes. However, their large-scale commercialization in the field of general lighting has turned out to be more challenging than initially expected, because efficiency and lifetime – in particular for the blue component – lag behind inorganic LEDs. But first and foremost, they are still too expensive for a cost-driven market [8]. Nevertheless, a strong benefit of OLEDs is their unique form factor, with the light being distributed homogeneously over large area and thus being glare-free. Hence, (almost) no additional fixtures are needed: the OLED already is the luminaire! Lately, this feature has been exploited, and a new opportunity for OLEDs has appeared, e.g. as tail lights in the automotive sector. Beyond general lighting, the compatibility of organic materials with various kinds of substrates and the ease of processing could open possibilities for new appliances in other fields as well. In particular the topics relevant for light extraction could have fruitful overlap with other fields of optics, e.g. cavity physics, photonic crystals, and plasmonics – and we are still waiting to see the electrically pumped organic laser diode.

Acknowledgement

The author thanks the German Ministry of Education and Research (BMBF) and the German Research Foundation (DFG) for funding over the last few years. The contributions of several PhD students and postdocs (J. Frischeisen, T. Lampe, C. Mayr, S. Nowy, N. Reinke, T.D. Schmidt, B. Scholz, T. Wehlus, and S. Wehrmeister) to this research are gratefully acknowledged. I also thank colleagues from Osram OLED GmbH (Regensburg, Germany) and Fraunhofer Institute for Applied Optics and Precision Engineering (Jena, Germany) as well as the groups of C. Adachi (Kyushu University, Japan), J.J. Kim (Seoul National University, Korea), and M.E. Thompson (University of Southern California, Los Angeles, USA) for their contributions to parts of this work.

279

280

65 Device Efficiency of Organic Light-Emitting Diodes

References 1. Humphreys, C.J. (2008). Solid-state 2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

13.

lighting. MRS Bull. 33: 459–470. Nobelprize.org. (2014). The Nobel Prize in Physics 2014 was awarded jointly to Isamu Akasaki, Hiroshi Amano and Shuji Nakamura for the invention of efficient blue light-emitting diodes which has enabled bright and energysaving white light sources. https://www .nobelprize.org/prizes/physics/ (accessed 15 May 2019). Pust, P., Schmidt, P.J., and Schnick, W. (2015). A revolution in lighting. Nat. Mater. 14: 454–458. Pope, M., Kallmann, H.P., and Magnante, P. (1963). Electroluminescence in organic crystals. J. Chem. Phys. 38 (8): 2042–2043. Helfrich, W. and Schneider, W.G. (1965). Recombination radiation in anthracene crystals. Phys. Rev. Lett. 14 (7): 229–231. Tang, C.W. and VanSlyke, S.A. (1987). Organic electroluminescent diodes. Appl. Phys. Lett. 51 (12): 913–915. Burroughes, J.H., Bradley, D.D.C., Brown, A.R. et al. (1990). Lightemitting diodes based on conjugated polymers. Nature 347: 539. US Department of Energy (2016). Solid-state lighting 2016 R&D plan. https://energy.gov/eere/ssl/downloads/ solid-state-lighting-2016-rd-plan (accessed 15 May 2019). D’Andrade, B.W. and Forrest, S.R. (2004). White organic light-emitting devices for solid-state lighting. Adv. Mater. 16 (18): 1585–1595. Schwartz, G., Reineke, S., Rosenow, T.C. et al. (2009). Triplet harvesting in hybrid white organic light-emitting diodes. Adv. Funct. Mater. 19 (9): 1319–1333. Gather, M.C., Köhnen, A., and Meerholz, K. (2011). White organic light-emitting diodes. Adv. Mater. 23 (2): 233–248. Tyan, Y.-S. (2011). Organic lightemitting-diode lighting overview. J. Photon. Energy 1: 011009. Xiao, L., Chen, Z., Qu, B. et al. (2011). Recent progresses on materials for

14.

15.

16.

17.

18.

19.

20.

21.

22.

23.

electrophosphorescent organic lightemitting devices. Adv. Mater. 23 (8): 926–952. Gather, M.C. and Reineke, S. (2015). Recent advances in light outcoupling from white organic light-emitting diodes. J. Photonics Energy 5 (1): 057607. Reineke, S. (2015). Complementary LED technologies. Nat. Mater. 14: 459–462. Hung, L.S., Tang, C.W., and Mason, M.G. (1997). Enhanced electron injection in organic electroluminescence devices using an Al/LiF electrode. Appl. Phys. Lett. 70 (2): 152–154. Coehoorn, R. and van Mensfoort, S.L.M. (2009). Effects of disorder on the current density and recombination profile in organic light-emitting diodes. Phys. Rev. B 80: 085302. Cottaar, J., Coehoorn, R., and Bobbert, P.A. (2012). Modeling of charge transport across disordered organic heterojunctions. Org. Electron. 13 (4): 667–672. Kuik, M., Koster, L.J.A., Wetzelaer, G.A.H., and Blom, P.W.M. (2011). Trapassisted recombination in disordered organic semiconductors. Phys. Rev. Lett. 107: 256805. Koch, N. (2007). Organic electronic devices and their functional interfaces. ChemPhysChem 8: 1438. van Mensfoort, S.L.M., Carvelli, M., Megens, M. et al. (2010). Measuring the light emission profile in organic lightemitting diodes with nanometre spatial resolution. Nat. Photonics 4: 329–335. Pfeiffer, M., Forrest, S.R., Leo, K., and Thompson, M.E. (2002). Electrophosphorescent pin organic light-emitting devices for very-high-efficiency flatpanel displays. Adv. Mater. 14 (22): 1633–1636. Baldo, M.A., O’Brien, D.F., You, Y. et al. (1998). Highly efficient phosphorescent emission from organic electroluminescent devices. Nature 395 (6698): 151–154.

References 24. Reineke, S., Lindner, F., Schwartz,

25.

26.

27.

28.

29.

30.

31.

32.

33.

G. et al. (2009). White organic lightemitting diodes with fluorescent tube efficiency. Nature 459 (7244): 234–238. Wang, Z.B., Helander, M.G., Qiu, J. et al. (2011). Unlocking the full potential of organic light-emitting diodes on flexible plastic. Nat. Photonics 5: 753–757. Kido, J., Hayase, H., Hongawa, K. et al. (1994). Bright red light-emitting organic electroluminescent devices having a europium complex as an emitter. Appl. Phys. Lett. 65: 2124. So, F., Krummacher, B., Mathai, M.K. et al. (2007). Recent progress in solution processable organic light emitting devices. J. Appl. Phys. 102 (9): 091101–091121. Tsutsui, T., Aminaka, E., Lin, C.P., and Kim, D.U. (1997). Extended molecular design concept of molecular materials for electroluminescence: sublimed-dye films, molecularly doped polymers and polymers with chromophores. Philos. Trans. R. Soc. London, Ser. A 355 (1725): 801–814. Adachi, C., Baldo, M.A., Thompson, M.E., and Forrest, S.R. (2001). Nearly 100% internal phosphorescence efficiency in an organic light-emitting device. J. Appl. Phys. 90 (10): 5048–5051. Baldo, M.A., O’Brien, D.F., Thompson, M.E., and Forrest, S.R. (1999). Excitonic singlet-triplet ratio in a semiconducting organic thin film. Phys. Rev. B 60 (20): 14422–14428. Segal, M., Baldo, M.A., Holmes, R.J. et al. (2003). Excitonic singlet-triplet ratios in molecular and polymeric organic materials. Phys. Rev. B 68 (7): 075211–075214. Kondakov, D.Y., Pawlik, T.D., Hatwar, T.K., and Spindler, J.P. (2009). Triplet annihilation exceeding spin statistical limit in highly efficient fluorescent organic light-emitting diodes. J. Appl. Phys. 106 (12): 124510–124517. Luo, Y. and Aziz, H. (2010). Correlation between triplet-triplet annihilation and electroluminescence efficiency

34.

35.

36.

37.

38.

39.

40.

41.

42.

43.

in doped fluorescent organic lightemitting devices. Adv. Funct. Mater. 20 (8): 1285–1293. Yokoyama, D., Park, Y., Kim, B. et al. (2011). Dual efficiency enhancement by delayed fluorescence and dipole orientation in high-efficiency fluorescent organic light-emitting diodes. Appl. Phys. Lett. 99 (12): 123303. Pu, Y.-J., Nakata, G., Satoh, F. et al. (2012). Optimizing the charge balance of fluorescent organic light-emitting devices to achieve high external quantum efficiency beyond the conventional upper limit. Adv. Mater. 24: 1765–1770. Endo, A., Sato, K., Yoshimura, K. et al. (2011). Efficient up-conversion of triplet excitons into a singlet state and its application for organic light emitting diodes. Appl. Phys. Lett. 98 (8): 083302. Goushi, K., Yoshida, K., Sato, K., and Adachi, C. (2012). Organic lightemitting diodes employing efficient reverse intersystem crossing for triplet-to-singlet state conversion. Nat. Photonics 6: 253–258. Wohlgenannt, M., Tandon, K., Mazumdar, S. et al. (2001). Formation cross-sections of singlet and triplet excitons in 𝜋-conjugated polymers. Nature 409 (6819): 494–497. Shuai, Z., Beljonne, D., Silbey, R.J., and Brédas, J.L. (2000). Singlet and triplet exciton formation rates in conjugated polymer light-emitting diodes. Phys. Rev. Lett. 84 (1): 131–134. Rothe, C., King, S.M., and Monkman, A.P. (2006). Direct measurement of the singlet generation yield in polymer light-emitting diodes. Phys. Rev. Lett. 97 (7): 076602–076604. van Mensfoort, S.L.M., Billen, J., Carvelli, M. et al. (2011). Predictive modeling of the current density and radiative recombination in blue polymer-based light-emitting diodes. J. Appl. Phys. 109: 064502. Wilson, J.S., Dhoot, A.S., Seeley, A.J.A.B. et al. (2001). Spin-dependent exciton formation in 𝜋-conjugated compounds. Nature 413 (6858): 828–831. Krummacher, B.C., Nowy, S., Frischeisen, J. et al. (2009). Efficiency

281

282

65 Device Efficiency of Organic Light-Emitting Diodes

44.

45.

46.

47.

48.

49.

50.

51.

52.

analysis of organic light-emitting diodes based on optical simulation. Org. Electron. 10 (3): 478–485. Nowy, S., Krummacher, B.C., Frischeisen, J. et al. (2008). Light extraction and optical loss mechanisms in organic light-emitting diodes: influence of the emitter quantum efficiency. J. Appl. Phys. 104 (12): 123109–123119. Becker, H., Burns, S.E., and Friend, R.H. (1997). Effect of metal films on the photoluminescence and electroluminescence of conjugated polymers. Phys. Rev. B 56 (4): 1893–1905. Wasey, J.A.E., Safonov, A., Samuel, I.D.W., and Barnes, W.L. (2001). Efficiency of radiative emission from thin films of a light-emitting conjugated polymer. Phys. Rev. B 64 (20): 205201–205209. Giebink, N.C. and Forrest, S.R. (2008). Quantum efficiency roll-off at high brightness in fluorescent and phosphorescent organic light emitting diodes. Phys. Rev. B 77: 235215. Greenham, N.C., Friend, R.H., and Bradley, D.D.C. (1994). Angular dependence of the emission from a conjugated polymer light-emitting diode: implications for efficiency calculations. Adv. Mater. 6 (6): 491–494. Smith, L.H., Wasey, J.A.E., Samuel, I.D.W., and Barnes, W.L. (2005). Light out-coupling efficiencies of organic light-emitting diode structures and the effect of photoluminescence quantum yield. Adv. Funct. Mater. 15 (11): 1839–1844. Nowy, S., Frischeisen, J., and Brütting, W. (2009). Simulation based optimization of light-outcoupling in organic light-emitting diodes. Proceedings Volume 7415, Organic Light Emitting Materials and Devices XIII; 74151C. SPIE. Saxena, K., Jain, V.K., and Mehta, D.S. (2009). A review on the light extraction techniques in organic electroluminescent devices. Opt. Mater. 32 (1): 221–233. Purcell, E.M. (1946). Spontaneous emission probabilities at radio frequencies. Phys. Rev. 69: 681.

53. Drexhage, K.H. (1970). Influence of

54.

55.

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

a dielectric interface on fluorescence decay time. J. Lumin. 1–2: 693–701. Drexhage, K.H. (1974). Interaction of light with monomolecular dye layers. Prog. Opt. 12: 165. Chance, R.R., Prock, A., and Silbey, R. (1974). Lifetime of an emitting molecule near a partially reflecting surface. J. Chem. Phys. 60 (7): 2744–2748. Chance, R.R., Prock, A., and Silbey, R. (1978). Molecular fluorescence and energy transfer near interfaces. Adv. Chem. Phys. 37: 1–65. Lukosz, W. (1981). Light emission by multipole sources in thin layers. I. Radiation patterns of electric and magnetic dipoles. J. Opt. Soc. Am. 71: 744–754. Neyts, K.A. (1998). Simulation of light emission from thin-film microcavities. J. Opt. Soc. Am. A 15: 962–971. Barnes, W.L. (1998). Fluorescence near interfaces: the role of photonic mode density. J. Mod. Opt. 45 (4): 661–699. Wasey, J.A.E. and Barnes, W. (2000). Efficiency of spontaneous emission from planar microcavities. J. Mod. Opt. 47: 725–741. Gather, M.C., Flämmich, M., Danz, N. et al. (2009). Measuring the profile of the emission zone in polymeric organic light-emitting diodes. Appl. Phys. Lett. 94 (26): 263301. Schmidt, T.D., Setz, D.S., Flämmich, M. et al. (2011). Evidence for non-isotropic emitter orientation in a red phosphorescent organic light-emitting diode and its implications for determining the emitter’s radiative quantum efficiency. Appl. Phys. Lett. 99: 163302. Schmidt, T.D., Lampe, T., Sylvinson, M.R.D. et al. (2017). Emitter orientation as a key parameter in organic light-emitting diodes. Phys. Rev. Appl. 8: 037001. Flämmich, M., Gather, M.C., Danz, N. et al. (2010). Orientation of emissive dipoles in OLEDs: Quantitative in situ analysis. Org. Electron. 11 (6): 1039–1046. Flämmich, M., Michaelis, D., and Danz, N. (2011). Accessing OLED emitter

References

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

properties by radiation pattern analyses. Org. Electron. 12 (1): 83–91. Lin, C.-L., Cho, T.-Y., Chang, C.-H., and Wu, C.-C. (2006). Enhancing light outcoupling of organic light-emitting devices by locating emitters around the second antinode of the reflective metal electrode. Appl. Phys. Lett. 88 (8): 081114-1–08114-3. Celebi, K., Heidel, T.D., and Baldo, M.A. (2007). Simplified calculation of dipole energy transport in a multilayer stack using dyadic Green’s functions. Opt. Express 15 (4): 1762–1772. Kim, S.-Y. and Kim, J.-J. (2010). Outcoupling efficiency of organic light emitting diodes and the effect of ITO thickness. Org. Electron. 11: 1010–1015. Flämmich, M., Gather, M.C., Danz, N. et al. (2009). In situ measurement of the internal luminescence quantum efficiency in organic light-emitting diodes. Appl. Phys. Lett. 95: 263306. Setz, D.S., Schmidt, T.D., Flämmich, M. et al. (2011). Comprehensive efficiency analysis of organic light-emitting devices. J. Photonics Energy 1 (1): 011006. Baldo, M.A., Lamansky, S., Burrows, P.E. et al. (1999). Very high-efficiency green organic light-emitting devices based on electrophosphorescence. Appl. Phys. Lett. 75 (1): 4–6. Frischeisen, J., Yokoyama, D., Endo, A. et al. (2011). Increased light outcoupling efficiency in dye-doped small molecule organic light-emitting diodes with horizontally oriented emitters. Org. Electron. 12 (5): 809–817. Flämmich, M., Frischeisen, J., Setz, D.S. et al. (2011). Oriented phosphorescent emitters boost OLED efficiency. Org. Electron. 12: 1663. Schmidt, T.D., Setz, D.S., Flämmich, M. et al. (2013). Comprehensive efficiency analysis of organic light-emitting diodes featuring emitter orientation and triplet-to-singlet up-conversion. Appl. Phys. Lett. 103 (9): 093303. Mayr, C., Lee, S.Y., Schmidt, T.D. et al. (2014). Efficiency enhancement of organic light-emitting diodes incorporating a highly oriented thermally

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

activated delayed fluorescence emitter. Adv. Funct. Mater. 24 (33): 5232–5239. Meerholz, K. and Müller, D.C. (2001). Outsmarting waveguide losses in thinfilm light-emitting diodes. Adv. Funct. Mater. 11 (4): 251–253. Brütting, W., Frischeisen, J., Schmidt, T.D. et al. (2013). Device efficiency of organic light-emitting diodes: progress by improved light outcoupling. Phys. Status Solidi A 210 (1): 44–65. Greiner, H. (2007). Light extraction from organic light emitting diode substrates: simulation and experiment. Jpn. J. Appl. Phys. 46: 4125–4137. Greiner, H. (2010). Light extraction from OLEDs, OLED100 Summer School (Krutyn, June 2010, unpublished). Meyer, J., Winkler, T., Hamwi, S. et al. (2008). Transparent inverted organic light-emitting diodes with a tungsten oxide buffer layer. Adv. Mater. 20 (20): 3839–3843. Agrawal, M., Sun, Y., Forrest, S.R., and Peumans, P. (2007). Enhanced outcoupling from organic light-emitting diodes using aperiodic dielectric mirrors. Appl. Phys. Lett. 90 (24): 241112. Riel, H., Karg, S., Beierlein, T. et al. (2003). Phosphorescent top-emitting organic light-emitting devices with improved light outcoupling. Appl. Phys. Lett. 82 (3): 466–468. Riel, H., Karg, S., Beierlein, T. et al. (2003). Tuning the emission characteristics of top-emitting organic light-emitting devices by means of a dielectric capping layer: an experimental and theoretical study. J. Appl. Phys. 94 (8): 5290–5296. Thomschke, M., Nitsche, R., Furno, M., and Leo, K. (2009). Optimized efficiency and 1angular emission characteristics of white top-emitting organic electroluminescent diodes. Appl. Phys. Lett. 94 (8): 083303. Hofmann, S., Thomschke, M., Freitag, P. et al. (2010). Top-emitting organic light-emitting diodes: influence of cavity design. Appl. Phys. Lett. 97 (25): 253308. Smith, L.H. and Barnes, W.L. (2006). Using a low-index host layer to

283

284

65 Device Efficiency of Organic Light-Emitting Diodes

87.

88.

89.

90.

91.

92.

93.

94.

95.

96.

increase emission from organic lightemitting diode structures. Org. Electron. 7 (6): 490–494. Nakamura, T., Tsutsumi, N., Juni, N., and Fujii, H. (2005). Thin-film waveguiding mode light extraction in organic electroluminescent device using high refractive index substrate. J. Appl. Phys. 97: 054505. Gärtner, G. and Greiner, H. (2008). Light extraction from OLEDs with (high) index matched glass substrates. Proceedings Volume 6999, Organic Optoelectronics and Photonics III; 69992T. SPIE. Mladenovski, S., Neyts, K., Pavicic, D. et al. (2009). Exceptionally efficient organic light emitting devices using high refractive index substrates. Opt. Express 17 (9): 7562–7570. Madigan, C.F., Lu, M.-H., and Sturm, J.C. (2000). Improvement of output coupling efficiency of organic lightemitting diodes by backside substrate modification. Appl. Phys. Lett. 76 (13): 1650–1652. Möller, S. and Forrest, S.R. (2002). Improved light out-coupling in organic light emitting diodes employing ordered microlens arrays. J. Appl. Phys. 91 (5): 3324–3327. Yamasaki, T., Sumioka, K., and Tsutsui, T. (2000). Organic light-emitting device with an ordered monolayer of silica microspheres as a scattering medium. Appl. Phys. Lett. 76 (10): 1243–1245. Riedel, B., Kaiser, I., Hauss, J. et al. (2010). Improving the outcoupling efficiency of indium-tin-oxide-free organic light-emitting diodes via rough internal interfaces. Opt. Express 18 (23): A631–A639. Lupton, J.M., Matterson, B.J., Samuel, I.D.W. et al. (2000). Bragg scattering from periodically microstructured light emitting diodes. Appl. Phys. Lett. 77 (21): 3340–3342. Sun, Y. and Forrest, S.R. (2008). Enhanced light out-coupling of organic light-emitting devices using embedded low-index grids. Nat. Photonics 2 (8): 483–487. Koo, W.H., Jeong, S.M., Araoka, F. et al. (2010). Light extraction from organic

97.

98.

99. 100. 101.

102.

103.

104.

105.

106.

107.

108.

light-emitting diodes enhanced by spontaneously formed buckles. Nat. Photonics 4 (4): 222–226. Wedge, S., Hooper, I.R., Sage, I., and Barnes, W.L. (2004). Light emission through a corrugated metal film: the role of cross-coupled surface plasmon polaritons. Phys. Rev. B 69 (24). https:// doi.org/10.1103/physrevb.69.245418. Raether, H. (1988). Surface Plasmons on Smooth and Rough Surfaces and on Gratings. Berlin: Springer-Verlag. Maier, S.A. (2007). Plasmonics: Fundamentals and Applications. Springer. Atwater, H.A. (2007). The promise of plasmonics. Sci. Am. 296 (4): 56–63. Andrew, P. and Barnes, W.L. (2004). Energy transfer across a metal film mediated by surface plasmon polaritons. Science 306 (5698): 1002–1005. Frischeisen, J., Scholz, B.J., Arndt, B.J. et al. (2011). Strategies for light extraction from surface plasmons in organic light-emitting diodes. J. Photonics Energy 1 (1): 011004. Barnes, W.L. (1999). Electromagnetic crystals for surface plasmon polaritons and the extraction of light from emissive devices. J. Lightwave Technol. 17: 2170–2182. Wedge, S., Giannattasio, A., and Barnes, W.L. (2007). Surface plasmonpolariton mediated emission of light from top-emitting organic lightemitting diode type structures. Org. Electron. 8 (2–3): 136–147. Frischeisen, J., Niu, Q., Abdellah, A. et al. (2011). Light extraction from surface plasmons and waveguide modes in an organic light-emitting layer by nanoimprinted gratings. Opt. Express 19 (S1): A7–A19. Hauss, J., Riedel, B., Gleiss, S. et al. (2011). Periodic nanostructuring for guided mode extraction in organic light-emitting diodes. J. Photonics Energy 1 (1): 011012. Kretschmann, Erwin. (1971). Die Bestimmung optischer Konstanten von Metallen durch Anregung von Oberflächenplasmaschwingungen. Z. Phys. 241: 313–324. Winter, G. and Barnes, W.L. (2006). Emission of light through thin silver

References

109.

110.

111.

112.

113.

114.

115.

116.

films via near-field coupling to surface plasmon polaritons. Appl. Phys. Lett. 88 (5): 051109. Scholz, B.J., Frischeisen, J., Jaeger, A. et al. (2011). Extraction of surface plasmons in organic light-emitting diodes via high-index coupling. Opt. Express 20: A205. Brütting, W., Frischeisen, J., Scholz, B.J., and Schmidt, T.D. (2011). More light from organic light-emitting diodes. Europhys. News 42 (4): 20. Tien, K.-C., Lin, M.-S., Lin, Y.-H. et al. (2010). Utilizing surface plasmon polariton mediated energy transfer for tunable double-emitting organic lightemitting devices. Org. Electron. 11 (3): 397–406. Kim, J.-S., Ho, P.K.H., Greenham, N.C., and Friend, R.H. (2000). Electroluminescence emission pattern of organic light-emitting diodes: implications for device efficiency calculations. J. Appl. Phys. 88 (2): 1073–1081. Frischeisen, J., Yokoyama, D., Adachi, C., and Brütting, W. (2010). Determination of molecular dipole orientation in doped fluorescent organic thin films by photoluminescence measurements. Appl. Phys. Lett. 96 (7): 073302. Weber, W.H. and Eagen, C.F. (1979). Energy transfer from an excited dye molecule to the surface plasmons of an adjacent metal. Opt. Lett. 4: 236–238. Yokoyama, D., Sakaguchi, A., Suzuki, M., and Adachi, C. (2009). Horizontal orientation of linear-shaped organic molecules having bulky substituents in neat and doped vacuum-deposited amorphous films. Org. Electron. 10 (1): 127–137. Meerheim, R., Furno, M., Hofmann, S. et al. (2010). Quantification of energy

117.

118.

119.

120.

121.

122.

123.

124.

loss mechanisms in organic lightemitting diodes. Appl. Phys. Lett. 97 (25): 253305. Furno, M., Meerheim, R., Hofmann, S. et al. (2012). Efficiency and rate of spontaneous emission in organic electroluminescent devices. Phys. Rev. B 85: 115205. Jurow, M.J., Mayr, C., Schmidt, T.D. et al. (2016). Understanding and predicting the orientation of heteroleptic phosphors in organic light-emitting materials. Nat. Mater. 15 (1): 85–91. Kim, S.-Y., Jeong, W.-I., Mayr, C. et al. (2013). Organic light-emitting diodes with 30% external quantum efficiency based on a horizontally oriented emitter. Adv. Funct. Mater. 23 (31): 3896–3900. Kim, K.-H., Liao, J.-L., Lee, S.W. et al. (2016). Crystal organic light-emitting diodes with perfectly oriented nondoped Pt-based emitting layer. Adv. Mater. 28 (13): 2526–2532. Komino, T., Sagara, Y., Tanaka, H. et al. (2016). Electroluminescence from completely horizontally oriented dye molecules. Appl. Phys. Lett. 108 (24): 241106. Dalal, S.S., Walters, D.M., Lyubimov, I. et al. (2015). Tunable molecular orientation and elevated thermal stability of vapor-deposited organic semiconductors. Proc. Natl. Acad. Sci. U.S.A. 112 (14): 4227–4232. Adachi, C. (2014). Third-generation organic electroluminescence materials. Jpn. J. Appl. Phys. 53 (6): 060101. Dias, F.B., Penfold, T.J., and Monkman, A.P. (2017). Photophysics of thermally activated delayed fluorescence molecules. Methods Appl. Fluoresc. 5 (1): 012001.

285

287

66 Dye-Sensitized Solar Cell Sara Pescetelli, Antonio Agresti, Angelo Lembo, and Aldo Di Carlo

66.1 Introduction

Because of the growing energy request for the increased industrialization, the possibility to convert solar energy, clean and endless, to generate electricity has always been an attractive and fascinating challenge. Particularly, in recent years, alternative energy sources are gaining considerable importance also taking into account the anthropogenic greenhouse effect and the limited amount of fossil fuels. Since their onset, after the first paper published by Grätzel and O’Regan in 1991 [1, 2], dye-sensitized solar cells (DSCs) received great interest proposing as a valid alternative to traditional solid-state solar cells, and nowadays, more than 3000 publications and 2000 patents have been disclosed [3–5]. This new generation of solar cells is considered as a promising technology that aims to achieve high efficiency at low cost per square meter, also including additional features such as color and shape tunability, light weight, roll-to-roll production, and easy scalability on large area [6–11]. In the past 20 years, DSCs have had a leading role in the third-generation photovoltaic (PV) scenario, and certainly, they can be regarded as a remarkable example of bioinspired photovoltaic system by mimicking the natural photosynthesis. By simplifying the operating principle of the DSC, we can consider it as composed of a sandwich of adjacent layers working in close synergy to convert light into electricity (Figure 66.1). Light is absorbed by the dye molecule stimulating the electron excitation (Figure 66.2), which is then transferred to an electron-transporting layer and subsequently to the electrode. The dye is regenerated by a redox process involving an electrolyte. The whole mechanism generates current, but it needs the correct functioning of each component of the cell and a good matching between them in order to work efficiently. In this prospective, it became more clear that the DSCs are an admirable example of surface engineering optimization, and many efforts have been made in recent years to optimize the various components of the layers by reaching a more stable and efficient device.

Surface and Interface Science: Applications of Surface Science I, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

66 Dye-Sensitized Solar Cell

Transparent conductive oxide Catalysts

O

NCS

NCS

N

O N

N

Ru

O

N

O

O

NCS NCS

N

N

Ru

O

O

N

Dye

N

O

O

O

O

N

Ru

N

N

O

N

O

NCS NCS

Electrolyte

O

O

O

O

O

O

O

O

O

O O

TiO2 Transparent conductive oxide Figure 66.1 Schematic showing the main components of a dye-sensitized solar cell.

TCO

TiO2

Dye

Electrolyte

Cathode

S* (LUMO) –0.5 E vs. NHE (V)

288

hν 0.0

TiO2 Fermi level

hν 3I–

ERedox I3–

0.5 S°/S+ (HOMO) Load Figure 66.2 Functioning mechanism of a typical DSC based on I− /I− redox couple. 3

DSCs differ from the existing thin-film photovoltaic technologies because the physical processes of the main device, such as light absorption and charge transport, are carried out by different materials within the cells. Because of this particular feature, a DSC can be considered as a synergetic matching between the constituent layers of a device that take place at the layers’ interfaces. The possibility to play with the physical and chemical properties of the material interfaces is the reading key to clearly understand the potentialities of the DSC technology. The physical, electrical, and chemical engineering of the interfaces can thus provide to the next renewable energy market a strongly competitive product with power conversion efficiencies and long-term stability comparable to that of thin film technologies

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

but accomplished by low production costs and a wider applicability. Although photoconversion efficiencies are not yet on par with commercially available thin film or silicon solar cells, the features of transparency, light weight, flexibility, conformability, workability under low-light conditions, and easy integration in buildings as solar windows compel further dwelling in DSC modules. In fact, DSC panels have been shown to deliver even more electricity per installed power (Wh/W) compared to silicon and thin film technologies because of their superior performance in low-light and diffuse-light conditions. This makes DSC a winning solution for Building Integrated Photovoltaics (BIPV) and for indoor light harvesters used, for example, in Internet of Things.

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

The DSC (Figure 66.1) is an electrochemical device mainly composed by different constituents sandwiched together, each having specific functions. The photoelectrode (PE) consisting in a glass substrate is covered by a transparent conductive oxide (TCO). This surface is overlaid by a mesoporous n-type semiconductor, typically TiO2 , that in turn is sensitized by a monolayer of dye molecules chemisorbed on the surface. The spongy mesoporous material has a very large effective area allowing a high density of chemisorbed dye. The dye layer makes the device photoactive by promoting the absorption of the light. The PE is immersed in an electrolyte, most typically liquid, which contains a redox pair. The electrolyte usually contains the iodide/triiodide (I− /I−3 ) redox couple, where the iodide acts as a reducer and triiodide acts as an oxidant. The cell is closed by a counter electrode (CE) containing a thin layer of catalysts (such as Pt), which facilitates the transfer of electrons from the CE to the electrolyte [3, 6, 12]. Lastly, the external load completed the circuit and extracted the photogenerated power. The schematic functioning behavior of a typical DSC is shown in Figure 66.2. Typical realizations of DSC made with several dyes are shown in Figure 66.3. The color of the cell is mainly ruled by the absorption properties of the dye even though the electrolyte color (here the yellow box) could change the final color rendering of the cell. In the following, the different layers forming the DSC, their operation conditions, and their peculiar characteristics are analyzed in detail. 66.2.1 Transparent Conducting Substrate

The most common substrates for DSCs are transparent conducting oxide glasses. A transparent conductive layer, commonly indium-doped tin oxide (ITO) or fluorinedoped tin oxide (FTO), is deposited on the bottom of the glass surface facing away from the sun. The role of this substrate is very critical if we consider that it works as a diaphragm allowing sunlight pass without significant attenuation but at the

289

290

66 Dye-Sensitized Solar Cell

Figure 66.3 Typical lab-scale DSC fabricated with different dyes.

same time conducting electrons to the outer circuit. Thus, the main requisites for this layer, as its acronym reveals, are high optical transparency in the visible and near-infrared regions of electromagnetic spectrum (transparency > 80%), low sheet resistance, stability at high temperature around 500 ∘ C required to the sintering of subsequent nanostructured layer, and finally inert material to avoid deterioration caused by electrolyte or other chemical components. Flexible substrates to produce flexible DSCs are another class of substrates that find several applications connected to their light weight and flexibility such as windows, baggage, portable devices, indoor and outdoor light-powered calculator, etc. These kinds of flexible substrates are prevalent to develop on plastic layers as polyethylene terephthalate (PET) having different characteristics. In this context, the PET/ITO conductive foil is the most employed flexible substrate because FTO needs elevated temperature to be processed. Despite that ITO is a rare earth metal with less chemical stability, it has lower sheet resistance (5 Ω/◽) with respect to FTO (12 Ω/◽). Moreover, ITO films have a transmittance of over 80%, whereas FTO film exhibits a transmittance of about 75% in the visible region. Nevertheless, a comparative study of FTO- and ITO-based DSCs conducted by Sima et al. shows that FTO is highly recommended for DSCs because of its temperature-stable sheet resistance [13]. In fact, the sheet resistance of glass/ITO increases after thermal treatments above 250 ∘ C, thus making conventional sintering processes of the mesoporous oxide not viable.

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

Different methods are used to deposit the TCO depending on dimensions, chemical material, etc. The most commonly used are spray coating, spray pyrolysis, sputtering, and chemical vapor deposition (CVD). Sputtering is a preferred mode to produce ITO films, whereas CVD is a more effective process for making FTO films. 66.2.2 Semiconductor Materials

The standard material used for this semiconductor layer is TiO2 , even if other kinds of semiconductor materials are commonly employed such as ZnO and SnO2 [14]. In contrast to other semiconductors with similar band gaps, TiO2 does not photodegrade upon excitation, shows an excellent thermal stability, and it is an inert material vs. chemicals or electrolytes present in DSC. Titanium dioxide can be mainly found in three different allotropic forms, which differ for their different crystallographic properties: anatase, rutile, and brookite. For dye-sensitized solar cells, the use of anatase has been proven superior to rutile allotrope because of the structure and chemical composition of the surface and the larger band gap. The most important difference between anatase and rutile is the distance between the Ti4 + ions inside their octahedral crystalline lattice, smaller in anatase compared to rutile making anatase thermodynamically less stable. However, the phase transition from anatase to rutile occurs at a significant temperature between 700 and 1000 ∘ C depending on the crystal size and the impurity content [15]. This lower sintering temperature, together with a higher electron conductivity compared to rutile, makes anatase the leading candidate for DSC photoanodes [16, 17]. The primary rule of semiconductor layer consists in providing a support scaffold to sensitizer molecules but contemporarily have to convey the photoexcited electrons from dye to external circuit. Therefore, a performance photoanode have to exhibit contemporarily two properties: a large surface area to maximize the amount of dye and a fast charge transport rate to ensure high electron collection efficiency [18]. Thus, the standard semiconductor used in the photonade of DSCs is TiO2 nanoparticles with 10–30 nm in size, sintered together to create a connected threedimensional network of spongy porous material with a very large effective area allowing for a high density of chemisorbed dye [2]. However, at the same time, the numerous grain boundaries are also the main disadvantage of this disordered network causing a decrease of electron mobility and a recombination of photoexcited electrons by lowering the power conversion efficiency (PCE) of the cells [19, 20]. Recently, many efforts have been made to realize a more efficient nanostructured photoanode [21] in terms of materials and morphology, principally focused on film preparation techniques such as sol–gel [22], hydrothermal/solvothermal [23, 24], electrochemical anodization [25], electrospinning [26, 27], spray pyrolysis [28], and atomic layer deposition [29]. On the other hand, several other structures of nanostructured semiconductor photoanodes (e.g. TiO2 , ZnO, SnO2 , and Nb2 O5 ) are considered including nanorod [30, 31], nanotube [32–34], nanosheet [35], mesoporous

291

292

66 Dye-Sensitized Solar Cell

structures [36], and 3D hierarchical architectures [37]. Most of them exhibited considerable efficiency improvements compared to nanoparticle systems. Finally, one of the factors to consider by choosing the most suitable material for the mesoporous semiconductor is the energy band matching between this layer and conductive oxide. The indirect band gap of anatase was determined to be 3.2 eV, and on absolute scale, the position of the anatase conduction band is given with −4.4 eV [38]. Most synthesis routes induce a significant amount of trap states because of oxygen vacancies, by changing the insulating character of titanium dioxide crystal due to type of bonding within anatase, partly covalent and partly ionic. As consequence, the conductivity can vary by several orders of magnitude. The oxygen vacancies cause the formation of Ti3+ -state, which dope the crystal negatively as showed below. 66.2.3 Dye Sensitizer

Owing to the large band gap of TiO2 , only 5% of the entire electromagnetic spectrum of the sun is absorbed by the mesoporous semiconductor. In order to extend the fraction of solar energy spectrum harvested by the device in the visible and infrared regions, the semiconductor material is sensitized with the dye molecule [39, 40]. The dye sensitizer has the main rule to absorb the light providing the photogenerated electron and inject it into the semiconductor layer. In order to make it happen efficiently, the conduction bands (CBs) of TiO2 should be below (in energy) the lowest unoccupied molecular orbital (LUMO) of dye molecule, thus facilitating the injection of electrons from the dye LUMO level into the CB of TiO2 through the chemical bonding between them. In fact, the dye molecule is designed in a way that specific groups, typically the –COOH group, can promote the binding between the dye and TiO2 , ensuring and electron transfer into the CB of semiconductor with a quantum yield close to unity. Beside the level alignment, an ideal sensitizer has to satisfy several other requirements such as a broad absorption spectrum, good stability/photostability, and nontoxicity. Moreover, the dye/electrolyte interface should have a good matching with the chemical potential of redox system of the electrolyte in order to be efficiently regenerated. The performance of DSCs highly depends on the molecular structure of the sensitizers. Many chemical compounds, such as the phthalocyanines [41–43], coumarin [44–46], carboxylated derivatives of anthracene [47, 48], and porphyrins [49–51], have been used for the sensitization of semiconductors. However, photosensitizers based on transition metals have been shown to be the best so far [52]. There are three classes of photosensitizers: metal complex sensitizers, metalfree organic sensitizers, and natural sensitizers (see some examples shown in Figure 66.4) [59]. The most common, first employed, and efficient class of sensitizers are the ruthenium-based dyes [39]. This class of dyes are widely used in DSCs and their chemical formula is [{(4,4′ CO2 H)2 (bipy)}2 RuX2 ] with (bipy, 2,2′ -bipyridyl; X = Cl, Br, I, CN, and NCS). By varying the ancillary and adhesion ligands, some of the most common dyes employed in the DSCs such as N719, Z907, and black dye can

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

COOH

C6H13O

C6H13O

TB

AO OC

Metal complex dyes

C8H17O

N

N

N N

C6H13O

Ru

COOH

SM371

A

Zn N

N

A=

N

N

BA

OT

CO

N

OC8H17

N

C

S

NCS

S NC

C8H17O

OC8H17

S

N

A=

C6H13O

COOH

SM315

Black dye Ru-terpyridine complex

Porphyrin dye

C6H13O

O

OC6H13 O S

S

Organic dyes

N

COOH CN

N C6H13 C6H13 O

S

O

CN COOH

OC6H13 C6H13O

D35

Y123

Y Y +

Natural dyes

O

B

+

HO

O

Y OH

OH

2-Phenyl-1-benzopyrilium (Flavyliu≪m)

Anthocyanidin (Y = H, OH, or OCH3)

Figure 66.4 Molecular structures of some dyes used as sensitizer in DSCs: metal complex dyes [53, 54], organic (metal free) dye [55, 56], and natural dyes [57, 58].

be possibly obtained. Figure 66.5 reports an example of two DSCs realized with N719 (red dye) and Ru505 (orange dye). In dye molecules, the kinetics of electrons transfer strongly depends by charge separations [60, 61]. In particular, when a photon is absorbed by Ru–polypyridyl complex dye, the metal-to-ligand charge transfer (MLCT) transition occurs (usually in the femtosecond time scale) by promoting the excited electrons on the functional group, which is directly attached to TiO2 , in an ultrafast way (picosecond time scale) [60, 61]. The positive charge density that remains on the dye is distributed over the metal and also to some extent over the NCS ligands. This results in a spatial separation of the positive charge density on the dye and the injected electrons has the crucial effect of retarding the rate of charge recombination between the injected electrons and the dye cation, which is a key loss mechanism. These sensitizers have strong visible absorption, long excitation lifetime, and efficient MLCT [62–65]. Although record efficiency and stability have been achieved with ruthenium-based sensitizers, the high cost, the scarcity of ruthenium, and

293

294

66 Dye-Sensitized Solar Cell

S

C

N N

S C N Ru

COOTBA

N

N COOH

Ru N

COOTBA N N COOH

HOOC

TBAOOC

COOH

COOH

N719 (red dye) (a)

C N

N

N

N C

Ru505 (Orange dye) (b)

Figure 66.5 N719 and Ru505 dye molecular structure and an example of solar cell realized with these two [66].

undesired environmental impact have necessitated the consideration of other options. Metal-free organic sensitizers have been used not only to replace the expensive ruthenium-based sensitizers but also to improve the electronic properties of devices. However, the efficiency of these sensitizers is still low when compared to devices based on ruthenium-based dyes. The literature shows that the efficiency of such dyes depends on the chemistry of the electrolytes. For example, efficiencies of metal-free organic dyes in liquid, ionic, and solid-state electrolytes are greater than 8%, 6%, and 4%, respectively [67]. Another alternative class of dyes used in DSCs is natural dyes because of their low cost, easy extraction, nontoxicity, and environmentally benign nature [68]. Anthocyanins is the group of plant pigments most extensively investigated as natural sensitizers and their extracts show maximum absorption in the range of 510–548 nm, depending on the fruit or solvent used [69]. These dyes are low cost, nontoxic, and easily available. Otherwise, the efficiency of natural dyes is very low because of the weak interaction between the semiconductor (TiO2 ) and dyes. Dye aggregation on the nanocrystalline film is another important cause of low efficiency. 66.2.4 Electrolyte

The main function of the electrolyte is to regenerate the dye after the injection of the electron into the conduction band of the semiconductor. On the other hand, the electrolyte acts as a charge transport medium to transfer positive charges toward the CEs. Thus, it plays an important role in the regulation of the electron transfer kinetics, by affecting the overall cell efficiency and by strongly influencing the longterm stability of DSCs.

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

In order to meet the different tasks, the electrolyte should be designed with specific characteristics [37, 70–73]. A high electrical conductivity joint to a low viscosity are necessary for faster diffusion of electrons, but a good interfacial contact with the nanocrystalline semiconductor and the CE is also requested. Thus, it should provide a fast oxidation of I− at the photoanode/electrolyte interface for efficient dye regeneration and slow reduction of I−3 at the electrolyte/CE interface for high carrier collection. At the same time, the electrolyte must not cause harmful desorption of the dye from the oxidized surface nor the degradation of the dye and finally should not absorb light in the visible region. Moreover, the electrolyte should exibit excellent infiltration, relative high stability, low cost, and easy preparation. [74]. Electrolytes used in DSCs are classified into three types: solid-state electrolytes, quasi-solid-state electrolytes, and liquid electrolytes. The latter are further classified into two types: organic solvent-based electrolytes and room-temperature ionic liquid (RTIL) electrolytes depending on the solvent used. The composition of the electrolytes includes organic solvent, redox couple, and some additive. Most commonly, volatile organic solvents used are acetonitrile, propionitrile (PN), methoxyacetonitrile, and methoxypropionitrile (MPN) while the redox couple is basically I− /I−3 , which supply a high ion conductivity, high dielectric constant, and dissolvability of electrolyte. Although to date the efficiency obtained by using the liquid electrolyte of I− /I−3 in volatile organic solvents is higher around 11% [75], some other redox couples have been used. Wang et al. [76] employed Br− /Br2 as a redox couple in eosin-sensitized solar cells, while in alternative the use of SCN− /(SCN)2 , SeCN− /(SeCN)2 [77, 78] or substituted bipyridyl cobalt (III/II) [79] redox couples were also proposed. In order to improve the conversion efficiency of the solar cells by reducing the recombination between triiodide and electron in the conduction band of a semiconductor, alkyl imidazolium cations and lithium cations were added in the electrolyte. In fact, the alkyl imidazolium cations may be adsorbed on the surface of the semiconductor film to form the Helmholz layer, by hindering the contact of triiodide and semiconductor films. As a direct consequence of the addition of these additives in the electrolytes, the light harvesting efficiency and photocurrent of the cell as well as the stability of the dye are improved. Commonly, the electrolytes also contain 4-tert-butylpyridine (TBP) and Nmethylbenzimidazole (NMBI). The main effects are to reduce the dark current and increase the fill factor (FF) by improving photoelectric conversion efficiency. However, in recent years, in order to overcome the disadvantage of organic liquid electrolytes, ionic liquid electrolytes were developed. Because of their unique physicochemical properties, several advantages have been shown by ionic liquid electrolytes in comparison with organic ones, such as good chemical stability, high thermal stability, nonflammability, negligible vapor pressure, high ionic conductivity and high solubility for organic or inorganic materials, and a wide electrochemical window, by confirming them as good alternative electrolytes for DSCs and other electrochemical devices. Kubo et al. [80] investigated the physical and physiochemical properties of 1-alkyl-3-methylimidazolium iodides. They showed that as a result of van der Waals forces, the viscosity of the

295

296

66 Dye-Sensitized Solar Cell

molten salts increases with the increase of alkyl chain length; on the other hand, the conductivity of the molten salts decreases with the increase of viscosity as the diffusion of ions in a liquid depends on its viscosity. However, several problems still exist for the I−3 /I− electrolyte. Disadvantages such as the absorption of visible light at 430 nm, corrosion of the noble metal CE (e.g. Pt and Au), and leakage and volatility of the solvent significantly affect the longterm performance and restrict further development of DSCs using this electrolyte system [81]. In fact, it may limit device stability because the liquid may evaporate when the cell is imperfectly sealed. Penetration of water or oxygen molecules and their reaction with the electrolyte may also worsen cell performance. Thus, in order to overcome such limitations, alternative electrolytes have been developed, including Co(II/III) polypyridyl complex, ferrocenium/ferrocene (Fc/Fc+ ) couple, Cu(I/II) complex, and thiolate/disulfide mediator [74]. Several related studies on Co(II/III) complex electrolytes have been performed [56, 82]. However, Co(II/III) complex electrolytes have some limitation to be overcome such as the slow diffusion of bulky Co(II/III)complex into photoanode films and the fast recombination of photoexcited electrons with the oxidized redox species along with the long-term stability concerns [81]. Recently, the group of professor Michael Grätzel demonstrated a highly efficient DSC structure comprising Co(II/III) complex and a molecularly engineered porphyrin dye, coded SM315, achieving a surprising PCE of 13% [54]. Moreover, that pioneering work demonstrated the feasibility in replacing Pt at the CE by drop casting a suspension of graphene nanoplatelets (GNPs) onto FTO glass. As a matter of fact, a record efficiency of 14% was reported in 2014 for a cobalt-based DSC by K. Kakiage et al. [83]. In more detail, the cophotosensitization here was used as a winning approach to realize PEs with an alkoxysilyl-anchor dye (coded ADEKA-1) and a carboxyanchor organic dye (coded LEG4), by enhancing the electron injection from the light-excited dyes to the TiO2 electrodes. Moreover, the use of [Co(phen)3 ]3+/2+ redox couple in the electrolyte solution allowed achieving high open-circuit voltage (V OC ) above 1 V while higher short-circuit current density (J SC ) and a FF were gained by employing FTO/Au/GNP as the counter electrode. Of no less importance is the recent attempt in employing copper phenanthroline complexes in the solid-state dye-sensitized solar cells, in which the electrolyte solution is replaced with a solid-state hole transporting material. In some recent works [84, 85], bis(2,9-dimethyl-1,10-phenanthroline)copper(I/II) (Cu(dmp)2 ) is coupled with the organic dye LEG4 by achieving PCE more than 8% under 1000 W/m2 AM1.5G (AM, air mass) illumination, with open-circuit potentials of more than 1.0 V. In this contest, M. Freitag and coworkers [86] employed new copper complexes as promising redox mediators in DSCs. In particular, they presented two copper bipyridyl complexes, Cu(II/I)(dmby)2TFSI2/1 (0.97 V vs. SHE (standard hydrogen electrode), dmby = 6,6′ -dimethyl-2,2′ -bipyridine) and Cu(II/I)(tmby)2TFSI2/1 (0.87 V vs. SHE, tmby = 4,4′ ,6,6′ -tetramethyl-2,2′ -bipyridine), in comparison with the previously reported Cu(II/I)(dmp)2TFSI2/1 (0.93 V vs. SHE, dmp = bis(2,9-dimethyl-1,10phenanthroline)) by using the Y123 dye. The optimized devices exhibit PCE above 10% under 1000 W/m2 AM1.5G illumination, testifying an efficient dye regeneration

66.2 DSCs: Material Interfaces, Energy Matching, and Layer Morphology

at minimized driving forces. Moreover, as the photovoltage remains high, above 1 V down to 0.2 SUN light intensity, these promising copper redox mediators are more attractive for indoor applications than the conventional iodide/triiodide or cobalt-based electrolytes. Indeed, the same authors reported in 2017 a recorded PCE of 28.9% for DSC based on Cu(II/I)(tmby)2 TFSI2/1 at 1000 LX [87]. The DSC fabricated by employing the proposed copper-based electrolyte in combination with dye sensitizer cosensitization (dyes coded D35 and XY1) showed one of the highest V OC (1.1 V) ever reported in the literature with a PCE of 11.3% under 1 SUN illumination. The authors demonstrated that the two sensitizers together form a more tightly packed monolayer at the surface than the individual dyes can. This blocks the approach of the Cu(II) complex to the surface, which retards the unwanted back-electron transfer reaction and thus increases the V OC of the device. Moreover, the close matching of the redox potentials of the sensitizers with that of the Cu(II/I) complexes used as a redox shuttle allowed that the V OC values achieved are much higher than those achieved with the conventional iodide/triiodide redox system. 66.2.5 Counter Electrode

The CE has two main functions: to collect the electrons from the external load and efficiently reduce the redox electrolyte. Usually, the conductive oxide (FTO) is functionalized with a thin platinum (Pt) film acting as an electrocatalyst to enhance the injection of electron into the cell by improving the reduction reaction [88]. Platinum (Pt) is considered as a preferred catalyst because of its high exchange current density, good catalytic activity, good transparency, and it is not corroded by the electrolyte. The performance of the CE depends on the method of Pt deposition on TCO substrate. Several deposition methods have been used; the most common are thermal decomposition of hexachloroplatinic salt in isopropanol [89], electrodeposition [90], sputtering [91], vapor deposition [92], and screen printing [92]. It has been found that the activity of the Pt catalyst decreases with time in the presence of iodide/triiodide redox couple [93]. Experiments show that two major factors are responsible for the deactivation of the Pt CE: alteration of its electrocatalytic properties and the removal of Pt from the substrate [89]. In the aim to reduce costs, many other materials are tested to replace the too expensive Pt such as graphite, graphene, activated carbon, and poly(3,4ethylenedioxythiophene) (PEDOT), but typically, a reduced efficiency with respect to Pt catalysts is found [94, 95]. The main difficulty in finding a cheap Pt substitute is that it has to be inert to high corrosive I− /I−3 redox couple in the electrolyte. To prevent the problem, some conductive films such as carbon, F-doped SnO2 , etc., are employed to cover the alternative metal CEs constituted by steel or nickel. Nevertheless, alternative catalysts have been very effective for electrolytes with redox couple different from I− /I−3 .

297

298

66 Dye-Sensitized Solar Cell

66.3 DSC: Working Principle

The idea behind dye-sensitized solar cells is similar to natural photosynthesis because light absorbance with consequent electronic excitation and charge separation are distinct processes occurring in different components of the cell. The regenerative processes in DSCs allow the generation of electricity from solar light without any permanent chemical transformation [96]. The full cycle of light-to-electricity conversion is performed following a number of subsequent steps (Figure 66.6). (1) When the photon impacts on the PE, it is absorbed by the dye and causes the molecule to be photoexcited (D* ). Generally, photons of different energy between 0.5 and 3.5 eV penetrate into the dye layer since because the transparent conductive film and the TiO2 layer are transparent to the visible and infrared light. (2) An electron is transferred from the ground state S∘ (higher occupied molecular orbital, HOMO) to higher energy levels (LUMO). However, only photons of energy equal to or more than the energy gap of the dye molecule produce an effective excitation of electron in the dye molecule from HOMO to LUMO.

TiO2 – Dye + hv

→

TiO2 – Dye∗

(dye excitation by photon) (1)

TiO2 – Dye∗

→

TiO2 – Dye+ + e– (CB)

(electron injection to CB of TiO2 in ps scale) (2)

TiO2 – Dye+ + 3I–

→

TiO2 – Dye + I3

I3– + 2e– (Pt)

→

3I

–

(dye regeneration by electrolytic reduction in ms scale) (3)

–

(reduction of ectrolyte by the counter electrode) (4)

While the dark reactions which may also happen are: I3– + 2e– (CB)

→

3I

(recombination to ectrolyte from ms to s scale) (5)

TiO2 – Dye+ + e–(CB)

→

TiO2 – Dye

(recombination to dye from ms to ms scale) (6)

–

FTO glass Dye sensitized TiO2 film Sensitizer dye 2

S+/S–

Pt counter electrode Redox couple in electrolyte

1 3 – –

4

I /3

S/S+ Load

Electron flow

Figure 66.6 Light and dark reactions in a DSC. (Source: From Reddy et al. 2014 [97].)

66.3 DSC: Working Principle

The time scale of excited electron injection ranges in the order of femtoseconds to picoseconds. Furthermore, photons below minimum threshold energy do not cause an effective excitation, whereas photons of higher excitation produce hot carriers or multiple electron excitations. Hot carriers are named so because their temperatures will be high compared to the corresponding lattice temperature. The electron photogenerated moves through the ligands between the dye molecule and the titanium oxide, and then, it is ultrafastly injected inside the oxide. Hence, it percolates through the network of nanoparticles until it reaches the TCO in micro to millisecond scale and the external contacts or recombines. The collected electrons in the CB of each TiO2 should then be transported all along the semiconductor film effectively without recombination to the electrolyte or other dye molecules. (3) Subsequently, the oxidized dye D+ is immediately reduced by electron donation of the I− of the electrolyte, usually the solution of an organic solvent or ionic liquid solvent containing the I−3 /I− redox system. The oxidized dye is regenerated back in the range of nanosecond. In this way, the original state of the dye is restored. This process changes the equilibrium ratio of the redox pair concentrations inside the cell creating a concentration gradient, which moves triiodide ions toward the platinum. (4) At the cathode, the equilibrium concentration is restored by regenerating the iodide, in turn, by reduction of triiodide and the circuit is closed through the external load [12]. Notably, the process control is governed by the kinetic competition of electron transfer that refers to the time difference occurring in the various processes, which drives the cell reaction forward. Because the electric field is negligible in the semiconductor film, the transport of electrons is governed by diffusion process; as a result, the diffusion length (DL) will be at least as long as the thickness of the TiO2 electrode. Thus, in the DSCs, the kinetic competition between electron transport and recombination is essentially due to the effective carrier DL Leff , given by Leff = [Deff ⋅𝜏]1/2 , where Deff is the effective electron DL depending on the position of the quasi-Fermi level in the TiO2 , 𝜏 is the electron lifetime related to the charge recombination reaction, and typically values for Deff are 1.5 × 10−2 cm2 /s at 1 SUN [98–100]. Generally, we can consider the DL to be independent from light intensity; in fact, Deff increases with light intensity, but on the other hand, 𝜏 proportionally decreases. Typical values for DL are 5–20 μm. In optimized systems, a favorable kinetic balance assures that loss mechanisms due to recombination processes are largely suppressed, but it is necessary to consider them to give a realistic picture of DSC operation. The probability of photoelectrons to recombine back with the electrolyte or dye molecule is generically called recombination reactions, which affected the DSC performances [101, 102]. In fact, in a semiconductor layer, the mesoporous TiO2 is in contact not only with the dye but

299

300

66 Dye-Sensitized Solar Cell

also with the liquid electrolyte intercalated between the pores. Thus, during the electron transport in the CB of TiO2 , two recombination mechanisms are possible. In fact, at the TiO2 –dye–electrolyte interface, the photoinjected electrons can recombine with the oxidized dye molecules in the range of micro–milliseconds but also with I−3 ions in the electrolyte in the order of milliseconds [101, 102]. In order to avoid this latter process, an additional thin layer of compact TiO2 was deposited on the FTO layer to block the recombination route between FTO and electrolyte [103]. Overall, the photogeneration process in DSCs does not involve permanent chemical transformation, but it is governed by kinetic completion, even if in optimized systems in terms of good interfaces, an advantageous kinetics balance allows to largely obtain the suppression of loss mechanisms such as the thermalization of the dye-excited state and other recombination processes (Figure 66.7). 66.3.1 Photovoltaic Characterization of DSCs

Overall, the three parameters that characterize the performance of a DSC are energy conversion efficiency, stability, and cost. The principal photovoltaic characterization of a DSC to estimate the energy conversion efficiency is the current–voltage (I–V ) characteristics. In Figure 66.8, typical I–V curves of a DSC are reported, black indicates that the curve when the cell is retained in the dark condition, while orange indicates when the cell is illuminated. Under standard illumination conditions (intensity of incident light 100 mW/cm2 , known as one sun or AM1.5), the I–V curve is used to determine the main parameters that determinate the proper operation of the cell such as maximum power point (MPP), FF, and efficiency (𝜂) or even known as PCE. The DSC efficiency is the measure of its total amount of electrical power per unit intensity of the incident solar light. The current and potential values are measured through an external variable resistance and under standard illumination conditions. Moreover, from the I–V curve, it is possible to derive the short-circuit current generated by TiO2 film, I SC (or short circuit current density, J SC generated per unit area) in short-circuit condition when applied voltage V = 0 V, while the open-circuit potential, V OC , is the maximum voltage obtained when there is no current flow I = 0. The voltage generated under illumination ideally corresponds to the difference between the Fermi level of the electron in the semiconductor electrode and the redox potential of the electrolyte; typical values for TiO2 are around 0.5–0.7 V. Furthermore, J SC is one more factor that determines the performance of the cell, and it is affected by many factors such as light harvesting efficiency, charge separation efficiency, and charge collection efficiency. Thus, the overall conversion efficiency of the dye-sensitized cell is determined by the photocurrent density measured under short-circuit conditions (I SC ), the opencircuit photovoltage (V OC ), the FF of the cell, and the intensity of incident light (I S ): 𝜂=

JSC VOC FF IS

66.3 DSC: Working Principle

Platinum-coated FTO (back contact)

Electrolyte

Dye-covered TiO2

FTO (front contact)

Sensitizing dye

Titania nanoparticle

(a) –3.5 eV

LUMO (D*)

Energy levels (compared with vacuum)

Electron injection

(b)

Electron injection overpotential

hν

–4.0 eV EF

TiO2/electrolyte recombination

–4.5 eV

VOC

I3–/I–

TiO2/dye recombination I·2 /I

– –

–5.0 eV

HOMO (D)

Dye regeneration

Sensitizing dye

Electrolyte

Dye regeneration overpotential

–5.5 eV

Mesoporous TiO2

Figure 66.7 Dye-sensitized solar cell device schematic and operation. (a) Liquid-based DSCs comprised of a transparent conducting oxide (such as fluorine-doped tin oxide, FTO) on glass, a nanoparticle photoanode (such as titania) covered in a monolayer of sensitizing dye, a hole-conducting electrolyte and a platinum-coated, FTO-coated glass backcontact. (b) Energy level and device operation of DSCs; the sensitizing dye absorbs a

photon (energy h𝜈), the electron is injected into the conduction band of the metal oxide (titania) and travels to the front electrode (not shown). The oxidized dye is reduced by the electrolyte, which is regenerated at the counter electrode (not shown) to complete the circuit. V OC is determined by the Fermi energy level (EF) of titania and the redox potential (I− /I− ) of the electrolyte [104]. 3

Under full sunlight (1 SUN, AM1.5 global), the incident light intensity I S is 100 mW/cm2 . The maximum output power of the solar cell is found where the product of current density (J MPP ) and voltage (V MPP ) reaches a particular point on J–V curve at which the power becomes maximum, known as the MPP.

301

66 Dye-Sensitized Solar Cell

The FF of a cell is defined as the ratio of its maximum power (Pmax ) to the product of its open-circuit voltage (V OC ) and short-circuit current density (J SC ). IMPP VMPP ISC VOC

FF =

The FF is a value between 0 and 0

TCO + Pt

Electrolyte

2e

2e–

+

(b)

+

– 3+

I

2e–

–

2e → 3I

3I V supply

V 100.000 M−1 /cm) and 500–700 nm (𝜀 > 20 000 M−1 /cm) because of the large 𝜋 aromatic molecule, excellent molecular stability, and appropriate energy levels with versatile structures [154]. By employing the cosensitized method, a record-breaking efficiency of 12.3% for DSCs was obtained in 2013 by using the porphyrin YD2-oC8 cosensitized with an organic dye (Y123) [155]. Recently, a record PCE overcoming 13% have been reached by Grätzel group by employing a molecularly engineered porphyrin dye (coded SM315) [54]. On the other hand, the higher molar extinction coefficient (𝜀 = 50.000– 200.000 M−1 /cm), the cost-effective synthesis processes, and the high flexibility of the molecule structures are the main advantages of metal-free organic dyes. The abovementioned dye molecules present the typical donor–P spacer–acceptor (D–P–A) structure [156, 157], where electron-rich moieties (e.g. triarylamines, carbazoles, and indulines) are usually employed as donor parts. Polyenes, thiophenes, and benzothiadiazole are usually employed as p-spacer parts, whereas cyanoacrylic acid and rhodamines have been selected as electron-withdrawing units [10]. Nowadays, the best devices employing a slightly modified D–D-𝜋-A-type organic sensitizers with duplex starburst triphenylamine and carbazole donors achieved a strikingly high efficiency of 10% in liquid DSCs [158]. The results in terms of PCE reached during the past 15 years have been resumed in Figure 66.11 for the three different dye categories treated in this paragraph. The graph has been readapted by Ref. [159].

309

66 Dye-Sensitized Solar Cell

23 15 21 22 14 6

14 12 PCE (%)

310

5

4 2

10

8

20 7 11 10 13 12

3 9

8

19

6

1

18 17

4

Ru dyes Organic dyes Phorphyrin dyes

16

2

90

19

95

19

00

20

05 20 Years

10

20

Figure 66.11 Plots of progress in PCEs of DSC from 1991 to 2013 based on five representative sensitizers, that is, Ru dyes (1–8), organic dyes (8–12), and porphyrin dyes (13–18). The labeled numbers represent different sensitizers: (1) trimeric Ru dye [2], (2) N3 [65], (3) N719 [160], (4) N719 [64], (5) CYC-B11 [161], (6) C106 [75], (7) black dye [152], (8) TUS38 [153], (9) indoline dye [162], (10) D149 [163], (11) D205

15

20

[156], (12) C219 [157], (13) JF419 [164], (14) cosensitization with ADEKA-1 and SFD-5 [165], (15) cosensitization with ADEKA-1 and a carboxy-anchor organic dye LEG4 [83], (16) Cu-2-a-oxymesoisochlorin [133], (17) tetra(4-carboxyphenyl)porphyrin (TCPP) [166], (18) Zn-3 [103], (19) ZnTPMA-2 [167], (20) YD-2 [168], (21) YD2-oC8 [155], (22) YD2oC8 cosensitized with an organic dye (Y123) [155], (23) SM315 [54].

66.4.1.2.4 Electrolyte Engineering Electrolyte is probably the DSC element that underwent the greater number of modifications during the optimization story of DSC technology. In fact, beside the redox couple, several additives have been considered in order to improve the device performance and stability. Additives could enormously affect the dye and the TiO2 properties such as the TiO2 /Dye/electrolyte interface. Their physical and chemical properties and their effect on the entire DSC system have been systematically investigated and, even regarding the most commonly used additives, the scientific discussion is still opened. Thus, the following paragraph does not have the aim to review all the tested electrolyte additives during the brief and rich DSC story but it will provide some useful examples that can help the reader to catch a possible way to engineer the electrolyte and the interfaces involving that component. Apart from the redox couple (typically I− /I−3 ) provided by inserting organic and inorganic salt (such as KI or 1-methyl-3-propylimidazolium iodide [PMII]) and an organic solvent (most commonly used solvents are acetonitrile and 3-MPN), several additives were explored in the past and now are widely used even in commercial electrolytes such as high stability electrolyte (HSE) from Dyesol or Livion from Merck [169, 170]. First of all, we can distinguish between salts (organic or inorganic) and nitrogen-containing heterocyclic compounds. Among these nitrogen heterocyclic compounds, the effect of TBP has been deeply investigated and some aspects are

66.4 DSC: Interface and Stability

still under debate. However, TBP negatively charging the film surface with the lone pair electrons of nitrogen resulting in an upward shift of the TiO2 conduction band is generally accepted. Furthermore, TBP can coordinate Ti3+ by passivating the surface states of TiO2 layer and by reducing the recombination rate. Some authors even reported the formation of Helmholtz layer by the TBP that could suppress the recombination between I−3 in the electrolyte and the injected electron into the TiO2 layer [171]. This leads to a significant increase in the device’s V OC even if the J SC is slightly penalized [171, 172]. The difficulty in understanding the physical effect of TBP is even owing to the interaction of that additive with dye molecules or the electrolyte by negatively affecting the overall J SC [173, 174]. Similar to TBP, NMBI is also retained to coordinate with Ti3+ by suppressing the charge recombination and consequently by increasing the V OC [175]. Among organic salts, guanidinium thiocyanate (GdmSCN) is the widely studied additive, and only very recently, the real effect on the device performance and stability has been disclosed. In the past, the cation Gdm+ has been claimed to passivate the TiO2 defects by reducing the recombination at the TiO2 /electrolyte interface. Following this direction, the J SC enhancement in DSC containing GdmSCN was explained with an higher driving force for electron injection because of a downshift in the TiO2 CB [176, 177]. This erroneous interpretation has been vanished by Z. Yu et al. in 2010 [178, 179] because no experimental evidence could be found about Gdm+ adsorbed to the TiO2 by using Raman spectroscopy. Only recently, X.A. Jeanbourquin et al. investigated the interaction between dye molecules (in particular, the authors tested N719 and D131) and Gdm+ cations by pointing out an effective binding between them that competes with the dye-iodine binding. This led to a reduced concentration of bound iodine close to TiO2 by significantly reducing the TiO2 –electrolyte recombination in the presence of GdmSCN. GdmSCN is usually preferred to lithium-based salt to prevent the intercalation of Li+ into the TiO2 crystal lattice. As a matter of fact, inorganic salt has been extensively used as additives in DSC with the aim to increase the device’s J SC . A general study about the effect of alkali cations adsorption onto the TiO2 surface was provided by C. Zhang et al. in 2012 [180]. The authors showed a progressive increase of J SC in DSC in the sequence Li+ > Na+ > K+ > Cs+ > DMPI+ (base electrolyte), while V OC gradually decreased by improving the cell photovoltaic performance. Intensitymodulated photocurrent spectroscopy (IMPS), electrical impedance spectroscopy (EIS), and incident photon to current convertion efficiency (IPCE) measurements elucidated that the electron transport time 𝜏 d depended on the radius of electrolyte cations, and it decreased from small Li+ cation to large Cs+ cation. Furthermore, the efficient binding of alkali cations onto the dyed TiO2 electrode could increase the distribution of trap states in the nanostructure TiO2 electrodes by decreasing the electron collection efficiency in the order of Li+ < Na+ < K+ < Cs+ < DMPI+ . This leads to an increasing electron lifetime 𝜏 r in the order of Li+ < Na+ ≈ K+ < Cs+ . The authors finally elucidated that the increased radius of cation from small Li+ to large 1,2-dimethyl-3-propylimidazolium ion (DMPI+ ) resulted in a decreased electron injection yield for DSCs and thus a reduced J SC .

311

312

66 Dye-Sensitized Solar Cell

When different cations are mixed in the electrolyte, the values of V OC appear to be determined by the cation with the higher concentration and with the stronger adsorption tendency toward the surface of the TiO2 film [181]. Despite the numerous attempts in introducing additives in the electrolyte solution I− /I−3 redox couple has been hardly replaced by other equally efficient redox couple. As a matter of fact, iodine-based electrolytes showed the higher PCE value [74] by using standard sensitizers. However, in order to overcome the record efficiencies and to make the DSC technology reliable and durable in time, an iodine-based redox couple should be replaced. In fact, it limits the overall device efficiency by absorbing part of visible light till 430 nm and by fixing the V OC upper limit at 0.9 V because of the I− /I−3 redox potential [81]. Furthermore, iodine induces corrosion of noble metal CEs such as platinum and gold by reducing the long-term stability of the device. Alternative redox couples such as Co(II/III) have been studied in detail and it showed enhanced stability even if the slow diffusion of bulky Co(II/III) complex into photoanode films and the fast recombination of photoexcited electrons with the oxidized redox species greatly limited the device performance [82, 182]. The goal of enhanced efficiency with higher stability can be reached by the synergetic engineering of all the device components and the interfaces. A masterly exemplum was provided by Grätzel group in 2014 [54]. The authors reported a DSC overcoming 13% in efficiency by employing an engineered porphyrin dye in combination with the cobalt(II/III) redox shuttle and a GNPs/FTO CE. The proposed device overcame 0.9 V as V OC and 18 mA/cm2 as J SC by opening a viable route to further boost the DSC technology. 66.4.1.2.5 CE Engineering The CE involves electrochemical process that mainly consists of two steps: redox reactions at the electrode–electrolyte interface and charge and mass transfer through the electrode. Because of their exceptional electrocatalytic activity, noble metals such as Pt, Au, and Ag are the most employed materials for the CE realization. Furthermore, the high surface area of the Pt nanoparticles is favorable for the fast reduction of I−3 ions by increasing the photocurrent [183]. Recently, in order to further improve the cell’s PCE, 3D nanostructures with high surface areas have been proposed, such as nanowires [184], nanoflowers [185], and nanotubes [186]. However, their prohibitive cost and the poor stability when in contact with iodine redox elements within the electrolyte solution [187] forced the scientific community to provide alternative solutions. In this context, carbon-based materials have attracted great attention because of their good electrocatalytic activity accomplished by low cost, high electrical conductivity, high thermal stability, and corrosion resistance [188]. Porous carbon [189], CNTs [190], and graphene [54, 191] demonstrated the possibility to replace the standard noble metal-based CE by retaining satisfying PCE. In addition, the combination of two carbon-based material such as CNT/graphene nanoribbons [192] or porous carbon/CNTs [193] could further enhance their catalytic activity.

66.5 From Cell to the Module: Module Structure Optimization

Finally, it is worth mentioning the numerous attempts in replacing the expensive Pt CE with low-cost materials such as conductive polymers [194], metal carbides [195], nitrides [196], oxides [197], and sulfides [71, 157]. Following this direction, very recently, X. Cui et al. developed high-performance DSCs based on Ag-doped SnS2 CEs that reached a PCE of 8.70%, which exceeded the efficiency of a Pt-based DSC (7.88%) by 10.41% [198]. The obtained performance was attributed to the effectively improved electrocatalytic activity of SnS2 and mixed conductivity resulting from the Ag dopant. 66.5 From Cell to the Module: Module Structure Optimization

High efficiency and long-term stability are two fundamental requirements in order to move the DSC technology from the lab scale to the market production. The interface improvements could help in madding very efficient solar cell, but the final marketable product has to satisfy power requirements achievable only with a module configuration. Despite that the record efficiency on small area solar cell reached the threshold of 13%, the scientific and semi-industrial progresses on module tell us a different story. As we will explore in this short paragraph, the problematics related to the DSC module are different from those related to the small area devices and should be addressed by the engineering of module layout, cell interconnection, and module fabrication procedure. Nowadays, the certified efficiency on module configuration reported by M.A. Green et al. [199] showed in their review table a PCE value of 11.9 ± 0.4% for a 1.005 cm2 DSC from R. Komiya et al. [200]. In particular, the abovementioned module was characterized by the monolithic integration of the composing cell consisting layer-by-layer realization rather than by assembling two electrodes together (Figure 66.12a). +

Monolithic series

–

+

–

+

(b) +

(a)

–

Parallel grid

W-type

Z-type

(c)

(d) Glass

FTO

Dye + TiO2

Electrolyte

Platinum

Sealing

Ag

Rutile carbon

Porous spacer

Figure 66.12 Schematic of series type of connections (a) monolithic, (b) parallel grid, (c) Z-type, and (d) W-type [205].

–

313

314

66 Dye-Sensitized Solar Cell

This kind of fabrication is suitable with roll-to-roll production and allows replacing the expensive platinum CE with the cheaper carbon black. Furthermore, as the monolithic configuration is realized on a single substrate (only the working electrode [WE] is employed), the module design is cost effective and it allows to reach a ratio between the active area and the total substrate area (aspect ratio) above 90%. As a main drawback, the monolithic integration of DSCs suffers from reduced J SC and low FF because of the high series resistance value. On the other hand, when the module cell is realized by sandwiching two glassbased electrodes, series and parallel connection can be performed and the possible module configuration are schematized in Figure 66.12b–d. The parallel grid-type module (Figure 66.12b) avoids the interconnections between counter and WE by resulting in an improved module’s FF. The major drawback consists in the low active area to the silver collector. Kumara et al. [201] developed a parallel module with 7.4% of PCE by employing the abovementioned silver collecting grids. Finally, W and Z design of the module employ series connection of the cells but differ between them for the module layout and realization. The Z-connection (Figure 66.12c) employs a vertical metallic conductor to seal and connect two adjacent equal solar cells. The most critical issue concerns the fabrication and realization of thermally stable vertical interconnections, which has been solved by an integrated vertical electrical connection with combined bonds [202]. Besides that, such interconnections suffer from corrosion by the iodine/iodide couple, which could come in contact with it because of the sealant failure and they need to be protected. An hermetic glass frit sealing was proposed by Sastrawan et al. [107], while those reported by Giordano et al. [203] achieve 𝜂 up to ∼7% on aperture area in modules of ∼45 cm2 area by optimizing the geometry of the PE and using a back reflector. Finally, a PCE of 9.4% on an active area under 1000 W/m2 was reached by employing Z-connection [203]. Despite the achievable voltage value by employing Z-type connections, the module suffers from low aspect ratio because of the vertical interconnections and the surrounding sealing and from low FF because of the increased series resistance introduced by the interconnections. In order to remove them, W-type connections for module have been developed by alternating working and CEs of cells on each substrate: namely, cell’s WE facing one direction with cell’s WE facing the opposite direction. The fabrication procedure is notably simplified and the module transparency is improved even if it is really difficulty to match the J SC between the cell because of the alternation between WE and CE and thus to the different operative conditions underwent from the front and the back illuminated cells. Yet, SHARP Co. reported the highest confirmed PCE in a W-type interconnected module (8.2% and 9.3% with respect to active and total area) [204]. As both the parallel and series cell connections in a module are limited from V OC and J SC , respectively, the combination between them could be a viable solution [206]. In particular, the best performing module showed 6% in PCE on a device area of about 90 cm2 . The major drawback is the very low active area (50% of the total substrate area) because of the additional interconnection.

66.6 DSC: Stability

Despite the several module types proposed in the literature, a lower J SC is ever reported with respect to the single solar cell. The J SC decrease is mainly due to the increased number of paths and path lengths of electron diffusion into the m-TiO2 PE during scaling up; even Halme et al. [207] suggested via device simulations that the possibility of electrons to be collected is higher if they are produced nearer to the WE substrate. Thus, vertically aligned 1D structures such as nanowires or nanotubes could be considered a viable route to avoid lateral diffusion and improve the charge collection efficiency (𝜂 cc ). Unfortunately, J SC reported for small area devices that employed such 1D structures did not show comparable J SC values with respect to nanocolloidal films because of their lower roughness and thus dye uploads. Finally, the module PCE is even strongly limited because of the increase of the series resistance that strongly reduces the FF. One of the main reasons is the incorrect with of the photoanode; the optimization of the TiO2 strip width (W S ) leads to 0.8 cm for optimum FF (∼0.65–0.62) by varying [208]. Even Giordano et al. suggested W S to be kept between 0.5 and 0.7 cm for PE length up to 15 cm by providing the best compromise between the aperture area and resistive loss [202].

66.6 DSC: Stability

The stability and durability of the DSC technology is still an open issue because of the lack of a standardized aging test protocol and the low number of publications focused on this topic. Furthermore, the investigation is complicated by the complexity of the electrochemical system under study: each cell component needs to be investigated in relation with the others in order to understand the entire chemical, electrochemical, and photochemical reaction interplay at the different material interfaces. The deep comprehension of the degradation mechanisms under several aging conditions (temperature, UV exposure, light soaking, RB, and so on) is the only viable route to both improve the device stability and close the gap between the “record” and “stable” efficiencies. Because of the lack of a well-defined aging protocol for the emerging newgeneration photovoltaic technologies, the standard protocol established for terrestrial thin film PV (IEC61646) dictates the general stability requirements even for DSCs. In particular, the devices should pass

• 1000 hours at 85 ∘ C (±2 ∘ C) in dark under 85% (±5%) humidity; • 1000 hours at 60 ∘ C under light illumination (800–1000 W/m2 ); • 300 thermal cycles between −40 ∘ C and + 85 ∘ C. The stability at high temperature still remains an unsolved issue for DSC technology. As a matter of fact, no consensual publication has shown devices retaining efficiencies above 5% after 1000 hours of thermal stress at 85 ∘ C in the dark. However, very impressive results about stability have been achieved by engineering the chemical components of the cell and the relative interfaces.

315

316

66 Dye-Sensitized Solar Cell

Very recently, in 2016, T. Stergiopoulos et al. reported a DSC passing harsh accelerated thermal aging test of 3000 hours light soaking followed by additional 2000 hours thermal aging at 85 ∘ C [209] by using high boiling point solvent-based electrolyte. In 2011, Dyesol Ltd. achieved as long as 25 600 hours at 55 ∘ C–60 ∘ C under continuous light soaking with preserving 4% efficiency [210], while Kato et al. in 2009 reported a module stable for 2.5 years in outdoor conditions without specification about the starting efficiency [211]. The weakness of DSCs stems from the stress-induced degradations of the composing layer and from the unfavorable binding or reactions/interactions at the interfaces. Thus, a deep investigation onto the degradation mechanisms is required by considering the different stress test conditions and the involved device components. In the following, the main degradations occurring under harsh stress test condition will be reviewed by focusing the attention onto the possible materials and interfaces engineering that can improve the stability of the stressed device. 66.6.1 Photoelectrode Stability

The huge difficulties in satisfying the stability requirements of IEC61646 protocol are derived from the instability of the TiO2 /dye/electrolyte interface under thermal stress conditions. As recently reviewed by Frédéric Sauvage [212], the degradations almost suffered by DSC under thermal aging conditions occur as a consequence of the strong interface interactions between the electrolyte and the sensitized TiO2 photoeletcrode. In fact, when a standard DSC undergoes harsh and prolonged thermal stress (85 ∘ C in the dark), the electrolyte additives (such as TBP and imidazolium-based organic salt) and triiodide ions suffered a strong depletion. Several studies suggested the unfavorable interaction between iodine and TBP or thiocyanate ligand of the dye by only partially explaining the marked depletion of I−3 that led to a visible bleaching in electrolyte coloration [213, 214]. Not least, the reduced triiodide concentration slows the redox ions mass transport kinetics and the dye regeneration rate by negatively affecting the device’s J SC and PCE. Several authors [12, 215] tried to justify the iodine consumption by hypothesizing a reaction between the I2 and the residual water within the electrolyte solution. However, such hypothesis nowadays has not been verified because it has not been possible to detect the IO−3 complex by spectroscopic methods such as UV–Vis, Raman, etc. Even the proposed sublimation of I2 [216, 217] or the reaction between I2 and glass frit sealant [218] seem not suitable hypothesis in explaining the iodine depletion. In fact, gas analysis performed by employing thermogravimetric analysis (TGA)/mass spectrometry (MS) or gas chromatography (GC)/MS could not detect any I2 gas formation and depletion of I−3 has been clearly detected even when polymeric sealants such as Bynel or Surlin (Dupont) were employed in the complete devices [105, 219, 220]. Even dye molecule layer is believed to suffer desorption phenomena when the devices undergo stress temperature above 60 ∘ C because of the ruthenium hexacoordination rupture accelerated when water is present within the electrolyte [221].

66.6 DSC: Stability

Furthermore, when thiocyanate ligands are present in heteroleptic ruthenium (+II) complexes, they tend to undergo substitution reactions (reviewed in Figure 66.13) with different external components. Tributsch, Hagfeldt, and Lund and coworkers [213, 214, 222, 223] became consensually alarmed on the vulnerability of the strong electron donor thiocyanate ligand because of the antibonding character of these orbitals. This ligand exchange reaction is triggered both by light action (photolysis) and by temperatures above 80 ∘ C. Indeed, in ruthenium complexes, the HOMO electron densities are mainly concentrated on ancillary ligands (thiocyanate SCN− in the case of N719 and Z907 while cyanide (CN− ) ligands in the case of Ru505 sensitizer), strongly influencing the oxidation potential of the molecule. It is mainly through these ligands that the reduction of the oxidized dye D+ is carried out by the iodide ions of the electrolyte [3, 224–227]. Thus, they play a fundamental role in the charge transfer mechanisms under both light and dark conditions and especially under RB as detailed in the following. When the ancillary ligands are released, the regeneration of dye molecules is definitely compromised and the device undergoes a remarkable performance decrease. The most stable element in DSC PE seems to be the TiO2 mesoporous layer. As a matter of fact, M. Flasque et al. [228] gave further inside investigation on the PE degradations under prolonged thermal stress conditions by underlining the role of

X

C

Y

N

C

S

N

C

S

N

C

S

O

C

or

N

O

C

or N

N N

C

N

C

S

N

C

S

C N

S N

N

X

N Ru N HOOC

Y

N

N

C

S

N

C

N

COOH

N

C

C

S

N

N

C

C

S

S

H2O

N

N

C

N

C N

C

S

Figure 66.13 Resume of the different dye side products characterized on either N719 or Z907 after aging. (Source: From Sauvage 2014 [212].)

317

318

66 Dye-Sensitized Solar Cell

TiO2 layer in catalyzing several decomposition reactions at the TiO2 /electrolyte interface, involving the electrolyte compounds. Indeed, when TiO2 /electrolyte system undergoes harsh thermal stress tests (above 60 ∘ C), the consumption of triiodide occurs, accomplished by the depletion of several additives such as GdmSCN, 1,3-dimethylimidazolium iodide (DMII), and N-butyl-1H-benzimidazole (NBB). The same degradation was not observed when only the electrolyte solution underwent the same prolonged thermal stress. On the contrary, the depletion of the electrolyte components was still observed, even if greatly reduced, when TiO2 is sensitized with dye molecules. Furthermore, the authors noted a change in the coloration of TiO2 film nanoparticles more pronounced at higher temperature and longer aging time even confirmed by the TEM images that underlined a change in the morphology of the film (Figure 66.14). On the other hand, diffuse reflectance measurements confirmed that no bulk band gap modification occurred at the TiO2 level. The application of high-resolution X-ray photoelectron spectroscopy (XPS) and time-of-flight secondary ion mass spectrometry (ToF-SIMS) carried out on TiO2 PE after the aging test revealed the formation of a solid electrolyte interphase (SEI), reported till now only at low potential upon redox-active electrode materials for lithium-ion batteries [229]. In particular, when prolonged thermal stress at 60 ∘ C is applied in the dark to the investigated PE, the SEI layer is composed by iodide, sulfur, and cyano ions that originate from the solvent. Moreover, the SEI layer seems to be significantly modified when prolonged 85 ∘ C stress test condition is applied. One main difference is the inclusion of imidazolium cations (C3 H3 N2 ), which results either from DMII and/or NBB degradation. Secondly, the degradation surface layer is much thicker (3–4 nm) if compared with that observed under 60 ∘ C. The schematic representation of SEI layer for both 60 and 85 ∘ C aging condition is provided by Figure 66.14a,b.

(a)

(b)

10 nm RT

(c)

10 nm DSC ageing temperature

85 °C

Figure 66.14 TEM micrographs of TiO2 particles (a) as prepared, (b) aged for 500 hours at 60 ∘ C/100 mW/cm2 , and (c) aged for 500 hours at 85 ∘ C/dark. Insets: schematics of the SEI layer composition. (Source: From Gorlov 2008 [211].)

66.6 DSC: Stability

The authors suggested that the SEI layer growth could start from the free space between the dye molecules that act as a thermally activated catalyst for electrolyte degradation. In this scenario, the presence of dye molecules could only slow down the SEI formation [230]. As a direct consequence, the occurrence of the SEI layer can easily explain the modification observed in terms of charge-transport and charge-recombination dynamics. Thus, it seems to favor the recombination process and would consequently explain the typical trend of V OC loss encountered during aging. Not least, the SEI layer can provide a likely explanation of the iodine consumption under prolonged thermal stress, with iodine or its reduced form of iodide being trapped in the SEI. Moreover, it could insulate the dye from the electrolyte, which raises issues for the dye regeneration process. Finally, an excessive growth of SEI layer (the author demonstrated till 20 nm thickness for longer aging and the TiO2 particles completely wrapped even when sensitized) could push the dye toward desorption and prevents its readsorption when the device returns to room temperature. In order to avoid the PE degradations discussed in this paragraph, there is not a univocal solution. Regarding the dye degradation due to the substitution of the weaker ancillary ligands such as SCN− in N719 and Z907 sensitizers, several alternative dye molecules have been proposed as extensively discussed in the previous paragraph “dye engineering,” showing an improved stability to the detriment of the final device’s PCE. For example, D-𝜋-A structures actually stand out from the others as they combine high power conversion efficiencies beyond 10% [157, 231] and stability standard passing even the accelerated aging test of 85∘C/dark (Y123 dye). The rate for this ligand exchange reaction for ruthenium-based dyes can also be lowered by a factor of 2 when including a buffer concentration of thiocyanate anions in the electrolyte (e.g. GdmSCN) [232, 233]. On the other hand, the degradation products based on 4-TBP can be prevented by replacing this latter with NBB or the closely related benzimidazole, which enhances the device stability [228]. The problem of dye desorption has a huge impact on the device J SC degradation and it is still an open issue. However, N. Heo et al. [234] proposed a new methodology in making DSC consisting in adding dye molecule within the electrolyte solution. In particular, it is possible to suppress the desorption of dye from the TiO2 surface by controlling thermodynamic equilibrium. Indeed, the added dye molecules in the electrolyte can suppress the driving forces for the adsorbed dye molecules to be desorbed from TiO2 nanoparticles. The stability of DSCs based on liquid electrolyte is even compromised by the volatility of the employed solvent; under thermal aging, the excessive vapor pressure causes the breaking of the sealant and the electrolyte leakage by leading to the final device’s breakdown. This issue can be addressed by substituting the most commonly employed solvent such as acetonitrile and 3-MPN with lower boiling point solvent such as propylene carbonate (PC), PN, sulfolane, butyronitrile (BN), and 𝛾-butyrolactone (GBL). An excellent example has been recently provided by T. Stergiopoulos et al. in 2016 [209]. The authors demonstrated that solvent-based DSCs using tetraglyme as a non-nitrile, high boiling point, organic solvent for the iodide/triiodide redox shuttle could pass a harsh accelerated thermal aging test

319

320

66 Dye-Sensitized Solar Cell

of 3000 hours light soaking followed by additional 2000 hours thermal aging at 85 ∘ C. The tetraglyme-based solar cells (in contrast to cells based on MPN) showed exceptional stability, compatible with the established IEC61646 protocol for thin film PVs, and keeping c. 90% of their initial performance under 1 SUN illumination. On this route, Perganti et al. [235] showed high stable DSCs prepared by employing Z907 dye combined with liquid electrolytes using ethyl isopropyl sulfone (EiPS) high boiling point solvent. The cells were tested for their durability under harsh thermal stressing conditions of 85 ∘ C and prolonged aging time, 3000 hours in the dark. The authors demonstrated that the use of EiPS outperforms stability-wise the typical MPN solvent, improving the cell stability from 38% to 75%. Finally, both the SEI layer formation and the I−3 depletion can be avoided by replacing the I− /I−3 redox couple with the cobalt-based redox couple. In particular, Gao et al. [236] optimized the electrolyte composition (tris(2,20bipyridine)Co(II)/Co(III)-based electrolyte) by minor adjustments involving a small increase in cobalt complex concentrations and the removal of a lithium-salt additive by obtaining a significant improvement in device stability. The authors reported that a standard stability test of 1000 hours under light-soaking conditions (60 ∘ C) were conducted successfully by losing only 8% of initial device PCE. 66.6.2 Counter Electrode Stability

The stability of the most commonly employed Pt/FTO CE can be evaluated only by taking into account both the platinum deposition technique and the redox couple employed within the electrolyte solution. However, the difficulty in asserting unambiguously the stability of this new class of catalysts for DSC’s CE is owing to the lack of a general assessment onto the CE characterization techniques, as extensively reported by Yun et al. [237]. EIS, cyclic voltammetry (CV), and dark current–voltage (I–V ) characteristics are only few of the several electrochemical investigation techniques used to assess the stability of the CE. In particular, as the photovoltaic performance of DSCs is influenced by corrosion or dissolution of CE materials in contact with the I−3 /I− electrolyte, dark current–voltage tests should show abnormal changes of the dark current of the DSCs in order to assert that a chemical reaction of CEs with the I−3 /I− electrolyte has occurred. On the other hand, EIS tests reveal whether cell aging has an influence on series ohmic resistance and mass transport in the redox electrolyte solution. Thus, the following reported stability for Pt-free CE should be considered in this context, where a direct comparison in stability between the different technologies can be performed only by comparing results obtaining from the same investigation technique and by taking into account the complete cell system. Despite that platinum has been considered as one of the most stable CEs for DSC technology, nowadays, several studies [93, 238] pointed out the limited durability when in contact with iodine-based electrolyte solution. G. Syrrokostas et al. in 2012 reported the severe degradations suffered Pt electrodes prepared by electrodeposition and thermal decomposition of hexachloroplatinic acid (H2 PtCl6 ) solutions

66.6 DSC: Stability

[93]. The electrodes were stored in an iodine-based electrolyte solution or in air, for up to 70 days. Surprisingly, the electrode stored in air showed a current density reduction in the range from 15% to 25% for both the electrodeposited and thermal realized Pt CE, while a drop of about 40% was pointed out for the same electrodes when in contact with the electrolyte solution. The authors clearly demonstrated that Pt electrodes suffer strong degradation consisting in a dissolution of platinum by iodine atoms and/or absorption of I2 on the Pt layer by poising the CE. This was proven to increase the redox activation energy, an effect that is present for both electrodeposited and thermal realized Pt electrode but more marked for the electrodeposited ones, because of the different morphology of the film (in the case of electrodeposited Pt layer, the Pt nanoparticles have a size ranging from 30 to 200 nm, while the thermal decomposition method yields continuous films with honeycomb patterns). Furthermore, Pt can be poisoned by air components during the deposition [239] and it is not effective for the I-free redox couple (such as Co3+ /Co2+ and T2 /T− ) electrolyte and polysulfide electrolytes [240, 241]. These disadvantages of the Pt CEs greatly affect the CE stability and result in deteriorated device performance. The attempts reporting Pt in composition with other material such as polymer and carbon (Pt nanoparticles supported on graphene-coated FTO (PtNPs/GR) [242] and Pt nanoparticles in multi-walled carbon nanotubes (PtNPs/MWCNTs) [243]) did not solve the CE stability problem but gave a route to save Pt by retaining a satisfying catalytic activity. On the contrary, several studies focus their attention in finding an alternative Ptfree material for CE with high catalytic activity and at the same time improved stability. Among the polymer materials, PEDOT exhibited high stability, electrochemical activity, transparency, and stability [244–246]. A promising alternative to Pt comes from the transition metal carbides, nitrides, and oxides owing to their low cost, high catalytic activity, and good thermal stability [94]. In particular, carbides showed an outstanding stability and a better catalytic activity than Pt for the regeneration of new organic redox couple of di-5-(1methyltetrazole)disulfide 5-mercapto-1-methyltetrazole N-tetramethylammonium salt T2 /T− . Moreover, even transition metal sulfides such as CoS and MoS2 have shown high potentialities in replacing Pt CE [247]. In particular, F. Wang et al. [248] investigated the high catalytic activity of CoS–graphene composite CE toward iodine reduction by demonstrating that the combination of the excellent catalytic activity of CoS and fast diffusion process for electrolyte species in the graphene composite CEs led to better photovoltaic performances than CoS and graphene alone. Furthermore, the authors reported a good CE stability after 100 cyclic voltammetry scan cycles. Graphene- and carbon-based materials have been even employed as efficient and stable CEs for DSC [249]. In fact, they showed distinctive structure and unique properties. For example, Kavan’s group found that the regeneration of I−3 /I− in an ionic liquid for graphene was better than that in the traditional organic solvent [250]. Moreover, the catalytic activity strongly depends from the concentration of edge defects and oxide groups [188] and/or from the C/O ratio in the functionalized

321

322

66 Dye-Sensitized Solar Cell

graphene sheets [251], and thus, it can be tuned. Furthermore, the final device efficiency has been found to depend from the structure of carbon-based material. As an example, three-dimension (3D) honeycomb-like structured graphene sheets exhibited superior catalytic performance as CEs in DSCs, obtaining a PCE of 7.8% [252] while I-mediated DSCs with CNTs showed the highest PCE value around 10.04% [253]. Even the combination between graphene-based material such as GNP and graphene oxide (GO) showed a PCE of 9.3% when used as CEs in DSCs with the Y123 dye and the Co2 + /Co3 + mediator [254]. The improvement gained in device PCE and the economic advantage in replacing the expensive Pt catalyst could be a real step forward the commercialization of DSC technology only if accomplished by an improved long-term stability. In this respect, very few studies reported the stability test and most of all were performed by the EIS measurements and subsequent CV scans. A good example of stable Pt-free CE was provided by S. Yun et al. [255]. The authors reported that mesoporous–graphitic–carbon-supported Hafnium(IV) oxide (HfO2 -MGC) CE materials were relatively stable (no clear current reduction or peak shifts were observed in the CV curves of 30 cycle successive scans) and could catalyze both T2 and I−3 reduction. L. Kavan et al. screened various other GO-based dummy cells by EIS measurement and confirmed that the increase of the charge transfer resistance (Rct) at the CE/electrolyte apparently stopped after approximately 6–10 days of aging [254]. Another example regards GNP nanocomposite CE films that showed better mechanical stability than pure GNP, and such stability was mainly attributed to the in situ carbonized poly(acrylonitrile) that forms a strong bond at the interface of the GNP and the FTO substrate [256]. Finally, a remarkable stability was reported by C. Bu et al. for highly transparent carbon CE prepared via an in situ carbonization method [257]. The authors showed notable efficiencies for both rear and front size illumination of the cell (5% and 6%, respectively, in this case works as a bifacial device). They even demonstrated that this device retained 86% (under front-illumination) and 87% (under rear-illumination) of initial PCE after 720 hours under ambient conditions and an irradiation of 75 mW/cm2 . 66.6.3 Stability of Dye-Sensitized Module

Dye-sensitized modules have to pass the standard aging protocols established for thin film photovoltaic, as a dedicated standard does not exists yet. For this reason, several protocols have been used during the past years to evaluate the DSC module stability. First of all, standard JIS C 8938 were successfully applied to evaluate the module stability when subjected to the endurance test [258], while DyePower showed successful UV-preconditioning, humidity freeze, and damp heat IEC 61646 tests carried out over large area (>500 cm2 , Figure 66.15) modules [259]. The described stability results refer only to indoor stress conditions without

66.6 DSC: Stability

Figure 66.15 Stabilized 20 × 30 cm2 DyePower DSC module.

considering the typical stress experienced by the module under real working outdoor conditions. UV stability and chemical stability should be considered to ensure a durability of almost 10 years and eventually to move the DSCs technology toward the photovoltaic market. Despite that a standard aging protocol is yet far to be established and represents one of the harshest issue to be addressed for the DSCs technology. Furthermore, even the aging test should be designed by considering the final application of the DSC product. A DSC module integrated in fixed or portable electronic products such as sensors in the home and personal computer peripherals should not be tested as a solar module integrated in a photovoltaic window as it experiences usually low-level power irradiation (indoor condition) and thus no harsh thermal or humidity conditions. On the other hand, BIPV applications require an improved and lasting encapsulation, more stable dyes, less volatile electrolytes with stabilizing additives, and quasi or fully solid-state carrier mediators. Finally, for either parallel or series or combined interconnections, diode protection strategies should be considered, depending on the type and size of the module, in order to prevent the module cell from heavy damages when shadowing or electrical mismatch phenomena will occur on it. 66.6.4 Thermal Stability

When exposed to ambient condition, organic materials can suffer strong degradation while electrolyte solution can break the sealant and it can corrode the module metallic connections.

323

324

66 Dye-Sensitized Solar Cell

Thus, a crucial point for the DSC module stability under outdoor conditions consists in improving the robustness of the sealant and parallel to design metal grids resistant to the electrolyte corrosion. In order to stabilize the electrolyte solution, the Energy research Centre of the Netherlands (ECN) added MgI2 and CaI2 by showing, for a parallel-type module, a long-term stability up to 8300 hours at 2.5 SUN by employing Surlyn as encapsulant material [260]; such time is equivalent to 10 years of outdoor operation. A remarkable thermal stability was reported by Grätzel group for a device employing a low volatile robust electrolyte in conjunction with an amphiphilic ruthenium dye (K19). Indeed, the authors reported a 8.2% efficient device retaining 98% of the original PCE value after soaking at 60 ∘ C for 1000 hours [5]. Dyesol Ltd. presented DSCs showing stability over 90% for 1000 hours at 95 ∘ C by employing stable dyes (N709 and Y123) and 3-MPN-based electrolyte [261]. Moreover, the substitution of Surlin with more stable sealants such as Bynel or UV-cured epoxy glue has given an additional impetus toward the DSC module commercialization. In this regard, Kroon et al. of ECN presented significantly stable (80%) performance at >80 ∘ C for 3000 hours. Finally, the use of alternative materials for collecting electrodes such as carbon, titanium, or tin (at cathode) could very much improve the stability of the DSC module. 66.6.5 Outdoor Stability

The outdoor stability of dye-sensitized module has been extensively tested by Mastroianni et al. [262]. In particular, a module composed by 30 cells with 3.6 cm2 active area were tested indoor (1 SUN and 85 ∘ C) and outdoor (both horizontally placed and tilted at 25 ∘ C) by showing an overall stability of 3200 hours with only 10% fall of the original PCE. However, the vertically oriented module showed a faster degradation probably because of uneven triiodide concentration by underlining the importance of the module orientation over the module stability. Regarding the outdoor stability, we can even cite Asghar et al. [263] that demonstrated Z907 and K19 as the most stable dyes for sensitization an polymer/ionic liquid electrolyte as a viable route to obtain great stability under light soaking test (60 ∘ C) and thermal stress at 80 ∘ C for 1000 hours. 66.6.6 Stability and Degradation Under Reverse Bias

Although the high-performance DSC devices meet the conversion efficiency criterion for market introduction, the requirement for high stability and the aging predictability are still open issues. In particular, the clarification of the degradation mechanisms under real working conditions is still under investigation. As aging test standardization is still missing, a stress test resembling the working conditions of a solar cell embedded in a module configuration is requested, especially in an up-scaled and commercialized technology.

66.6 DSC: Stability

During the real working conditions, series-connected cells in modules can have different performances because of extrinsic phenomena such as shadowing, leading to current mismatch. From the stability point of view, the mismatch can force the least performing cells to work under RB and to behave as a dissipating load [264]. The prolonged RB condition leads to severe degradations of chemical compounds composing the cell (both sensitizer molecules and electrolyte solution) and finally to the complete device breakdown [105]. Thus, the detection and the clarification of the degradation mechanisms under real working conditions are crucial steps in order to improve the device stability and to design more stable chemical compounds. In this optic, the in situ resonance Raman spectroscopy (RRS) can be applied on each device layer as a useful and nondestructive characterization technique in order to follow the device chemical degradations during the aging time [66, 265, 266]. In particular, in this paragraph, the attention is focused on the degradation mechanisms involving the sensitizer, the electrolyte solution, and the interface between them in a device based on the most commonly used and studied materials for this technology: cis-dithiocyanate-N,N 0 -bis-(4-carboxylate-40tetrabutilammoniumcarboxy-late-2,20-bipyridine) ruthenium(II) as dye molecule (even known as N719) and the commercially available HSE from Dyesol as a electrolyte solution employing the iodine-based redox couple (I− /I−3 ) [267]. When composing cells in a module are series connected, the shadowed cell suffers the harshest stress when the module is working at short-circuit condition. In fact, short-circuit current (I SC ) is more sensitive than open-circuit voltage (V OC ) to the shadowing as it is directly proportional to the incident photon flux, whereas V OC is logarithmically dependent. A similar behavior occurs when the cell is damaged and it suffers a partial active area loss. In order to simulate the RB condition experienced by a series connected cell in a module, a fixed current is applied to the device in the dark as reported in Figure 66.16a. The applied current value is about three times higher than the typical cell’s shortcircuit current under 1 SUN illumination condition, in order to accelerate the stress test: the stress-induced degradation mechanisms have been demonstrated to be the same even for lower current values [264]. The degradation processes induced by RB stress involve both the electrolyte solution and the sensitizer during the aging time. In particular, in RB regime, the electrochemical stress induces at PEs an increased concentration of I−3 and of oxidized dye molecules favoring the formation of the [D+ –I−3 ] adduct (Figure 66.16b), clearly detectable by Raman spectroscopy. The stress-induced subproduct at TiO2 /dye/electrolyte interface inhibits efficient dye regeneration by resulting in the formation of a permanent population of oxidized dyes molecules. Meanwhile, polyiodide formation at FTO/Pt/electrolyte interface induces a strong unbalance in the redox couple consisting in triiodide depletion, pointed out by means of fluorescence spectroscopy. The slowing down of the dye regeneration process and the I−3 depletion mechanism results in a gradual increase of anodic operating voltage (V RB ) and in a decrease in the diffusion-limited current (I lim ), respectively. In particular, as V RB is strongly dependent on the chemical structure

325

326

66 Dye-Sensitized Solar Cell

Reverse bias

Glass

e–

e–

TCO e– P E

Power supply

Electrolyte

l3 –

l–

e– l–

l3 –

+

(a)

Pt D+

–

Platinum

e–

3l –

+

TIO2/Dye

Dye

C E –

TiO2

(b) Figure 66.16 (a) Schematic overview of a DSC under RB condition and (b) schematic representation of the charge transfer processes under dark in a complete DSC for positive voltages at TiO2 (RB). The physicists’ convention is here used: (i) under

illumination the photogenerated current is negative at positive voltages and (ii) under the dark by applying a positive potential to the TiO2 , the voltage and the current are negative (reverse bias and anodic currents) [267].

of the dye [105, 264], its reduction is symptomatic of changes involving dye chromophores in the N719 molecular structure. When I lim approaches the current forced on the reverse-biased cell, the device starts to work under limited diffusion current condition and the electrolyte solution starts to produce gas bubbles causing sealing failure, electrolyte leakage from active area, and strong color heterogeneity on PE. Raman spectroscopy on a completely degraded solar cell clarified that the apparent N719 dye bleaching in particular zones of the cell active area can be reasonably attributed to the irreversible structural changes that affect dye molecule in some zones of PE which are not in contact with electrolyte solution. In fact, during the turbulent final stages of RB stress test, the absence of electrolyte in the bleached zone prevents dye regeneration and, consequently, promotes the detachment of the ancillary group (SCN− ) from ruthenium. This translates in the formation of a relevant population of [D+ –I−3 ] adducts, observable by monitoring the intensity of the characteristic Raman band over the device’s active area, as reported in Figure 66.17 [66]. Thus, by clarifying the degradation mechanisms occurring in a DSC under real working conditions, the present study provides a viable route to improve the device long-term stability and to move this technology toward the next commercialization. The presented degradation mechanism investigation point out the relevance of DSC interface onto long-term stability. Thus, the design of new, stable, and efficient solvent-based electrolytes as well as sensitizers should take into account the interplay between different cell constituents at their interfaces, toward promising photovoltaic technology.

References 1.00 0.95 0.90 0.85

0.75 0.70 0.65

Intensity

0.80

0.60 0.55 0.50 0.45

10 μm

0.40

10 μm

0.35

(a)

(b)

Figure 66.17 (a) [D+ –I− ] Raman band intensity map on RB stressed N719-sensitized cell 3 and (b) microscopic detail of scanned area [66].

References 1. Snaith, H.J. and Schmidt-Mende, L. 2. 3. 4.

5. 6. 7. 8. 9. 10.

11. 12. 13. 14.

(2007). Adv. Mater. 19: 3187. O’Regan, B.C. and Grätzel, M. (1991). Nature 353: 737. Hagfeldt, A. and Grätzel, M. (2000). Acc. Chem. Res. 33: 269. Baxter, J.B. and Vac, J. (2012). Sci. Technol. A Vacuum, Surfaces, Film. 30: 20801. ´ Humphry-Baker, R. Wang, P., Klein, C., et al. (2005). Appl. Phys. Lett. 86: 1. Grätzel, M. (2003). J. Photochem. Photobiol., C 4: 145. Hagfeldt, A., Boschloo, G., Henrik, S.L. et al. (2010). Chem. Rev.: 110. Peter, L.M. (2011). J. Phys. Chem. 2: 1861. Tulloch, G.E. (2004). J. Photochem. Photobiol., A 164: 209. Zhang XD Numata, S.F.Y. and YH Han, L.Y. (2013). Energy Environ. Sci. 6: 1464. Grätzel, M. (2009). Acc. Chem. Res. 42: 1788. Hagfeldt, A., Boschloo, G., Sun, L. et al. (2010). Chem. Rev. 110: 6595. Sima, C., Grigoriu, C., and Antohe, S. (2010). Thin Solid Films 519: 595. Jose, R., Thavasi, V., and Ramakrishna, S. (2009). J. Am. Ceram. Soc. 92: 289.

15. Park, N.-G., van de Lagemaat, J., and

16. 17. 18.

19. 20. 21.

22. 23. 24. 25.

26.

27.

Frank, A.J. (2000). J. Phys. Chem. B 104: 8989. Diebold, U. (2003). Appl. Surf. Sci. 48: 53. Ovenstone, J. and Yanagisawa, K. (1999). Chem. Mater. 11: 2770. Maçaira, J., Andrade, L., and Mendes, A. (2013). Renewable Sustainable Energy Rev. 27: 334. Feng, X., Zhu, K., Frank, A.J. et al. (2012). Sol. Cells: 2781. Wang, J. and Lin, Z. (2012). Chem. Asian J. 7: 2754. Colodrero, S., Forneli, A., Lòpez-Lòpez, C. et al. (2012). Adv. Funct. Mater. 22: 1303. Bahramian, A. (2013). Ind. Eng. Chem. Res. 52: 14837. Wu, H.P., Lan, C.M., Hu, J.Y. et al. (2013). J. Phys. Chem. Lett. 4: 1570. Bao, Z.Q., Xie, H., Zhu, Q. et al. (2013). CrystEngComm 15: 8972. Mir, N., Lee, K., Paramasivam, I., and Schmuki, P. (2012). Chem. Eur. J. 18: 11862. Chen, H.Y., Zhang, T.L., Fan, J. et al. (2013). ACS Appl. Mater. Interfaces 5: 9205. Kumar, E.N., Jose, R., Archana, P.S. et al. (2012). Energy Environ. Sci. 5: 5401.

327

328

66 Dye-Sensitized Solar Cell 28. Huo, J., Hu, Y., Jiang, H. et al. (2013). 29.

30. 31. 32. 33. 34. 35. 36. 37. 38.

39.

40. 41. 42.

43. 44. 45. 46.

47. 48.

49.

50.

Ind. Eng. Chem. Res. 52: 11029. Son, H., Prasittichai, C., Mondloch, J.E. et al. (2013). J. Am. Chem. Soc. 135: 11529. Lv, M., Zheng, D., Ye, M. et al. (2013). Energy Environ. Sci. (6): 1615. Zhang, H., Wang, Y., Yang, D. et al. (2012). Adv. Mater. 24: 1598. Ye, M., Zheng, D., Lv, M. et al. (2013). Adv. Mater. 25: 3039. Ye, M., Xin, X., Lin, C., and Lin, Z. (2011). Nano Lett. 3214. Anh, V., Thinh, T., Anh, T. et al. (2017). Appl. Surf. Sci. 399: 515. Shi, Y., Zhu, C., Wang, L. et al. (2013). Chem. Eur. J. 19: 282. K. Li, Y. Wang, Y. Xu, H. Chen, C. Su, D. Kuang, 2013. Sauvage, F., Di Fonzo, F., Li Bassi, A. et al. (2010). Nano Lett. 10: 2562. Scanlon, D.O., Dunnill, C.W., Buckeridge, J. et al. (2013). Nat. Mater. 12: 798. Nazeeruddin, M.K., Humphry-Baker, R., Liska, P., and Grätzel, M. (2003). J. Phys. Chem. B 107: 8981. Reynal, A., Forneli, A., and Palomares, E. (2010). Energy Environ. Sci. 3: 805. Giribabu, L., Singh, V.K., Jella, T. et al. (2013). Dyes Pigm. 98: 518. Balraju, P., Kumar, M., Roy, M.S., and Sharma, G.D. (2009). Synth. Met. 159: 1325. Huang, H., Cao, Z., Li, X. et al. (2012). Synth. Met. 162: 2316. Hara, K., Sato, T., Katoh, R. et al. (2003). J. Phys. Chem. B 107: 597. Seo, K.D., Song, H.M., Lee, M.J. et al. (2011). Dyes Pigm. 90: 304. Hara, K., Tachibana, Y., Ohga, Y. et al. (2003). Sol. Energy Mater. Sol. Cells 77: 89. Justin Thomas, K.R., Singh, P., Baheti, A. et al. (2011). Dyes Pigm. 91: 33. Mikroyannidis, J.A., Tsagkournos, D.V., Sharma, S.S. et al. (2010). Sol. Energy Mater. Sol. Cells 94: 2318. Xiang, N., Zhou, W., Jiang, S. et al. (2011). Sol. Energy Mater. Sol. Cells 95: 1174. Panda, M.K., Ladomenou, K., and Coutsolelos, A.G. (2012). Coord. Chem. Rev. 256: 2601.

51. Zhou, W., Zhao, B., Shen, P. et al.

(2011). Dyes Pigm. 91: 404. 52. Kalyanasundaram, K. and Grätzel, M.

(2009). Materials Matter. 4.4 (88): 1. 53. Wang, Z.S., Yamaguchi, T., Sugihara,

54. 55.

56. 57.

58.

59.

60. 61. 62.

63.

64. 65.

66. 67.

68. 69. 70. 71. 72.

H., and Arakawa, H. (2005). Langmuir 21: 4272. Mathew, S., Yella, A., Gao, P. et al. (2014). Nat. Chem. 6: 242. Jiang, X., Karlsson, K.M., Gabrielsson, E. et al. (2011). Adv. Funct. Mater. 21: 2944. Yum, J.-H., Baranoff, E., Kessler, F. et al. (2012). Nat. Commun. 3: 631. Calogero, G., Sinopoli, A., Citro, I. et al. (2013). Photochem. Photobiol. Sci. 12: 883. Richhariya, G., Kumar, A., Tekasakul, P., and Gupta, B. (2017). Renewable Sustainable Energy Rev. 69: 705. Mehmood, U., Rahman, S., Harrabi, K. et al. (2014). Adv. Mater. Sci. Eng. 2014: 1. Tachibana, Y., Moser, J.E., Gratzel, M. et al. (1996). J. Phys. Chem. 100: 20056. Haque, S.A., Tachibana, Y., Willis, R.L. et al. (2000). J. Phys. Chem. B 104: 538. Polo, A.S., Itokazu, M.K., and Murakami Iha, N.Y. (2004). Coord. Chem. Rev. 248: 1343. Takahashi, Y., Arakawa, H., Sugihara, H. et al. (2000). Inorg. Chim. Acta 310: 169. Yanagida, M., Islam, A., Tachibana, Y. et al. (2002). New J. Chem. 26: 963. Nazeeruddin, M.K., Kay, A., Rodicio, I. et al. (1993). J. Am. Chem. Soc. 115: 6382. Agresti, A., Pescetelli, S., Quatela, A. et al. (2014). RSC Adv. 4: 12366. Mishra, A., Fischer, M.K.R., and Büuerle, P. (2009). Angew. Chem. Int. Ed. 48: 2474. Dai, Q. and Rabani, J. (2001). Chem. Commun. 2: 2142. Narayan, M.R. (2012). Renewable Sustainable Energy Rev. 16: 208. Law, M., Greene, L.E., Johnson, J.C. et al. (2005). Nat. Mater. 4: 455. Greene, L.E., Law, M., Tan, D.H. et al. (2005). Nano Lett. 5: 1231. Varghese, O.K., Paulose, M., and Grimes, C.A. (2009). Nat. Nanotechnol. 4: 592.

References 73. Kang, T.-S., Smith, A., Taylor, B., and 74.

75. 76. 77.

78.

79.

80. 81. 82.

83. 84. 85. 86.

87. 88.

89. 90.

91. 92.

93.

94.

Durstock, M. (2009). Nano Lett. 9: 601. Wang, M., Grätzel, C., Zakeeruddin, S.M., and Grätzel, M. (2012). Energy Environ. Sci. 5: 9394. Yu, Q., Wang, Y., Yi, Z. et al. (2010). ACS Nano 4: 6032. Wang, Z.S., Sayama, K., and Sugihara, H. (2005). J. Phys. Chem. B 109: 22449. Bergeron, B.V., Marton, A., Oskam, G., and Meyer, G.J. (2005). J. Phys. Chem. B 109: 937. Oskam, G., Bergeron, B.V., Meyer, G.J., and Searson, P.C. (2001). J. Phys. Chem. B 105: 6867. Sapp, S.A., Elliott, C.M., Contado, C. et al. (2002). J. Am. Chem. Soc. 124: 11215. Kubo, W., Kitamura, T., Hanabusa, K. et al. (2002). Chem. Commun.: 374. Yen, Y.-S., Chou, H.-H., Chen, Y.-C. et al. (2012). J. Mater. Chem. 22: 8734. Kashif, M.K., Nippe, M., Duffy, N.W. et al. (2013). Angew. Chem. Int. Ed. 52: 5527. Kakiage, K., Aoyama, Y., Yano, T. et al. (2015). Chem. Commun. 51: 15894. Freitag, M., Giordano, F., Yang, W. et al. (2016). J. Phys. Chem. C 120: 9595. Freitag, M., Daniel, Q., Pazoki, M. et al. (2015). Energy Environ. Sci. 8: 2634. Saygili, Y., Söderberg, M., Pellet, N. et al. (2016). J. Am. Chem. Soc. 138: 15087. Freitag, M., Teuscher, J., Saygili, Y. et al. (2017). Nat. Photonics: 1. Thomas, S., Deepak, T.G., Anjusree, G.S. et al. (2014). Mater. Chem. A 2: 4474. Papageorgiou, N. (1997). J. Electrochem. Soc. 144: 876. Tsekouras, G., Mozer, A.J., and Wallace, G.G. (2008). J. Electrochem. Soc. 155: K124. Fang, X., Ma, T., Guan, G. et al. (2004). J. Photochem. Photobiol., A 164: 179. Khelashvili, G., Behrens, S., Weidenthaler, C. et al. (2006). Thin Solid Films 511–512: 342. Syrrokostas, G., Siokou, A., Leftheriotis, G., and Yianoulis, P. (2012). Sol. Energy Mater. Sol. Cells 103: 119. Wu, M., Lin, X., Wang, Y. et al. (2012). J. Am. Chem. Soc. 134: 3419.

95. Andrade, L., Ribeiro, H.A., and

96. 97.

98. 99.

100. 101.

102.

103. 104. 105.

106.

107.

108. 109. 110. 111.

112.

113. 114. 115.

Mendes, A. (2011). Encycl. Inorg. Bioinorg. Chem. Zhang, S., Yang, X., Qin, C. et al. (2014). J. Mater. Chem. A (2): 5167. Reddy, K.G., Deepak, T.G., Anjusree, G.S. et al. (2014). Phys. Chem. Chem. Phys. 16: 6838. Kay, A. and Grätzel, M. (1996). Sol. Energy Mater. Sol. Cells 44: 99. Dloczik, L., Ileperuma, O., Lauermann, I. et al. (1997). J. Phys. Chem. B 101: 10281. Bisquert, J.V. and Vikhrenko, V.S. (2004). J. Phys. Chem. B 108: 2322. Bisquert, J. and Fabregat-santiago, F. (2009). J. Phys. Chem. 113 (40): 17278–17290. Huang, S.Y., Schlichthrlo, G., Nozik, A.J., Graetzel, M. and Frank, A. J. (1997) J. Phys. Chem. B, 101: 2576–2582. Wang, Q., Moser, J.-E., and Grätzel, M. (2005). J. Phys. Chem. B 109: 14945. Hardin, B.E., Snaith, J., and Mcgehee, M.D. (2012). Nat. Photonics 6: 162. Mastroianni, S., Lembo, A., Brown, T.M. et al. (2012). ChemPhysChem 13: 2964. Fabregat-Santiago, F., Bisquert, J., Palomares, E. et al. (2007). J. Phys. Chem. C 111: 6550. Sastrawan, R., Renz, J., Prahl, C. et al. (2006). J. Photochem. Photobiol., A 178: 33. Peter, L.M. (2007). Phys. Chem. Chem. Phys. 9: 2630. Ito, S., Murakami, T.N., Comte, P. et al. (2008). Thin Solid Films 516: 4613. Chen, T., Hu, W., Song, J. et al. (2012). Adv. Funct. Mater. 22: 5245. Xia, J., Masaki, N., Jiang, K., and Yanagida, S. (2007). J. Phys. Chem. B 110: 25222. Zhang, Y., Wu, L., Liu, Y., and Xie, E. (2009). J. Phys. D: Appl. Phys. 42: 85105. Jung, H.S., Lee, J.-K., Nastasi, M. et al. (2005). Langmuir 21: 10332. Law, M., Greene, L.E., Radenovic, A. et al. (2006). J. Phys. Chem. 101: 22652. Nguyen, T.V., Lee, H.C., Alam Khan, M., and Yang, O.B. (2007). Sol. Energy 81: 529.

329

330

66 Dye-Sensitized Solar Cell 116. Kim, H., Park, S.Y., Kwak, S. (2011).

117. 118. 119. 120.

121. 122.

123.

124. 125.

126. 127.

128. 129. 130.

131. 132. 133. 134.

135. 136.

2010 International Conference on Biology, Environment, and Chemistry, volume 1, p. 178 (28 to 30 December 2010) Hong Kong, China. Sudhagar, P., Asokan, K., Jung, J.H. et al. (2010). Nanoscale Res. Lett. 6: 30. Duong, T.-T., Choi, H.-J., He, Q.-J. et al. (2013). J. Alloys Compd. 561: 206. Xin, X., Scheiner, M., Ye, M., and Lin, Z. (2011). Langmuir 27: 14594. Bazzan, G., Deneault, J.R., Kang, T.S. et al. (2011). Adv. Funct. Mater. 21: 3268. Wu, S., Han, H., Tai, Q. et al. (2008). Nanotechnology 19: 215704. Lee, S., Jun, Y., Kim, K., and Kim, D. (2001). Sol. Energy Mater. Sol. Cells 65: 193. O’Regan, B.C., Durrant, J.R., Sommeling, P.M., and Bakker, N.J. (2007). J. Phys. Chem. C 111: 14001. Peng, T., Fan, K., Zhao, D., and Chen, J. (2010). J. Phys. Chem. C 114: 22346. Wooh, S., Kim, T.-Y., Song, D. et al. (2015). ACS Appl. Mater. Interfaces 7: 25741. Ozawa, H., Okuyama, Y., and Arakawa, H. (2012). Dalton Trans. 41: 5137. Chandiran, A.K., Tetreault, N., Humphry-baker, R. et al. (2012). Nano Lett. 12: 1. Lim, S.P., Pandikumar, A., Lim, H.N. et al. (2015). Sci. Rep. 5: 11922. Schaadt, D.M., Feng, B., and Yu, E.T. (2005). Appl. Phys. Lett. 86: 1. Kang, S.H., Kim, H.S., Kim, J.Y., and Sung, Y.E. (2010). Mater. Chem. Phys. 124: 422. Kim, J.-Y., Kang, S.H., Kim, H.S., and Sung, Y.-E. (2010). Langmuir 26: 2864. Seo, H., Wang, Y., Ichida, D. et al. (2013). Jpn. J. Appl. Phys. 52: 11NM02. Kay, A. and Grätzel, M. (1993). J. Phys. Chem. 97: 6272. Neale, N.R., Kopidakis, N., Van De Lagemaat, J. et al. (2005). J. Phys. Chem. B 109: 23183. Cacialli, F., Wilson, J.S., Michels, J.J. et al. (2002). Nat. Mater. 1: 160. Nair, A.S., Peining, Z., Babu, V.J. et al. (2011). Phys. Chem. Chem. Phys. 13: 21248.

137. Kay, A. and Grätzel, M. (2002). Chem.

Mater. 14: 2930. 138. Zhai, T., Fang, X., Liao, M. et al. (2009).

Sensors 9: 6504. 139. Zhang, Q. and Cao, G. (2011). Nano

Today 6: 91. 140. Karthik, S., Gopal, K.M., Haripriya, E.P.

et al. (2007). Nanotechnology 18: 65707. 141. Krishnamoorthy, T., Tang, M.Z., Verma,

142. 143.

144. 145. 146. 147.

148. 149. 150.

151. 152. 153. 154. 155. 156. 157. 158. 159. 160. 161.

A. et al. (2012). J. Mater. Chem. 22: 2166. Lee, T.Y., Alegaonkar, P.S., and Yoo, J.B. (2007). Thin Solid Films 515: 5131. Lee, W.J., Ramasamy, E., Lee, D.Y., and Song, J.S. (2009). ACS Appl. Mater. Interfaces 1: 1145. Nilsing, M., Lunell, S., Persson, P., and Ojamae, L. (2005). Surf. Sci. 582: 49. Wang, X., Kitao, O., Zhou, H. et al. (2009). J. Phys. Chem. C 113: 7954. Li, X., Lin, H., Zakeeruddin, S.M. et al. (2009). Chem. Lett. 38: 322. Choi, H., Kim, S., Kang, S.O. et al. (2008). Angew. Chem. Int. Ed. 120: 8383. Dawoud, B., Amer, E., and Gross, D. (2007). Int. J. Energy Res. 31: 135. Robertson, N. (2006). Angew. Chem. Int. Ed. 45: 2338. Huang, W.-K., Wu, H.-P., Lin, P.-L. et al. (2012). J. Phys. Chem. Lett. 3: 1830. Wang, S.-W., Wu, K.-L., Ghadiri, E. et al. (2013). Chem. Sci. 4: 2423. Han, L., Islam, A., Chen, H. et al. (2012). Energy Environ. Sci. 5: 6057. Ozawa, H., Sugiura, T., Kuroda, T. et al. (2016). J. Mater. Chem. A 4: 1762. Li, L.-L. and Diau, E.W.-G. (2013). Chem. Soc. Rev. 42: 291. Yella, A., Lee, H.-W., Tsao, H.N. et al. (2011). Science 334: 629. Ito, S., Miura, H., Uchida, S. et al. (2008). Chem. Commun. (Camb).: 5194. Zeng, W., Cao, Y., Bai, Y. et al. (2010). Chem. Mater. 22: 1915. Tan, L.-L., Huang, J.-F., Shen, Y. et al. (2014). J. Mater. Chem. A 2: 8988. Ye, M., Wen, X., Wang, M. et al. (2015). Mater. Today 18: 155. Barbe, C.J., Arendse, F., Comte, P. et al. (1997). J. Am. Ceram. Soc. 80: 3157. Chen, C., Wang, M., Li, J. et al. (2009). ACS Nano 3: 3103.

References 162. Horiuchi, T., Miura, H., Sumioka, K.,

163.

164.

165. 166. 167.

168.

169.

170.

171. 172. 173. 174. 175. 176.

177. 178.

179.

and Uchida, S. (2004). J. Am. Chem. Soc. 126: 12218. Ito, S., Zakeeruddin, S.M., Humphry-Baker, R. et al. (2006). Adv. Mater. 18: 1202. Yella, A., Humphry-Baker, R., Curchod, B.F.E. et al. (2013). Chem. Mater. 25: 2733. Kakiage, K., Aoyama, Y., Yano, T. et al. (2015). Chem. Commun. 51: 6315. Cherian, S. and Wamser, C.C. (2000). J. Phys. Chem. B 104: 3624. Campbell, W.M., Jolley, K.W., Wagner, P. et al. (2007). J. Phys. Chem. C 111: 11760. Bessho, T., Zakeeruddin, S.M., Yeh, C.Y. et al. (2010). Angew. Chem. Int. Ed. 49: 6646. Yadav, S.K., Agresti, A., Pescetelli, S. et al. (2015). Proceedings of the 15th IEEE International Conference on Nanotechnology (27 July 2015), Rome, Italy, pp. 381–384. Yadav, S.K., Ravishankar, S., Pescetelli, S. et al. (2017). Phys. Chem. Chem. Phys. 19: 22546. Boschloo, G., Ha, L., and Hagfeldt, A. (2006). J. Phys. Chem. B: 13144. Long, H., Zhou, D., Zhang, M. et al. (2011). J. Phys. Chem. C 115: 14408. Zhang, K., Zhang, S., Sodeyama, K. et al. (2012). Appl. Phys Express: 5. Kusama, H., Sugihara, H., and Sayama, K. (2009). J. Phys. Chem. C 113: 20764. Boschloo, G. and Hagfeldt, A. (2003). Chem. Phys. Lett. 370: 381. Kopidakis, N., Neale, N.R., and Frank, A.J. (2006). J. Phys. Chem. B 110: 12485. Zhang, C., Huang, Y., Huo, Z. et al. (2009). J. Phys. Chem. C 113: 21779. Yu, Z., Gorlov, M., Boschloo, G., and Kloo, L. (2010). Synergistic effect of N-methylbenzimidazole and guanidinium thiocyanate on the performance of dye-sensitized solar cells based on ionic liquid electrolytes. J. Phys. Chem. C 114: 22330–22337. Jeanbourquin, X.A., Li, X., Law, C. et al. (2014). J. Am. Chem. Soc. 136: 7286.

180. Zhang, C.N., Huang, Y., Chen, S.H.

181. 182.

183.

184. 185. 186. 187. 188. 189. 190. 191. 192. 193. 194. 195. 196.

197. 198. 199.

200.

201.

et al. (2012). J. Phys. Chem. C 116: 19807. Wang, H., Bell, J., Desilvestro, J. et al. (2007). J. Phys. Chem. C: 15125. Mosconi, E., Yum, J.-H., Kessler, F. et al. (2012). J. Am. Chem. Soc. 134: 19438. Calogero, G., Calandra, P., Irrera, A. et al. (2011). Energy Environ. Sci. 4: 1838. Wang, R., Higgins, D.C., Hoque, M.A. et al. (2013). Sci. Rep. 3: 2431. Hsieh, T.-L., Chen, H.-W., Kung, C.-W. et al. (2012). J. Mater. Chem. 22: 5550. Wu, J., Tang, Z., Huang, Y. et al. (2014). J. Power Sources 257: 84. Wu, M. and Ma, T. (2012). ChemSusChem 5: 1343. Kavan, L., Yum, J.-H., and Gratzel, M. (2011). Nano Lett. 11: 5501. Yang, W., Ma, X., Xu, X. et al. (2015). J. Power Sources 282: 228. Hwang, S., Batmunkh, M., Nine, M.J. et al. (2015). ChemPhysChem 16: 53. Casaluci, S., Gemmi, M., Pellegrini, V. et al. (2016). Nanoscale 8: 5368. Yang, Z., Liu, M., Zhang, C. et al. (2013). Angew. Chem. Int. Ed. 52: 3996. Jo, Y., Cheon, J.Y., Yu, J. et al. (2012). Chem. Commun. (Camb). 48: 8057. Hong, W., Xu, Y., Lu, G. et al. (2008). Electrochem. Commun. 10: 1555. Liao, Y., Pan, K., Wang, L. et al. (2013). ACS Appl. Mater. Interfaces 5: 3663. Li, G.R., Song, J., Pan, G.L., and Gao, X.P. (2011). Energy Environ. Sci. 4: 1680. Wang, H., Wei, W., and Hu, Y.H. (2014). Top. Catal. 57: 607. Cui, X., Xu, W., Xie, Z., and Wang, Y. (2015). Electrochim. Acta 186: 125. Green, M.A., Emery, K., Hishikawa, Y. et al. (2015). Prog. Photovoltaics Res. Appl. 23: 805. Komiya, R., Fukui, A., Murofushi, N. et al. (2011). Technical digest – 21th International Photovoltaic Science and Engineering Conference (28 November-2 December 2011), Fukuoka. Kumara, G.R.A., Kawasaki, S., Jayaweera, P.V.V. et al. (2012). Thin Solid Films 520: 4119.

331

332

66 Dye-Sensitized Solar Cell 202. Brown, T., Giordano, F., Guidobaldi,

203.

204. 205.

206.

207.

208.

209.

210. 211.

212. 213. 214.

215.

216. 217.

218.

219.

A., et al. (2012). Vertical electrical connection of photoelectrochemical cells. Patent number US20130240014A1, field H01G9/2081 Serial interconnection of cells, issue date 2011-10-04. Giordano, F., Guidobaldi, A., Petrolati, E. et al. (2012). Prog. Photovoltaics Res. Appl. 21: 1653. Han, L., Fukui, A., Chiba, Y. et al. (2009). Appl. Phys. Lett. 94: 1. Fakharuddin, A., Jose, R., Brown, T.M. et al. (2014). Energy Environ. Sci. 7: 3952. Mariani, P., Vesce, L., and Di Carlo, A. (2015). Semicond. Sci. Technol. 30: 104003. Halme, J., Vahermaa, P., Miettunen, K., and Lund, P. (2010). Adv. Mater. 22: E210. Jun, Y., Son, J.H., Sohn, D., and Kang, M.G. (2008). J. Photochem. Photobiol., A 200: 314. Stergiopoulos, T., Kontos, A.G., Jiang, N. et al. (2016). Sol. Energy Mater. Sol. Cells 144: 457. Harikisun, R. and Desilvestro, H. (2011). Sol. Energy 85: 1179. Kato, N., Takeda, Y., Higuchi, K. et al. (2009). Sol. Energy Mater. Sol. Cells 93: 893. Sauvage, F. (2014). Adv. Chem. 2014: 1. Greijer, H., Lindgren, J., and Hagfeldt, A. (2001). J. Phys. Chem. B 105: 6314. Hansen, P.E., Nguyen, P.T., Krake, J. et al. (2012). Spectrochim. Acta, Part A 98: 247. Macht, B., Turrion, M., Barkschat, A. et al. (2002). Sol. Energy Mater. Sol. Cells 73: 163. Gorlov, M. and Kloo, L. (2008). Dalton Trans.: 2655. Matsui, H., Okada, K., Kitamura, T., and Tanabe, N. (2009). Sol. Energy Mater. Sol. Cells 93: 1110. Jensen, K.F., Veurman, W., Brandt, H. (2012) et al. 27th European Photovoltaic Solar Energy Conference and Exhibition , 3. Kontos, A.G., Stergiopoulos, T., Likodimos, V. et al. (2013). J. Phys. Chem. C 117: 8636.

220. Berginc, M., Topiˇc, M., and

221. 222. 223.

224.

225.

226.

227.

228.

229. 230.

231. 232.

233. 234. 235.

236.

237. 238.

239. 240.

Opara Krašovec, U. (2014). Phys. Chem. Chem. Phys. 16: 12940. Junghänel, M. and Tributsch, H. (2006). C.R. Chim. 9: 652. Thomalla, M. and Tributsch, H. (2006). C.R. Chim. 9: 659. Tuyet Nguyen, P., Rand Andersen, A., Morten Skou, E., and Lund, T. (2010). Sol. Energy Mater. Sol. Cells 94: 1582. Andersen, A.R., Halme, J., Lund, T. et al. (2011). J. Phys. Chem. C 115: 15598. H. Rensmo, S. Sdergren, L. Patthey, K. Westermark, L. Vayssieres. (1997). Chem. Phys. Lett. 4: 51–55. Rensmo, H., Lunell, S., and Siegbahn, H. (1998). J. Photochem. Photobiol. A 114: 117. Schiffmann, F., Vandevondele, J., Hutter, J. et al. (2010). PNAS 107: 4830. Flasque, M., Van Nhien, A.N., Swiatowska, J. et al. (2014). ChemPhysChem 15: 1126. Peled, E. (1979). J. Electrochem. Soc. 126: 2047. Asghar, M.I., Miettunen, K., Mastroianni, S. et al. (2012). Sol. Energy 86: 331. Joly, D., Pellejà, L., Narbey, S. et al. (2014). Sci. Rep. 4: 4033. Nguyen, H.T., Ta, H.M., and Lund, T. (2007). Sol. Energy Mater. Sol. Cells 91: 1934. Grätzel, M. (2006). C.R. Chim. 9: 578. Heo, N., Jun, Y., and Park, J.H. (2013). Sci. Rep. 3: 1712. Perganti, D., Kontos, A.G., Stergiopoulos, T. et al. (2015). Electrochim. Acta 179: 241. Gao, J., Bhagavathi Achari, M., and Kloo, L. (2014). Chem. Commun. (Camb). 50: 6249. Yun, S., Lund, P.D., and Hinsch, A. (2015). Energy Environ. Sci. 8: 3495. Olsen, E., Hagen, G., and Eric Lindquist, S. (2000). Sol. Energy Mater. Sol. Cells 63: 267. Hauch, A. and Georg, A. (2001). Electrochim. Acta 46: 3457. Ramasamy, E., Jo, C., Anthonysamy, A. et al. (2012). Chem. Mater. 24: 1575.

References 241. Radich, J.G., Dwyer, R., and Kamat, P.V. 242. 243. 244. 245.

246.

247.

248. 249. 250.

251.

252. 253. 254.

255.

(2011). J. Phys. Chem. Lett. 2: 2453. Xu, C., Wang, X., and Zhu, J. (2008). J. Phys. Chem. C 112: 19841. Huang, K.C., Wang, Y.C., Dong, R.X. et al. (2010). J. Mater. Chem. 20: 4067. Ahmad, S., Deepa, M., and Singh, S. (2007). Langmuir 23: 11430. Ha, Y.H., Nikolov, N., Pollack, S.K. et al. (2004). Adv. Funct. Mater. 14: 615. Biancardo, M., West, K., and Krebs, F.C. (2007). J. Photochem. Photobiol., A 187: 395. Wang, M., Anghel, A.M., Marsan, B. et al. (2009). Am. Chem. Soc. 131: 15976. Wang, F., Wu, C., Tan, Y. et al. (2015). J. Nanosci. Nanotechnol. 15: 1180. Guo, X., Lu, G., and Chen, J. (2015). Front. Energy Res. 3: 1. Kavan, L., Yum, J.-H., Nazeeruddin, M.K., and Gräetzel, M. (2011). ACS Nano 5: 9171. Roy-mayhew, J.D., Bozym, D.J., Punckt, C., and a Aksay, I. (2010). ACS Nano 4: 6203. Wang, H., Sun, K., Tao, F. et al. (2013). Angew. Chem. Int. Ed. 52: 9210. Nam, J.G., Park, Y.J., Kim, B.S., and Lee, J.S. (2010). Scr. Mater. 62: 148. Kavan, L., Yum, J.-H., and Graetzel, M. (2012). ACS Appl. Mater. Interfaces 4: 6999. Yun, S.N., Pu, H.H., Chen, J.H. et al. (2014). ChemSusChem 7: 442.

256. Stefik, M., Yum, J.-H., Hu, Y., and

257.

258. 259.

260.

261.

262.

263.

264.

265.

266. 267.

Grätzel, M. (2013). J. Mater. Chem. A 1: 4982. Bu, C.H., Liu, Y.M., Yu, Z.H. et al. (2013). ACS Appl. Mater. Interfaces 5: 7432. Arakawa, H., Yamaguchi, T., Okada, K. et al. (2009). Fujikura Tech. Rev.: 55. Barolo, C.B.C., Boaretto, R., Brown, T. et al. (2014). HOPV2014 book of abstract, 72. nanoGe. Sommeling, P.M., Späth, M., Smit, H.J.P. et al. (2004). J. Photochem. Photobiol., A 164: 137. Jiang, N., Sumitomo, T., Lee, T. et al. (2013). Sol. Energy Mater. Sol. Cells 119: 36. Mastroianni, S., Lanuti, A., Penna, S. et al. (2012). ChemPhysChem 13: 2925. Asghar, M.I., Miettunen, K., Halme, J. et al. (2010). Energy Environ. Sci. 3: 418. Mastroianni, S., Lanuti, A., Brown, T.M. et al. (2012). Appl. Phys. Lett. 101: 123302. Quatela, A., Agresti, A., Mastroianni, S. et al. (2012). Microelectron. Reliab. 52: 2487. Agresti, A., Pescetelli, S., Gatto, E. et al. (2015). J. Power Sources 287: 87. Agresti, A., Cinà, L., Pescetelli, S. et al. (2016). Vib. Spectrosc. 84: 106.

333

335

67 Electronic Nose: Current Status and Future Trends1) Anna Staerz, Frank Roeck, Udo Weimar, and Nicolae Barsan

67.1 Introduction to the Device

Olfaction is one of the oldest senses, enabling organisms to identify food, potential mating partners, dangers, and enemies [1]. The human olfaction system is highly complex and can be trained to (e.g. oenologists) distinguish thousands of different odors. Although instruments to identify odors had been developed earlier, it was not until the 1980s that sensor arrays were used for odor identification [2]. In 1982, Dodd and Persaud of Warwick University first compared the ability of a sensor array to differentiate odors to the mammalian olfaction system [3]. Similar work was done in 1987 by Kaneyasu et al. from Hitachi in Japan [4]. Although both groups compared the sensor’s ability to detect odors to the mammalian sense of smell, the term electronic nose (e-nose) was only coined in 1987 [2]. Of existing odor detection methods, an e-nose based on a sensor array most closely resembles the mammalian olfaction system. In this case, gas sensors are the counterparts of biological receptors. The sensors are not selective allowing for broad detection. Different odors can only be identified by comparing the signals of several sensors. Since the work of Dodd and Persaud, e-noses have been tested as a replacement for the human nose in a wide variety of applications. In addition, however, e-noses have also been used to supplement mammalian olfaction by identifying non-odorous gases. Today the term e-nose is no longer as fitting, as the devices are used primarily as compact, easy-to-use, and inexpensive alternatives to traditional analytical methods and not necessarily as replacements for human olfaction. Here all aspects of working with an e-nose will be considered: sampling procedure, detection types, applications, and data evaluation.

1) This chapter is an updated version of an article “Electronic Nose: Current Status and Future Trends” published in Chemical Reviews, Copyright (2008) American Chemical Society. Surface and Interface Science: Applications of Surface Science I, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

336

67 Electronic Nose: Current Status and Future Trends

67.2 Sample Handling

Different applications require different sampling methods. High-quality results can only be attained if the sampling method is appropriate [5]. This begins with a representative selection of samples, continues with an appropriate pretreatment, possibly including pre-concentration and separation steps, and ends with a reproducible sample delivery procedure to the detector. 67.2.1 Sampling

Although each additional step can result in statistical and systematic errors, proper sample preparation can also dramatically increase the sensitivity of the system and reduce problems caused by background interferences. E-noses were developed as simple replacements for complex analytical instruments. For this reason, complex sampling methods were initially avoided. The importance of accurate sampling has, however, become evident over the last years. Different sample types must be handled differently. For instance, the concentration of volatiles in the headspace of aqueous samples can be increased through stirring, heating, salting out, or pH variation [6]. To make the system even more sensitive and not dependent on direct vapor partitioning, a pre-concentration step is inevitable [7]. The enrichment of the analytes can be divided into two major categories: active and passive air sampling [8]. In active sampling the gaseous sample is drawn through an adsorbent material. A flow meter is used to measure the flow rate and the total volume. The advantage is that lower concentrations can be monitored in a shorter time. Passive sampling is, however, much simpler and there is no need for additional technical equipment [9]. Here, the analytes simply follow the concentration gradient according to Fick’s first law to the sorbent. Therefore, the only driving forces are diffusion and the partition coefficient between the two phases. The applicability of each method must be considered for a particular application. The combination of e-noses with different pre-concentration methods has been compared. For example, Schaller et al. analyzed the ripening grades of Swiss Emmental cheese with the help of a mass spectrometer-based electronic nose (SMart Nose) [10]. They compared several different extraction methods:

• Static headspace extraction. • Purge-and-trap extraction with a mixture of Carbosieve SIII and Carbopack B60/80 as adsorbent materials.

• Solid-phase microextraction (SPME) [11] with a 65 μm CW/DVD-coated fiber. The authors conclude that the static headspace measurement is useful for high levels of volatile compounds for which the two pre-concentration methods do not result in increased sensitivity. However, both techniques extract approximately the same class of compounds with a higher mass-to-charge (m/z) ratio. In direct comparison, the SPME method shows better repeatability, usability, and concentrating

67.2 Sample Handling

ability for medium to high molecular masses. Ampuero et al. confirmed this finding in work done on identifying the botanical origin of unifloral honeys with the same electronic nose [12]. In this study static headspace extraction and SPME were performed under similar conditions. Instead of the classical purge-and-trap technique with continuous gas flow, they used inside-needle dynamic extraction (INDEx) as the active sampling procedure. Compared with SPME this device has a higher mechanical robustness, needs half of the analysis time, and is simple to use [12, 13]. SPME, however, showed a better extraction capacity for heavier volatiles with an m/z > 110. The benefits of pre-concentration methods for sensor-based electronic noses are often not apparent from the sensor signal alone but become clear after data evaluation. Examples are the identification of lampante virgin olive oils [14], the differentiation between apple varieties, the identification of the ripeness of pineapples, and the detection of an off-flavor in sugar with an SPME/surface acoustic wave (SAW) sensor array [15]. Based on a tin dioxide multisensor, Lozano et al. tested the ability of different SPME fiber coatings for wine discrimination [16]. Particularly for quantification tasks, the influence of the coating thickness has to be considered as even low variations have a strong influence on the analyte response [17, 18]. Therefore, a lack of interfiber comparability depending on the production process used can adulterate the results. 67.2.2 Filtration and Analyte Separation

The comparison between different extraction techniques has already shown that the ratio of the detected compounds is dependent on the method. This gives the potential to deliberately increase not only the sensitivity but also the selectivity of the system by varying the sampling conditions. For example, by adapting the polymer coating of the SPME fiber of the Gerstel Twister , used for stir bar sorptive extraction (SBSE) [8, 19] or an appropriate filling for the adsorbent tubes. In addition, ingenious solutions can be found in literature for specific applications. The following examples show possible approaches and demonstrate that there are no boundaries between separation techniques with or without simultaneous sample pre-concentration:

™

• Villanueva et al. discriminated red wines, differing only in the variety of grapes, by a system based on SPME and a metal oxide sensor array [20]. In a two-step desorption process, they first “dried” the polar absorbent fiber at low temperatures in order to eliminate the influence of water and ethanol. • Instead of taking discrete temperature steps, Morris et al. desorbed the volatiles from a Tenax TA bed using a temperature program [21]. The bed had previously been exposed to the headspace of groundwater and to urban air. The temperature profile over time means that water is eluted first separately from the interesting volatiles. Instead of obtaining a steady-state sensor signal, a complex spectrum is created that contains information about the boiling point of the particular substances (elution time) and the functional type (peak width). A similar approach was previously used by Strathmann et al. [22].

®

337

338

67 Electronic Nose: Current Status and Future Trends

• Ali et al. used a heated pre-concentration tube as a dispersive element for a quartz

•

•

•

•

microbalance (QMB) array [23]. Water interferences were eliminated by using the different breakthrough times of water and toluene, the target substance. Investigating off-flavor detection in wine, Ragazzo-Sanchez et al. proposed backflush gas chromatography to remove water and ethanol from the other volatiles [24]. Off-flavor-doped wines were discriminated by using a FOX 4000 electronic nose data. The group of the University of Tübingen characterized packaging emissions with the help of four metal oxide gas sensors connected to a chromatographic column. For this purpose a very simple packed column was sufficient to separate water from the residual solvents and to determine the total amount of solvent in paper and paperboard in a reliable way [25]. The hardware of the zNose is a complete gas chromatograph with a SAW sensor as the detector. A similar approach was used by Zampolli et al. with a micromachined gas chromatographic column connected to a solid-state gas sensor. In this case the use of a single sensor means that the conventional 2D data evaluation approaches can be used. A further possibility to enhance selectivity was demonstrated using mass transport phenomena across a membrane [26]. Organophilic pervaporation can be used to discriminate wine model solutions in the presence of ethanol. After this pretreatment, both conducting polymer-based sensors [27] and metal oxide sensors [28] are able to overcome ethanol interference.

™

The presented examples demonstrate different strategies to eliminate interferences and enhance the whole e-nose system. In contrast to sensor-based improvements of the selectivity, they all result in an increase setup complexity and analysis time. But in contrast to highly selective sensors, reversibility is still maintained for most of these approaches. This has practical implications: when training the system on calibration sets, these approaches do not suffer from instrumental drift as in the case of high selectivity sensors. The stability of the system is preserved and there is no need for drift correction in the subsequent data analysis. A direct comparison of the improvements and the additional costs brought by the different sampling strategies is difficult. Each application has its own requirements and the sample preparation cannot be considered in isolation. In the examples shown, the information obtained often increases at the expense of additional time dependency. Therefore, an adapted data evaluation strategy is necessary to maximize the benefit gained.

67.3 Sensors

In this section, different chemical sensors traditionally used for e-noses will be considered. Stetter et al. simply define chemical gas sensors as “a small device that as the result of a chemical interaction or process between the analyte gas and the sensor device, transforms chemical or biochemical information of a quantitative or qualitative type into an analytically useful signal” [29].

67.3 Sensors

67.3.1 Gravimetric Sensors

Simplistically, in gravimetric sensors, the adsorbed mass of the analyte onto a microsystem induces a detectable shift in its resonance frequency [30]. There are two types of piezoelectric sensors: those based on bulk acoustic waves (BAWs) and those based on SAWs [31]. In 1880, Pierre and Jacques Curie observed that the compression of a quartz crystal produced an electric potential [31]. Acoustic waves can be generated by applying an alternating electric potential to the piezoelectric material. The characteristics of the wave are strongly dependent on the physical properties and dimensions of the material through which it travels. Changes in the acoustic wave propagation are correlated to the amount of analyte captured on the surface or bulk [31]. Selectivity can be obtained by coating the sensor surface with certain thin films [31]. 67.3.2 Chemiresistive Sensors

In the 1960s, semiconducting metal oxide (SMOX)-based sensors first became commercially available in Japan for the domestic detection of natural gases. The sensors are usually operated at temperatures between 150 and 500 ∘ C. Commercially available SMOX-based sensors consist of a heated ceramic substrate coated by either an n- or p-type metal oxide [32]. Gases react with the surface of metal oxide particles causing a detectable change in conductivity [33]. Depending on the thickness of the metal oxide layer, the devices are classified as either thin (6–100 nm) or thick (10–300 μm) film sensors. Although thin film sensors usually respond faster, most commercially available SMOX-based devices are thick film sensors due to the better production reproducibility [32]. Major manufacturers of SMOX-based sensors include AMS (Sweden), FIGARO Engineering Inc. (Japan), SGX Sensortech (Switzerland), and FIS Inc. (Japan) [34]. Chemiresistive sensors based on conducting polymers operate very similarly to those based on SMOX. The conducting polymer is coated to an insulating substrate between two electrodes and a constant voltage is applied [35]. The conducting polymer interacts with the analyte gas, resulting in a charge transfer. This can be detected as a change in resistance [36]. 67.3.3 Chemically Sensitive Field Effect Transistors (CHEMFETS)

A chemically sensitive field effect transistor (CHEMFET)-based sensor consists of a doped (either p- or n-type) silicon semiconductor with two highly doped regions (correspondingly n or p type), a silicon oxide insulation and a catalytic metal “gate.” The electric flow between the source and the drain is controlled via the gate. If no voltage is applied at the gate, then there is no electric flow between the source and the drain. If, however, a positive voltage is applied at the gate, then a positive electric field

339

340

67 Electronic Nose: Current Status and Future Trends

is induced, attracting the electrons and repelling the holes in the (for this example) p-doped bulk. The electrons in the bulk create a conductive channel for electrons to flow between the source and the gate. The electric flow between the source and drain is dependent on the voltage applied at the gate. In a CHEMFET, the gate is coated with a gas-responsive material, e.g. polymer or metal oxide. 67.3.4 Amperometric Gas Sensors

The first amperometric oxygen sensor was developed in 1956 by Clark [37]. Typically commercially available amperometric gas sensors have three electrodes: a counter, a reference, and a working electrode [38]. The target gas is either oxidized or reduced, which results in a current between the sensing and the counter electrode. The current is proportional to the concentration of the target gas. The reference electrode maintains a stable potential and is used to eliminate interferences from side reactions and increase the selectivity of the cell [39]. By using different filters, electrodes, and electrolytes, amperometric gas sensors can be made more specific. Major manufacturers of amperometric sensors include City Technology (United Kingdom), Alphasense (United Kingdom), SGX Sensortech (Switzerland), and Dräger Safety AG & Co. KGaA (Germany) [34]. 67.3.5 Optical Sensors

In optical sensors, modulation in light is the measured signal [39]. Colorimetric sensors are the simplest optical sensors. Typically, colorimetric arrays are composed of multiple colored dyes that change color as a result of their interaction with analyte gases [40]. Colorimetric sensor arrays have been inkjet printed onto paperlike substrates [41]. In addition to being easily manufactured, colorimetric sensors are inexpensive, simply operated, and portable and have a short response time [40]. An example of a colorimetric array before and after exposure to different white wine samples is shown in Figure 67.1. There are numerous other optical gas sensor operation techniques. In his overview, Bogue divided the techniques into two categories: non-absorptive and absorptive techniques [43]. Chemiluminescence is a non-absorptive optical method. It is considered the “gold standard” for gaseous NO detection [44]. NO molecules are reacted with ozone to create electronically excited NO∗2 . When the excited molecules return to their ground state, electromagnetic energy is released (between 600 and 3000 nm). This emitted light can be detected. This first device approved by the Food and Drug Administration (FDA) for the medical detection of fractional exhaled NO was based on chemiluminescence, the desktop NIOX (currently NIOX FLEX) from Aerocrine [45]. Photoionization detectors (PIDs) are typically small handheld devices that are well suited for the detection of volatile organic compounds (VOCs).

®

67.4 Classic Analytical Methods

25 mm

50 mm

Before reacted with wine sample

After three different wine reacted Figure 67.1 A colorimetric sensor array containing eight dye beads is shown before and after exposure to different white wine samples [42]. (Source: The figure is reprinted from Ref. [42].)

The PID contains a UV lamp that photoionizes gaseous organic compounds in a small cell, forming positively charged ions and electrons. The ions and electrons are propelled to electrodes, resulting in a current proportional to the gas concentration. Through variation of the UV lamp, different compound ranges can be detected [46]. For example, by using a 9.5 eV lamp, amines, benzene, and aromatic compounds are detectable. A 10.6 eV lamp additionally detects ammonia, ethanol, and acetone, whereas acetylene, formaldehyde, and methanol are only to be detected by using a 11.7 eV lamp. Major manufacturers of PIDs include Ion Science Ltd. (United Kingdom), Alphasense (United Kingdom), Baseline (United States), and Dräger Safety AG & Co. KGaA (Germany) [34]. Optical absorption techniques are based on the Beer–Lambert law and include techniques like non-dispersive infrared (NDIR), UV absorption, and photoacoustic spectroscopy. Infrared will be thoroughly explained in the classic analytical methods.

67.4 Classic Analytical Methods 67.4.1 Infrared Spectroscopy

Infrared spectroscopy (IR) can be used to detect gases that absorb in the infrared region of the electromagnetic spectrum of light. IR can also be considered as

341

342

67 Electronic Nose: Current Status and Future Trends

an electronic nose [47–49]. In a range between 4000 and 200 cm−1 , molecular vibrations and higher energy levels are excited. Through characteristic absorption bands, the type of chemical bonds can be determined, and pure chemicals can be identified by their unique fingerprint spectrum. The spectrum corresponding to mixtures is evaluated by classical e-nose algorithms. For the detection of substances in the gas phase, two affordable methods for mobile devices are established. In photoacoustic infrared spectroscopy a modulation of the intensity of an IR source causes a temperature variation, and the resulting expansion and contraction of the gas will be measured as audible frequencies with a microphone. Alternatively, the absorbed energy of a narrow bandpass infrared beam is measured in filter-based infrared spectroscopy. Commercially available devices (e.g. MIRAN SapphIRe from Thermo Scientific) are mostly used for absolute measurements of concentration either in detection of a single species that has a unique absorbance wavelength or by analysis at multiple wavelengths for a known gas mixture. However, where the constituents of the gas mixture are unknown, these instruments can also be combined with pattern recognition and used as an electronic nose. Despite confirmed feasibility [47], the infrared-based nose has not become popular, and commercially available devices like the MIRAN SapphIRe from Thermo Scientific are more generally considered portable analytic tools rather than as e-noses. 67.4.2 Mass Spectrometer

Combined with gas chromatographs (GC), mass spectrometers (MS) are often applied for lab analytics or as stand-alone devices for the identification of pure chemicals. After ionization of the compounds through thermionic emitted electrons (electron ionization) or through interaction with reagent ions (chemical ionization), the molecule ions and their fragment ions are separated according to their m/z ratio. This takes place with an electric and/or magnetic field. Today many different detection principles are used for mass spectrometry. In order to mention only a few of them, the sector instrument is the classical approach with tunable static fields, whereas the quadrupole mass analyzer consists of four parallel metal rods and filters the several ions by oscillating electrical fields. Finally, the ions collide at the electron multiplier and the current is measured. The disadvantage of all types of mass spectrometers is that they require a vacuum and are therefore not as convenient as the solid-state sensor arrays described previously; it also introduces additional costs. When used as e-noses, the system is fed with the gaseous sample without previous separation (the sample is not pretreated using chromatography). Each m/z ratio can be treated as a separate virtual sensor and analyzed by a pattern recognition algorithm [50, 51]. Despite its higher technical complexity, this approach is, in general, not better suited for odor detection when compared with the classical e-noses but has advantages for defined tasks. For example, the mass spectrometer has proved its ability to detect peptides in a higher mass range and was used for mixtures of peptide pheromones.

67.5 Application Areas

67.4.3 Ion Mobility Spectrometer

The working principle of ion mobility spectrometry (IMS) is also the filtering of ions as in the case of mass spectrometry. In IMS this is more easily realized, because the aim is not to separate the target molecules exclusively by their differences in the mass to charge ratio, but based on their different mobilities. That means that, as well as their reduced mass and their charge, the different collision cross section, determined by size and shape, has a direct influence on the ability to separate ions. Therefore, the collisions between the ions and the ambient air molecules are utilized, and the measurement can be performed under normal pressure [21]. The most common agent for ionization is a radioactive β-emitter like 63 Ni or 241 Am. After a series of ion-molecule reactions, a sample molecule with a high proton affinity reacts in humid air under proton transfer to a positively charged ion. By doping the drift gas with NH3 vapor, acetone, chlorinated solvents, or others, selectivity can be modified. Substances with electron-capturing capabilities, like halogenated compounds, can be detected by potential inversion as negative ions as well. Another often used alternative for compounds with sufficiently low ionization potential is the UV photoionization. It is appropriate for selective measurements of molecules with an ionization potential less than 8–12 eV. After ionization of the air sample, the ions are pulsed through a shutter into a drift tube, which is isolated from atmospheric air. The drift tube has a uniform weak electric field, which accelerates the ions along the tube. The movement is hindered by collisions, until the ions reach the detector at the end. Depending on the ion impact, a current is generated and measured over the time of flight. For a manageable and calibrated component amount, this gives information about the identity and concentration. If the composition is too complex, however, this often fails because of ion–ion interactions or overlapping peaks. In this case, classical e-nose data evaluation algorithms (adapted from spectroscopy) [22, 23] can be applied to gain a maximum of information out of the measurements. Compared to the mass spectrometry, the virtual sensor array is not given by discrete mass-to-charge relations, but by the signal integration over definable time intervals.

67.5 Application Areas

Over the years, e-noses have been created and tested in a vast variety of applications. In this section, a broad overview of the different application fields will be given, and selected examples will be explained in detail. 67.5.1 Food Analysis

E-noses have been used in the food industry to monitor quality, origin, manufacturing processes, and spoilage [52]. To prevent fraud, it is sometimes necessary to

343

344

67 Electronic Nose: Current Status and Future Trends

differentiate between varying product quality and origin, e.g. olive oil [53], green tea [54], and wine [55, 56]. For example, the price of green tea varies strongly with quality [54]. Yu et al. examined if the quality of green tea could be accurately determined by using a commercially available e-nose. The e-nose contained 10 different SMOX-based sensors, and the headspace of different tea samples was pumped into the sensor chamber at a constant rate of 50 ml/min. Each measurement lasted only 60 seconds. Classification of the teas using linear discriminant analysis (LDA) was very promising. In the future, it may be possible to use an e-nose instead of the costly human tasting panel to identify tea quality [54]. E-noses can also be used to monitor certain production processes. For example, the roasting of coffee is an important process step that causes physical, structural, and flavor changes. During the roasting process, the aroma profile changes [57]. Traditionally the quality of the beans is determined based on moisture measurements, weight loss, density, color, and flavor after roasting. Ideally, to ensure high quality, the roasting process should be continually monitored and automated [57]. Romani et al. determined the roasting degree of coffee beans using the PEN2 (AIRSENSE Analytics, Milano, Italy), an e-nose containing 10 SMOX-based sensors and an artificial neural network [57]. This successful evaluation marks the first step in an in-line monitoring system for the coffee bean roasting process [57]. The ripening and eventual decomposition process of fresh food (fresh fruits, vegetables, refrigerated meats, and non-frozen processed meat and eggs) is often accompanied by the development of odors. Fruits like guavas, bananas, and oranges were found to expel different gases during the ripening process [58]. Capone et al. tracked the dynamic evolution of milk rancidity using an electronic nose containing semiconductor thin film sensors and dynamic principal component analysis (PCA) [59]. Meat is an ideal growth medium for bacteria. The spoilage of different meat types has been monitored by using e-noses, e.g. beef [60], poultry [61], and pork. The spoilage of fish is particularly well researched. Different spoilage organisms dominate depending on the fish species and the storage conditions [52]. The odors produced by different microorganisms vary; spoiled marine temperaturewater fish have an offensive H2 S odor, while some tropical and freshwater types stored in air reportedly begin to smell fruity when spoiling [62]. Although the ratio between total volatile basic nitrogen (TVBN) and trimethylamine (TMS) can be used to monitor the quality of sea fish, the method is imprecise and should only be used as an orientation method [63–65]. Not only does the odor vary between different fish types and storage conditions, but also different parts of the fish show variation in the produced VOCs [66]. The shelf life of food is dependent on many different factors. If food is ideally stored, then it may still be fresh long after the expiration date; on the other hand, if storage conditions are not ideal, food deterioration may begin prematurely. In the last 10 years, consumption of fresh food has drastically increased in the United States. The National Purchase Diary) (NPD) Group predicts that fresh food consumption will continue to grow at a rate of 4% faster than population growth until 2018 [67]. The US Department of Agriculture estimates that food waste makes up between 30% and 40% of the American food supply [68]. The ability to continually

67.5 Application Areas

See

whe

n it

is r

ipe

Read the sensor Crisp Firm Juicy

™ is

Figure 67.2 Here a schematic depiction of the smart label from RipeSense shown [75]. (Source: The figure is reprinted from Ref. [75].)

monitor the freshness of food would drastically reduce the amount of waste. Today, the FOODsniffer, a small portable e-nose that can be used to determine if food has gone bad, is already on the market [69]. Using a broadband Mach–Zehnder interferometer (BB-MZI), the detector can be used to identify spoiled raw fish, pork, or beef [70]. Although the gadget has only been on the market shortly, consumers report that it properly identifies spoilage [71, 72]. In the future, ideally the freshness monitor would be directly integrated into the packaging of the food. In the last five years, the research interest in smart packaging has drastically increased [73]. Ripeness indicator labels based on colorimetric sensors already exist for fruit, e.g. ripeSense for pears and avocados [74]. The label is shown schematically in Figure 67.2 [75]. In order to be integrated directly into packaging, sensors must be inexpensive and flexible. There are already reports of inkjet-printed metal oxide-based sensors [76, 77]. Rieu et al. successfully inkjet printed SnO2 sensors onto polyimide foil [77]. In addition to being flexible, printing reduces the production costs and allows for extremely fast deposition times (a fraction of a second) [76]. Unlike the colorimetric sensors already in use for smart food packaging, however, most other gas sensors additionally require printed electronics. A great deal of research has been done on printed electronics, and although they do not have the reliability and high-performance ability of their non-printed counterparts, they could be inexpensively mass-produced [78].

™

345

346

67 Electronic Nose: Current Status and Future Trends

Another issue with the integration of non-colorimetric gas sensors into packaging is that they require a power supply. Exploratory work has been done on the integration of sensors onto radio-frequency identification (RFID) tags [79]. Simplistically, RFID systems consist of two major parts: a tag and a reader [75]. The tag can either be passive or active. Active tags have a local power source, while passive tags collect the energy from the reader. The RFID reader is a two-way radio transmitter that sends and receives a signal from the tag. Humidity sensors have successfully been integrated into RFID tags [79, 80]. In 2009, Oprea et al. successfully integrated an entire sensor platform onto flexible polyimide foil. The platform contained a Pt thermometer, electrodes, and connection pads. The interdigital electrode structures included two plane capacitive transducers, one used for sensing and one for reference. Measurements were done using chemiresistive sensors based on polyether urethane (PEUT) and polydimethylsiloxane (PDMS). By subtracting the signal of the uncoated (reference) capacitor from the signal of the sensing capacitor, it was possible to eliminate the effects of the substrate, e.g. significant sensitivity to humidity. The preliminary measurements showed promising responses to several test gases [81]. The measurement methods and the operation mode reduced the amount of recorded data and even lowered power consumption. This is the first step in the integration onto an RFID tag. At the moment, however, RFID technology is still expensive and therefore not yet viable for large-scale smart packaging. Overall, the monitoring of food products with e-noses is promising. Different processing steps can be monitored, the ripening phases observed, and decomposition identified. Although not yet widely implemented as technology advances, e-noses could very well be an important component of smart packaging in the future. 67.5.2 Environmental Comfort and Safety Monitoring Applications

Gas sensors are believed to have developed out of necessity as a result of the hazardous working environments in coal mines. As a result of the industrial revolution, coal mining drastically increased in the twentieth century. In Britain in 1900, 224 million tons of coal was mined, over 10 times as much as in 1700 [82]. Many hazardous gases can be found in mines, e.g. H2 S and methane [83, 84]. In order to address the rising death toll of miners, a compact inexpensive gas detector was needed. John Scott Haldane, a Scottish physiologist, suggested that miners take canaries with them to work [85]; if the bird stopped chirping, it was time to leave the mine. More sophisticated gas sensing methods followed: in the 1920s, Dr. Jiro Tsuji, founder of Riken Keiki, developed sensors based on light wave interference [86]; in the 1950s, Oliver Johnson patented combustion-type sensors [87]; in the 1960s, Naoyoshi Taguchi brought the first commercial SMOX-based sensor on the market; and in 2012, an e-nose for the detection of air quality in coal mines was patented [88]. Coal mining is only one of many different fields in which hazardous gases present a risk to workers. Oil refineries are another dangerous work environment. Neri et al. developed an e-nose that monitors the concentration of methane, hexane, pentane,

67.5 Application Areas

and H2 S for use in oil refineries [89]. The developed system contains five commercially available sensors from Figaro USA Inc. (TGS-825, TGS2611, TGS-6810, TGS-6812, and KE-50) [89]. Preliminary results obtained with the e-nose in a refinery showed promising results [89]. The odors produced by these industries are not only dangerous for their workers but can also be a nuisance for the cities where they are located. Recently, work has begun on e-noses that continuously monitor odors within cities. The Port of Rotterdam is the largest port in Europe, with 4 world-scale oil refineries and over 40 chemical/petrochemical companies [90]. Due to the high industrialization and specifically the large number of chemical/petrochemical companies, the odor levels increased. An average 5000–6000 odor complaints are filed yearly by residents [90]. In an attempt to reduce the odor nuisance and identify the sources, companies, municipalities, and residents joined together to form the We-Nose network. E-noses, containing four semiconductor-based sensors, were installed at various locations throughout the 10 km × 10 km port. The e-noses were trained to the residential complaints [91, 92]. In Rotterdam, over 150 e-noses are distributed throughout the port, making it possible to identify the origin of the gases. After a training period, the sensors matched over 90% of odor complaints [90]. The detected odor patterns were compared with an existing database. The placement of the e-noses within the city must consider topography and meteorology in order to gain spatial knowledge of the odors’ origin. An example of an e-nose system designed to continuously monitor environmental odors is shown in Figure 67.3. The figure also includes a map depicting where the e-noses where located in regard to water treatment plants [93]. The system helps to lower the accidental spread of emissions, and the number of odor complaints from residents has already drastically decreased [94]. Similar projects exist in the Port of Tallinn in Estonia, the Schiphol International Airport in Amsterdam, and the Port of Durban in South Africa [95].

picture it is clear how compact the environFigure 67.3 Dentoni et al. did field tests with the EOS 507 electronic nose, which was mental e-noses are [93]. (Source: The figure is reprinted from Ref. [93].) specifically developed for the continuous monitoring of environmental odors. From the

347

348

67 Electronic Nose: Current Status and Future Trends

An open sampling system, like that used for the online e-nose monitoring in the Port of Rotterdam, has the advantage of allowing the continuous measurement of odors [91]. Unlike in a closed system, however, environmental factors can only be monitored and not controlled. The dissertation of Trincavelli examines the difficulties of evaluating the results of an open system (in particular of a mobile open sensing system) [96]. Another source of uncomfortable odors for residents is local landfills. Nicolas et al. presented a simple approach to estimate the odor emission rate of a compost hall [97]. The sensor signal of a single Figaro metal oxide sensor (TGS 822) was correlated to the odor concentration measured by olfactometry. In a straightforward way the calibration was directly used to predict the malodorousness and the possible odor annoyance for the neighboring area. Knowing that volatiles and gases from other sources also cause a sensor response, Nicolas used a sensor array of six metal oxide sensors to determine time intervals when interferences occur. Using this approach, it is possible to find out when the odor predictions are reliable and when they are influenced, for example, by exhaust gases emitted from trucks or machinery or conversely when an odor-neutralizing product is sprayed in the hall. Dickert et al. monitored the composting procedure with the long-term aim to ensure ideal transformation and avoid strong smells from a very early stage [98]. With six QMB resonators, coated with different molecularly imprinted polymers, they traced four key analytes: water, 1-propanol, ethyl acetate, and limonene. The concentration pattern of the organic compounds showed strong similarities with GC–MS measurements. Thus, it is possible to determine the state and the advancement of the degradation process throughout its different phases to completion. In 2016, Lasigna et al. did infield testing of a waste management center in Italy by using two different e-nose technologies. Different odor sources were evaluated: biogas, sludge, and municipal solid waste. The results indicate that the e-nose containing SMOX-based sensors could more accurately discriminate between sources than an e-nose based on polymer/black carbon sensors. The authors conclude that ideally in the future an odor monitoring system should contain e-noses with specialized sensors for different odors [99]. Wastewater is another potential odor source. In 2009, in an attempt to reduce the odor nuisance to residents, the Pima County Regional Wastewater Reclamation Department in Arizona, USA, began to implement a system-wide odor control plan. An integral part of the system is the widespread use of e-noses from Odotech [100]. The OdoWatch system couples an e-nose containing 16 metal oxide sensors with information from a weather station. Today, numerous other wastewater plants are using the OdoWatch system, e.g. Hampton Roads S.D. in Virginia Beach, USA, and Ginestous in Toulouse, France [101]. In 2011, Mohamed et al. compared the odor fingerprints of influent and effluent water in treatment facilities in Alexandria, Egypt [102]. The water influent is the water entering the facility, while the effluent water is that after treatment. Mohamed et al. evaluated the results of the PEN3, AIRSENSE Analytics GmbH, Schwerin, Germany, to different influent and effluents samples from three treatment plants by using PCA [102]. A seasonal dependence of the components within the wastewater

™

™

™

67.5 Application Areas

was described. In addition, both the influent and the effluent samples had H2 S and NH3 [102]. These results highlight the need for monitoring systems to ensure the quality of reclaimed water. The harsh measurement environment makes the complete and continuous monitoring of wastewater treatment plants with traditional methods very difficult [103]. Guz et al. compared results attained from headspace measurements using a homemade e-nose to traditional physicochemical methods [103]. The e-nose contained eight sensors: TGS Figaro metal oxide-based sensors, a Dallas DS18B20 for temperature, and a Honeywell Analytics HIH-4000 humidity sensor. The data from the e-nose was evaluated using artificial neural networks [103]. The authors report a good correlation between the physicochemical methods and the results of the enose. The advantage of the e-nose is that it could be implemented for continuous monitoring of the wastewater treatment process [103]. In addition to effluent wastewater, it is also necessary to check freshwater sources for pollutants. In particular cyanobacteria, originating from sewage and agricultural waste, are a serious threat to potable water sources [104]. Cyanobacteria are usually blue-green algae, some of which produce toxins that are not only poisonous to animals but also to humans [104]. In 2000, Gardner et al. used a modified FOX 2000 (Alpha MOS, France) e-nose that contains six metal oxide-based sensors to continually monitor the growth of cyanobacteria for 40 days via headspace measurements [104]. The goal was to differentiate between two closely related cyanobacteria: toxin producing Microcystis aeruginosa PCC 7806 and nontoxic PCC 7941. Using MLP LVQ (linear vector quantization) and Fuzzy ARTMAP (predictive adaptive resonance theory), 100% of the unknown toxic samples were recognized [104]. The strains of bacteria were grown in laboratory conditions [104]. Another source of contamination for potable water is fecal matter, which can result in the presence of Escherichia coli. In a study similar to Gardner et al., Estefania et al. used a modified EOS835 (SACMI IMOLA Scarl, Imola, Italy) e-nose to monitor the growth of E. coli cultures in laboratory conditions via headspace measurements [105]. Nayak et al. compared samples containing water from the Swarna river during rainy season, water from the Swarna river during summer season, water after preliminary filtration (Swarna river water), and filtered to chlorinated drinking water supplied from Udupi Municipal Corporation (Swarna river water) [106]. The authors found that the e-nose gave results comparable to the traditional bacterial count in the most probable number) (MPN) index [106]. Unlike pollutants that are only a nuisance to residents, toxins released intentionally can pose a real threat. In 1988, 5000 people were killed in an Iraqi nerve attack; in 1995, a Japanese cult group released sarin on parts of the Tokyo subway [107]; and most recently there was allegedly repeated use of chemical warfare agents (CWAs) in Syria [108]. In addition to IMS, several e-nose formats are already used for field detection of CWAs, e.g. colorimetric and SAW sensors [109]. The HAZMATCAD from MSA – The Safety Company is a portable detection system based on an array of SAW sensors and electrochemical detectors [110]. The system is only 8 in. long and runs from six to nine hours in two SONY NFP-500 lithium-ion (Li-ion) rechargeable battery packs [111]. In a thorough study ordered by the US Environmental

™

349

350

67 Electronic Nose: Current Status and Future Trends

Protection Agency, the HAZMATCAD Plus almost always provided an audible and visual alarm within 3–20 seconds after exposure to hydrogen cyanide, phosgene, arsine, or Cl2 (tested background ranges: 5–35 ∘ C; 80% relative humidity) [RH]) [110]. Research has also been done on the use of e-noses containing metal oxide-based sensors for the detection of CWAs [112, 113]. In addition to the early identification of CWAs, the detection of explosives and illicit drugs is of paramount importance for civilian safety. Already in 2003, Yinon reviewed the applicability of several detection methods for explosives [114]. He predicted that e-noses would be used more and more by law enforcement and security agencies in the future [114], but today systems based on IMS remain the gold standard [115]. Mueller et al. cite the misassumption that gas sensors must be able to directly detect small amounts of gases emerging from the illicit materials as the reason why sensors are not widely used [115]. Explosives and illicit drugs are typically low vapor pressure solids. Mueller et al. show that good sensitivity and selectivity can be attained if the samples are collected in solid form and then flash evaporated [115]. Prior to vaporization, the target molecules can be electrostatically separated from lower affinity background matter by using atmospheric-pressure chemical ionization [115]. Within the EU project SNIFFER, a portable electrostatic precipitation was built into a handheld vacuum cleaner. For increased efficiency, larger fractions were discarded using centrifugal filtering before reaching the corona discharge region [115, 116]. Although Mueller et al. used the sampler with SMOX-based sensors, they acknowledge that there may be other better suited gas sensing methods [115]. The work of Mueller et al. was also funded by the EU project SNOOPY. While the objective of the SNIFFER project was broadly to capture and analyze odors from persons and illegal substances, in particular explosives, that are significant for border security, the goal of the SNOOPY project was to create a handheld e-nose to identify living hidden persons [117]. The difficulty of sniffing living humans is apparent from the work of Haedrich et al. In 2008, they examined if decomposing buried coney carcasses could be identified using the Multigas-SENSORiCCARD (JENASENSORIC e.V./MULTISENSORIC GmbH, Jena, Germany), a commercial e-nose containing three metal oxide-based gas sensors [118]. The signals were very low initially, but rose with time, and after four weeks the decomposition was detectable [118]. From the research of Haedrich et al., the difficulty of detecting living persons is apparent (the buried carcasses were only detectable once decomposition had begun). In order to detect living humans, the sensors must be sensitive to gases naturally produced by humans and not those that result from decomposition. The human body releases many different odors as a result of different aspects, for example, food intake, health, or hormonal status [119]. The human armpit is a region where a large number of glands and bacteria produce a penetrating smell and is considered one of the best places to sample body odor [119]. Wongschuk et al. used an e-nose containing five metal oxide-based Figaro sensors to analyze the armpit odors of two male subjects, right after waking up and after eight hours. By evaluating the results of the e-nose with PCA, it was possible to differentiate between the two subjects. The sampling method was, however, tedious. A cotton pad was in direct contact with the armpit for 10 minutes and then stored

®

67.5 Application Areas

Principal component analysis 4 Left armpit Right armpit Left armpit Right armpit

3 Person B

PC2 (21.7%)

2 1 0 −1 −3 −3

Person A Left (deodorant-free) Right (deodorant used) −6

−4

−2

0

2

4

6

PC1 (74.0%) Figure 67.4 The 2D-PCA of armpit odors from two persons as measured in the afternoon. The sensor response was used [119]. (Source: The figure is reprinted from Ref. [119].)

in a closed glass from which the air is sampled [119] (see Figure 67.4). From this research the difficulty of replacing search and rescue dogs with e-noses is apparent. Environmental comfort and safety monitoring is a very broad application field in which there is a wide use for sensors. In several cases e-noses have proven to be a viable option for applications where traditional analytical methods cannot be used, e.g. to monitor wastewater plants and detect the source of pollution in highly industrialized areas. In other application fields, more research is still need. For some, more sophisticated sampling methods could potentially allow more expensive analytical devices to be replaced by e-noses, e.g. detection of explosives and illicit drugs. As technology advances and cities/industries become smarter, e-noses will most likely become even more widely used. 67.5.3 Indoor and In-Cabin Air Monitoring

Most modern humans spend most of their time indoors; the average American, for example, spends 69% of their time indoors [120], making indoor air quality monitoring important. Traditionally, CO2 has been used as a marker for indoor air quality. It is not harmful to humans [121]. All humans exhale a certain amount of CO2 dependent on their age and activity level. CO2 concentrations are, however, related to human occupancy and are unaffected by the buildup of other contaminants [122]. There are many different sources of indoor air pollution, the outgassing of certain building materials or decorations, fungal contamination, and human activities [123]. In an attempt to make modern homes more energy efficient, they are often made more airtight. To prevent the buildup of indoor pollution, active ventilation is necessary [124]. Although programmable ventilation systems

351

352

67 Electronic Nose: Current Status and Future Trends

exist, most residents use the manual mode [125]. Ideally, ventilation could be automatically controlled based on the results of an e-nose. A great deal of research exists on the applicability of e-noses for monitoring indoor air quality. He et al., for example, used a homemade e-nose (QS-01 from FIS, TGS2600 and TGS2602 from FIGARO, temperature and humidity sensor SHT10 from Sensirion) with an embedded microprocessor coupled with a pattern recognition algorithm to examine indoor air quality [126]. An Xbee WiFi module from Digi, XBee S6B model, allowed for wireless connectivity to the e-nose [126]. The ability of the e-nose to detect formaldehyde in laboratory settings was examined [126]. The measurement accuracy was good for conditions under 1 ppm [126]. Formaldehyde is considered dangerous for adults over 1 ppm [127]. It is a contributor of “sick building syndrome” and is a known carcinogen [128]. Lu et al. also reported good results for low concentrations of formaldehyde using an e-nose containing 32 sensors based on doped and pristine carbon nanotubes [128]. Benzene is another organic toxin found in higher concentrations indoors versus outdoors [129]. The European directive for air quality makes the monitoring of benzene in ambient air mandatory. Within the realm of the EU project EURAMET, Spinelle et al. examined the applicability of different commercially available sensors for the detection of sub-ppb concentrations of benzene [34]. The research found that several prototypes and commercially available systems containing UV spectrometers or multiple metal oxide-based sensors are capable of detecting sub-ppb levels of benzene [34]. Specifically the prototype e-nose from De Vito et al. containing seven sensors based on metal oxides was cited. After a short neural calibration, the e-nose was able to successfully carry out infield benzene measurements for six months [130]. In the final Result In Brief of the EU project SENSINDOOR (ended December 2016), project members report the successful development of a nanotechnology-based microsystem that can sensitively detect benzene, formaldehyde, and naphthalene [131]. The system contains sensors based on SMOX and silicon carbide-based gas-sensitive field effect transistors. Pre-concentrators based on metal–organic frameworks, which absorb the target gases, were used to increase sensitivity and selectivity [131]. The low cost of the sensor systems could allow them to be installed in each room and is the first step to fully automated ventilation systems. To ensure the comfort of occupants, the monitoring of air quality in confined spaces like car cabins is important. In the 1980s, work began on the commercial sensor modules to automatically prevent exhaust pollution from entering the passenger cabin. Over the years, the sensor modules for the electronic control unit were modified [132]. The modern miniaturized module now marketed by AMS AG is based on microelectromechanical systems (MEMS) technology. The system contains two sensors, one for CO and VOCs and one for NO2 . These are the gases predominantly present in the exhaust of gasoline- and diesel-powered automobiles [133]. Work has been done on a system that monitors the in-cabin air as well as the air outside and controls the flap based on a comparison of the air quality [134]. Odor events in the passenger cabin result in poor air quality, e.g. consumption of food or cigarette smoke [134]. Promising results were attained using an array containing

67.5 Application Areas

three SMOX-based sensors and two environmental sensors (temperature and humidity) [134]. Ensuring in-cabin air quality at high altitudes is even more complex. A commercial airliner can have several hundred passengers [135]. In order to ensure passenger comfort, the air on planes is constantly circulated and renewed. At high altitudes, the partial pressure of oxygen is too low; therefore, air from outside must be pressurized before it can enter the passenger cabin. In most modern airplanes, the pressurized bleed air is taken from the engines and circulated through the cabin [135]. The air quality is not monitored before entering the passenger cabin nor is the air quality on board monitored. Recently, there have been several reports of fume incidents on planes [136]. A study done by Kansas State University found an average of 0.2 incidents of smoke/oil/fumes per 1000 flights [136]. Helwig et al. make the argument that IR-based and photoacoustic sensors may be particularly well suited to monitor the air quality on passenger airplanes. They stress that for application on a commercial airliner, sensors must be small, inexpensive, and stable over long periods of time and have a self-test capability [135]. Specific Range Solutions Ltd. is currently developing an aircraft cabin air quality monitoring system [137]. Large aerospace companies, like Boeing, have already expressed interest in reliable and accurate aircraft air quality monitoring systems. The need for cabin air quality monitoring is exacerbated in space. Due to the difficult atmospheric conditions in space, the air in the spacecraft is filtered and recycled during the mission. It is critical for the health of the shuttle occupants that the buildup of toxins is detected. In addition, the presence of microorganisms aboard the space station can also represent a serious danger to not only the crew but also the station. Both size and weight are severely limited on the spacecraft. A miniature e-nose would be ideally suited to monitor the air quality aboard a space shuttle. Already in October 1998, an e-nose weighing only 1.4 kg with a volume of 1700 cm3 was aboard the STS-95 [138]. The device was developed by the Jet Propulsion Laboratory (JPL) and the California Institute of Technology (Caltech) [139]. The e-nose was controlled by a HP 200LX Palmtop and contained a sensor array that could identify 10 different toxins [139]. The sensors were based on a conductometric polymer and a carbon sensing media developed by Caltech [138]. Over six days in-cabin air was collected for post-flight sensing. At the time of collection an alcohol wipe was opened as a marker for the e-nose measurements. A baseline measurement was done every three hours. The e-nose continually monitored the air in the cabin taking a data point every 30 seconds [138]. The e-nose data was then analyzed post-flight and compared to the results attained for the cabin samples [139]. The only events detected by the e-nose (outside of the initial alcohol marker measurements) were changes in humidity. This result is in accordance with the analysis of the collected cabin air samples, in which no target gases were detected (over the e-nose detection limit) [138]. This experiment aboard the STS-95 showed that the e-nose functions in microgravity and could be used in future space missions to monitor the atmosphere [138]. In 2003, Ryan et al. presented the second-generation e-nose at the Bioastronautics Investigators’ Workshop and Advanced Environmental Monitoring [140]. The presented prototype weighed only 800 g and contained

353

354

67 Electronic Nose: Current Status and Future Trends

optimized polymer sensors. The second-generation model was extensively tested in ground measurements at different temperatures and humidities [141]. In 2009, the third-generation e-nose was tested for six months at the International Space Station (ISS) [142]. During the test period, the e-nose continuously monitored the air of the space station for 10 contaminants. It successfully recognized gases like formaldehyde and methanol (all at harmlessly low levels). In the past, similar experiments have been done by other countries on the ISS [143]. In 2012 and 2013, an e-nose created by Airbus was tested in the Russian sector of the ISS. The e-nose was designed to monitor the microbial contamination level on the ISS. Space stations are the ideal environment for microorganisms. The contamination can be dangerous for crew members and can result in microbial destruction of the space station [144]. After 15 years of operation, hundreds of different bacterial and fungal species were found aboard the Mir space station [144]. In order to allow for proper counter-measurements and prevent contamination on the ISS, a reliable detection means for microbial contamination was needed. The e-nose had been developed since 2009 by the German Aerospace Center (DLR) [145] and has been aboard the ISS since 2009 [146]. The e-nose system contains 10 metal oxide-based sensors. It is a modified version of the Portable Electronic Nose from AIRSENSE Analytics GmbH, Schwerin, Germany [144]. Today, e-noses are successfully implemented to monitor air quality indoors and in-cabins. They are widely used to recognize exhaust odor events allowing exhaust fumes to be kept out of the passenger cabin of automobiles. In addition, an e-nose is now used in space to monitor the microbial contamination on the ISS. As home ventilation systems become fully automated, e-noses may play an important role. As a result of their small size and low cost, e-noses based on sensors could be widely distributed through the home to make sure that indoor quality is ensured through the living space. Overall e-noses appear promising as tools to monitor indoor air quality. 67.5.4 Medical Diagnostics

E-noses have been used in a wide array of biomedical applications, ranging from disease diagnostics to drug purity testing [147]. It was already known in ancient Greece that a change in the odors of bodily excretions could be indicative of a patient’s health [147]. These excretions include breath, urine, and sweat. In their review paper, Wilson et al. list many different aromas that are characteristic for different diseases, e.g. stale beer for tuberculous lymphadenitis [148]. In 1995, Parry et al. were able to identify the presence of beta-streptococcal infections in leg ulcers by analyzing the bacterial cultures of 21 patients [149]. During the metabolism, fungi and bacteria produce VOCs known as MVOCS [150]. Different bacteria or fungi produce varying MVOCS. In 1998, Gardner et al. successfully discriminated between six pathogenic bacteria: Clostridium perfringens, Proteus, Haemophilus influenzae, Bacteroides fragilis, Oxford staphylococcus, and Pseudomonas aeruginosa by analyzing the headspace of bacterial culture with an e-nose containing four commercially available metal oxide sensors [151]. In 1997, Chandiok

67.5 Application Areas

et al. used an e-nose containing 32 polymer sensors to identify bacterial vaginosis in vaginal swabs. The initial results were promising, and in 2004 Chaudry et al. monitored treatment of bacterial vaginosis by tracking the acetic acid concentration with a conducting polymer array [152, 153]. By far, the most widely tested e-nose for medical applications is the Cyranose 320. In 2000 the NASA spinoff Cyrano Sciences (since 2004 Smiths Detection–Pasadena Inc., a subsidiary of Smiths Detection Group Limited) developed the e-nose. The Cyranose 320 is a small handheld odor-detecting instrument. The device contains the patented NoseChip , an array containing 32 single-wall carbon nanotube, metallic nanoparticle, and conductive composite-based sensors. A modular plugand-play version of the Cyranose, in which different thin film nanosensors can easily be exchanged, is also available. Sensigent offers software packages with many different data analysis options. As of 2014 there were over 200 industrial and medical publications in which the Cyranose 320 was used [154]. Specifically, the e-nose has been used by several different groups to evaluate the headspace of bacteria. In 2004, Dutta et al. used the Cyranose 320 to examine the headspace of six different bacteria responsible for eye infections [155]. The authors concluded that, overall, the data analysis and feature extraction was very tedious and difficult. Using a combination of three different nonlinear methods, the performance of the Cyranose 320 was enhanced, and the bacteria classes could be predicted with 94%/98% accuracy [155]. Ideally, e-noses could be used to test samples in real time directly from the patient [148]. In 2004, swabs from ear, nose, and throat infections were immediately analyzed with Cyranose 320. Although the accuracy of the sensor array was only 88.2%, the authors marked this as the first time an e-nose was used for point-of-care testing [156]. The Cyranose 320 has also been used for a number of studies done on breath samples. Breath analysis is one of the most promising future application areas for e-noses [148]. It was already known in ancient Greece that the smell of human breath can reflect a person’s health [157]. In 1971, Cary et al. identified over 250 different VOCs in human breath and 280 different substances in urine using gas–liquid partition chromatography [158]. Today over 1000 different substances have been identified in breath. Using the Cyranose 320, De Heer et al. found that patients suffering from invasive pulmonary aspergillosis during prolonged chemotherapyinduced neutropenia had very different breath profiles than healthy controls [159]. More recently the breath fingerprints of patients suffering from chronic obstructive pulmonary disease (COPD) were recorded using the Cyranose 320 in 2011 by Fens et al. [160] and in 2015 by Shafiek et al. [161]. Using the e-nose, it was possible to discriminate the breath fingerprints of patients with COPD versus noninfected subjects with a success ratio of 75% [161]. Dragonieri et al. reported that the Cyranose 320 can distinguish between the pattern of VOCs present in the breath of lung cancer patients and that of those with COPD [162]. Already in 1985, Gordon et al. reported, based on the results of GC/MS measurements, that patients with lung cancer have specific breath patterns that vary from those of negative controls [163]. In 2001, Di Natale et al. evaluated whether an e-nose could be used to identify patients suffering from lung cancer [164]. Using an e-nose based on QMB sensors and partial

™

355

356

67 Electronic Nose: Current Status and Future Trends

least squares discriminant analysis, 100% of patients affected with lung cancer were identified [164]. Over the years, a great deal of research has been done on the applicability of e-noses for lung cancer screening [165–169]. In 2013, by using the Cyranose 320, Paff et al. found that children suffering from cystic fibrosis and primary ciliary dyskinesia have different exhaled molecular profiles from healthy controls [170]. The research done by Nakhleh et al. is by far the most comprehensive study done using an e-nose and breath analysis [171]. Breath was tested from 1404 different subjects using a sensor array containing 20 different nanomaterial-based sensors. Over half of the subjects had been diagnosed with either chronic kidney failure (CKD), idiopathic Parkinson’s disease (IPD), atypical Parkinsonism (PDISM), multiple sclerosis (MS), Crohn’s disease (CD), ulcerative colitis (UC), irritable bowel syndrome (IBS), pulmonary arterial hypertension (PAH), pre-eclampsia in pregnant women (PET), head and neck cancer (HNC), lung cancer (LC), colorectal cancer (CRC), bladder cancer (BC), kidney cancer (KC), prostate cancer (PC), gastric cancer (GC), and ovarian cancer (OC). The data evaluation was highly complex. Four sensing features were read out for each sensing response (the relative change of the sensors’ resistance at the peak [beginning], middle, and end of the exposure and the area under the curve of the whole measurement). In total, 59 sensing features were used for the statistical analysis. In total, 13 VOCs were found for all the diseases, but differed significantly in abundance between the diseases and the controls. The authors report a 86% discrimination accuracy [171]. The vast majority of the studies did not measure breath directly from the patient but instead collected the sample in tedlar bags. Accurate sampling is historically one of the largest hindrances to breath analysis for medical diagnostics. The concentration of disease biomarkers in breath is usually very low, making accurate sampling crucial. The monumental approval by FDA for the NIOX FLEX in 2003 for clinical breath testing has been attributed to the successful standardization of the measurement procedure [45]. The device from Aerocrine Inc. (Solna, Sweden) is designed to detect fraction exhaled nitric oxide (FENO ) [45]. People suffering from stable asthma are reported to have between 20 and 25 ppb FENO in breath [172]. Children are reported to have nasal NO concentrations over 400 ppb [173], and ambient NO concentrations can be in the ppb range [174]. In order to avoid inaccuracies, the breath sampling system used by Aerocrine is highly sophisticated. NO is removed from the inhalation air and only the end of exhalation is taken [175]. Today, modular samplers exist for certain gases that can be adjusted and coupled with different detection methods [176]. Righettoni et al. successfully coupled the commercially available SOFIA breath sampler with a Si-doped WO3 -based sensor to detect acetone (a biomarker for diabetes) in breath [176]. Breath analysis devices for ethanol (detection of alcohol consumption) and volatile sulfuric compounds (halitosis) are already commercially available [177, 178]. In each of these cases, the origin of the biomarker in the body is well researched (in the case of ethanol external origin). For some diseases, no single gas can be correlated but instead a unique fingerprint of different gases is indicative, e.g. COPD [179]. Although a great deal of research has been done on the identification of breathprint for COPD, their

®

67.5 Application Areas

exact origin within the body is not known. This is clear from the research of Biller et al. In order to test the efficiency of anti-inflammatory drugs, healthy subjects are exposed to ozone. This exposure is known to result in a reversible impairment of the lung function (persists between 18 and 24 hours) [180]. The authors were unable to explain why there was no change in the VOC profile, but the study highlights the necessity to understand where the disease biomarkers in breath originate [181]. Despite the large amount of research done on e-noses for breath analysis in the last decades and in particular the years of research done using the Cyranose, only one e-nose (NIOX FLEX) has been approved by the FDA for medical diagnostics. Accurate sampling is cited as a possible source of error [181] in e-nose breath measurements, and until the origin of the components found in the breathprints is not known, accurate sampling is impossible. Breath analysis is highly attractive as a painless and noninvasive method for medical diagnostics, but before its widespread application is possible, more basic research will be necessary. Until this information is available, other noninvasive methods could be useful, such as urine headspace analysis. Capelli et al. did a very comprehensive review on the analysis of urine with e-noses for medical diagnostics [182]. They divided the research into several subsections: identification of urinary tract infections [183–185], cancer diseases [186, 187], diabetes [188], and bowel diseases [182, 189]. Persaud et al. used an array of conducting polymers to detect the metabolites produced by bacteria in clinical urine samples [185]. The sampling method was simple and straightforward. The samples were sealed in 22 ml headspace vials with a polytetrafluorethylene (PTFE) rubber septum, which could be pierced by the sampling needle. After being sealed, the samples were loaded onto an autosampler and held at a constant temperature. The samples were then lowered to a preheated plate. The sample temperature is held constant for a defined period of time before a dual concentric needle is inserted into the vial through the septum. The dynamic headspace is then extracted and transferred across the array using a continuous humidified gas flow [185]. Similar methods were used in many other cases, regardless of sensing method and disease. Bernabei et al., for example, kept urine samples at 25 ∘ C for a predetermined period of time and then extracted 10 ml of headspace, which was added to a sterile bag filled with N2 . The contents of the bag were then pumped through the sensor chamber at a constant rate. The goal of the research was to identify bladder cancer by using an array containing eight sensors based on QMB [186]. The application of e-noses for medical diagnostics is one of the most intriguing fields. The ability to identify diseases via breath or urine headspace would be noninvasive and would revolutionize modern medicine. Despite the great deal of research, the application remains difficult. Odors emitted from the human body are usually found in low concentrations and in the presence of a complicated background matrix. In the future, high-responding sensors will be needed, and standardization of sampling and data evaluation methods is necessary. A great deal of testing, both in the lab and under real conditions, will be required before a device can be approved for use in medical diagnostics.

357

358

67 Electronic Nose: Current Status and Future Trends

67.6 Data Evaluation

Dodd and Persaud used the ratio of the steady-state sensor responses for data evaluation 25 years ago [3], whereas in current research the data obtained is often so complex that it cannot be manually evaluated. Furthermore, data evaluation is not only limited to pattern recognition; it begins with the data acquisition step [190]. This includes the choice of the appropriate sensors, feature selection, scaling, and normalization. Finally, pattern recognition and classification techniques can be model free or model based and supervised or unsupervised. Each of these functions can be performed by a variety of different approaches that is more or less suitable for a specific application. Unfortunately, no general guidelines to determine the appropriate strategy exist. For this reason, in several publications, these factors are a product of chance, or, if they were done more systematically, a product of trial and error. In the latter case, however, the danger of overfitting and therefore false classification is high for operators lacking deeper understanding on this field, as Goodner was able to demonstrate [191]. Additionally, the lack of knowledge on which substances may be encountered hinders an adequate selection of the sensors and the training of the array to each possible analyte. An overview of the analysis of data is given in the review article of Scott et al. [192]. Because of the need to have real experimental data, current research in this field is, in most cases, specific to the application and the e-nose used. Therefore, there is a need to compare existing pattern recognition processes on the same data set [193], to adapt and improve existing algorithms [194–196], and to transfer data evaluation methods from other research areas [197–199]. The latter is especially important for the new types of e-nose setups that produce additional time-dependent information [197, 198]. However, in handling large amounts of data, it is important to consider redundancy. As these new techniques increase the dimensions of the data set, the number of theoretical features becomes large – and hence, selection of the right features becomes challenging [51]. For e-noses based on a sensor array, these are principally:

• • • •

Transient sensor response [200–202]. Temperature modulation of metal oxide sensors [203–206]. Partial pre-separation of the compounds [21, 23, 25, 207]. Slight differences in the sensors caused by a gradient over temperature, doping concentration, sensitive layer thickness, or membrane thickness [208, 209].

Modern approaches may also have a high-dimensional output data, for example, the MS-based Smart Nose with its high amount of m/z ratios, IMS with the timedependent measurement [210], or high-density optical sensor arrays [211, 212]. However, for any given training set, there exist an optimum number of features. In case it is too high, overfitting or computational ill-conditioning will take place, and generalization will fail with the consequence of poor validation performance [213]. Therefore, a lot of work has been carried out recently in order to select the best features [213–217] or even the most appropriate sensors [218, 219]. There is

67.8 Conclusion and Outlook

progress in the direction of having solid features and consequently reliable results from data evaluation instead of fitting the noise [51].

67.7 Commercially Available E-Nose Systems

There are numerous different companies that offer commercially available e-noses. The devices contain different sensor arrays and data evaluation programs. Today, e-noses are not only implemented for the detection of odorous substances but are also used to identify gases that humans are incapable of detecting. The range of e-noses on the market spans from military, security and safety applications, food processing, and medical applications to the pharmaceutical industry and even include mass markets like automotive applications or white goods. The border between classical analytical systems, e-nose technology, and detectors for specific substance classes or even single compounds has become blurry. Some manufacturers call their devices e-nose, whereas others avoid mentioning this term even if their product operates in a similar way. The table gives an overview of e-noses on the market according to the criteria above listing their manufacturers and technology basis (see Table 67.1).

67.8 Conclusion and Outlook

In the last decades, there has been an immense amount of research done on e-noses and, specifically, on their application in a wide array of fields. Today, there are numerous companies that offer e-noses, and for certain applications e-noses are a widely used solution. They are used to identify sources of pollution in highly industrialized areas, e.g. the Port of Rotterdam or as a flap control system hindering the entrance of exhaust into the passenger cabin of a car. In both cases, an odor event must be recognized, and it is not necessary that the exact composition of the odor is identified. In both cases, the low cost and compact size of e-noses makes them advantageous over more sophisticated analytical methods. E-noses are also successfully used in highly specialized applications, e.g. to monitor microbial contamination on the ISS. By comparing the results of measurements on the ISS with the odor prints of certain microbes known to grow on space stations, it is possible to recognize contamination and take preventative results. The easy operation of the e-nose and the practically instant readout allowing for onboard data evaluation, in addition to the small size, make an e-nose ideally suited for this application. In many other applications, despite promising research results, e-noses are not widely implemented, for example, intelligent food packaging. Although smart packaging could potentially drastically reduce the amount of waste and some smart labels containing colorimetric sensors exit, smart packing containing e-noses remains uncommon. This has many reasons. Among other things, most gas sensors are still too expensive and require a power source, making their integration into disposable

359

Table 67.1

Companies manufacturing e-noses are listed.

Manufacturer

Model

Technology

Use

Altitude Tech (https://store.altitude.tech)

The Sensly HAT air quality

Indoor air quality

AirVisual (https://airvisual.com) AIRSENSE Analytics GmbH (http://www.airsense.com)

Node Portable Electronic Nose GDA

Semiconductor metal oxide (SMOX) sensor array Humidity, particle, CO2 sensor SMOX sensor Ion mobility spectrometer (IMS), photoionization detector (PID), electrochemical cell (EC) and two SMOX sensors

AIRSENSE Analytics GmbH in cooperation with Lufthansa Technik (http://www.lufthansa-technik.com) AIRSENSE Analytics GmbH in cooperation with PCA Technologies (http://www.pcatechnologies.com)

aerotracer

Hybrid sensor array

OlfoSense PEN3

SMOX sensor array, PID, and EC SMOX sensor array

Alpha MOS (http://www.alpha-mos.com) AMS Sweden

Heracles

Flash GC

Indoor Air Quality

SMOX sensor

Indoor air quality

First response (hazardous substances) Explosives (chemical and explosive identification) Fumigation (for import containers) Personal (CWAs and TICs) Reduce downtime in aviation Environmental monitoring Food and pharmaceutical industry Various Personal monitor for indoor air quality

AltraSens (http://www.altrasens.de)

OdourVectorTM

Aryballe Technologies ( http://www.aryballe-technologies.com)

NeOse

CDx Inc. (www.cdxlife.com) CSIRO (www.csiro.au) Dr. Födisch Umweltmesstechnik AG (http://www.foedisch.de)

O’cell O-sense NeOse AeroDx MyDx CannaDX CYBERNOSE MCA 12

Six gas sensors on a mass-sensitive transducer Surface plasmon resonance prism technique Biosensors Optical sensors Unknown Unknown Unknown Unknown Colorimetric

SGA 16

EC, IR photometer, and paramagnetic sensor Gas filter correlation, frequency measurement method and ZrO2 cell Gas filter correlation, frequency measurement method and ZrO2 cell Photometer

MGA

EC sensors

MCA 14

Gas filter correlation, frequency measurement method and ZrO2 cell

MCA 10

MCA 04

Environmental monitoring Various

Various Various Portable odor detector Air quality sensor analyzer Portable cannabis profile Cannabis profile Various Gas emission monitoring (cold) Gas emission monitoring (hot) Gas emission monitoring (hot) Continuous industrial measurements Mobile industrial measurements under 650 ∘ C Industrial measurements (hot) (continued overleaf )

Table 67.1

(Continued)

Manufacturer

Model

Technology

Use

Dräger Safety AG & Co. KGaA (http://www.draeger.com)

X-am 5000

Catalytic sensor and EC

Mobile safety sensor

Dräger tube

Colorimetric

Aerotest 5000/Aerotest Alpha Aerotest Simultan HP/Navy MultiTest med. Int. Simultan Test CO2

Colorimetric

Detection of fumigation agents Breath air quality in low pressure Breath air quality in high pressure Test purity of medical gases Detection of CO2 in low pressure Various

Electronic Sensor Technology (www.estcal.com) Environics Oy (http://www.environics.fi)

7100 zNose/4600 zNose/4300 zNose/4200 zNose ChemPro100i/ ChemProFXi

ChemProDM FOODSniffer (www.myfoodsniffer.com)

The FOODsniffer

Colorimetric Colorimetric Colorimetric GC/SAW (surface acoustic waves) IMS, SMOX sensor array, and temperature, humidity, pressure, and mass flow sensors Open-loop IMS plus additional sensors Ten optoelectronic sensors (LED self-aligned to a broadband Mach–Zehnder interferometer and a photodetector array)

TICs and CWAs

TICs and CWAs Meat freshness

Gerstel GmbH & Co. KG (http://www.gerstel.de) GSG Mess- und Analysengeräte GmbH (http://www.gsg-analytical.com) Honeywell Analytics (http://www.honeywellanalytics.com)

ChemSensor 4440A

Headspace GC and quadrupole MS

Routine quality control and measurements of flavors Various Industry

IAQPoint2 AirscanTM IR-F9 Sensepoint XCD

EC and catalytic bead sensor NDIR Diffusion sensor EC

Midas

Unknown

GasAlertQuattro SPM Flex Signalpoint

Unknown Unknown Polyphenylene sulfide sensor IR

MOSES II E3Point

®

Searchpoint Optima Plus ACM 150 Searchline Excel Sensepoint HT

®

Chemcassette IR-148

FTIR Open-path IR Unknown Colorimetric IR

Air comfort Refrigerant gas monitoring Flammable and toxic and oxygen deficiency Toxic and flammable gases during semiconductor production Industrial safety Industrial safety Combustible gases Explosive gases Explosive gases Combustible gases Explosive gases at high temperature Toxic gases Toxic and combustible gases (continued overleaf )

Table 67.1

(Continued)

Manufacturer

Model

Technology

Use

Indiegogo (https://www.indiegogo.com/projects/atmotubethe-portable-air-pollution-monitorenvironment#/) International Gas Detectors Ltd. (www.internationalgasdetectors.com)

Atmotube

Unknown

Personal air pollution monitor

TOC-30

IR, EC, and catalytic sensors Catalytic, IR, and EC Sensor array with eight SAW sensors and one reference sensor Unknown Unknown Unknown Unknown

Safe area gas detectors

Karlsruher Institut für Technologie (http://www.kit-technology.de)

TOCSIN 103 SAGAS

MSA (http://de.msasafety.com/landing) Odotech (www.odotech.com)

HAZMATCAD Altair SulfNoseTM MultiNoseTM

Owlstone Inc. (www.owlstonenanotech.com)

Lonestar

PCA Technologies (www.pcatechnologies.com)

ECOPROBE5 EXPLONIX

GDA

Field asymmetric ion mobility spectrometry (FAIMS) PID, 4-channel selective IR detector Infrared vapor mode and particulate sensor IMS, SMOX sensors, PID, and EC

Industrial gas sensors

CWAs Various H2 S Different environmental pollutants Various

Chemical and physical condition of soil Explosives and radioactive materials

CWAs and TICs

Pem-Tech Inc. (http://www.pem-tech.com) Proengin Inc. (http://www.proengin.com) RAE Systems by Honeywell (www.raesystems.com)

Ultra1000 Series PT605-Dual AP2C/AP4C MultiRAE AreaRAE EntryRAE IAQRAE MeshGuard MicroRAE MiniRAE ppbRAE 3000 QRAE 3 RAEGuard 2 PID RAEGuard IR RAEGuard LEL RAEGuard S ToxiRAE 3 UltraRAE VRAE

EC, catalytic bead, and IR EC and NDIR Flame spectrophotometry Catalytic bead, PID, NDIR, EC PID, combustible sensors, EC Catalytic bead, PID, EC PID, NDIR, EC Catalytic bead, EC, PID Catalytic bead, EC PID PID Catalytic bead, liquid electrolyte, EC PID NDIR Catalytic bead Smart NDIR, catalytic sensor, or EC 3-electrode pinless microsensor PID with vacuum UV lamp Catalytic sensor, thermal conductivity sensor, EC

Toxins and oxygen deficiency H2 S and CO2 Chemical contamination/CWAs Worker safety Personal detection Personal detection Indoor air quality Various Various Various Various Hazardous gases for workers Hazardous work areas Hazardous work areas Hazardous work areas Hazardous work areas CO and H2 S Industrial safety and leak detection (marine spills) Various (continued overleaf )

Table 67.1

(Continued)

Manufacturer

Model

RipeSense Limited (by Jenkins Group Limited) (www.ripesense.co.nz) SACMI

ripeSense

Sensidyne LP (http://www.sensidyne.com)

Sensigent (www.sensigent.com)

®

EOS Aroma EOS Ambiente EOS 101 Portable SensAlert ASI Point Hydrocarbons SensAir Point Gas Detector Cyranose 320

eNose Aqua

SENSIT Technologies (www.gasleaksensors.com)

®

SENSIT GOLD/SENSIT GOLD G2/SENSIT GOLD CGI SENSIT GOLD 100 TRAK-IT Illa

®

SENSIT HXG-3P/SENSIT HXG-2d

®

Technology

Use

Colorimetric

Food ripeness

6 SMOX sensors 6 SMOX sensors SMOX sensors IR, catalytic bead, and EC Dual wavelength NDIR Catalytic bead

Flavors and aromas Odorous emissions Short-term indoor Personnel safety industry

NoseChipTM Nanocomposite sensor array NoseChipTM Nanocomposite sensor array Semiconductor sensor plus optional EC Semiconductor sensor Catalytic bead sensor plus optional EC Semiconductor sensor

Hydrocarbons Various Various

Contamination in refillable water bottles Combustible leak

Gas leak portable Underground gas leaks Combustible gas leak

®

SENSIT TKX SENSIT P400 SGX Sensortech (http://www.sgxsensortech.com) Smiths Detection (www.smithsdetection.com)

The eNose Company (www.enose.nl) Vaporsens (http://www.vaporsens.com/products)

SGX-4DT MICS-4512 MICS-6814 GID-3 SABRE 5000 LCD-NEXUS MMTD

Solid-state sensor Catalytic, galvanic fuel cell, and EC Dual EC Dual SMOX sensor Triple SMOX sensor Unknown IMS Mini IMS IMS

LCD 3.2E Aeonose Aetholab Pilot 4.2 Acorn Process Monitoring (Release 2017)

IMS Unknown Unknown Unknown Unknown Unknown

Air Quality Monitoring (Release 2017)

Unknown

Gas leak Personal gas monitor Industrial toxic gas Automobile (NO2 and CO) Pollution CWAs Explosives and CWAs CWAs and TICs TICs and CWAs, narcotics, and explosives CWAs Breath analysis Microbe headspace Various Various Aging, spoilage, leak detection, and quality control Indoor air quality

368

67 Electronic Nose: Current Status and Future Trends

packing currently impossible. New sensing methods and advanced technology are required to allow the widespread integration of gas sensors into packaging. The need for entirely optimized systems, from sampling to data evaluation, becomes apparent from the use of e-noses in medical diagnostics. Despite the great excitement surrounding the application field and the vast amount of research, as of today, only the FENO monitor from Aerocrine has been approved by the FDA for use in medical diagnostics. The concentration of most bodily odors is low and part of a complicated matrix (many other gases and humidity). Regardless of whether the odor is taken from breath or sweat, the sampling method can have a dramatic effect. Before e-noses can be used for medical diagnostics, standardization of the sampling method and of the data evaluation is required. An example of how smart sampling can allow e-noses to be successfully used in a once seemingly inviable application is the detection of explosives. IMS-based detectors are very common for the detection of explosives. Mueller et al. attribute the lack of e-nose-based detectors on the low vapor pressure of the samples. By collecting the low vapor samples as solids and then flash evaporating them, detection was possible using an e-nose. Due to the small size and low cost of e-noses, interest in their use will remain high in the future. In addition to optimizing sensing material, the standardization and modernization of sampling methods will be required before e-noses can be used for certain applications.

References 1. Sarafoleanu, C., Mella, C., Georgescu,

2.

3.

4.

5.

M., and Perederco, C. (2009). The importance of the olfactory sense in the human behavior and evolution. J. Med. Life 2: 196–198. Gardner, J.W. and Bartlett, P.N. (1994). A brief history of electronic noses. Sens. Actuators, B 19: 211–220. Persaud, K. and Dodd, G. (1982). Analysis of discrimination mechanisms in the mammalian olfactory system using a model nose. Nature 299: 352–355. Kaneyasu, M., Ikegami, A., Arima, H., and Iwanaga, S. (1987). Smell identification using a thick-film hybrid gas sensor. IEEE Trans. Compon. Hybrids Manuf. Technol. 10: 267–273. https:// doi.org/10.1109/TCHMT.1987.1134730. Capelli, L., Sironi, S., and del Rosso, R. (2013). Odor sampling: techniques and strategies for the estimation of odor emission rates from different source types. Sensors (Switzerland) 13: 938–955. https://doi.org/10.3390/ s130100938.

6. Friant, S.L. and Suffet, I.H. (1979).

7.

8.

9.

10.

Interactive effects of temperature, salt concentration, and pH on head space analysis for isolating volatile trace organics in aqueous environmental samples. Anal. Chem. 51: 2167. Landy, P., Mielle, P., and Ray, J.L. (2002). Proceedings of International Symposium Olfaction & Electronic Noses (ISOEN), Rome, Italy. Rome: Aracne, p. 19. Baltussen, E., Cramers, C.A., and Sandra, P.J.F. (2002). Sorptive sample preparation – a review. Anal. Bioanal. Chem. 373: 3–22. https://doi.org/10 .1007/s00216-002-1266-2. Seethapathy, S., Górecki, T., and Li, X. (2008). Review: passive sampling in environmental analysis. J. Chromatogr. A 1184: 234–253. https://doi.org/10 .1016/j.chroma.2007.07.070. Schaller, E., Zenhäusern, S., Zesiger, T. et al. (2000). Use of preconcentration techniques applied to a MS-based

References

11.

12.

13.

14.

15.

16.

17.

18.

“Electronic Nose”. Analusis 28: 743–749. Ouyang, G. and Pawliszyn, J. (2006). Recent developments in SPME for onsite analysis and monitoring. Trends Anal. Chem. 25: 692–703. https://doi .org/10.1016/j.trac.2006.05.005. Ampuero, S., Bogdanov, S., and Bosset, J.-O. (2004). Classification of unifloral honeys with an MS-based electronic nose using different sampling modes: SHS, SPME and INDEX. Eur. Food Res. Technol. 218: 198–207. https://doi.org/ 10.1007/s00217-003-0834-9. Bosset, J.O., Pillonel, L., Altieri, D., and Tabacchi, R. (2004). Comparison of efficiency and stability of two preconcentration techniques (SPME and INDEx) coupled to an MS-based “electronic nose”. Mitt. Leb. Hyg. 95: 85–98. García-González, D.L., Barie, N., Rapp, M., and Aparicio, R. (2004). Analysis of virgin olive oil volatiles by a novel electronic nose based on a miniaturized SAW sensor array coupled with SPME enhanced headspace enrichment. J. Agric. Food Chem. 52: 7475–7479. Barié, N., Bücking, M., and Rapp, M. (2006). A novel electronic nose based on miniaturized SAW sensor arrays coupled with SPME enhanced headspace-analysis and its use for rapid determination of volatile organic compounds in food quality monitoring. Sens. Actuators, B 114: 482–488. https://doi.org/10.1016/j.snb.2005.06 .051. Lozano, J., Santos, J.P., Aleixandre, M. et al. (2006). Identification of typical wine aromas by means of an electronic nose. IEEE Sens. J. 6: 173–178. Shirey, R. (2004). Improvements in SPME Reproducibility Through Enhancements in the Fiber Manufacturing Process, Report 22. 7–8. Paschke, A. and Popp, P. (2004). Diffusion-based calibration for solidphase microextraction of benzene, toluene, ethylbenzene, p-xylene and chlorobenzenes from aqueous samples. J. Chromatogr. A 1025: 11–16. https:// doi.org/10.1016/j.chroma.2003.08.059.

19. Baltussen, E., Sandra, P., David, F., and

20.

21.

22.

23.

24.

25.

26.

27.

Cramers, C. (1999). Stir bar sorptive extraction (SBSE), a novel extraction technique for aqueous samples: theory and principles. J. Microcolumn Sep. 11: 737–747. Villanueva, S., Vegas, A., Fernandez-Escudero, J.A. et al. (2006). SPME coupled to an array of MOS sensors reduction of the interferences caused by water and ethanol during the analysis of red wines. Sens. Actuators, B 120: 278–287. https://doi.org/10.1016/j .snb.2006.02.042. Morris, L., Caruana, D.J., and Williams, D.E. (2002). Simple system for part-perbillion-level volatile organic compound analysis in groundwater and urban air. Meas. Sci. Technol. 13: 603–612. Strathmann, S., Hahn, S., and Weimar, U. (2001). Proceedings of ISOEN 2001, Washington, USA, 65–75, 1-56677-3210. Ali, Z., James, D., O’Hare, W.T. et al. (2003). Gas-phase pre-concentration for a quartz crystal microbalance based electronic nose. J. Therm. Anal. Calorim. 71: 163–171. Ragazzo-Sanchez, J.A., Chalier, P., and Ghommidh, C. (2005). Coupling gas chromatography and electronic nose for dehydration and desalcoholization of alcoholized beverages: Application to off-flavour detection in wine. Sens. Actuators, B 106: 253–257. https://doi.org/10.1016/j.snb.2004 .08.006. Röck, F., Gurlo, A., and Weimar, U. (2005). Multisensor system for characterization of packaging emissions: prediction of total solvent amount and odor scores. Sens. Actuators, B. 77: 2762–2769. https://doi.org/10.1016/j .snb.2004.05.040. Schäfer, T., Vital, J., and Crespo, J.G. (2004). Coupled pervaporation/mass spectrometry for investigating membrane mass transport phenomena. J. Membr. Sci. 241: 197–205. https://doi .org/10.1016/j.memsci.2004.05.014. Pinheiro, C., Schäfer, T., and Crespo, J.G. (2005). Direct integration of pervaporation as a sample preparation

369

370

67 Electronic Nose: Current Status and Future Trends

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

method for a dedicated “electronic nose”. Anal. Chem. 77: 4927–4935. Schäfer, T., Serrano-Santos, M.B., Rocchi, S., and Fuoco, R. (2006). Pervaporation membrane separation process for enhancing the selectivity of an artificial olfactory system (“electronic nose”). Anal. Bioanal. Chem. 384: 860–866. https://doi.org/10.1007/ s00216-005-0259-3. Stetter, J.R., Penrose, W.R., and Yao, S. (2003). Sensors, chemical sensors, electrochemical sensors, and ECS. J. Electrochem. Soc. 150: 11–16. https:// doi.org/10.1149/1.1539051. Fanget, S., Hentz, S., Puget, P. et al. (2011). Gas sensors based on gravimetric detection – a review. Sens. Actuators, B 160: 804–821. https://doi.org/10.1016/j.snb.2011 .08.066. Chang, S.M., Muramatsu, H., Nakamura, C., and Miyake, J. (2000). Principle and applications of piezoelectric crystal sensors. Mater. Sci. Eng., C 12: 111–123. https://doi.org/10.1016/ S0928-4931( 00)00167-3. Schaller, E., Bosset, J.O., and Escher, F. (1998). “Electronic noses” and their application to food. LWT Food Sci. Technol. 31: 305–316. https://doi.org/10 .1006/fstl.1998.0376. Wilson, A.D. and Baietto, M. (2009). Applications and advances in electronic-nose technologies. Sensors 9: 5099–5148. https://doi .org/10.3390/s90705099. Laurent, S., Michel, G., Gertjan, K., and Tilman, S. (2015). Sensitivity of VOC sensors for air quality monitoring within the EURAMET key-VOC project. Fourth Scientific Meeting EuNetAir. pp. 10–13. doi:https://doi .org/10.5162/4EuNetAir2015/02. Janata, J. and Josowicz, M. (2003). Conducting polymers in electronic chemical sensors. Nat. Mater. 2: 19–24. https:// doi.org/10.1038/nmat768. Bai, H. and Shi, G. (2007). Gas sensors based on conducting polymers. Sensors 7: 267–307. https://doi.org/10.3390/ s7030267. Clark, L.C. Jr., (1956). Monitor and control of blood and tissue oxygen

38.

39.

40.

41.

42.

43.

44.

45.

46.

47.

tensions. Trans. Am. Soc. Artif. Intern. Organs 2 (1): 41–48. Warburton, P.R., Pagano, M.P., Hoover, R. et al. (1998). Amperometric gas sensor response times. Anal. Chem. 70: 998–1006. https://doi.org/10.1021/ ac970644y. Patel, H.K. (2014). The Electronic Nose: Artificial Olfaction Technology. Springer. Kangas, M.J., Burks, R.M., Atwater, J. et al. (2016). Colorimetric sensor arrays for the detection and identification of chemical weapons and explosives. Crit. Rev. Anal. Chem. 47: 1–16. https://doi .org/10.1080/10408347.2016.1233805. Soga, T., Jimbo, Y., Suzuki, K., and Citterio, D. (2013). Inkjet-printed paper-based colorimetric sensor array for the discrimination of volatile primary amines. Anal. Chem. 85: 8973–8978. https://doi.org/10.1021/ ac402070z. Chung, S., Park, T., Park, S. et al. (2015). Colorimetric sensor array for white wine tasting. Sensors 15: 18197–18208. https://doi.org/10.3390/ s150818197. Bogue, R. (2015). Detecting gases with light: a review of optical gas sensor technologies. Sens. Rev. 35: 133–140. https://doi.org/10.1108/SR-09-2014696. Maniscalco, M., Vitale, C., Vatrella, A. et al. (2016). Fractional exhaled nitric oxide-measuring devices: technology update. Med. Devices Evid. Res. 9: 151–160. https://doi.org/10.2147/ MDER.S91201. Paschke, K.M., Mashir, A., and Dweik, R.A. (2010). Clinical applications of breath testing. Med. Rep. 2: 1–6. https://doi.org/10.3410/M2-56. RAE Systems (2013). The PID Handbook: Theory and Applications of Direct-Reading Photoionization Detectors. Honeywell. van Kempen, T.A.T.G., Powers, W.J., and Sutton, A.L. (2002). Technical note: Fourier transform infrared (FTIR) spectroscopy as an optical nose for predicting odor sensation. J. Anim. Sci. 80: 1524–1527.

References 48. Cynkar, W., Cozzolino, D., Dambergs,

49.

50.

51.

52.

53.

54.

55.

56.

B. et al. (2007). Feasibility study on the use of a head space mass spectrometry electronic nose (MS e_nose) to monitor red wine spoilage induced by Brettanomyces yeast. Sens. Actuators, B 124: 167–171. https://doi.org/10.1016/j .snb.2006.12.017. Armenta, S., Coelho, N.M.M., Roda, R. et al. (2006). Seafood freshness determination through vapour phase Fourier transform infrared spectroscopy. Anal. Chim. Acta 580: 216–222. https://doi .org/10.1016/j.aca.2006.07.070. Hálová, J., Strouf, O., Zák, P. et al. (1998). Short communication QSAR of catechol analogs against malignant melanoma using fingerprint descriptors. Quant. Struct.-Act. Relat. 17: 37–39. Hájek, P. and Havránek, T. (1978). Mechanizing Hypothesis Formation: Mathematical Foundations for a General Theory. Berlin, Heidelberg, New York: Springer-Verlag. Röck, F., Barsan, N., and Weimar, U. (2008). Electronic nose: current status and future trends. Chem. Rev. 108: 705–725. https://doi.org/10.1021/ cr068121q. Cosio, M.S., Ballabio, D., Benedetti, S., and Gigliotti, C. (2006). Geographical origin and authentication of extra virgin olive oils by an electronic nose in combination with artificial neural networks. Anal. Chim. Acta 567: 202–210. https://doi.org/10.1016/j.aca.2006.03 .035. Yu, H., Wang, J., Zhang, H. et al. (2008). Identification of green tea grade using different feature of response signal from E-nose sensors. Sens. Actuators, B 128: 455–461. https://doi.org/ 10.1016/j.snb.2007.07.048. Cynkar, W., Dambergs, R., Smith, P., and Cozzolino, D. (2010). Classification of Tempranillo wines according to geographic origin: combination of mass spectrometry based electronic nose and chemometrics. Anal. Chim. Acta 660: 227–231. https://doi.org/10.1016/j.aca .2009.09.030. Berna, A.Z., Trowell, S., Clifford, D. et al. (2009). Geographical origin of Sauvignon Blanc wines predicted by

57.

58.

59.

60.

61.

62.

63.

64.

mass spectrometry and metal oxide based electronic nose. Anal. Chim. Acta 648: 146–152. https://doi.org/10.1016/j .aca.2009.06.056. Romani, S., Cevoli, C., Fabbri, A. et al. (2012). Evaluation of coffee roasting degree by using electronic nose and artificial neural network for off-line quality control. J. Food Sci. 77: 960–965. https://doi.org/10.1111/j .1750-3841.2012.02851.x. Kanade, A. and Shaligram, A.D. (2014). Development of an E-nose using metal oxide semiconductor sensors for the classification of climacteric fruits. Int. J. Sci. Eng. Res. 5: 467–472. Capone, S., Epifani, M., Quaranta, F. et al. (2001). Monitoring of rancidity of milk by means of an electronic nose and a dynamic PCA analysis. Sens. Actuators, B 78: 174–179. https://doi .org/10.1016/S0925-4005( 01)00809-7. Panigrahi, S., Balasubramanian, S., Gu, H. et al. (2006). Neural-networkintegrated electronic nose system for identification of spoiled beef. LWT Food Sci. Technol. 39: 135–145. https:// doi.org/10.1016/j.lwt.2005.01.002. Timsorn, K., Thoopboochagorn, T., Lertwattanasakul, N., and Wongchoosuk, C. (2016). Evaluation of bacterial population on chicken meats using a briefcase electronic nose. Biosyst. Eng. 151: 116–125. https://doi .org/10.1016/j.biosystemseng.2016.09 .005. Gram, L. and Dalgaard, P. (2002). Fish spoilage bacteria – problems and solutions. Environ. Biotechnol. 13: 262–266. https://doi.org/10.1016/ S0958-1669( 02)00309-9. Duflos, G., Coin, V.M., Cornu, M. et al. (2006). Determination of volatile compounds to characterize fish spoilage using headspace/mass spectrometry and solid-phase microextraction/gas chromatography/mass spectrometry. J. Sci. Food Agric. 86: 600–611. https:// doi.org/10.1002/jsfa.2386. Ruiz-Capillas, C., Gillyon, C.M., and Horner, W.F.A. (2001). Determination of different volatile base components as quality control indices in fish by official

371

372

67 Electronic Nose: Current Status and Future Trends

65.

66.

67.

68.

69.

70.

71.

72.

73.

74.

75.

76.

methods and flow injection analysis. J. Food Biochem. 25: 541–553. Rodríguez-Jérez, J.J., Hernández-Herrero, M.M., and Roig-Sagués, A.X. (2000). New methods to determine fish freshness in research and industry. CIHEAM Glob. Qual. Assesment Mediterr. Aquac. 51: 63–69. http://om.ciheam.org/article .php?IDPDF=600292. Mansur, M.A., Bhadra, A., and Takamura, H. (2003). Volatile flavor compounds of some sea fish and prawn species. Fish. Sci. 69: 846–866. The NPD-Group (2014). Where Is Fresh Headed? An excerpt from the Future of Eating Report. United States Department of Agriculture (2011). The U.S. Food Waste Challenge, 24–27. The United States Government https://www.usda.gov/ foodlossandwaste. (2017). FOODsniffer – a smart kitchen tool to enjoy meals safely 1–7. www .myfoodsniffer.com (accessed 1 February 2017). Raptis, I., Gogolides, E., and Tserepi, A. (2012). FOODSNIFFER - EU, FP7-ICT, STREP - Contract No 318319. O’Brien, C. (2015). Review: Food Sniffer. https://steamykitchen.com/ 42865-foodsniffer-review-giveaway.html. Cliff, D., Moul, R., and O’Brien, C. (2015) https://www.irishtimes.com/ business/technology/review-food-sniffer1.2091257). Review: food sniffer. The Irish Times: 1–10. Vanderroost, M., Ragaert, P., Devlieghere, F., and De Meulenaer, B. (2014). Intelligent food packaging: the next generation. Trends Food Sci. Technol. 39: 47–62. https://doi.org/10 .1016/j.tifs.2014.06.009. Yam, K.L. and Lee, D.S. (2012). Emerging Food Packaging Technologies: Principles and Practice. Cambridge: Woodhead Publishing. Fuertes, G., Soto, I., Carrasco, R. et al. (2016). Intelligent packaging systems: sensors and nanosensors to monitor food quality and safety. J. Sens. 2016 https://doi.org/10.1155/2016/4046061. Kukkola, J., Mohl, M., Leino, A.-R. et al. (2012). Inkjet-printed gas sensors:

77.

78.

79.

80.

81.

82.

83.

84.

85.

86.

metal decorated WO3 nanoparticles and their gas sensing properties. J. Mater. Chem. 22: 17878–17886. https:// doi.org/10.1039/c2jm32499g. Rieu, M., Camara, M., Tournier, G. et al. (2015). Inkjet printed SnO2 gas sensor on plastic substrate. Procedia Eng. 120: 75–78. https://doi.org/10 .1016/j.proeng.2015.08.569. Rosa, P., Câmara, A., and Gouveia, C. (2015). The potential of printed electronics and personal fabrication in driving the internet of things. Open J. Internet Things 1: 16–36. Kang, H., Park, H., Park, Y. et al. (2014). Fully roll-to-roll gravure printable wireless (13.56 MHz) sensorsignage tags for smart packaging. Sci. Rep. 4: 5387. https://doi.org/10.1038/ srep05387. Deng, F., He, Y., Zhang, C., and Feng, W. (2014). A CMOS humidity sensor for passive RFID sensing applications. Sensors (Switzerland) 14: 8728–8739. https://doi.org/10.3390/s140508728. Oprea, A., Courbat, J., Bârsan, N. et al. (2009). Temperature, humidity and gas sensors integrated on plastic foil for low power applications. Sens. Actuators, B 140: 227–232. https://doi.org/10 .1016/j.snb.2009.04.019. Coal Mines Industrial Revolution, CB Media. (2010) 1–5. http:// industrialrevolution.org.uk/coal-minesindustrial-revolution (accessed 22 December 2016). Srivastava, S.K. (2015). Real Time Monitoring System for Mine Safety Using Wireless Sensor Network (MultiGas Detector). National Institute of Technology Rourkela. Harvey, T.J.M., Cory, S., Kizil, M.S., and Gillies, A.D.S. (1998). Mining through H2S seam gas zones in underground coal mines. Proc. Counc. Min. Met. Congr., Toronto, pp. 1–10. Sekhar, K.C. and Rao, S.S.C.C. (2014). John Scott Haldane: the father of oxygen therapy. Indian J. Anaesth. 58: 350–352. https://doi.org/10.4103/00195049.135087. Kleven, B. (2016). A Summary of Gas Detection, 1–12. Advanced Calibration Designs http://www.goacd.com.

References 87. Johnson, O. (1957). Gas detection appa88.

89.

90.

91.

92.

93.

94.

95.

96.

97.

98.

99.

ratus. US 2786350. Electronic nose for detection of underground coal mine air quality. CN103592358A (2014). https://patents .google.com/patent/CN103592358A/en. Pace, C., Khalaf, W., Latino, M. et al. (2012). E-nose development for safety monitoring applications in refinery environment. Procedia Eng. 47: 1267–1270. https://doi.org/10.1016/ j.proeng.2012.09.384. Milan, B., Bootsma, S., and Bilsen, I. (2012). Advances in odour monitoring with E-noses in the Port of Rotterdam. Chem. Eng. Trans. 30: 145–150. https:// doi.org/10.3303/CET1230025. Bootma, S. and Milan, B. (2010). Odour monitoring with e-noses in the Port of Rotterdam. Chem. Eng. Trans. 23: 147–152. Gilbert, R.O. (1987). Statistical Methods for Environmental Pollution Monitoring. Wiley. Dentoni, L., Capelli, L., Sironi, S. et al. (2012). Development of an electronic nose for environmental odour monitoring. Sensors (Switzerland) 12: 14363–14381. https://doi.org/10.3390/ s121114363. We-nose, (n.d.). 1–2. https://www .portofrotterdam.com/en/our-port/ourthemes/a-safe-port/e-noses-for-a-safeport (accessed 23 February 2017). Common Invent, (n.d.) 4–7. http:// www.comon-invent.com/projects (accessed 18 March 2017). Trincavelli, M. (2011). Gas discrimination for mobile robots. KI - Künstliche Intelligenz 25: 351–354. https://doi.org/ 10.1007/s13218-011-0104-0. Nicolas, J., Romain, A.-C., and Ledent, C. (2006). The electronic nose as a warning device of the odour emergence in a compost hall. Sens. Actuators, B 116: 95–99. https://doi.org/10.1016/j .snb.2005.11.085. Dickert, F.L., Lieberzeit, P.A., Achatz, P. et al. (2004). QCM array for on-linemonitoring of composting procedures. Analyst 129: 432–437. Giungato, P., de Gennaro, G., Barbieri, P. et al. (2016). Improving recognition of odors in a waste management plant

100.

101.

102.

103.

104.

105.

106.

107.

108.

109.

by using electronic noses with different technologies, gas chromatographymass spectrometry/olfactometry and dynamic olfactometry. J. Clean. Prod. 133: 1395–1402. https://doi.org/10 .1016/j.jclepro.2016.05.148. Odor Control - Pima County (2015). 6500. http://webcms.pima.gov/ government/wastewaterreclamation/ odor_control (accessed 18 March 2017). Page, T. (n.d.). OdoWatch: Real-time Odor Emission and Impact Monitoring Using Electronic Noses, www.Odotech .com (accessed 2 April 2019). Mohamed, E.I., Taha, M.M., Zaki, S.M. et al. (2014). Electronic noses for surveillance of wastewater from secondary treatment plants in Alexandria, Egypt, 2014 World Congr. Adv. Civ., Emviromental, Mater. Res., Busan, Korea, 2014. Guz, Ł., Łagód, G., Jaromin-Gle´n, K. et al. (2015). Application of gas sensor arrays in assessment of wastewater purification effects. Sensors (Switzerland) 15: 1–21. https://doi .org/10.3390/s150100001. Gardner, J.W., Shin, H.W., Hines, E.L., and Dow, C.S. (2000). An electronic nose system for monitoring the quality of potable water. Sens. Actuators, B 69: 336–341. https://doi.org/10.1016/ S0925-4005( 00)00482-2. Estefanía, A., Carmona, N., Carmona Núñez, E. et al. (2014). Detection of microbial contamination in potable water by Nanowire technology. Proc. 8th Int. Conf. Sens. Technol. pp. 2–4. Nayak, K., Supreetha, B.S., Deccaraman, M., and Nayak, V. (2012). E-nose system to detect E. coli in drinking water of Udupi district. Int. J. Eng. Res. Dev. 1: 58–64. Ganesan, K., Raza, S.K., and Vijayaraghavan, R. (2010). Chemical warfare agents. J. Pharm. BioAllied Sci. 2: 166–178. https://doi.org/10.1016/ S1522-8401( 02)90036-4. Schmitt, E. (2016). ISIS used chemical arms at least 52 times in Syria and Iraq report says. New York Times. 19–22. Cliff, D., Moul, R., and Jugieux, A. (2013). Chemical Weapons Detection :

373

374

67 Electronic Nose: Current Status and Future Trends

110.

111.

Inspecting Syria. Vetic Brief https:// www.files.ethz.ch/isn/169046/VB22.pdf . Hofarce, K., Derringer, T., Folsom, D. et al. (2004). Environmental Technology Verification Report. Cajigas, J.C., Longworth, T., Davis, N., Ong, K.Y. (2003). Testing of HAZMATCAD detectors against chemical warfare agents: summary report of evaluation performed at soldier biological and chemical command (SBCCOM), EDGEWOOD Chem. Biol. Cent. ABERDEEN PROVING Gr. MD. Olguin, C., Laguarda-Miro, N., Pascual, L. et al. (2014). An electronic nose for the detection of Sarin, Soman and Tabun mimics and interfering agents. Sens. Actuators, B 202: 31–37. https:// doi.org/10.1016/j.snb.2014.05.060. Lee, S.C., Choi, H.Y., Lee, S.J. et al. (2009). The development of SnO2 -based recoverable gas sensors for the detection of DMMP. Sens. Actuators, B 137: 239–245. https://doi.org/10.1016/j.snb .2008.12.051. Yinon, J. (2003). Detection of explosives by electronic noses. Anal. Chem. 75: 98 A–105 A. https://doi.org/10.1021/ ac0312460. Mueller, G., Hackner, A., Beer, S., and Goebel, J. (2016). Solid-state gas sensors: sensor system challenges in the civil security domain. Materials (Basel) 9: 1–30. https://doi.org/10 .3390/ma9010065. Beer, S., Müller, G., and Wöllenstein, J. (2012). Development and characterization of an electrostatic particle sampling system for the selective collection of trace explosives. Talanta 89: 441–447. https://doi.org/10.1016/j .talanta.2011.12.059. SNOOPY Project (2011). Novel sensors increase air cargo security (2014–2016). 3–4. https://cordis.europa.eu/project/ rcn/111313/brief/en (accessed 20 May 2019). Hädrich, C., Ortmann, C., Reisch, R. et al. (2010). An electronic bodytracking dog? Int. J. Legal Med. 124: 43–47. https://doi.org/10.1007/s00414009-0338-7. Wongchoosuk, C., Lutz, M., and Kerdcharoen, T. (2009). Detection

120.

™

112.

113.

114.

115.

116.

117.

118.

119.

121.

122.

123.

124.

125.

126.

127.

128.

129.

and classification of human body odor using an electronic nose. Sensors 9: 7234–7249. https://doi.org/10.3390/ s90907234. Hodgson, A.T., Nabinger, S.J., and Persily, A.K. (2004). Volatile organic compound concentrations and emission rates measured over one year in a new manufactured house. Indoor Environ.: 1–34. https://doi.org/10.2172/838617. Schell, M. and Int-Hout, D. (2001). Demand control ventilation using CO2 . ASHRAE J. 43: 18–29. Schell, M.B., Turner, S.C., and Shim, R.O. (1998). Application of CO2 -based demand-controlled ventilation using ASHRAE standard 62: optimizing energy use and ventilation. ASHRAE Trans. 104: 1213–1225. Sironi, S., Eusebio, L., Capelli, L. et al. (2014). Use of an electronic nose for indoor air quality monitoring. Chem. Eng. Trans. 40: 73–78. https://doi.org/ 10.3303/CET1440013. Derbez, M., Berthineau, B., Cochet, V. et al. (2014). Indoor air quality and comfort in seven newly built, energy-efficient houses in France. Build. Environ. 72: 173–187. https://doi.org/ 10.1016/j.buildenv.2013.10.017. Furchgott, R. (2016). Smart doesn’t always mean an easy home. https:// www.nytimes.com/2016/07/10/ realestate/smart-doesnt-always-meanan-easy-home.html. He, J., Xu, L., Wang, P., and Wang, Q. (2016). A high precise E-nose for daily indoor air quality monitoring in living environment. Integr. VLSI J.: 1–10. https://doi.org/10.1016/j.vlsi.2016.12 .010. European Commission (2004). The INDEX project (Critical Appraisal of the Setting and Implementation of Indoor Exposure Limits ind the EU): Summary on recommendations and management options. Lu, Y., Meyyappan, M., and Li, J. (2011). A carbon-nanotube-based sensor array for formaldehyde detection. Nanotechnology 22: 55502. https://doi .org/10.1088/0957-4484/22/5/055502. Health Canada (2013). Benzene in indoor air. http://www.hc-sc.gc.ca/

References

130.

131. 132.

133.

134.

135.

136.

137.

138.

139.

140.

ewh-semt/alt_formats/pdf/pubs/air/ benzene_fs-fi/benzene_fs-fi-eng.pdf (accessed 4 April 2019). De Vito, S., Massera, E., Piga, M. et al. (2008). On field calibration of an electronic nose for benzene estimation in an urban pollution monitoring scenario. Sens. Actuators, B 129: 750–757. https://doi.org/10.1016/j.snb.2007.09 .060. SENSINDOOR Result In Brief, 2017. Kalman, E.-L., Winquist, F., Löfvendahl, A. et al. (2002). Indoor Built Environ. 11: 105–110. https://doi.org/10.1159/ 000064246. Barsan, N., Gauglitz, G., Oprea, A. et al. (2016). Ullmann’s Encyclopedia of Industrial Chemistry. WILEYVCH Verlag https://doi.org/10.1002/ 14356007.a15_559. Blaschke, M., Tille, T., Robertson, P. et al. (2006). MEMS gas-sensor array for monitoring the perceived car-cabin air quality. IEEE Sens. J. 6: 1298–1308. Müller, G., Beer, S., Paul, S., and Helwig, A. (2011). Novel chemical sensor applications in commercial aircraft. Procedia Eng. 25: 16–22. https:// doi.org/10.1016/j.proeng.2011.12.005. Dooley, E. (2017). American Airlines Jet Has 3rd Fume Incident in 3 Months, 7 Flight Attendants Transported to Hospital and Released. ABC News http:// abcnews.go.com. Majeed, O. (2015). Design and Development of an Aircraft Cabin Air Contaminant Sensor. Specific Range Solutions Ltd. Ryan, M.A., Buehler, M.G., Homer, M.L. et al. (2000). Results from The Space Shuttle STS-95 Electronic Nose Experiment. Jet Propulsion Lab., California Inst. of Tech. https://ntrs.nasa .gov/archive/nasa/casi.ntrs.nasa.gov/ 20000056918.pdf . Kneeskern, B.W. and Buy, W. (2010). The HP 200 LX Powers an Electronic Nose in Space, 2–3. The HP Palmtop Paper http://www.palmtoppaper.com/ ptphtml/51/51000019.htm. Ryan, M.A., Homer, M.L., Manfreda, A. et al. (2003). Second generation electronic nose: real-time air quality

141.

142.

143.

144.

145.

146.

147.

148.

monitoring in human habitat. Bioastronautics Investig. Work. Adv. Environ. Monit., Galveston, Texas. Shevade, A.V., Homer, M.L., Zhou, H. et al. (2007). Development of the third generation JPL electronic nose for International Space Station Technology demonstration. SAE Tech. Pap. 1 3149–3156. doi:https://doi.org/10 .4271/2007-01-3149. The Jet Propulsion Laboratory, California Institute of Technology (2009). Electronic nose to return from space. 1–3. https://www.jpl.nasa.gov/ news/news.php?feature=2309 (accessed 3 April 2019). Expedition, B., Engineer, F., Pettit, D. (2011). Italian-Electronic NOse for Space exploration-(I-ENOS)-11.22.16, 1–10. doi:10.1128/mSphere.0022716.Editor. Stenzel, C. (2016). Deployment of precise and robust sensors on board ISS — for scientific experiments and for operation of the station. Anal. Bioanal. Chem. 408: 6517–6536. https://doi.org/ 10.1007/s00216-016-9789-0. Airbus Defence and Space (2013). On its way to the ISS: the E-nose gets a trunk 1–2. https://www.airbus.com/ newsroom/news/en/2016/03/on-itsway-to-the-iss-the-e-nose-gets-a-trunk .html (accessed 30 December 2016). E-nose, T. (2016). Electronic artificial nose will “ flexibly ” search for bacterial contamination on board the International Space Station About Airbus Defence and Space. http://www .airbusgroup.com/int/en/news-media/ press-releases/Airbus-Group/Financial_ Communication/2016/03/20160317_ airbus_defence_and_space_e-nose.html (accessed 22 March 2017). Wilson, A.D. (2011). Future Applications of Electronic-Nose Technologies in Healthcare and Biomedicine, 267–290. Wide Spectra of Quality Control https://doi.org/10.5772/20836. Wilson, A.D. and Baietto, M. (2011). Advances in electronic-nose technologies developed for biomedical applications. Sensors 11: 1105–1176. https://doi.org/10.3390/s110101105.

375

376

67 Electronic Nose: Current Status and Future Trends 149. Parry, A.D., Chadwick, P.R., Simon, D.

150.

151.

152.

153.

154.

155.

156.

157.

158.

159.

et al. (1995). Leg ulcer odour detection identifies beta-haemolytic streptococcal infection. J. Wound Care 4: 404–406. Korpi, A., Jarnberg, J., and Pasanen, A.L. (2009). Microbial volatile organic compounds. Crit. Rev. Toxicol. 39: 139–193. https://doi.org/10.1080/ 10408440802291497. Gardner, J.W., Craven, M., Dow, C., and Hines, E.L. (1998). The prediction of bacteria type and culture growth phase by an electronic nose with a multi-layer perceptron network. Meas. Sci. Technol. 9: 120–127. https://doi .org/10.1088/0957-0233/9/1/016. Hay, P., Tummon, A., Ogunfile, M. et al. (2003). Evaluation of a novel diagnositc test for bacterial vaginosis: “the electronic nose,”. Int. J. STD AIDS 14: 114–118. Chaudry, A.N., Travers, P.J., Yuenger, J. et al. (2004). Analysis of vaginal acetic acid in patients undergoing treatment for bacterial vaginosis. J. Clin. Microbiol. 42: 5170–5175. https://doi.org/10 .1128/JCM.42.11.5170. Cyranose 320 (2014). Sensigent: intelligent sensing solutions. www.sensigent .com (accessed 3 April 2019) p. 1438. Dutta, R., Hines, E.L., Gardner, J.W., and Boilot, P. (2002). Bacteria classification using Cyranose 320 electronic nose. Biomed. Eng. Online 7: 1–7. Shykhon, M.E., Morgan, D.W., Dutta, R. et al. (2004). Clinical evaluation of the electronic nose in the diagnosis of ear, nose and throat infection: a preliminary study. J. Laryngol. Otol. 118: 706–709. https://doi.org/10.1258/ 0022215042244660. Gouma, P. (2011). Selective oxide sensors as non-invasive disease monitors. SPIE Newsroom: 10–12. https://doi.org/ 10.1117/2.1201001.003427. Pauling, L., Robinson, A.B., Teranishi, R., and Cary, P. (1971). Quantitative analysis of urine vapor and breath by gas-liquid partition chromatography. Proc. Natl. Acad. Sci. U.S.A. 68: 2374–2376. https://doi.org/10.1073/ pnas.68.10.2374. De Heer, K., Van Der Schee, M.P., Zwinderman, K. et al. (2013). Electronic

160.

161.

162.

163.

®

164.

165.

166.

167.

nose technology for detection of invasive pulmonary aspergillosis in prolonged chemotherapy-induced neutropenia: a proof-of-principle study. J. Clin. Microbiol. 51: 1490–1495. https://doi.org/10.1128/JCM.02838-12. Fens, N., De Nijs, S.B., Peters, S. et al. (2011). Exhaled air molecular profiling in relation to inflammatory subtype and activity in COPD. Eur. Respir. J. 38: 1301–1309. https://doi.org/10.1183/ 09031936.00032911. Shafiek, H., Fiorentino, F., Merino, J.L. et al. (2015). Using the electronic nose to identify airway infection during COPD exacerbations. PLoS One 10: 1–16. https://doi.org/10.1371/journal .pone.0135199. Dragonieri, S., Schot, R., Mertens, B.J.A. et al. (2007). An electronic nose in the discrimination of patients with asthma and controls. J. Allergy Clin. Immunol. 120: 856–862. https://doi .org/10.1016/j.jaci.2007.05.043. Gordon, S.M., Szldon, J.P., Krotoszynski, B.K. et al. (1985). Volatile organic compounds in exhaled air from patients with lung cancer. Clin. Chem. 31: 1278–1282. Di Natale, C., Macagnano, A., Martinelli, E. et al. (2003). Lung cancer identification by the analysis of breath by means of an array of non-selective gas sensors. Biosens. Bioelectron. 18: 1209–1218. https://doi.org/10.1016/ S0956-5663( 03)00086-1. Yu, H., Xu, L., Cao, M. et al. (2003). Detection volatile organic compounds in breath as markers of lung cancer using a novel electronic nose. IEEE Sens. J. 2: 1333–1337. Chen, X., Cao, M., Li, Y. et al. (2005). A study of an electronic nose for detection of lung cancer based on a virtual SAW gas sensors array and imaging recognition method. Meas. Sci. Technol. 16: 1535–1546. https://doi.org/10.1088/ 0957-0233/16/8/001. Machado, R.F., Laskowski, D., Deffenderfer, O. et al. (2005). Detection of lung cancer by sensor array analyses of exhaled breath. Am. J. Respir. Crit. Care Med. 171: 1286–1291. https://doi .org/10.1164/rccm.200409-1184OC.

References 168. Mazzone, P.J., Hammel, J., Dweik, R.

169.

170.

171.

172.

173.

174.

175.

et al. (2007). Diagnosis of lung cancer by the analysis of exhaled breath with a colorimetric sensor array. Thorax 62: 565–568. https://doi.org/10.1136/thx .2006.072892. Neustein, M.E. (n.d.). Polymer thin film colorimetric gas sensor for lung cancer analytes. Master thesis, Drexel University. Paff, T., van der Schee, M.P., Daniels, J.M.A. et al. (2013). Exhaled molecular profiles in the assessment of cystic fibrosis and primary ciliary dyskinesia. J. Cyst. Fibros. 12: 454–460. https://doi .org/10.1016/j.jcf .2012.12.010. Nakhleh, M.K., Amal, H., Jeries, R. et al. (2016). Diagnosis and classification of 17 diseases from 1404 subjects via pattern analysis of exhaled molecules. ACS Nano acsnano.6b04930. doi:https://doi.org/10.1021/acsnano .6b04930. Dweik, R.A., Boggs, P.B., Erzurum, S.C. et al. (2011). An official ATS clinical practice guideline: interpretation of exhaled nitric oxide levels (FENO) for clinical applications. Am. J. Respir. Crit. Care Med. 184: 602–615. https://doi .org/10.1164/rccm.912011ST. Struben, V.M.D., Wieringa, M.H., Mantingh, C.J. et al. (2005). Nasal NO: normal values in children age 6 through to 17 years. Eur. Respir. J. Off. J. Eur. Soc. Clin. Respir. Physiol. 26: 453–457. https://doi.org/10.1183/ 09031936.05.00015205. Respiratory, L., Oxide, N., and Nitric, N. (1999). Recommendations for standardized procedures for the online and offline measurement of exhaled lower respiratory nitric oxide and nasal nitric oxide in adults and children—1999. Am. J. Respir. Crit. Care Med. 160: 2104–2117. https://doi.org/10.1164/ ajrccm.160.6.ats8-99. Aerocrine AB (2011) Mastering the measurement of F eno Nitric Oxide Monitoring System NIOX Flex – Nitric Oxide Monitoring System. Righettoni, M., Ragnoni, A., Güntner, A.T. et al. (2015). Monitoring breath markers under controlled conditions. J.

177.

178.

179.

180.

181.

182.

183.

184.

®

176.

185.

Breath Res. 9: 1–6. https://doi.org/10 .1088/1752-7155/9/4/047101. Rosenberg, E. (2015). A Study of the Accuracy and Precision of Selected Breath Alcohol Measurement Devices (“ Breathalyzers ”), Rep. TUW CTA 2015/27EN. 1–45. Hughes, F.J. and McNab, R. (2008). Oral malodour - a review. Arch. Oral Biol. 53: S1–S7. https://doi.org/10 .1016/S0003-9969( 08)70002-5. Phillips, C., Parthaláin, N.M., Syed, Y. et al. (2014). Short-term intra-subject variation in exhaled volatile organic compounds (VOCs) in COPD patients and healthy controls and its effect on disease classification. Metabolites 4: 300–318. https://doi.org/10.3390/ metabo4020300. Holz, O., Tal-Singer, R., Kanniess, F. et al. (2005). Validation of the human ozone challenge model as a tool for assessing anti-inflammatory drugs in early development. J. Clin. Pharmacol. 45: 498–503. https://doi.org/10.1177/ 0091270004273527. Biller, H., Holz, O., Windt, H. et al. (2011). Breath profiles by electronic nose correlate with systemic markers but not ozone response. Respir. Med. 105: 1352–1363. https://doi.org/10 .1016/j.rmed.2011.03.002. Capelli, L., Taverna, G., Bellini, A. et al. (2016). Application and uses of electronic noses for clinical diagnosis on urine samples: a review. Sensors (Switzerland) 16: 1–23. https://doi.org/ 10.3390/s16101708. Pavlou, A.K., Magan, N., McNulty, C. et al. (2002). Use of an electronic nose system for diagnoses of urinary tract infections. Biosens. Bioelectron. 17: 893–899. Kodogiannis, V.S., Lygouras, J.N., Tarczynski, A., and Chowdrey, H.S. (2008). Artificial odor discrimination system using electronic nose and neural networks for the identification of urinary tract infection. IEEE Trans. Inf. Technol. Biomed. 12: 707–713. https:// doi.org/10.1109/TITB.2008.917928. Persaud, K.C., Pisanelli, A.M., Evans, P., and Travers, P.J. (2006). Monitoring urinary tract infections and bacterial

377

378

67 Electronic Nose: Current Status and Future Trends

186.

187.

188.

189.

190.

191.

192.

193.

194.

vaginosis. Sens. Actuators, B 116: 116–120. https://doi.org/10.1016/j.snb .2005.12.062. Bernabei, M., Pennazza, G., Santonico, M. et al. (2008). A preliminary study on the possibility to diagnose urinary tract cancers by an electronic nose. Sens. Actuators, B 131: 1–4. https://doi.org/ 10.1016/j.snb.2007.12.030. Weber, C.M., Cauchi, M., Patel, M. et al. (2011). Evaluation of a gas sensor array and pattern recognition for the identification of bladder cancer from urine headspace. Analyst 136: 359–364. https://doi.org/10.1039/c0an00382d. Siyang, S., Kerdcharoen, T., Wongchoosuk, C. (2012). Diabetes diagnosis by direct measurement from urine odor using electronic nose. 5th 2012 Biomed. Eng. Int. Conf. BMEiCON 2012. 23–25. doi:https://doi.org/10 .1109/BMEiCon.2012.6465441. Arasaradnam, R.P., Quaret, N., Thomas, M.G. et al. (2012). Evaluation of gut bacterial populations using an electronic e-nose and field asymmetric ion mobility spectrometry: further insights into “fermentonomics,”. J. Med. Eng. Technol. 36: 33–337. Chapman, P., Clinton, J., Kerber, R. et al. (2000). CRISP-DM 1.0: Step-byStep Data Mining Guide. SPSS https:// www.the-modeling-agency.com/crispdm.pdf . Goodner, K.L., Dreher, J.G., and Rouseff, R.L. (2001). The dangers of creating false classifications due to noise in electronic nose and similar multivariate analyses. Sens. Actuators, B 80: 261–266. Scott, S.M., James, D., and Ali, Z. (2007). Data analysis for electronic nose systems. Microchim. Acta 156: 183–207. https://doi.org/10.1007/ s00604-006-0623-9. Ciosek, P. and Wróblewski, W. (2006). The analysis of sensor array data with various pattern recognition techniques. Sens. Actuators, B 114: 85–93. https:// doi.org/10.1016/j.snb.2005.04.008. Wang, M., Perera, A., and Gutierrez-Osuna, R. (2004). Principal discriminants analysis for small-samplesize problems: application to chemical

195.

196.

197.

198.

199.

200.

201.

202.

203.

sensing. Proc. IEEE Sens. 2: 591–594. https://doi.org/10.1109/ICSENS.2004 .1426234. Raman, B., Sun, P.A., Gutierrez-Galvez, A., and Gutierrez-Osuna, R. (2006). Processing of chemical sensor arrays with a biologically inspired model of olfactory coding. IEEE Trans. Consum. Electron. 17: 1015–1024. Raman, B., Yamanaka, T., and Gutierrez-Osuna, R. (2006). Contrast enhancement of gas sensor array patterns with a neurodynamics model of the olfactory bulb. Sens. Actuators, B 119: 547–555. https://doi.org/10.1016/j .snb.2006.01.035. Skov, T. and Bro, R. (2005). A new approach for modelling sensor based data. Sens. Actuators, B 106: 719–729. https://doi.org/10.1016/j.snb.2004.09 .023. Bro, R. (2006). Review on multiway analysis in chemistry — 2000 – 2005. Crit. Rev. Anal. Chem. 36: 279–293. https://doi.org/10.1080/ 10408340600969965. Gutierrez-Galvez, A. and Gutierrez-Osuna, R. (2006). Increasing the separability of chemosensor array patterns with Hebbian/anti-Hebbian learning. Sens. Actuators, B 116: 29–35. https://doi.org/10.1016/j.snb.2005.11 .081. Gutierrez-Osuna, R., Nagle, H.T., and Schiffman, S.S. (1999). Transient response analysis of an electronic nose using multi-exponential models. Sens. Actuators, B 61: 170–182. Snopok, B.A. and Kruglenko, I.V. (2002). Multisensor systems for chemical analysis: state-of-the-art in electronic nose technology and new trends in machine olfaction. Thin Solid Films 418: 21–41. Zhang, W.-M., Hu, J.-S., Song, W.-G., and Wan, L.-J. (2007). Detection of VOCs and their concentrations by a single SnO2 sensor using kinetic information. Sens. Actuators, B 123: 454–460. https://doi.org/10.1016/j.snb .2006.09.051. Hiranaka, Y., Abe, T., and Murata, H. (1992). Gas-dependent response in the

References

204.

205.

206.

207.

208.

209.

210.

211.

212.

temperature transient of SnO2 , gas sensors. Sens. Actuators, B 9: 177–182. Lee, A.P. and Reedy, B.J. (1999). Temperature modulation in semiconductor gas sensing. Sens. Actuators, B 60: 35–42. Burresi, A., Fort, A., Rocchi, S. et al. (2005). Dynamic CO recognition in presence of interfering gases by using one MOX sensor and a selected temperature profile. Sens. Actuators, B 106: 40–43. https://doi.org/10.1016/j.snb .2004.05.035. Heilig, A., Barsan, N., Weimar, U. et al. (1997). Gas identification by modulating temperatures of SnO2 -based thick film sensors. Sens. Actuators, B 43: 45–51. Mielle, P., Souchaud, M., Landy, P., and Guichard, E. (2006). A direct thermal desorber as a sampling device for application specific sensor system. Sens. Actuators, B 116: 161–167. https://doi .org/10.1016/j.snb.2005.12.070. Althainz, P., Goschnick, J., Ehrmann, S., and Ache, H.J. (1996). Multisensor microsystem for contaminants in air. Sens. Actuators, B 33: 72–76. Goschnick, J. (2001). An electronic nose for intelligent consumer products based on a gas analytical gradient microarray. Microelectron. Eng. 57–58: 693–704. Tuovinen, K., Kolehmainen, M., and Paakkanen, H. (2001). Determination and identification of pesticides from liquid matrices using ion mobility spectrometry. Anal. Chim. Acta 429: 257–268. https://doi.org/10.1016/ S0003-2670( 00)01290-3. White, J., Kauer, J.S., Dickinson, T.A., and Walt, D.R. (1996). Rapid analyte recognition in a device based on optical sensors and the olfactory system. Anal. Chem. 68: 2191–2202. Michael, K.L., Taylor, L.C., Schultz, S.L., and Walt, D.R. (1998). Randomly ordered addressable high-density

213.

214.

215.

216.

217.

218.

219.

optical sensor arrays. Anal. Chem. 70: 1242–1248. Perera, A., Yamanaka, T., Gutiérrrez-Gálvez, A. et al. (2006). A dimensionality-reduction technique inspired by receptor convergence in the olfactory system. Sens. Actuators, B 116: 17–22. https://doi.org/10.1016/j .snb.2005.11.082. Pardo, M. and Sberveglieri, G. (2007). Comparing the performance of different features in sensor arrays. Sens. Actuators, B 123: 437–443. https://doi .org/10.1016/j.snb.2006.09.041. Gualdrón, O., Llobet, E., Brezmes, J. et al. (2006). Coupling fast variable selection methods to neural network-based classifiers: application to multisensor systems. Sens. Actuators, B 114: 522–529. https://doi.org/10.1016/j .snb.2005.04.046. Llobet, E., Gualdron, O., Brezmes, J. et al. (2005). An unsupervised dimensionality-reduction technique. Sens. IEEE 6. Llobet, E., Gualdrón, O., Vinaixa, M. et al. (2007). Efficient feature selection for mass spectrometry based electronic nose applications. Chemom. Intell. Lab. Syst. 85: 253–261. https://doi.org/10 .1016/j.chemolab.2006.07.002. Gardner, J.W., Boilot, P., and Hines, E.L. (2005). Enhancing electronic nose performance by sensor selection using a new integer-based genetic algorithm approach. Sens. Actuators, B 106: 114–121. https://doi.org/10.1016/j.snb .2004.05.043. Ballabio, D., Cosio, M.S., Mannino, S., and Todeschini, R. (2006). A chemometric approach based on a novel similarity/diversity measure for the characterisation and selection of electronic nose sensors. Anal. Chim. Acta 578: 170–177. https://doi.org/10.1016/j .aca.2006.06.067.

379

381

68 Surface Science in Batteries Yuping Wu and Rudolf Holze

A brief survey of electrochemical systems for energy conversion and storage with attention to currently popular and widespread systems as well as current trends of development is followed by a look at experimental methods suitable for studies of surfaces and interfaces in such systems with particular attention to in situ methods. The relevance of obtained results for further research and development is illustrated based on a brief introduction of electrode overpotentials as a fundamental concept of electrode kinetics. Selected examples of investigations of electrodes are presented as evidence of the central role of surface science in batteries.

68.1 Batteries: An Overview

The term battery in daily communication – although rather incorrect, because its French origin implies only the combination of several devices of the same kind, e.g. cannons in armament – names a wide variety of electrochemical devices for energy conversion and storage [1–8]. Their size ranges from tiny button cells powering wristwatches (and even smaller microbatteries incorporated into electronic solid-state devices powering sensors) up to huge collections of lead-acid batteries stabilizing the electric grid of cities and states. Equally wide is the range of principles and systems upon which these devices are based. A first overview can be obtained by a simple classification: 1. Primary systems: devices that cannot be recharged and provide only conversion of stored chemical energy into electrochemical energy (popular terminology: zinc carbon, Leclanché, zinc alkaline, zinc air, lithium). 2. Secondary systems: devices that can be recharged electrochemically, i.e. by turning the galvanic cell operating during discharge into an electrolytic cell during charge (popular terminology: lead acid, lithium ion, nickel cadmium, nickel metal hydride).

Surface and Interface Science: Applications of Surface Science I, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

382

68 Surface Science in Batteries

3. Fuel cells: devices that serve only as converters, but not as storage devices (popular terminology: phosphoric acid fuel cell, direct methanol fuel cell, solid oxide fuel cell, molten carbonate fuel cell) [9] (See also Chapter 69 in this volume.) 4. Supercapacitors (popular terminology: electrochemical double-layer capacitor [EDLC]). This classification is only a rough one and a starting point. Boundaries between primary and secondary systems are porous, e.g. the established zinc-alkaline primary cell can be modified easily, yielding a rechargeable secondary cell. Many materials used in primary and secondary cells can also be used in supercapacitors (e.g. MnO2 ). The same applies to operating principles: supercapacitors initially based exclusively on ion accumulation at electrode/solution interfaces at highly porous carbon electrodes (EDLCs) now also utilize surface redox processes just as with materials in secondary batteries [10–14]. Metal-air batteries employing zinc as the negative electrode (the terms anode and cathode are not used here because they cause only confusion) and a dioxygen-reducing air-fed gas diffusion electrode as the positive merge principles of a primary cell with those of a fuel cell. In a redox flow battery (RFB) wherein the reactants are present as dissolved salts of the respective redox-active ions, features of a fuel cell (in being only a converter in the cell stack) are combined with those of a secondary battery when storing charge and energy by reversing the electrode reactions proceeding initially during discharge. The overlap becomes even more obvious when a zinc electrode is employed as the negative electrode and a soluble redox system at the positive in a RFB. In a similar way a supercapacitor utilizing ion accumulation at the positive electrode and lithium-ion insertion/deinsertion at the negative electrode is a merger of concepts from secondary batteries and double-layer capacitors with the result called lithium-ion capacitor. In all devices heterogeneous processes occur at buried interfaces1) (i.e. interfaces between condensed phases: liquid/solid, solid/solid, liquid/liquid). This heterogeneous nature is a fundamental and inherent drawback because it seriously limits all devices in terms of space–time yield, i.e. in terms of the rate of conversion between electric energy and energy stored in chemical compounds (“chemical energy”), and has resulted in close interactions with surface science. This limitation is well known from catalysis when comparing heterogeneous and homogeneous catalysis. But this unwelcome fact is the foundation of all electrochemical processes: it provides the only conceivable coupling between electronic currents and ionic currents. The rate of electrochemical conversion reactions (i.e. electrode reactions), whether expressed with a rate constant k or as an exchange current density j0 , has huge implications on the performance of every electrochemical device. Its understanding and control, in most cases its increase, but sometimes also its decrease, are central challenges in electrochemical research and development. But this drawback comes 1) Commonly the term surface is applied to solid/gas or solid/vacuum interfaces, whereas interface is a more general term denominating phase boundaries. Both terms are used in this chapter synonymously.

68.1 Batteries: An Overview

with a potentially huge advantage: the efficiency is not limited to that derived from the Carnot cycle. Some general considerations of thermodynamic, kinetic, economic, and practical aspects have always guided selection and subsequent combination of materials to be used as active masses in systems treated in this chapter:

• The electrode potentials of both electrodes should be as far apart as possible (see the electrochemical series [15]).

• The electrode processes should be fast (rapid kinetics; see Section 68.3). • The materials should be cheap and abundantly available; their production (if they do not occur in nature already) and handling should be environmentally friendly, or at least it should cause no harm to the environment. • Materials should be mechanically stable showing no transformation into less easily convertible forms by crystal growth (Ostwald ripening) or separation from a current collector (shedding). • The materials should be chemically stable with respect to the electrolyte (solution, melt, or whatever); otherwise self-discharge will occur. • Materials for secondary system should enable high yield electrode reactions in both directions, i.e. during charge and discharge. Although a large number of conceivable combinations have been suggested and investigated for all types of systems, only a few are of commercial importance (Figure 68.1). Primary cells are mostly of the zinc-alkaline type (AlMn); in some countries the significantly cheaper zinc-carbon cells (ZnCs) still dominate. In professional and medical applications (for an example, see Figure 68.2), primary lithium batteries based on various chemistries are widespread. Secondary cells omnipresent in most types of mobile and portable applications are lithium-ion batteries of various chemistries. Lead-acid cells are almost exclusively used in vehicles with an internal combustion engine for starting, lighting, and AlMn Lead acid NiCd NiMh Li-ion others ZnC

2% 3%

6% 13% 2%

13% 61% Figure 68.1 Fractions of various battery systems in the total of 42 531 tons of batteries sold in Germany in 2010.

383

384

68 Surface Science in Batteries

Figure 68.2 Lithium-battery-powered data logger.

ignition. Because of its relatively low price (when compared with lithium-ion systems), it is also frequently found in cars (mostly hybrid cars combining a secondary battery with an internal combustion engine). Nickel-cadmium cells are slowly disappearing because of environmental concerns, whereas nickel metal hydride cells enjoy popularity because they are cheaper than lithium-ion cells and easier to handle (Figures 68.3–68.5). Portable and mobile devices like mobile phones, portable computers of all types, and many kinds of electronic gadgets are powered with lithium-ion batteries (Figure 68.6). Compared with lead-acid accumulators, they provide significantly higher energy density at a much higher price. When used in electric vehicles, this results in higher overall costs. Small batteries used in bicycles and scooters (Figures 68.7 and 68.8) are already competitive. When cars are equipped with lithium-ion batteries, driving range and price are currently not competitive at the time of writing with vehicles powered with internal combustion engines.

68.1 Batteries: An Overview

Figure 68.3 Components of a dissembled primary lithium cell (button type). Left to right: plastic sealing ring, metal grid current collector, cell top, separator, copper foil current collector, and cell can.

Figure 68.4 Electric scooter with a lead-acid battery.

Figure 68.5 Low-speed lead-acid battery-powered electric vehicle, China.

385

386

68 Surface Science in Batteries

Figure 68.6 Mobile phone powered with lithium-ion battery.

Figure 68.7 Electric bicycle with a lithium-ion battery.

Current developments aim at improved energy density at significantly reduced costs. Because of the limited supply of lithium, secondary batteries employing alternative chemistries, e.g. based on sodium or aluminum ions, are also studied intensely. Secondary batteries employing a bifunctional dioxygen electrode (lithium air [16–19], sodium air [20–28], zinc air [29–36]) have attracted huge interest

68.1 Batteries: An Overview

Figure 68.8 Electric scooter with a lithium-ion battery.

(a)

(b)

(c)

Figure 68.9 Mobile data bank (a), power tool (b), and active touch pens (c) powered with lithium-ion batteries.

because of their inherently given high energy densities: the dioxygen does not affect the energy density calculation (Figures 68.9 and 68.10). Particularly suitable for large-scale and long-term storage of electric energy are RFBs [37–45]. As schematically depicted in Figure 68.11, a galvanic flow-through cell is combined with two tanks containing the same supporting electrolyte with different redox systems added into both tanks. The electrode potentials established at the electrodes in the galvanic cell when the solutions are feed into the cell yield the cell voltage; as an example the all-vanadium cell with the electrode reactions Charge

+ + − −−−−−−−−− → VO2+ + H2 O ← − VO2 + 2 H + e E0,SHE = 1.00 V Discharge

and Discharge

3+ − −−−−−−−−− → V2+ ← − V + e E0,SHE = −0.26 V Charge

387

388

68 Surface Science in Batteries

Figure 68.10 Electric cars powered with lithium-ion batteries (Inset: charge plug cover).

proceeding at electronically conducting materials like carbon acting only as electron sink/source may be considered. Mixing of both electrolyte solutions must be avoided as perfectly as possible because it will result in a chemical comproportionation reaction between redox ions replacing the heterogeneous reactions of the respective ions at their designated electrodes with a homogeneous reaction, i.e. a “chemical short circuit.” Currently ion-exchange membranes permitting transfer of one type of electrolyte ions only are used. The size of the tanks for the electrolyte solutions and thus the storage capability can be scaled easily as required; the power capability can be adopted as easily by adding cells or increasing the size of the cells in the stack (multiple cells are connected in series to obtain a practically useful voltage). A commercial example is shown in Figure 68.12.

68.1 Batteries: An Overview

– Tank for negative electrode/ electrolyte V(II)/V(III)

Pump

+

Cell stack

Tank for positive electrode/ electrolyte V(IV)/V(V)

Pump

Figure 68.11 Scheme of an all-vanadium redox flow battery.

Figure 68.12 All-vanadium RFB, 8 kW maximum output, 10 kW maximum input, storage capability 16 kW h, Yinfeng New Energy, China.

389

390

68 Surface Science in Batteries

A major limitation of all secondary devices is their limited power capability, i.e. capability to deliver large currents. Certainly lead-acid batteries for vehicles provide relatively large currents for short times even at low temperatures, but their capability to receive large charging currents is very limited. The same applies to all other secondary systems. This drawback becomes highly unwelcome when fast charging of batteries in vehicles is required, and it is even more unwelcome when regenerative braking of trains, tramways, and other large loads is required. This feature, which permits braking by operating the electric motors as generators, is basically a well-established concept utilized by Swiss railways for decades already. In trains used for urban transport, this concept where generated electricity is fed back into the overhead wire is not suitable. Because savings of consumption of electric energy of as much as 30% are possible, other means have been considered. Supercapacitors are a solution frequently applied already. These capacitors have been initially simply capacitors utilizing the large double-layer capacity of the electrochemical interface between electronically conducting phase/ionically conducting phase (e.g. carbon and electrolyte solution). Typical small ones (left), medium-sized (top right), and large ones (bottom right) are shown in following (Figures 68.13 and 68.14). Even larger storage capabilities (by a factor of up to 100 and higher with respect to the electrochemically active surface area) can be achieved when utilizing superficial

Figure 68.13 Supercapacitors of the EDLC type from various manufacturers. Left: capacitor containing two cells in series and all others single cells. (Source: Top right: Photograph courtesy of Maxwell Technologies, Inc.)

68.1 Batteries: An Overview

Figure 68.14 Supercap bus on Renmin Rd., Shanghai, China. Powered with EDLC supercapacitors. Charging every few stops (as shown here) provides enough energy for the next kilometers. Compared with

battery-operated vehicles, cost reductions of 40% at purchase and US$ 200 000 savings on fuel expenses during expected lifetime are calculated.

(surface-bound) redox reactions of metal oxides [46–51] or intrinsically conducting polymers [52–58] in contact with an electrolyte solution. The relationship between gravimetric energy density, i.e. the amount of electric energy stored per mass of a storage and conversion system, and the gravimetric power density, i.e. the power per mass a device can deliver, is frequently displayed in a so-called Ragone plot [59]; a typical example representing approximately the state of 2014 is depicted in Figure 68.15. It shows the high gravimetric energy density of fuel cells, where the fuel itself (hydrogen, alcohols, etc.) and oxidant (dioxygen from air) are not taken into the calculation, and the high power density of electrolytic capacitors. Quite obviously an ideal system would be found in the top right corner of the plot. As of today in this corner only internal combustion engines are found. These plots should be taken with precaution despite their omnipresent popularity: frequently it is not specified which parts of a system are taken into account – only the active masses in the electrodes, or the complete mass of the cell content and housing, or even external components like battery management systems. In addition the performance of a system is reduced just to one property, i.e. one dimension. This may be inappropriate and insufficient in many cases.

391

68 Surface Science in Batteries

1000 0.1c 10 h

Fuel cells 100 Energy density (Wh/kg)

392

1c 1h

10c 6 min

Li ion 100c 36 s

NiCd/NiMeH 10

Pb acid

1000c 3.6 s

Ultracaps 1

10000c 0.36 s

EDLC

105c

0.1 Electrolytics 0.01 10

100

1000

10000

Power density (W/kg) Figure 68.15 Ragone plot for selected electrochemical energy storage and conversion systems, labels: discharge rating in values of c (rated capacity) and theoretical discharge time based on rated capacity.

68.2 Experimental Methods

All interfaces in electrochemical systems related to energy conversion and storage (with the notable exception of solid oxide fuel cells not treated in this chapter) are buried interfaces [60] between two condensed phases. Accordingly many methods well known from surface science – where more frequently gas/solid or vacuum/solid interfaces are studied – cannot be used at all or only after considerable adaption for electrochemical in situ studies [61]. This approach is always preferable because it avoids all possible pitfalls and artifacts conceivably caused by transfer of the object to be studied from an electrochemical cell into the (mostly UHV) system of the selected surface analytical method. Many well-established methods (see Vols. 1 and 2 of this work) are frequently employed ex situ nevertheless either because the risk of artifacts, contamination, etc. can reliably be excluded or because the desired information cannot be obtained otherwise. A look at available probes and signals [36, 62] easily enables a first overview of probes, signals, and resulting methods that can only be applied ex situ (non-italic) and those also applicable in situ (italics). Although the latter appear to be in a small minority, the rapidly developed and still growing field of spectroelectrochemistry [36, 63, 64] has convincingly demonstrated the remarkable capabilities of numerous methods to study electrode reactions and to elucidate their steps and the associated structures and dynamics of the respective electrified interfaces (Figure 68.16). The application under in situ conditions adds one more problem when applying most methods: how to separate signals from the bulk of the adjacent condensed

68.2 Experimental Methods

e–

W

h∗ν

i+/0/– H

h∗ν E

E

i+/0/– H e– W Electrolyte solution

Electrode Figure 68.16 Probes and signals in surface spectroscopies. e− : electrons, i: ions, h*𝜈: electromagnetic radiation, W: thermal energy, E: electric field, and H: magnetic field.

phases from those originating at the interface to be studied. Obviously this can be afforded by directing the probe in an appropriate way (grazing angle) or by selecting only signals generated at the interface. Another way is the preparation of thin films or finely dispersed materials with most atoms/species of interest being at the surface. The desired information from an electrochemical interface can be associated with the following properties:

• • • • •

Chemical identity Chemical composition State of oxidation of species Chemical and crystallographic environment of a particular species Crystallographic orientation

Further properties like reflectivity or optical transmission in various spectral regions are interesting in other fields of research but not for the subject of this chapter; they are treated in detail elsewhere [36]. The chemical composition of the surface of an electrode material is most frequently studied ex situ with XPS [65, 66]; Auger electron spectroscopy [65, 66] is applied less frequently. Beyond the identification of the chemical species, their state of oxidation and their chemical environment, i.e. chemical bonds the species are engaged in, can be determined. Spatial distribution of elements on a surface/at an interface can be studied with various scanning techniques. Scanning Auger electron spectroscopy is rarely applied; the popularity of scanning electron microscopes with attached instrumentation enabling energy-dispersive X-ray spectroscopy (EDX)/analysis [67] makes this the dominating method (Figure 68.17). A potential disadvantage of EDX is the larger depth of penetration; possibly undesired information from the bulk of the sample might contribute to the recorded spectrum as shown below for graphite particles coated with Al2 O3 acting as a synthetic solid electrolyte interface (SEI) [68] with a strong carbon peak (Figures 68.18 and 68.19).

393

68 Surface Science in Batteries

Figure 68.17 EDX mapping image of carbon-coated VO2 suggested as electrode material for 5 V lithium-ion batteries; (for details see [68].)

12000

CK Element Weight (%) Atomic (%)

10000 Counts

394

8000 6000 4000

CK

91.77

93.92

OK

7.44

5.71

Al K

0.8

0.36

Al K

2000

OK

0 0

1 μm (a)

1 2 Energy (keV)

18 19 20

(b)

Figure 68.18 TEM image of Al2 O3 a-coated graphite (a) and EDX spectrum (b) [69].

Results pertaining to both qualitative and quantitative elemental composition are helpful in understanding electrode surface properties and electrode reaction mechanisms. Beyond mere identification of elements themselves from deconvolution of XPS peaks, the chemical environment or the location of an element in a particular functional group can be deduced, further supporting the already addressed understanding (Figure 68.20). When combined with a sputtering option, profiles of chemical composition can be obtained helpful in studies of layer formation on working electrodes. Even when initially only the topmost atoms of one (solid) phase are interacting with species from the adjacent liquid phase resulting in formation of an adsorbate layer, this interaction might result in rearrangement of these topmost atoms (surface reconstruction [72, 73]) and formation of a specific spatial arrangement of the adsorbed species.

68.2 Experimental Methods

2p3/2

Ru 3d3/2 Co 2p

Ru 3d5/2

2p1/2 2+ Satellite

(b)

Binding energy (eV)

2+ 3+ Satellite

810 805 800 795 790 785 780 775

278 280 282 284 286 288 290 292

(a)

3+

Intensity (a.u.)

Intensity (a.u.)

Ru 3d and C 1s

Binding energy (eV)

O 1s

Intensity (a.u.)

529.4 eV

531.3 eV 530.8 eV

532.4 eV

536 535 534 533 532 531 530 529 528 527

(c)

Binding energy (eV)

Figure 68.19 XPS results from ultrathin mesoporous RuCo2 O4 nanoflakes prepared by electrodeposition. The Ru 3d5/2 peak in (a) could be fitted with three components all typical of Ru(IV); for further details see [70].

Intensity (a.u.)

C1s

294

292

290 288 Binding energy (eV)

286

284

Figure 68.20 C 1s XPS results from the surface of natural graphite after oxidation treatment (LS2); for details see [71].

395

396

68 Surface Science in Batteries

These two layers (in the simplest case) form a new, third phase, which is frequently called an interphase. The essential 2D interface has expanded into something 3D; because of its very small z-dimension, it has been called sometimes 2.5D [74]. The term interphase is even more fittingly applied to thicker new phases, e.g. corrosion or passivation layers, deposits, and films. Consequently information may now result from somewhere inside this phase or from locations closer to one of the adjacent phases. Provided sufficient spatial resolution is possible with an experimental method, such localized information may be accessible; it might be helpful in understanding transformation reactions inside interphases during reduction or oxidation of intrinsically conducting polymer films. Crystallographic surface-specific information can be extracted preferably ex situ by low energy electron diffraction (LEED) [75, 76]. X-ray diffraction can be applied also in situ, but this method is inherently not surface sensitive. Using grazing incidence or using very thin films that cause only small bulk contributions can be an option [77, 78]. All diffraction methods require a minimum of periodicity, i.e. a minimum size of crystalline particles or regions inside an interphase. X-ray transparent cell windows and thin electrolyte solution layers are desirable to minimize X-ray intensity attenuation. A simplified cross section shown below outlines the experimental arrangement (Figures 68.21 and 68.22). The presence and absence of X-ray diffraction peaks associated with the charge and discharge products of TiS2 and LiTiS2 can be identified and can be related to the electrochemical processes inside the cell. Because of the penetration depth of the X-ray, the obtained results are hardly surface specific. An arrangement using specular reflectance providing higher surface sensitivity is depicted in the following picture (Figure 68.23). Spatial information about the structural and chemical environment of atoms on or at interfaces, e.g. a distance between adsorbed atoms/ions and surface electrode atoms, can be extracted from absorption spectra of X-rays at high resolution close to the absorption edge of the atoms under investigation [82–84]. Measurements at energies (wavelengths) slightly below the absorption edge are called X-ray absorption near-edge structure/spectroscopy (XANES) or near-edge X-ray absorption fine structure structure/spectroscopy (NEXAFS), whereas measurements around the absorption edge and slightly above it are called extended X-ray absorption fine structure/spectroscopy (EXAFS). Because of the penetration depth of X-rays, these methods are not intrinsically surface sensitive. Depending on the experimental Cu Kα radiation Window Cathode Separator Anode Figure 68.21 In Situ X-ray diffraction electrochemical cell. (Source: Based on [79].)

68.2 Experimental Methods

1 Charged Relative intensity I/I0

0.8 0.6 Discharged 0.4 0.2 0 28

29

30

31

32

33

34

35

2θ (°) Figure 68.22 In situ X-ray diffraction pattern of TiS2 recorded during charging/discharging experiments. (Source: Based on data in [52, 80].)

Mylar film

Working electrode Reference electrode

Counter electrode

Electrolyte solution

(a)

X-ray Mylar film

Counter electrode (b)

Working electrode Reference electrode Electrolyte solution

Figure 68.23 Schematic spectroelectrochemical cell for in situ X-ray specular reflection measurements. (a) Position of central working electrode down for electrochemical measurements. (b) Working electrode up for X-ray measurements. (Source: Based on [81].)

setup and/or the selection of the X-ray energy and in turn the atoms to be probed, species at the interface can be studied specifically. The absorption coefficient 𝜇 of an atom in a molecule or a solid has a fine structure2) amounting to about 15% of the absorption step at the edge extending up to several hundred electron volts above the absorption edge, whereas a free atom shows a smooth absorption only as seen below (Figure 68.24). 2) This absorption is sometimes called AXAFS (atomic X-ray absorption fine structure).

397

68 Surface Science in Batteries

Absorption

398

NEXAFS SEXAFS

500

600

700

800

900

Photon energy (eV) Figure 68.24 Characteristic SEXAFS spectra of atomic oxygen adsorbed on Cu(110) in a Cu(110)-(2 × 1)-O arrangement; smooth line shows absorption of free atoms. (Source: Based on data in [85].)

The oscillations of the observed absorption are caused by interference between the photon emitted from the absorbing atom and photon waves emitted from neighbor atoms, as depicted schematically in Figure 68.25. When the neighbor atoms are surface atoms, the distance between the atoms of interest and surface atoms rAS can be determined (Figure 68.26).

Photon

Figure 68.25 Principle of photon interference in X-ray absorption.

α rAB rAS

rAS

Figure 68.26 Structural features determined with X-ray absorption spectroscopy.

Absorption

68.2 Experimental Methods

NEXAFS SEXAFS

500

600 700 Photon energy (eV)

800

Figure 68.27 Characteristic SEXAFS spectra of NO adsorbed on Ni(111) in a Ni(111)c(4 × 1)-NO arrangement. (Source: Based on data in [87].)

This distance determines the frequency of the absorption oscillation, whereas the amplitude is influenced by the number and identity of the neighbors and their distance. The polarization of the incident X-rays strongly influences the observed oscillations. Bonds with interatomic vectors in the plane of polarization contribute significantly, whereas bonds perpendicular to this plane do not (for further details, see [86]). In the case of atomic species, the near-edge structure is of no further interest. With molecular adsorbates the near-edge X-ray absorption is dominated by intramolecular transitions (μ- and σ-resonances). Their dependency on the polarization provides information on the orientation of the molecule. From the energy of the σ-resonance intramolecular bond lengths, rAB can be estimated for simple molecules. The X-ray absorption spectrum of NO adsorbed on a nickel surface is shown in Figure 68.27. These arguments are of course also valid for species on top of a surface. The previous figure shows the structural information available from both spectroscopies. Distances can be determined with a precision of about ±1 pm, and the number of atoms with about 15% uncertainty. Since no diffraction is involved, samples without any long-range ordering can be studied. As already indicated X-ray absorption spectroscopy is inherently not surface sensitive. By adjusting the photon energy to a value matching the absorption edge of species sitting on a surface, this sensitivity can be easily obtained. With photons impinging at a grazing angle below the angle of total reflection, contributions from the substrate are suppressed to a large extent [88], and the depth of information is reduced to about a few nanometers only. The method is called grazing incidence X-ray absorption fine structure/spectroscopy (GIXAFS). A general experimental layout is shown in Figure 68.28. The state of oxidation of some elements showing the Mössbauer effect (the resonant absorption of gamma radiation by certain atomic nuclei, 57 Fe, 119 Sn, several other isotopes of technical interest like 57 Co can also be studied in the emission

399

400

68 Surface Science in Batteries

Synchrotron

Monochromator

Electrochemical cell

Focusing mirror Detector Figure 68.28 Setup for an X-ray absorption experiment in the fluorescence mode.

Fe(I) S = 3/2 Fe(I) S = 1/2 Fe(II) S = 2 Fe(II) S = 1 Fe(II) S = 0 Fe(III) S = 5/2 Fe(III) S = 3/2 Fe(III) S = 1/2 Fe(IV) S = 2 Fe(IV) S = 1 Fe(VI) S = 1

–1

–0.5

0

1 0.5 Isomer shift (mm/s)

1.5

2

2.5

Figure 68.29 Correlation between state of oxidation of iron ions and the observed isomer shift in Mössbauer spectroscopy. (Source: Based on literature data [90].)

mode) can be determined with Mössbauer spectroscopy [89, 90]; in addition this method can provide information about the crystallographic environment and the symmetry of the surrounding [91]. The energetic shift of the absorption and the splitting of lines and their multiplicity are indicative of the nature of the chemical bond between the investigated atom and its neighbors or the state of oxidation. Figure 68.29 provides a rough correlation between the state of oxidation of iron and the isomer shift as an example. A cell for spectroelectrochemical in situ studies is shown below in cross section (Figure 68.30). Prussian blue formed as a highly colored colloidal precipitate by adding Fe(III) ions to a solution of Fe(II) hexacyanide has been identified as being electrochromic and suitable for electro-optic applications; more recently some members of the large hexacyanometallate family have attracted attention as possible sodium ion storage materials in batteries and supercapacitors (Figure 68.31).

68.2 Experimental Methods

Lead shield

Working electrode

Reference electrode

Counter electrode

57Co-source

Aluminium cell mount

Detector

Plexiglass plunger

Plexiglass cell body

Arbitrary units

Figure 68.30 Electrochemical cell for in situ Mössbauer spectroscopy in the transmission mode [92].

–1.5

–0.7

0.1

0.9 v (mm/s)

1.7

2.5

3.3

Figure 68.31 In Situ Mössbauer spectra of a film of Prussian blue deposited on a glassy carbon electrode in an aqueous solution of 1 M KCl (pH = 4) at E SCE = 0.6 V (top) and E SCE = −0.2 V (bottom). (Source: Based on [93].)

Tin has been suggested as a negative electrode material for lithium-ion batteries because of its large theoretical storage capabilities. Large volume changes during charge/discharge reactions have limited progress in the application so far. Mössbauer spectroscopy has been employed in studies of tin related to this application [94, 95]. Identification of molecular species being adsorbed on an electrode surface as a reactant, as an intermediate, or a surface poison or being a constituent of an interphase is most frequently done using vibrational spectroscopies [40, 96]. Except for high-resolution electron energy loss spectroscopy (HREELS) [97, 98] these spectroscopies can basically be employed in situ. Unfortunately both

401

402

68 Surface Science in Batteries

principles – infrared (IR) absorption and Raman scattering – face major obstacles accompanying the advantages: with IR spectroscopy, strong IR absorption of most solvents used for electrolyte solutions is registered, and Raman spectroscopy is marred by a very low scattering yield (about 10−6 ). Both drawbacks can be overcome to some extent: thin films of electrolyte solutions (thin layer cells [TLCs]) combined with modulation (of electrode potential or polarization of the electric field of the IR radiation) techniques have enabled widespread application of in situ IR spectroscopy. The low photon yield of Raman spectroscopy can be hugely increased by surface effects associated with microscopic roughness features, related effects are operative with surface enhanced IR absorption [99, 100]. Combined with advanced instrumentation for both spectroscopies, even the unwelcome consequences of the use of TLCs and of roughened electrodes can be limited significantly. Various arrangements and cell constructions for in situ IR spectroelectrochemistry are shown in Figure 68.32; further setups have been reported in particular for IR spectroscopy [101]. Arrangements for external (top right) and internal reflection (attenuated total reflectance, bottom right) are depicted. For external reflection all sufficiently reflective materials of basically every thickness can be used. The IR beam has to pass through the electrolyte solution twice with associated losses in intensity. In the internal reflection mode, these losses are avoided; the beam arrives “from the backside” through an ATR element (IR transparent crystal like ZnSe) and the electrode, which is present as a very thin film on top of the ATR element. Single reflection and multiple reflections are possible. This limits the selection of electrode materials, only materials that can be sputtered, evaporated, or deposited as thin films otherwise. In all methodological variations, measured Raman spectra are absolute spectra; they do not require a reference spectrum because they are based on scattered radiation. Quite different IR spectroscopy being an absorption method always requires a reference spectrum needed to eliminate all unwanted spectral contributions of solvents, windows, energy distribution of the light source, and sensitivity distribution of the detector. In standard IR spectroscopy this is routinely done by recording a reference spectrum without the compound of interest. In surface spectroscopies, recording spectra without and with the (adsorbed) species may be an option, but most frequently polarization modulation is employed. Polarized light reflected at a surface shows a phase shift depending on the angle of incidence and the orientation of the electric vector of the incoming beam (probe). p-Polarized light with its electric vector oscillating perpendicular to the reflecting surface shows a phase shift, resulting in a substantial electric field at the point of reflection, whereas s-polarized light has no resulting electric field (Figures 68.33 and 68.34) [102]. The phase shift at angles close to 90∘ is at a value yielding a maximum effective field strength. Calculations of absorption coefficients Ap for p-polarized and As for s-polarized light based on the Maxwell equations assuming a layer of 1 nm of acetone on a reflecting surface show basically the close relationship between angle of incidence and phase shift (Figure 68.35).

c ZnSe window

a

b

f

From IR source

Brass flange To detector Threaded bolt d

PTFE insert e

Glass cell PTFE-guide Infrared beam

O-ring

Brass flange

Metal grid

ZnSe ATR element

Nut Counter electrode

Working electrode Fine thread screw

(a)

Infrared beam

(b)

Electrolyte solution Reference electrode

ZnSe ATR element

Figure 68.32 (a) TLC for external reflection in situ IR spectroscopy. (b) Top right: optical path used with the cell shown left. Bottom right: cell for in situ attenuated total reflection ATR in situ IR spectroscopy in the multiple reflection mode.

68 Surface Science in Batteries

Figure 68.33 Schematic illustration of phase shift of the electric field vector of polarized light reflected at a surface. (a) s-Polarized light. (b) p-Polarized light.

0 P

Δφ (°)

–50

–100

–150 S –200 0

22.5

45

67.5

90

θ (°) Figure 68.34 Calculated phase shift for s- and p-polarized at a reflecting surface. (Source: Based on data in [75].)

30 25 Absorption · 103

404

20 15 10 Ap

5

As

0 –5

0

25

50

75

100

θ (°) Figure 68.35 Calculated absorption coefficients As and Ap for a layer of 1 nm thickness of acetone on a reflecting surface as a function of the angle of incidence 𝜃. (Source: Based on data in [103].)

68.2 Experimental Methods

Adsorbed species showing a change of an electric dipole during a molecular vibration in the plane of reflection and correspondingly perpendicular to the reflecting surface can absorb energy from the IR light causing an absorption band. The difference in IR radiation absorption between s- and p-polarized light yields the desired absorption spectrum of the adsorbate [40]. This approach can also be applied in electrochemical in situ studies. Because the adsorption behavior of species at electrochemical interfaces frequently changes as function of electrode potential, modulation of this experimental variable is also employed. Interpretation of observed spectra is less straightforward. Two single-beam spectra (displays of reflectivity as a function of wavenumber) are ratioed against each other; the obtained difference spectrum is interpreted having in mind this procedure. The following possible combinations are shown schematically, and more combinations can be found elsewhere (Figure 68.36) [41, 104]. In case a, the absorption at the measurement potential Em is smaller than at Er (because of a lower degree of coverage) resulting in an upward pointing band in the display of ΔR/R. Generally resulting IR reflection absorption spectra (not singlebeam spectra as available with some methods) are displayed with bands above the baseline indicating higher absorption of any IR-active species at Er and downward pointing bands indicating higher absorption at Em . In case b, the situation is just reversed, resulting in a downward pointing band. In both cases determination of the band position is easy. Case c shows the major problem of this experimental approach: the obtained band shows the typical “differential shape,” making determination of the band position almost impossible. A rather rare case with a band considerably more narrow at Er than at Em is displayed in d. When complex mixtures of adsorbates or adlayers are present on an electrode, interpretation becomes even more difficult even when spectra of the single compounds are available for comparison. Sensitivities and detection limits of polarization and electrode potential

b

c

d

Reflection R or ΔRr/Rm

a

Wavenumbers (cm–1) Figure 68.36 Schematic infrared reflection absorption spectra, __.__.__ reflectance at E r ; _._. at E , ____ resulting spectrum; for details see text. m

405

406

68 Surface Science in Batteries

Laser light

Scattered light

Raman microscope Window Seal Working electrode Counter electrode

Figure 68.37 Schematic cross section of a spectroelectrochemical cell for in situ Raman spectroelectrochemistry using an optical microscope and backscattering geometry.

modulation approaches have been compared and discussed frequently; with respect to battery research, the polarization modulation approach has been suggested as more sensitive [105]. Experimental setups for Raman spectroscopy are rather straightforward in particular when aqueous electrolyte solutions are employed because water is a weak Raman scatterer, and thus thin layer arrangements are not requested. Both backscattering geometries are easily adaptable to common Raman microscopes and arrangements where the laser beam impinges on the electrode at an angle of incidence of about 60∘ , which has been identified as particularly effective in surface Raman spectroscopy [106–109]. There are numerous methodological variations of surface Raman spectroscopy depending on the properties of the adsorbate/surface modifier (e.g. surface resonance Raman spectroscopy in the case of strongly UV–Vis absorbing materials) [62]. Some of the experimental details are addressed below when experimental examples are presented (Figures 68.37 and 68.38). During some electrode reactions, in particular during corrosion of cell components, self-discharge/overcharge, and decomposition of components, volatile species may be formed. Mass spectrometry established as a powerful and versatile method for identification of compounds operates under UHV conditions and does not seem to be compatible with electrochemical in situ studies. Placing a hydrophobic porous membrane allowing transfer of low-molecular-weight volatile species keeping the electrolyte solution out has been suggested first by Bruckenstein and Gadde [110, 111]. Between the onset of a current flow caused by the electrode reaction and the appearance of a signal in the mass spectrometer, a delay time of about 20 seconds was reported. Wolter and Heitbaum suggested a more efficient method of vacuum generation (differential pumping), and further development enabled much shorter delay times, making the registration of a cyclic

68.2 Experimental Methods

GE

N2 AE

BE z x y K E

S

Z S

z x y

OA Figure 68.38 Spectroelectrochemical setup inlet; S: mirror, Z: cylindrical lens, K: camera and cell for in situ surface Raman studies. AE: lens; E: entrance slit of spectrometer; OA: working electrode, BE: reference electrode; optical axis of spectrometer. GE: counter electrode, N2 : nitrogen purge

voltammogram (CV) and the mass signal of selected mass values up to scan rates of 1 V/s. Because of the applied pumping technique and because of the proportionality between electrochemical current (i.e. derivative of consumed charge) and mass signal, the method was called differential electrochemical mass spectrometry (DEMS). A schematic cross section of a cell is shown in Figure 68.39. Working electrode contact

Electrochemical cell with electrolyte solution

Porous PTFE membrane Porous electrode Porous glass frit

Shrinkable PTFE tube

Ground glass joint NS 14.5

Bottom of electrochemical cell

To inlet system and vacuum Figure 68.39 Schematic cross section of a DEMS cell.

407

408

68 Surface Science in Batteries

Numerous other cell design enabling investigations of solid electrodes have been reported [40]. Given the large number of spectroscopies adopted for electrochemical (and preferably in situ) investigations, this brief overview must be focused on the most frequently employed one. Readers interested in a complete overview including rarely used methods and techniques applicable only for very rare tasks will consult the literature [40].

68.3 Overpotentials

All electrode reactions will be a sequence of some steps from the following generic scheme with the electronic conductor acting only as a sink/source for electrons (Figure 68.40). Of particular interest in the present context are charge transfer and adsorption/desorption steps because they tend to depend on the chemical identity of the surface, its chemical composition, its crystallography, roughness, and cleanliness – all properties amenable to surface science methods. Their rate may control in the case of particularly low values the overall rate of the heterogeneous reaction (rate-determining step [RDS]) and thus the electric current measured at the electrode. In the case of electrode reactions involving metal deposition (e.g. the negative electrode in a lead-acid battery), a different sequence will be observed (Figure 68.41). Basically all steps preceding the charge transfer reaction can be the same. Once the metal ion in its adsorbed state is reduced, the created adatom has to be integrated into the crystal lattices as a step of crystallization. From a surface energy point of view, not all places on a metal surface are equal. Edges, kinks, and local distortions Electronic conductor (electrode)

Homogeneous reaction

S

Charge transfer

S

Sbulk

Sad Adsorption/ desorption

Sad–

S–

S–

Diffusion

S–bulk–

Double layer Figure 68.40 General scheme of possible steps of an electrochemical reaction.

68.3 Overpotentials

Electronic conductor (electrode)

Homogeneous reaction

S

S

Charge transfer

Sbulk

Sad Diffusion Sad,red

Sred,dep

Adsorption/ desorption Adatom Double layer

Figure 68.41 General scheme of possible steps during a metal deposition/dissolution electrode reaction.

are in most cases energetically favored, and nucleation as the initial step of formation of a new lattice plane is the least attractive option. Consequently the adatom has to move around in search for an energetically attractive location. Surface diffusion may be a slow step, but in most cases only the energy requested for nucleation has been measured so far [112, 113]. The effect these steps have on the electrode potential is commonly observed as a deviation of the electrode potential E at a flowing current I from the rest potential E0 . This difference is named the overpotential 𝜂, and the effects of the various steps are the respective overpotentials: 𝜂 CT : the overpotential caused by the finite rate of the charge transfer. 𝜂 diff : the overpotential caused by slow diffusion of species either consumed or generated at the electrode surface. Based on the consideration provided above, it is given for a reduced species diffusing to the electrode for an electrooxidation by cred,0 R⋅T ln 𝜂diff = 𝛼a ⋅ n ⋅ F cred,s 𝜂 react : the overpotential caused by a slow chemical reaction preceding or following the charge transfer, thus limiting supply or removal of reactants. 𝜂 conc = 𝜂 diff + 𝜂 react : the overpotential caused by concentration differences between the electrode surface and a point in the bulk of the electrolyte solution. It is either due to slow diffusion (𝜂 diff ) or slow chemical reactions (𝜂 react ) or both. 𝜂 ads : the overpotential caused by slow adsorption or desorption reactions. 𝜂 cryst : the overpotential caused by slow crystallization of lead on a lead electrode in the charging mode of the negative electrode.

409

410

68 Surface Science in Batteries

As the sum 𝜂 = ΔE = E0 –E = 𝜂CT + 𝜂conc + 𝜂ads + 𝜂cryst is obtained. Because the voltage U of an electrochemical cell is composed of the electrode potentials E, the deviation of the cell voltage ΔU from its rest value U 0 observed without a flow of current and overvoltage can be defined: ΔU = U0 –U The current I flowing as an ionic or electrolytic one through the cell has to pass through the electrolyte (solution) between the electrodes. The ohmic resistance R (or more precisely Rsol ) of this component causes a voltage drop I⋅R, increasing the overvoltage ΔU and consequently decreasing the available cell voltage U: ΔU = U0 –U = (E0,2 –E0,1 )–(E2 –E1 –I ⋅ R) Of particular interest to surface science are the charge transfer, the adsorption/desorption, and the crystallization step at the electrode surface. The relationship between 𝜂 CT and the current j flowing across the electrochemical interface is described by the Butler–Volmer [114] equation: ( 𝛼a ⋅n⋅F ) 𝛼c ⋅n⋅F j = j0 e R⋅T  − e− R⋅T  with the exchange current density j0 given by j0 = nFk s cox e−

𝛼c ⋅n⋅F 𝜂CT R⋅T

= nFk s cred e

𝛼a ⋅n⋅F 𝜂CT R⋅T

The symmetry factors (or transfer coefficients) 𝛼 a and 𝛼 c specify the effect a change of electrode potential has on anodic and cathodic partial current adding up to j; they add up according to 𝛼 a + 𝛼 c . = 1. At rest electrode potential (j = 0 A/cm2 ), i.e. 𝜂 = 0 V, a simplification follows: j0 = nFk s cox 1−𝛼c = nFk s cred 𝛼a An increase of j0 can be effected by an increase of reactant concentration (not always a viable option) or by an increase of the rate constant k s . Depending on the type of electrode reaction, surface modifications in numerous different ways are possible; this is the field of electrocatalysis [115–117]. When evaluating improved performance noticed in terms of smaller 𝜂 CT , always plain increases of electrochemically active surface area should be kept in mind because j0 = j/A implies that smaller values of 𝜂 CT may be due to larger values of j0 at constant A and/or larger values of A at constant values of j0 [49].

68.4 Electrode Materials

Active masses in electrochemical systems are prepared from a wide variety of raw materials. Zinc and lithium are used as negative electrodes in primary batteries, while lead is used as the negative electrode in lead-acid batteries. The

68.4 Electrode Materials

electrochemical activity (a rather popular and imprecise term mostly designating large exchange current densities [15]) of zinc is large in neutral and slightly acidic electrode solutions as employed in the primary Leclanché cell; thus a smooth metal sheet can be used – the can of the battery. Because zinc is less noble than protons, the metal might react with water from the electrolyte solution, resulting in its dissolution (corrosion) and hydrogen evolution. This is equivalent to self-discharge and thus highly unwelcome. Although the exchange current density of the hydrogen electrode at the zinc electrode is low, the corrosion must be limited to values as small as possible. Initially this was done by adding the mercury salts to the electrolyte solution. Mercury is deposited by a cementation reaction and forms an alloy, and because the exchange current density of the hydrogen electrode at a mercury electrode is extremely small, hydrogen evolution is suppressed. Because of environmental concerns, this addition and alloying lead or cadmium into the zinc for the same purpose have been stopped. Now organic compounds adsorbing onto the zinc surface reducing the hydrogen evolution rate or various metal salts forming inhibiting deposits are used. Lead is used in elementary form as the negative electrode in the lead-acid battery. Large cell currents, i.e. high power, are possible at acceptably low electrode overpotential by increasing the electrochemically active surface area by using highly porous electrodes [118], thus enabling low actual current densities. When reducing the lead sulfate formed during discharge, redeposited lead is less porous with larger particle sizes. As a result the surface area decreases; in addition the larger particles may drop of the metal grid used as current collector. Expanders (containing also carbon black and barium sulfate in various amounts depending on the intended use of the battery) are added either to the battery acid solution or to the mixture prepared for the negative electrode. Lignosulfonic acid is derived from pulp control and direct metal deposition by adsorption on the lead surface. It also prevents passivation of the lead surface by coverage with lead sulfate. The very poorly soluble barium sulfate being isostructural with lead sulfate provides nuclei for lead sulfate deposition. Lithium very attractive as negative electrode material because of its extremely negative electrode potential [15] and its low weight (resulting in a high gravimetric charge density) is chemically highly reactive. Water-based electrolyte solutions cannot be used. Even electrolyte solutions based on aprotic solvents, both organic and inorganic ones, react with lithium, forming a surface layer (the SEI) protecting the lithium surface against further corrosion. Formation of this layer – which may also be considered as a special case of corrosion – consumes some lithium metal. In primary batteries the SEI limits the current flow immediately after connecting a load to the battery; when applying a constant load or current, the voltage will drop to low values initially (voltage delay). The SEI coating will be removed from the lithium surface upon discharge because of the intense transport processes associated with the flow of current at sites with thin SEI, pinholes, or other defects. After the load is disconnected, the SEI will be restored. Optimization of the electrolyte solution aiming at a highly protective coating with a minimum of lithium consumption is attempted. In secondary batteries these surface processes have dramatic, possibly devastating consequences. Lithium deposition during charging proceeds at defect sites of the

411

412

68 Surface Science in Batteries

1001.88 nm 57 nm

0.00 nm 500.94 nm

0 nm 0 nm

500.94 nm

Figure 68.42 Local nonuniformities during lithium deposition from an electrolyte solution of 0.5 M LiAsF6 in propylene carbonate recorded with an atomic force microscope.

1001.88 nm

(Source: Reprinted with permission from Cohen et al. 2000 [119]. Copyright 2000, American Chemical Society.)

SEI, and not in a homogeneous way, resulting in a desirably smooth electrode surface. The formed deposits are of a poor quality as electrode material. Even worse is the risk of dendrite formation (see light area in Figure 68.42) when localized deposition proceeds at high rates. When dendrites perforate the separator and reach the positive electrode, short circuits will result with high local current densities and local heating possibly high enough to cause decomposition of LiCoO2 . Thermal runaway, ignition, and explosion may follow. Numerous metal oxides are used in the positive electrode of batteries and supercapacitors. Different electrochemical reactions and associated changes of the materials are observed. In case of MnO2 in zinc-carbon and zinc-alkaline batteries reduction yielding Mn(III) and Mn(II) compounds are found. Compounds like LiCoO2 or LiFePO4 in the positive electrode of lithium-ion batteries serve only as host for lithium ions that intercalate during discharge of the battery with a corresponding reduction of the cobalt or iron. These various modes of operation can be summarized in the following overview. Although all processes are volume reactions, they always start at the electrochemical interface, and details of these heterogeneous reaction steps control the following reaction and the products. Consequently surface science methods are employed in their investigations.

68.5 Examples

Mode

Insertion Alloying Conversion

Degree of structural

Utilization

change

of material

Low Medium High

Low Medium High

Capacity

Example

Low Medium High

Graphite, LiCoO2 Tin Lead, lithium

Carbon in various forms [120] is used as electrode material for supercapacitors of the EDLC type. In graphitic form it is the negative electrode in lithium-ion batteries acting as the host for lithium-ion insertion/deinsertion [121, 122] and as simple electron source/sink-type electrode in RFBs [69].

68.5 Examples

Following selected examples of application of surface science, in particular of surface science methods, are presented. Studies reported so far do not provide a complete representation of all methods. Instead they reflect to some extent the scientific interest in practical challenges, and in a deeper understanding of observed problems, they also indicate the range of applicability of methods. Because in many battery electrodes those employed in RFBs are a notable exemption, electrode reactions may proceed at the surface, i.e. at the interface between the electrolyte solution and the topmost layers of the electrode, whereas the subsequent transformations extend far into the interior of the mass of the electrode, and a clear-cut separation between surface and volume aspects is impossible. Regarding the latter aspect supercapacitor electrodes with their redox reactions limited in particular at high currents to the top layers are an exemption again; in the EDLC type only surface processes occur (and no spectroelectrochemical studies with this type of supercapacitor are known). With supercaps employing redox reactions, whether they are correctly called supercapacitive or not, things might be different. At the time of writing only few spectroelectrochemical studies of working supercapacitor electrodes have been reported. 68.5.1 Aqueous Systems

The lead-acid battery providing a cell voltage substantially larger than the “electrochemical window of stability” of water (1.23 V) is enabled by the large hydrogen

413

414

68 Surface Science in Batteries

overpotential of the negative electrode. Instead of thermodynamically favored corrosion reaction between lead and protons yielding hydrogen gas and corrosion products, lead is oxidized during the discharge reaction and redeposited again during charge. Participating species have been identified with in situ IR spectroscopy [123–125]; in the external reflection mode, quantitative compositional changes could be monitored. The nickel hydroxide–oxyhydroxide electrode has been and still is an attractive object for intense research because of its remarkably fast electrode reaction when employed as a positive electrode in secondary batteries. It has been coupled successfully with various negative electrodes like metal hydride or cadmium [3]. In an attempt to reveal details of the electrode reaction mechanism(s), in situ Raman spectroscopy has been used [126]. For enhanced signal intensity, surface enhancement was obtained by depositing silver (see, e.g. Refs. [127–131]) into the freshly prepared Ni(OH)2 . Spectral deconvolution implied that both the α-phase and the β-phase were present in the freshly prepared material. During electrode potential (imitating charge/discharge) changes, in particular, a transition from the initially predominant α-phase into a disordered β-phase could be observed. An additional peak indicating formation of a further phase was seen. Distinct differences between chemically prepared α- and β-nickel hydroxides and electrochemically prepared samples were found. The effects of various additives to hydrous nickel hydroxide have been studied in the same way [132]. The presence of cobalt lowered the threshold of Ni(II) oxidation and inhibits the phase transitions mentioned above. Lithium improves the reversibility of the nickel electrode reaction and prevents capacity fading. According to obtained Raman spectra, cobalt ions occupy nickel ion sites, whereas lithium addition apparently causes more fundamental structural modifications, yielding additional Raman bands. Addition of both lithium and nickel results in a superposition of the effects mentioned above. 68.5.2 Lithium-Ion Batteries

Processes at battery electrodes for lithium-ion batteries and inside of them have been studied frequently using various spectroelectrochemical methods [133, 134]. Because of the rich chemistry started by the reductive reaction between lithium metal [134–140] and lithium ions intercalating into the negative graphite electrode and the organic electrolyte solution containing mostly a mixture of various solvents and electrolytes like LiPF6 , the chemical composition of the SEI tends to be extremely complex. Effects of traces of moisture possibly introduced during cell assembly via the various cell components and the repeated cycling of a cell add to this complexity. Already ex situ studies with ATR-IR spectroscopy of used electrodes (postmortem analysis) yielded no conclusive evidence regarding chemical compounds actually present [141]. In situ studies mostly in external reflection geometry have been reported frequently. Although metallic lithium is the candidate material of choice as the negative electrode in a rechargeable lithium battery, safety problems related to the dendritic

68.5 Examples

deposition of lithium metal from practically all relevant electrode solutions have limited its practical application. Because of the significant impact of a deeper knowledge of processes and structures at the lithium/organic solvent-based electrolyte solution, numerous studies of the lithium surface employing IR and Raman spectroscopy have been reported. Naudin et al. [142] employed confocal Raman spectroscopy and polarization-modulated IR reflection absorption spectroscopy in an investigation of metallic lithium exposed to dry argon and to air containing traces of carbon dioxide and water. In the latter case use of the laser Raman method yielded evidence of the formation of lithium acetylide (Li2 C2 ) because of local laser heating. Without heating, in the case of IR spectroscopy, carbonates and hydroxides were detected. In situ spectra of lithium and noble metal electrodes in contact with propylene carbonate (PC)-based (PC, γ-butyrolactone, various ethers) electrolyte solutions with 1 M LiClO4 were studied using external reflection and electrode potential modulation by Goren et al. [143]; the results demonstrate the feasibility of this experimental approach. Using IR spectroscopy and IR spectroscopy, lithium surfaces in contact with dry air and with a battery electrode were studied [105]. When exposed to dry air, only LiOH and LiCO3 were identified. When brought in contact with battery electrolyte, spectral features of lithium alkylcarbonate were observed. Taking into account further evidence from ex situ XPS, localization of the various reaction products could be verified. Lithium first cycle insertion and deinsertion into/out of microcrystalline graphite in contact with ordinary battery electrolyte solution have been studied in situ with Raman spectroscopy by Sole et al. [144]. The double resonance 2D band recorded for the first time in this study shifted from 2681 to 2611 cm−1 accompanied with a shape transformation into a single Lorentzian line when the electrode potential moved from 0.24 to 0.15 V, i.e. with an increasing degree of lithiation. Together with further observations this could be correlated with the staging process (observed with in situ Raman spectroscopy for the first time by Inaba et al. [145]), and the results provided further information on electronic structure and C–C bonding in particular of stage 3 and 4 graphite intercalation compounds (Figure 68.43). Electrochemically induced intercalation assuming flexible graphene layers in the host graphite deformed around lithium ions as suggested by the Daumas-Hérold model was found to be in agreement with the changes of the 2D band. For an overview, see [133, 146, 147]. Data on thickness and density of films formed

Stage 4 Graphene layer

Stage 3

Stage 2

Lithium

Figure 68.43 Stages of lithium intercalation into graphene according to the Daumas-Hérold model.

Stage 1

415

416

68 Surface Science in Batteries

on carbon in contact with organic solvent-based electrolyte solutions could be obtained from combined spectroscopic ellipsometry and electrochemical quartz crystal microbalance studies [148, 149]. Lithium intercalation and deintercalation at graphite [150, 151] (for another example of a layered compound see above) have been studied with vibrational spectroscopies to obtain an improved understanding of fundamental aspects of the ongoing processes. Because the complex chemistry of formation of the SEI and because of its major importance [152, 153] for the functioning of this type of batteries, further methods for analysis of gaseous reaction products have been employed (DEMS [19, 147, 154–164]). During formation of the SEI at the negative electrode in lithium and lithium-ion batteries and to a lesser extent at the positive electrode, decomposition of the various electrolyte solutions into solid deposits (the SEI) and to gaseous products has been observed. Hydrogen and ethylene were detected with DEMS [165]. Humidity introduced into lithium-ion batteries with various components provides an unwanted reactant H2 O also capable of reacting with the cell inventory releasing gaseous products [98]. A first total estimate can be obtained by pressure measurements [23]. More specific information can be obtained from mass spectrometric analysis of the products. Although this can basically be done by just studying the effluent of the cell, a more specific analysis of chemical identity and amount of gas evolved in particular as a function of electrode potential and time of operation is desirable. This can be achieved by coupling the working electrode – in the case of a sodium dioxygen secondary battery with sodium superoxide as the discharge product at the positive electrode [23]. At both beginning and end of charge process, undesirable formation of H2 , H2 O, and CO2 was detected at electrode potentials: ENa/Na+ > 3.5 V. Similar studies at lithium-dioxygen batteries have been reported [166]. Lithium peroxide (Li2 O2 ) could be verified as discharge product different from the sodium superoxide found in the sodium-air battery. Changes of crystalline structure and morphology of materials employed as positive electrodes [167] (mostly lithium transition metal oxides) have been studied with vibrational spectroscopies in order to improve the understanding of changes of the material and because uncontrolled reactions during overcharge have been of particular interest since they may become relevant for cell safety. Products of oxidation of PC and a mixed electrolyte solution solvent of dimethyl carbonate (DMC) and ethylene carbonate possibly being present during their way to the final CO2 have been monitored in situ with IR spectroscopy taking into account the influence of various amounts of moisture [168]. Numerous bands could be assigned to modes of various molecular fragments; beyond assignments to the solvents, no specific further intermediates could be verified. No CO2 was observed during oxidation of DMC. Inaba et al. [169] have used Raman spectroscopy to study structural changes of LiCoO2 . Results could be matched with results of X-ray diffraction data. After deintercalation, i.e. when the cell is charged again, remaining lithium ions stay randomly distributed on lithium lattice planes in the layered rock-salt structure. At the positive electrode of a lithium-ion battery, surface layer formation may also proceed, and reaction pathways and mechanisms differ from those observed

68.5 Examples

at the negative electrode. Matsui et al. [101] have investigated such layers formed on LiMn2 O4 and LiNi1/2 Mn3/2 O4 thin film electrodes prepared on gold substrates with an electrolyte solution of 1 M LiPF6 in a 1 : 1 mixture of ethylene carbonate and diethyl carbonate using in situ IR spectroscopy in the external reflection mode. Dependence of band intensities of the various carboxylic acid derivatives constituting the film on electrode potentials strongly suggests the participation of redox couples (Mn3+/4+ and Ni2+/4+ ). Stripping of the unstable surface film at electrode potentials 2000 m2/g)

Hybrid superstructures

Spin logic circuits Ultrafast low-power electronics

THz imager Mobility > 20 000 cm2/Vs on large area R2R Transparent (> 80%) conductive (Rs < 100 Ω/ ) films Transparent (> 80%) conductive (Rs ∼10 Ω/ ) films Conductive inks

Optical Medical repair modulation (0.05–10dB) Ultrafast optical Self-powered flexible response (< 5 fs)

mobile devices

THz oscillators RF A/D converters

Distributed sensor networks Fiber-optic communication system

Flexible electronics and optoelectronics

System integration

Spin valves Nanomagnets

kits

Lightweight batteries High-performance supercaps High-efficiency solar energy converters

Nonvolatile memories Interconnects in ICs Vertical tunnel FET arrays

Prostheses Mode-locked solid-state laser Modulators Photodetectors

Printed RF tags Foldable OLED Rollable E-paper Touchscreen and displays

Artificial retina

Food quality and safety biosensors Environmental DNA sensors sensors

Figure 71.3 Graphene application timeline. (Source: Ferrari et al. 2015 [11]. Reproduced with permission of Royal society for Chemistry.)

71.2 Mechanical

It is not surprising that mechanical properties are among the first to be utilized in real-world graphene applications. Although sheet density of single-layer graphene (SLG) is only 0.77 mg/m2 , SLG is difficult to implement in applications, so all mechanical applications so far have relied on GNPs. Submicron-thick sheets made by filtration of GNPs can have density on the order of 1.8 g/cm3 , which is comparable to the value for bulk graphite [12]. For comparison, aluminum has a density of 2.7 g/cm3 , titanium 4.5 g/cm3 , and steel 8.05 g/cm3 . Low density coupled to low cost is the reason graphite has been the material of choice for tennis rackets and other sports equipment for decades. However graphene offers additional advantages, like an extremely high Young’s modulus (leading to low deformation under tensile stress) and high breaking strength. The reported Young’s modulus for SLG is 1 TPa [13], while for structural steel it is “only” 200 GPa, iron 210 GPa, molybdenum 329 GPa, bulk graphite 27.6 GPa, and carbon nanotubes up to 950 GPa [14] (see Figure 71.4). Ultimate tensile strength (intrinsic strength, breaking strength) for SLG is 130 GPa [13], compared with 400 MPa [15] for steel and tens of megapascals for bulk graphite.

71.2 Mechanical

Graphene Carbon nanotubes 950 270 Ceramics 60

>1000

Porous ceramics 8 100 Metals and alloys 400

13 Composites 8

200

Wood 0.08

25 Polymers 10

Rubber 0.1 Polymer foams 0.5 0.01

0.1

1 10 Young’s modulus (GPa)

100

1000

Figure 71.4 Young’s modulus of various materials. Data for graphene from [13], for carbon nanotubes from [14]. (Source: Changgu Lee 2008 [13]. Reproduced with permission of American Association for the Advancement of Science.)

71.2.1 Strength

The first graphene-enhanced product on the market was a tennis racket from HEAD, introduced as early as 2012. The racket was used around that time by Novak Djokovic and Maria Sharapova, the world’s best tennis players at the time. Later the company Victor produced a graphene-enhanced badminton racket, which was used by the Korean Olympic team at the Olympics in London. In both cases GNPs were added in small amounts to the racket prepreg resin that formed the racket frame, although it seems that the composite material was primarily introduced into the rackets’ T-joints (the place where the shaft and the face of the racket meet). Presumably the carbon fibers traditionally used in that section of the racket align with the long axis of the racket, allowing for undesired face torque. The graphene-enhanced material should have omnidirectional strength enhancement, leading to less face torque. HEAD also claimed that by substituting some of the material in the shaft (middle part of the racket) with lightweight graphene nanoparticles, the weight is shifted toward the tip and handle, leading to enhanced ease of use [16]. Graphene technology is still applied to HEAD’s higher-end rackets.

517

71 Applications of Graphene

For reasons of mechanical superiority and ease of integration into polymer matrices, but also for reasons related to popularity of graphene, graphene-based materials are being pushed forward as lightweight strengtheners across the premium sports world. The University of Manchester collaborated with watchmaker Richard Mille to produce the first watch that contains graphene in the case that houses the watch mechanism, as well as in the rubber strap. The wristwatch is the lightest of its kind, weighing a mere 40 g. For comparison, high-end titanium watches weigh about 80 g, while standard steel ones weigh about 120 g. The racecar maker McLaren also participated in this development; thus we can expect some professional formula 1 drivers to be wearing this watch on their wrists [17]. The model RM 50-03 is extremely luxurious, with only 75 pieces made and a price tag of US$ 1 million. 71.2.2 Acoustic

The low density and high stiffness of graphene make it an ideal candidate for acoustic membranes. Vibrating membranes have been the cornerstone of acoustic technology in the last half a century, forming the essential components of microphones and speakers. Acoustic membranes need to be lightweight because a lightweight membrane leads to a smaller required acoustic pressure, which in turn can yield energy savings. They also need to be stiff to avoid unwanted oscillations. Also, a large tensile strength allows for large strain to be placed on the membrane, leading to extended microphone bandwidth. Graphene microphones have already shown an up to 32 times higher response to acoustic excitation compared with high-end commercial microphones (see Figure 71.5), for a membrane just 25 nm thin [18] and bandwidth extending to the ultrasonic part of the spectrum [19], which could lead to new applications in recording wildlife, for example, bats.

90 dB (comparison method)

518

80 70 60 50 40 30 102

B&K Type 4134 Graphene microphone 103

104

f (Hz) Figure 71.5 Acoustic response of a graphene microphone compared to high-end commercial microphone. (Source: Dejan Todorovi´c 2015 [18]. Reproduced with permission of IOP Publishing.)

71.3 Thermal

Although graphene acoustic membranes are a rather novel concept with only a few scientific publications on the topic [18–21], there is already a product on the market featuring real graphene oxide (GO) membranes. Namely, the company ORA Sound launched the sales of headphones that feature membranes made of GrapheneQ, a material that the company claims is based on GO. Because of the low mass of graphene membranes and thus the reduced need for energy required to actuate them, these wireless headphones have the longest battery life on the market. ORA Sound is still tweaking the composition of GrapheneQ, with the goal of producing a very stiff material for the tweeters, a very-low-density material for the midrange and full-range cones, and a thermally conductive material for micro-speakers. However it is already obvious that compared with commonly used Mylar and other acoustic membranes, graphene-based membranes have a higher Young’s modulus, a lower density, and much better thermal conductivity, which all result in an overall superior figure of merit (see Table 71.1). Also, the first deformation mode for GrapheneQ membranes is at a four times higher frequency than for Mylar membranes (see Figure 71.6).

71.3 Thermal

Aside from outstanding and unique electrical, mechanical, and optical properties, graphene has unprecedented thermal properties. Namely, the experimentally determined thermal conductivity of graphene is on the order of 5000 W/mK [22], which is significantly higher than for any other material (see Figure 71.7 [23, 24]). The extremely high thermal conductivity of graphene (and some other 2D materials) is likely due to the lack of out-of-plane scattering channels for acoustic phonons [25]. In fact, it has even been shown that the thermal conductivity of graphene keeps increasing with sample length, logarithmically diverging even for samples longer than the average phonon mean free path [26]. Table 71.1 Acoustic membrane material properties.

Titanium Aluminum Mylar Beryllium CVD diamond GrapheneQ

™

E (GPa)

𝝆 (kg/cm3 )

V S (m/s)

kT (W/mK)

120 70 3 300 1 050 30–130

4.5 2.7 1.39 1.84 3.5 0.800–1.800

5 164 5 092 1 897 12 769 17 321 6 120–8 500

16 240 0.15 216 1 800 1 500 (r-GQ)

FOM =

vs 𝝆

=

√

E 𝝆3

1.14 1.89 1.07 6.94 4.95 4.72–7.65

E is Young’s modulus, 𝜌 is density, V S is the speed of sound, k T is thermal conductivity, and FOM is figure of merit based on criteria established by the company. Source: Courtesy of ORA Sound.

519

520

71 Applications of Graphene

Mylar

Titanium

1

1.1

Aluminium Carbon fiber 1.8

GrapheneQ™ Ultra fidelity

3.2

CVD diamond

Berrylium

GrapheneQ

4.6

5.1

4.4–7.1

Mylar Speaker breakup

Figure 71.6 The first deformation mode of GrapheneQ is at frequencies more than four times higher than in Mylar membranes. (Source: Courtesy of ORA Sound.)

In most practical applications, such as heat sinks for silicon chips, graphene will be supported by and be integrated with insulators [27, 28]. Until researchers find a way to decouple graphene in-plane acoustic phonons from the substrate, scattering of phonons with the underlying surface will limit the practical application of graphene in thermal management applications. Nevertheless, some examples of early applications have already hit the market. Recently, China’s mobile giant Huawei announced the successful development of a graphene-assisted Li-ion battery (LIB) that holds charge for twice as long as ordinary LIBs [29]. Although the technology is proprietary, the company has indicated that graphene plays a thermal cooling role in those devices. It is likely that the spectacular thermal management capabilities of graphene are partially recovered by functionalization [27], which could play a leading role in graphene thermal management in composites. The battery remained functional at a temperature of 60 ∘ C, a temperature 10 ∘ C higher than the existing upper limit. Under the same operating conditions, the graphene-assisted high-temperature LIB is 5 ∘ C cooler than ordinary LIBs. Over 70% of the graphene-assisted battery’s capacity is left after it is recharged 2000 times at a temperature of 60 ∘ C. Less than 13% of its capacity is lost after being kept in a 60 ∘ C environment for 200 days. Angstron Materials, the company that first started to sell GNPs, markets a range of thermal management products based on graphene. Thermal pastes and foils are on standard offer in the catalog, with thermal conductivities on the order of 10 and 1700 W/mK, respectively. Although the thermal conductivity value for the paste may seem small, it is still higher than regular commercial thermal grease and thermal epoxy (see Figure 71.7). The company can also work with the customer

71.3 Thermal

Graphene 5000 Diamond 1000 Silver 406 Copper 385 Gold 314 Aluminum Stainless steel 11 35

205

Thermal epoxy 7 1 Thermal grease 0.7 3 Wood 0.3 0.1

1

10

100

1000

10000

Thermal conductivity (W/mK) Figure 71.7 Thermal conductivity of materials.

to formulate custom graphene/polymer composites for thermal management in applications such as LED housings, electronics housings, and heat sinks [30]. On offer are also paints and coatings for barrier and thermal applications. The company claims that they can engineer the alignment of the basal plane of the GNPs depending on application, making use of the fact that in-plane (i.e. along the surface) thermal conductivity of graphene is dominant compared with out of plane. The ability to engineer the alignment of GNPs is proving to be crucial for many applications, including mechanical (see tennis rackets, Section 71.2.1) and multifunctional (see bicycle tires, Section 71.6.1). Among graphene products that make use of thermal properties, graphene light bulbs are probably furthest up the value chain. Graphene Lighting, a Manchesterbased start-up run by an ex-student of Geim and Novoselov, produces lighting solutions enhanced with graphene [31]. The company uses graphene for heat dissipation of LEDs, getting rid of bulky metal heat sinks or complicated electronics. This allows the engineers to play with the design of bulb filaments, but more importantly the innovation is said to increase the LED circuit lifetime, to provide 360∘ uniform lighting (which is possible because of the removal of bulky heat sinks), to be of low cost and low energy, and to have high brightness. Commercial sales of the graphene LED bulbs have commenced in mid-2017.

521

522

71 Applications of Graphene

71.4 Electronic, Optical, and Optoelectronic

Electronic uses of graphene were the first that came to minds of many when it was first isolated in 2004 [2]. Properties such as a strong ambipolar field effect, large carrier mobility, and exotic effects like the quantum Hall effect [32] promised applications in high-speed electronics and new ways of computing. Although the electronic properties of graphene samples with high perfection of the atomic lattice are exceptional, structural defects often appear during the fabrication process and can limit the performance of real-world devices [33]. Furthermore, it was found that the substrate, especially polarizable substrates such as the standard SiO2 , can limit carrier dynamics in graphene [34]. Due to the need for high-quality graphene on nonstandard substrates, the high-frequency electronics applications are now being expected later than initially thought, beyond 2024 [11]. Nevertheless, certain applications that involve optical and electro-optic effects are already reaching the prototype stage of development. 71.4.1 Optical Communications

Optical communication formed the backbone of the Internet age and is expected to be equally pivotal for the developing 5G networks. Modern communications rely on optical links that fly information at the speed of light and on circuitry such as photodetectors and modulators that is able to encode a wealth of information onto these light beams. Although silicon is the material of choice for photonic waveguides on optical chips, photodetectors are made from other semiconductors such as GaAs, InP, or GaN, because silicon is transparent at standard telecom wavelengths. Integrating these other semiconductors with silicon is difficult, complicating fabrication processes and raising expenses. Also, thermal management is becoming a problem as photonic devices keep shrinking while using more power. Graphene is a promising material for telecom photodetectors, because it absorbs light over a large bandwidth, including standard telecom wavelengths. It is also compatible with CMOS technology [35], which means it can be technologically integrated with silicon photonics. Furthermore, graphene is an excellent heat conductor, promising a reduction in heat consumption of graphene-based photonic devices. For these reasons, graphene for optical communications has been an intense field of research, which is now gaining fruition in full working prototypes (see Figure 71.8). The first graphene photodetectors were developed in a research lab at IBM already in 2009 [36]. These transistor-based photodetectors had bandwidths exceeding 25 GHz and were subsequently used to transfer data over a 10 Gbit/s optical data link [37]. The efficiency of detection in those devices was improved by employing an asymmetric metal–graphene–metal transistor configuration. Analysis suggests that the bandwidth of such graphene photodetectors may ultimately exceed 500 GHz [36].

71.4 Electronic, Optical, and Optoelectronic

Graphene photodetector bandwidth 80 70 Speed (GHz)

60 50

CMOS compatible

40 30 20 10 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Year

Figure 71.8 Graphene photodetector historical bandwidth.

2013 was a productive year for graphene photodetector results. Several teams reported graphene photodetectors of different geometries [38–40], resulting in CMOS-compatible photodetectors that covered all communication bands at bandwidths up to 18 GHz [39]. In all these new realizations, graphene was positioned directly on top of silicon waveguides, and light was absorbed as it propagated down the waveguide. These were the first truly CMOS-compatible graphene photodetectors. In 2016, the bandwidth of CMOS-compatible graphene photodetectors reached 65 GHz [41], utilizing graphene/silicon pn junctions with potential bit rates of ∼90 Gbit/s. Already in 2017, graphene photodetectors with a bandwidth exceeding 75 GHz were fabricated in a 6′′ wafer process line [42]. These record-breaking devices were showcased at the Mobile World Congress in Barcelona in 2018, where visitors could experience the world’s first all-graphene optical communication link operating at a data rate of 25 Gbit/s per channel [43]. In this demonstration, all active electro-optic operations were performed on graphene devices. A graphene modulator processed the data on the transmitter side of the network, encoding an electronic data stream to an optical signal. On the receiver side, a graphene photodetector did the opposite, converting the optical modulation into an electronic signal. At the same show, Ericsson showed the first graphene-based optical ultrafast interconnection in mobile access networks, with a graphene-based photonic switch. Graphene-based integrated photonics are seen as a key area of future development, with potential for high-speed optical networks that use less energy than networks based on semiconductor photonics, while keeping the costs low and providing integration with existing technology. Rumors have spread in the second half of 2016 that Nokia will soon launch smartphones that feature graphene-enhanced cameras, supposedly to be manufactured by Foxconn. Nokia’s interest in graphene cameras is no secret, but market realization is yet to come. Apart from high sensitivity in low light conditions, graphene cameras are expected to provide infrared (IR) operation, which would be a novel feature in

523

524

71 Applications of Graphene

smartphones. The state of the art of graphene camera sensor technology is being developed at the Institute of Photonic Sciences (ICFO) near Barcelona, where it relies on arrays of hundreds of thousands of photodetectors based on graphene and quantum dots (QDs). The QDs make up for the low light absorption of graphene, while strong coupling between the QDs and graphene ensures a fast transfer of energy. In 2017, the team demonstrated a CMOS-integrated camera sensor that spans the ultraviolet, visible, and IR parts of the spectrum [35]. Rumors spread again in mid-2018 that Nokia 9, to be launched in 2019, will host a graphene camera, alongside its other four cameras. There has been no official confirmation from Nokia. 71.4.2 Batteries

Rapid development of mobile communication devices, electric vehicles, and other energy-hungry machines detached from landlines is stretching the capabilities of current battery technology (see also Chapter 68 in Volume 9). LIBs are today’s dominant technology due to their excellent cycle stability and good charge/discharge rates. However, the energy density packed in LIBs has reached its peak and is becoming a limiting factor for widespread use by mobile energy consumers. Energy density translates into charging speed, which is also highly sought after by consumers. Lithium has been a preferred component for portable energy solutions due to its low cost and natural abundance. Although research into improving LIBs certainly does not cease, future batteries are expected to be based on lithium–sulfur and lithium–air technology. Lithium–sulfur batteries, compared with today’s LIBs, hold potential to quadruple the specific energy stored, yielding substantially longer battery lifetime for all the world’s portable devices. Lithium–sulfur is leading the way for next-gen battery technology, due to initially simpler manufacturing, lower costs, and good performance. Moreover, lithium–sulfur batteries are technologically a step closer to lithium–air batteries that are expected to perform even better in the midrange future. The development of such new powerful batteries runs into challenges while using established electrode materials, because the new batteries require electrodes that support larger energy fluxes, to fully exploit the potential offered by the energy source itself. For example, the capacity of current electrode technology drops sharply when used with stronger batteries, so much that they offer only about 100 discharge cycles. Furthermore, cathode volume expansion and cathode cracking due to strong currents is a safety concern. GO offers an opportunity as a host in Li–S cathodes: S/graphene composites (see Figure 71.9). In GO/sulfur composites, graphene plays a significant role in improving the electronic conductivity of sulfur, inhibiting the shuttle effect of soluble polysulfides that causes cathode cracking in traditional cathodes. Here in particular the extremely large surface-to-volume ratio of graphene offers large benefits, because it offers more opportunities for sulfur atoms to bind per unit volume. GO-based cathodes have shown to be more durable and efficient than traditional ones in lithium–sulfur battery technology [45]. Apart from electrode material, graphene holds potential as a component of other battery parts [46].

71.4 Electronic, Optical, and Optoelectronic

S-rGO

Sulfur Figure 71.9 The role of graphene in lithium–sulfur batteries. (Source: Vinayan and Zhirong 2016 [44]. Reproduced with permission of the Royal Society of Chemistry.)

In practice though, there are few commercial battery products that contain graphene. Vorbeck advertised a graphene-enhanced flexible battery strap for laptop carrier bags, the Vor-Power battery strap already in 2014 [47], but the company seems to have discontinued the product since. Vorbeck has applied for several US patents to protect the use of graphene in electrodes; thus it is very likely that graphene formed part of the electrode in this product. 71.4.3 Transparent Conductors

TCs are ubiquitous today, because they are used as the top touch sensitive layer in all touchscreens. These layers need to be optically as transparent as possible with electrical sheet resistance as small as possible, made from naturally abundant materials that do not cost much, are easy to deposit in thin sheets, and are chemically stable over time. The predominantly used ITO, a metal oxide that is transparent throughout the visible part of the spectrum, has been the TC of choice for decades because it has satisfactory transparency, has a small sheet resistance, and is chemically stable. However, ITO is said to be among the costliest parts of mobile devices, in part because indium is one of the least abundant metals on Earth. Furthermore, ITO is brittle, which is not in line with the planned development of flexible electronic devices; thus the search for ITO replacements is becoming increasingly important. Graphene naturally arises as a candidate substitute for ITO. The high optical transparency (2.3% absorbance by a single layer), flexibility, and low sheet resistance make graphene an ideal candidate for TC applications, in theory. In fact, depending on how it is produced, graphene reaches sheet resistance nearly as low as that of ITO and main alternative competing material silver (see Figure 71.10, data from [48], and references therein). The problem thus far, as with all other electronic and optoelectronic applications of graphene, has been the production of sheets with consistently high quality and the cost of making graphene. As production costs keep decreasing and consistency increasing, it is to be expected that graphene becomes a plausible solution for TC applications.

525

71 Applications of Graphene

1 Optical transmittance (%)

526

0.8

0.6 Graphene ITO Ag nanowires PEDOT: PSS CNTs Thin Ag layer

0.4

0.2 1

10

100

1000

Sheet resistance (Ω/ ) Figure 71.10 Suitability of some materials for transparent conductor applications. (Source: Cao et al. 2014 [48]. Reproduced with permission of Society of Photo-Optical Instrumentation.)

In terms of real-world applications, there have been numerous rumors of phones featuring graphene in their screens (including one by Moxi, a Chinese company that announced in mid-2016 a smartphone that one can wrap around a wrist like a watch); however it is still unclear whether any of those rumors will materialize in the near future.

71.5 Barrier, Membrane, and Chemical

2D materials are a perfect platform for chemical and biological sensing because of their extreme surface-to-volume ratios. Sheets of 2D material consist practically of only the surface, which makes, for example, charge transport through narrow 2D devices extremely sensitive to adatoms or species that attach to defects in the 2D crystal lattice. The same physical principle is used for graphene-based capture of greenhouse gases, such as CO2 – the passing molecules of greenhouse gas get attached to the graphene surface, allowing clean air to flow through a graphene filter. Trace gas detection and capture with graphene sensors are fast-developing fields of research, with the bulk body of applications expected still in the longer term, mainly due to difficulty of controlling device quality and defect density. The most striking demonstration of trace gas detection with graphene came in 2007 by Schedin and Geim [49], who demonstrated detection of single molecules adsorbed on graphene. The adsorbed molecules change the carrier concentration in graphene on a single-electron level, which is detected in a Hall bar configuration (see Figure 71.11). The sensitivity of such graphene gas detectors is 1 ppb, on par with other sensor types. Graphene sensors have evolved toward defect-based sensors, in which gas molecules get attached to defects in graphene. In this respect

71.5 Barrier, Membrane, and Chemical

a

Changes in ρxy (Ω)

Adsorption 30

1e 20

1e Desorption

10

0 0

200

400

600

t (s) Figure 71.11 Graphene trace gas sensor. (Source: Schedin and Geim 2007 [49]. Reproduced with permission of Society of Springer Nature.)

graphene of a lower quality, such as reduced GO, makes for a better sensor than pristine graphene [50]. Most graphene gas detectors to date have been tested in controlled conditions only, mainly because humidity and components of air strongly affect the Fermi level in graphene, completely overshadowing the effect that trace gases exert. Although the automaker Honda is working on self-cleaning graphene sensors [51] that promise sensitivities in the range of parts per quadrillion, results are still in the form of scientific achievements [52] and not in real products. However, machine learning is now increasingly being used to improve sensor selectivity [53]. The company Graphenea is one of a few to market a graphene sensor platform aimed at lowering the barriers to adoption for researchers that wish to work with sensors but do not wish to spend resources developing their own devices (see Figure 71.12). The sensors are based on graphene field effect transistors (GFETs) in a configuration much like that demonstrated by Schedin et al. (see Figure 71.13).

Figure 71.12 GFET sensors from Graphenea on a wafer.

527

528

71 Applications of Graphene

Graphene channel Metal electrode (source)

Metal electrode (drain)

SiO2 Vs

I Silicon substrate

VG

Figure 71.13 GFET schematic. (Source: Courtesy of Graphene Frontiers.)

The resistance of the graphene channel changes in response to a changing environment. The company suggests uses in graphene device research, bioelectronics, photodetectors, chemical sensors, magnetic sensors, and biosensors. To satisfy the needs of a broad spectrum of researchers, Graphenea makes several standard chips with 2-probe, 4-probe, and Hall bar devices, with optional custom design per request. This product holds an interesting position on the graphene market – its technology readiness level is geared somewhere between bare graphene and product for mass use. An example of graphene gas capture research, EU-funded project GRAMOFON (2016–2020) aims to use graphene nanostructures for capturing CO2 emission from power plants and other industrial polluters [54]. The pollutant is captured by adsorption onto the graphene surface. Most other methods of capturing CO2 emission are based on absorption in liquid, which subsequently needs to be subjected to further chemical processing in order to isolate the CO2 and ship it to storage. The extra chemistry step is energy consuming and costly and would nearly double the cost of electricity for plants that employ liquid-based CO2 capture [55]. The claim of the project GRAMOFON is that the adsorbed CO2 can be removed from graphene in a facile and economical manner by treatment with microwave radiation. The project lies at technological readiness level 2–5, i.e. somewhere between research to prove feasibility and early technology demonstration. In other environmental applications, graphene sheets perforated with tiny holes are being studied as water desalination sifts. Already in 2012 scientists used molecular dynamics to predict [56] that graphene sheets with pores of diameters on the order of 1 nm would be very effective in filtering NaCl from water. Water permeability of such membranes was predicted to be several orders of magnitude higher than conventional reverse osmosis membranes, which would offer tremendous energy savings if such membranes could be made. To date, however, controllably introducing nanometer-sized pores in graphene remains technologically impossible. It is interesting that Lockheed Martin, the infamous defense company, hurried to patent this technology and advertise the product Perforene, a nanoporous graphene membrane for water treatment. The membrane promises

71.6 Multifunctional/Hybrid

over 5× improvement in flux, 10–20% energy reduction, and a fouling reduction up to 80% for improved membrane lifetime [57]. It seems, however, that Perforene is still at the moment a technological dream, waiting for new fabrication techniques capable of producing 1 nm holes at will over large areas. Such fabrication methods are quickly ripening though, with recent reports of size-tunable atomic pores in suspended graphene [58].

71.6 Multifunctional/Hybrid 71.6.1 Sporting and Aerospace

Sporting goods are definitely proving to be fertile ground for harvesting the advantages of graphene, in particular the material’s lightweight, flexibility, and high breaking strength. Among the first companies to tap into graphene’s strength was Italian bicycle tire maker Vittoria. Incorporating GNPs made by Directa Plus (also an Italian manufacturer, partly owned by Vittoria) into its tire rubber apparently made Vittoria’s bike tires better for heat dissipation, lighter, more impact resistant, flexible, and laterally stiff. The manufacturer claims that GNPs, when aligned with the tire rolling direction, reduce friction between the tire and the surface, offering a lower rolling resistance. When orthogonal to the surface, however, the exposed nanoplatelet edges offer better grip by interacting strongly with the surface (it is well known that edges of graphene have an abundance of chemically reactive sites, which is probably what causes better grip). Nanoplatelets can be purposely oriented parallel to the road surface for minimum rolling resistance during cruising, but as braking causes tire deformation, the platelets rotate toward a direction orthogonal to the substrate, enhancing braking efficiency. Depending on the application and riding conditions (mountain or road bike, dirt, dry, or wet), the company can also engineer the orientation and density of GNPs in the tread and body, thus offering an entire product range of different graphene-enhanced tires. In recent months, Directa Plus has applied its G+ graphene products in lightweight bicycle wheels, smart textiles, and rubber hoses. The same physical principles and similar composites are being applied to enhanced sporting shoes. Vorbeck applied to patent graphene-enhanced footwear. Graphene is embedded in standard polymers and elastomers to enhance the strength of the shoe while maintaining a small mass and keeping costs down compared with, for example, adding more polymer or enhancing with metal. The UK-based company inov-8 has taken a different approach, applying graphene as part of the sole rubber composite in order to improve grip (see Figure 71.14). The rubber composite was developed in collaboration with the University of Manchester, and apart from improving grip it is claimed to improve strength, flexibility, and wear resistance.

529

530

71 Applications of Graphene

Figure 71.14 Graphene-enhanced shoe. (Source: Courtesy of inov-8.)

Vittoria is also incorporating graphene into the rims and looking for ways to use it in the body of bicycles as well. Meanwhile the UK-based company Dassi Bikes [59] announced the introduction of graphene-laced bicycle frames. Introducing graphene decreased weight, improved stiffness, increased strength, and boosted aerodynamics of the first 25 specialty bikes produced. The company is now considering using the electrical conductivity of graphene to turn certain parts of the frame into sensors that monitor the biomechanical efficiency of the rider, effectively a step in the direction of “smart bikes.” Graphene in sports equipment seems to come natural to Italian manufacturers. An Italian-built graphene-enhanced motorcycle helmet, launched in October 2016, also combines the mechanical and thermal benefits of graphene. Collaboration of the Italian Institute of Technology (IIT) and the luxury design company Momodesign yielded a helmet that was produced in a series of 3000 pieces using a modified process on the company’s existing production line. Graphene is incorporated as a coating on the outer part of the helmet, providing better heat dissipation that leads to more user comfort. Perhaps even more importantly, the manufacturer claims that the graphene adds to the mechanical protection against impact. The manufacturing process uses graphene powder with few- and multilayer graphene flakes, or GNPs. This powder is subsequently mixed into a solution and then spray-coated on the exterior shell of the helmets. In the near future, the researchers will be working to add graphene into the inner plastic materials of the helmet, aiming to achieve the same level of safety with a thinner helmet and improve comfort for the wearer. The graphene-enhanced unmanned aerial vehicle (UAV or drone) developed by two UK universities is a poster example of how graphene can be used in the vast aerospace industry. The UAV named Prospero (see Figure 71.15) houses graphene in its wings and propeller, yielding drag reduction (most likely by the same principle

71.6 Multifunctional/Hybrid

Figure 71.15 Prospero, a graphene-enhanced UAV on its first public flight in July 2016. (Source: Courtesy of the University of Central Lancashire.)

by which friction is reduced in graphene bicycle tires), reduced weight, and improved thermal management compared to conventionally skinned carbon fiber wings. The wings also show increased impact resistance of up to 60%, as claimed by the Engineering Department of the University of Central Lancashire, which made the drone together with the University of Manchester’s National Graphene Institute. Coating an aircraft with conductive graphene could improve lightning strike protection in the near future. With aerospace giants like Airbus and Boeing having already shown interest in graphene through patents of graphene composite materials and public showings, it is to be expected that graphene will play an important role in aircraft construction materials within just a few short years. The same combination of lightweight, reduced friction, and improved thermal management holds promise for applications in other types of vehicles as well as windmills and hydro turbines. 71.6.2 Printing and 3D Printing

The demand for printed electronics is growing rapidly as technology visionaries eye applications such as wearable electronics, printed radio-frequency (RF) tags, flexible mobile devices, wearable healthcare sensors, etc. In parallel and at the same time, 3D printing enthusiasts are promising a new industrial revolution that will enable the production of goods in local micro-factories that use collections of custom-made printing filaments that match their specific product needs. Interestingly, those two markets, both relying on emergence of novel materials, are expanding exactly at a time when graphene technology is maturing and becoming ready for use in masterbatches. The requirements that flexible electronics and 3D printing place on new materials – flexibility, conductivity, optical transparency, mechanical stiffness, and small density – are very much compatible with properties of graphene (see transparent conductors). Graphene for printed electronics is dominantly produced with LPE, a method of separating the layers of graphene from bulk powdered graphite [60]. During LPE, graphite powder is immersed in a solvent and bombarded with ultrasound, which aids in separating the individual layers of graphene (see Figure 71.16). The exfoliated material is separated from the leftover bulk material by centrifuge. The graphene

531

532

71 Applications of Graphene

Layered material in bulk

Sonication in high boiling point organic solvent

Sonication

Liquid dispersion of 2D material

Centrifugation

Figure 71.16 Liquid-phase exfoliation.

flakes produced with this method are typically 1–3 nm thick, with lateral dimensions of 50–150 nm [61]. The graphene flakes dispersed in solution are used as ink for piezoelectric inkjet printing or gravure/screen printing. Various suppliers now offer graphene ink in different solvents, each having different characteristics such as viscosity, volatility, toxicity, graphene density, flake size, etc. Printed graphene electronics has found applications in radio-frequency identification (RFID) tags, RF antennas, and printed sensors. The perceived key advantages that graphene offers for these applications are optical transparency, flexibility, and a potential for low cost. Nevertheless, graphene technology for printed electronics faces competition that will be hard to overcome, such as ITO-on-PET, metal mesh, and silver nanowires, which share many of the same benefits but have established supply chains, product lines, and lower cost. Another related market is graphene-enhanced filaments for 3D printing. These are filaments from standard 3D printing polymers with GNPs embedded into the polymer matrix. The graphene tends to be used as an element that provides electrical conductivity, allowing for 3D printed electric circuitry and printed sensors. Again, a direct comparison with competing technology reveals that graphene does not add much in terms of specific features, such as volume resistivity, which is on the order of ∼0.6–0.8 Ω cm for all materials, whereas the price of graphene filaments is still several times higher than other solutions.

71.7 Outlook

Graphene in its purest form, i.e. a monolayer of carbon atoms arranged in a hexagonal lattice, holds promise for a myriad of applications due to its superior electrical, mechanical, thermal, and optical properties. If one considers also GNPs that have high chemical reactivity and can be made in large quantities at a low cost and also GO that shares many qualities with graphene but is even easier and cheaper to make and integrate into composites, the possibilities for application become extremely plentiful, so much so that graphene has rightfully been called “the new plastic.” Beyond 2019, the Graphene Flagship anticipates [11] applications such as foldable OLED, rollable e-paper, flexible touchscreens, printed RF tags, highly sensitive

References

IR photodetectors, food quality and safety biosensors, environmental sensors, DNA sequencers, lightweight batteries, high-performance supercapacitors, highefficiency solar energy converters, artificial limbs, RF A/D converters and THz oscillators, ultrafast computer memory, graphene interconnects and field effect transistor (FET) arrays, and finally spin valves and nanomagnets. Such applications that mostly rely on high-quality CVD graphene should witness rapid growth in the coming years, as the price of graphene is no longer a barrier for market adoption. CVD graphene has reached the same range of prices as semiconductor silicon (Jesus de la Fuente, private communication). Although the temporal development of technology is impossible to predict, the Flagship roadmap should be trusted as a source of realistic applications to come in the next decade. The timeline of the roadmap will be shifted due to unexpected technological breakthroughs, such as the use of graphene as an additive to 3D printing filaments, which opens the doors of research to many private innovators, also known as “basement inventors.” Some other achievements may be delayed due to technological obstacles or may never reach the market because of their cost and the rise of more competitive technologies. Nevertheless, it is safe to say that graphene has already entered products in several niche markets, exploiting virtually all of the material’s beneficial properties. Real large-scale use will come in due time, especially as the aerospace, automotive, electronics, and construction markets seriously start using new graphene-based products.

References vapor deposition and transfer process. Appl. Phys. Lett. 102 (2): 023112. et al. (2016). Accurate thickness measurement of graphene. Nanotechnology 7. Hernandez, Y., Nicolosi, V., Lotya, M. 27: 1257e04. et al. (2008). High-yield production of graphene by liquid-phase exfoliation of Novoselov, K.S., Geim, A.K., Morozov, graphite. Nat. Nanotechnol. 3: 563–568. S.V. et al. (2004). Electric field effect in atomically thin carbon films. Science 306 8. Kovtyukhova, N.I., Ollivier, P.J., Martin, (5696): 666–669. B.R. et al. (1999). Layer-by-layer assembly of ultrathin composite films from Somani, P.R., Somani, S.P., and Umeno, micron-sized graphite oxide sheets M. (2006). Planer nano-graphenes from and polycations. Chem. Mater. 11 (3): camphor by CVD. Chem. Phys. Lett. 771–778. 430: 56–59. Bae, S., Kim, H., Lee, Y. et al. (2010). 9. Wolf, E.L. (2014). Practical productions Roll-to-roll production of 30-inch of graphene, supply and cost. In: Appligraphene films for transparent eleccations of Graphene (ed. E.L. Wolf), trodes. Nat. Nanotechnol. 5 (8): 19–38. Cham, Switzerland: Springer 574–578. International Publishing. Li, X., Cai, W., An, J. et al. (2009). 10. Wei, J., Atif, R., Vo, T., and Inam, F. Large-area synthesis of high-quality and (2015). Graphene nanoplatelets in epoxy uniform graphene films on copper foils. system: dispersion, reaggregation, and Science 324: 1312–1314. mechanical properties of nanocomposites. J. Nanomater. 2015: 561742. Kobayashi, T., Bando, M., Kimura, N. et al. (2013). Production of a 100-m11. Ferrari, A.C., Bonaccorso, F., Fal’ko, long high-quality graphene transparent V. et al. (2015). Science and technolconductive film by roll-to-roll chemical ogy roadmap for graphene, related

1. Shearer, C.J., Ashley, S.D., Stapleton, A.J.

2.

3.

4.

5.

6.

533

534

71 Applications of Graphene

12.

13.

14.

15.

16.

17.

18.

19.

20.

21.

22.

23. 24. 25.

two-dimensional crystals, and hybrid systems. Nanoscale 7: 4598–4810. Applied Nanotech, Inc. (2018). Graphene films and foils. http://www .appliednanotech.net/graphene-fims/ (accessed 24 May 2019). Changgu Lee, X.W. (2008). Measurement of the elastic properties and intrinsic strength of monolayer graphene. Science 321 (5887): 385–388. Yu, M.-F. and Lourie, O. (2000). Strength and breaking mechanism of multiwalled carbon nanotubes under tensile load. Science 287 (5453): 637–640. Callister, W. (2006). Materials Science and Engineering: An Introduction, 7e. Wiley Publishers. HEAD (2017). Discover a new era of graphene. http://www.head.com/en/ sports/tennis/technology/graphene-xt (accessed 24 May 2019). Richard Mille (2017). RM 50-03. https:// www.richardmille.com/en/collections/ rm-50-03 (accessed 24 May 2019). Dejan Todorovi´c, A.M. (2015). Multilayer graphene condenser microphone. 2D Mater. 2 (4): 045013. Qin Zhou, J.Z. (2015). Graphene electrostatic microphone and ultrasonic radio. Proc. Natl. Acad. Sci. U.S.A. 112 (29): 8942–8946. Woo, S.T., Han, J.-H., Lee, J.H. et al. (2017). Realization of a high sensitivity microphone for a hearing aid using a graphene–PMMA laminated diaphragm. ACS Appl. Mater. Interfaces 9 (2): 1237–1246. Zettl, Q.Z. (2013). Electrostatic graphene loudspeaker. Appl. Phys. Lett. 102 (22): 223109. Balandin, A.A., Ghosh, S., Bao, W. et al. (2008). Superior thermal conductivity of single-layer graphene. Nano Lett. 8 (3): 902–907. Haynes, W.M. (2014). CRC Handbook of Chemistry and Physics, 95e. CRC Press. Young, H.D. (1992). University Physics. Reading, MA: Addison Wesley. Yan, Z., Nika, D.L., and Balandin, A.A. (2015). Thermal properties of graphene and few-layer graphene: applications in electronics. IET Circuits Devices Syst. 9 (1): 4–12.

26. Xiangfan Xu, L.F. (2014). Length-

27.

28. 29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

dependent thermal conductivity in suspended single-layer graphene. Nat. Commun. 5: 3689. Han, H., Zhang, Y., Wang, N. et al. (2016). Functionalization mediates heat transport in graphene nanoflakes. Nat. Commun. 7: 11281. Prasher, R. (2010). Graphene spreads the heat. Science 328 (5975): 185–186. Huawei (2016). Huawei achieves major breakthrough in graphene-assisted high temperature Li-ion batteries. http:// www.huawei.com/en/news/2016/12/ Graphene-Assisted-Li-ion-Batteries (accessed 24 May 2019). Global Graphene Group (2018). Graphene thermal paste. https://www .theglobalgraphenegroup.com/thermalpaste (accessed 24 May 2019). Graphene Lighting (2017, February 23). Graphene Lighting – See the world in a better light. http://graphenelighting.co (accessed 24 May 2019). Zhang, Y., Tan, Y.-W., Stormer, H.L., and Kim, P. (2005). Experimental observation of the quantum Hall effect and Berry’s phase in graphene. Nature 438: 201–204. Banhart, F., Kotakoski, J., and Krasheninnikov, A.V. (2011). Structural defects in graphene. ACS Nano. 5 (1): 26–41. Fratini, S. and Guinea, F. (2008). Substrate-limited electron dynamics in graphene. Phys. Rev. B 77: 195415. Goossens, S., Navickaite, G., Monasterio, C. et al. (2017). Broadband image sensor array based on graphene–CMOS integration. Nat. Photonics 11: 366–371. Xia, F., Mueller, T., Lin, Y.-m. et al. (2009). Ultrafast graphene photodetector. Nat. Nanotechnol. 4: 839–843. Mueller, T., Xia, F., and Avouris, P. (2010). Graphene photodetectors for high-speed optical communications. Nat. Photonics 4: 297–301. Gan, X., Shiue, R.-J., Gao, Y. et al. (2013). Chip-integrated ultrafast graphene photodetector with high responsivity. Nat. Photonics 7: 883–887.

References 39. Pospischil, A., Humer, M., Furchi,

40.

41.

42.

43.

44.

45.

46.

47.

48.

49.

M.M. et al. (2013). CMOS-compatible graphene photodetector covering all optical communication bands. Nat. Photonics 7: 892–896. Wang, X., Cheng, Z., Xu, K. et al. (2013). High-responsivity graphene/siliconheterostructure waveguide photodetectors. Nat. Photonics 7: 888–891. Schuler, S., Schall, D., Neumaier, D. et al. (2016). Controlled generation of a p–n junction in a waveguide integrated graphene photodetector. Nano Lett. 16 (11): 7107–7112. Schall, D., Porschatis, C., Otto, M., and Neumaier, D. (2017). Graphene photodetectors with a bandwidth >76 GHz fabricated in a 6′′ wafer process line. J. Phys. D: Appl. Phys. 50: 124004. Graphene Flagship (2018). Ultra-fast graphene photonics for next generation datacomms at the Graphene Pavilion at the Mobile World Congress. https://graphene-flagship.eu/ultra-fastgraphene-photonics (accessed 24 May 2019). Vinayan, B.P. and Zhirong, Z.K. (2016). Performance study of magnesium–sulfur battery using a graphene based sulfur composite cathode electrode and a nonnucleophilic Mg electrolyte. Nanoscale 8: 3296–3306. Nitze, F., Agostini, M., Lundin, F. et al. (2016). A binder-free sulfur/reduced graphene oxide aerogel as high performance electrode materials for lithium sulfur batteries. Sci. Rep. 6: 39615. CheapTubes (2016, December 21). Understanding graphene batteries. http://www.azonano.com/article.aspx? ArticleID=4326 (accessed 24 May 2019). Vorbeck (2014, June 12). Vorbeck materials debuts flexible battery straps. https://vorbeck.com/pages/zpr_ 20140512_batterystraps (accessed 24 May 2019). Cao, W., Li, J., Chen, H., and Xue, J. (2014). Transparent electrodes for organic optoelectronic devices: a review. J. Photonics Energy 4 (1): 040990. Schedin, F. and Geim, A.K. (2007). Detection of individual gas molecules adsorbed on graphene. Nat. Mater. 6: 625–655.

50. Ricciardella, F. and Massera, E. (2016). A

51.

52.

53.

54.

55.

56.

57.

58.

59.

calibrated graphene-based chemi-sensor for sub parts-per-million NO2 detection operating at room temperature. Appl. Phys. Lett. 104: 183502. Graphene Tracker (2014, November 12). Honda applies for patent for graphene-based gas sensors. http://www .graphenetracker.com/honda-applied-forpatent-for-graphene-based-gas-sensors (accessed 24 May 2019). Lv, R., Chen, G., Li, Q. et al. (2015). Ultrasensitive gas detection of B-doped graphene. Proc. Natl. Acad. Sci. USA 112 (47): 14527–14532. Wiederoder, M.S., Nallon, E.C., Weiss, M. et al. (2017). Graphene nanoplateletpolymer chemiresistive sensor arrays for the detection and discrimination of chemical warfare agent simulants. ACS Sens. 2 (11): 1669–1678. European Commission (2017, August 4). New process for efficient CO2 capture by innovative adsorbents based on modified graphene aerogels and MOF materials. https://cordis.europa.eu/ project/rcn/205816_en.html (accessed 24 May 2019). Metz, B., Davidson, O., de Coninck, H. et al. (eds.) (2005). IPCC special report on carbon dioxide capture and storage. In: Working Group III of the Intergovernmental Panel on Climate Change, 343. Cambridge: Cambridge University Press. Grossman, D.C.-T. (2012). Water desalination across nanoporous graphene. Nano Lett. 12 (7): 3602–3608. Lockheed Martin (2015). Perforene datasheet. http://www.lockheedmartin .com/content/dam/lockheed/data/ms2/ documents/Perforene-datasheet.pdf (accessed 1 March 2017). Madauss, L., Schumacher, J., Ghosh, M. et al. (2017). Fabrication of nanoporous graphene/polymer composite membranes. Nanoscale 9: 10487–10493. Dassi (2016, January 17). Dassi strengthens innovation in UK bike building, with graphene. https://dassi.com/dassi-blog/ item/167-dassi-strengthens-innovationin-uk-bike-building-with-graphene (accessed 24 May 2019).

535

536

71 Applications of Graphene 60. Tomaševi´c-Ili´c, T., Peši´c, J., Miloševi´c, I.

et al. (2016). Transparent and conductive films from liquid phase exfoliated graphene. Opt. Quantum Electron. 48 (6): 319. 61. Secor, E.B. and Hersam, M.C. (2014). Graphene Inks for Printed electronics.

Sigma Aldrich https://www.sigmaaldrich .com/technical-documents/articles/ technology-spotlights/graphene-inksfor-printed-electronics.html (accessed 24 May 2019).

537

72 Industrial Heterogeneous Catalysis Ekkehard Schwab

72.1 Introduction

Catalysis is a prominent example where the properties of surfaces are put into practical application. Catalysis was not invented. Actually, it is a phenomenon which was observed and exploited long before the word was used to describe its underlying scientific reality. It is not only the often cited biocatalytic transformation of sugar into ethanol that goes back far into history but also the invention of soap making by boiling animal fat in water and in the presence of wood or seaweed ashes: the alkaline nature of the ashes playing the role of the catalyst [1]. Interestingly, the first commercial application of a heterogeneously catalyzed process dates much later, namely, the famous “Döbereiner Feuerzeug” in the early nineteenth century, which used the ability of platinum to catalyze the oxidation of hydrogen, thus generating a gentle flame to light fires and smoking pipes. Today, the vast majority of the manufacturing processes in the chemical and the refining industries are heterogeneously catalyzed, which does not mean that homogeneous or bio-catalysts aren’t important. They will however not be considered here in the context of “surface and interface science.” Catalysis is also the key technology for cleaning of off-gases before venting them into the environment, be it in the tailpipe of individual cars or in the off-gas streams of power stations or chemical plants. According to various sources that can be found in the Internet, the business of producing and selling catalysts for the chemical and refining industry is estimated to grow from currently approximately US$ 15 billion to approximately US$ 20 billion by 2020. On top of that comes the catalytic converter business, which is estimated to grow from US$ 8 billion in 2015 to US$ 13 billion by 2025 [2]. For the car industry, the catalytic converter is mandatory for the “license to operate” because it is the only technology that allows a combustion engine to cope with today’s environmental regulations. For the chemical and refining industry catalysts are the lever to generate value: estimates are that US$ 1 in catalyst value on average generates US$ 300 in chemicals. Another way to illustrate the leverage of catalysts is the calculation of the specific catalyst consumption per ton of manufactured product. One of the oldest Surface and Interface Science: Applications of Surface Science II, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

538

72 Industrial Heterogeneous Catalysis

heterogeneously catalyzed industrial processes is the vanadium oxide-catalyzed sulfuric acid process, also known as “Kontaktverfahren.” It was started in 1888 at BASF, and the catalyst, although many improvements happened since, is at its core still the same today. From 1888 to 1994 about 25 million tons of sulfuric acid were produced at BASF, and during that time about 1.800 tons of catalyst were consumed, thus 75 g catalyst per ton of acid. Given the tremendous importance of catalysts for industrialized societies, it goes without saying that a sound technical and scientific basis in this field is of utmost importance.

72.2 Discovery and Understanding

The application of surface science tools in the field of heterogeneous catalysis started in the 1960s with the pioneering work of Jochen Block at the Fritz Haber Institute in Berlin and his investigations of catalytically active surfaces using the technique of field ion emission microscopy [3]. This work for the first time was able to show catalytic surfaces with atomic resolution. A large step forward followed in the 1980s with the work of Gerhard Ertl and coworkers in Germany and inter alia the group of Gabor Somorjai in the United States [4]. Gerhard Ertl picked up an observation that had been discovered in the 1970s in the laboratory of Ewald Wicke and coworkers [5, 6]: the rate of CO oxidation over Pt catalysts starts to oscillate under isothermal conditions (Figure 72.1), when the ignition temperature is reached (205 ∘ C). The frequency of the oscillations increases with increasing temperature, and at the same time the amplitude decreases. Finally, at around 246 ∘ C a quasi-stationary behavior is observed. Many proposals, among others, based upon mass transport effects were made to explain the phenomenon. It was Ertl’s group that could show by means of several analytical surface science tools that the structure and properties of the platinum surface changes in the presence of the reactants, which in turn changes its adsorption properties so that after awhile the original surface appears again and a new cycle is started [7, 8]. The experimentally proven insight that the surface of a working catalyst normally is not a static structure but in fact is only created in the presence and with the participation of the reacting species reflects a somewhat sloppy long-known saying among catalyst researchers: “You need to show the catalyst what it is supposed to do, and then it will do so,” also empirically known as the “catalyst formation phase.” This term describes that many catalysts need some time under reaction conditions in the technical reactor before they reach their full performance. Thus, the surface of a working catalyst is not a static structure but a highly dynamic and disordered state as clearly shown by Ertl’s work on model systems. Probably the most famous example for the application of surface science technologies is Ertl’s work on the ammonia synthesis catalyst, which was awarded the Nobel Prize in 2007 https://www.nobelprize.org/nobel_prizes/chemistry/laureates/2007/ ertl_lecture.pdf . Ertl’s Nobel lecture was entitled “Reactions at surfaces: from atoms to complexity” (Figure 72.2).

72.2 Discovery and Understanding

U 16

U 17

U 19

5s

8

13 t

t υ = 212 °C

υ = 220 °C

5 t

υ = 207 °C

U(%)

U 25

31 30 19 t (s) 5 υKat = 246 °C

235 °C

t

υ

228 °C

Figure 72.1 Oscillating conversion as a function of temperature for CO oxidation over Pt. (Source: Keil and Wicke 1980 [6]. Reproduced with permission of Wiley.)

Figure 72.2 Oscillating surface patterns of CO (dark) and Oxygen on Pt(110) as imaged by photoemission electron microscopy. (Source: From [8].)

It is about 100 years ago that the catalyzed synthesis of ammonia from the elements was done on industrial scale for the first time. Alwin Mittasch, the father of the ammonia catalyst, has described in 1951 [9] the guidelines along which the catalyst was developed.

539

540

72 Industrial Heterogeneous Catalysis

Physicochemical measurements set the process window in terms of temperature and pressure; economic considerations led to the concept of partial conversion per pass and recycling the unconverted feed. Still a catalyst was missing that could enhance the reaction rate sufficiently to make the whole process economic – the concept of “space–time yield” was introduced. The first material that fulfilled the technical target was osmium – however due to its scarcity this could never be an economic solution. At this point Mittasch started a systematic screening of possible catalysts. The basic guideline was that catalyst candidates should be able to form “semi-stable” nitrides, which were assumed to be an intermediate in the reaction. As a second principle there were attempts to improve the intrinsic properties of the selected material by combining them with additives. Only at a later point in time, it was realized that there are not only promoting but also poisoning additives, sulfur being the most prominent among them. Yet, still an empirical discovery during the screening phase turned out to become the final key for success: catalysts made from a certain Swedish magnetite outperformed all others, and subsequent experiments were aimed – finally successful – at the synthetic reproduction of this material. All experiments had to be run under technically relevant conditions, which in this case meant under high pressure. In order to speed up the screening, many parallel identical test units were built (Figure 72.3). Finally, also a method for the catalyst production had to be developed, which in the case of the ammonia catalyst ended up in a rather unusual technology, namely, an oxidative smelting of the required catalyst composition. The historic example of the ammonia process to a large extent still today is the way to go for the development of an industrial catalyst.

Figure 72.3 Test equipment used by Mittasch to screen ammonia catalyst formulations. (Source: BASF.)

72.2 Discovery and Understanding

• • • •

Identify the physicochemical and economic boundaries of the intended process. Try to identify the rate determining step in the supposed catalytic cycle. Look for elements that are known to form possibly suited intermediates. Tune the properties of the catalytic material by additives (promotors), and check for poisons. • Run systematic tests under industrially relevant conditions (“high throughput testing”). • Make sure that the cost of the catalyst is reasonable. • Develop a process for the catalyst production on industrial scale. Mittasch was sure THAT his catalyst reproducibly worked; he did not know in detail HOW this performance was achieved. It took about 100 years to answer the “HOW” by means of Ertl’s work: Figure 72.4 shows the potential energy diagram for the ammonia synthesis catalyzed by an iron surface [10]. The largest energy barrier along this pathway is in the formation of NHad from Nad and Had . Although the numbers have some uncertainties, they do at least qualitatively illustrate why the heterogeneous, catalyzed reaction is so much favored over the homogeneous reaction. It took another 20 years for the proof that at least for the ammonia catalyst, the mechanistic insights that were gathered under ultra-high vacuum (UHV) conditions are essentially identical with the situation under technical high pressure conditions. For that purpose in situ neutron diffraction was used to study the structure of a technical catalyst under reaction conditions [11]. N + 3H

314 NH + 2H

389

1129

1400

~960

NH2 + H

460 ΔH = 46 50

543 17

~21

1/2 N2 + 1/2 N2ad 3/2 H2 + 3/2 H2

~33 259 106 Nad + 3Had

NHad + 2Had

~41 NH2ad + H2ad

NHad

NH3

Figure 72.4 Potential energy diagram for ammonia synthesis catalyzed by an iron surface (energy values in kcal/mol). (Source: Ertl 1980 [10]. Reproduced with permission of Taylor & Francis.)

541

542

72 Industrial Heterogeneous Catalysis

72.3 Single Crystals and Particles

Surface science investigations of catalytic materials are done on model systems, often planar single crystal surfaces. Industrial catalysts on the other hand are composed of a variety of small crystallites, and at a first glance, the relevance of single crystal studies for the behavior of “real catalysts” may be questioned. However, crystal faces terminate also small crystallites. Typically, smaller crystallites are terminated with higher indexed surfaces. In 1969 researchers from DSM published a systematic investigation on the statistics of surface atoms and sites with a certain local environment as a function of crystal size [12]. The authors note that they intend to explain the empirical observation: “It has often been observed that catalyst samples having the same qualitative composition (for instance, nickel-on-silica) but differing in mode of preparation, show pronounced dissimilarities in sorptive and catalytic behavior, even if allowance is made for the differences in active surface area. … Different methods of preparation will, generally, yield catalysts differing in crystallite size, crystallite size distribution and shape of the metal crystals.” In their work the number of nearest neighbors for the three close-packed metal structures is analyzed as a function of crystal size and crystal shape. Depending upon the crystal morphology, the number of these sites changes as a function of crystal size as well as of the crystallographic plane, which terminates the crystal. Figure 72.5a,b shows the statistics of surface atoms as a function of crystallite site for (a) an fcc (face centered cubic) octahedron and (b) a bcc (body centered cubic) rhombic dodecahedron. It can be seen from Figure 72.5a,b that the fraction of surface atoms with a certain local environment (N(Cj )/NS ) changes as a function of crystallite size (drel ) and that this change is different for different crystal types. In both cases significant differences in catalytic activity can be expected if either only the corner atoms (C4 ) or only the atoms on the face (C9 and C6 , respectively) are responsible for the activity and, moreover, the sample consists predominantly N(Cj)

N(Cj)

NS

1.0

1.0

NS

4

C4 C9,10 4 C97

0.5

6

C4 C35

0.5

2

C39

C6

drel

0 0

5

10

15

20

drel 25

0 0

(a)

5

10

15

20

25

(b)

Figure 72.5 (a) fcc octahedron: statistics of surface atoms and (b) bcc rhombic dodecahedron: statistics of surface atoms. (Source: van Hardeveld and Hartog 1969 [12]. Reproduced with permission of Elsevier.)

72.3 Single Crystals and Particles

of either very small (5 nm) particles. The concept has also been known under the earlier terminology of “structure-sensitive” and “structureinsensitive” reactions, which was introduced by M. Boudart (see also Chapter 39 in Volume 5). This principle is the basis of the recently rather prominent “nanoengineering” of heterogeneous catalysts where people try to synthesize catalysts that are composed of crystallites with a specific shape and narrow crystallite size distribution. Recently the topic was nicely reviewed by G. Somorjai and coworkers [13, 14]. An example, where this strategy has led to a successful industrial development is the BRIM catalyst family of Haldor Topsoe [15]. The catalysts are used in the desulfurization of hydrocarbon streams, the active phases are Co–MoS2 or Ni–MoS2 structures. It could be shown by means of scanning tunneling microscopy, that it is the edges of MoS2 that are interacting with thiophene molecules and that led to catalyst design routes where the amount of these edge structures is maximized (Figure 72.6). “The fundamental insight and new preparation procedures have allowed Haldor Topsoe to introduce several new high performance Co–Mo and Ni–Mo BRIM catalysts. The improved fundamental understanding of the nature of the reactions and inhibitions has helped tailor-make the catalysts for different hydrotreating applications including the ULSD production” [15]. The example shows where surface science tools (and modeling) are currently placed in the development process for industrial catalysts: they provide a much deeper understanding and mechanistic insight into the interaction of the catalyst surface with the reactants. In some cases this lead to improvements of the catalysts on a rational basis; cases where a not yet known and theoretically predicted catalytic material made it into industrial practice are still to come.

™

Figure 72.6 STM image of MoS2 nanocluster showing the interaction of thiophene molecules with the metallic brim sites. (Source: Topsoe 2007 [15]. Reproduced with permission of Elsevier.)

543

544

72 Industrial Heterogeneous Catalysis

72.4 Catalyst Manufacturing

Most of the literature which covers the production of catalysts is disclosed in patents. This is due to the fact that in many cases very specific and unique conditions during the manufacturing process are required to obtain high quality materials in a reproducible manner. Except for the case of the ammonia catalyst where a smelting process is applied (see Section 72.2), essentially three technology families are established: (1) The active ingredients, in many cases precious metals, are impregnated on preshaped porous inorganic support structures. (2) The active phase is formed as an inorganic precipitate or under hydrothermal conditions, followed by a thermal transformation often into an oxide (the socalled calcination) and finally shaped into the catalyst body by technologies like extrusion or tableting. The support materials used in (1) are also made in this way. (3) The third technology comprises the coating of a slurry of dispersed particles as a layer of a few microns thickness on multichannel brick-like structures, the so-called monoliths (Figure 72.10), which are used in off-gas treatment applications. For (1) as well as for (2), it is important to get control over nucleation and growth [16]. It has been demonstrated above that the surface architecture of particles changes as a function of size and shape – and both are determined by the conditions during these critical phases. Figure 72.8 shows schematically the thermodynamics during the nucleation phase: when the nucleating “particle” exceeds a certain size, the gain in bulk energy exceeds the surface contribution. Since this process is linked to exceeding the solubility limit of the respective material, it becomes clear that a precise control of local concentrations is key for a stable and robust production process. Impregnated catalysts often are manufactured in a way that the active ingredient is only present in a certain part of the total pellet. In the majority of cases this is close to the surface, the thickness of this “shell” can be varied in a controlled manner.

100 μm (a)

250 μm (b)

Figure 72.7 Supported metal catalyst with varied shell thickness. (Source: BASF.)

72.4 Catalyst Manufacturing

+

Free enthalpy

Surface contribution = 4πr 2ΔGsurface Sum

Threshold for spontaneous growth:

rc

dSum

rc =

=0

dr

2ΔGsurface ΔGbulk

Bulk contribution = 4πr 3ΔGbulk 3
Figure 72.8 Thermodynamics during the nucleation phase of solids from solution. (Source: Schüth et al. 2001 [16]. Reproduced with permission of Elsevier.)

The underlying synthesis principle are electrostatic interactions between the support surface and the used metal compounds. Figure 72.9b shows how the surface charge of oxidic carrier materials change as a function of pH. Figure 72.9a shows the species that are present in a PdCl2 solution as a function of pH and concentration. For a strong interaction between support and metal species, both should be present with opposite charges. It can be concluded that at pH 4 such a situation does exist for alumina carriers but not for silica [17]. Thus, when using Pd chloride as a precursor, it can be expected in this simple example that a narrow Pd shell is formed on alumina, but not on silica.

0.8

0.4

PdCl2

0.0

0.8

PdCl+ Pd(OH)2

0.4

0.0

(b)

PdCl3(OH)–

2

4

0.0 –20.0

–60.0

0.05 m

PdCl(OH) PdCl2

0.2

20.0

–40.0

PdCl3–

0.6

0

(a)

PdCl(OH)

PdCl42–

: SiO2 : γ–Al2O3 : (α+ γ)–Al2O3

40.0

0.005 m

0.2

1.0 c(PdLx)/co(PdCl42–)

PdCl3–

0.6

–

Pd(OH)2

PdCl42–

Zeta potential (mV)

c(PdLx)/co(PdCl42–)

1.0

6

8

10

12

2

4

6

8

10

pH

14

pH

Figure 72.9 Impregnation chemistry: electrostatic interactions as a function of pH impregnations.

545

546

72 Industrial Heterogeneous Catalysis

72.5 Catalyst Bodies

Industrial catalysts come in many different geometric shapes: powder catalysts for liquid slurry phase or fluidized bed gas phase reactors, tableted or extruded pellets in various shapes for industrial fixed bed reactors or brick like structures, so-called monoliths, which dominate the off-gas treatment applications in cars or industrial plants. Figure 72.10 shows some illustrative examples. Usually the catalyst bodies are porous, with typical specific surface areas of several ten to several hundred square meters per gram. Moreover, often the catalytically active species are not evenly distributed across the pellet, but located in a certain fraction of the total volume. Pellet shape and distribution of the active component are dictated by the chemical engineering rules of heat and mass transfer: the active sites need to be accessible for the reacting species. The flow of reactants and products through the catalyst bed should be as easy as possible, and last but not least it must be made sure that the macroscopic catalyst structure remains mechanically stable for a long time: attrition and breaking of the bodies is absolutely undesirable. Mechanical strength and high porosity are clearly opposing targets, and sometimes it is the most difficult task during the development of an industrial catalyst to find a solution in this field. The time scales involved in heterogeneously catalyzed reactions range from femtoseconds for the catalyzed event at the active site over milliseconds for diffusion processes to seconds or minutes for the residence time in the catalytic reactor. Similarly, also a wide range of distances is covered: from the atomic scale of the active site to dimensions of 10–100 nm of the catalyst pore structure and finally several millimeters up to centimeters of the catalyst pellet.

(a)

(b)

Figure 72.10 Examples for industrial catalysts – pellets (a) and monoliths (b). (Source: BASF.)

72.6 Catalyst Performance: Activity, Selectivity, and Lifetime

Whereas the volume of catalytic converters in the tailpipe of combustion engines range from 100 ml (motor cycles) to 100 l (large trucks), the catalyst inventories in petrochemical reactors go up to hundreds of tons in refineries. Fluidized catalytic crackers are able to handle up to 500 m3 of feedstock per hour and have an inventory of up to 1000 metric tons of (fluidized) catalyst.

72.6 Catalyst Performance: Activity, Selectivity, and Lifetime

The performance of industrial catalysts is measured by the three parameters activity, selectivity, and lifetime. Unless the catalyzed reaction can only run to thermodynamic equilibrium without the possibility of side reactions, selectivity is by far the most important criterion. The ammonia synthesis is an example for the first class of reaction: the more active the catalyst, the lower the temperature, which is required to overcome the activation barrier, and the higher the conversion per pass for the exothermic reaction. More often, however, the reactants can go into different reaction channels, and it is then the task of the catalyst to selectively lower the activation barrier only for the desired reaction. As an example, Figure 72.11 shows the different molecules which can be obtained from the 𝛼,𝛽-unsaturated aldehyde citral, when different catalysts are used for the selective hydrogenation of one of the three double bonds in the molecule or its derivatives. There is a rich literature on this topic, for example, reviewed in [18]. Essentially, the underlying principle is to make use of differences in the adsorption strength of different functional groups on different metal surfaces. Catalysis, in this case, enables the cost-efficient synthetic access to odorous molecules for the flavor and fragrance industry, finding their use mainly in home and personal care products, fine fragrances, and the food industry. They are replacing (expensive) natural flavorings, which have been extracted for long times from a wide range of fruits, vegetables, nuts, bark, leaves, herbs, spices, and oils. They are referred to as being nature identical flavorings. This category of flavorings is identical to natural flavors but manufactured synthetically. Starting from citral as depicted in Figure 72.11, the class of so-called acyclic monoterpene alcohols is accessible. Depending upon which of the three double bonds is hydrogenated, the scent of the molecules differs: geraniol/nerol and citronellol present rose scents, while citral and citronellal have lemon scent. Geraniol/nerol can be further converted via an isomerization step (not shown in Figure 72.11) to linalool and tetrahydrolinalool with lavender and bergamotte scent, and finally menthol (accessible by cyclization of citronellal) as the most important cyclic terpene alcohol has the smell of mint. The synthesis of the latter one comprises an enantioselective, homogeneous hydrogenation step. Another principle that is used for selective transformations is the so-called shape selectivity. Here one takes advantage of the fact that the pore openings of zeolite structures restrict the size of molecules that can enter or leave the catalytically active cavity in the material. A prominent application is the acid-catalyzed isomerization

547

548

72 Industrial Heterogeneous Catalysis

H2

O

H2

OH

O

H2

OH

ENAL 3,7-Dimethyl-2-octenal

ENOL 3,7-Dimethyl-2-octenol

H2

O

H2

OH

H2

O

OH O

Citral (cis/trans)

UOL Geraniol and Nerol H2

H2

H2 H2 O

H2 OH

CAL Citronellal

H2

DHC Dihydrocitronellal

COL Citronellol

OH

DMO 3,7-Dimethyloctanol

H2 OH

IP Isopulegol

OH

MTL Menthol

Figure 72.11 Reaction network of citral hydrogenation (color codes mark the hydrogenated groups).

of toluene into benzene and p-xylene by means of zeolite catalysts. Only p-xylene molecules fit through the openings of the zeolite pores whereas o- and m-xylene molecules do not. Finally, in order to achieve a profitable process, industrial catalysts need to reach service times of at least one year. Service time is to be distinguished from the cycle time between regenerations. Both phenomena are summarized under the term catalyst deactivation. Depending on the process used, the catalyst cycle time may vary from a few seconds, as in fluid catalytic cracking (FCC), to several years, as in ammonia synthesis. Figure 72.12 shows representative examples that cover the mentioned time span [19]. Obviously, a short period between two regenerations does not exclude the development of a technical process but requires adapted reactor technology, as demonstrated most prominently by the fluidized catalytic cracking process, which is the most important unit in a modern refinery. Regeneration and active phase both occur within seconds, and the technical solution is the circulation of huge amounts of catalysts between the reactor and the regenerator. This points to the fact that industrial catalysts always are an integral part of the system, which is made up by reactor and

72.7 Summary

10−1 100 101 102 103 104 105 106 107 108

FCC

Hydrocracking HDS Catalytic reforming

EO MA Formaldehyde Aldehydes Hydrogenations Acetylene Oxychlorination

C3 dehydrogenation

Fat hardening NH3 oxidation Time (s)

TWC

SCR

10−1 100 101 102 103 104 105 106 107 108 1hour 1 day 1year Figure 72.12 Time scale of deactivation of various catalytic processes. (Source: Moulijn et al. 2001 [19]. Reproduced with permission of Elsevier.)

catalyst. Process development typically happens in an iterative manner, until the best compromise between the requirements and/or possibilities of both elements is achieved. 72.7 Summary

The current overview could only briefly touch into the many aspects of the multidisciplinary applied science of industrial heterogeneous catalysis. Although having the reputation of being rather a (black) art than a science, it is the intention of this article to demonstrate that there are many links between fundamental science and applied industrial catalysis. However, it needs to be stated as well that so far the discovery of new catalytic materials still to a large extend follows the approach of Alwin Mittasch at the beginning of the last century as outlined above. Surface science has been extremely valuable to understand (in hindsight) the complex behavior of catalyst surfaces. Based on that knowledge, already existing catalysts can be further improved. Surface science data intrinsically suffer from the fact that they are not obtained under industrially relevant pressure conditions. Therefore, in all cases it needs to be critically checked whether this leads to different surface structures (and consequently wrong conclusions). Equally important is an increased understanding of the processes that happen during the synthesis of catalytic materials. Only with more progress in this field, it will be possible to move away from a situation in which the properties of a product are defined on a “product by process” level, meaning that the same product is obtained in a reproducible manner only if the manufacturing parameters are closely watched. Last but not least, reference is made to much more comprehensive textbooks and encyclopedia dealing with this fascinating field [20].

549

550

72 Industrial Heterogeneous Catalysis

References 1. Plinius d.Ä. Historia naturalis, 23–70 2.

3.

4.

5.

6.

7.

8.

9. 10.

11.

n.Chr. Johnson Matthey and the Evolving Powertrain (2016). Presentation to Analysts/Investors (4 February 2016). For example, Cocke, D.L. and Block, J.H. (1978). Field ion and field desorption mass spectrometry of inorganic compounds. Surf. Sci. 70: 363, and other contributions in the same volume. Somorjai, G.A. (1992). The experimental evidence of the role of surface restructuring during catalytic reactions. Catal. Lett. 12: 17. Beusch, H., Fieguth, P., and Wicke, E. (1972). Thermisch und kinetisch verursachte Instabilitäten im Reaktionsverhalten einzelner Katalysatorkörner. Chem. Ing. Tech. 44: 445. Keil, W. and Wicke, E. (1980). Über die kinetischen Instabilitäten bei der CO Oxidation an Platin Katalysatoren. Ber. Bunsen Ges. Phys. Chem. 84: 377. Imbihl, R., Cox, M.P., and Ertl, G. (1986). Kinetic oscillations in the catalytic CO-oxidation on Pt(100): experiments. J. Chem. Phys. 84: 3519. Jakubith, S., Rotermund, H.H., Engel, W. et al. (1990). Spatiotemporal concentration pattern in a surface reaction: propagating and rotating spirals, and turbulence. Phys. Rev. Lett. 65: 3013. Mittasch, A. (1951). Geschichte der Ammoniaksynthese. Verlag Chemie. Ertl, G. (1980). Surface science and catalysis – studies on the mechanism of ammonia synthesis: the PH Emmett award address. Catal. Rev. 21: 201. Kandemir, T., Schuster, M.E., Senyshyn, A. et al. (2013). The Haber-Bosch process revisited: on the real structure and stability of “ammonia iron” under working conditions. Angew. Chem. Int. Ed. 52: 12723.

12. van Hardeveld, R. and Hartog, F. (1969).

13.

14.

15.

16.

17.

18.

19.

20.

The statistics of surface atoms and surface sites on metal crystals. Surf. Sci. 15: 189. An, K. and Somorjai, G.A. (2015). Nanocatalysis I: synthesis of metal and bimetallic nanoparticles and porous oxides and their catalytic reaction studies. Catal. Lett. 145: 233. Alayoglu, S. and Somorjai, G.A. (2015). Nanocatalysis II: in situ surface probes of nano.catalysts and correlative structure-reactivity studies. Catal. Lett. 145: 249. Topsoe, H. (2007). The role of Co–Mo–S type structures in hydrotreating catalysts. Appl. Catal., A 322: 3. Schüth, F., Bussian, P., Agren, P. et al. (2001). Techniques for analyzing the early stages of crystallization reactions. Solid State Sci. 3: 801. Mang, Th. (1996). Präparation, Charakterisierungund Aktivität von Palladiumkatalysatoren mit Konzentrationsprofil. Dissertation. München. Stolle, A., Gallert, T., Schmoeger, C., and Ondruschka, B. (2013). Hydrogenation of citral: a wide-spread model reaction for selective reduction of 𝛼,𝛽-unsaturated aldehydes. RSC Adv. 3: 2112. Moulijn, J.A., van Diepen, E.A., and Kapteijn, F. (2001). Catalyst deactivation: is it predictable? What to do? Appl. Catal., A 212: 3. For example (a) Thomas, J.M. and Thomas, W.J. (1997). Principles and Practice of Heterogeneous Catalysis. Wiley-VCH. (b) Bartholomew, C.H. and Farrauto, R.J. (2005). Fundamentals of Industrial Catalytic Processes. WileyVCH. (c) Ertl, G., Knözinger, H., Schüth, F., and Weitkamp, J. (eds.) (2004–2014). Handbook of Heterogeneous Catalysis. Wiley-VCH.

551

73 Automotive Catalysis Feng Gao and János Szanyi

Increased fuel efficiency and decreased emission of environmentally harmful pollutants have been the two driving forces in the development of new, highly efficient, selective, and durable catalysts for mobile internal combustion engine (ICE) applications. The rapid increase in the number of automobiles created a shortage in cheap fossil energy resources in the second half of the twentieth century and demanded the development of highly fuel-efficient combustion engines. Meanwhile, growing societal consciousness about the negative consequences of automotive pollution on our environment demanded the development of new catalytic technologies for mobile applications [1]. Due to the fuel economy advantages of lean-burn engines, diesel-powered engines have been continuously gaining popularity, which, in turn, demanded renewed research aimed at developing novel catalytic approaches for lean NOx reduction. The task seemed daunting, since any feasible catalyst is needed to reduce NOx to N2 in the presence of very large excess of O2 . Traditional three-way catalysts (TWCs) worked rather efficiently at stoichiometric air-to-fuel ratios and removed both CO and unburned hydrocarbons (HCs) by oxidation and concomitantly reduced NOx to N2 [2]. The optimum operational air-to-fuel ratio window of these types of catalysts, however, is very narrow (close to stoichiometry). Besides their limited window of operation, they also require the use of expensive platinum group metals (PGMs): Pd, Pt, and Rh. The initial work of Iwamoto et al. on the direct catalytic decomposition of NO on Cu–ZSM-5 (MFI) seemed to provide a new solution for lean NOx reduction [3]. Subsequently a huge surge in lean NOx reduction research then focused on the catalytic reduction of NO either by NH3 or HCs on metal ion-exchanged zeolite catalysts [4]. In particular, Cu and Fe ion-exchanged MFI showed very promising catalytic performance. The major drawbacks of these catalysts that eventually prevented their introduction in practical emission control systems were their insufficient hydrothermal stability and nonresistance to poisoning. Although these catalysts failed to provide the practical solution for automotive lean NOx emission control it originally intended, the tremendous research effort provided large amount of invaluable information about reaction mechanisms, structure–activity relationships, and catalyst degradation mechanisms. As the practical viability of zeolite-based catalysts seemed to fade in the 1990s, new approaches toward lean NOx emission were developed. One of these Surface and Interface Science: Applications of Surface Science II, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

552

73 Automotive Catalysis

new technologies, nonthermal plasma-assisted NOx reduction [5], took advantage of the use of cold plasma that served two key roles in the process of reducing NOx under O2 -rich conditions: it converted all the NO to NO2 and oxidized some of the unburned HCs to aldehydes [6]. This allowed the utilization of a relatively simple, cheap catalyst for the actual NO2 reduction process: Ba–Y, FAU. Although this catalyst showed great promise under laboratory testing conditions, it failed to survive the harsh practical testing requirements. The two main deactivation mechanisms of these catalysts are poisoning by unburned HCs (reversible deactivation) and the collapse of the zeolites structure under high temperature hydrothermal conditions (irreversible deactivation). In spite of the practical infeasibility of this lean NOx reduction technology, the knowledge gained from fundamental mechanistic studies on these systems proved invaluable in later research. In the mid-1990s Toyota developed a new approach toward lean NOx reduction that combined NOx storage and NOx reduction in one catalyst [7]. This currently commercialized technology, called NOx storage/reduction (NSR) or lean NOx trap (LNT), required the very precise sensing of the NOx level at the catalyst outlet, which, in turn, was used to control the composition of the exhaust gas at the catalyst inlet. The catalyst consisted of three key components: a high surface area, thermally durable support (e.g. γ-Al2 O3 ), a NOx storage component (a basic oxide, e.g. BaO), and a PGM (mostly Pt) that provided both NO oxidation (in the lean cycle) and NO reduction (in the rich cycle). This technology requires either a lean/rich cyclic engine operation or a controlled reductant introduction into the exhaust gas stream under constant lean engine operation [8] (and references therein). The operational complexity of this technology and the insufficient sulfur poisoning resistance of the catalyst limited the widespread introduction of this system into practical application. The limited exhaust handling capacity restricts the applicability of this technology to light- and medium-duty applications. The most important breakthrough in commercial lean NOx emission control catalysis came from the discovery of the superior NH3 selective catalytic reduction (SCR) performance of small pore metal (Cu and Fe) ion-exchanged zeotypes with specific crystal structures. These materials exhibited very high catalytic activity and N2 formation selectivity, as well as hydrothermal stability under lean NH3 SCR conditions. The synthesis of SSZ-13, a zeolite with chabazite (CHA) structure by Chevron, opened the path to the large-scale production of Cu ionexchanged catalysts that are highly active in a wide temperature window and exhibit very high N2 selectivity and superior hydrothermal stability during NH3 SCR. The isostructural silicoaluminophosphate (SAPO-34) has very similar attributes to the corresponding zeolite catalysts. These materials are able to comply with the current NOx emission levels mandated by regulations, although further research is needed to extend their operational temperature window as new highly fuel-efficient diesel engines with exhaust emission gas temperatures below 150 ∘ C are being developed. In this chapter we are going to focus on these two key technologies of automotive exhaust control catalysis developed in the past two decades and gained practical introduction. We will highlight primarily our own research and discuss it in the context of work from other research groups. We will focus on fundamental surface science studies on model NOx storage systems that allowed us to use sample

73.1 NOx Storage Reduction Catalysis

preparation and characterization techniques that requires ultrahigh vacuum (UHV) conditions. On the NH3 SCR system, our focus will be on the basic mechanistic understanding of the working of these catalysts. At the end of the chapter, we will also briefly summarize some of the remaining and new challenges of this field.

73.1 NOx Storage Reduction Catalysis 73.1.1 Introduction

NOx storage reduction catalysts work under periodic operating conditions. Under oxygen-rich (lean) conditions, all of the NO that is converted to NO2 by the precious metal component of the catalyst reacts with the storage material (a basic oxide) to form nitrites and nitrates. As the storage material gradually saturates with NOx , a brief rich period is required for its regeneration. Under reducing conditions NOx is released from the basic oxide and simultaneously reduced on the precious metal component of the catalyst. Thus the precious metal (primarily Pt) plays a critical role in both the storage and the reduction cycle: during lean operation (NOx storage), it acts as an oxidation catalyst, while in the rich cycle it carries out the reduction of NOx to N2 . The operational principle of an NSR catalyst is shown in Scheme 73.1. In the following sessions we discuss some of the key processes taking place on/in these complex catalysts during NSR and the effect of impurities in the gas stream on these processes. We focus primarily on studies conducted on model systems that allowed the use of modern surface analytic techniques to investigate the fundamentals of these processes.

HC, CO, H2

NO, NO2, O2

N2, CO2, H2O

NO2(g)

NO2(g) Nitrate

Nitrate Pt

NO2

Ba

γ-Al2O3 NOx storage cycle

Pt

NO2

Ba

γ-Al 2O3 NOx reduction cycle

Scheme 73.1 Scheme of NOx storage and reduction under lean and rich engine operating conditions, respectively.

553

554

73 Automotive Catalysis

73.1.2 Cooperative NO2 Adsorption: Nitrite/Nitrate Formation and Stability

Since this catalyst operates on the principle of NOx storage and release, first the interaction/reaction between NOx (NO and NO2 ) and the BaO storage material needs to be addressed. Both NO and NO2 are odd-electron molecules, and they do not behave as classic Lewis acids [9]. Since NOx can undergo either one-electron oxidation or reduction to generate their common Lewis acidic (NO+ , NO2 + ) or Lewis basic (NO2 − , NO3 − ) forms, respectively, if charge transfer occurs between two neighboring adsorbates, they can exhibit a strong attractive interaction. This charge transfer-based lateral interaction (i.e. cooperativity) results in a greater affinity between the individual adsorbates and the surface adsorption sites than that between the corresponding neutral adsorbates and the surface. This cooperative adsorbate stabilization was first proposed by Schneider for the adsorption of NOx on MgO(001) surface, and then it was extended to other adsorbates (COx and SOx ) on alkali metal oxides [10]. Density functional theory (DFT) calculations have suggested that enhanced surface binding of the cooperative pairs more than compensates the energetic cost arising from the charge transfer. As the alkali metal oxides are neither good electron donors nor electron acceptors, the formation of charged NOx species by the cooperative interaction explains their abilities to bind NOx species mostly as nitrites and nitrates. The binding configuration is represented for NO2 on MgO(001) in the scheme below. e–

M2+

NO2

e–

O2–

M2+

O M2+

N O2–

e–

NO2+

O M2+

M2+

O2–

N– O M2+

O O2–

M2+

The energy gain due to the cooperative adsorption decreases from MgO to BaO, but even in BaO it is significant. On MgO it is about 100%, while on BaO it is around 25%. This pairwise adsorption of NO2 was shown to be substrate mediated by the DFT calculations of Broqvist et al.: nitrite and nitrate interact through the oxide surface [11]. The adsorption of the first NO2 (nitrite formation) creates an electronic defect that is healed by the adsorption of a second NO2 (formation of a nitrate), producing an ion pair. Due to this electron–hole separation, the mechanism of nitrite/nitrate pair formation is non-localized; therefore, pairs may form at long distances. This, in turn, will minimize the repulsive interaction between the two charged NOx ions facilitating high nitrite/nitrate coverages without loss of stability. Our work on BaO/Al2 O3 systems (both on model BaO/Al2 O3 /NiAl(110) systems and high surface area BaO/γ-Al2 O3 storage materials) provided the first experimental confirmation of the above discussed cooperative adsorption mechanism of NO2 . Processes taking place during both the cooperative adsorption (nitrite/nitrate ion pair formation) and thermal decomposition were followed by temperature-programmed desorption (TPD), X-ray photoelectron spectroscopy (XPS), and Fourier transform infrared spectroscopy (FTIR). The adsorption of

73.1 NOx Storage Reduction Catalysis

NO2 on a thick BaO layer at cryogenic temperatures resulted in the formation of nitrite/nitrate ion pairs, suggesting a low activation barrier for the ion pair formation [12]. TPD experiments revealed that nitrites decomposed first, leaving an oxygen atom behind, thus resulting in the formation of BaO2 . This BaO2 phase is very stable and was seen to be present even after all the nitrates decomposed. The thermal decomposition of nitrates takes place in two distinct steps: first, at a lower temperature, they decompose by releasing an NO2 molecule, and then at high temperatures, they decompose by releasing NO + O2 . The series of FTIR spectra collected during the exposure of a ∼20-monolayer (ML) BaO film supported on an Al2 O3 layer grown on NiAl(110) (20 ML BaO/Al2 O3 /NiAl(110)) at 90 K is displayed in Figure 73.1a. During the first two NO2 doses, only bridge-bound nitrites (1225 cm−1 ) and ionic nitrates (1342 and 1429 cm−1 ) are formed. From the third NO2 dose on, besides the increases of the intensities of both nitrite and nitrate species, the presence of adsorbed N2 O4 (1259, 1750, and 1772 cm−1 ) can also be seen. Spectral signatures of adsorbed N2 O4 appear even before the saturation of the intensities of infrared (IR) bands originated from the nitrite/nitrate ion pairs. This is due to the very high sticking coefficient (practically unity) of NO2 at the low sample temperature of 90 K. At very low NO2 exposures (first two spectra), all the incoming NO2 molecules find empty surface sites and form nitrite/nitrate ion pairs. However, with increasing NO2 exposure (even at rather low nitrite/nitrate surface coverages), not all of the incoming NO2 molecules can react with the BaO surface; rather they condense on top of the nitrite/nitrate “patches” and form N2 O4 ice. The formation of nitrite/nitrate ion pairs at 90 K is also confirmed by the N 1s XPS spectra (Figure 73.1b) collected during the stepwise annealing of an NO2 -exposed BaO film (the one used in the IR measurements). Two peaks are seen 1772 1765 0.1%

1272 1259

1342 1429

XPS intensity (arb. unit)

IRAS intensity (arb. unit)

1737

100 K 200 K 300 K 400 K 500 K 575 K 620 K 650 K 800 K 900 K

403.8

407.6

1225

(a)

1800 1700 1600 1500 1400 1300 1200 1100 Wavenumber (cm–1)

Figure 73.1 (a) IRAS spectra: NO2 adsorption on a thick BaO film at 90 K. (b) XPS data of the N 1s region after adsorption of a NO2 at 90 K on a thick BaO film and

414

(b)

412 410 408 406 404 402 400 398 Binding energy (eV)

subsequent annealing. (Source: Adapted with permission of Yi et al. 2007 [12]. Copyright 2007, American Chemical Society.)

555

556

73 Automotive Catalysis

in the XPS spectrum after a high dosage of NO2 is applied at low temperature, representing nitrite + N2 O4 ice (403.8 eV) and nitrate species (407.6 eV). Heating the sample to 125 K results in the complete evaporation of the N2 O4 ice. At that point the intensities of the two N 1s XPS peaks are identical, substantiating the 1/1 stoichiometric ratio of the nitrite/nitrate ion pairs. The pairwise NO2 adsorption mechanism was also confirmed on high surface area BaO/Al2 O3 NOx adsorber materials. The nature of nitrates, however, was somewhat different on the model (BaO/Al2 O3 /NiAl(110)) and practical (BaO/γ-Al2 O3 ) systems: on the model system only bulk nitrates were observed, while on the practical ones both surface and bulk nitrates were detected. This difference originates from the vastly different BaO coverages over these two systems: on the model systems the BaO coverage was kept very high (>20 ML) in order to avoid contribution to the chemistry of the BaAl2 O4 phase that readily forms in the reaction between BaO and the very thin alumina substrate. In contrast, in the practical systems, the nominal BaO coverage never exceeded 2.5 ML, and BaO was present either as small (BaO)x clusters or BaO particles. The nature of the different types of nitrate species was examined in a combined TPD/FTIR/15 N-NMR (nuclear magnetic resonance) study on BaO/γ-Al2 O3 systems with BaO coverages of 8 and 20 wt% [13]. TPD spectra following NO2 saturation of these samples at 300 K showed two distinct desorption features (Figure 73.2a): an NO2 peak at 670–690 K and an NO + O2 feature at 770–790 K. The intensity ratios of these two NOx desorption peaks varied systematically with the BaO loading: at low BaO loadings, the low-temperature peak dominated, while at high BaO coverages the high-temperature NO desorption feature was the more intense one. Consistent with the TPD results, the 15 N NMR spectra of 15 NO2 -saturated 8 and 20 wt% BaO/γ-Al2 O3 also showed the presence of two distinct nitrate species at 337 and 340.5 ppm chemical shifts (Figure 73.2b). In the NMR spectrum of each sample, components of both of these peaks are present, but with vastly different intensity 340.5

NO(20%) NO2(20%) NO(8%) NO2(8%) NO(Al2O3) NO2(Al2O3)

337 313 20 wt% BaO/Al2O3 8 wt% BaO/Al2O3

Al2O3

300 (a)

700 500 Temperature (K)

900

370 (b)

350

330 310 290 Chemical shift (ppm)

270

Figure 73.2 (a) TPD spectra from Al2 O3 and after NO2 saturation 673 K. (Source: Adapted 8 and 20 wt% BaO/Al2 O3 after NO2 saturawith permission of Szanyi et al. 2005 [13]. tion at 300 K. (b) 15 N solid-state NMR spectra Copyright 2007, American Chemical Society.) from Al2 O3 and 8 and 20 wt% BaO/Al2 O3

73.1 NOx Storage Reduction Catalysis

ratio: for the 8 wt% sample, the 337 ppm peak was more intense, while for the 20 wt% sample the 340.5 ppm peak dominated the spectrum. Consistent with the results of both the TPD and NMR measurements, FTIR spectra from these NO2 -saturated samples also substantiated the presence of two nitrate species. Although the IR absorption features in the spectrum of each sample exhibit IR features at the same peak positions, the intensity ratios of the 1574 and 1425 cm−1 peaks varied systematically with the BaO loading: in the spectrum of the 8 wt% sample the 1574 cm−1 peak, while in the spectrum of the 20 wt% sample 1425 cm−1 peak dominated. Based on the results of this multi-spectroscopy study, we suggested the existence of two different types of nitrates: one that is bound to monolayer BaO interacting strongly with the alumina support and ionic type of nitrates formed on the BaO particles. A subsequent combined spectroscopy/microscopy study on BaO/γ-Al2 O3 revealed that at low BaO loadings BaO was very highly dispersed on the alumina surface due to its specific interaction with the pentacoordinated Al3+ surface sites located on the (100) facets of the support [14]. The properties of these very small (BaO)x units are fundamentally different from those of the bulk BaO particles. Their strong interaction with the alumina surface alters their chemical properties in their reaction with NO2 , consequently altering their spectroscopic characteristics. The results of DFT calculations were instrumental in assigning the different vibrational features observed experimentally for the surface and bulk nitrates on BaO/γ-Al2 O3 . The IR signature vibrations of nitrates on NO2 -exposed BaO/Al2 O3 /NiAl(110) were assigned to different binding configurations and were supported by the results of DFT calculations. On the alumina support only bridging nitrates were present, while on BaO both bridging and monodentate nitrates were present [15]. Furthermore, temperature-dependent NO2 adsorption studies on these model NOx storage materials revealed that NO2 exposure at 300 K sample temperature primarily produced monodentate and bridging surface nitrates, in contrast to NO2 exposure at 500 K where the formation of ionic (bulk) nitrates took place [16]. Figure 73.3 shows two series of RAIR (reflection absorption infrared) spectra collected on a model NOx storage material (BaO/Al2 O3 /NiAl(110)) at 300 and 500 K sample temperatures. At low NO2 exposure (1 × 10−5 Torr, nitrites were converted to nitrates. On thick BaO films, cooperative NO2 adsorption (i.e. formation of both nitrites and nitrates) took place even at low NO2 exposures. At high NO2 partial pressures (>1 × 10−4 Torr), the thin BaO layers ( crystalline phase(s) > active center/elemental arrangement. We have already discussed the catalyst and here we focus on the reactor. The reactor can be operated in continuous, semicontinuous, or batch mode and under adiabatic or isothermal conditions. A commercial-scale reactor can contain up to several hundred tons of catalyst. Unit design and operating conditions are chosen based on kinetic and thermodynamic considerations and, obviously, economic factors such as capital and operational costs and productivity metrics. Examples of common reactor types include fixed beds (generally gas-phase plug-flow operation utilizing shaped solid catalyst particles in the millimeter size range), trickle beds (similar to fixed beds but designed for three-phase reactions), captive and moving fluidized beds (for example, gas-phase utilizing 50–150 μm diameter fluidizable catalyst particles), continuous stirred tank reactors (fully back-mixed systems using slurried catalyst powders), and various configurations of gauze-type, monolith, or tube reactors having catalyst-coated walls [10, 19]. Often, a production process contains two or more reactors in a sequential configuration. Major industrial processes employing such reactors include petroleum refining (e.g. hydroprocessing, fluid catalytic cracking, alkylation, and reforming), chemical processing (e.g. polyolefin synthesis, oxidation/ammoxidation, and hydrogen, ammonia, and methanol synthesis), emissions applications (e.g. automotive and stationary), and emerging fields such as those utilizing renewable feedstocks and other “green” chemistries. 74.1.2 History and Impact of High-Throughput Heterogeneous Catalysis

The need for catalyst companies to improve and develop new industrial catalysts and processes is constant, driven by the requirement to compete in the marketplace, grow their businesses, respond to customers’ needs, adapt to new regulatory requirements, and obtain and maintain intellectual property protection. Furthermore, in today’s competitive economic environment, companies are forced to perform the requisite R&D faster (to reduce time to market) and more efficiently (to minimize investment on a per-project basis). The pharmaceutical industry was faced with these challenges in the 1980s and responded by implementing

74.1 Introduction

combinatorial chemistry and high-throughput screening research methodologies that greatly accelerated the drug discovery process, which, at the time, was plagued by high costs and long development times. They developed a variety of high-dimensional experimental methodologies including split-pool synthetic approaches using mixtures of thousands of compounds prepared on polymer beads, parallel syntheses to produce “libraries” of related organic molecules, and new property-screening techniques for evaluating large collections of potential leads efficiently and accurately. It was this work that inspired the later development of hierarchical approaches for the high-throughput preparation, characterization, and testing of inorganic and organometallic materials and catalysts [20–23] that are today successfully used in many catalyst R&D laboratories [24–37]. Compared with pharmaceuticals, catalyst R&D presents a very different set of challenges and requirements. Heterogeneous catalysts are chemically and physically complex formulations, in many cases with ill-defined structures and mechanisms of action. The parameter space that must be addressed to optimize such catalyst systems is exceedingly large. The ability to predict the necessary catalyst composition, structure, formulation, shape, and operational process conditions, as well as the ideal catalyst synthesis variables such as metal precursors, wet and dry synthesis conditions, and methods of posttreatment/calcination for a given chemical transformation, is low and for complex multicomponent catalysts, almost nonexistent. The result is that traditional methods of synthesis, characterization, and testing have been manual, slow, and inefficient, and catalyst improvement and discovery protocols have necessarily been primarily trial-and-error processes, albeit with the inclusion of significant know-how and theory. Catalyst researchers have confronted these problems by developing various integrated solutions generally comprising automated, software-supported, high-throughput technologies for synthesizing, characterizing, and testing large numbers of materials for a variety of catalytic applications. Rapid advances in the technology and application of high-throughput R&D to heterogeneous catalysis took place in the mid-1990s, led by start-up companies Symyx Technologies, hte GmbH, and Avantium, as well as several academic laboratories in the United States and Europe. They initiated many of the developments that have taken place since then and that are being used today for industrial catalyst development. The number of experiments per unit time that can now be performed using state-of-the-art high-throughput workflows can be up to 2 orders of magnitude or more higher than using conventional methods. For example, a high-throughput research effort can yield 15 000–50 000 experiments per year compared with 500–1000 experiments using traditional methods [36, 38]. In addition, the field has advanced to the point where the quality of the resulting data and the commercial relevance of the experiments in many cases equal that of conventional experimental procedures, which allows its application to areas such as process development, scale-up, and even commercial production support. 1) Although the term “combinatorial” is also used in the catalyst R&D, our preference is to use “high throughput” to refer to synthesis, characterization, and testing workflows because it is more descriptive of the tools and techniques used in the field.

615

616

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

The use of high-throughput synthesis, characterization, and screening (testing) methodologies to accelerate R&D in heterogeneous catalysis has now become well established in many industrial and academic laboratories around the world. These methodologies are providing significant advantages to their practitioners, including a reduction in time to market for new and optimized catalysts, increased probabilities of success due to the ability to perform far greater numbers of experiments than in the past, better intellectual property protection because of the thoroughness with which a given technical area can be explored, shorter and more projects executed per unit time, and increased organizational efficiency due to improved data storage, access, analysis, and sharing. In addition, and perhaps unexpectedly, the very large numbers of, in many cases, empirical experiments are providing the basis for a more thorough understanding of catalyst structure–performance relationships for many different catalytic systems. Many catalysts have now been commercialized that were developed using highthroughput R&D. However, due to industry secrecy there are few details available even in the patent literature, which rarely discloses if high throughput was used to invent a catalyst or process. The continued investments in this area made by companies, including Clariant in its Palo Alto laboratories, certainly strongly suggest its value to their businesses. It should be noted, though, that because of high equipment costs and the specialized expertise required to establish and run high-throughput programs that can produce the high-quality results necessary for successful catalyst R&D, it has not become as deeply established in industry nor academia as has been the case in the pharmaceutical and related fields [39]. The situation is improving with better and more efficient technologies, know-how, and economies of scale, though only gradually. This chapter will describe representative high-throughput synthesis, characterization, and testing workflows developed and utilized today in heterogeneous catalysis with a focus on gas-phase reactants converted over solid catalysts (one of the most commercially prevalent processes at large scale) and liquid-phase batch processes (also important in areas such as specialty, fine, and pharmaceutical chemistries). We describe the infrastructure, operational procedures, preparation, characterization, testing, and analytical methods that enable high-throughput catalytic materials discovery and optimization along with case studies that demonstrate its utility. A basic but thorough overview is presented focusing on the latest advances and important but not-yet-widely-practiced applications, including the use of workflows correlated with industrial processes and high-throughput-assisted catalyst scale-up and production support.

74.2 Comparison of High-Throughput and Conventional R&D

As discussed in the introduction, the complexities and large variable spaces associated with the preparation of heterogeneous catalysts have resulted in catalyst R&D and scale-up being traditionally, and by necessity, a slow, manual process

74.2 Comparison of High-Throughput and Conventional R&D

often employing trial-and-error approaches. Theory-based methodologies have been limited due to the general lack of fundamental knowledge about active sites and reaction mechanisms, which are difficult (many times impossible) and expensive to probe even using today’s sophisticated analytical tools. As a result, scientists and engineers have been constrained by a catalyst development process characterized by lengthy commercialization times, incremental improvements, and slow invention and innovation. Typically, experiments are performed in a sequential manner and often last many days or weeks. This is because scales are large and unit operations slow, and there is often limited access to the necessary equipment and manpower due to competing priorities and equipment scarcity. It is also expensive, and thus limiting, to perform many experiments, especially as the scale goes up. It is remarkable that given these limitations researchers have nonetheless developed important industrial processes with enormous impact on society by relying on their chemical know-how, intuition, and hard, often tedious, work. It must be noted, however, that many of these processes came about many years or decades ago during a time of rapid catalyst technology development, when the growth vs. time s-curve was steeply increasing. Today, the field can be considered mature and therefore requires new approaches to accelerate technology development. Many of the limitations inherent in traditional heterogeneous catalyst R&D can be overcome by the application of high-throughput technologies. The process generally consists of the following sequential steps: (1)

Software-assisted design of high-density, multi-sample, arrays of potential catalytic materials (referred to as “libraries”). (2) Parallel or fast-serial synthesis of catalyst libraries at gram or sub-gram scales using industrially relevant and up-scalable preparation techniques. (3) Characterization of the resulting catalytic materials using a variety analytical tools operated in a rapid mode. (4) Catalyst performance testing at gram or sub-gram scales using parallel or fastserial reactor systems designed to produce high-quality data or, at minimum, an accurate and reproducible catalyst ranking. (5) Data work-up and analysis optionally applying computational algorithms where appropriate. (6) Feedback of the results to the subsequent library design. Together these steps are commonly referred to as the high-throughput “workflow,” the specifics of which are tailored to the requirements of the chemistry of focus and the stage/phase of the high-throughput project. Specialized hardware and software is integral to each workflow, and the processes are often automated. It is critical that all equipment and procedures are thoroughly validated against catalysts of known performance before a project is initiated. Data from each of the above steps must be captured, connected, and databased for analysis. Note that the high-throughput process can also be used to evaluate catalyst pilot plant or even production plant samples to provide rapid feedback that can be used to modify or optimize those processes (see Section 74.4).

617

618

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

Equipment miniaturization, robotics, custom equipment designs, and integrated data management systems are all hallmarks of high-throughput R&D, and this requires the application of specialized chemistry, engineering, and software expertise. Until recently, these have not been available at chemical companies or catalyst manufacturers. Accordingly, high-throughput R&D is optimally performed by interdisciplinary teams made up of chemists and chemical, mechanical, and software engineers. The support of skilled technicians, machinists, and pilot and commercial plant operators is also critical. Compared with conventional methods, where up to hundreds of materials can be made and tested in a traditional laboratory in a year, a high-throughput laboratory’s productivity can be 50–100 or more times higher. This can translate to an acceleration of catalyst discovery phases by three to four years. This is achieved primarily by working at much smaller scales (0.1–1.0 g for high throughput vs. 10 g to 10 kg for a conventional lab); by performing syntheses, characterization, and performance testing in a parallel or rapid-serial manner; and by automating certain steps in the workflow. For example, a parallel fixed-bed reactor may test 48 catalysts per unit time, whereas a bench unit would only test one catalyst in two to five times that same unit time. The methods and technologies to achieve this are described in detail in Sections 74.5 and 74.6. It should be noted that although the throughputs discussed here are lower than those that the reader may be familiar with from the combinatorial approaches utilized in pharmaceutical or biological research, the proper comparison that should be made is between the number of high-throughput experiments performed and the next best practiced alternative in the field. The large throughput difference is mainly because the workflows necessary for high-throughput heterogeneous catalysis R&D are highly complex in comparison with those used in pharmaceuticals or the life sciences. There are other benefits for using high-throughput techniques beyond the obvious ones related to the fast commercialization of new higher-performing catalysts. Because of the small scales, safety hazards due to reagent amounts in use or in storage are lower, as are reagent and waste disposal costs. There is clearly a sustainability advantage. As discussed in Section 74.1.2, the main drawback of high-throughput heterogeneous catalyst R&D is the cost, effort, and expertise required for implementation and operation.

74.3 High-Throughput Strategies and Methods

Although high-throughput R&D for heterogeneous catalysis relies heavily on advanced hardware and software technologies, it is the manner in which these tools are used, the strategies with which they are applied, that often determines their effectiveness. Certain general approaches and principles that have been developed and proven themselves are discussed below.

74.3 High-Throughput Strategies and Methods

74.3.1 High-Throughput R&D Phases: The Hierarchical Workflow

The most important strategy used in high-throughput R&D programs to maximize throughput and cover the desired experimental variable spaces efficiently and effectively is the hierarchical workflow, depicted in Figure 74.1. It is broadly divided into phases: primary, secondary, tertiary, pilot, and production, each of which includes integrated catalyst preparation, characterization, and performance testing unit operations.2) Together, the high-throughput phases (in bold) address the traditional stages of research, development, scale-up, and commercial catalyst production support. Note that in the figure, the secondary, tertiary, pilot, and production phases are separated by dotted rather than solid lines, which differs from the inverted triangle figure we have previously published (the last version is in reference [23]). This is intended to reflect the current state of high-throughput technologies, particularly at Clariant, where the secondary phase workflows have advanced to the point that they can be applied with success to catalyst scale-up, process optimization, and even production support (see Sections 74.4 and 74.7). The primary screening phase is most often focused on new catalyst “hit” discovery and in some cases optimization of existing commercial compositions. It typically

Data precision, commercial conditions

Number of samples

Variables

Primary phase Hit/lead discovery

Secondary phase Lead validation and optimization; often the starting point for HT development program Secondary/tertiary phase Catalyst and process optimization, scale-up, plant support Pilot plant

Composition/phase Precursors Supports Pretreatments Formulation Composition, synthesis, support optimization Commercial raw materials Formulation method Test conditions Real reactor feeds Final formulation Particle scale-up Commercializable synthesis Plant raw materials, feeds Catalyst line-out, stability, tolerance to upsets, lifetime HT tests of actual plant catalysts at each unit operation

Commercial production

Figure 74.1 Phases of high-throughput catalyst discovery, optimization, and scale-up showing specific activities and variables addressed. Note the overlap of the secondary and later phases made possible by recent technological advances. 2) A note on terminology: Phases in the hierarchical high-throughput workflow are also interchangeably referred to as “stages,” “scales,” or “screens” depending on the preferences of various authors. The term “screening” does not necessarily indicate binary (“yes/no”) or purely qualitative-type testing, and continuing advances in high-throughput technologies have brought high-quality quantitative information to primary and higher screens.

619

620

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

consists of ultrahigh-throughput qualitative, semiquantitative, or even quantitative experiments performed on very small samples of materials that can, in advance, reasonably be expected to catalyze the desired chemical transformation. Catalyst library diversity is high in this phase, and the focus is on varying elemental composition and arrangement (e.g. crystalline phases), although formulation and process conditions are also often explored. The goal is to eliminate from consideration those materials and families of materials that do not have the potential to move to the next synthesis/testing phase. Several hundred materials can be made and tested per day at the 0.1–1.0 mg scale, and up to 50 000 experiments per year can be routinely performed per project. Because of the scale and throughput, unconventional syntheses and reactor designs are often necessary (see Section 74.5.1). In most cases performance testing is not for the precise property or data that will be eventually required, but for a surrogate or a partial data set that is a necessary but not sufficient subset of the information of eventual interest [40]. For example, catalyst yields to a desired product may be probed but not the full selectivity profile. Because of this, it is usually necessary to rely on catalyst rankings in comparison to a standard and to each other to define hits. The objective in the primary phase is to balance throughput with precision such that the probability of success is reasonably high while the generation of false positives and especially false negatives is minimized. Typical methods employed in the primary stage for catalyst evaluation include mass spectrometry (MS) (rapid-serial) and infrared (IR) thermography and colorimetric methods (both parallel) – all yielding primarily, although not exclusively, activity information. Notwithstanding the above, it is critically important here, and in all stages of high-throughput work, that catalyst preparation and testing methods are not so far removed from the real commercial processes that they lose their relevance. It must be ensured that the results correlate to the real process for both catalyst manufacture and catalytic reaction via a “validation” process by which known catalysts can be “rediscovered” using the high-throughput workflows. Hits discovered in the primary phase are next taken to the secondary phase for confirmation and optimization. Variable space that has been identified in the primary phase to not be useful is discarded. In contrast to the primary phase, here the variables under consideration are more numerous and either more closely resemble or actually are those that are or would be used in commercial production processes (see “Variables” column in Figure 74.1). Catalyst formulation, form and size, method(s) of preparation, and the testing process, reactor, and analytics are designed to be similar to those of the commercial-scale materials and reaction processes. The use of commercial raw materials for catalyst preparation as well as real production feeds for the catalysis is critical at this stage so that a good correlation can be established with the commercial scale. Because of these close-to-commercial conditions, the secondary phase is often the starting point for the high-throughput development and/or optimization of catalysts for existing processes; the primary phase is skipped in these cases. The catalyst performance data generated in the secondary phase should be of at least equivalent quality as that of a standard laboratory bench reactor or even a pilot

74.3 High-Throughput Strategies and Methods

unit, since the goal is to observe small improvements in performance as a function of the catalyst variables. Accurate, precise, and comprehensive data sets (e.g. activity and selectivity with mass balance) are required. Thus test reactors closely resemble scaled-down versions of larger units, for example, small gram-scale fixed beds with gas chromatography (GC) analytics. Similarly, synthetic procedures should produce materials nearly indistinguishable from commercial production (with the exception of bulk scale), and therefore preparative methods such as incipient wetness impregnation, coprecipitation, hydrothermal synthesis, and various pre- and posttreatment methods are employed. Most of these challenges have today been met as high-throughput technology has developed and matured over the last several years, and its application to a variety of diverse cases has proven successful. We detail two examples from recent high-throughput developments at Clariant in Section 74.7. “Lead” catalysts from the secondary phase are moved to the tertiary phase where the focus is on scale-up to commercial production. Tertiary synthesis and testing is performed using conventional bench, pilot, semiworks, and/or production facilities. Here the focus is on late-stage development issues such as final formulation, minimization of costs, use of plant feeds and raw materials, particle scale-up, catalyst life, effects of bulk scale, generation of commercial performance data (important for the adoption of new catalysts by customers/operators), and production of quantities of catalyst for customer testing and qualification. With the recent availability of highquality secondary synthesis and testing, many of the activities that in the past had to be performed in the tertiary phase can now be accomplished in the secondary phase, as illustrated in Figure 74.1. Increasingly, at Clariant, pilot plant studies and catalyst production trials can proceed directly from the secondary phase without tertiary phase work. In these cases the role of the conventional pilot plant can change from a tool used for the iterative optimization of catalyst formulations into one that is focused on the later-stage work described above. In summary, while the hierarchical high-throughput workflow and its phases of catalyst development remain an important aspect of the overall process, in our experience it is the secondary synthesis and testing equipment and procedures that have emerged as the most flexible and useful to support practical catalyst discovery, development, scale-up, and production. In essence, it can be used to fulfill the role of the other phases in the hierarchy and has thus become the main tool for high-throughput catalyst R&D at Clariant. However, because secondary phase equipment and procedures can be complex and expensive and require special expertise and know-how to develop, operate, and apply, only organizations that are able to make or access these investments can fully benefit from the technology. 74.3.2 High-Throughput Workflow Integration and Execution

A diagram depicting a generalized high-throughput workflow is presented in Figure 74.2. The cycle of library design; synthesis, characterization, and testing in each phase; data collection and storage; and data analysis and visualization is illustrated. Underlying the workflow is an enabling software infrastructure,

621

622

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

Data analysis and visualization Catalyst library design Inventory Metal precursors Promotors Supports Formulators

Discovery

Development or optimization

Primary HT phase

Secondary/tertiary HT phase

Semiquantitative data

Quantitative data

Pilot plant Scale-up Production support Production

Database

Figure 74.2 High-throughput workflow diagram. Each phase includes a synthesis, characterization, and testing component. At Clariant, we have closely integrated high throughput, pilot, and production plants as is shown.

discussed in detail in Section 74.6, which includes a capability for statistical and/or predictive design of experiments. Figure 74.2 depicts a cycle that begins with the design of a catalyst library. Once the corresponding high-throughput experiments are completed and the data is extracted from the database and analyzed, the cycle begins again using the results as feedback for the subsequent catalyst library design. The speed of the process allows for the constant, almost “real-time” adjustment of the catalyst libraries based on the results of previous tests. This is a very powerful approach for catalyst discovery and development and is most effective when combined with scientific and technical personnel that have expertise not only in high-throughput methodologies but also in heterogeneous catalysis. Additionally, it is critical that there be close, continuous interaction between the high-throughput team and the pilot plant and production teams, as well as the sales and legal teams and the eventual catalyst customer(s). This ensures that the final catalyst that is developed is able to be manufactured on commercial production lines and that it meets market and customer requirements. For these reasons, a high-throughput group located within a catalyst producer is an ideal arrangement.

74.4 High-Throughput Applied to Catalyst Scale-Up and Production

In recent years, high-throughput methods and technologies have successfully begun to be applied to catalyst and process scale-up in pilot units and even in commercial

74.4 High-Throughput Applied to Catalyst Scale-Up and Production

catalyst production plants. For example, at Clariant we have made significant progress in designing high-throughput equipment and procedures that allow the rapid evaluation of production processes and the resulting catalysts’ performance to allow fast tuning of the large-scale process to maximize value. This had led to several commercial successes, two of which are presented in Section 74.7. One approach to couple high-throughput to production is to scale down plantscale unit operations. In many cases detailed information about the operation of a piece of plant equipment has been collected over many years. This includes the operation window for the equipment as well as the temporal history of the material as it passes through a processing step. These conditions can often be accurately modeled in small-scale lab equipment. The difficulty is that plant data is usually proprietary and is specific to one particular piece of equipment. A close collaboration between production and R&D is thus vital to achieve a close tie between high-throughput experiments and the production process. Once a catalyst recipe has been developed, it is scaled up in a pilot plant on a multikilogram scale and then scaled up again in a plant trial to a multi-ton scale. Often steps that were batch processes in the lab are continuous processes in the plant. Each process step must be verified, at a minimum, and, ideally, optimized. However, the time available to optimize a process, especially in the plant, is very limited. Pilot and plant trials must be carefully planned with a list of variations for each process step and a list of samples to be collected after each process step. An analogous series of intermediate materials is made in the lab with an optimized procedure. The physical properties of the intermediate materials such as surface area, XRD, and composition are measured and used as targets for the large-scale preparation. These target properties are then used to optimize the plant process. Unfortunately, easily measureable physical properties of the materials often do not directly correlate with the final catalyst performance. Normally the catalytic performance test is the most sensitive test for catalyst quality. An effective strategy is to finish processing the intermediate samples collected in the plant using highly repeatable high-throughput tools. A catalytic performance measurement can then be used to determine what process step is problematic and what the optimum condition is for each step. At Clariant, we have advanced our secondary phase workflows far downstream in the catalyst development process. Whenever possible, commercial reagents are used very early in the process. Raw materials used for production must be available in multi-ton quantities and have a reasonable cost. The impurities in the raw materials can vary by supplier, production process, and number of purification steps. High throughput is ideally suited to quickly determine the effect of the raw material impurities on catalyst performance and to develop mitigating actions. Extrusion and tableting are two common methods to produce shaped catalysts on a commercial scale. These forming operations have been scaled down to the 1–10 g scale in our high-throughput laboratory as have the associated operations such as milling, dry blending, and kneading (see Section 74.5.2.1). This way, complete commercial formulations can be prepared and tested that include binders, pore formers, lubricants, and extrusion aids. The goal in each research program is to match the

623

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

density, crush strength, and porosity of the commercial catalyst. This lowers the chance of scale-up problems. Another processing step in the high-throughput workflow that has been tied closely to the commercial process is the heat treatment. In laboratory research it is very popular to calcine samples “overnight.” Commercially, such long heat treatments are uneconomical. Most commercial belt calciners and rotary furnaces have a maximum catalyst residence time of one hour, and the catalyst is only at the target temperature for a fraction of that time. Modeling the temperature profile of a commercial calciner with small-scale calcination can eliminate lengthy re-optimizations of the heat treatment during pilot studies. Figure 74.3 shows an example of a small-scale calciner being used to model a plant calciner. A laboratory tube furnace is often used to calcine multiple catalyst samples in parallel in a high-throughput workflow. An illustration of the successful application of high throughput to the scale-up and production of an actual catalyst product developed at Clariant is shown in Figure 74.4. The graph indicates catalyst percent yield measured in a highthroughput secondary gas-phase reactor, validated against a commercial unit, over the term of the project, which includes a high-throughput development period, a high-throughput-assisted scale-up and production optimization period, and a high-throughput second-generation catalyst development period. Catalysts’ performance prepared using high-throughput methods, in a pilot unit, and in the production plant are shown. Over the life of the project, a target-exceeding catalyst was developed first using high-throughput methods and then transferred to the pilot unit (in this case with little loss of performance) and to the production plant where initially performance dropped significantly. By applying our high-throughput methodologies, however, the plant-made catalyst performance was increased to exceed the reference performance and meet the program target. This catalyst became a successful new product for the company. Finally, a second-generation catalyst with even higher performance was developed using high throughput and Oven temperature vs. time

Temperature

Catalyst temperature vs. time

Temperature

624

Time (a)

Time (b)

Figure 74.3 Modeling of a (b) plant calciner with an (a) laboratory tube furnace used for heat treating high-throughput catalyst samples. (a) Catalyst temperature vs. time and (b) oven temperature vs. time are shown.

74.5 High-Throughput Equipment and Methods

Second-generation catalyst production in progress

Project time

Target Reference

Yield (%)

Performance in: High-throughput Pilot plant Production

HT development

HT-assisted scale-up and production optimization

Figure 74.4 Example of the application of high-throughput technologies to support and accelerate catalyst scale-up and production. In this case nine process steps were optimized, which necessitated the

HT development second generation

testing of 150 pilot plant samples and 102 production samples. Without highthroughput this volume of testing would have been slower by 50 times and realistically not possible.

transferred to the pilot and production plants using a similar process. In all, more than 250 pilot and production samples were tested, representing a 50× reduction in time to the commercial catalyst compared with a traditional approach. In reality, this amount and type of testing would not be possible without high-throughput technologies.

74.5 High-Throughput Equipment and Methods

The hardware and methods used for high-throughput catalyst preparation, characterization, and testing can vary substantially depending on the individual unit operations required for the particular type of material to be synthesized and reaction to be evaluated. For example, a supported noble metal catalyst requires different high-throughput synthesis procedures than does a zeolite catalyst. Similarly, a liquid-phase batch catalytic hydrogenation reaction requires a completely different reactor than does a gas-phase fixed-bed process. Even within a certain process, e.g. gas-phase fixed bed, there are a multitude of process parameters that affect reactor design. There are, however, basic components and steps that are common to the majority of high-throughput workflows. These unit operations include liquid and solid handling hardware/robots for accurate dispensing and combining of catalyst precursor chemicals used during synthesis and formulation steps, posttreatment

625

626

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

(e.g. calcination) equipment, shaping tools (e.g. tableting, extrusion) for preparing particle bodies, material characterization instrumentation, and batch or continuous test reactors with integrated analytics for evaluating the catalytic process. Each of these operations may be performed in a parallel or rapid-serial fashion depending on the required throughput and the complexity and compatibility of the processes. Many specific examples of various high-throughput technologies and their applications can be found in the academic and patent literature and in reviews. There have been many important contributions over the years ranging from the early work in materials science by Hanak [41], describing the preparation and characterization of thin films having continuous compositional gradients via simultaneous co-sputtering, and Schultz and coworkers [20], describing sequential sputtering through masks to generate thin films with individual areas of defined compositions, to the various latest advances in the field of catalysis [24, 30]. Interestingly, the focus of technology development during the evolution of high-throughput heterogeneous catalysis has been mainly on the test reactors rather than on the commercial relevance of the catalyst syntheses. This may be because many of the technologies have been developed by organizations that do not manufacture catalysts, i.e. academic institutions or companies specializing in high-throughput equipment development and with a chemical engineering focus. There are exceptions, of course, and catalyst manufacturers such as Clariant, BASF, Haldor Topsoe, Johnson Matthey, UOP, and others have in recent years significantly advanced high-throughput methods for preparing and formulating catalysts using a variety of industrially relevant methods. High-throughput reactor designs range from parallel units that can evaluate hundreds of catalysts simultaneously to scaled-down versions of traditional laboratory reactors that test several to hundreds of catalysts at a time. A wide range of analytical techniques have been reported including GC, IR thermography, scanning mass spectrometry, resonance-enhanced multiphoton ionization, fluorescence or colored-dye assays, thin layer chromatography, photoacoustic analysis, and gas sensors [36]. In many cases, multiple methods of analysis are required to rapidly screen libraries, which produce complex mixtures of products. In this section, the focus is on diverse examples of equipment and methods applied to heterogeneous catalysis that illustrate the general concepts of highthroughput synthesis, characterization, and testing and that are based on the authors’ experiences in industrial high-throughput R&D and prior published work. 74.5.1 Primary Phase Workflows 74.5.1.1 Catalyst Preparation and Characterization

In the primary high-throughput phase, where the number of catalyst variables and process conditions is very high, it is usually most efficient and productive to synthesize and test catalyst candidates in an array format on the surface of a twodimensional substrate, often referred to as a “wafer” or “wafer-formatted library.” Each resulting spot in the array has a discrete and well-defined composition or

74.5 High-Throughput Equipment and Methods

formulation that can later be correlated to characterization and testing information. The arrays can vary in format depending on the choice of the characterization technique and reactor to be used for screening. For example, Symyx has published several papers describing the use of 256-element (16 × 16) arrays on 4-in. square individual quartz or glass substrates using approximately 1 mg of each catalyst for each element (Figure 74.5) [22, 36, 38, 42–45]. Catalyst precursor solutions in most cases are premixed separately in microtiter plates (according to the desired library design) before being transferred to the catalyst wafers in either a serial or parallel fashion using automated liquid handling robots. Dispensing is by volume, and high precision is required because of the small amounts of solutions used, about 10–100 μl. The depositions are performed by either dispensing and then evaporating solutions of the precursors from the wafer to yield solid “thick films,” the direct addition of a slurry mixture onto the wafer with subsequent drying, or the impregnation of a precursor solution onto supports already present on the wafer followed by drying. In the case of incipient wetness impregnation, the catalyst supports (e.g. silica, alumina, other oxides, or carbon) are pre-dispensed onto the wafers as slurries, followed by drying, before the addition of the impregnation solution. After drying, the wafers are calcined, reduced, or otherwise posttreated to form the final catalyst oxides, metals, etc. Additional cycles of impregnation and/or physical posttreatment may be performed. Depending on the specific heat pretreatment, it can be done on multiple wafers simultaneously in ovens or in tube furnaces under controlled atmospheres. At this point the individual library elements may be characterized, although usually at reduced throughput and quality, using common techniques that can be applied to deposited materials including XRD, SEM, XRF, and EDS (energy-dispersive X-ray spectroscopy). This Precursor solution preparation in microtiter plate

Deposition/impregnation/precipitation

Serial

Parallel

Drying, calcination/reduction (multiple wafers in parallel)

Characterization and testing Figure 74.5 Primary phase synthesis workflow showing photos of serial and parallel preparations. Catalysts may be prepared by deposition, impregnation, or precipitation.

627

628

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

provides information about structural phases, morphologies, and compositions. The next step is testing as described in Section 74.5.1.2. The above methods have been designed to produce catalysts that, as much as possible in a primary phase, resemble those used on larger laboratory and commercial scales. Variations can be used for precipitation, sol–gel, and solvent evaporation syntheses. There have even been primary synthesis workflows developed for hydrothermal synthesis (e.g. to prepare zeolites), although not on substrates but using arrays of small sealed reactors. Other approaches such as those based on vapor deposition via evaporation or sputtering (PVD (Physical vapor deposition), CVD (chemical vapor deposition)) generally yield materials that differ too much from those that are or can be produced commercially and therefore have limited utility in heterogeneous catalysis R&D. Methods that produce materials whose composition varies continuously across a substrate, i.e. where there are no well-defined compositions at discrete points on the substrate, are limited because it is difficult to correlate actual composition with performance and the preparation is difficult using techniques that are not relevant to any type of commercial heterogeneous catalyst production. However, these types of primary syntheses have found use in the high-throughput evaluation of applications such as dielectrics and phosphors. Pooled synthesis techniques have also been used successfully in the primary highthroughput phase. This approach differs from the spatially addressable methods described above. Mallouk and coworkers published early proof-of-concept experiments on the synthesis of material libraries made up of noble metals supported on gamma alumina beads using a split-pool approach [46]. In this process, small carrier beads are placed in vials and then impregnated with different catalyst precursor solutions. After drying and thermal treatment/calcination, the beads in each vial, consisting of a single unique catalyst composition, are mixed together and then split into a separate set of vials. This process is repeated numerous times to generate a catalyst library. To calculate the theoretical number of different compositions on the beads, the following equation is used: N(nm) = (n + m − 2)!/(n − m)!(m − 1)!, where n is the number of components present and m is the number of split-pool operations. Identification of the beads in this example was accomplished using fluorescent dyes and micro-X-ray fluorescence after the synthesis. hte GmbH has also published extensively on the split-pool approach and has published an overview of this technique [47]. 74.5.1.2 Catalyst Testing

There are numerous approaches for testing catalysts in the primary high-throughput phase using a variety of nonconventional reactor and analytical equipment designs. The focus here, as in the primary synthesis section, is on screening two-dimensional catalyst arrays, because of the exceptionally high throughputs that can be achieved. Rapid-serial and parallel approaches have been developed ranging from those that simply rank catalysts’ activities to those that also provide selectivity information. Illustrative examples are presented below. An important limitation to these wafer-based screens is that the contact times between the reactant gas and solid catalyst surface are short; they are “flow-over” rather than “flow-through” systems. This means that conversions are necessarily low in most cases, and therefore

74.5 High-Throughput Equipment and Methods

selectivities cannot be measured at high conversions. Moreover, time-on-streamdependent properties of the catalyst, such as lifetimes, cannot be probed. There are cases, however, where the geometry of these two-dimensional primary screens is well suited to real-world conditions, including coated catalysts used in automotive catalytic converters and other similar applications. One of the most straightforward, yet effective, techniques for screening catalyst libraries for gas/solid catalysis is IR thermography [48], first described by Willson and coworkers [49] and Maier and coworkers [50] in the late 1990s. Here, a catalyst library is heated to reaction conditions in a chamber that contains the reactant gas. At the top of the chamber is an IR-transparent window with an IR camera located behind it and pointed at the catalyst array. During the experiment the IR camera is used to simultaneously measure the temperature of each catalyst under reaction conditions, which is compared to the background, a standard, and/or other experimental catalysts. Exotherms and/or endotherms are calculated from the temperature differences, thereby yielding activity information rapidly and in parallel (Figure 74.6). This technique has been shown to be well suited for screening reactions such as oxidations [51, 52] and emission control [38]. Recently, high-throughput Fourier transform infrared (FTIR) imaging has been reported [53]. Symyx has published detailed descriptions and applications of several platforms that make use of semiquantitative primary catalyst screens operating in either rapidserial or fully parallel modes. For example, a scanning mass spectrometer-based reactor system (Figure 74.7) has been used to evaluate catalyst libraries prepared on wafer substrates in a rapid-serial manner [22, 42–45, 54, 55]. The system operates by locating a movable feed/sampling probe over each candidate catalyst in the array, scanning from sample to sample sequentially, and analyzing reaction products at each location using a quadrupole mass spectrometer. The sample wafer is mounted on an x–y moving stage that locates each sample under the probe. The probe is moved perpendicular to the wafer (on the z-axis), allowing the tip to be located above each sample. During the test, the location being probed is heated to the 4.5

3.5

IR camera

3 2.5 2 1.5

Reactant gas in chamber

1

Temperature change (K)

4

0.5 0

Thermography image

Heated catalyst wafer Figure 74.6 Diagram of a high-throughput IR thermography reactor and catalyst testing image for measuring catalyst activity. Temperatures of individual catalysts are compared to background, each other, and a reference.

629

630

74 High-Throughput Heterogeneous Catalyst Research, Development, Scale-Up, and Production Support

Products to mass spectrometer Reaction gas in Gas to waste

Probe head z-axis nozzle control

Catalyst element

Catalyst wafer

CO2 laser heating

x–y scanning

IR thermometer

Figure 74.7 Diagram showing the fluidic routing for a high-throughput scanning mass spectrometer.

desired temperature(s) using a CO2 laser located beneath the quartz wafer, with the temperature being monitored and controlled by a proportional-integral-derivative (PID)-based feedback loop. The design is such that neighboring catalysts are not affected by the heating. Simultaneously, reactant gas is flowed onto the catalyst through the probe via a set of concentric tubes as shown in Figure 74.7. At the center of the head, a small fraction of the product gas is transferred through a capillary to the online detector. Due to the serial nature of the experiment, initial conversions and selectivities can be measured with each sample being exposed to the same time on stream. Again, because of the flow-over nature of this method, conversions are in most cases very low, meaning that initial reaction kinetics are monitored, not integral productivities at high conversions. Representative testing durations with this system are about one sample per minute. In Refs. [56–58], Senkan et al. describes related MS- and REMPI (Resonance-enhanced multiphoton ionization)-based primary screening reactors where the catalysts are supported on beads in stacked two-dimensional arrays. These innovative early reactors were able to test 80 catalysts using a parallel feed and serial analytics. More recently, Klein et al. has combined primary IR thermography with MS or GC/MS for collection of more detailed information on the hits generated from IR thermography [59]. An example of a unique and very powerful parallel primary screening reactor system utilizing parallel product detection was developed by Symyx researchers in the early 2000s and is shown in Figure 74.8 [36, 38, 42–44, 60]. It is designed to test two-dimensional catalyst arrays where all the samples are exposed to the same feed and process conditions simultaneously. The reactor is based on a microfluidic gas

74.5 High-Throughput Equipment and Methods

(a)

(b) Reaction station

(c) Imaging station

Spray station

CCD camera

Cooling plate Sorbent wafer Wafer distributor

(d)

Light sources

Spray nozzle

Insulation

Wafer distributor Catalyst wafer Heating plate

(g)

TLC plate

(e)

Figure 74.8 Photograph of the (a) microfluidic parallel screening reactor, (b) gas distribution wafer, and (c) example catalyst wafer. Diagram of the full system showing the (d) reaction station with the gas feed

TLC plate

(f)

Backlight source

in and out indicated with arrows, (e) spray station, and (f ) imaging station. An example image of a developed product sorbent wafer is shown in (g).

flow distribution and delivery device, a 16 × 16 256-element catalyst sample array, and a colorimetric detection method. The system provides identical gas flows to each of the catalysts via the fluidic device, which is made using microfabrication technologies (Figure 74.8b). “Microreactors” are formed over each catalyst sample by contacting the catalyst wafer and gas distribution device together (Figure 74.8d). The single gaseous feed stream is divided into 256 separate and independent streams by the binary splitting pattern (bifurcation) of channels in the distribution wafer. Each individual stream is contacted with the 2 mm diameter × approximately 0.2 mm deep well on the heated sample wafer that contains approximately 1 mg of catalyst. The precision of the microfabricated channel structures and the bifurcation strategy for the division ensure the accuracy of the flow division. The channel-to-channel flow difference is 0.9 V causes a gradual increase of the still passivating Fe content within the film with increasing potential. The strong increase of Cr(III) in the passive range is a consequence of the extremely small dissolution rate of Cr3+ ions as also found for pure Cr. Fe dissolves with current densities of 7 μA/cm2 in 0.5 M H2 SO4 , which is more than 1 order of magnitude larger than that of Cr3+ . As a consequence Cr is enriched within the passive layer. In alkaline solutions, the composition of the passive film is different. Due to insolubility, Fe(III) oxide preferential dissolution does not take place and, consequently, Cr(III) is accumulated more moderately with a maximum of c. 50% compared to 80% in 0.5 M H2 SO4 . The total thickness in the passive range in 0.5 M H2 SO4 is 1.5–2 nm [84, 86], similar to that obtained on pure chromium in the same potential range [71]. It has been observed that, for short polarization times (lesser than or equal to two hours), the crystallinity of the passive films decreases with increasing Cr content of the alloy [82, 83]. Structural changes also occur during ageing under anodic polarization. The major modification is an increase of the crystallinity of the film and the coalescence of Cr(III) oxide nanocrystals in the inner oxide as observed on Fe–22Cr [84] and Fe–18Cr–13Ni alloys [84, 85] studied over time periods of up to 65 hours. This is illustrated by the images shown in Figure 78.10. The comparison of the rates of crystallization of Fe–22Cr(110) and Fe–18Cr–13Ni(100) revealed that the rate of crystallization is more rapid on the austenitic stainless steel than on the ferritic one. This was tentatively explained by a regulating effect of Ni on the supply of Cr on the alloy surface, a lower rate of Cr enrichment being in favor of a higher degree of crystallinity [127].

815

816

78 Application of Surface Science to Corrosion

(a)

(b)

1 nm

1 nm (d)

(c)

1 nm Figure 78.10 (a, c) Structure of Fe–22Cr(110) and (b, d) Fe–18Cr–13Ni(100) surfaces observed by STM after passivation in 0.5 M H2 SO4 (aq) at +0.5 V/SHE for 2 hours (a, b) and for 22 hours (c, d). The nearly hexagonal lattice is marked. The effect of ageing under polarization is evidenced

2 nm by the extension of the observed crystalline areas. (Source: (a, c) Maurice et al. 1996 [84]. Reproduced with permission of The Electrochemical Society. (b, d) Maurice et al. 1998 [86]. Reproduced with permission of The Electrochemical Society.)

78.6 Breakdown of the Passive Film and Initiation of Localized Corrosion

Oxide passive films are sensitive to local breakdown eventually leading, in the presence of aggressive species (e.g. chlorides), to accelerated dissolution of the metallic substrate at localized sites (e.g. pitting), whereas the rest of the surface remains well protected [128, 129]. Pitting is a phenomenon that affects passivable metallic materials. It is particularly insidious because a component, otherwise well protected by the passive film, can be perforated locally in a short time with no appreciable forewarning sign. Passivity breakdown is the first step in the initiation of pitting corrosion [128, 129]. Dissolution of the oxide passive film, enhanced in the presence of chlorides, is one of the mechanisms proposed for passivity breakdown [94, 128, 129], and a surface science approach has demonstrated that the steps at the surface of the passive film play a key role in the dissolution mechanism in the passive state as revealed

78.6 Breakdown of the Passive Film and Initiation of Localized Corrosion [011]

[101] [110]

15 nm (a)

15 nm (b)

(111) Terrace [011]

[111] [101]

(010) Step 15 nm (c)

(d)

Figure 78.11 (a–c) Sequence of EC-STM images showing the localized dissolution of the passivated Ni(111) surface at +0.85 V/SHE in 0.05 M H2 SO4 + 0.095 M NaOH (pH 2.9). The oxide crystallographic directions are indicated. The circles show the areas of localized dissolution. (d) Model of a (111)-oriented

facet delimited by edges oriented the closepacked directions of the NiO lattice. Oxygen (surface hydroxyls) and Ni are represented by large and small spheres, respectively. (Source: Maurice et al. 2007 [108]. Reproduced with permission of Woodhead Publishing Ltd.)

for passivated Ni(111). Figure 78.11 shows an EC-STM imaging sequence of this process. It is observed without and with chloride anions in the electrolyte [108, 130]. The passive film dissolves at the edges of the facets produced by the tilted epitaxy between the NiO(111) and Ni(111) lattices. The process is similar to that of active dissolution of metal surfaces at moderate potential (see Section 78.2). The resulting 2D step flow process is dependent on the step orientation: the step edges oriented along the closed-packed directions of the oxide lattice dissolve much less rapidly, due to the higher coordination of their atoms. This process leads to the stabilization of the facets with edges oriented along the close-packed directions of the oxide lattice and produces steps that are oriented along the {100} planes, the most stable orientations of the NiO structure. A model for such a surface is presented in Figure 78.11b. Corrosion modeling using DFT methods has been applied to study the Cl− interaction with hydroxylated NiO(111) [106, 107, 131] and their effect on the dissolution of the oxide. Based on experimental results, a (533)-oriented NiO periodic model

817

818

78 Application of Surface Science to Corrosion

including monoatomic (010) step edges and (111) terraces was built to model the effect of steps at the surface of the passive film (Figure 78.8b). Cl adsorption was modeled by substituting the surface OH groups by Cl atoms at 25%, 50%, 75%, and 100% coverages and sub-surface insertion was modeled by exchanging one adsorbed Cl of the topmost anionic layer with one O atom of the first inner anionic layer of the oxide [106]. After DFT optimization, substructures of Ni(OH)2 , Ni(OH)Cl, or Ni(Cl)2 composition were observed to form and detach from the step edges of the adsorbed structures, as shown in Figure 78.12, confirming the major role of the step edges in the dissolution of the oxide. The calculated energies of detachment of the substructures revealed that the Cl-containing substructures are easier to detach, showing that dissolution at the step edges can be promoted by the adsorption of Cl and confirming a Cl adsorption-induced thinning as a possible breakdown mechanism of the oxide passive film [128, 129]. Unlike the adsorbed structures, the sub-surface inserted structures do not lead to the formation of substructures of Ni(OH)Cl or Ni(OH)2 type detaching from the step edges, suggesting that, after sub-surface insertion of the Cl atoms, film dissolution is not promoted. At surface saturation, the sub-surface inserted structures become more stable than the adsorbed structures, indicating After optimization

Before optimization H Cl O Ni

25%

50%

75%

100%

(a)

(b)

Figure 78.12 Side view models of the hydroxylated NiO(533) surface including a (010) step edge and a (111) terrace. Cladsorbed structures are shown before and

after optimization at coverages of 25%, 50%, 75%, and 100%. (Source: Bouzoubaa et al. 2009 [106]. Adapted with permission of Elsevier.)

78.6 Breakdown of the Passive Film and Initiation of Localized Corrosion

a possible bifurcation from the Cl adsorption-induced oxide thinning mechanism to a penetration-induced mechanism of passivity breakdown. This requires the saturation in adsorbed Cl of the step edges and their immediate vicinity but not necessarily of the extended defect-free terraces. Intergranular sites, i.e. grain boundaries for a well-crystallized passive film, play a major role in passivity breakdown. STM and AFM studies performed on well-defined substrate surfaces revealed that they act as preferential nanostructural defects where nanopits nucleate both without and with aggressive anions (Cl− ) present in the electrolyte [94, 95, 108, 132, 133]. The presence of chlorides promotes the growth of the nanopits. A model has been proposed to take into account the effect of oxide grains boundaries on the mechanisms of passivity breakdown and localized corrosion [94, 95]. On alloys such as stainless steels, the film also consists of grains separated by intergranular boundaries, but with crystallinity depending on formation conditions as discussed in Section 78.5.2. However, not only do the structure exhibit nanoscale defects such as intergranular sites, but the most recent studies suggest that the Cr enrichment, which is a key parameter for the corrosion resistance, is not homogeneous at the nanoscale and varies between the nanograins themselves [134] and also depends on the coordination sites (steps vs. terraces) of the substrate [88] as a result of the growth mechanism of the passive film. This is illustrated by Figure 78.13 for a FeCrNiMo(100) single crystal alloy surface [88]. The presented STM image (Figure 78.13), obtained after passivation in sulfuric acid solution (E = 0.5 V/SHE, t = 2 hours), displays the terrace and step topography of the alloy surface with the covering oxide film presenting a nanogranular morphology. The increase of the lateral grain size, 11.5 ± 2.6 nm for the passive film vs. 5.3 ± 0.9 nm for the native air-formed oxide film, evidences a coalescence phenomenon of the oxide grains induced by electrochemical passivation. Noteworthy is the presence of depressions. Their penetration depth (2.27 ± 0.25 nm) is not only larger than before passivation (1.02 ± 0.20 nm) but exceeds the thickness of 1.9 nm

100 nm Figure 78.13 Nanoscale morphology of the Fe–l7Cr–l4.5Ni–2.3Mo(100) stainless steel surface observed by STM after passivation in 0.05 M H2 SO4 (aq) at +0.5 V/SHE for two hours. The oxide passive has a granular morphology (some grains are pointed) and covers the substrate terraces and step edges. Substrate terraces display depressions (some are circled) evidencing local

protection failure caused by competing transient dissolution during passivation. Substrate step edges (marked by dashed lines) are better corrosion resistant owing to preferential local Cr enrichment of the passive film. (Source: Maurice et al. 2015 [88]. Adapted with permission of The Royal Society of Chemistry.)

819

820

78 Application of Surface Science to Corrosion

measured for this passive oxide film [88]. These depressions result from transient dissolution that compete with the transformation of the oxide film induced by passivation. Hence nanoscale roughening occurs due to local competition between dissolution and oxide transformation during passivation. The oxide becomes further enriched in chromium after passivation: the Cr(III) cation fraction increases from 57% to 67% and the Fe(II)–Fe(III) cation fraction decreases from 41% to 29% as obtained from XPS analysis [88]. It can be concluded that dissolution is a marker for the least Cr-enriched areas of the passivated surface. Further examining the surface also reveals that dissolution is preferentially located on the terraces and not at the step edges of the substrate. This may appear counter intuitive from a surface science perspective and also based on the mechanism of active dissolution of pure metals showing preferential reactivity at step edges (Section 78.2). This difference arises from the competing passivation by Cr enrichment of the oxide film, favored at step edges and thus promoting the protection of these substrate sites. As a result, dissolution is relocated on the substrate sites where the protection provided by the oxide film is less effective because of a lower local Cr enrichment on the oxide film. These data suggest that the homogeneity of the Cr enrichment in the passive film is a key issue of the local corrosion resistance.

78.7 Conclusion

The application of surface analytical methods to investigate model interfaces with well-defined solid surfaces exposed to corroding liquid environments in wellcontrolled conditions enables to study the very early stages triggering a corrosion process before its propagation. In this chapter, selected examples on metals (Cu, Ag, Ni, Fe) and Cr-containing alloys (stainless steels) have been presented to illustrate how insightful such a surface science approach is to understand the mechanisms of dissolution and self-protection against corrosion by passive oxide films. The discussed aspects show that

• Atomic defects (i.e. step edges) play a major role in the active dissolution of metals occurring at the surface of crystalline grains. They are preferential sites of etching leading to a step flow mechanism with anisotropic effect induced by the surface atomic structure. • The mechanisms of active dissolution is not altered by strongly adsorbed layers of anions but its anisotropy is influenced. • For polycrystals, grains boundaries often are preferential sites for the initiation of active dissolution but in some cases (i.e. coherent twins) they can be as resistive as grains. • Hydroxide ions adsorption precedes 3D oxide growth at oversaturation for anodic oxidation. 2D overlayer structures of adsorbed hydroxyl groups act as structural precursors for the growth of 3D passive films by inducing reconstruction of the topmost metal plane that adopts the structural arrangement found in the 3D passivating oxide.

References

• Metal surfaces covered by 2D passivating oxide/hydroxide layers have a corro-

•

•

•

•

sion resistance sensitive to the local structure. Dissolution is then promoted at the weakest non-ordered sites, initiating 2D nanopits on the substrate terraces and subsequent 3D nanopit growth by a repeated mechanism on the newly exposed terraces. Oxide passive films are in most cases crystalline. The substrate structure influences the crystallographic orientation in which the oxide grows. The oxide surface is hydroxylated. Ageing under polarization is critical to the crystallization of chromium-rich passive films. Oxide grains of passive films expose a facetted surface owing to a few degree tilt of the oxide lattice with respect of the metal lattice. The step edges of the passive film surface are preferential sites of dissolution in the passive state. Grain boundaries are numerous in crystalline passive films on metals. They are preferential nanostructural defects for passivity breakdown and localized corrosion initiation. On stainless steel, the Cr enrichment of the passive films may be inhomogeneous at the nanoscale, influencing the local resistance to localized corrosion initiation.

References 1. Marcus, P. (ed.) (2011). Corrosion

2.

3.

4. 5. 6. 7.

8.

9.

10.

Mechanisms in Theory and Practice, 3e. CRC Press/Taylor & Francis, 941 pp. Marcus, P. and Mansfeld, F. (eds.) (2006). Analytical Methods in Corrosion Science and Engineering. CRC Press/Taylor & Francis, 760 pp. Taylor, C.D. and Marcus, P. (eds.) (2015). Molecular Modeling of Corrosion Processes – Scientific Development and Engineering Applications. Wiley, 256 pp. Suggs, D.W. and Bard, A.J. (1994). J. Am. Chem. Soc. 116: 10725. Suggs, D.W. and Bard, A.J. (1995). J. Phys. Chem. 99: 8349. Moffat, T.P. (1997). Mater. Res. Soc. Symp. Proc. 451: 75. Vogt, M.R., Lachenwitzer, A., Magnussen, O.M., and Behm, R.J. (1998). Surf. Sci. 399: 49–69. Magnussen, O.M., Vogt, M.R., Scherer, J., and Behm, R.J. (1998). Appl. Phys. A 66: S447. Broekmann, P., Anastasescu, M., Spaenig, A. et al. (2001). J. Electroanal. Chem. 500: 241–254. Magnussen, O.M., Zitzler, L., Gleich, B. et al. (2001). Electrochim. Acta 46: 3725.

11. Broekmann, P., Hai, N.T.M., and

12. 13. 14. 15. 16. 17.

18. 19.

20.

21. 22.

Wandelt, K. (2006). J. Appl. Electrochem. 36: 1241–1252. Phan, T.H., Kosmala, T., and Wandelt, K. (2015). Surf. Sci. 631: 207–212. Oppenheim, I.C., Trevor, D.J., Chidsey, C.E.D. et al. (1991). Science 254: 687. Moffat, T.P., Fan, F.-R., and Bard, A.J. (1991). J. Electrochem. Soc. 138: 3224. Chen, S.J., Sanz, F., Ogletree, D.F. et al. (1993). Surf. Sci. 292: 289. Stratmann, M. and Rohwerder, M. (2001). Nature 410: 420. Martín, H., Carro, P., Hernández Creus, A. et al. (2002). J. Phys. Chem. B 104: 8239. Renner, F.U., Stierle, A., Dosch, H. et al. (2006). Nature 439: 707. Pareek, A., Borodin, S., Bashir, A. et al. (2011). J. Electrochem. Soc. 133: 18264–18271. Shrestha, B.R., Bashir, A., Ankah, G.N. et al. (2015). Faraday Discuss. 180: 191–204. Suzuki, T., Yamada, T., and Itaya, K. (1996). J. Phys. Chem. 100: 8954. Dietterle, M., Will, T., and Kolb, D.M. (1995). Surf. Sci. 327: L495.

821

822

78 Application of Surface Science to Corrosion 23. Ando, S., Suzuki, T., and Itaya, K.

24.

25.

26.

27.

28.

29.

30. 31. 32. 33. 34.

35.

36. 37. 38.

39. 40.

(1997). J. Electroanal. Chem. 431: 277–284. Itaya, K. (1999). Atomic-scale aspects of anodic dissolution of metals: studies by in situ scanning tunneling microscopy. In: Interfacial Electrochemistry – Theory, Experiments and Applications, 187 pp. (ed. A. Wieckowski). New York, NY: Marcel Dekker. Ando, S., Suzuki, T., and Itaya, K. (1996). J. Electroanal. Chem. 412: 139–146. Teshima, T., Ogaki, K., and Itaya, K. (1997). J. Phys. Chem. B 101: 2046–2053. Sashikata, K., Matsui, Y., Itaya, K., and Soriaga, M.P. (1996). J. Phys. Chem. 100: 20027–20034. Kim, S.H., Erb, U., Aust, K.T., and Palumbo, G. (2001). Scr. Mater. 44: 835–839. Miyamoto, H., Yoshimura, K., Mimaki, T., and Yamashita, M. (2002). Corros. Sci. 44: 1835–1846. Hu, C., Xia, S., Li, H. et al. (2001). Corros. Sci. 5: 1880–1886. Randle, V. (2006). Scr. Mater. 54: 1011–1015. Gerrtsman, V.Y. and Gruemer, S.M. (2001). Acta Mater. 49: 1589–1598. Liu, M., Qiu, D., Zhao, M.-C. et al. (2008). Scr. Mater. 58: 421–424. Martinez-Lombardia, E., Lapeire, L., Maurice, V. et al. (2014). Electrochem. Commun. 41: 1–4. Lapeire, L., Martinez Lombardia, E., Verbeken, K. et al. (2014). J. Mater. Sci. 49: 3909–3916. Strehblow, H.-H. and Titze, B. (1980). Electrochim. Acta 25: 839. Kautek, W. and Gordon, J.G. (1990). J. Electrochem. Soc. 137: 2672. Millet, B., Fiaud, C., Hinnen, C., and Sutter, E.M.M. (1995). Corros. Sci. 37: 1903. Feng, Y., Siow, K.-S., Teo, W.-K. et al. (1997). Corrosion 53: 389. Sander, U., Strehblow, H.-H., and Dohrmann, J.K. (1981). J. Phys. Chem. 85: 447.

41. De Chiavo, M.R.G., Salvarezza, R.C.,

42. 43.

44.

45.

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

56.

57.

58.

and Arvia, A. (1984). J. Appl. Electrochem. 14: 165. Mayer, S.T. and Muller, R.H. (1992). J. Electrochem. Soc. 139: 426. Chan, H.Y.H., Takoudis, C.G., and Weaver, M.J. (1999). J. Phys. Chem. B 103: 357. Melendres, C.A., Bowmaker, G.A., Leger, J.-M., and Beden, B.J. (1998). J. Electroanal. Chem. 449: 215. Härtinger, S., Pettinger, B., and Doblhofer, K. (1995). J. Electroanal. Chem. 397: 335. Maurice, V., Strehblow, H.-H., and Marcus, P. (2000). Surf. Sci. 458: 185–194. Kunze, J., Maurice, V., Klein, L.H. et al. (2001). J. Phys. Chem. B 105: 4263–4269. Kunze, J., Maurice, V., Klein, L.H. et al. (2003). Electrochim. Acta 48: 1157–1167. Kunze, J., Maurice, V., Klein, L.H. et al. (2003). J. Electroanal. Chem. 554–555: 113–125. Kunze, J., Strehblow, H.-H., and Staikov, G. (2004). Electrochem. Commun. 6: 132. Maurice, V., Klein, L.H., Strehblow, H.-H., and Marcus, P. (2007). J. Phys. Chem. C 111: 16351. Zemlyanov, D.Y., Savinova, E.R., Scheybal, A. et al. (1998). Surf. Sci. 418: 441. Jovic, B.M., Jovic, V.D., and Stafford, G.R. (1999). Electrochem. Commun. 1: 247. Savinova, E.R., Zemlyanov, D., Pettinger, B. et al. (2000). Electrochim. Acta 46: 175. Savinova, E.R., Kraft, P., Pettinger, B., and Doblhofer, K. (1997). J. Electroanal. Chem. 430: 47. Savinova, E.R., Zemlyanov, D., Scheybal, A. et al. (1999). Langmuir 15: 6546. Lützenkirchen-Hecht, D., Waligura, C.U., and Strehblow, H.-H. (1998). Corros. Sci. 40: 1037. Macdonald, D.D. (1999). Passivity – the key to our metals-based civilization. Pure Appl. Chem. 71: 951–978.

References 59. Olsson, C.-O.A. and Landolt, D. (2003). 60.

61.

62. 63. 64.

65. 66. 67.

68.

69.

70.

71.

72. 73.

74.

75. 76.

Electrochim. Acta 48: 1093–1104. Strehblow, H.-H., Maurice, V., and Marcus, P. (2011). Passivity of metals. In: Corrosion Mechanisms in Theory and Practice, 3e (ed. P. Marcus), 235–326. CRC Press/Taylor & Francis. Marcus, P. and Maurice, V. (2012). Oxide passive films and corrosion protection. In: Oxide Ultrathin Films. Science and Technology (eds. G. Pacchioni and S. Valeri), 119–144. Weinheim: Wiley-VCH. Maurice, V., Talah, H., and Marcus, P. (1993). Surf. Sci. 284: L431. Maurice, V., Talah, H., and Marcus, P. (1994). Surf. Sci. 304: 98. Yau, S.-L., Fan, F.-R., Moffat, T.P., and Bard, A.J. (1994). J. Phys. Chem. 98: 5493. Zuili, D., Maurice, V., and Marcus, P. (2000). J. Electrochem. Soc. 147: 1393. Hirai, N., Okada, H., and Hara, S. (2003). Mater. Trans., JIM 44: 727. Scherer, J., Ocko, B.M., and Magnussen, O.M. (2003). Electrochim. Acta 48: 1169. Nakamura, M., Ikemiya, N., Iwasaki, A. et al. (2004). J. Electroanal. Chem. 566: 385. Seyeux, A., Maurice, V., Klein, L.H., and Marcus, P. (2005). J. Solid State Electrochem. 9: 337. Seyeux, A., Maurice, V., Klein, L.H., and Marcus, P. (2006). J. Electrochem. Soc. 153: B453. Maurice, V., Yang, W., and Marcus, P. (1994). J. Electrochem. Soc. 141: 3016. Zuili, D., Maurice, V., and Marcus, P. (1999). J. Phys. Chem. B 103: 7896. Bhardwaj, R.C., Gonzalez-Martin, A., and Bockris, J.O’.M. (1991). J. Electroanal. Chem. 307: 195. Ryan, M.P., Newman, R.C., and Thompson, G.E. (1995). J. Electrochem. Soc. 142: L177. Li, J. and Meier, D.J. (1998). J. Electroanal. Chem. 454: 53. Diez-Pérez, I., Gorostiza, P., Sanz, F., and Müller, C. (2001). J. Electrochem. Soc. 148: 307.

77. Rees, E.E., Ryan, M.P., and MacPhail,

78. 79. 80. 81. 82.

83.

84.

85. 86. 87.

88. 89.

90.

91.

92.

93.

94.

95.

96.

D.S. (2002). Electrochem. Solid-State Lett. 5: B21. Deng, H., Nanjo, H., Quian, P. et al. (2007). Electrochim. Acta 52: 4272. Ikemiya, N., Kubo, T., and Hara, S. (1995). Surf. Sci. 323: 81. Kunze, J., Maurice, V., Klein, L.H. et al. (2004). Corros. Sci. 46: 245. Foelske, A., Kunze, J., and Strehblow, H.-H. (2004). Surf. Sci. 554: 10. Ryan, M.P., Newman, R.C., and Thompson, G.E. (1994). Philos. Mag. B 70: 241. Ryan, M.P., Newman, R.C., and Thompson, G.E. (1994). J. Electrochem. Soc. 141: L164. Maurice, V., Yang, W., and Marcus, P. (1996). J. Electrochem. Soc. 143: 1182. Nanjo, H., Newman, R.C., and Sanada, N. (1997). Appl. Surf. Sci. 121: 253. Maurice, V., Yang, W., and Marcus, P. (1998). J. Electrochem. Soc. 145: 909. Massoud, T., Maurice, V., Klein, L.H., and Marcus, P. (2013). J. Electrochem. Soc. 160: C232–C238. Maurice, V., Peng, H., Klein, L.H. et al. (2015). Faraday Discuss. 180: 151–170. Machet, A., Galtayries, A., Zanna, S. et al. (2004). Electrochim. Acta 49: 3957. Magnussen, O.M., Scherer, J., Ocko, B.M., and Behm, R.J. (2000). J. Phys. Chem. B 104: 1222. Toney, M.F., Davenport, A.J., Oblonsky, L.J. et al. (1997). Phys. Rev. Lett. 79: 4282. Davenport, A.J., Oblonsky, L.J., Ryan, M.P., and Toney, M.F. (2000). J. Electrochem. Soc. 147: 2162. Reikowski, F., Maroun, F., Di, N. et al. (2016). Electrochim. Acta https://doi .org/10.1016/j.electacta.2016.01.052. Marcus, P., Strehblow, H.-H., and Maurice, V. (2008). Corros. Sci. 50: 2698. Seyeux, A., Maurice, V., and Marcus, P. (2009). Electrochem. Solid-State Lett. 12: C25. Islam, M.M., Diawara, B., Maurice, V., and Marcus, P. (2009). Surf. Sci. 603: 2087–2095.

823

824

78 Application of Surface Science to Corrosion 97. Mac Dougall, B. and Cohen, M. (1976). 98. 99.

100.

101. 102.

103. 104. 105.

106. 107. 108.

109. 110. 111. 112. 113.

J. Electrochem. Soc. 123: 191. Mac Dougall, B. and Cohen, M. (1976). J. Electrochem. Soc. 123: 1783. Dickinson, T., Povey, A.F., and Sherwood, P.M.A. (1977). J. Chem. Soc., Faraday Trans. 1 F 73: 327. Marcus, P., Oudar, J., and Olefjord, I. (1979). J. Microsc. Spectrosc. Electron. 4: 63. Hoppe, H.-W. and Strehblow, H.-H. (1980). Corros. Sci. 20: 167. Mitchell, D.F., Sproule, G.I., and Graham, M.J. (1985). Appl. Surf. Sci. 21: 199. Wagner, F.T. and Moylan, T.E. (1989). J. Electrochem. Soc. 136: 2498. Hoppe, H.-W. and Strehblow, H.-H. (1989). Surf. Interface Anal. 14: 121. Zuili, D., Maurice, V., and Marcus, P. (1997). In situ investigation by ECSTM of the structure of the passive film formed on Ni(111) single-crystal surfaces. In: Passivity and Its Breakdown, The Electrochemical Society Proceedings Series, PV 97-26 (eds. P. Natishan, H.S. Isaacs, M. Janik-Czachor, et al.), 1013. Pennington, NJ: The Electrochemical Society. Bouzoubaa, A., Diawara, B., Maurice, V. et al. (2009). Corros. Sci. 51: 2174. Bouzoubaa, A., Diawara, B., Maurice, V. et al. (2009). Corros. Sci. 51: 941. Maurice, V., Nakamura, T., Klein, L., and Marcus, P. (2007). Initial stages of localised corrosion by pitting of passivated nickel surfaces studied by STM and AFM. In: Local Probe Techniques for Corrosion Research, vol. 45 (eds. R. Oltra, V. Maurice, R. Akid and P. Marcus), 71. Cambridge: Woodhead Publishing Ltd./CRC Press LLC. Lu, Z. and Macdonald, D.D. (2008). Electrochim. Acta 53: 7696–7702. Keller, P. and Strehblow, H.-H. (2004). Corros. Sci. 46: 1939. Haupt, S. and Strehblow, H.-H. (1995). Corros. Sci. 37: 43. Marcus, P. and Olefjord, I. (1988). Corros. Sci. 28: 589. Olefjord, I. and Brox, B. (1983). Passivity of Metals and Semiconductors (ed. M. Froment), 561. Amsterdam: Elsevier.

114. Mitchell, D.F. and Graham, M. (1987).

Surf. Interface Anal. 10: 259. 115. Mischler, S., Mathieu, H.J., and

116. 117. 118. 119.

120. 121.

122. 123. 124. 125. 126. 127.

128.

129.

130.

Landolt, D. (1988). Surf. Interface Anal. 11: 182. Calinski, C. and Strehblow, H.-H. (1989). J. Electrochem. Soc. 136: 1328. Kirchheim, R., Heine, B., Fischmeister, H. et al. (1989). Corros. Sci. 29: 899. Castle, J.E. and Qiu, J.H. (1989). Corros. Sci. 29: 591. Hubschmid, C., Mathieu, H.J., and Landolt, D. (1993). Proceedings of the 12th International Congress, 3913. Houston, TX: NACE. Yang, W.P., Costa, D., and Marcus, P. (1994). J. Electrochem. Soc. 141: 2669. Oblonsky, L.J., Ryan, M.P., and Isaacs, H.S. (1998). J. Electrochem. Soc. 145: 1922. Hamm, D., Ogle, K., Olsson, C.-O.A. et al. (2002). Corros. Sci. 44: 1443. Olefjord, I. and Elfström, B.-O. (1982). Corrosion 38: 46. Brox, B. and Olefjord, I. (1988). Surf. Interface Anal. 13: 3. De Vito, E. and Marcus, P. (1992). Surf. Interface Anal. 19: 403. Olsson, C.-O.A. and HörnstrÖm, S.E. (1994). Corros. Sci. 36: 141. Marcus, P. and Maurice, V. (1998). Comparison of atomic structures of passive films on chromium and on ferritic and austenitic stainless steels. In: Passivity and Its Breakdown, The Electrochemical Society Proceedings Series, PV 97-26 (eds. P. Natishan, H.S. Isaacs, M. Janik-Czachor, et al.), 254. Pennington, NJ: The Electrochemical Society. Frankel, G.S. (1998). Pitting corrosion of metals. A review of the critical factors. J. Electrochem. Soc. 145: 2186–2198. Strehblow, H.-H. and Marcus, P. (2011). Mechanisms of pitting corrosion. In: Corrosion Mechanisms in Theory and Practice, 3e (ed. P. Marcus), 349. CRC Press/Taylor & Francis. Maurice, V., Klein, L.H., and Marcus, P. (2002). Surf. Interface Anal. 34: 139–143.

References 131. Bouzoubaa, A., Costa, D., Diawara,

B. et al. (2010). Corros. Sci. 52: 2643–2652. 132. Maurice, V., Inard, V., and Marcus, P. (1999). STM investigation of the localized corrosion of passivated Ni(111) single-crystal surfaces. In: Critical Factors in Localized Corrosion III, The Electrochemical Society Proceedings Series, PV 98-17 (eds. P.M. Natishan,

R.G. Kelly, G.S. Frankel and R.C. Newman), 552–562. Pennington, NJ: The Electrochemical Society. 133. Maurice, V., Klein, L.H., and Marcus, P. (2001). Electrochem. Solid-State Lett. 4: B1–B3. 134. Massoud, T., Maurice, V., Wiame, F. et al. (2014). Corros. Sci. 84: 198–203.

825

827

79 Electrons, Electrodes, and the Transformation of Organic Molecules Robert Francke, Luisalberto Gonzalez, R. Daniel Little, and Kevin D. Moeller

79.1 Introduction

Electrochemistry is an outstanding method for transferring electrons to and from a wide variety of molecules [1, 2]. It allows for the manipulation of molecular oxidation states, the generation of highly reactive intermediates, the use of normally stoichiometric electron transfer agents as catalysts, the recycling of chemical waste products, and the initiation of an impressive array of chemical reactions that stem from a reversal in the polarity of a functional group, substrate, or catalyst [3]. In this way, electrochemistry as a field of endeavor has had a large positive impact on the analytical chemistry community, the probing of molecular interactions of biological relevance, the development of new, more sustainable means for energy conversion, and the synthesis and characterization of new organometallic species. Yet while these methods have been very successful, and a host of very talented individuals have shown how electrochemical methods can be applied to organic systems and used in synthetic efforts, electrochemistry has to date seen only limited use by the larger synthetic community. This is changing, and the use of electrochemistry as a synthetic method has been growing with a number of more traditional synthetic groups not only utilizing reactions developed by others but also advancing new electrochemical methods of their own. This recent surge in the use of electrochemistry is being driven by our increased understanding of electrode processes and the “downstream” chemistry that they trigger. As organic chemists become more familiar with the physical organic chemistry principles that play a role in determining the outcome of electrochemical reactions, they become more comfortable with the design and implementation of their own reactions. With this in mind, the review that follows is aimed at both summarizing recent synthetic advances in the general area of electroorganic synthesis and providing insights into the mechanistic considerations one needs to design reactions on their own.

Surface and Interface Science: Applications of Surface Science II, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

828

79 Electrons, Electrodes, and the Transformation of Organic Molecules

79.1.1 Electrochemical Basics

To fully understand the synthetic and mechanistic chemistry highlighted below, it is helpful to have an understanding of a few requirements of an electrochemical reaction [4]. Relative to what is often assumed by synthetic chemists at large, the reactions are most often exceedingly simple to perform. At their core is the passage of current through a reaction solution. There are typically two ways in which this can be accomplished. The first makes use of a controlled potential at the working electrode (CPE). In this type of experiment, three electrodes are used: a cathode, an anode, and a reference electrode. The potential at either the anode or the cathode is then held constant relative to the reference electrode (typically Ag/AgCl or Hg/Hg2+ ). The potential selected matches that of a substrate in solution that is to be either oxidized (anode) or reduced (cathode). The reaction proceeds by a transfer of an electron between the working electrode and the substrate, while a sacrificial group is either oxidized or reduced at the counter electrode. As the substrate is consumed, the current flowing between the working and counter electrodes drops. Electrolysis reactions run in this fashion are advantageous because they are highly selective for the substrate to be oxidized or reduced because the potential at the working electrode remains fixed. The reactions have disadvantages in that they require a reference electrode and they have difficulty going to completion because the rate of the reaction drops as the substrate is consumed. The second method for conducting an electrolysis reaction makes use of a constant current electrolysis (CCE) or galvanostatic experiment. In a CCE, only a working electrode and a counter electrode are required. A constant current is then passed between these two electrodes, and the potential at each is allowed to vary. In a CCE, the potential at the working electrode initially climbs until it matches that of the most easily oxidized (anode) or reduced (cathode) substrate in solution. The potential of the electrode then remains at that potential as long as there is enough substrate present to satisfy the constant current that must be passed through the cell. When enough of the substrate is consumed so that it cannot satisfy the current that is being passed through the cell, then the potential at the electrode climbs again until it reaches a value consistent with oxidation or reduction of the substrate in solution with the next highest potential. The electrode potential then holds constant at that value and the cycle repeats. Constant current reactions are advantageous in that they are easy to run. They require no reference electrode and no knowledge of a substrate’s potential. The system simply adjusts to match whatever oxidation/reduction potential is needed. The reactions have a major disadvantage in that they will lose selectivity as the reaction proceeds, the substrate is consumed, and the working potential at the electrode needs to climb. This disadvantage can be minimized by keeping the current density at the electrode low (thus requiring less substrate to consume the current), but it remains a problem if one needs to push a reaction to 100% completion. Because each method has advantages and disadvantages, both are used depending on the particular needs of a reaction. With that said, constant current reactions have

79.1 Introduction

been of particular interest in many modern reactions because of the ease with which they can be run. Both types of reactions can be run in either a divided or an undivided cell. In a divided cell (Figure 79.1), the anodic chamber is separated from the cathodic chamber by either a glass frit or a permeable membrane. This makes it impossible for the cathode to interfere with the reaction at the anode and vice versa. In an undivided cell (Figure 79.2), the two chambers are not separated. The reaction medium remains

Figure 79.1 A divided cell.

Figure 79.2 An undivided cell.

829

830

79 Electrons, Electrodes, and the Transformation of Organic Molecules

neutral since any acid generated at an anode can react with the corresponding base generated at the cathode. Both reactions are used, but if the interaction of a substrate or product with the counter electrode is not a problem, then the use of an undivided cell is frequently favored because of its simplicity. 79.1.1.1 Half Reactions, Electrolytes, and Electrodes

One common misunderstanding about electrochemical reactions is that charge flows through the reaction cell. While it is true that electrons are introduced into solution at the cathode and removed from the solution at the anode, this net electron transfer is accomplished with a pair of “half reactions.” For a complete electrochemical reaction, one substrate is reduced at the cathode, while a second is oxidized at the anode. In this way, there is a net electron transfer and the formation of two different products from two different substrates. With this in mind, the electrolyte present in an electrolysis reaction does not “carry charge” through the cell. Instead, the electrolyte provides counterions for the reactions that occur at the electrodes. At the cathode, electrons are transferred from the electrode to a substrate. The result is the formation of anionic intermediates and products. The electrolyte provides the positively charged counterions for those anions. The opposite process occurs at the anode where electrons are transferred from the substrate to the electrode, an event that leads to the formation of cationic intermediates and products. The electrolyte provides the negatively charged counterions for those cations. In this way, the electrolyte helps maintain a low resistance of the cell to the current. Without the electrolyte, positive charge builds at the anode, and negative charge builds at the cathode. As this charge builds, it becomes harder to conduct the oxidation and reduction reactions necessary at the electrodes, and the cell resistance climbs. The electrolyte can play additional roles in an electrolysis reaction. When a charged electrode is placed in a solution containing an electrolyte, the solution near its surface becomes ordered. This ordering is frequently referred to as a “double layer” or an outer Helmholtz layer in physical chemistry terms [1, 5]. In its simplest form, the double layer can be viewed as an initial coating of a positively charged anode or negatively charged cathode with the oppositely charged counterion. In other words, when one places a positively charged anode into a solution with electrolyte, it initially becomes coated with a layer of anion. The opposite happens at the cathode. At this point, the anode appears negative to the solution, so a second positive layer of electrolyte is attracted to the surface. The process is repeated for multiple layers, but each time becomes less ordered and the charge more diffuse. Since much of the chemistry that happens at the anode occurs in the two most organized layers closest to the electrode, the concept of a double layer became popular. The presence of this double layer has real consequences with respect to reactions that occur at the electrode surface. First, the double layer can exclude solvent from the area close to an electrode and slow diffusion. The net result is that reactive intermediates generated at the electrode surface are frequently protected from both solvent trapping and dimerization. For example, the Monsanto hydrodimerization of acrylonitrile to form adiponitrile made a radical anion of

79.1 Introduction

the acrylonitrile in water solvent without the anion being protonated [6]. This occurred because of the use of tetraethylammonium tosylate as the electrolyte for the reaction. The tetraethylammonium tosylate formed a hydrophobic double layer that excluded water from the surface of the electrode and in so doing extended the lifetime of the radical anion intermediate. Similar effects have been seen for the generation of radical cations at an anode [7]. In fact, double-layer effects have been used to optimize the chemoselectivity of radical cation reactions governed by the Curtin–Hammett principle (Scheme 79.1) [8]. In this case, a cyclization reaction was favored over solvent trapping by using a hydrophobic electrolyte to exclude the methanol solvent from the region surrounding the anode.

S

RVC anode Pt cathode

S

OMe

OMe

S

S

+

S

+

30% MeOH/ THF 2,6-Lutidine 8 mA/ 2.5 F/mol

OMe

OMe

OMe

Fast cyclization

S

Fast MeOH trapping

S

MeO

OMe

+ OMe

MeO

OMe OMe 2 20% 50% 60%

OMe 1b

1a

1

0.1 M LiClO4: 0.1 M Et4NOTs: 0.5 M Et4NOTs:

S

+

MeO

OMe

OMe

MeO

OMe OMe 3 21% 10% 7%

OMe

OMe 4 15% 6% 2%

Scheme 79.1

Because electrolytes play such a critical role, significant effort is being made to optimize them. While organic electrochemistry is ideally suited for the use of electrolytes because they are easy to wash away during a workup, their presence can hurt the overall sustainability of an electrolysis reaction. For this reason, a number of approaches have been developed in an effort to eliminate or at least greatly reduce the amount of supporting electrolyte used in an electrolysis. In so doing, waste is reduced and product isolation is simplified. One of these approaches has focused on the development of polymer-based electrolytes. Wend and Steckhan, for example, devised a clever system wherein a redox mediator was covalently attached to a polyelectrolyte backbone [9]. Yoshida et al. have appended reagents that are capable of epoxidizing alkenes while simultaneously serving as a conducting medium [10]. A similar polymer-based approach allowed Zupan and Dolenc to mediate benzylic oxidation [11]. In another important

831

832

79 Electrons, Electrodes, and the Transformation of Organic Molecules

variation on this theme, Tajima and Fuchigami established that in the presence of methanol, silica gel-supported piperidine bases equilibrate to form methoxide and the protonated amine, whose role is to serve both as a base in subsequent acid–base chemistry and as the supporting electrolyte [12]. In a second approach, the use of microflow reactors also allows one to minimize the use of a supporting electrolyte [13], occasionally allowing it to be avoided entirely [14]. In a similar vein, thin layer flow cells as well as laminar flow systems have also been developed and utilized. They, too, permit electrolyses to be conducted in the absence of supporting electrolyte [15]. Thirdly, a recyclable composite dispersion that is stable at room temperature for more than 1.5 years has been shown to behave as an easily recovered and reused surrogate for a traditional supporting electrolyte. The composite is formed by sonicating a slurry consisting of PDDA-triflimide (PDDA, polymer of diallyldimethylammonium chloride), Super P carbon black, and an organic solvent. The medium has served exceedingly well in a variety of different oxidative processes, including the simple oxidation of benzylic alcohols, a Friedel–Crafts-like heteroarylation reaction [16], and the α-heteroarylation of N-vinyl amides [17]. While the composite dispersion has proven an effective medium for oxidations, its use in reductive processes is limited by virtue of the fact that the polymeric framework (PDDA) is a polycation and can compete with substrate for access to the cathode, thereby shutting down electroreduction of the substrate [18]. In an electrolysis reaction, the electrolyte selected for the reaction is matched with a solvent that is typically easy to select. Two features guide that selection. The solvent must be either inert to the electrolysis or must play a defined role in the reaction, and it must solubilize both the substrates and the electrolyte. Water, methanol, acetonitrile, THF, benzene, dichloromethane, and ionic liquids have all been employed as solvents in electrochemical reactions. One of the key statements in the preceding paragraph is that the solvent can play a defined role in an electrolysis reaction. For example, consider again the chemistry illustrated in Scheme 79.1. In this reaction, the methanol solvent used for the electrolysis played two key roles. First, it was the substrate for the half reaction conducted at the cathode. This reaction reduced the methanol to 2 equiv. of methoxide and 1 equiv. of hydrogen gas for each equivalent of substrate oxidized. Second, the methoxide ion traps the cations generated by the oxidation. This dual role for the solvent is common for many oxidations, and it is the reason that so many of the reactions are run in methanol. In addition to the choice of an electrolyte and solvent, the selection of an electrode can be important. For oxidation reactions, common anode materials include carbon, platinum, and more recently boron-doped diamond [19]. In many cases, the cathode used is simply a material that allows for a facile reduction of protons (carbon or platinum) since that is the most common half reaction at the auxiliary electrode. For reductions, copper, lead, carbon, and mercury are common cathode materials since they have a higher hydrogen overpotential (essentially the kinetic barrier to reducing protons at the surface of the electrode). This allows for the desired reduction reaction. Carbon, platinum, or sacrificial Mg anodes are often used for the other half

®

79.1 Introduction

reaction. Sacrificial Mg anodes are particularly attractive for the generation of anion intermediates because they lead to the formation of Grignard-type species [20, 21]. 79.1.2 Direct Electrochemical Methods

With the basics in place, there are two types of electrochemical reactions commonly used in modern organic electrochemistry. In the first, electrons are directly transferred to and from substrates to make reactive radical, radical cation, radical anion, cation, or anion intermediates. The reactions are very powerful in that they allow for the generation of the intermediates from a wide variety of substrates utilizing the same reaction conditions. This enables structure–reactivity studies to be conducted on the intermediates in a manner that allows for development of the mechanistic insight needed to fully capitalize on the intermediates. Since electrochemical reactions remain at the initial pH selected for the reaction (equal amounts of acid and base are generated at the anode and cathode, respectively), the conditions utilized can be systematically varied so that one can study the oxidation and reduction of neutral, anionic, and protonated species. While direct electrochemical reactions are very powerful, they do have limitations. The reactions select based on potential. Whether that potential is set (controlled potential) or allowed to set itself (constant current), the material that oxidizes or reduces in the reaction is the one that oxidizes or reduces at the lowest potential. That is not true of many chemical oxidation and reduction reactions where the selectivity of the reaction is based on molecular interactions and not just the electron richness or poorness of a substrate. For this reason, indirect electrochemical reactions are an attractive alternative. 79.1.3 Indirect Electroorganic Synthesis: Basic Principles

A special case of electroorganic synthesis is represented by the indirect electrolysis. Here, a so-called mediator is electrochemically activated and serves as a redox shuttle between electrode and substrate (see Scheme 79.2) [22]. Therefore, the Mediator (inactive)

Heterogeneous electron transfer

Substrate

Homogeneous electron transfer

Mediator (activated)

Reactive intermediate

Electrode Electrolyte

Scheme 79.2 Concept of the indirect electrolysis.

Chemical reaction(s)

Product

833

834

79 Electrons, Electrodes, and the Transformation of Organic Molecules

heterogeneous electron transfer between electrode and substrate is replaced by a homogeneous redox reaction. Ideally, the oxidized form and the reduced form of the mediator act as a reversible redox couple and can therefore be employed in catalytic quantities. In this sense, an indirect electrolysis can be considered as a hybrid between a direct electrolysis and a classical homogeneous redox reaction. The electrochemical activation and regeneration of the mediator can either be conducted in situ in a so-called in-cell process or separately (“ex-cell process”) [23]. Using the in-cell approach, the redox agent can be employed catalytically, since a continuous regeneration at the electrode is possible during electrolysis. In this case, the activated form of the mediator must only be sufficiently stable to react with the substrate in situ. In contrast, the active species must have a sufficient stability for a transfer to a second reaction vessel if it is prepared ex-cell. In addition, a more difficult separation protocol may result, since the mediator has to be employed in stoichiometric quantities. Compared with direct electrolysis, the use of redox mediators in indirect processes offers several advantages and is therefore receiving a growing significance [22]: 1. Redox mediators can feature higher or completely different selectivity. In fact, the use of mediators creates an immense number of opportunities to lead the selectivity of the electrochemical reaction in the desired direction and therefore constitutes an important tool for the organic electrochemist. 2. Kinetic inhibitions (the overpotential) typically related to the heterogeneous ET between electrode and starting material can be reduced, and a smooth conversion therefore facilitated. 3. The use of a mediator can be helpful when direct oxidation or reduction causes passivation of the electrode, since direct interaction of the electrode surface with the starting material is avoided. Mechanistically, two types of homogeneous ET processes can be differentiated (see Scheme 79.3) [22]. The type of mechanism depends both on the nature of the mediator and on the kind of substrate. In the first case, an outer-sphere (nonbonded) ET proceeds between the previously activated mediator and the substrate. If at an Med ) is less positive than indirect oxidation the oxidation potential of the mediator (Eox Sub of the substrate (conversely for indirect reductions), the the oxidation potential Eox ET equilibrium is located on the side of the starting material (whereby the exact Med − E Sub ). In other words, the mediator has to location is a function of ΔE = Eox ox Case 1: Outer-sphere ET

Medox + Substrate

Case 2: Inner-sphere ET

Medox + Substrate

Electron transfer

Bond formation

Medred + [Substrate]n+

[Med Substrate]

Chemical reaction(s)

Medred + [Substrate]n+

Chemical reaction(s)

Scheme 79.3 Comparison between outer-sphere (Case 1) and inner-sphere ET (Case 2) in the case of indirect anodic oxidation.

79.1 Introduction

react against the potential gradient. In such reactions, a full conversion is only possible if an irreversible chemical reaction follows the homogeneous ET, shifting the position of the equilibrium toward the product side. In fact, redox-mediated electrolyses involving outer-sphere ET have been reported with potential differences of up to 0.5 V (particular examples for such cases will be discussed in Section 79.3.2.1) [23–25]. It is important to keep in mind that such an “unfavorable” constellation of the redox potentials is a precondition for an in-cell process, since otherwise the substrate would be converted directly despite the presence of the mediator (including the possible disadvantages of direct conversion of substrates, which was discussed above). Conducting the electrolysis at potentials lower than the redox potential of the starting material also means that the reaction is carried out under milder conditions. This can become particularly important when sensitive functional groups are present on the substrate molecule, and undesired side reactions at these positions have to be avoided. In the second case depicted in Scheme 79.3, the so-called inner-sphere mechanism, the process is initiated by formation of an adduct between starting material and activated mediator. Subsequent cleavage of this adduct renders the inactive form of the mediator and the oxidized (reduced) form of the starting material. A further possibility is represented by an inner-sphere ET via formation of a charge transMed and E Sub of up to 1.5 V can be overcome fer complex. In case 2, a ΔE between Eox ox (some examples will be discussed in Section 79.3.2.2). In addition, such reactions are generally more selective compared with those involving outer-sphere ET, since the selectivity is rather determined by the chemical reactivity than the redox potentials of substrate and mediator. Since the mediator plays a crucial role in the successful implementation of an indirect electrolysis, one has to select this species carefully. In addition to the consideration of the aspects discussed above, one has to ensure that both the oxidized form and the reduced form of the mediator are inert to all components of the system except for the desired electron transfer(s), since irreversible side reactions reduce the catalytic activity dramatically. Furthermore, a fast reaction between mediator and substrate is clearly advantageous. In the case of a slow homogeneous reaction, the use of a divided cell (or alternatively the implementation in an ex-cell process) is required in order to avoid a discharge of the mediator at the counter electrode. In this sense, the optimization of an indirect electrolysis can be more challenging, but on the other hand the results are typically rewarding with regard to selectivity and overall efficiency. For preliminary assessment of the feasibility of a mediated process, cyclic voltammetry is particularly useful. First, the ΔE between catalyst and substrate can be measured, and the potential at which the mediated reaction is supposed to proceed can be determined. Furthermore, the efficiency of the redox mediation can be evaluated by comparing the voltammetric response at the potential of the mediator Emed in the presence and absence of substrate. In the case of a fast electron transfer and follow-up reaction, a so-called catalytic current can be observed. Typical voltammograms for the case of an indirect anodic oxidation are exemplified in

835

79 Electrons, Electrodes, and the Transformation of Organic Molecules

0.6

(3)

Med Med + S-H Med + S–H + B S–H + B

0.5 0.4 Current, i(mA)

836

0.3

(2) Medox + S–H Anode

0.1

(2) (1)

(3) Medox + S–H

0.0 −0.1 −0.2 0.4 (a)

Anode k0

(1) Medred

(4)

Effect of base

0.2

Corresponding processes:

Anode

ΔE

Med

E ox 0.6

0.8

1.0

Potential, E(V)

S–H

E ox 1.2

(4) S–H

kf,1

Medred + [S–H]

kb,1 kf,1 kb,1

k0′ Anode

1.4

Medox

[S–H]

+

Medred + [S–H]

+

+ kf,2 +

S

–H

k

f,3

+

–H

S

(b)

Figure 79.3 (a) The effect of mediator (Med) and base (B) on the voltammetric profile of the electrooxidation of a substrate S–H. (b) The processes corresponding to the individual voltammograms.

Figure 79.31) . Here a single ET from substrate (S−H) to a mediator (Medox ) is followed by deprotonation of the resulting radical cation [S−H]•+ . The deprotonation of such reactive intermediates is very common and represents a typical example for a rapid and thermodynamically favorable consecutive reaction (see equations in Figure 79.3). Compared to the voltammogram corresponding to the mediator in absence of substrate (black line, Eq. (1)), the anodic peak current increases slightly upon addition of an excess amount of S−H (red line, Eq. (2)). Due to partial consumption in the equilibrium reaction, the active form Medox is not entirely available for cathodic reduction in the reverse scan, leading to a diminished cathodic peak and a rather irreversible behavior. As a consequence of the addition of a base (B), the catalytic current is significantly increased (dashed line, Eq. (3)), since the rate of the following deprotonation step is dramatically accelerated. The opportunities for studying the mechanism of the reaction and for quantification of the catalytic performance are manifold, and for an in-depth treatment, the interested reader is referred to Refs. [26–28]. As pointed out above, for indirect in-cell oxidations, the redox potential of the mediator has to be less positive than the potential of the substrate (less negative in the case of indirect reductions) to allow for a selective electron transfer from the electrode to the redox catalyst. This means that under potentiostatic conditions, the potential of the working electrode is adjusted to the potential of the mediator, not to the one of the substrate. In a CCE, the compounds present in the reaction mixture are reacting sequentially in the order of increasing redox potential. Accordingly, the mediator is selectively oxidized/reduced in the presence of starting material as 1) The voltammetric profiles were simulated using the DigiElch package (Elchsoft). Sweep rates, diffusion coefficients and mediator concentrations for all simulations: v = 20 mV/s; D = 10–5 cm2 /s; c(Medred ) = 1 mM. Specific parameters for (1): E0 = 0.7 V; 𝛼 = 0.5; k 0 = 10 cm s−1 . For (2): c(S–H) = 10 mM; k f,1 = k b,1 = 10 L2 /mol2 /s. For (3): k f,2 = 500 L/mol/s. For (4): E0 = 1.3 V; 𝛼 = 0.5; k 0 ′ = 10 cm/s. c(S–H) = 2 mM; k f,3 = 500 L/mol/s.

79.2 Reductions and the Cathodic Half Reaction

long as the mediator is consumed at a rate that does not exceed diffusion of starting material to the electrode surface.

79.2 Reductions and the Cathodic Half Reaction 79.2.1 Direct Reductions at the Cathode

The intent of this section is to provide the reader with an overview that focuses on the cathodic reduction of electron-deficient alkenes. It is by no means intended to be comprehensive. The reader interested in such a treatment is referred, in particular, to the fifth edition of Organic Electrochemistry, edited by Hammerich and Speiser [1]. 79.2.1.1 Formation and Umpolung Property of Radical Anions

Reduction of a 10−3 M solution of enone 5 dissolved in dry dimethylformamide (DMF) containing 0.1 M n-Bu4 NClO4 at 25–30 ∘ C, leads to a well-resolved EPR spectrum corresponding to the radical anion 6 [29]. Analysis of the hyperfine splitting indicates that 40–50% of the unpaired spin density resides at the β-carbon, none at the alpha position, and the remainder dispersed between the carbon and oxygen of the carbonyl group (Scheme 79.4). The synthetically most useful characteristic of the radical anions derived from electron-deficient alkenes is that the β-carbon behaves as though it is nucleophilic [3]. O (H3C)3C

O −

+e C(CH3)3

(H3C)3C

5

C(CH3)3 6

Scheme 79.4 Reduction of an olefin.

The nucleophilic behavior is exemplified by radical anion dimerization, a process that is referred to as electrohydrodimerization (EHD) [30]. Thus, reduction leads to an umpolung at the β-carbon and does so without the need to modify the structure of the starting material. This differs, of course, from non-electrochemical approaches. The well-established use of Breslow–Stetter intermediates, for example [31], necessitates the formation of thiazolium ylides in order to achieve nucleophilic addition to formally electrophilic sites. 79.2.1.2 Electrohydrodimerization (EHD)

The best-known example of EHD involves the electrochemical β,β-coupling of acrylonitrile to afford adiponitrile (7), a precursor of Nylon 6,6 [6]. Remarkably, the reaction occurs via the coupling of two radical anions, despite the fact that water is used as the solvent. Thus, thanks to the double layer, formed by the interaction

837

838

79 Electrons, Electrodes, and the Transformation of Organic Molecules

of the tetraethylammonium cation with the cathode surface, the radical anions are born near one another and are screened from the aqueous environment long enough to permit the coupling to occur (Scheme 79.5). 2H2C=CHCN

+2e−, +2HA NC Et4NOTs, H2O

CN 7

Scheme 79.5 Electrohydrodimerization.

Both homocoupling and mixed coupling reactions have been investigated, as have asymmetric variations [32, 33]. Recently, cathodic reductive coupling of methyl cinnamate on boron-doped diamond electrodes has been successfully employed as a convenient synthetic route to neo-lignin-like substances [34]. 79.2.1.2.1 Radical Anion Lifetime The lifetime of radical anions depends critically upon the conditions under which they are generated. Not unexpectedly, the presence of acidic protons within the substrate or in the reaction medium reduces the lifetime. For example, that of the di-t-butyl ketone 5 is >10 seconds at ambient temperature in dry DMF and diminishes to 2 V substrate and the mediator. Consequently, even aliphatic alcohols with Eox vs. Ag/AgNO3 can be efficiently oxidized.

79.3 Oxidations and the Anodic Half Reaction

Conveniently, N-oxyl radical-based mediator systems provide many opportunities to direct the selectivity toward the desired outcome. For instance, the choice of the electrolyte represents an important tool to influence the chemoselectivity of the reaction: whereas under anhydrous conditions, the overoxidation to carboxylic acids can be entirely avoided, electrolysis in aqueous basic electrolyte leads to the carboxylic acid exclusively [122]. In this way, the reactions exactly mimic the selectivity of purely chemical reagent-based alcohol oxidations. Another interesting feature of this mediator system is the possibility for control of the chemoselectivity. Since primary alcohols react much faster than secondary alcohols, a selective transformation of primary alcohols in the presence of secondary hydroxy groups is generally feasible [122, 123]. Again, like other chemical oxidants that use sterics to achieve chemoselectivity, the course of the reaction can be channeled away from the more electron-rich alcohol that would be favored by a pure electron transfer reaction. A further tool for control of reaction rate and selectivity is the variation of the N-oxyl species itself. For the oxidation of sterically hindered secondary alcohols, the use of a mediator with a smaller steric profile such as 9-azabicyclo[3.3.1]nonane N-oxyl (commonly abbreviated as ABNO; see Figure 79.6) is advantageous with regard to the reaction rate. For less hindered secondary and primary alcohols, Med plays a more important role than Stahl and coworkers recently found that Eox the steric environment of the N-oxyl radical (“driving force trumps sterics”) [124]. Med = 0.52 V vs. Consequently, 4-acetamido-TEMPO (ACT, see Figure 79.6) with Eox Ag/AgNO3 exhibits much higher catalytic currents for the indirect electrooxidaMed = 0.35 V (despite the more tion of alcohols than the ABNO mediator with Eox benign steric profile of the latter species). It should be noted that the superior electrocatalytic performance of ACT is attributed to the enhanced electrophilicity of the corresponding oxoammonium species ACT+ , which facilitates the formation Med generally reflects the electrophilicity of the of intermediate adduct 135. Since Eox oxoammonium ion, it can be used as a readily accessible parameter to assess the electrocatalytic performance of an N-oxyl mediator. Another development with respect to the use of N-oxyl radical mediators involves the control of stereoselectivity in an oxidation reaction. The work centers on the use of optically active N-oxyl mediators such as compounds 136 and 137. These systems have been used for anodic oxidation of racemic secondary alcohols, whereby a kinetic resolution was achieved. In such cases, the product mixture consists of ketone and an enantiomerically enriched alcohol [125, 126]. iPr H N N

ABNO

O

N O N

O

ACT

O

N AcHN

136

137

Figure 79.6 Selection of several N-oxyl-based mediators for different applications.

O

871

872

79 Electrons, Electrodes, and the Transformation of Organic Molecules

Very recently, Baran and coworkers broke new ground in the field of N-oxyl radical mediated electrosynthesis with their allylic C–H oxidation using a mediator system based on tetrachloro-N-hydroxyphthalimide (138; see Scheme 79.46), a compound that is readily available due to its widespread use as (nontoxic) flame retardant [127]. Using catalytic amounts of 138 and stoichiometric amounts of t BuOOH as oxygen donor, olefins 140 can be converted to enones 143 using an extremely practical and straightforward procedure (undivided cell, CCE). In contrast to the example depicted in Scheme 79.5 where the oxoammonium ion represents the active species in the catalytic cycle, here the anodically formed N-oxyl radical 139 reacts with the starting material. Compared to TEMPO and related mediators, phthalimide and especially the tetrachloro-substituted derivative 138 exhibit much higher oxidation potentials, which renders the formation of the oxoammonium species difficult. On the other hand, the electron-withdrawing keto and chloro functionalities provide the N-oxyl radical 139 with sufficient driving force to be a SET mediator. The mechanism of the transformation is presumably very similar to chemical oxidations with N-hydroxyphthalimide catalysts in combination with terminal oxidants [128]. Radical 139 oxidizes olefin 140 to the relatively stable allylic radical 141, which reacts with anodically formed t BuO• to give peroxide 142 (see Scheme 79.46). The latter finally decomposes in a non-electrochemical step to form t BuOH and enone 28. Cl

CCE Acetone/LiClO4 Pyridine

O

Cl N O Cl Cl 139

−

−e − H+

O

R 140

R 143

− tBuOH Cl Cl

O tBuOOH

N OH Cl Cl Anode

20 mol% mediator 23

O

138

O

R

OOtBu

− e−, − H+ R

141

142

Scheme 79.46 Indirect allylic C–H oxidation using tetrachloro-N-hydroxyphthalimide (138).

The Stahl group has shown how N-oxyl radical mediators can be used in concert with Cu(II) mediators to expand the scope of more sustainable electrochemical oxidations (Scheme 79.47) [62]. In this case, the doubly mediated system (bipyCu(I)OTf and TEMPO) was used to carefully control the transition state of alcohol oxidations (Scheme 79.47) in order to improve the efficiency and expand the scope of benzylic alcohol oxidations. While the reactions occurred at the potential associated with oxidation/recycling of the Cu(I) mediator (more than 0.5 V below the oxidation potential of TEMPO), there was clear cooperativity between the mediators. The Cu(I) mediator was responsible for a one-electron oxidation, while TEMPO nitroxyl

79.3 Oxidations and the Anodic Half Reaction

− 2e− pyrCu(I)OTf TEMPO

OH

Et3N, Bu4NClO4 CH3CN

X

O

X

X = OMe, Me, H, Br, CF3, NO2 Scheme 79.47 A doubly mediated approach to benzylic alcohol oxidation.

served as both a proton and electron acceptor in the overall oxidation that required the loss of one proton and two electrons in the key step. With the use of the doubly mediated system, the reactions were accelerated by the presence of an electron-poor aryl ring in direct contrast to reactions conducted either with a direct oxidation or the use of only TEMPO. This observation was consistent with deprotonation of the alcohol and formation of a copper-alkoxide intermediate prior to the oxidation being rate limiting. As in the earlier examples, the presence of the chemical mediators in this reaction introduced new mechanistic pathways, and in so doing new chemical selectivities, that were not possible for the direct electrochemical method. In another example where an N-oxyl radical mediator has been used to improve the scope of an electrochemical method, the Xu group has used both electrochemically generated TEMPO and iron-based chemical oxidants to trigger amidyl radical cyclizations that cannot be accomplished with a direct oxidation (Scheme 79.48) [129–131]. Using direct oxidation conditions, amidyl radicals can only be coupled to olefins that are activated with electron-donating groups. The donating groups aid in the removal of a second electron from the cyclized product, a step that is required for formation of the final product [132]. The mediated reactions have no such limitation. The reactions proceed fine even with unactivated olefins, a situation that leads to a much broader range of substrates being compatible with the method.

Ph (1)

O

RVC anode, Pt cathode 10 mA/ 1.5 F/mol

NH

Bu4NBF4, Na2CO3 H2O/ CH3CN TEMPO

O

Ph O

(2)

R1 N O

H N R4

RVC anode, Pt cathode 5 mA, Bu4NBF4 R1 R3 MeOH/THF/ Na2CO3 reflux Cp2Fe

Scheme 79.48 Oxidative generation of N-centered radicals.

N

O

O

R2 H

O

Ph N

N

O 90% R2 R3

N N O

R4 Yields = 59–94%

873

874

79 Electrons, Electrodes, and the Transformation of Organic Molecules

79.3.2.3 The Generality of the Inner-Sphere Approach

Of course the use of inner-sphere mediated oxidations is not limited to N-oxyl radicals. It can be used to improve the sustainability of a wide variety of chemical oxidations, each of which uses a chemical reagent to control the selectivity of a reaction in a manner not possible for a direct electrolysis. In the interest of space, two representative examples will be illustrated here (Scheme 79.49) [66a]. In the first, a sunlight-driven electrolysis was used to conduct an asymmetric dihydroxylation by recycling a chiral Os(VIII) catalyst at the anode. A doubly mediated system was used to improve the efficiency of the reaction as suggested by the Torii group [133]. The reaction led to the same levels of asymmetric induction and same yield of product obtained with the more traditional use of stoichiometric iron to recycle the osmium. Of course, the asymmetric formation of diol 145 was completely dependent on the presence of the chemical mediator and its participation in the transition state leading to the desired product. In the second example, an oxidative Heck reaction was conducted by employing Pd(II) as a catalyst. Once again, a doubly mediated system was used to aid in the electron transfer reaction with hydroquinone serving as the catalytic cooxidant. The success of the reaction is completely dependent on the Pd(II) catalyst that first mediates a selective CH activation of the aryl ring followed by a Heck reaction with an electron-poor olefin, a reaction that generates a Pd(0) species that is then converted back into Pd(II) by the anodically generated hydroquinone. RVC anode Cat. K2OsO2(OH)4 Cat. K3Fe(CN)6

(1)

OH OH

(DHQD)2-PHAL t-BuOH, K2CO3, H2O

144

145

Literature Yield = 95%, 97% ee Yield with sunlight: 92%, 94% ee

Me (2)

H N +

Me O

Carbon anode Me Pt-foil cathode Pd(OAc)2, Bu4NBF4 Hydroquinone AcOH, 1–3 mA, 2.2 F/mol

O Ot-Bu

H N

Me O

146

CO2t-Bu

Literature yield = 78% (NMR) Yield with sunlight = 68% Scheme 79.49 Solar-driven reagent generation.

Both examples illustrate how mediated electrochemical reactions can capitalize on the special chemical reactivity and selectivity of a chemical reagent without giving up the opportunity for more sustainable processes associated with the electrochemical method. In both cases, the reactions consume only sunlight as the source of energy and generate hydrogen gas as the only by-product.

79.4 Paired Electrochemical Reactions

79.4 Paired Electrochemical Reactions 79.4.1 Basic Principles

While much of the chemistry described in Section 79.3 focused on chemistry at the anode and how oxidations can be used to generate both reactive intermediates and reagents at that electrode, it is important to remember that all such oxidation reactions do require an opposing half reaction at the cathode. For an oxidation in protic medium, the counterreaction is frequently ignored because the reduction leads to the formation of hydrogen gas. Hence, the reactions afford the desired oxidation products while producing a benign by-product as the only stoichiometric by-product. However, for oxidations in aprotic medium and for reduction reactions, the counter reaction is frequently not so easily dismissed. In such cases, the counterreaction most often involves either electrolyte degradation or the use of a so-called depolarizer (an intentionally added sacrificial substrate). Both approaches can be problematic from the economic and ecologic point of view (including the resulting difficulties for the product purification). While these issues can be resolved ((i) by sacrificial anodes for a reduction that generate the counterions for the anions generated at the anode or (ii) by the generation of protons at an anode to neutralize a base formed at the cathode), such analyses ignore the energy associated with the reaction at the counter electrode. This is an issue because the total energy required to drive any electrolysis reaction is defined by the potential difference needed for accomplishing both the oxidation and the reduction half reactions (Figure 79.7). In most electrolysis reactions, the part of this energy associated with the counterreaction is simply thrown away. If one wants to optimize both the atom and energy economy of a reaction, then this cannot be the case. For this reason, the development of paired electrochemical reactions has been undertaken. Overpotential Required Cell Potential

Thermodynamic oxidation potential

Overpotential Figure 79.7 Energy requirements for an electrolysis.

Thermodynamic reduction potential

875

876

79 Electrons, Electrodes, and the Transformation of Organic Molecules −

+

(a)

A

B

[A]*

[B]*

P1

P2

−

+

A [A1]* (b)

P1

[A2]* P2

−

+

(c)

A

B

[A]*

[B]* P

Figure 79.8 Schematic illustration of the three types: (a) parallel paired electrosynthesis, (b) divergent paired electrosynthesis, and (c) covergent paired electrosynthesis.

In a paired electrochemical reaction, both the anodic reaction and the cathodic reaction are used for the generation of high-value products. Such paired electrosyntheses can be quite difficult to realize, and therefore, only few examples have been reported thus far. Some of these examples have already been reviewed [134]. Conceptually, the examples reported in literature can be divided in three categories (a schematic representation is depicted in Figure 79.8). In a parallel paired electrosynthesis, substrates A and B are simultaneously converted to the reactive intermediates [A]* and [B]*. Independently from each other, both intermediates undergo a chemical follow-up reaction under formation of products P1 and P2 (Figure 79.8a). For realization of such processes, a spatial separation of the half reactions using a separator may be necessary. This does not only lead to a simpler separation process but also assures that interactions between starting materials, the intermediates, and the products with the opposite half reaction (including the counter electrode) are avoided. A further possibility for a paired process is represented by the divergent paired electrolysis, where a single substrate A is converted simultaneously on cathode and anode (Figure 79.8b). Via formation of intermediates [A1 ]* and [A2 ]*, two different products, P1 and P2 , are obtained. Again, a separation of the two half reactions may be advantageous. In contrast to the parallel and the divergent process, an undivided cell is absolutely necessary for a convergent paired electrosynthesis (Figure 79.8c). Here, two different substrates, A and B, are separately converted to intermediates [A]* and [B]* (for instance, an electrophile and a nucleophile), which react with each other to form a single product, P. 79.4.2 Examples

A remarkable example for a parallel paired electrolysis is depicted in Scheme 79.50. In this process introduced by BASF SE, p-tert-butyltoluene (147) is anodically converted to the corresponding benzaldehyde dimethyl acetal (148), while dimethyl phthalate (149) is simultaneously reduced to phthalide (150) [134]. It should be noted that the electrolysis is free of side products and that it was optimized to a very simple setup (undivided cell and methanol as solvent). Products 148 and 150 play

79.4 Paired Electrochemical Reactions

tBu + 2 MeOH

COOMe COOMe

147

− 4e− − 4H+

149

+ 4e− + 4H+

tBu 2 MeOH + MeO

O O

OMe 148

150

Scheme 79.50 Anodic methoxylation of p-tert-butyltoluene (147) and cathodic synthesis of phthalide (150) in a parallel paired process.

important roles for the BASF product portfolio: whereas 148 represents an intermediate in the production of pharmaceuticals, crop protection agents, and fragrances (“Lysmeral”), 150 is used for the synthesis of strobilurin fungicides. Despite the elegancy of this paired process, its profitability strongly depends on a balanced demand for 148 and 150. In Scheme 79.51, an example for a divergent paired electrolysis is depicted. Here, the oxidation of glucose to gluconic acid and the reduction of glucose to sorbitol have been paired in an undivided cell [135]. Due to the high demand for both products from both the food and the pharmaceutical industry, intensive studies have been

OH

2Hads + 2OH−

HO HO

O OH

Br2

OH

Glucose

+2e−

+2OH−

2Br−

2H2O

Ni cathode

−2e−

H HO H H

CH2OH OH H OH OH CH2OH

Sorbitol

H HO H H

COOH OH H OH OH CH2OH

Graphite anode

Gluconic acid

Scheme 79.51 Divergent paired electrosynthesis of sorbitol and gluconic acid from glucose.

877

878

79 Electrons, Electrodes, and the Transformation of Organic Molecules

carried out on this system, although this process has finally not been commercialized [136]. The electrolysis has been realized on the upper kilogram scale in an undivided flow cell with packed-bed electrodes using Raney Ni powder as catalytically active cathode material and with graphite chips serving as anode. An intriguing illustration of a convergent paired electrolysis is represented by the generation of formamide-protected homoallyl alcohols 153 in a three-component synthesis (see Scheme 79.52). On the cathodic side, allyl bromide (154) is reductively dehalogenated to the allyl anion (155), which subsequently reacts with benzaldehyde under formation of homoallyl alcoholate 156. The solvent DMF (151) is simultaneously oxidized on the anode to the corresponding acyl iminium ion 152 (see Section 79.3.1.2). Compared to the other compounds in solution, DMF is very hard to oxidize, and therefore, a high anodic current density has to be realized. In this case, the high current density was achieved by using an undersized platinum wire as anode (quasi-divided cell). Finally, product 153 is formed upon nucleophilic attack of 156 on 152. It should be noted that the situation created in this case is quite remarkable: in two parallel umpolung processes, a nucleophile (155) and an electrophile (152) are generated simultaneously in a single reaction vessel. Such a constellation is extremely hard to realize by means of conventional organic synthesis, and typically, such reactive species would have to be prepared separately prior to the reaction. + Br

O N

151

154

+2e− −H+

O

O

O

+N

+2e− −Br−

Ph

Ph 156

152

155

N Pt-wire anode

O

O Ph

153

Glassy carbon cathode

Scheme 79.52 Synthesis of protected homoallyl alcohols in a convergent paired synthesis.

79.4.3 On-Site Reagent Generation

In spite of the obvious benefits of the paired electrochemical approach illustrated above, the reactions and overall process have not captured the imagination of most synthetic chemists. There are a number of reasons for this, but one that stands out is the specialized nature of the reactions examined. While the examples are impressive, most synthetic chemists do not make use of specifically designed reactions. They

79.4 Paired Electrochemical Reactions

make use of generally useful reactions that allow them to construct products they need to make. In other words, the reactions used are not selected because of an interesting technique or an optimized use of energy. They are dictated by the nature of a specific product needed for a specific reason. With this in mind, a new look at paired electrochemical reactions is needed that takes advantage of the generality of the constant current method. As noted earlier, in a constant current reaction, the potential at the working electrode automatically matches that of the substrate that needs to be oxidized or reduced. That situation occurs at both the anode and the cathode. Therefore, in principle, any oxidation reaction can be paired with any reduction reaction. With this in mind, one can readily imagine pairing an oxidation reaction with the generation of a chemical reagent or substrate at a cathode for use elsewhere in a synthetic sequence. For example, in a recent paper on the valorization of lignin-derived materials, the oxidation of an electron-rich benzylic alcohol (157) was paired with the electrochemically driven hydrogenation of an olefin (159) needed for the synthesis of indanones from lignin-derived material (Scheme 79.53) [137]. A similar reaction was conducted that utilized the hydrogen generated at the cathode to remove a Cbz protecting group from an amine. Constant current

O + 2H+

−2e−

158

2MeOH Cathode

Anode

MeO OMe

+2e− OH

MeO

H2 + 2MeO−

157 OMe O

O MeO

MeO

OH

OH

Pd/C MeO OMe

159

1: 1MeOH/EtOAc

MeO OMe

Yield 160: 82% Yield 158: 80%

Scheme 79.53 Generation of reagents at the cathode.

Of course, cathodic reactions are not restricted to the production of hydrogen gas. They can be used to generate a variety of substrates and reagents (see Section 79.2.1.6 for example). The use of a cathode in this manner has the potential to dramatically change the way in which we utilize electrochemical oxidation reactions. In principle, they can be used to not only improve the sustainability of the desired oxidation process but also aid in the sustainability of other processes within a larger synthetic effort.

879

880

79 Electrons, Electrodes, and the Transformation of Organic Molecules

79.5 Conclusions and a Look Toward the Future

Electrochemical reactions are gaining in popularity and that popularity is accelerating their development in exciting ways. This current increase mainly originates from two different aspects: first, electrosynthesis has the potential to impact a wide range of problems, many of which are not typically associated with standard synthetic efforts. A good illustration of this versatility is represented by the microelectrode array technology, which can be used for the real-time study of binding events between small molecules and receptors (an insight is provided in Section 79.5.1). Second, the emerging interest in electrosynthesis is at least in part due to the local abundance of electric energy associated with the increasing use of renewable energy sources (a more detailed discussion if this point can be found in Section 79.5.2). 79.5.1 Addressable Molecular Libraries and the Total Synthesis of Molecular Surfaces

As an increasing number of synthetic chemists explore electrochemical methods, their creativity and insights are leading to new electrochemical techniques and the application of those techniques to new problems that were previously unimaginable. Take, for example, the development of addressable molecular libraries for use in the “real-time” probing of binding events between small molecules and biological receptors [138]. The challenge associated with such an effort involves not only the construction of the small molecules to be studied but also the placement or synthesis of those molecules on the surface of a device in a manner that leaves them both spatially isolated from each other and individually addressable. The end result is the need to carefully construct a complex molecular surface with “spatial selectivity.” What are the tools available for this task? The answer to this question involves a new application of the very reactions described above. The key to the challenge is the ability of microelectrode arrays to function as the analytical device for addressing the molecules to be analyzed. Microelectrode arrays contain thousands spatially isolated electrodes that can each be individually addressed and can each be used to probe the chemistry of molecules attached to their surface. So, if one can place or build molecules on the arrays such that they are located proximal to specific electrodes in a microelectrode array, then the molecules will be spatially isolated from each other and individually addressable. But how does one build or place a molecule by a specific electrode in a microelectrode array when the array has over 12 000 electrodes per square centimeter? Fortunately, the array has electrodes and as we have seen above, electrodes are excellent tools for conducting synthetic reactions. So the same electrode in the array used to monitor a molecule’s behavior can also be used to construct or place the molecule to be studied on its surface. With this in mind, the arrays are coated with a porous polymer that provides attachment points for the molecules [139] and then the electrodes in the arrays used to generate chemical reagents and catalysts that allow for synthetic chemistry to

79.5 Conclusions and a Look Toward the Future

be conducted on that polymer. The chemistry to make those reagents and catalysts is identical to the chemistry employed for the mediated electrochemical reactions described above. It is simply scaled down in size. In fact, the first site-selective array reaction took advantage of the exact chemistry developed by Torii and coworkers for the electrochemical Wacker oxidation [140]. Since electrochemical reactions can be used to make a wide variety of chemical reagents, the approach allows for much of modern synthesis to be used to build the surface of the array. To date, Pd(0), Pd(II), Cu(I), Cu(II), Ce(IV), DDQ, Os(VIII), Ru(VII), triarylamine radical cations, H+ , methoxide, H2 , and Sc(III) have all been used to conduct syntheses on an array. Of course, for the reactions to work on an array for the desired purpose, the reagents generated at the electrodes must be confined to those same electrodes. Otherwise, the reactions would occur everywhere on the array, and the spatial isolation of the molecules on the device would be lost. Fortunately, the reactions can be nicely confined to specific locations on the array by adding a second reagent to the reaction that will destroy the electrochemically generated reagent. One of the nicest examples of this approach is shown in Scheme 79.54 [141]. The reaction involves an electrochemically mediated N-oxyl radical mediator (TEMPO) alcohol oxidation that is very closely related to the mediated oxidations described in Section 79.3.2.2. For this reaction, the substrate for the oxidation was an agarose coating on the surface of the array. The TEMPO mediator was generated by a “T-pattern” of electrodes in the array and then confined to those electrodes with the use of the electron-rich 4-methoxybenzyl alcohol in solution. The 4-methoxybenzyl alcohol was rapidly oxidized by any TEMPO that migrated away from the surface of the electrode, thereby consuming that oxidant. The success of the reaction and its confinement was determined by using the carbonyls generated from the oxidation reaction in a reductive amination reaction with a fluorescent amine (Texas Red hydrazide). The array was then examined with a fluorescence microscope to give rise to the image shown in Scheme 79.54. The electrodes used for the T-pattern can be clearly observed. As a side note, the array is composed of a series of vertical and horizontal wires. By turning on one vertical wire and one horizontal wire, any electrode in the array can be addressed. Hence any pattern desired can be made. This particular example of an oxidation reaction was shown here because it shows just how well confinement of the reaction to the selected electrodes can be controlled. For the reaction shown, the Agarose TEMPO, Bu4NBr 4-methoxybenzyl alcohol OH

0.4 M Na2CO3/0.3 M NaHCO3 2:7:1 DMF/CH3CN/H2O “T” pattern

NaBH3CN Texas Red-Hydrazide O MeOH

Scheme 79.54 An array-based alcohol oxidation.

HN

Texas red

881

882

79 Electrons, Electrodes, and the Transformation of Organic Molecules

agarose coating that serves as the substrate for the oxidation covers the entire surface of the array including the area of the array in between the electrodes. However, no reaction occurs on the array remote from the selected electrodes, even in the region between two electrodes that were both used. The confinement was perfect. A second example of a site-selective oxidation run on the array is shown in Scheme 79.55. In this case, selected electrodes in the array were used to generate Cu(II) from a Cu(I) precursor and in so doing trigger a Chan–Lam-type coupling reaction between a thiol nucleophile and a borate ester functionalized polymer coating the array [142]. Excess thiol in solution was used to reduce the Cu(II) back to Cu(I) before it could migrate to neighboring electrodes in the array leading to a dithiane in solution. The thiol in this experiment was labeled with a pyrene group so that fluorescence microscopy could again be used to determine the success of the reaction. A C-pattern of electrodes (C for cysteine) was used and the reaction was perfectly confined to those electrodes. For this reaction, a series of control experiments did show that the thiol was coupled to the array in preference to the amine nucleophile. The rate of the thiol reaction is simply faster than the rate of the corresponding amine coupling reaction. This example was shown because of the generality of the reaction. Cu(II) has now been used to add thiols, alcohols, amines, acetylenes, and azides to the surface of a borate ester coated array. Of course, the chemistry is not restricted to the use of the array electrodes as anodes. If one reverses the direction of the current used in the experiment, then the array can be used to conduct site-selective reduction reactions. In Scheme 79.56, a reaction is illustrated that is opposite of the reaction shown in Scheme 79.55. In this case, a Cu(II) precursor in solution was reduced to Cu(I) at selected electrodes in the array in order to trigger a coupling reaction between a thiol and an aryl bromide functionalized polymer coating the array [143]. Air was used to oxidize any Cu(I) in solution back to Cu(II), thereby reversing the electrochemical process and preventing the Cu(I) from migrating to neighboring sites on the array. The chemistry placed a cysteine-labeled peptide on the array that was used to then probe the signaling capabilities of the arrays. Once again, note that the chemistry conducted on the microelectrode array is no different than any other CCE that generates a chemical reagent. The only difference is the use of the solution phase confining agent. To this end, it has been shown that reactions developed for the arrays can be scaled to perform preparative reactions. The reaction shown in Scheme 79.57 has been used to both synthesize benzimidazoles on an array [141] and benzimidazole building blocks from lignin-derived platform chemicals on a preparative scale [137]. The point of the microelectrode array work and its inclusion here is to illustrate how the creative work currently being done in synthetic electrochemistry and described in the sections above can be applied to a growing number of challenges. Electrochemical methods have the potential to improve the sustainability of existing chemical reactions, trigger entirely new synthetic transformations that cannot be accomplished with a chemical reagent, and allow us to tackle total synthesis challenges that range from complex molecules to complex molecular surfaces in entirely new ways. With the field expanding and with new investigators bringing

OFF

O

O

B

B O

ON

O

O

B O O B O

Scheme 79.55 Cu(II)-mediated reactions.

N H SH

NH2 NH

S

O H2N

OFF

Pyrene

Cu(OAc)2, DMF, Bu4NPF6 +2.0 V for 20 cycles (30 s on,10 s off) O Pyrene O B O

884

79 Electrons, Electrodes, and the Transformation of Organic Molecules

7:2:1 MeCN/DMF/H2O CuSO4, Bu4NBr, Air − 1.7 V, 2 times 90 s 10 blocks of 12 electrodes Br

O

H N Peptide O

S

O

HO

HO

O

O

CO2H

H2N

NH

H N

H2N O

O H N

O N H

O

NH

O

OH

HN

O N H

H N O

O OH SH

HO

RGDyK(5FAM)C Scheme 79.56 A site-selective Cu(I) reaction.

Agarose

O O

NH2 NH2

Pyrene-CHO CAN/ Bu4NPF6 2:1 DMF/ water

O

+2.4 V, 0.5 s on/ 0.1 s off 300 cycles

MeO

O

N Pyrene N H

O N H “Confining agent”

NO2

Scheme 79.57 A scalable oxidative condensation.

their creativity to the development of electrochemical methods, it appears that we finally have a chance to fully realize all that electrochemistry can contribute the larger fields of organic and biological chemistry. 79.5.2 Flow Chemistry and the Construction of Metabolites

In another intriguing application of available electrosynthetic methods to a problem lying outside of the “normal” set of synthetic challenges, Stadler and Roth

79.5 Conclusions and a Look Toward the Future

have used an electrochemical microflow cell to selectively generate oxidative metabolites (Scheme 79.58) [144]. The reactions allowed for unique selectivities and the formation of between 10 and 100 mg/h of the metabolites. Six cases were studied. In the example shown, the oxidation led to the formation of products having new functional groups added by both the oxidation (marked with a *) and a subsequent conjugate addition (glutathione [GSH] incorporation). In this way, the electrochemistry was used to generate metabolites that could be compared with those generated from the biological decomposition of drug candidates. O Cl

O

− e− 10 F/mol

OH

H N

OH N

0.5 equiv. NaHSO3 1:1 CH3CN/H2O

Cl

O

O GSH =

Cl

HO NH2

N H

SH H N

Cl

O OH

O

GSH, 50 °C 1:1 CH3CN/H2O

O O

Cl

O OH

H N

Cl +

OH*

Cl GS

H N

OH GS

Cl

OH*

Scheme 79.58 Oxidative metabolite formation.

79.5.3 Electrosynthesis and Energy Management

A further aspect of electrosynthesis that may obtain increasing attention in the future is associated with contemporary energy management issues. Since renewable energy is typically not available where it is mostly demanded, its transport over longer distances is problematic and leads to significant losses (not to speak of the necessity of an enormous supply grid). The integration of decentralized storage systems for stabilization of the grid and for the security of the supply is therefore inevitable, and electrochemical battery systems are currently in the focus of these considerations. However, a diversification of buffer technologies in future grids is of fundamental importance, and electrosynthesis in production plants situated in the same areas where the renewable energy is produced could represent an interesting supplement to other energy storage/conversion methods. In the context of energy management, the energy efficiency of the electrolysis will obtain increasing attention, with electrolyte conductivity, cell design (divided/undivided), and reaction overpotential as key parameters for minimization of ohmic losses. Future research will therefore not only be directed toward

885

886

79 Electrons, Electrodes, and the Transformation of Organic Molecules

expanding the scope of synthetic applications but also toward the development of innovative electrolyte concepts, cell designs, electrocatalysts, and recycling strategies. A further important point is to realize processes with a long lifetime of the cell components (low reactor maintenance). In order to serve as an efficient buffer system in a smart grid, the scale of the electrosynthesis (and the demand for the chemical) would have to be appropriate in order to provide a significant stabilization. With regard to the economic efficiency, a production process where electricity consumption represents a major expense would have to be used. Consequently, the production of bulk chemicals appears to be a more promising buffer than the generation of fine chemical or pharmaceuticals. Aside from the synthesis of organic compounds, the electrochemical conversion of abundant small molecules such as CO2 , H2 O, and N2 to fuels could play an important role and is currently in discussion [145, 146]. 79.5.4 Summary

The future organic electrochemistry lies in both an expansion in the scope of synthetic methods that can be employed and an increase in the efficiency and sustainability of those reactions. Electrochemistry has the potential to revolutionize the way existing redox reactions are conducted, as well as to introduce entirely new methods for building and manipulating both complex molecules and complex molecular surfaces. Yet while the range of problems that can be tackled with electrochemistry is both diverse and rapidly expanding, all of the chemistry is governed by a few basic principles (Section 79.1) that once understood can guide the development of new methods in a manner that makes the chemistry easy to interpret and design. With the rise of readily available, and often very simple, equipment, the application of electrochemistry to problems in organic chemistry has never been more accessible.

References 1. For a general discussion of electro-

chemical methods see:Hammerich, O. and Speiser, B. (eds.) (2016). Organic Electrochemistry: Revised and Expanded, 5e. Boca Raton, FL: CRC Press. 2. For reviews of recent efforts see:(a) Sperry, J.B. and Wright, D.L. (2006). Chem. Soc. Rev. 35 (7): 605. (b) Yoshida, J., Kataoka, K., Horcajada, R., and Nagaki, A. (2008). Chem. Rev. 108 (7): 2265. (c) Frontana-Uribe, B.A., Little, R.D., Ibanez, J.G. et al. (2010). Green Chem. 12 (12): 2099. (d) Waldvogel, S.R. and Janza, B. (2014). Angew. Chem. Int. Ed. 53: 7122. (e)

3. 4.

5.

6. 7. 8.

Horn, E.J., Rosen, B.R., and Baran, P.S. (2016). ACS Cent. Sci. 2: 302. Little, R.D. and Moeller, K.D. (2002). Electrochem. Soc. Inter. 11: 36–42. For a discussion written specifically for synthetic chemists see:Moeller, K.D. (2000). Tetrahedron 56: 9527. Fry, A. J. Synthetic Organic Electrochemistry: 2nd Ed; Wiley, New York, NY; 1989. Baizer, M.M. (1964). J. Electrochem. Soc. 111: 215–222. Xu, G. and Moeller, K.D. (2010). Org. Lett. 12: 2590. Redden, A. and Moeller, K.D. (2011). Org. Lett. 13: 1678.

References 9. Wend, R. and Steckhan, E. (1997). Elec10.

11. 12. 13.

14.

15.

16.

17.

18.

19. 20.

21.

22. 23. 24. 25.

trochim. Acta 42: 2027–2039. Yoshida, J.-I., Hashimoto, J., and Kawabata, N. (1982). J. Org. Chem. 47: 3575–3577. Zupan, M. and Dolenc, D. (1991). Tetrahedron 47: 5025–5028. Tajima, T. and Fuchigami, T. (2005). Chem. Eur. J. 11: 6192–6196. Horii, A.M., Fuchigami, T., and Marken, F. (2005). Electrochem. Commun. 7: 35–39. Horcajada, R., Okajima, M., Suga, S., and Yoshida, J.-I. (2005). Chem. Commun.: 1303–1305. (a) Horii, D., Amemiya, F., Fuchigami, T., and Atobe, M. (2008). Chem. Eur. J. 14: 10382–10387. (b) Amemiya, F., Fuse, K., Fuchigami, T., and Atobe, M. (2010). Chem. Commun. 46: 2730–2732. Yoo, S.J., Li, L.-J., Zeng, C.-C., and Little, R.D. (2015). Angew. Chem. Int. Ed. 54: 3744–3747. Li, L., Jiang, Y., Zeng, C.-C. et al. (2015). J. Org. Chem. 80 (21): 11021–11030. Herold, S., Waldvogel, S.R., Little, R.D., and Yoo, S.J. (2016). Electrochim. Acta 196: 735–740. Waldvogel, S.R., Mentizi, S., and Kirste, A. (2012). Top. Curr. Chem. 320: 1–31. For a reviews see reference 1 as well asChaussard, J., Folest, J.C., Nedelec, J.Y. et al. (1990). Synthesis: 369. For selected examples see:(a) Nishiguchi, I., Yamamoto, Y., and Maekawa, H. (2003). Proc. Electrochem. Soc. 73: 2003–2012. (b) Wang, H., Zhang, K., Liu, Y.-Z. et al. (2008). Tetrahedron 64 (2): 314–318. (c) Ohkoshi, M., Michinishi, J.-Y., Hara, S., and Senboku, H. (2010). Tetrahedron 66 (39): 7732–7737. Francke, R. and Little, R.D. (2014). Chem. Soc. Rev. 43: 2492. Steckhan, E. (1986). Angew. Chem. Int. Ed. Engl. 25: 683–701. Steckhan, E. (1987). Top. Curr. Chem. 142: 1–69. Zhang, N.-T., Zeng, C.-C., Lam, C.M. et al. (2013). J. Org. Chem. 78: 2104–2110.

26. Rountree, E.S., McCarthy, B.D.,

27. 28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38. 39. 40.

Eisenhart, T.T., and Dempsey, J.L. (2014). Inorg. Chem. 53: 9983–10002. Savéant, J.-M. (2008). Chem. Rev. 108: 2348–2378. Savéant, J.-M. (2006). Elements of Molecular and Biomolecular Electrochemistry: An Electrochemical Approach to Electron Transfer Chemistry. Hoboken, NJ: Wiley-Interscience. Bowers, K.W., Giese, R.W., Grimshaw, J. et al. (1970). J. Am. Chem. Soc. 92: 2783–2799. For a review of reductive coupling, see:Utley, J.H.P., Little, R.D., and Nielsen, M.F. (2016). Reductive coupling, Chapter 17. In: Organic Electrochemistry, 5e Ed: Revised and Expanded, (eds. O. Hammerich and B. Speiser), 621–704. CRC Press. Read de Alaniz, J., Kerr, M.S., Moore, J.L., and Rovis, T. (2008). J. Org. Chem. 73: 2033. (a) Petrovich, J.P., Baizer, M.M., and Ort, M.R.J. (1969). Electrochem. Soc. 116: 749–756. (b) Baizer, M.M. and Anderson, J.D. (1965). J. Org. Chem. 30: 3138–3141. Kise, N., Mashiba, S.-I., and Ueda, N. (1998). J. Org. Chem. 63 (22): 7931–7938. Kojima, T., Obata, R., Saito, T. et al. (2015). Beilstein J. Org. Chem. 11: 200–203. Mikesell, P., Schwaebe, M., DiMare, M., and Little, R.D. (1999). Acta Chem. Scand. 53: 792–799. Utley, J.H.P., Nielsen, M.F., and Wyatt, P.B. (2016). Electrogenerated bases and nucleophiles, Chapter 43. In: Organic Electrochemistry, 5e EdRevised and Expanded (eds. O. Hammerich and B. Speiser), 1303–1434. CRC Press. Little, R.D. and Schwaebe, M.K. (1997). Reductive cyclizations at the cathode. Top. Curr. Chem. 185: 1–48. Handy, S.T. and Omune, D. (2007). Tetrahedron 63 (6): 1366–1371. Moëns, L., Baizer, M.M., and Little, R.D. (1986). J. Org. Chem. 51: 4497. (a) Frontana-Uribe, B.A. and Little, R.D. (2005). Electrochim. Acta 50: 1383–1390. (b) Andreu, R. and

887

888

79 Electrons, Electrodes, and the Transformation of Organic Molecules

41. 42. 43. 44.

45. 46. 47. 48.

49.

50.

51. 52.

53. 54. 55.

56.

57.

58.

Pletcher, D. (2003). Electrochim. Acta 48: 1065–1071. Bode, H.E., Sowell, C.G., and Little, R.D. (1990). Tetrahedron Lett. 31: 2525. Little, R.D., Fox, D.P., Van Hijfte, L. et al. (1988). J. Org. Chem. 53: 2287. Yee, R., Mallory, J., Parrish, J.D. et al. (2006). Electroanal. Chem. 593: 69–73. (a) Andrieux, C.P. and Savéant, J.M. (1974). Electroanal. Chem. 53: 165. (b) Parker, V.D. (1986). Topics in Organic Electrochemistry, Chapter 2 (eds. A.J. Fry and W.E. Britton), 35–79. New York, NY: Plenum. Tabakovic, I. (1997). Top. Curr. Chem. 185: 87–139. Fry, A.J., Little, R.D., and Leonetti, J.J. (1994). Org. Chem. 59: 5017–5026. Crimmins, M.T. and Nantermet, P.G. (1993). Org. Prep. Proced. Int. 25: 41. Sowell, C.G. (1990). Application of the electroreductive cyclization to organic synthesis: quadrone, prostaglandins, and phorbol. PhD thesis. University of California, Santa Barbara. Shono, T., Nishiguchi, I., and Ohmizu, H. (1997). J. Am. Chem. Soc. 99: 7396–7397. Shono, T., Ohmizu, H., Kawakami, S., and Sugiyama (1980). Tetrahedron Lett. 21: 5029–5032. Sowell, C.G., Wolin, R.L., and Little, R.D. (1990). Tetrahedron Lett. 31: 485. Fuchigami, T., Suzuki, K., and Nonaka, T. (1990). Electrochim. Acta 35 (1): 239–244. Oh, Y. and Hu, X. (2013). Chem. Soc. Rev. 42: 2253. Windle, C.D. and Perutz, R.N. (2012). Coord. Chem. Rev. 256: 2562–2570. Wagoner, E.R., Baumberger, C.P., Gerroll, B.H., and Peters, D.G. (2014). Electrochim. Acta 132: 545–550. Miranda, J.A., Wade, C.J., and Little, R.D. (2005). J. Org. Chem. 70 (20): 8017–8026. Du, P., Mubarak, M.S., Karty, J.A., and Peters, D.G. (2007). J. Electrochem. Soc. 154: F231. Foley, M.P., Du, P., Griffith, K.J. et al. (2010). J. Electroanal. Chem. 647: 194–203.

59. Fuchigami, T., Kasuga, M., and Konno,

60.

61. 62. 63.

64. 65.

66.

67. 68.

69.

70.

A. (1996). J. Electroanal. Chem. 411: 115–119. Hosoi, K., Inagi, S., Kubo, T., and Fuchigami, T. (2011). Chem. Commun. 47: 8632. Francke, R. and Little, R.D. (2014). Chem. Soc. Rev. 43: 2492–2521. Badalyan, A. and Stahl, S.S. (2016). Nature 535: 406–410. Yoo, S.J., Li, L.-J., Zeng, C.-C., and Little, R.D. (2015). Angew. Chem. Int. Ed. 127: 3815–3818. Kroto, H.W., Heath, J.R., O’Brien, S.C. et al. (1985). Nature 318: 162–163. Echegoyen, L. and Echegoyen, L.E. (2001). The electrochemistry of C60 and related compounds. In: Organic Electrochemistry, 4e, Revised and Expanded: (eds. H. Lund and O. Hammerich), 323–339. New York, NY: M. Dekker. (a) Nguyen, B.H., Perkins, R.J., Smith, J.A., and Moeller, K.D. (2015). Beilstein J. Org. Chem. 11: 280–287. (b) Smith, J.A., Feng, R., Janetka, J., and Moeller, K.D. (2016). Electroanalysis https://doi .org/10.1002/elan.201600200. Tang, F., Chen, C., and Moeller, K.D. (2007). Synthesis: 3411. For reviews of radical cation initiated cyclizations derived from electrochemical oxidations see:Moeller, K.D. (1997). Top. Curr. Chem. 185: 49. For an earlier review of the specific reaction shown in Scheme 25 see:Moeller, K.D. (2009). Synlett (Invited Account) 8: 1208. For total synthesis efforts using the reaction shown in Scheme 25 see: (a) Liu, B., Duan, S., Sutterer, A.C., and Moeller, K.D. (2002). J. Am. Chem. Soc. 124: 10101. (b) Duan, S. and Moeller, K.D. (2001). Org. Lett. 3: 2685. (c) Mihelcic, J. and Moeller, K.D. (2004). J. Am. Chem. Soc. 126: 9106. (d) Wright, D.L., Whitehead, C.R., Sessions, E.H. et al. (1999). Org. Lett. 1: 1535. (e) Hughes, C.C., Miller, A.K., and Trauner, D. (2005). Org. Lett. 7: 3425. (f ) Miller, A.K., Hughes, C.C., Kennedy-Smith, J.J. et al. (2006). J. Am. Chem. Soc. 128: 17057. (g) Wu, H. and Moeller, K.D. (2007). Org. Lett. 9: 4599.

References

71.

72.

73.

74. 75.

76. 77. 78. 79.

80. 81.

82. 83.

84.

85.

86.

(h) Xu, H.-C., Brandt, J.D., and Moeller, K.D. (2008). Tetrahedron Lett. 49: 3868. (a) Tang, F. and Moeller, K.D. (2009). Tetrahedron 65: 10863. (b) Tang, F. and Moeller, K.D. (2007). J. Am. Chem. Soc. 129: 12414. (c) Huang, Y. and Moeller, K.D. (2006). Tetrahedron 62: 6536. Campbell, J.M., Xu, H.-C., and Moeller, K.D. (2012). J. Am. Chem. Soc. 134: 18338. (a) Xu, H.-C. and Moeller, K.D. (2008). J. Am. Chem. Soc. 130: 13542. (b) Xu, H.-C. and Moeller, K.D. (2010). J. Am. Chem. Soc. 132: 2839. Smith, J.A. and Moeller, K.D. (2013). Org. Lett. 15: 5818. Perkins, R.J., Xu, H.-C., Campbell, J.M., and Moeller, K.D. (2013). Beilstein J. Org. Chem. (9): 1630. Brandt, J.D. and Moeller, K.D. (2005). Org. Lett. 7: 3553–3556. Ashikari, Y., Nokami, T., and Yoshida, J. (2011). J. Am. Chem. Soc. 133: 11840. Hayashi, R., Shimizu, A., and Yoshida, J. (2016). J. Am. Chem. Soc. 138: 8400. (a) Okada, Y., Akaba, R., and Chiba, K. (2009). Org. Lett. 11: 1033. (b) For a review see:Chiba, K. and Okada, Y. (2012). Recent Trends Electrochem. Sci. Technol.: 91–106. Goetz, A.E. and Boydston, A.J. (2015). J. Am. Chem. Soc. 137: 7572–7575. Morofuji, T., Shimizu, A., and Yoshida, J. (2015). J. Am. Chem. Soc. 137: 9816–9819. Herold, S., Möhle, S., Zirbes, M. et al. (2016). Eur. J. Org. Chem.: 1274–1278. Lips, S., Wiebe, A., Elsler, B. et al. (2016). Angew. Chem. Int. Ed. 55 https://doi.org/10.1002/anie.201605865. For a review see:Ogawa, K.A. and Boydston, A. (2015). J. Chem. Lett. 44: 10–16. Green, R.A., Pletcher, D., Leach, S.G., and Brown, R.C.D. (2016). Org. Lett. 18: 1198–1201. (a) See reference 1, Chapter 28 and reference 4 as well as (b) Jones, A.M. and Banks, C.E. (2014). Beilstein J. Org. Chem. 10: 3056–3072. (c) Moeller, K.D. (2004). The electrochemistry of nitrogen containing compounds. In: Encyclopedia of Electrochemistry,

87.

88.

89.

90. 91.

92. 93.

94.

95.

vol. 8 (ed. H.J. Schäfer), 277–312. Wiley/Verlag Chemie. (a) Ross, S.D., Finkelstein, M., and Peterson, C. (1964). J. Am. Chem. Soc. 86: 4139. (b) Ross, S.D., Finkelstein, M., and Peterson, C. (1966). J. Org. Chem. 31: 128. (c) Ross, S.D., Finkelstein, M., and Peterson, C. (1966). J. Am. Chem. Soc. 88: 4657. (a) Shono, T., Matsumura, Y., and Tsubata, K. (1985). Org. Synth. 63: 206. (b) Shono, T. (1988). Top. Curr. Chem. 148: 131. A detailed experimental procedure for the oxidation of amino acids can be found inFobian, Y.M. and Moeller, K.D. (1999). Methods Mol. Med. 23 (Peptidomimetic Protocols): 259–279. Wong, P.L. and Moeller, K.D. (1993). J. Am. Chem. Soc. 115: 11434–11445. (a) Yudin, A.K. and Siu, T. (2001). Curr. Opin. Chem. Bio. 5: 269–272. (b) Tsiu, T., Li, W., and Yudin, A.Y. (2000). J. Comb. Chem. 2: 545–549. Miller, L.L., Stermitz, F.R., and Flack, J.R. (1973). J. Am. Chem. Soc. 95: 2651. (a) Yoshida, J. and Isoe, S. (1987). Tetrahedron Lett. 28: 6621. (b) For a review see:Yoshida, J. (1994). Top. Curr. Chem. 170: 39. (c) Yoshida, J., Wtanabe, M., Toshioka, H. et al. (1998). Novel Trends in Electroorganic Synthesis (ed. S. Torii), 99. Tokyo: Springer. For selected applications see: (a) Kamada, T. and Oku, A. (1998). J. Chem. Soc., Perkin Trans. 1: 3381. (b) Suga, S., Watanabe, M., and Yoshida, J. (2002). J. Am. Chem. Soc. 124: 14824. (c) Yoshida, J., Suga, S., Fuke, K., and Watanabe, M. (1999). Chem. Lett.: 251. (d) Suzuki, S., Matsumoto, K., Kawamura, K. et al. (2004). Org. Lett. 6: 3755. (e) Cao, Y., Hidaka, A., Tajima, T., and Fuchigami, T. (2005). J. Org. Chem. 70: 9614. (f ) Suga, S., Watanabe, M., Song, C.-H., and Yosida, J. (2006). Electrochemistry 74: 672. (g) Kesselring, D., Maurer, K., and Moeller, K.D. (2008). Org. Lett. 10: 2501–2504. (h) Nokami, T., Ohata, K., Inoue, M. et al. (2008). J. Am. Chem. Soc. 130: 10864–10865. Sun, H., Martin, C., Kesselring, D. et al. (2006). J. Am. Chem. Soc. 128: 13761.

889

890

79 Electrons, Electrodes, and the Transformation of Organic Molecules 96. (a) Shoji, T., Kim, S., Yamamoto, K.

97.

98.

99. 100.

101.

102. 103.

104.

105.

106.

107.

et al. (2014). Org. Lett. 16: 6404. (b) Shoji, T., Haraya, S., Shokaku, K., and Kazuhiro, C. (2016). Electrochim. Acta 200: 290. (a) Nokami, T., Isoda, Y., Sasaki, N. et al. (2015). Org. Lett. 17: 1525. (b) Nokami, T., Hayashi, R., Saigusa, Y. et al. (2013). Org. Lett. 15: 4520. (c) Nokami, T., Nozaki, Y., Saigusa, Y. et al. (2011). Org. Lett. 13: 1544. (d) Nokami, T., Saito, K., and Yoshida, J. (2012). Carbohydr. Res. 363: 1. For recent examples see:(a)Kuribayashi, S., Shida, N., Inagi, S., and Fuchigami, T. (2016). Tetrahedron 72: 5343. (b) Yin, B., Inagi, S., and Toshio, F. (2015). Beilstein J. Org. Chem. 11: 85. Horii, D., Fuchigami, T., and Atobe, M. (2007). J. Am. Chem. Soc. 129: 11692. Kashiwagi, T., Amemiya, F., Fuchigami, T., and Atobe, M. (2012). J. Flow Chem. 3: 17. (a) For an early review and comparison to the cation flow method see:Yoshida, J. and Suga, S. (2002). Chem. Eur. J. 8: 2650. (b) For a recent reviews see: Suga, S. (2010). Electrochemistry 78: 202, as well as reference 1, Chapter 9. Yoshida, J., Suga, S., Suzuki, S. et al. (1999). J. Am. Chem. Soc. 121: 9546. For an extension to oxonium ions see: Suga, S., Suzuki, S., Yamamoto, A., and Yoshida, J. (2000). J. Am. Chem. Soc. 122: 10244. For a lead reference please see:Suga, S., Okajima, M., Fujiwara, K., and Yoshida, J. (2001). J. Am. Chem. Soc. 123: 7941. (a) Yoshida, J., Nagaki, A., and Yamada, T. (2008). Chem. Eur. J. 14: 7450. (b) Yoshida, J., Kim, H., and Nagaki, A. (2011). ChemSusChem 4: 331. For a review see:Matsumoto, K., Suga, S., and Yoshida, J. (2586). Org. Biomol. Chem. 2011: 9. For the cyclization shown see: (a) Nad, S. and Breinbauer, R. (2004). Angew. Chem. Int. Ed. 43: 2297. For related reactions see (b) Suga, S., Matsumoto, K., Ueoka, K., and Yoshida, J. (2006). J. Am. Chem. Soc. 128: 7710. (c) Matsumoto, K., Ueoka, K., Suzuki, S. et al. (2009). Tetrahedron 65: 10901.

108. (a) Fujie, S., Matsumoto, K., Suga, S.

109.

110. 111.

112. 113. 114.

115.

116. 117. 118. 119.

120. 121.

122. 123. 124.

125.

126.

et al. (2010). Tetrahedron 66: 2823. (b) For intramolecular cyclizations:Suga, S., Nishida, T., Yamada, D. et al. (2004). J. Am. Chem. Soc. 126: 14338. (c) Matsumoto, K., Fujie, S., Ueoka, K. et al. (2008). Angew. Chem. Int. Ed. 47: 2506. Morofuji, T., Shimizu, A., and Yoshida, J. (2012). Angew. Chem. Int. Ed. 51: 7259. Platen, M. and Steckhan, E. (1984). Chem. Ber. 117: 1679–1694. Dapperheld, S., Steckhan, E., Brinkhaus, K.-H.G., and Esch, T. (1991). Chem. Ber. 124: 2557–2567. Schmidt, W. and Steckhan, E. (1980). Chem. Ber. 113: 577–585. Fuchigami, T., Tetsu, M., Tajima, T., and Ishii, H. (2001). Synlett: 1269–1271. Fuchigami, T., Mitomo, K., Ishii, H., and Konno, A. (2001). J. Electroanal. Chem. 507: 30–33. Shen, Y., Hattori, H., Ding, K. et al. (2006). Electrochim. Acta 51: 2819–2824. Wu, X., Davis, A.P., and Fry, A.J. (2007). Org. Lett. 9: 5633–5636. Wu, X., Davis, A.P., Lambert, P.C. et al. (2009). J. Tetrahedron 65: 2408–2414. Park, Y.S. and Little, R.D. (2008). J. Org. Chem. 73: 6807–6815. Zeng, C.-C., Zhang, N.-T., Lam, C.M., and Little, R.D. (2012). Org. Lett. 14: 1314–1317. Weinberg, N.L. and Weinberg, H.R. (1968). Chem. Rev. 68: 449–523. Semmelhack, M.F., Chou, C.S., and Cortes, D.A. (1983). J. Am. Chem. Soc. 105: 4492–4494. Schnatbaum, K. and Schäfer, H.J. (1999). Synthesis: 864–872. Inokuchi, T., Matsumoto, S., and Torii, S.J. (1991). Org. Chem. 56: 2416–2421. Rafiee, M., Miles, K.C., and Stahl, S.S. (2015). J. Am. Chem. Soc. 137: 14751–14757. Kuroboshi, M., Yoshihisa, H., Cortona, M.N. et al. (2000). Tetrahedron Lett. 41: 8131–8135. Kashiwagi, Y., Kurashima, F., Kikuchi, C. et al. (1999). Tetrahedron Lett. 40: 6469–6472.

References 127. Horn, E.J., Rosen, B.R., Chen, Y. et al. 128. 129. 130.

131.

132. 133. 134.

135.

136.

(2016). Nature 533: 77–81. Recupero, F. and Punta, C. (2007). Chem. Rev. 107: 3800–3842. For TEMPO see:Xu, F., Zhu, L., Zhu, S. et al. (2014). Chem. Eur. J. 20: 12740. For iron as a mediator see:(a)Hou, Z.-W., Mao, Z.-Y., Zhao, H.-B. et al. (2016). Angew. Chem. Int. Ed. 55: 1–6. (b) Zhu, L., Xiong, P., Mao, Z.-Y. et al. (2016). Angew. Chem. Int. Ed. 55: 2226–2229. For a related synthetic transformation using a hypervalent iodine mediator see: Amano, Y. and Nishiyama, S. (2006). Tetrahedron Lett. 47: 6505–6507. Xu, H.-C., Campbell, J.M., and Moeller, K.D. (2014). J. Org. Chem. 79: 379–391. Torri, S., Inokuchi, T., and Sugiura, T. (1986). J. Org. Chem. 51: 155. (a) Paddon, C.A., Atobe, M., Fuchigami, T. et al. (2006). J. Appl. Electrochem. 36: 617–634. (b) Steckhan, E., Arns, T., Heineman, W.R. et al. (2001). Chemosphere 43: 63–73. Park, K., Pintauro, P.N., Baizer, M.M., and Nobe, K. (1985). J. Electrochem. Soc. 132: 1850. Pütter, H. (2001). Industrial electroorganic chemistry. In: Organic

137.

138.

139. 140.

141.

142. 143.

144. 145. 146.

Electrochemistry, 4e , Revised and expanded: (eds. H. Lund and O. Hammerich), 1259–1308. New York, NY: M. Dekker. Nguyen, B.H., Perkins, R.J., Smith, J.A., and Moeller, K.D. (2015). J. Org. Chem. 80: 11953–11962. For a review see:Graaf, M.D. and Moeller, K.D. (2015). Langmuir 31: 7697–7706. Hu, L., Graaf, M.D., and Moeller, K.D. (2013). J. Electrochem. Soc. 160: G3020. Tesfu, E., Maurer, K., Ragsdale, S.R., and Moeller, K.D. (2004). J. Am. Chem. Soc. 126: 6212–6213. Nguyen, B.H., Kesselring, D., Tesfu, E., and Moeller, K.D. (2014). Langmuir 30: 2280. Graaf, M.D. and Moeller, K.D. (2016). J. Org. Chem. 81: 1527–1534. Stuart Fellet, M., Bartles, J.L., Bi, B., and Moeller, K.D. (2012). J. Am. Chem. Soc. 134: 16891. Stalder, R. and Roth, G.P. (2013). ACS Med. Chem. Lett. 4: 1119–1123. DuBois, D.L. (2014). Inorg. Chem. 53: 3935–3960. Benson, E.E., Kubiak, C.P., Sathrum, A.J., and Smieja, J.M. (2009). Chem. Soc. Rev. 38: 89–99.

891

893

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications Alla Synytska, Astrid Drechsler, Ralf Frenzel, Jens Harenburg, and Cornelia Bellmann

Self-cleaning surfaces are of great interest for a broad variety of applications, e.g. facades, windows, solar cell panels, food containers, kitchen equipment, or oil tubings. Soil-repellent behavior is well known from many natural surfaces. Nonetheless it is a great challenge to transfer the principle to technically relevant materials or coatings in order to create self-cleaning surfaces with a good durability and mechanical stability. This chapter presents three different techniques for the creation of technically relevant superhydrophobic or even superomniphobic surfaces. Based on a fundamental understanding on the mechanisms of self-cleaning, it furthermore elucidates the correlations between chemical composition, morphology, and roughness of the surfaces and their wettability and self-cleaning behavior. (See also Chapter 54 in Volume 7.)

80.1 Fundamental Aspects

The self-cleaning properties of plant leaves, e.g. the lotus leaf, have been known for more than 2000 years. In the 1970s, the effect was studied first by Barthlott and Ehler [1] by investigating the surface of water-repellent plants and animal skins with scanning electron microscopy (SEM). In papers published in 1997 [2, 3], they showed that the self-cleaning mechanism of many biological surfaces results from an interplay of surface roughness, reduced particle adhesion, and water repellency. Water droplets roll off of highly structured hydrophobic surfaces taking all soil particles with them. This mechanism is called Lotus effect or superhydrophobicity. Although the first patent application on superhydrophobic surfaces made of hydrophobic fumed silica was filed already in 1974 [4], the principle was transferred successfully to self-cleaning technical surfaces only in the 1990s. In 1999 the first commercial superhydrophobic varnish Lotusan was introduced to the market by Sto SE & Co. KGaA. The correlations between roughness and wetting were studied, however, much earlier [5–9]. The chemistry of a surface determines its surface free energy and

®

®

Surface and Interface Science: Applications of Surface Science II, First Edition. Edited by Klaus Wandelt. © 2020 Wiley-VCH Verlag GmbH & Co. KGaA. Published 2020 by Wiley-VCH Verlag GmbH & Co. KGaA.

894

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

thus its intrinsic contact angle 𝜃 in , which is the thermodynamic contact angle of a perfectly smooth, homogeneous surface. It can be calculated from the solid/vapor, solid/liquid, and liquid/vapor interfacial tensions 𝛾 sv , 𝛾 sl , and 𝛾 lv using the Young equation [10]: 𝛾 − 𝛾sl (80.1) cos 𝜃in = sv 𝛾lv Most solid surfaces are neither perfectly smooth nor homogeneous. The roughness of the surface affects the propagation of a liquid. On microscopic level, the liquid drop tries to adopt the intrinsic contact angle on each point of the solid/liquid/vapor contact line. A rough surface has, however, different tilt angles at each point. This results in an apparent macroscopic contact angle different from the intrinsic contact angle. Wenzel [9] introduced a roughness factor (or Wenzel roughness) rW that represents the ratio between the true surface area and the area projected on the xy plane. According to Wenzel, the apparent contact angle 𝜃 app can be calculated from the intrinsic contact angle 𝜃 in by cos 𝜃app = rW cos 𝜃in

(80.2)

This equation can be applied under the condition that the surface under the drop is wetted completely (the homogeneous or “Wenzel regime,” cf. Figure 80.1a) and has a uniform roughness on an order of magnitude that is much smaller than the dimensions of the liquid drop. Cassie and Baxter [5] revealed in 1945 that liquids with an intrinsic contact angle 𝜃 in > 90∘ can adopt large apparent contact angles on rough and porous surfaces if air is entrapped between the drop and the solid surface (cf. Figure 80.1a). In this so-called heterogeneous or “Cassie–Baxter regime,” the apparent contact angle is determined by the Cassie–Baxter equation: cos 𝜃app = f1 cos 𝜃in − f2

(80.3)

where f 1 is the wetted solid area fraction and f 2 is the area fraction of the interface to entrapped air, with f 1 + f 2 = 1. This equation is valid if the true wetted area is considered – a value that is often not easily accessible [12]. Usually, liquids on rough surfaces adopt different contact angles, while the liquid front advances or recedes, the advancing contact angle 𝜃 a and the receding contact angle 𝜃 r . Johnson and Dettre were the first to study this roughness-induced contact angle hysteresis both experimentally [6] and theoretically [7, 8]. They stated that the wetting of rough surfaces is often determined by metastable states similar to the Cassie–Baxter regime. Provided that the aspect ratio of the surface asperities is sufficiently high, the slope of the side walls prevents the liquid from entering the voids. Therefore the entrapment of air under the droplet may be energetically favorable. They thus gave the first explanation of the effect of superhydrophobicity. David and Neumann [13, 14] performed numerical calculations to study the wetting of randomly structured surfaces. While the theories by Wenzel and Cassie predict one global minimum of the interfacial free energy leading to the contact angles given in Eqs. (80.2) and (80.3), David’s calculations for real structures yielded

Contact angle

cos θrough

1 Wenzel Impregnation

θr (b)

0

θa

0

cos θin

Contact angle

–1 –1

Wicking

Increasing roughness

Cassie-Baxter

(a)

θa

1

(c)

Composite surface θr

Increasing roughness

Figure 80.1 (a) Wetting diagram for a fluid on a geometrically rough surface. The solid lines correspond to different wetting states according to the macroscopic theories of Wenzel and Cassie (Eqs. (80.2) and (80.3)). Dashed portions of the curves represent metastable states. (b, c) Qualitative prediction of the wetting of rough surfaces (b) for liquids with intrinsic contact angle 𝜃 in < 90∘ and (c) for liquids with intrinsic contact angle 𝜃 in > 90∘ as function of the roughness of the wetted surface. (Source: (b, c) Garbassi et al. 1998 [11]. Reprinted with permission of Wiley.)

896

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

“a free energy landscape with local minima.” Advancing and receding contact angles are the highest and lowest possible contact angles corresponding to the outer minima in this “landscape.” The authors found that the wetting behavior of such irregular surfaces is basically determined by the Wenzel roughness rW . In the literature, there are different ways to display the relations between roughness and wetting behavior schematically. Two common examples are shown in Figure 80.1. The wetting diagram in Figure 80.1a [15, 16] gives the cosine of the contact angle on a rough surface 𝜃 rough related to the cosine of the contact angle on a flat surface, i.e. the intrinsic contact angle 𝜃 in . The three solid lines show the behavior predicted by the theories of Wenzel and Cassie–Baxter (Eqs. (80.2) and (80.3)) and a state of complete impregnation of the rough surface; dashed lines are the corresponding metastable wetting conditions [17]. Different morphologies and roughness values result in varying slopes of the lines. This schematic presentation is frequently used to compare experimental results with theoretical predictions. It does, however, not consider the contact angle hysteresis. Garbassi et al. [11] give a qualitative survey of the relations between roughness and contact angle for two qualitatively different cases of intrinsic contact angles (Figure 80.1b,c). Figure 80.1b predicts advancing and receding contact angles on a rough surface as functions of the roughness for liquid/solid pairings with an intrinsic contact angle 𝜃 in < 90∘ . If the roughness increases, the advancing contact angle first rises, while the receding contact angle approaches zero. The contact angle hysteresis Δ𝜃 = 𝜃 a − 𝜃 r is high. At high roughness the liquid spreads on the surface due to capillary effects (wicking). The liquid drop is always in the Wenzel regime; the apparent contact angle can be calculated according to Eq. (80.2), which, however, does not consider the contact angle hysteresis. If the intrinsic contact angle exceeds 90∘ , the influence of the roughness is different (Figure 80.1c). At low roughness a similar effect is observed as for 𝜃 in < 90∘ . For higher roughness, 𝜃 a and 𝜃 r do, however, not go to zero. Beyond a certain threshold of the roughness, 𝜃 r increases and approaches 𝜃 a ; the contact angle hysteresis vanishes. The liquid drop is in the “Cassie–Baxter regime”; the apparent contact angle (which in the case of negligible hysteresis represents both advancing and receding contact angles) is determined by Eq. (80.3). This effect is used for the preparation of superhydrophobic surfaces. Very high apparent contact angles can be reached if f 1 ≪ f 2 . This is the case with hierarchically structured or fractal surfaces having roughness on various orders of magnitude when the liquid sits on the very top of small surface asperities. For a theoretical description of the wetting of hierarchical surfaces, all roughness levels larger than the hydrodynamic diameter of the liquid molecules [18] and smaller than the size of the liquid drop [19] have to be considered. This may result in more complicated, combined wetting regimes as Wenzel-in-Wenzel, Cassie-in-Wenzel, or Wenzel-inCassie [20]. A theory of Onda et al. [15] approximates the apparent contact angle of fractal surfaces at the transition between Wenzel and Cassie–Baxter regime by means of

80.2 Materials for Technical Applications

their fractal dimension D and the upper and lower limits of fractality, L und l: cos 𝜃app =

( )D−2 L cos 𝜃in l

(80.4)

This theory was confirmed experimentally by fractal superhydrophobic surfaces made of alkyl ketene dimer with hysteresis-free water contact angle of up to 174∘ [16]. A necessary condition for superhydrophobicity is, besides the hierarchical or fractal structure, a high intrinsic contact angle of different liquids, i.e. a high surface energy. Typical low energy surfaces consist of fluorinated organic or organic–inorganic compounds. Examples that are very interesting for applications by sol–gel technique are the so-called F-ORMOSILS or POSS [21, 22], polyhedral oligomeric silsesquioxanes with fluorinated side chains. With these compounds, intrinsic water contact angles up to 120∘ can be reached; the intrinsic contact angles of unpolar liquids remain, however, below 90∘ [22]. Nonetheless, repellency of all liquids, i.e. superoleophobic or superomniphobic behavior, can be obtained if the surfaces exhibit structures that allow entrapment of air by preventing the liquid from entering the voids [18, 22, 23]. In nature, superomniphobicity is known from small insects living in flower pots, the springtails. The skin of these insects is hierarchically structured – bumps covered with smaller T- or mushroom-shaped structures with undercuts [24, 25]. Such so-called reentrant structures with negative slopes of the structure sidewalls act as energy barriers also for liquids with intrinsic contact angle 𝜃 in < 90∘ and create a Cassie–Baxter-like wetting regime with entrapped air. This effect has been used by several groups to prepare superomniphobic surfaces with and even without fluorine chemistry on lab scale [23]. Werner and coworkers successfully transferred the springtail structure on membranes produced by reverse imprint lithography [26].

80.2 Materials for Technical Applications 80.2.1 Superhydrophobic Surfaces Based on Uniform and Janus Core–Shell Particles

Precondition for the technological applicability of superhydrophobic surfaces is the preparation of a reproducible rough morphology and a defined surface chemical composition with a low surface energy and high intrinsic contact angle on a large scale. A versatile and experimentally simple method is the self-organization of core–shell particles. Core–shell particles are micro- or nanosized particles consisting of a core of one material covered by a shell of another material. Both core and shell can be made of a variety of materials, soft or hard, inorganic and organic substances; the shell can be chemically grafted or physically adsorbed. Thus properties of the core and the shell are combined, which makes core–shell

897

898

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

particles interesting for manifold applications. Particles used for the preparation of superhydrophobic surfaces are modified by a hydrophobic shell. Very low surface energies and high intrinsic contact angles are reached with fluoro-organic compounds with high fluorine content where the fluorine moieties are enriched at the layer surface [27, 28]. A special variety of core–shell particles are Janus particles. Named after the twofaced god Janus, their shell consists of two (or more) parts with different properties, e.g. hydrophilic/hydrophobic, hard/soft, charged/uncharged, etc. [29]. The term “Janus beads” was established by Casagrande et al. for spherical glass particles with one hydrophilic and one hydrophobic hemisphere and their behavior at oil/water interfaces [30]. Currently much work is done to create Janus particles for a broad variety of applications, e.g. the tailored design of surfaces and interfaces with special wettability, optical or electronical features, and catalytic or biologic activity [29]. Core–shell particles deposited on a solid surface form ordered structures with a well-defined morphology and roughness [31, 32]. Hierarchical structures are obtained by unordered particle deposition [33], by microparticles covered by nanoparticles (“raspberry particles”) [34], by mixing particles of different dimensions [35], or by self-organization of Janus particles [36]. Particle assemblies can also be used as templates for the creation of inverse structures with sharp edges [37] or reentrant geometry [38]. Using particles of different sizes with the same shell allows studying the influence of the roughness and morphology while the chemistry is unchanged. 80.2.1.1 Preparation of Uniform Core–Shell Particles

In the present study, hybrid core–shell particles with a hard inorganic core and a grafted polymer shell were applied. Silica spheres of different radius (100 nm–10 μm) were used as cores. Their surface chemical composition was modified by immobilization of low-molecular-weight compounds or polymers forming the particle shell. Three different hydrophobizing agents were used: (tridecafluoro-1,1,2,2-tetrahydrooctyl)dimethylchlorosilane (FSI), polystyrene (PS), and a styrene-2,3,4,5,6-pentafluorostyrene copolymer (FPS). FSI was attached directly to the particle surface by reaction with the silica hydroxyl groups. For grafting PS and FPS, first an anchor layer of 3-glycidoxypropyl trimethoxysilane (GPS) was bound covalently to the particle surface. PS and FPS molecules endfunctionalized with carboxylic acid groups were then “grafted to” the epoxy groups of the GPS layer. This condensation leading to the formation of ester groups was verified by attenuated total reflection-Fourier transform infrared spectroscopy (ATR-FTIR) spectra. The surface concentration of the grafted polymers was in the order of 4–5 mg/m2 , referring to a film thickness of 4–5 nm [31]. The effect of the modification was studied by wettability measurements. The particle surface tension 𝛾 sv was determined by measuring the capillary penetration of a packed column of the particles by liquids with different surface tensions, and analysis according to the modified Washburn equation [39]. While the surface tension of unmodified silica particles was about 39 mN/m, the surface tension of the PS-grafted PS particles (𝛾 sv ≈ 28 mN/m) coincided well with the surface tension

80.2 Materials for Technical Applications

of plain PS particles [31]. That proves that the surface properties of the core–shell particles are determined by the shell material. 80.2.1.2 Roughness and Contact Angle of Assemblies of Uniform Core–Shell Particles

In a first approach, core–shell particles were deposited by sedimentation on slightly tilted silica wafers modified by the same polymer as the particles. They formed wellordered, densely packed two-dimensional hexagonal sphere packings [31, 32]. Some examples are shown in Figure 80.2 by SEM (a), atomic force microscopy (AFM) (b), and confocal white light sensor (MicroGlider ) images (c). X-ray scattering images (insets in Figure 80.2) confirm the perfectly hexagonal assembly. Figure 80.3 presents the root-mean-square roughness Rq of the hexagonal particle assemblies as function of the particle radius. The symbols denote experimental values obtained by MicroGlider and AFM; lines show the results of mathematical modeling of the hexagonal package [31]. The analysis of AFM images and the corresponding model leads to lower roughness values since – in contrary to the light beam used by MicroGlider – the AFM tip does not reach the bottom of the voids between the particles. Despite this quantitative difference, the coincidence of experiment and theory proves the hexagonal structure and reveals a linear relation between particle radius and roughness. In contrast to Rq, the Wenzel roughness rW (ratio between true surface and projected area) is constant for the hexagonal sphere packing (rW = 1.91) and independent of the particle radius. Figure 80.4 shows the advancing and receding contact angles of water on twodimensional hexagonal arrays of core–shell particles with different surface modification as a function of the particle radius. For comparison, the data at the left-hand side of Figure 80.4a give the contact angles on smooth surfaces modified with FPS, PS, or FSI. With increasing sphere radius (and roughness), the advancing contact angle of water increases slightly. The differences between the modifications are small; only for the PS-modified particles, a sudden decrease of the advancing contact angle was observed for large radii r = 10 μm (Figure 80.4b). With increasing sphere radius, however, the contact angle hysteresis, i.e. the difference between

®

(a)

(b)

5 μm

50 μm

Figure 80.2 (a) SEM, (b) AFM, and (c) MicroGlider image of particle assemblies on silica wafers modified by the same polymer. Particle size: (a) 1.0 μm, (b) 2.3 μm, and (c) 5.0 μm. The numbers in the images mark

®

(c)

50 μm the length of the scale bars. The insets show the corresponding X-ray scattering images. (Source: Synytska et al. 2006 [31]. Reprinted with permission of Springer.)

899

900

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

Rq (μm) 0.2

4

0.1 0 0 0.3

2

0 0

2.5

5

7.5

10 r (μm)

Figure 80.3 Root-mean-square roughness Rq of the particle assemblies as function of the particle radius r. Symbols denote the experimental values (•, AFM; ○, MicroGlider ); lines show the results of corresponding mathematical models. (Source: Synytska et al. 2006 [31]. Reprinted with permission of Springer.)

®

θadv °

θadv °; θrec°

160

180 150

150 120

FPS FPS PS 140 PS FSI FSI 130

90 60

PS FPS FSI

30 0

(a)

0.09 0.81 Particle radius (μm)

7.29

Figure 80.4 (a) Advancing (𝜃 a , filled symbols) and receding (𝜃 r , open symbols) water contact angles on surfaces modified with ordered layers of core–shell particles as function of the particle radius r (in logarithmic presentation). The shell of the particles

120

(b)

0.1

1 Particle radius (μm)

10

consists of FPS (•, ○), PS (◾, ◽), and FSI (▴, △). (b) Advancing contact angles on a stretched contact angle scale. (Source: Synytska et al. 2006 [31]. Reprinted with permission of Springer.)

advancing and receding contact angle, grows. This effect is most pronounced for particles with a PS shell, a moderately hydrophobic, non-fluorinated polymer (surface tension of smooth PS films: 𝛾 sv = 28.9 mN/m). This increase of the contact angle hysteresis with the roughness is typical for the homogeneous or Wenzel wetting regime (cf. Figure 80.1). Due to the lower surface tension, water may penetrate the voids between the particles and wet large parts of the particle surface. Particle assemblies with shells made of the more hydrophobic, fluorinated agents FPS (𝛾 sv = 24.3 mN/m) or FSI (𝛾 sv = 18.3 mN/m) exhibit a noticeably smaller roughness-induced contact angle hysteresis compared with PS, FSI less than FPS. Obviously the higher surface tension of the particle shells leads to a transition from the homogeneous (Wenzel) to the heterogeneous (Cassie–Baxter) wetting regime where the liquid (water) does not enter the cavities between the particles completely and air is entrapped in the voids. The increasing advancing contact angle for FPS and FSI is a result of the increasing curvature radii at the three-phase contact line between particle, water, and air.

80.2 Materials for Technical Applications

A theoretical description of the wetting of regular hexagonally packed structures in comparison with earlier theories and experimental results is given in [32]. In the wetting diagrams the contact angles from the wetting experiments showed deviations from the lines predicted by Wenzel and Cassie–Baxter models (cf. Figure 80.1a) that increased with increasing particle radius, decreasing intrinsic contact angle, and increasing solid free energy of the particle “shell.” This provides experimental evidence for the existence of metastable wetting configurations as proposed by Johnson and Dettre [6–8]. Although the roughness of the surface increases with growing particle radius, no superhydrophobic behavior (high advancing and receding contact angles, negligible hysteresis) was observed with two-dimensional hexagonal assemblies of core–shell particles. Obviously, surfaces with a higher roughness and a hierarchical structure are needed. Such structures were obtained by an unordered multilayer deposition of uniform core–shell nanoparticles (r = 100 nm) on planar substrates. SEM images of two examples of particle layers are shown in Figure 80.5. The layer shown in Figure 80.5a comprises FSI-covered particles and was obtained by evaporation of concentrated suspensions in ethanol (5 wt%). The layer made of PS-covered core–shell particles was prepared by strong suppression of the particle layer using the Langmuir-Blodgett layer (LBL) technique. Topographical analysis of the particle layers with FSI shell by AFM and MicroGlider showed that they adopt a fractal-like morphology with a fractal dimension D ≥ 2,5, a root-mean-square roughness Rq = 1.5 μm (which is much higher than the Rq obtained with regular hexagonal assemblies of particles with a radius of 100 nm), and a Wenzel roughness rW ∼ 1.7. They exhibited a clearly superhydrophobic behavior. Advancing and receding contact angle of water were in the order of 160∘ ; the roll-off angle (minimum tilt angle at which the water drop rolls off the surface) for water droplets with a volume of 10–20 μl was 1–2∘ . In Figure 80.6 the relation between experimentally determined contact angles of water/methanol mixtures on flat FSI-coated surfaces and fractal FSI-coated particle layers is compared with theoretical predictions of the wetting behavior on generic

H 185 nm H 152

× 10.000

(a)

H 185 nm

2 μm

× 20.000

1 μm

(b)

Figure 80.5 SEM images of unordered layers of core–shell particles (r = 100 nm) with shells made of FSI (a) and PS (b) on planar surfaces. (Source: Reprinted with permission of Synytska et al. 2009 [33]. Copyright 2009, American Chemical Society.)

901

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

1.0 COS θa, rough/fractal (COS θr,rough/fractal)

902

0.5

0.0

–0.5

–1.0 –1.0

–0.5

0.0

0.5

1.0

COS θa, flat (COS θr, flat) Figure 80.6 Experimentally measured and theoretically predicted wetting behavior on the fractal surface prepared from FSIcoated 200 nm diameter silica particles. Cosine of the (◾) advancing (𝜃 a,fractal/flat ) and (○) receding (𝜃 r,fractal/flat ) contact angles of water/methanol mixtures of the fractal surface vs. those on the flat surface. The blue solid lines indicate the wetting

behavior predicted according to the model of Onda–Shibuichi, and the red dashed lines that of the Cassie–Baxter model. The pink area comprises the wetting area according to the Cassie–Baxter model at f s = 0.03–0.11. (Source: Reprinted with permission of Synytska et al. 2009 [33]. Copyright 2009, American Chemical Society.)

rough and fractal composite surfaces by Cassie and Baxter [5] (Eq. (80.3), dashed line) and by Onda et al. [15] (Eq. (80.4), solid line). For intrinsic contact angles with cos 𝜃 in < 0, a good correlation with both theories was found. For 0 < cos 𝜃 in < 0.5, however, the measured contact angle on the fractal surfaces was – in contradiction to theoretical predictions – higher than 90∘ . As for the regular hexagonal particle assemblies, this can be explained as formation of metastable wetting configurations [6–8]. These metastable states can lead to superhydrophobic behavior with freely rolling water droplets even on unordered particle layers with a moderately hydrophobic PS shell (𝜃 in ∼ 89.5∘ ) [33]. Obviously, a moderate surface tension of the particle surface is sufficient for superhydrophobicity if the surfaces are highly structured; fluorinated surfaces are not necessary. 80.2.1.3 Preparation of Core–Shell Hybrid Janus Particles

For the creation of self-cleaning surfaces with hybrid Janus particles, the controlled aggregation and self-organization of amphiphilic particles with one hydrophobic and one hydrophilic side is used. For this purpose mono- and bipolymeric hybrid hairy Janus particles were prepared. They consist of an inorganic silica core and a shell comprising a hydrophobic polymer on one side and hydrophilic groups or a hydrophilic polymer on other side. The preparation procedure of the Janus particles is depicted schematically in Figure 80.7 [40].

80.2 Materials for Technical Applications

Wax

SiO2

Wax

ATRP

COOH

Figure 80.7 Schematic illustration of the hybrid hairy Janus particle synthesis via “grafting from”/“grafting to” approaches. (Source: Reprinted with permission of Berger et al. 2008 [40]. Copyright 2008, American Chemical Society.)

Amino-modified silica spheres (cores) are enriched at the water–wax interfaces of a wax-in-water Pickering emulsion. One side of the particles is masked by wax, while the other side is modified with an initiator for atomic transfer radical polymerization (ATRP). After dissolution of the wax, the first polymer is synthesized by a “grafting from” approach starting from this initiator. In this way, monopolymeric Janus particles are obtained; the surface of the ungrafted side is covered by amino groups. For the preparation of bipolymeric Janus particles, a second –COOH endfunctionalized polymer is bound to these amino groups by “grafting to.” The hybrid hairy Janus particles synthesized for superhydrophobic coatings were covered by a moderately hydrophobic shell (PS or poly(tert-butyl acrylate), PtBA) on one hemisphere and with amino groups or a hydrophilic polymer (PS-poly(2vinylpyridine), P2VP or PtBA-P2VP) on the opposite hemisphere of the silica particles. Furthermore, bicomponent Janus particles with a hydrophilic P2VP side and a highly hydrophobic poly(3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluorodecyl methacrylate) (PHDFMA) side were prepared [36]. 80.2.1.4 Preparation and Properties of Janus Particle Layers

For the preparation of ultra- or superhydrophobic surfaces, the controlled aggregation of the amphiphilic Janus particles in dispersion was used. Particles and

903

904

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

aggregates were deposited from the dispersion on flat substrates modified by hydrophilic P2VP or hydrophobic PS. The aggregates are distributed randomly on the particle-covered substrate and form hierarchically structured rough layers with a root-mean-square roughness Rq varying between 165 nm (PS-Janus particles) and 353 nm (PtBA-P2VP-Janus particles) and a fractal dimension D between 2.13 and 2.85 (Figures 80.8 and 80.9a). The wettability of the substrate affected both the morphology and the wettability of the Janus particle layers. So the fractal dimension of Janus particle layers on P2VP-modified substrates was higher than that on the PS-modified substrates. Furthermore the particle layers on P2VP-modified substrates were more hydrophobic than those formed on the PS-modified ones. This is a result of the interaction between the Janus particles and the substrate. On P2VP-modified substrates, the particles adsorb preferentially with their hydrophilic side; the hydrophobic side is exposed to the water droplet. On PS, the particles adsorb vice versa leading to a more hydrophilic particle layer surface. Moreover, the wetting properties of the PtBAP2VP-Janus particle layers exhibit an almost superhydrophobic behavior (with advancing and receding water contact angles >140∘ ) that arises from metastable wetting states induced by the three levels of surface roughness of the layers ranging from the nanometer scale up to the dimension of several hundreds of microns (Figure 80.8B). Using bicomponent Janus particles with a hydrophilic P2VP side and a fluorinated, highly hydrophobic (PHDFMA) side, ultrahydrophobic surfaces with advancing and receding water contact angles >150∘ were created (Figure 80.9). Recently, also the group of Liang and Yang et al. dealt with the preparation of surfaces with tunable wettability that could be changed from highly adhesive to water to superhydrophobic by changing the size distribution of cone-like Janus particles [41]. The same group used strawberry-like Janus particles for the fabrication of superhydrophobic coatings [42]. Amphiphilic Janus particles have been used not only to create superhydrophobic surfaces but also for other closely related applications. Silica particles (r = 100 nm/500 nm) covered with shells of hydrophilic (3-aminopropyl)triethoxysilane (APTES) on one side and hydrophobic octadecyltrichlorosilane (OTS) on the other side were applied to make textile materials water repellent (Figure 80.10) [43]. The hydrophilic side of the particles was bound chemically to polyethylene terephthalate (PET) fabric pretreated with poly(glycidyl methacrylate) (PGMA). At concentrations >10 wt% in the dispersion, the smaller Janus particles (r = 100 nm) distributed homogeneously on the PET fibers (Figure 80.10b), leading to static contact angles of >130∘ and small roll-off angles of water (Figure 80.10a,c). The larger particles (r = 500 nm) accumulated in the voids between the fibers; their hydrophobizing effect was lower than that of the small particles (Figure 80.10d). Robust surfaces with anti-icing properties based on a hydrophilic/hydrophobic heterogeneity were prepared by casting 1 μm large hybrid hairy Janus particles composed of a silica core covered with a poly(poly(ethylene glycol) methyl ether methacrylate) (PEGMA) and poly(dimethylsiloxane) (PDMS) polymer shell (Janus ratio 2 : 1) on silicon wafers pre-modified with PGMA, serving as an adhesion

80.2 Materials for Technical Applications

Janus particles

Aggregate of Janus particles

Water droplet

(A) PS-P2VP-JPs

PtBA-P2VP-JPs (a)

(b)

Macro

100 μm

100 μm (c)

(d) Micro

50 μm

50 μm

(e)

(f)

1 μm

1 μm (g)

Nano

(h)

400 nm

400 nm

(B) Figure 80.8 (A) Schematic illustration of the design of functional surfaces using bicomponent Janus particles (JPs); JPs form aggregates in solution, which, after deposition on a substrate, possess ultrahydrophobic properties. (B) Illustration of the multilevel hierarchy of the JP layers. First, second, and third levels of surface roughness are formed by particle agglomerates, spherical

shape of the particles, and an intrinsically rough PtBA layer, respectively (cartoons); microscopy images of PtBA-P2VP-JPs (a, c, e, g) and PS-PVP-JPs (b, d, f, h), wherein (a, b) optical microscopy images, (c, d) SEM images, (e–h) atomic force microscopy (AFM) images. (Source: Reproduced with permission of Berger et al. 2011 [36]. Copyright 2011, Wiley-VCH Verlag GmbH & Co. KGaA.)

905

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

(a) 5 μm

(b)

(c)

50 μm

Figure 80.9 (a) Topography (representative SEM images) of bipolymeric PHDFMAP2VP-Janus particles on a P2VP substrate. Snapshots of (b) static and (c) receding water droplets on the layer formed by

PHDFMA-P2VP-Janus particles. (Source: Reproduced with permission of Berger et al. 2011 [36]. Copyright 2011, Wiley-VCH Verlag GmbH & Co. KGaA.)

140

Water contact angle (°)

906

Textile modified by 10 wt% 200 nm APS-OTS-JPs

100 80 60 40

Unmodified textile

50 μm

20 0

(a)

Unmodified PET

120

0.01

1

0.1

10

Time (s)

(b) PET + 1 μm Jps

PET + 200 nm Jps

2 μm (c) Figure 80.10 Wetting properties of the textiles modified with Janus particles. (a) Time-resolved water contact angle on the native textile (orange) and the textile modified with JPs (green). (b–d) SEM morphology images and optical images (insets) of the

10 μm (d) water droplets on the native textile and textile modified with 200 nm large JPs and with 1 μm large JPs, respectively. (Source: Reprinted with permission of Kirillova et al. 2014 [43]. Copyright 2014, American Chemical Society.)

80.2 Materials for Technical Applications

promoter [44]. SEM images showed that these particles exposed both hydrophilic and hydrophobic areas at the surface, leading to inhomogeneous ice nucleation and low ice growth rates. Another application of Janus particles closely related to self-cleaning is the prevention of marine biofouling [45]. The particles prepared for this purpose were composed of either a spherical silica or a platelet-like kaolinite core, covered with hydrophilic P(PEGMA) and hydrophobic PDMS (P(PDMSMA)) at the opposite sides of the core (PDMSMA = monomethacryloxypropyl terminated polydimethylsiloxane). The spherical Janus particles were synthesized as described above; the platelet-like Janus particles were synthesized via one-step simultaneous grafting of both polymers from the opposite sides of inorganic kaolinite core particles [43]. These particles were casted on PGMA-modified silicon wafers. After deposition they exposed randomly their hydrophilic and hydrophobic sides to air. This chemical heterogeneity of the surfaces allowed a significant lowering of bacterial adhesion compared with the native control surfaces. Especially the Janus particle layers based on flat kaolinite particles were proven to be very robust and may be prepared through a reduced number of steps, allowing the fabrication of large area coatings. 80.2.1.5 Summary

Superhydrophobic behavior can be obtained with assemblies of uniform and Janus core-shell particles if the resulting layers have a hierarchical structuring, i.e. a roughness on micro- and nanoscale. The surface of the particles should be modified with a hydrophobizing agent. The use of fluorinated substances is helpful but not always necessary. Layers of amphiphilic Janus particles allow not only the creation of superomniphobic surfaces but a variety of applications preventing contamination or adhesion, e.g. by soil, ice or biofouling. 80.2.2 Superhydrophobic Aluminum Surfaces

Due to a number of beneficial properties, e.g. good processability, density, stability, and corrosion resistance, aluminum materials are currently used for a broad variety of applications and rank highly under the available materials. They are usually exposed to atmospheric conditions; decorative aspects often play a great role. Therefore, corrosion and staining should be avoided; repellency of water would be a desirable property. Metal and metal oxide surfaces exhibit, however, a high surface free energy and polarity; they are usually wetted very well by water. Hydrophobic surface modifications lowering the surface free energy of the aluminum surface are an effective tool to reduce the wettability by water. Complete repellency of water and good self-cleaning properties are obtained by superhydrophobic surfaces [46, 47]. Superhydrophobicity requires not only a low surface free energy but also a rough, hierarchically structured surface morphology.

907

908

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

80.2.2.1 Preparation of Surface Roughness

To obtain superhydrophobic aluminum surfaces, it is necessary to create an appropriate surface roughness that can be introduced either by a structured coating [48] or – as in this study – by structuring the aluminum surface itself. For the structuring of aluminum on a technically relevant scale, different techniques are available: mechanical methods, laser treatment, or etching. Mechanical structuring (e.g. micro-embossing) leads to rather coarse structure that has to be refined by an additional treatment, e.g. by etching [49]. Laser treatment [50] is a scanning technique. It is suitable for the structuring of small areas and creates a regular surface structure. Larger areas can be roughened more easily by etching [51]. This method produces an irregular surface morphology. It can be performed using acidic or basic media. Another possibility is the anodic oxidation, which is an established technique for the corrosion protection of aluminum materials. By variation of the process parameters (intensified anodic oxidation [52, 53]), this method was performed in a way that simultaneous growth and partial dissolution of the aluminum oxide layer led to a mountain-like oxidic surface profile (Figure 80.11). Besides these specimens that were prepared in sulfuric acid, some samples were roughened using phosphoric acid [55]. These surfaces exhibited a thinner oxide layer and a finer microprofile. 80.2.2.2 Selection and Properties of Hydrophobizing Layers

Freshly oxidized aluminum surfaces are very hydrophilic and are considered as high surface energy surfaces according to the classification by Zisman [56]. To obtain superhydrophobic properties, they have to be coated completely by hydrophobizing organic or organic/inorganic materials. The thickness of these coatings should ensure the formation of a closed film without affecting the surface morphology. Especially the cavities created by etching must not be filled or leveled out by the coating. Dense, molecularly thin hydrophobizing films can be created by molecules undergoing a molecular self-organization during the deposition or adsorption process [27]. Such molecules consist of an alkyl chain with 10–20 carbon atoms (which may be partially fluorinated) and an anchor group allowing electrostatic or chemical bonding to the surface. After deposition, the alkyl chains align parallel, similar to the bristles of a brush (cf. Figure 80.12). Methyl or trifluoromethyl groups dominate (a)

(b)

20 μm

(c)

2 μm

20 μm

Figure 80.11 SEM images of an anodically oxidized and roughened aluminum substrate. (a) Top view, (b) tilted (35∘ ) and magnified image, and (c) polished cross section. (Source: Reprinted with permission of Frenzel et al. 2003 [54]. Copyright 2010, Wiley.)

80.2 Materials for Technical Applications

the surface. Their low surface free energy ensures the desired hydrophobic surface properties. The thickness of the coating corresponds to the length of the molecule and is small compared with the roughness of the aluminum surface, thus preventing a leveling of the structure. Several classes of compounds can be applied for that purpose. The use of alkylthiols as hydrophobizing agents requires a thin base layer of noble metals, usually gold, on the substrate surface [57, 58]. Alkyl trichlorosilanes and monoalkyl phosphates can be deposited directly on aluminum oxide surfaces [59]. In our group, additionally alkyl trialkoxy silanes and alkanephosphonic acids were applied to prepare superhydrophobic aluminum surfaces (Figure 80.12) [53, 54]. A possible disadvantage of monomolecular films is their low thickness that challenges the mechanical stability of the hydrophobizing layer. Another weak point is the monomolecular connection with the substrate. Breaking the singular bonding leads to irreversible damage of the film. It is therefore desirable to develop hydrophobizing layers made of larger molecules, i.e. suitable polymeric systems. The proof of concept was performed with Teflon AF (DuPont), an amorphous fluoropolymer – poly[4,5-difluoro-2,2bis(trifluoromethyl)-1,3-dioxole-co-tetrafluoroethylene], 65 : 35 mol/mol [53, 54]. In contrast to the commonly known Teflon or or [poly(tetrafluoroethylene)], PTFE, Teflon AF is soluble in the fluoro-organic solvent Fluorinert FC-75 by 3M Company (perfluoro(2-n-butyltetrahydrofuran)) and can thus be applied from solution. Since Teflon AF is quite expensive and its solubility is limited to fluoro-organic solvents, its use on a technical scale is rather unlikely. The coating of pretreated aluminum sheets with other commercially available and soluble partially fluorinated copolymers, e.g. poly(tetrafluoroethylene-co-vinylidene fluoride-co-propylene), did, however, not yield satisfactory results. Polymethacrylic acid esters with various alcohol components are a good basis for copolymers with variable composition of the side chains [60]. Thus different properties can be combined in one polymer molecule. For superhydrophobic coatings, the rather hydrophobic polymer poly(tert-butyl methacrylate) (PtBMA) appeared to be convenient [61]. This polymer did, however, not wet the hydrophilic

®

®

CH3

CH3

CH3

CH3

®

CH3 Alkyl chains C12…C18 Adsorbed molecules of phosphonic acids

OPO OPOH OPO OPO OPOH O

O

O

O

O

Oxide layer Aluminum Figure 80.12 Schematic composition of a roughened and oxidized aluminum surface with a superhydrophobic modification by adsorbed alkyl phosphonic acid (not in scale). The scheme shows the molecular

self-organization, leading to an enrichment of the methyl end groups at the interface to air. (Source: Reprinted with permission of Frenzel et al. 2003 [54]. Copyright 2010, Wiley.)

909

910

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

aluminum oxide surface sufficiently; no closed films were formed on the roughened surfaces. The wetting properties of PtBMA could be improved by copolymerization of tert-butyl methacrylate monomers with methacrylates containing more polar alcohol residues. Additional integration of fluoro-organic monomers in the copolymers ensured a low surface free energy and hydrophobicity. The composition of the copolymers is shown schematically in Figure 80.13. The combination of hydrophilic and hydrophobic sequences in one polymer molecule thus allows both good adhesion and dense film formation on hydrophilic substrates and the formation of hydrophobic, water-repellent surfaces. 80.2.2.3 Surface Properties of Chemically Modified Aluminum Surfaces

Copolymer films of methacrylic acid esters with a varying number of hydrophilic sequences and strongly hydrophobic fluorinated sequences were first prepared on flat glass substrates. The number of hydrophilic sequences (marked B in Figure 80.13) had only little effect on the solid surface tension determined experimentally by contact angle measurements. Variation of the number of hydrophobic sequences (marked C in Figure 80.13), however, led to strong changes of the surface tension. This reduction of the surface tension by an increasing content of fluoro-organic components relies on the fact that fluoro-organic side chains enrich F F F

F

F F

F F

C

F F

F

F

F

F

F

F

F

O

O

O O

O O

O

O

O O

O

O

O O

A

O

O

O O

O O

OH B

Figure 80.13 Schematic presentation of a statistic copolymer made of tertbutyl methacrylate (A), hydrophilizing 2hydroxyethyl methacrylate (B, alternatively

2-(methacryloyloxy)ethyl acetoacetate), and strongly hydrophobizing Zonyl TM monomers (C, Zonyl TM is a mixture of perfluoralkylethyl methacrylates).

80.2 Materials for Technical Applications

at the interface with air in order to reduce the free enthalpy of the system. In the case of methacrylate copolymers, this effect was verified by X-ray photoelectron spectroscopy (XPS), a very surface-sensitive element-analytical method (maximum information depth ca. 8 nm) [62, 63]. As an example, XPS spectra of both sides of a fluoro-modified tert-butyl methacrylate copolymer film prepared on a glass slide and pulled off afterwards show significant differences in the typical peaks. The spectrum taken at the interface to air shows a much higher F 1s peak with a relative atomic concentration of fluorine [F] = 13.9% than the spectrum taken at the bottom side, the former interface with the polar glass (atomic concentration [F] = 7.6%). In contrast, the atomic concentrations of C and H are higher at the interface to glass [60]. That proves that the fluorinated groups segregate to the interface polymer/air while the polar sequences are oriented more toward the polar glass surface. This preferential orientation explains the low influence of the number of polar sequences on the surface tension (𝛾 sv ) of these copolymer films. To obtain superhydrophobic aluminum surfaces with a high roughness and low surface energy, the copolymer films described above were prepared on the aluminum surfaces roughened by anodic oxidation shown in Figure 80.11. Hydrophilic sequences in the copolymer ensured an improved wetting of the rough substrates and the formation of sufficiently closed films. Their thickness could be adjusted to very low values such that the surface topography of the etched aluminum surfaces was maintained and even small cavities were not filled or leveled out. Figure 80.14 compares the surface free energy (𝛾 sv ) and the advancing and receding contact angles of water on tert-butyl methacrylate-based copolymers deposited on smooth oxidized silicon wafers and on anodically oxidized and roughened aluminum sheets. The surface free energy was calculated from the water contact angles on the flat surface using the equation-of-state model [64, 65]. Copolymerization of the fluorocarbon Zonyl TM decreased the surface free energy of the copolymers from about 30 to 20 mJ/m2 . Not only were the advancing contact angles of water on flat surfaces increased from ∼90∘ to >100∘ , but also the contact angle hysteresis became higher with Zonyl TM. Interestingly, superhydrophobic surfaces characterized by high advancing and receding contact angles and a negligible contact angle hysteresis were obtained only with the copolymers 11a (without Zonyl TM), 11b, und 12b (with Zonyl TM) on roughened aluminum. This demonstrates that a high surface energy of the copolymer is not the only precondition for superhydrophobic behavior on rough aluminum substrates. It is essential, too, to introduce hydrophilic sequences into the copolymers to get a sufficiently high compatibility of the aluminum surface and the copolymer as prerequisite for complete wetting and coverage. Comparison of the contact angles on rough aluminum substrates and smooth silicon wafers underlines the significance of the roughness for superhydrophobicity. Obviously the intensified anodic oxidation provides a morphology with high roughness and cavities, leading to a heterogeneous wetting regime (Cassie–Baxter regime [66]) in contact with water. The surface segregation of implemented partially fluorinated alkyl chains allows to create superhydrophobic aluminum surfaces even with the much more hydrophilic polymer poly(methyl methacrylate) (PMMA). For PMMA-based copolymers, a low

®

911

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications 50

Surface free energy

Contact angles on smooth silicon substrate

160

θa

140

40

Contact angles on rough AlMg1 substrate

160 140

0

(b)

Sample number

Figure 80.14 (a) Surface free energy 𝛾 sv of tert-butyl methacrylate-based copolymers on flat silicon wafers calculated from the water contact angles in (b). (b, c) Advancing contact angle 𝜃 a and receding contact angle 𝜃 r of water on the copolymers (b) on smooth silicon wafers (measured by axisymmetric drop shape analysis-profile ADSA-P [64]) and (c) on anodically oxidized and roughened aluminum sheets (measured by DSA 10, Krüss GmbH Hamburg).

10 a

12

a 10

a

b 12

b

12

a

11

11

a

b

10

10

(a)

b

0

0

12

20

b

20

11

40

12 a

40

10

b

60

10

60

a

80

12 b

80

11 b

100

b

100

11 a

20

120

10

30

120

11 a

θr Contact angle θ (°)

Surface free energy γsv (mJ/m2)

912

(c)

Explanation of the sample numbers: 10a, poly(tert-butyl methacrylate) (PtBMA); 10b, poly(tert-butyl methacrylate-co-Zonyl TM), 9 : 1; 11a, poly[tert-butyl methacrylate-co-2(methacryloyloxy)ethyl acetoacetate], 9 : 1; 11b, poly[tert-butyl methacrylate-co-Zonyl TM-co-2-(methacryloyloxy)ethyl acetoacetate], 8 : 1 : 1; 12a, poly(tert-butyl methacrylateco-2-hydroxyethyl methacrylate), 9 : 1; 12b, poly(tert-butyl methacrylate-co-Zonyl TM-co2-hydroxyethyl methacrylate), 8 : 1 : 1.

surface free energy and advancing water contact angles >90∘ on flat substrates were reached if they contained the fluorocarbon Zonyl TM (Figure 80.15), but only the copolymer 21b deposited on roughened aluminum exhibited superhydrophobic behavior. In contrary, PMMA without fluorinated comonomers (20a) formed a hydrophilic film on roughened aluminum. Obviously, superhydrophobicity can be reached only if a delicate balance between hydrophilic and hydrophobic groups is maintained. 80.2.2.4 Summary

Superhydrophobic aluminum surfaces were created by using the fundamental thermodynamic principle of energy minimization of a material system by minimizing the specific free surface energy. The polar oxidized and roughened aluminum surfaces were coated with polymers containing hydrophilic and hydrophobic sequences. The method is not limited to polymers with native hydrophobic sequences. Polymers with mainly hydrophilic groups can be used to create superhydrophobic coatings after copolymerization of small amounts of partially fluorinated monomers. The incorporation of hydrophilic comonomers appears to be advantageous for the wetting and film formation on polar metal and metal oxide substrates. Because of the segregation of the hydrophobic sequences to the interface to air, the influence of the hydrophilic moieties on the solid surface tension was low for all investigated copolymers.

80.2 Materials for Technical Applications 50

Contact angles on smooth silicon substrate

Surface free energy

160

Contact angles on rough AlMg1 substrate

160

θa

20

0

0

(a)

(b)

Figure 80.15 (a) Surface free energy 𝛾 sv of methyl methacrylate-based copolymers on flat silicon wafers calculated from the water contact angles in (b). (b, c) Advancing contact angle 𝜃 a and receding contact angle 𝜃 r of water on the copolymers (b) on smooth silicon wafers (measured by axisymmetric drop shape analysis-profile ADSA-P [64]) and (c) on anodically oxidized and roughened aluminum sheets (measured by DSA 10, Krüss GmbH Hamburg).

20 a

22

21

20 a

b

a

22

b 21

22

20 b

21 a

20 a

0

Sample number

b

20

a

40

22

40

10

22

60

a

60

21 b

80

21

80

a

100

b

100

a

120

21

20

120

20 b

Contact angle θ (°)

Surface free energy γsv (mJ/m2)

30

140

θr

20 b

140

40

(c)

Explanation of the sample numbers: 20a, poly(methyl methacrylate) (PMMA); 20b, poly(methyl methacrylate-co-Zonyl TM), 9 : 1; 21a, poly[methyl methacrylate-co-2(methacryloyloxy)ethyl acetoacetate], 9 : 1; 21b, poly[methyl methacrylate-co-Zonyl TMco-2-(methacryloyloxy)ethyl acetoacetate], 8 : 1 : 1; 22a, poly(methyl methacrylate-co2-hydroxyethyl methacrylate), 9 : 1; 22b, poly(methyl methacrylate-co-Zonyl TM-co-2hydroxyethyl methacrylate), 8 : 1 : 1.

80.2.3 Superhydrophobic and Oleophobic Sol–Gel Coatings

During the last 20 years, sol–gel processing has become a promising technical method for the large-scale preparation of hydrophobic coatings. The sol–gel process bases on monomer alkoxy compounds of silicon or metals that are catalytically hydrolyzed and cross-linked by condensation [67]. Sol–gel coatings combine the specific advantages of organic and inorganic coatings and can be applied on a technical scale by industrially relevant techniques, e.g. dipping, coating with a doctor blade, or spraying. A high mechanical stability is obtained with increased content of inorganic material. The properties of sol–gel layers (e.g. charge or wettability) can be modified by functional groups, additives, and fillers. Sol–gel formulations including fluorine compounds as fluorinated polyethers and organosilanes with fluorinated side chains were developed for use as antiadhesive, “easy-to-clean,” and anti-fingerprint varnishes [68, 69]. Starting from these commercially available sol–gel coatings, new antiadhesive nano-varnishes have been developed to create “easy-to-clean” systems allowing the removal even of very adhesive soils. As basis for the new coatings, the formulations H 1006 (fluorine-free) and H 5055 (with fluorine) (FEW Chemicals GmbH, Wolfen)

913

914

80 Self-Cleaning Surfaces: From Fundamental Aspects to Real Technical Applications

were used. These sol–gel coatings are sprayed on metal substrates, and dried and cross-linked at 150 ∘ C. They form smooth, homogeneous films with a thickness of 1.5–3 μm (depending on the sprayed mass) and a root-mean-square roughness Rq ∼ 25 nm. The fluorine-free H 1006 is oleophilic and moderately hydrophilic. H 5055 is hydrophobic; the contact angle of hexadecane is above 65∘ (cf. Table 80.1). A noticeable increase of the contact angles of both water and dodecane (as model oil) was obtained by addition of fluorine compounds. The addition of the perfluoropolyether Fluorolink S10 (Solvay Specialty Polymers Italy S.p.A.) to H 1006 led to water contact angles above 100∘ and a static contact angle of dodecane above 50∘ . Therefore Fluorolink S10 was selected for the development of superhydrophobic coatings. Comparable contact angles were observed by other groups using spin-coated sol–gel layers including perfluoropolyether with alkoxysilane functionalities having very low solid surface tensions of 14–16 mN/m [71, 72].

®

80.2.3.1 Development of Rough Superhydrophobic Sol–Gel Coatings

Superhydrophobic behavior can only be reached with rough surfaces. A common method to create rough superhydrophobic layers is the incorporation of silica nano- and microparticles into fluorinated sol–gel dispersions [73–76]. Shirtcliffe et al. [77] produced superhydrophobic foams using a sol−gel phase separation process. Thermally stable, transparent, and corrosion-resistant fluorine-free silica films with superhydrophobic properties were developed by Mahadik and coworkers by a special sol–gel process forming ball-like silica particles [78, 79]. Li et al. created superhydrophobic fluorinated silica sol–gel layers by dip coating and calcination [80]. Taurino et al. [81] added perfluoropolyethers and metal alkoxide precursors to a sol–gel formulation. Layers obtained by air brushing exhibited a root-mean-square roughness up to 3 μm, hysteresis-free water contact angles of about 150∘ , and in some cases even contact angles of the unpolar liquid hexadecane above 100∘ with low hysteresis. Sprayed sol–gel multilayer assemblies with a fluoropolymer on top feature a higher scratch resistance [82]. In a study by the authors in cooperation with FEW Chemicals, Wolfen, Germany, three different strategies were checked to increase the roughness of the sol–gel layers: (1) Addition of various micro- and nanosized pigments (corundum, SiOx , SiC, TiOx , blue pigment) and Fluorolink S10 to H 1006 increased the root-meansquare roughness of the layers depending on the size and concentration of the pigments. The static water contact angle was – due to the addition of Fluorolink S10 but independent of the pigments – in the order of 105∘ , but no superhydrophobic behavior was observed. (2) A thin layer of H 5055 was sprayed on base coats made of H 1006 with incorporated SiOx and SiC particles and ultrahigh-molecular-weight polyethylene GUR 2126 (Ticona GmbH). The roughness of the base coats was controlled by varying size, mixing ratio, and concentration of the particles and polyethylene. H 5055 formed a smooth thin film without any microstructure on these

®

Table 80.1 Elemental ratios (from XPS), root-mean-square roughness Rq, and advancing and receding contact angles of water and hexadecane on sprayed sol–gel layers (Fllk, Fluorolink S10; FP, fluoropolymer dispersion; n.m., not measured).

®

Elemental ratios (XPS)

Roughness

Sample

Additives

[F]:[C]

[O]:[C]

[Si]:[C]

Rq (𝛍m)

H 1006 H 5055 HB1_FP HB5_FP1 HB5_FP2

None None FP Fllk + FP Fllk + FP

0.76 1.475 0.68

0.75 0.05 0.43

0.325 0.01 0.15

(26 ± 3) × 10−3 (25 ± 2) × 10−3 14.8 ± 1.2 16.9 ± 2.6 18.4 ± 1.5

(Source: Drechsler et al. 2017 [70]. Reprinted with permission of Elsevier.)

𝜽adv, H

2O

(∘ )

77 ± 1 114 ± 1 146 ± 2 134 ± 3 138 ± 3

𝜽rec,

Contact angles (∘ ) 𝜽

H2 O

52 ± 7 97 ± 1 145 ± 1 134 ± 3 135 ± 2

adv, HD

(∘ )

40 mN/m [70]. All HB5_FP1 layers exhibit high advancing contact angles of hexadecane. The receding contact angle of hexadecane is, however, low until a Wenzel roughness of 5 is exceeded. This reflects the transition from the homogeneous Wenzel to the heterogeneous Cassie–Baxter wetting regime as shown in Figure 80.1c [11]. Since the intrinsic contact angle of hexadecane is