Terahertz Science And Technology For Military And Security Applications 9789812771803, 9789812771797

Citation preview

TERAHERTZ SCIENCE AND TECHNOLOGY FOR MILITARY AND SECURITY APPLICATIONS

SELECTED TOPICS IN ELECTRONICS AND SYSTEMS

Editor-in-Chief: M. S. Shur

Published Vol. 30: Terahertz Sensing Technology - Vol. 1 Electronic Devices and Advanced Systems Technology eds. D. L. Woolard, W. R. Loerop and M. S. Shur Vol. 31: Advanced Device Modeling and Simulation ed. T. Grasser Vol. 32: Terahertz Sensing Technology - Vol. 2 Emerging Scientific Applications and Novel Device Concepts eds. D. L. Woolard, W. R. Loerop and M. S. Shur Vol. 33: GaN-Based Materials and Devices eds. M. S. Shurand R. F. Davis

Vol. 34: Radiation Effects and Soft Errors in Integrated Circuits and Electronic Devices eds. R. D. Schrimpf and D. M. Fleetwood Vol. 35: Proceedings of the 2004 IEEE Lester Eastman Conference on High Performance Devices ed. Robert E. Leoni 111 Vol. 36: Breakdown Phenomena in Semiconductors and Semiconductor Devices M. Levinshtein, J. Kostamovaara and S. Vainshtein Vol. 37: Radiation Defect Engineering Kozlovski V. and Abrosimova V. Vol. 38: Design of High-speed Communication Circuits ed. R. Harjani Vol. 39: High-speed Optical Transceivers eds. Y, Liuand H. Yang Vol. 40: Sic Materials and Devices - Vol. 1 eds. M. S. Shur, S. Rumyantsev and M. Levinshtein Vol. 41 : Frontiers in Electronics Proceedings of the WOFE-04 eds. H Iwai, Y. Nishi, M. S. Shurand H. Wong

Vol. 42: Transformational Science and Technology for the Current and Future Force eds. J. A. Parmentola, A. M. Rajendran, W. Btyzik, B. J. Walker, J. W. McCauley, J. Reifman, and N. M. Nasrabadi Vol. 43: Sic Materials and Devices - Vol. 2 eds. M. S. Shur, S. Rumyantsev and M. Levinshtein

Vol. 44: Nanotubes and Nanowires ed. Peter J. Burke Vol. 45: Proceedings of the 2006 IEEE Lester Eastman Conference on Advanced Semiconductor Devices eds. Michael S.Shur, P. Maki and J. Kolodzey

Selected Topics in Electronics and Systems - Vol. 46

TERAHERTZ SCIENCE AND TECHNOLOGY FOR MILlTARY AND SECURITY APPLICATIONS Editors

Dwight 1. Woolard US Army Research Office, USA

James 0. Jensen U.S. Army Edgewood Chemical Biological Center, USA

R. Jennifer Hwu Innosys, Inc.

Michael S. Shur Rennsselaer Polytechnic Institute, USA

N E W JERSEY

*

LONDON

*

Scientific 1: World -

SINGAPORE

BElJlNG

*

SHANGHAI

.

HONG KONG

*

TAIPEI

*

CHENNAI

Published by World Scientific Publishing Co. Re. Ltd. 5 Toh Tuck Link, Singapore 596224 USA ofice: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601 (IKofice: 57 Shelton Street, Covent Garden, London WC2H 9HE

British Library Cataloguing-in-PublicationData A catalogue record for this book is available from the British Library.

TERAHERTZ SCIENCE AND TECHNOLOGY FOR MILITARY AND SECURITY APPLICATIONS Selected Topics in Electronics and Systems - Vol. 46 Copyright Q 2007 by World Scientific Publishing Co. Pte. Ltd. All rights reserved. This book, or parts thereox may not be reproduced in any form or by any means, electronic or mechanical, including photocopying, recording or any information storage and retrieval system now known or to be invented, without written permission from the Publisher.

For photocopying of material in this volume, please pay a copying fee through the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to photocopy is not required from the publisher.

ISBN- 13 978-981-277-179-7 ISBN-10 981 -277-179-4

Editor: Tjan Kwang Wei

Printed in Singapore by Mainland Press Pte Ltd

FOREWORD In recent years, the field of Terahertz (THz) science and technology has entered a completely new phase of unprecedented expansion that is generating every growing levels of broad-based international attention. Indeed, the plethora of activities that have arisen recently in both the technology and scientific arenas associated with the THz frequency domain - i.e., usually defined as the portion of the submillimeter-wavelength electromagnetic (EM) spectrum between approximately 1 millimeter (300 GHz) and 100 micrometers (3 THz) - suggest that the field might be attempting to undergo a dramatic transition that could lead to long-awaited payoffs in a number of application areas. The inherent advantages and potential payoffs of the THz regime for military and security relevant applications have long stood as an important driver of interest in this science and technology area. For example, this extremely expansive and spectrally unique portion of the EM spectrum was initially of high interest for such applications as space-based communications, upper atmospheric sensing and communications, and potentially for short-range terrestrial communications and non-intrusive package screening. However, the very rapid growth in more recent years is arguably most closely linked to the potential payoffs of THz sensing and imaging (THz-S&I) for an array of military and security applications. These applications include the spectroscopic-based detection identification and characterization of chemical and biological (CB) agents and materials, remote and standoff early-warning for CB warfare threats, and video-rate imaging of concealed weapons and explosives, just to name a few. In addition, these same THz-S&I capabilities have a close synergy and dual-use potential for private-sector application areas as biological science, medical diagnostics, pharmaceutical characterization and security screening. While recent developments in the THz field offers promise that a broad spectrum of commercial and scientific payoffs may be on the horizon, there still remain significant technological challenges that will need to be resolved. In particular, the refinement of THz systems is needed to enable very near-term payoffs such as security screening and the detailed characterization of materials such as explosives and pharmaceuticals; a significant advancement is needed in THz source and detector technology to enable medium-range applications such as remoteMandoff detection and identification of biological and chemical agents; and, new breakthroughs in sensor architectures and probing techniques will be required for enabling far-future applications such as detailed spectroscopic characterizations of biological molecules and nanoscale systems. This special issue presents some of the leading fundamental research efforts that working towards the realization of practical THz-S&I capabilities for military and security applications. Specifically, the papers that follow span an array of pertinent THz Science

V

vi

Fore word

and Technology (THz-S&T) subjects that will have important ramifications to the future success of THz-S&I applications. Relevant subjects included the theoretical prediction and/or measurement of THz spectroscopic phenomenon in solid-state materials such as high explosives (e.g., HMX, PETN, RDX, TNT, etc.), carbon-fiber composites, biological agents (e.g., DNA, RNA, proteins, amino acids) and organic-semiconductor nanostructures. Individual papers in this special issue also address the effective utilization of state-of-the-art THz-frequency technology in military and security relevant scenarios such as standoff S&I, screening of packages and personnel, and perimeter defense. Technical papers included in this volume also introduce novel devices and/or concepts that enhance THz source and detector performance, that enable completely new types of sensor functionality at THz frequency (e.g., detection at nanoscale/molecular levels), and that define new and innovative sensing modalities (e.g., remote personnel identification) for defense and security. Therefore, the collective research presented in this special issue represents a valuable source of information on the evolving field of THz-S&I for Military and Security Applications.

Editors: Dwight L. Woolard, U.S. Army Research Office R. Jennifer Hwu, Innosys, Inc. James 0. Jensen, U.S. Army Edgewood Chemical Biological Center Michael S. Shur, Rensselaer Polytechnic Institute

vi

CONTENTS

Foreword

V

Development of Computational Methodologies for the Prediction and Analysis of Solid-state Terahertz Spectra D. G. Allis and T. M. Korter

1

Fire Damage on Carbon Fiber Materials Characterized by THz Waves N . Karpowicz, D. Dawes, M. J . Perry and X.-C. Zhang

21

An Analysis of the THz Frequency Signatures in the Cellular Components of Biological Agents A. Bykhovski, T. Globus, T. Khromova, B. Gelmont and D. Woolard

33

Standoff Sensing and Imaging of Explosive Related Chemical and Bio-Chemical Materials Using THz-TDS H. Zhong, A . Redo-Sanchez and X.-C. Zhang

47

Fingerprinting Insulins in the Spectral Region from Mid-IR to THz R. Song, Y. J. Ding and Y. B. Zotova

59

Ambient Air Used as the Nonlinear Media for THz Wave Generation X . Xie, J . Dai, M. Yamaguchi and X.-C. Zhang

69

Time Domain Terahertz Imaging of Threats in Luggage and Personnel D. Zimdars, J. White, G. Stuk, G. Sucha, G. Fichter and S. L. Williamson

79

Experimental and Density Functional Theory Study on THz Spectra of 4-NT and 2, 6-DNT Y. Chen, H. Liu and X.-C. Zhang Interactions of THz Vibrational Modes with Charge Carriers in DNA: Polaron-Phonon Interactions D. Ramadurai, T. Yamanaka, Y. La, M. Vasudev, V. Sankar, M. Dutta, M. A. Stroscio, T. Rajh and Z. Saponjic Designed Self-organization for Molecular Optoelectronic Sensors M. Norton Study of Nano-Structured Silicon-Phenyl Nanoclusters Towards Single Molecule Sensing C. Herrera and J. M. Seminario

vii

91

101

119

135

...

viii

Contents

An Optically-Triggered I-RTD Hybrid THz Oscillator Design D. Woolard, W. Zhang, E. Brown, B. Gelrnont and R. Trerw

147

Terahertz Phonon-Polariton Imaging for the Application of Chemical Detection M. Yamaguchi, M. Wang and P. Suarez

163

New Technique to Suppress Sidelobe Clutter in Perimeter Security Systems G. W . Webb, I. V. Minin and 0. V. Minin

175

Millimeter and Sub-Millimeter Wave Performance of an ErAShAlGaAs Schottky Diode Coupled to a Single-Turn Square Spiral E. R. Brown, A. C. Young, J. E. Bjarnason, J . D. Zimmerman, A. C. Gossard and H. Kazemi

191

Long-Wave Infrared and Terahertz-Frequency Lasing Based on Semiconductor Nanocrystals V. I. Rupasov and S. G. Krivoshlykov

203

Multi-Spectral Terahertz Imaging Using Reflected and Scattered Radiation M. C. Kemp, A . Glauser and C. Baker

211

Remote Identification of Foreign Subjects A . Sokolnikov

223

Terahertz Interferometric and Synthetic Aperture Imaging A . M. Sinyukov, A. Bandyopadhyay, A . Sengupta, R. B. Barat, D. E. Gary, Z.-H. Michalopoulou, D. Zimdars and J. F. Federici

239

International Journal of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 193-212 @ World Scientific Publishing Company

World Scientific www.worldscientific.com

DEVELOPMENT OF COMPUTATIONAL METHODOLOGIES FOR THE PREDICTION AND ANALYSIS OF SOLID-STATE TERAHERTZ SPECTRA Damian G. Allis

Department of Chemistty, Syracuse Universiq, I I I College Place, Syracuse, New York 13244, USA damian@somewhereviNe. corn Timothy M. Korter

Department of Chemistv, Syracuse University, I I I College Place. Syracuse, New York 13244, USA [email protected]

The analytical applications of terahertz (THz) spectroscopy for the characterization of molecular solids have been limited by the lack of information concerning the assignment of observed spectral features to specific internal (intramolecular) and external (intermolecular) atomic motions. Computational methodologies addressing the assignment of spectral data are the enabling technology for moving THz spectroscopy to the forefront of available detection methods for both imaging and spectroscopic applications. Solid-state density functional theory (DFT) studies have been performed on the high explosives cyclotetramethylenetetranitramine (HMX) and pentaerythritol tetranitrate (PETN) in order to address the dependencies of the predictions of solid-state vibrations in the terahertz (3 to 120 cm-') region on the choice of basis set and integration grid size, building on previous work that examined this dependency on the choice of density hnctional. DFT THz simulations reveal that both the choice of basis set and grid size have important influences on the reproduction of spectral features. The sensitivity to basis set choice is most pronounced in the calculation of vibrational intensities, where it is found that THz absorption intensities are most accurately reproduced when derived from basis set-sensitive Mulliken atomic charges as opposed to basis set-insensitive atomic charges generated by the Hirshfeld partitioning method. Keywords: Terahertz spectroscopy; solid-state density functional theory, HMX, PETN.

1. Introduction

The broad application of terahertz (THz) spectroscopy as a detection, analysis, and characterization technique for the identification of chemical threat agents', pharmaceutical polymorphs2, and a host of other molecular-based materials has been slowed by the lack of a rigorous means for unambiguous spectral assignment. The THz studies performed on molecular crystals and solution-phase samples have revealed that many molecules have distinct spectral fingerprints in this region, often attributable to

194

D. G. illlis t3 T. M.Korter

intramolecular motions (such as large-amplitude vibrations and bond torsions) and intermolecular motions (such as coordinated oscillations in complexes and molecular motions in the solid-state and in solution). As with all molecular, aggregate, and solidstate vibrational studies, computational methods greatly reduce the complexity of vibrational mode assignments by providing predictions of normal mode energies and atomic motions. In solution-phase THz studies, the agreement between theory and experiment is often excellent because the single molecule in solution exists in a disordered, often weakly-interacting environment that is reasonably well approximated in the calculations by its complete neglect (the “gas-phase’’ limit). Attempts to employ these same isolated-molecule calculations for solid-state THz spectral assignments of molecular crystals have proven far less successful. On the one hand, the reason for the inaccurate reproduction of solid-state spectra from isolatedmolecule calculations seems obvious. Solid-state interactions between molecules often include strong non-covalent interactions (such as hydrogen-bonding) in addition to the hard-sphere packing that define the molecular environment, together constraining single molecules and their motions. This periodic chemical environment, more restrictive than a molecule would experience in solution, would be expected to have a significant influence on the energies of molecular vibrations and the atomic motions. In the absence of information about the molecular vibrational motions and their energies, even the correlation of solution-phase and solid-state THz spectra would prove difficult in those cases where both sets of data exist, as the actual influence the crystal cell environment has on the motions and energies of molecules is not at all clear. Given the agreement between theory and solution-phase THz measurements, the use of isolated-molecule calculation in solid-state studies seems the most likely cause for any theoretical disagreements. Recent density functional theory (DFT) studies attempting the reproduction of crystalline sample THz spectra have demonstrated that the inclusion of both the crystal cell interactions and the constrained environment of the cell have considerable influence on the positions of the vibrational modes and the molecular motions. The assignment of the THz spectra of the molecular high explosives HMX3 and PETN4, have both been reported recently for room temperature crystalline samples. These two molecules are shown in Figure 1. In both cases, agreement between theory and experiment was found to be adequate for assigning all of the THz features between 3 and 120 cm-’, and that this observed agreement was dependent on the choice of density functional used. The empirical nature of DFT (the development of density functionals to account for electron correlation) has resulted in the development of numerous density functionals that all have their strengths and weaknesses with respect to the prediction andlor reproduction of molecular properties. The reproduction of solid-state vibrational modes by DFT shows great variation with respect to the functional choice. In the study of HMX, the VWNBP6-’ density functional was found to best reproduce the solid-state THz spectrum, with the BP’, density functional also agreeing well with experiment. In the case of PETN, the BP density functional was found to best reproduce the measured THz spectrum, while the VWN-BP functional showed clear deficiencies in aspects of the spectral reproduction. The remaining functionals examined in these studies indicated that all were limited in their abilities to simulate major features of the THz peak positions and intensities despite accurately reproducing the molecular geometries in their crystal environments.

’

2

Analysis of Solid-State Terahertz Spectra

195

0.b .PO N

I

oJJ\.

0

PETN

HMX

Figure 1. Molecular geometry and symmetry-unique atom labels (for charge analyses in Figure 7) for the high explosives HMX (left, C, symmetry) and PETN (right, S4 symmetry). CPK renderings performed with VMD9.

The identification of clear trends in the computational simulations of HMX and PETN have bearing on other solid-state THz simulations, although much remains to be done to confirm or counter that these two studies are, in fact, diagnostic of THz simulations in general. Among the near-term issues related to the choice of theory in THz simulations is the comparison of results from the functional-exhaustive HMX and PETN studies and the choice of functionals employed in other recent solid-state THz simulations. In two studies of P-oligopeptide sheets" and various trialanine crystals", the PW91" density functional was used with the DNP basis set (comparable to a 6-31G(d,p) Gaussian-type basis set and the same as used for the HMX and PETN studies, see below) to simulate measured solid-state THz spectra. While it was possible to assign some of major features in the two spectra based on the calculations, this was largely due to the well-resolved experimental spectra and not, specifically, to agreement in calculated peak positions and intensities. In the case of the trialanine structures, agreement was generally quite poor, an aspect of the calculations that received critical discussion in the text. While the PW91 calculations in the HMX calculations were reasonable relative to other functionals, the PW91 calculations on PETN were found to either over- or underestimate the positions of the two major features in the sub-120 cm-' region. In contrast, the BP functional reproduced both the shape and the relative positions of these features adequately enough to confidently assign the THz spectrum. While the choice of density functional has been shown to be key in the reproduction of THz spectra, there are other parameters in the calculations that warrant consideration. Among these are the choice of basis set, or the accuracy of the description of the atomic orbitals from which chemical bonding and the force constants between atoms in their vibrations are generated, and density functional integration grid size, the atomic mesh from which matrix elements in DFT are evaluated for the accurate description of electron

3

196 D. G. Allis t 3 T . M . Korter

correlation in a molecule during geometry optimization and normal mode analyses. The study presented here addresses these two parameters for HMX and PETN, building on the published functional comparisons to examine how basis set and grid size influence the simulated THz spectra, thereby completing the computational survey of these two systems.

2. Methods The experimental room temperature THz spectra of crystalline PETN (PETN I) and HMX (P-HMX) were obtained from TeraView Limited (Cambridge, UK)I3. Isolatedmolecule and solid-state DFT calculations were performed with the program DMo13 (version 3.2)143 l 5 using the DN (double numerical), DND (double numerical with d polarization), and DNP (double numerical with d and p polarization) basis sets (comparable to 6-31G, 6-31G(d), and 6-31G(d,p) Gaussian-type basis sets, re~pectively),'~ and one functional from among the available BLYP,7,l 6 BOP,17 PW9112, HCTH," BP,7, l 9 PBE,20321 RPBE,22 and VWN-BP6? 79 l 9 generalized gradient approximation (GGA) density fimctionals. Unlike many plane-wave DFT packages, DMo13 does not optimize lattice constants, meaning the molecules within the unit cell were optimized within the cell parameters specified by the room temperature X-ray diffraction studies as reported in the Cambridge Structural Database. Crystal cell parameters for PETN I are as follow: Space Group P-421~(Z = 2), a = 9.3027 A, b = 9.3027 A, c = 6.6403 A, a = 90.0', P = 90.0', y = 90.0'. 23-25 Crystal cell parameters for P-HMX are as follow: Space Group P21/c (Z = 2), a = 6.54 A, b = 11.05 A, c = 8.70 A, a = 90.0', p = 124.3', y = 90.0'26. The infrared (IR) intensities calculated in the DMo13 normal mode analyses are the result of dipole derivative calculations. Isolated-molecule dipole derivatives for each normal mode are provided by default in DMo13, but are unavailable in the solid-state calculations. The solid-state IR intensities reported in this work are calculated from the square of the change in dipole moments for the unit cell that result from atomic displacements along each normal mode coordinate (d,ddQ) using M ~ l l i k e nand ~~ Hirshfeld2' population analyses, thereby considering only the IR intensity dependence on the change in dipole moment to estimate the actual intensity.

3. Results 3.1. Adjustable Parameters in the DMof Code

The DMo13 program was the first basis set-based DFT program to perform normal mode analyses in both the solid-state and for single molecules, making it ideal for examining the affects of the inclusion of the crystal cell interactions on the predicted frequencies of molecular crystals. Other quantum chemical programs (such as Gaussian03, CrystalO6, and NWChem) are either newly capable or are developing solidstate functionality (including vibrational analyses), meaning the inclusion of crystal forces in all solid-state computational studies may soon be de rigueur. DMo13 provides the necessary means for exploring theoretical dependencies in the solid-state, and has been the mainstay in all of the solid-state THz studies performed in our group and by

4

Analysis of Solid-State Terahertz Spectra

197

many others. The description of basis sets and grid sizes will, therefore, follow naming conventions established for this program. Basis Set: The choice of basis set determines the accuracy of the electronic description used in the calculation of chemical bonds and forces between atoms in the normal mode analyses. The three basis sets used in this study are the DN, DND, and DNP numerical basis sets. An exhaustive survey of the qualities and accuracies of these basis sets is beyond the scope of this study and has been presented elsewhere. Of particular interest for the normal mode analyses performed here are the dependence of the Mulliken population analyses on the choice of basis set (see below) and the size of each basis set as it relates to the calculation of intermolecular interactions. The polarization functions included in the DND and DNP basis sets provide a larger effective orbitalspace by which longer range interactions between molecules are better described. Although studies have not been undertaken, these larger basis sets would also presumably reduce basis set superposition error in the solid-state calculations. Just as larger basis sets are considered “always better” in common single molecule calculations, the order of accuracy for the DMo13 calculations goes as DNP > DND > DN. Integration Grid Size: The accuracy of the electronic description of a molecule is dependent on integration grid quality in grid-based DFT, with the increase in computational time the trade-off with increasing accuracy that must be taken into account when performing these calculations. The DMo13 program includes five pre-defined grid sizes (XCoarse, Coarse, Medium, Fine, XFine), of which the lowest-quality (XCoarse, Coarse) options were not used in the performed THz simulations. The DFT grid in DMo13 is defined using spherical coordinates to define both the extent of the mesh (radial distance (rmuxp) from each atom center) and the number of points generated for matrix element evaluation (grid points are generated using spherical harmonic functions, with the number of points increasing quadratically with the angular momentum (0 of the function used to generate these points. Actual fit: [Grid Points] = 0.3351[l12 + 0.5552[4 + 3.9277). The remaining three, “Medium” (rmaxp = 10.0 au (5.3 A); up to 1 = 6), “Fine” (rmaxp = 12.0 au (6.36 A); up to I = 6), and “XFine” (rmaxp = 15.0 au (7.95 A); up to 1 = 7), were found to provide a range of accuracy and computation time in the THz simulations and make up the final parameter selections in this study.

3.2. Intensity Calculations from Mulliken and Hirshfeld Charges There is no single “best way” to partition electron density in a molecule. The utility of a charge partitioning scheme is largely dependent on the molecular property being considered or, more specifically, the ability of the population analysis to explain a particular trend in reactivity, intermolecular interaction strength, etc. Because of this, many quantum chemical programs offer multiple charge partitioning methods. Much has been written on the advantages and limitations of various partitioning schemes29and the details are not provided here. DMo13 provides Mulliken and Hirshfeld electron densities. Mulliken population analyses are the oldest and most widely known of the partitioning approaches. Their primary advantage is their commonality to all programs. Among their known limitations are the considerable basis set-dependencies found among Mulliken population analyses for molecules with otherwise identical parameters. In general use, population analysis comparisons between calculations not employing identical

5

198 D. G. Allis B T. M .

KOTteT

parameters are immediately suspect. In the case of THz reproduction, the Mulliken charges are being used to calculate both static dipole moments for the crystal cells and dipole changes for the calculation of intensity. The differences between theory levels are important, but the primary focus of the analysis is the reproduction of the experimental THz spectra, meaning individual differences in charge partitioning at the atomic level will not be considered in significant detail in the section examining the simulation results. The second partitioning method available in DMo13 is that proposed by Hirshfeld, which defines electron density relative to the difference between natural atomic electron densities and the deformed charge density resulting from chemical bonding. The magnitude of charge differences is typically smaller in Hirshfeld analyses and is largely basis set-independent. As with the Mulliken electron density results, the use of the Hirshfeld charges will be gauged relative to the reproduction of THz spectra from the calculation of difference-dipoles. 3.3. A Review of the Functional Comparisons The results of the functional comparisons for the solid-state normal mode analyses of HMX and PETN are summarized in Figure 2. For the P-HMX crystal cell, the VWN-BP and BP density functionals were found to adequately reproduce the observed THz spectrum (the VWN-BP/DNP calculations were favored due to better agreement with calculated intensities), accounting for both peak positions and relative intensities. Agreement among the other six functionals varied in position (HCTH) and relative intensities of peaks associated with the major THz features (RPBE, PW91, PBE, BOP, BLYP). The PETN functional study showed a much larger range of agreement with experiment, with the BP simulation providing the best agreement in peak position and relative intensity with the two major features in the measured spectrum. The remaining functionals resulted in simulations either disagreeing in the positions of the simulated THz features (RPBE, PW91, PBE) or disagreement in both position and intensity (VWNBP, HCTH, BOP, BLYP). In many of these cases, some straightforward post-processing of the calculated results could bring the simulations into correspondence with the experimental results, using those calculated peaks with large intensities as a guide to direct frequency shifts. The goal of these two functional studies was not the identification of scaling methods to achieve agreement, but the identification of a functional that would provide intrinsic agreement for mode assignments. In effect, the functional studies revealed that a form of “coarse adjustment” exists for the calculations by the choice of functional. There exist no exhaustive comparisons of functionals, their performances, and the mathematical foundations for their observed differences in vibrational predictions for DMo13. As solidstate DFT is a very new functionality in the quantum chemistry toolbox, such studies are still ongoing. The HMX and PETN studies indicated that the differences in functional performance are important factors to consider, and the results are driving further studies.

6

3JE Ejpsds ZHX lEiirauiuadxg -f( suuoj ]BjsAjo 3m joj

ro o

U/YVOIJS OZl °> OZ uio-g (pajoauo ui jijgu) i Nigj pUB (pajuauo ui apoui IBULIOU jjs[a/[[Buoipunj] VOO 3)Eis-p;ios jt(3i3 aijx

HMX Intensities

"D T3

ffl |§ 3 cr Q

f8

PETN Intensities

a^n^g-p^og /o sisfyvny

66T

200

D. G. Allis & T . M . Korter

3.4.

A Review of the Isolated-Molecule vs. Solid-Stute Normal Mode Frequencies

The correlation of calculated solid-state vibrational modes to their isolated-molecule origins in the HMX and PETN studies revealed the extent to which the crystal cell interactions and the relative motions between molecules contributed to the observed structure of the solid-state THz measurements. These results for the VWN-BPDNP calculations of HMX and the BPiDNP calculations of PETN are summarized in Figures 3 (HMX) and 4 (PETN). The number assignments for the isolated-molecule calculations (top plot in each) are the IR-active mode numbers as they are obtained from the calculations (those in the 0 to 120 cm-' range). In HMX, modes 3 and 4 are IR-inactive in the VWN-BP/DNP calculations. In PETN, modes 1 and 7 are IR-inactive in the BPiDNP calculations. The Hirshfeld (middle) and Mulliken (bottom) plots show the calculated solid-state vibration positions (narrow lines), their simulated broadening using Lorentzian line shapes (5 cm-'), and the assignment of mode combinations (with Z = 2 for each crystal cell, each single molecule vibration occurs twice in the solid-state, with the two vibrations occurring as in-phase (,'+") and out-of-phase ("-") linear combinations of the two isolated-molecule modes. In PETN, a number of degenerate low-frequency modes exist that complicate the assignment of solid-state modes to linear combinations. In the plots, these are referred to in quotes). In the solid-state, the rotational and translational modes that are neglected in isolated-molecule (gas phase) calculations appear as external vibrational modes between molecules. The unlabeled peaks in the HMX and PETN spectra correspond to these relative motions (see Refs. 3 and 5 for more information). In HMX, these external modes are intermixed with internal mode combinations in the sub-120 cm-' region, providing additional structure to the THz simulations. In PETN, the external modes all appear in the 40 to 60 cm-' region and are responsible for the first major THz feature in its entirety. An isolated-molecule assignment of the PETN spectrum would fail to correctly account for this feature, which is of significance given that the isolated-molecule PETN calculations at the same level of theory as the solid-state calculations predict that the first nine vibrational isolatedmolecule modes occur in this region. The functional dependencies and the normal mode assignments together demonstrated the utility of solid-state DFT for predicting and assigning THz spectra while also revealing that the accuracy of the results is dependent on the functional. Another important result to come from the PETN study was the identification of significant differences in the calculated Mulliken and Hirshfeld intensities despite the use of identical program parameters, a clear indicator of the differences that exist between the various charge partitioning methods. These two computational studies did not address either the basis set or the integration grid size dependencies. The findings of these additional studies are provided below. 3.5. Basis Set and Integration Grid Size Dependencies

The simulation of the THz spectra for HMX and PETN are considered below with variation in the choice of basis set and grid size, leading to nine total combinations in each case. All calculations employ the density functional found to best reproduce the experimental spectra (VWN-BP for HMX, BP for PETN). This focus on individual functionals does ignore the possibility that other combinations of basis set or grid size, combined with a different functional, would also yield the level of agreement found for

8

Analysis of Solid-state Teruhertz Spectra

201

--Intensity PETN THz Spectrum (kmlmol) -- Mulliken Intensity (Debye) Hirshfeld Intensity(Debyet)

20

40

60

80

100

120

wavenumber (cm-3) Figure 3. The isolated-molecule (top) and solid-state Hirshfeld (middle) and Mulliken (bottom) simulated THz spectra of HMX. Solid-state mode labels identify in- and out-of-phase ("+" and "-",respectively) combinations of the identified isolated-molecule modes. Remaining unlabeled modes are external modes between molecules in the B-HMX crystal cell.

the original VWN-BP/DNP and BP/DNP simulations. The results of these basis set and integration grid studies do, in part, address the possibility that variation in THz simulation with choice of either leads to changes large enough to warrant further studies across the range of available density functionals. 3.6. Analysis of the P H M X Calculations The simulated spectra for the solid-state P-HMX THz measurement are provided in Figure 5 . The vibrational modes for all calculations are provided with normalized Mulliken and Hirshfeld intensities in Table 1. The agreement with the experimental THz spectrum as a function of basis set and grid size is found to be more subtle than the functional comparisons, with a number of clear trends and at least on unexpected result appearing. Quantitative similarities are found in the calculations employing the DN basis sets, with the calculated normal modes grouping in the Lorentzian fits at approx. 66 cm-' and 100 cm-'. Both the Hirshfeld and Mulliken intensities are similar throughout. While

9

202

D. G. Allis & T. M . K o r t e r

--

PETN THz Spectrum -Intensity (km/mol) Mulliken Intensity (Debye2) Hirshfeld Intensity (Debyes)

0

20

40

60

80

100

120

wavenumber (cm-I) Figure 4. The isolated-molecule (top) and solid-state Hirshfeld (middle) and Mulliken (bottom) simulated THz spectra of PETN. Solid-state mode labels identify in- and out-of-phase ("+" and "-", respectively) combinations of the identified isolated-molecule modes. Remaining unlabeled modes are external modes between molecules in the PETN I crystal cell. Modes identified as "3,4" and "8,Y are combinations of these degenerate modes. See text.

in agreement with one another, the simulated spectra do not accurately reproduce the structure of the THz spectrum, with only the most intense peak in the THz spectrum being accounted for throughout. The remaining two THz features are either overestimated in peak position (59 cm-') or greatly underestimated in intensity (82 cm-'). The results of the VWN-BPIDNP calculations are in much better agreement overall. The VWN-BP/DNP calculations with the Fine grid size, used in the original study to assign the THz spectrum, reproduce the splitting of the major THz features at 80 and 95 cm-', account for intensity in the first major THz peak at 45 cm-', and even account for shoulder structure on the experimental spectrum. The variation with choice of grid size is greater than observed for the DN calculations. The DNP/Medium spectrum shares similarities with the DN results, including a lack of splitting in the most intense calculated modes with the employed 5 cm" Lorentzian line shapes. The

10

Analysis of Solid-state Terahertz Spectra VWN-BP/DNP

20

,rl

e5

203

VWN-BPIDNP

80

1M

120 0

80

1w

120

M

40

80

80

IW

120

VWN-BPIDND

20

40

63

wavenumber (cm-I) Figure 5 . The 0 to 120 cm-' simulated Hirshfeld (red) and Mulliken (blue) solid-state THz spectra of P-HMX using the VWN-BP density functional and choice of basis set (DNP, DND, DN) and grid size (Medium Fine, XFine). The experimental THz P-HMX spectrum is shown in grey.

DNP/XFine calculations fare better for the Hirshfeld analysis, with the Mulliken intensities underestimating both the first and third THz peaks. The VWN-BP/DND calculations show considerable differences with choice of grid size, a surprising result that is also produced in the PETN calculations. While the DNDiMedium and DND/XFine calculations are similar to the DNP/Medium and DNP/XFine results, respectively, the DND/Fine simulated spectra differ considerably from all other simulated spectra both in terms of peak positions and calculated intensities.

11

204

I). G. Allis B T. M . Korter

Table 1. Solid-state P-HMX vibrational modes (between 0 and 120 cm-') and normalized Mulliken (M) and Hirshfeld (H) intensities using the VWN-BP density functional with the DNPIDNDIDN basis sets and XFineiFineIMedium integration grid sizes. DNP/XFine H em-' M

DNP/Fine em-' M H

DNP/Medium em-' M H

48.4 64.7 78.9 87.2 87.5 100.2 100.2 117.4

41.5 60.7 78.0 85.5 86.2 95.4 96.4 117.6

34.1 66.8 79.8 89.8 90.0 96.9 99.1 119.5

0.01 0.03 0.17 0.80 0.01 0.23 0.15 1.00

0.05 0.40 0.38 0.18 0.68 0.67 0.15 1.00

0.03 0.11 0.12 0.37 0.06 0.59 0.02 1.00

0.01 0.38 0.29 0.77 0.18 0.75 0.16 1.00

0.00 0.03 0.10 0.75 0.05 0.00 0.08 1.00

0.00 0.41 0.37 0.33 0.58 0.52 0.24 1.00

DND/XFine em-' M H

DND/Fine em-' M H

49.7 66.1 79.0 87.8 89.3 100.5 101.4 119.9

47.5 64.4 64.9 72.4 82.2 92.0 98.3 105.7 114.0

DND/Medium em-' M H

DN/XFine cmd M H

em-'

M

H

42.6 69.8 82.0 92.7 93.3 100.0 101.5

55.6 67.7 80.3 91.8 92.3 95.5 99.7

57.0 65.8 81.1 91.3 91.4 94.0 99.0

0.00 0.01 0.03 1.00 0.09 0.22 0.53

0.00 0.20 0.10 0.70 1.00 0.18 0.25

0.01 0.06 0.22 1.00 0.00 0.01 0.04

0.01 0.99 0.67 0.27 0.90 1.00 0.29

'

0.05 0.27 0.35 0.48 0.54 0.27 1.00

DN/Fine

0.01 0.15 0.10 1.00 0.16 0.57 0.16

0.01 0.00 0.31 0.84 0.05 0.26 0.23 1.00

0.03 0.45 0.37 0.11 0.61 0.16 0.68 1.00

0.01 0.01 0.78 0.38 0.06 0.12 0.09 0.18 1.00

0.00 0.49 1.00 0.41 0.34 0.63 0.93 0.86 0.49

DN/Medium cm-' M H 55.1 67.3 68.4 90.1 94.1 95.6 98.1

0.00 0.29 0.09 0.97 0.20 0.38 1.00

0.00 0.20 0.09 0.16 1.00 0.85 0.25

The disagreement in the Hirshfeld and Mulliken intensities in the 80 to 120 cm-' region are quite unexpected, indicating a clear lack of correspondence in charge distribution changes for the same atomic displacements. Both peak splittings and relative intensities are in disagreement with experiment, yielding simulated spectra quantitatively dissimilar to the experimental measurement. 3.7. Analysis of the PETN I Calculations

The simulated spectra for the solid-state PETN I THz measurement are provided in Figure 6. The vibrational modes for all calculations are provided with normalized Mulliken and Hirshfeld intensities in Table 2. The trends are similar in some respects to the HMX results, with the increased number of vibrational modes in the PETN calculations yielding more structure in the simulations to account for. The Hirshfeld intensities across all but the BPIDNPiFine simulation are in reasonable agreement with one another, with the intensities generated from the Mulliken dipole calculations varying considerably across different methods but, ultimately, yielding the simulation in better agreement with experiment. The internal modes that account for the remainder of the vibrational structure vary considerably in the nine calculations and are the focus of the specific analyses below. The BPDN simulations are demonstrative of the overall disagreement that can result from each piece of the simulation not accurately reproducing observed features. While the Hirshfeld intensities significantly underestimate the intensities of the internal molecular modes in the 80 to 120 cm-' region, the ratios of the Mulliken intensities of the 67 and 96 cm-' THz features are similar to the experimental results. In the case of the BPiDNiMedium and BP/DN/Fine results, the primary disagreement with the 96 cm-' experimental peak lies in the calculated splitting of the vibrational modes that combine to account for this feature. In the BNIDNiXFine results, both the splitting and downfield

12

Analysis of Solid-State Terahertz Spectra

0

m

40

Bo

80

100

1230

m

m

40

40

80

80

loo

60

eQ

100

,200

20

40

8c

eQ

tw

205

120

I20

wavenu mber (cm-1) Figure 6 . The 0 to 120 cm-' simulated Hirshfeld (red) and Mulliken (blue) solid-state THz spectra of PETN I using the BP density functional and choice of basis set (DNP, DND, DN) and grid size (Medium Fine, XFine). The experimental THz PETN I spectrum is shown in grey.

shift of the two internal mode features lead to reduced agreement with experiment, yielding a single peak at 70 cm-' in good agreement with the major THz feature and a prominent shoulder feature on the calculated peak at 77 cm-'. While the relative intensities in the DNP calculations do vary as a function of both grid size and charge partitioning method, the results from these calculations are the most accurate from among the parameter choices. In these calculations, the Mulliken intensity plots yield simulated spectra that show the correct grouping of vibrational modes to the experimental features and reasonable relative intensities. The peak groupings of the

13

206

D. G. Allis & T. M. Korter

Table 2. Solid-state PETN I vibrational modes (between 0 and 120 cm-') and normalized Mulliken (M) and Hirshfeld (H) intensities using the BP density functional with the DNP/DND/DN basis sets and XFineiFineiMedium integration grid sizes. DNPiXFine cm-' M H 37.0 0.0 0.4 38.3 0.0 0.0 57.0 1.0 1.0 57.8 0.3 0.0 62.9 0.0 0.0 63.6 0.2 0.0 70.6 0.0 0.0 73.9 0.2 0.0 81.5 0.5 0.0 84.8 0.0 0.0 85.6 0.1 0.0 92.9 0.1 0.1 93.7 0.1 0.0 93.9 0.0 0.0 97.9 0.0 0.0 99.3 0.1 0.1 101.1 0.6 0.0 104.4 0.0 0.0 108.1 0.0 0.0 109.5 0.4 0.1

DNPiFine cm-l M 39.4 0.0 39.7 0.1 60.4 0.2 1.0 60.7 0.0 62.2 73.0 0.1 73.1 0.0 81.4 0.0 0.1 85.3 0.0 85.4 93.2 0.0 97.6 0.2 0.1 97.6 0.1 97.8 100.6 0.1 100.8 0.2 105.0 0.0 111.2 0.0 111.4 0.1

DNDiMedium cm-' M H 27.7 0.0 0.0 28.1 0.1 0.4 0.0 0.0 56.2 61.1 0.0 1.0 63.8 1.0 0.3 67.7 0.0 0.2 71.5 0.5 0.1 82.9 0.3 0.0 88.4 0.0 0.0 88.6 0.0 0.0 96.5 0.0 0.0 97.2 0.0 0.0 97.2 0.2 0.0 98.0 0.0 0.0 107.1 0.0 0.1 110.4 0.0 0.0 111.7 0.1 0.0 115.8 0.0 0.1 119.0 0.3 0.0

H 0.2 0.0 0.0 1.0 0.0 0.1 0.1 0.0 0.0 0.0 0.0 0.1 0.0 0.0 0.1 0.0 0.0 0.0 0.0

DNPiMedium M H cm-l 27.8 0.0 0.0 31.5 0.0 0.2 53.6 0.0 0.0 0.0 0.6 59.7 1.0 1.0 61.8 68.0 0.0 0.0 69.7 0.1 0.1 81.2 0.1 0.0 87.4 0.1 0.0 87.4 0.0 0.0 94.6 0.1 0.0 94.8 0.0 0.0 94.9 0.2 0.1 96.1 0.0 0.0 104.9 0.0 0.0 108.6 0.1 0.0 109.6 0.3 0.0 114.9 0.1 0.0 118.0 0.2 0.0

DNiXFine cm-' M 39.1 0.1 0.2 39.3 67.2 0.0 0.0 69.2 1.0 69.4 0.2 69.6 76.9 0.8 77.2 0.0 89.0 0.0 91.4 0.6 91.5 0.2 95.5 0.2 0.0 95.6 0.2 97.5 0.1 98.8 106.0 0.2 106.1 0.0 113.7 0.1 115.4 0.1 115.4 0.1

H 0.0 0.3 0.0 0.0 1.0 0.0 0.2 0.1 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.1 0.0 0.0 0.0 0.1

DNDlXFine cm-l M H 37.2 0.2 0.3 38.0 0.6 0.0 57.3 0.0 1.0 58.2 1.0 0.0 63.5 0.2 0.0 71.2 0.4 0.0 73.8 1.0 0.1 81.9 0.0 0.0 85.4 0.1 0.0 86.7 0.7 0.1 94.9 0.6 0.1 95.0 0.2 0.0 95.6 0.7 0.0 99.4 0.2 0.0 100.6 0.6 0.1 103.3 0.0 0.0 106.9 0.3 0.0 109.2 0.1 0.0 110.7 0.5 0.0 117.4 0.1 0.0

DNiFine cm-l M 43.0 0.0 43.2 0.0 73.1 0.1 73.4 1.0 0.0 73.5 79.6 0.1 87.8 0.1 87.9 0.5 91.3 0.3 96.0 0.2 96.1 0.0 99.8 0.0 104.4 0.0 105.3 0.1 105.3 0.1 113.4 0.5 113.5 0.2

H 0.1 0.4 1.0 0.7 0.1 0.0 0.1 0.2 0.0 0.0 0.0 0.0 0.0 0.1 0.0 0.4 0.0

DNDiFine cm-' M 39.2 0.0 39.5 0.2 59.3 0.8 62.4 0.5 64.6 0.0 73.0 0.1 76.2 1.0 83.0 0.2 86.7 0.0 88.4 0.1 96.1 0.0 98.1 0.3 99.5 0.2 100.6 0.2 103.0 0.4 103.4 0.3 106.8 0.2 112.1 0.1 112.6 0.0

H 1.0 0.1 0.7 0.0 0.0 0.2 0.4 0.0 0.1 0.1 0.0 0.2 0.0 0.1 0.5 0.1 0.2 0.2 0.1

DNiMedium cm-' M H 29.1 0.0 0.0 38.3 0.0 0.2 70.9 0.6 1.0 71.3 1.0 0.1 79.3 0.0 0.0 88.2 0.1 0.1 89.0 0.2 0.0 91.9 0.1 0.0 93.8 0.5 0.0 94.0 0.0 0.0 95.4 0.0 0.0 103.7 0.3 0.0 105.0 0.0 0.0 109.4 0.3 0.0 113.0 0.6 0.0

BP/DNP/Fine calculations are best among all calculations. The BP/DNP/XFine Mulliken intensities are better overall, but do not reproduce the grouping of the internal mode features at 96 cm-' as well as the DNPiFine results. The BPIDNPiMedium Mulliken intensity results are similar to those of the DNPXFine results, with the inaccurate grouping of the internal modes reducing the agreement with the shape of the 96 cm-' THz feature.

14

Analysis of Solid-State Terahertz Spectra

207

As was also the case of the HMX analyses, the Mulliken and Hirshfeld PETN agreement varies most in the BPiDND calculations. In the BP/DND/XFine calculations, the Hirshfeld and Mulliken intensities are very different, with the Mulliken plots reproducing the observed peak intensities very well but not reproducing the grouping of the external modes at 67 cm-’ as accurately as the other calculations. This same peak splitting of the external modes occurs in the BP/DND/Fine calculations, reducing the simulation accuracy of an otherwise excellent Mulliken intensity agreement with experiment. The BP/DND/Medium calculations are similar to the DN/Medium and

DNPiMedium results both in terms of the underestimation of internal mode (96 cm-’) peak intensities and overestimation of group mode splittings. 4. Discussion

The HMX and PETN THz spectral reproductions using solid-state DFT are dependent on a number of factors, including the electronic description (basis set choice), the calculated energies of vibrational modes at the various levels of theory, the choice of charge partitioning used for the difference-dipole calculations, the relative atomic displacements in each mode that depend on crystal packing interactions, and the accuracy of the difference-dipole method itself for reproducing IR intensities. To these factors are included the choice of functional, as addressed in the previous reports on these two systems. Aspects of these variables (neglecting functional choice) are discussed below. 4.1. Basis Set and Integration Grid Dependencies

The theoretical THz simulations performed for HMX and PETN are, perhaps, nonstandard in their use. The goal of the simulations was the best possible reproduction of the experimental spectra. This does not necessarily require the use of the highest available level of theory and, given the many variables that contribute to the production of these spectra, may, in fact, benefit more from the use of intermediate levels of theory. This notion of “best theory for the application” has some foundation in the empirical nature of density functional development in DFT. In the geometry analyses of both HMX and PETN, the HCTH density functional was found to best reproduce the molecular bond lengths, but was clearly ineffective as a functional for THz simulations. In the HMX and PETN THz simulations, the calculated intensities and peak positions were found to vary as a function of basis set and integration grid size, with the simulations using the DNP basis set and Fine grid size providing the best reproductions of the experimental spectra. Based only on the reproduction of the THz spectra, the Fine and XFine grid sizes both offered good agreement, but revealed limited accuracy in either the peak positions (XFine) or the relative intensities obtained in the difference-dipole calculations (Fine). No single calculation provided ideal agreement, but the various simulations revealed that higher quality basis sets and the higher quality grid sizes do perform better at THz simulations. While more work is required to confirm their abilities to reproduce THz spectra in other systems, the choices of the DNP basis set and Fine grid size appear to be adequate for THz simulations and mode assignments based on the calculated agreement with experiment. The VWN-BP functional was found to be superior in both peak position and intensity calculations for HMX, but the BP density functional provides the best agreement with experiment with HMX and PETN taken together. The differences between VWN-BP and BP for HMX were small enough that a subsequent vibrational mode analysis with the BP density functional is not warranted,

15

208

D. G. Allis 6 3 T. M . K o r t e r

especially in light of the interest in reproducing the measured THz spectrum and not the use of the single test case, in the absence of many other systems to consider in such analyses, for performing a benchmark-type study with the detail performed here. With the higher-level calculations performed on these systems, the choice of population analysis was found, specifically for the PETN THz simulations, to play a critical role in the reproduction of the experimental spectra. 4.2. Mulliken and Hirshfeld Charge Dependencies

The variety in calculated intensities from Hirshfeld and Mulliken charges demonstrate the differences in these two charge partitioning methods. More important, the calculated differences demonstrate how the utility of each charge partitioning method is dependent on the property being calculated. While the Hirshfeld charges are commonly considered to be superior for describing chemical structure and reactivity, their use across all calculations in the difference-dipole analyses show that their basis set-independent quality leads to similar spectral reproductions in many cases that do not agree with the THz measurements. The considerable differences in vibrational frequencies across different basis sets and grid sizes does, in some way, lead one to the conclusion that the lack of variation in the Hirshfeld charge partitioning is a limitation of the method for dipole calculations in the solid-state. Mulliken difference-dipole calculations, despite their obvious dependence on the basis set, do yield reasonable simulations of THz spectra in HMX and PETN provided an adequate electronic description (basis set) is used in the normal mode analysis. As a further reinforcement of the charge partitioning differences, the Mulliken and Hirshfeld charges for the minimum energy symmetry-unique atoms in the solid-state calculations of HMX and PETN are provided in Figure 7. While the Hirshfeld charges are effectively constant throughout, the effects of polarization function inclusion in the basis sets are significant in the Mulliken charge analyses, with the addition of these functions increasing the electron density differences between the N and 0 atoms in the NOz groups of both molecules (with the DND and DNP basis sets) and the inclusion of p polarization functions on the H atoms (DNP) in both molecules resulting in significant electron density shifts away from the H-bound carbon atoms. 4.3. Timings

The timings for single processor geometry optimizations and normal mode analyses of HMX and PETN are summarized in Figure 8. While the computational expense should not be the deciding factor in the accurate assignment of THz spectra, the HMX and PETN crystal cells are very small systems on which to perform DFT calculations. The extension of the solid-state assignment methods described above to larger systems, such as the high explosives RDX and TNT, constitute significant increases in the number

or crystal cell atoms and, consequently, computational cost. The trends for HMX and PETN show some variety in calculation time due largely to longer geometry optimization processes (the PETN BPiDNDIMedium calculations required significantly more steps than the other BP/DND calculations to find a stable minimum), but the overall increases in compute time with increasing basis set quality and integration grid size are clear. The choice of integration grid was found to be most significant, leading to, for instance, a near-doubling in compute time for each step in grid quality for the DNP calculations. Among all, it is found that the DNPiFine calculations offer reasonable compute times and, importantly, adequate agreement with experiment for performing THz assignments.

16

Analysis of Solid-state Ternhertz Spectra

209

Figure 7. Variation in Mullibn (upper plots) and Hirshfeld (lower plots) population analyses with basis set and grid size for the individual molecules in the P-HbLX (top, VWN-BP functional) and PETN I (bottom, BE' functional) crystal cells. Atom labels correspond to those in Figure 1.

DNPl Medium

DNPI Fine

DNPI DNDI DNDI XFine Medium Fine

DNDl DN/ XFine Medium

BN/ Fine

BNI XFine

Figure 8. Single-processor compute times (in minutes) for the HMX and PETN crystal cells as a function of basis set and grid size. Calculations are for an AMD Opteron (280) workstation in single-processor mode rnnning the SuseLinux 10.0 operating system.

17

210

D. G. AZZis B T. M . Korter

5. Conclusions

The computational factors influencing the theoretical reproduction of the THz spectra of the solid-state explosives HMX and PETN have been examined. Specifically, the choice of basis set and integration grid size were explored to determine their importance in the accurate simulation of the observed experimental THz spectra. This study of computational parameter space builds on previous studies of these two molecules that considered only the simulation dependencies on the choice of density functional. This investigation reveals that the two considered experimental spectra are best reproduced using the DNP basis set and Fine integration grid size option (program options specific to the DMo13 program) which are the settings employed in the original functional comparisons. Perhaps the most significant finding in this study was the determination of the most appropriate mechanism for the calculation of THz absorption intensities. The spectral intensities in the THz simulations are found to be reproduced best using difference-dipole calculated intensities utilizing atomic charges derived from Mulliken population analyses. The calculated charges from Mulliken analyses are found to be basis set-dependent while in marked contrast, the Hirshfeld population analyses were found to be insensitive to changes in basis set across all calculations. Overall, the best THz intensity reproductions were found using Mulliken atomic charges with the DNP basis set. The Hirshfeld method was reasonably accurate for difference-dipole calculations of HMX, but not PETN. Variations in integration grid size made little difference in the spectra simulations and both the Fine and XFine values provided for adequate reproduction of the HMX and PETN THz spectra when used with the DNP basis set. This extensive study of the computational parameters in the simulation of solid-state THz spectra by DMo13 leads us to recommend the BP fimctional, DNP basis set, Fine integration grid size, and Mulliken population analyses for THz absorption intensities. The completed analyses of functional, basis set, and grid size are to be used as guides in subsequent simulations of the THz spectra of molecular crystals as a means for both spectral assignments and further tests of the above results.

6. Acknowledgments We acknowledge the support of the National Science Foundation (PHY-0442188) and the Intelligence Community (IC) Postdoctoral Research Fellowship Program. The authors thank TeraView Limited (Cambridge, UK) for providing the experimental THz spectra and Ms. Darya A. Prokhorova for assistance with manuscript preparation.

References 1. Kemp, M. C.; Taday, P. F.; Cole, B. E.; Cluff, J. A,; Fitzgerald, A. J.; Tribe, W. R., Security applications of terahertz technology. Proceedings of SPIE-The International Society for Optical Engineering 2003, 5070,44. 2. Taday, P. F., Applications of terahertz spectroscopy to pharmaceutical sciences. Philosophical Transactions of the Royal Society of London, Series A: Mathematical, Physical and Engineering Sciences 2004, 362, (1815), 351-364.

18

Analysis of Solid-State Terahertz Spectra

211

3. Allis, D. G.; Prokhorova, D. A.; Korter, T. M., Solid-state Modeling of the Terahertz Spectrum of the High Explosive HMX. Journal ofPhysical Chemistry A 2006, 110, ( 5 ) , 1951-1959. 4. Allis, D. G.; Prokhorova, D. A,; Fedor, A. M.; Korter, T. M., First principles analysis of the terahertz spectrum of PETN. Proceedings of SPIE-The International Society for Optical Engineering 2006, 6212, (Terahertz for Military and Security Applications IV), 62120Fil62120Fi11. 5. Allis, D. G.; Korter, T. M., Theoretical Analysis of the Terahertz Spectrum of the High Explosive PETN. Chem. Phys. Chem. 2006, submitted. 6. Vosko, S. H.; Wilk, L.; Nusair, M., Accurate spin-dependent electron liquid correlation energies for local spin density calculations: a critical analysis. Canadian Journal of Physics 1980, 58, (8), 1200-11. 7. Becke, A. D., Density-functional exchange-energy approximation with correct asymptotic behavior. Physical Review A: Atomic, Molecular, and Optical Physics 1988, 38, (6), 3098-100. 8. Perdew, J. P.; Wang, Y., Accurate and simple density functional for the electronic exchange energy: Generalized gradient approximation. Physical Review B: Condensed Matter and Materials Physics 1986,33, 8800. 9. Humphrey, W.; Dalke, A,; Schulten, K., VMD: visual molecular dynamics. Journal of molecular graphics 1996, 14, (I), 33-8,274. 10. Franzen, S., Use of Periodic Boundary Conditions To Calculate Accurate Beta-Sheet Frequencies Using Density Functional Theory. Journal of Physical Chemistry A 2003, 107, (46), 9898-9902. 11. Siegrist, K.; Bucher, C. R.; Mandelbaum, I.; Walker, A. R. H.; Balu, R.; Gregurick, S. K.; Plusquellic, D. F., High-Resolution Terahertz Spectroscopy of Crystalline Trialanine: Extreme Sensitivity to b-Sheet Structure and Cocrystallized Water. Journal of the American Chemical Society 2006, 128, (17), 5764-5775. 12. Perdew, J. P.; Wang, Y., Accurate and simple analytic representation of the electron-gas correlation energy. Physical Review B: Condensed Matter and Materials Physics 1992,45, 13244. 13. Lo, T.; Gregory, I. S . ; Baker, C.; Taday, P. F.; Tribe, W. R.; Kemp, M. C., The very far-infrared spectra of energetic materials and possible confusion materials using terahertz pulsed spectroscopy. Vibrational Spectroscopy in press. 14. Delley, B., An all-electron numerical method for solving the local density functional for polyatomic molecules. Journal of Chemical Physics 1990, 92, (I), 508-17. 15. Delley, B., From molecules to solids with the DMol3 approach. Journal of Chemical Physics 2000, 113, (18), 7756-7764. 16. Lee, C.; Yang, W.; Parr, R. G., Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density. Physical Review B: Condensed Matter and Materials Physics 1988, 37, (2), 785-9. 17. Tsuneda, T.; Suzumura, T.; Hirao, K., A new one-parameter progressive Colle-Salvetti-type correlation functional. Journal of Chemical Physics 1999, 110, (22), 10664-10678. 18. Boese, A. D.; Handy, N. C., A new parametrization of exchange-correlation generalized gradient approximation functionals. Journal of Chemical Physics 2001, 1 14, (13), 5497-5503. 19. Perdew, J. P.; Wang, Y., Physical Review B: Condensed Matter and Materials Physics 1986, 33, 8800. 20. Perdew, J. P.; Burke, K.; Ernzerhof, M., Generalized gradient approximation made simple. Physical Review Letters 1996,77, (18), 3865-3868. 21. Perdew, J. P.; Burke, K.; Emzerhof, M., Generalized gradient approximation made simple. [Erratum to document cited in CA126:51093]. Physical Review Letters 1997, 78, (7), 1396. 22. Hammer, B.; Hansen, L. B.; Norskov, J. K., Improved adsorption energetics within densityfunctional theory using revised Perdew-Burke-Emzerhof functionals. Physical Review B: Condensed Matter and Materials Physics 1999, 59, (1 I), 7413-7421.

19

212

D. G. Allis & T. M . Korter

23. Booth, A. D.; Llewellyn, F. J., The crystal structure of pentaerythritol tetranitrate. Journal cf the Chemical Society, Abstracts 1947, 837-46. 24. Cady, H. H.; Larson, A. C., Fentaerythritol tetranitrate 11. Its crystal structure and transformation to PETN I. Algorithm for refinement of crystal structures with poor data. Acta Crystallographica, Section B: Structural Crystallography and Crystal Chemistry 1975, B3 1, (7), 1864-9. 25. Trotter, J., Bond lengths and angles in pentaerythritol tetranitrate. Acts Cryst. 1963, 16, (Pt. 7), 698-9. 26. Choi, C. S.; Boutin, H. P., Study of the crystal structure of bcyclotetramethyleneteetranitramineby neutron diffraction. Acta Crystallographica, Section B: Structural Crystallography and Crystal Chemistry 1970,26, (Pt. 9), 1235-40. 27. Mulliken, R. S., Electronic population analysis on LCAO-MO [linear combination of atomic orbital-molecular orbital] molecular wave functions. I. Journal of Chemical Physics 1955, 23, 1833-40. 28. Hirshfeld, F. L., Bonded-atom fragments for describing molecular charge densities. Theoretica Chimica Acta 1977,44, (2), 129-38. 29. Fonseca Guerra, C.; Handgraaf, J.-W.; Baerends Evert, J.; Bickelhaupt, F. M., Voronoi deformation density (VDD) charges: Assessment of the Mulliken, Bader, Hirshfeld, Weinhold, and VDD methods for charge analysis. Journal of Computational Chemistry 2004,25, (2), 189-210.

20

International Journal of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 213-224 @ World Scientific Publishing Company

World Scientific www.worldsciefltific.com

FIRE DAMAGE ON CARBON FIBER MATERIALS CHARACTERIZED BY THZ WAVES NICHOLAS KARPOWICZ Department of Physics, Rensselaer Polytechnic Institute I10 Street, Troy, NY 12180 zhangxc@rpi. edu DAVID DAWES Battelle Memorial Institute 505 King Drive, Columbus, OH 43201 MARK J. PERRY Battelle Memorial Institute 505 King Drive, Columbus, OH 43201 perrymj@battelle. org X.-C. ZHANG Department of Physics, Rensselaer Polytechnic Institute I10 8IhStreet, Troy, NY I2180 [email protected] We apply THz imaging technology to evaluate fire damage to a variety of carbon fiber composite samples. The majority of carbon fiber materials have polarization-dependent reflectivities in the THz frequency range, and we show how the polarization dependence changes versus the hum damage level. Additionally, time domain information acquired through a THz time-domain spectroscopy (TDS) system provides further information with which to characterize the damage. The technology is discussed in terms of non-destructive testing applications to the defense and aerospace industries. Keywords: Terahertz; Non-destructive testing; Non-destructive evaluation; composites

1. Introduction

Imaging in the THz frequency range has been a subject receiving considerable attention in recent years, with hopes of access to new methods in diverse subjects ranging from medicine to manufacturing to security. The impetus for this interest springs from three aspects of radiation in this part of the electromagnetic spectrum: 1) The preponderance of spectroscopic signatures of large molecules lying within this range, 2) that imaging systems based on THz waves can achieve resolutions easily amenable to human interpretation (i.e., the detail acquired does not require leaps of imagination on the part of the operator in order to understand the information therein) and 3) the transparency of many commonly-used materials within this range (including plastics, fabrics and, to a degree, ceramics). The types of THz radiation used for imaging falls into two categories,

21

214

N.Karpowicz et al.

pulsed’ and continuous-wave2-6(CW). A pulsed system sends a short (typically less than one picosecond) transient onto or through a sample, and coherently records the resulting waveform, which can be analyzed in both time and frequency domains. CW imaging works with a single frequency, which in the absence of rapid tuning may inhibit spectroscopic imaging, but has the advantages that it is simpler to reduce into a compact imager and can be simpler to operate since the signal does not need to be sought in the time domain. As one of the major applications of THz imaging is in the field of non-destructive testing, we have applied it to an area of critical need: the testing of aerospace materials. Composite materials such as carbon fiber are widely used in this industry. The nature of their use requires technologies that are able to differentiate between safe and unsafe materials, either due to manufacturing tolerance or damage acquired while in use. In this paper, we discuss the applicability of terahertz (THz) imaging systems to this purpose, focusing on graphite fiber composite materials. We have previously demonstrated the use of both continuous-wave and pulsed THz imaging systems for the evaluation of space shuttle insulating foam6, which is an ideal target for such systems due to its low absorption and index of refraction in this spectral range. Carbon fiber-based materials present more of a challenge, as the fibers are conductive and therefore exhibit a high THz reflectivity. As a result, our measurements are performed in a reflection imaging geometry, which does have the advantage of more accurately simulating the type of measurement that could be performed in a real-world setting.

The material used in this study falls somewhat outside of the range of typical THz imaging subjects as it is not highly transparent in bulk form. The conducting fibers will reflect or absorb the incoming signal either immediately at the surface or within a few sample layers, depending on the polarization of the incoming waves. The usefulness of this type of information depends on the type of damage one is attempting to detect and the specifics of the structure being studied. We will show that in some cases the information provided by a THz image gives a more complete outline of the damaged area than what is apparent in a visible light image, and that time-domain processing with data from a pulsed THz system can reveal changes to the structure of the surface of the material.

2. Experimental Setups Two systems were used to study the samples in these measurements: a THz time-domain spectroscopy (TDS) system and a CW THz imager operating at 0.6 THz, both of which have been described el~ewhere”~. In summary, the THz TDS system begins with a Ti:Sapphire oscillator (Spectra-Physics Mai Tai), which produces pulses of 800 nm central wavelength, 80 fs duration, and 80 MHz repetition rate with an average power of 800 mW. The laser beam is split into pump and probe beams, the former being incident

22

Fire Damage o n Carbon Fiber Materials

215

on a p-type InAs wafer, which produces the THz pulses and the latter is recombined with the THz signal after it is focused onto and returned from the sample. The THz and probe pulses travel collinearly through a ZnTe crystal, which serves as an electro-optic sensor of the full amplitude and phase of the THz pulse. The system is shown schematically in Fig. 1.

IT0 Wallaston ZnTe QWP prism glass

detector polarizer

t

beam

translation

HWP

time delay stage Fig. 1. Schematic diagram of the THz TDS system used in the experiment.

The CW system, a diagram of which is shown in Fig. 2, consists of a Gunn diode oscillator operating at 100 GHz, which is then frequency doubled and tripled to 0.6 THz. The beam is focused by a hyperbolic polyethylene lens to a 500 pm spot, collected by the same lens after being reflected by the sample and is measured by a Golay cell. The sample is mounted on a pair of translation stages and moved to produce the image.

23

216

N . Karpowicz et al.

600 GHz Gunn diode assembly

translation stages Fig. 2. Schematic diagram of the continuous-wave THz imaging system used in the experiment

3. Continuous-wave imaging

The CW images reveal features that are important for useful application: damaged areas show a strong change in reflectivity, and the reflectivity is highly dependent on the polarization of the incoming radiation. The latter is quite similar to the function of a wire grid polarizer, which reflects waves whose electric field is parallel to the grid and allows those whose field is perpendicular to pass. Thus, the image with parallel electric field shows mainly information about the topmost surface, while the perpendicular field image penetrates several layers further. The reflectivity can be described as

where & describes the polarization-sensitive reflectivity of the fiber grid, and RB describes the background reflectivity of the sample, which arises from the fact that the fibers are not entirely unidirectional and are densely spaced in a dielectric resin. These two parameters can be extracted via measurements at orthogonal polarizations. Once measured, and RB give indications of the severity and nature of the damage to the sample. If there is a significant reduction of & after a sample is damaged, this indicates that there is damage to the fibers, e.g. through reorientation or a chemical change

24

Fire Damage on Carbon Fiber MateTials

resulting in lower conductivity such as residual surface char &om the matrix. reduction of RBcan attributed to damage to the resin, or increased surface roughness.

217

A

In Fig. 3, we show the results of imaging a severely damaged sample and applying this chara~te~~%ation fkamework. The four images shown are the optical image, its well as ~ ~ tthe y electric field parallel to the fibers (Q, nt;, and RB images of the ~ e ~ e c t iwith (which is equivalent to the reflectivity ~ e ~ e n d ~ c utol athe r fibers). It can be seen that the areas of change to &, and RB differ, indicating that the nature of the damage is not uniform across the damaged region. It can also be seen that the contrast in the Tlllz images is muck higher than that of the optical image (although, the resolution is lower). ~ l direction, a ~ One may also observe that when the polarization is ~ e ~ e ~ dto~thec fiber more structure is visible, due to increased penetration of the beam into the sample.

tical

Pig. 3. Imaging results for a sample of carbon fiber composite with severe bum damage (the imaged area is 28 x 31 mm). R// is formed with the THz wave polarized such that the electric field is parallel to the fibers of the material, resulting in the highest reflectivity. & and Rs are images of the parameters of Equation 1, where Rc; is the po1~zation"sensitivereflectivity of the grid of wires and RB is the background reflectivity. Rn is equal to the reflectivity obtained when the polarization of the THz wave causes the electric field to be perpendicular to the fibers.

25

218

N . Karpoiuicz et a3

An important consideration for a non-des~ctiveevaluation technique i s the rate o f false calls. This is mitigated in this imaging modality by the fact that the images provided are easily recognizable by human vision. To test the case of minor damage that does not have an effect on the strength of the material, but it still apparent in a visual inspection of the material, we imaged a lightly burned sample of material, which is shown in Fig. 4. It can be seen that the scorch mark that is apparent in the optical image does not appear in the TMz image since the burning did not affect the ~ ~ e r l ~ i i ~ g structure.

G Fig. 4. Results of a p ? l y ~ gCW imaging (in an identical matter to that of Fig. 3) to a sample with ~ n s i ~ ~ ~ ~ c a n t bum damage (the imaged area measures 80 x 35 mm). It can be seen that the scorch mark visible in the optical image (indicated by the white arrow) does not appear in the TMz images. It is noteworthy that this sample has a more complicated fiber structure than that of Fig. 3, which is apparent in the THz image, but not in the optical image since the fibers are covered by an optically-opaquedielectric layer.

In order to apply this CW imaging technique to real-world ap~lications,there are several i ~ ~co~si~erations. o ~ ~ First t is the geometry of the material being studied. In o incidence ~ ac o n d~~ ~ ~ o nIfs . these examples, a flat sample of material i s used under ~ the material is curved, or if normal incidence is impossible, a more complicated system will be required in order to ensure that the beam enters the detector. One must also know the fiber orien~a~iQn~ which for many carbon fiber materials is quite obvious, but if the

26

Fire Damage o n Carbon Fiber Materials

219

polarization is not aligned precisely parallel and perpendicular to the fiber direction, the extraction of RG and RB becomes more difficult (and impossible at a 45" polarization angle).

4. THz Time-Domain Spectroscopy Measurements

Measurements with a coherently detected THz pulse through THz time-domain spectroscopy (TDS) yield additional information about the material. Since the measurement is time-resolved, it is possible to extract the presence and depth of multiple reflections from the surface. There are numerous methods of extracting high-resolution ranging information from this data8-l2. The method utilized herein is a deconvolution of the reference waveform from the waveform acquired from the damaged region. The desired data can be pictured as a series of delta functions in the time domain corresponding to the reflection events of the incident pulse on the sample surface. In the acquired data, each of these delta functions is convolved with the THz pulse. By utilizing a reference waveform, which in this case is simply the waveform reflected by an undamaged portion of the sample or from an undamaged sample of the same composition, it is possible to remove this broadening effect and recover the positions of the reflection events. An example of THz TDS waveforms from a damaged and undamaged sample surface are shown in Fig. 5 . There is a clear double-peak structure in the waveform reflected by the damaged surface, indicating delamination. This shows that time-domain THz measurements are able to add a third dimension to imaging of carbon fiber composites, further characterizing the nature of the damage. Since TDS measurements are resolved in the time domain, it adds another dimension of sensitivity to the material geometry, however. If the distance to the target changes, the position of the pulse in the time domain will shift, and if the detection window is too small, disappear. One can simply increase the size of the time window, but this increases acquisition time, which may not be acceptable for a given application.

27

220

N . Karpowicz et al.

40 -

30

--gP

~

20-

.P 10-

6 Time delay (ps)

8

4 6 Time delay (ps)

8

10

101

-4'

0

2

Fig. 5. Time-domain waveforms taken from the material shown in Fig. 3, (a) undamaged and (h) damaged. The splitting of the peak in the time domain indicates the presence of multiple reflecting surfaces. In the case of this material, the large separation between the peaks (in this geometry, the 2.5 ps separation is equivalent to a 130 pm separation between the reflecting surfaces), indicates delamination.

To correlate the measured THz data with the strength of the damaged materials, the samples were destructively tested using a flex modulus test in accordance with American Society for Testing and Materials D790 procedures (Standard Test Methods for Flexural Properties of Unreinforced and Reinforced Plastics and Electrical Insulating Materials). While this testing method does not provide detailed information about the position of the weakened material, it allows for a rough correlation between samples that were found to have a strongly modified THz signature and flexure strength. In the flexure tests, a force is applied to the center of the sample, which is supported only on its ends. As the center

28

Fire Damage o n Carbon Fiber Materials

221

flexes out, the relationship between the force and displacement characterizes the sample; the slope of the curve indicates the rigidity of the sample (a rigid sample will have a large slope) and the maximum value of the curve indicates the overall capacity of the material. When multiple layers in a sample fail successively, jagged edges will be apparent. Examples of the data from these tests are shown in Figure 6 for two types of samples: one with a structure that fails in a single event (a through c) and one which fails layer by layer (d through f). Each sample showed evidence of delamination resulting from the fire damage. The data indicate that in cases where there is a marked decrease in the strength of the material, there is a strong difference in the THz time-domain waveform versus a reference waveform. Additionally, in cases where there were visible burn marks on the surface, but small effects in the THz data, the loss of strength was generally less severe. However, the loss of strength shown by samples of the type shown in Fig. 6 (d) through 6 (f) was relatively minor compared to the distortion of the waveform. The burned sample was heavily damaged across the entire sample surface, however the structure of the material was such that it contained very distinct layers such that when the top few layers were damaged, the rest of the structure remained relatively intact. Since the THz pulse does not penetrate deeply into the material, the information retrieved by THz TDS only indicates the condition of the top layer, to a depth of no more than several hundred microns.

29

222

N.Karpowicz e t al.

-

400

100

50-

40 -

0.2

Displacement (rnm)

0.3

50

-unburned - - - burned

(b)

0.1

"0

Displacement (mm)

40

5 30~ s

(el

- - - burned

5 30 9

7J 20

0 20~ Qa = .-2 10

a, .c

.I! 10 I-

,-.: '

+ 4 .

-10-

-200

-0.5

-0.25

0 0.25 Position (mm)

0.5

-0.5

2

-0.25

4

Time (ps)

0 0.25 Position (mm)

6

0.5

Fig. 6. (a) Load vs. displacement characteristics for a two graphite fiber composites of the same kind, one burned (shown in Fig. I (b)) and one unburned. The rising slope of the plot indicates the rigidity of the structure, while the height of the graph indicates the overall strength of the sample. Jagged edges indicate successive failures of layers within the sample. (h) THz time-domain waveforms reflected from the surface of these samples. (c) Result of a deconvolution of the wavefonn from the damaged sample with a reference waveform from the undamaged sample. (d) Load vs. displacement characteristics for a different pair of samples (not shown in Fig.1). (e) THz waveforms reflected from their surfaces. (f) Result of a deconvolution of the waveform from the second damaged sample with a reference waveform from the second undamaged sample.

Real-world application of these techniques requires speed as well as specificity. The examples shown in this paper utilize a raster scanning technique to form the images, wherein a THz beam is focused and scanned across the sample, with the detected reflected signal being recorded point by point. Such an imaging modality is time-

30

Fire Damage o n Carbon Fiber Materials

223

consuming and most-likely will be inappropriate for large-scale applications or testing where time is an issue. For these cases, “camera-like’’ operation would be much preferred. Steps have been taken in this direction, for both pulsed and CW THz techniques. At its most basic level, camera functionality requires an array of detectors and focusing elements capable of forming an image upon it. For CW THz systems, detector array options include commercially available arrays of pyroelectric detectors, which unfortunately suffer from high noise-equivalent power, and microbolometer arrays, which have been shown to produce usable, real-time imaging combined with quantum cascade 1ase1-s’~.Combined with appropriate optics, which can include systems made of reflective optics (e.g. parabolic mirrors) or lenses made of THz-transparent materials such as polyethylene, Teflon or Picarin glass, such detector arrays can be readily adapted to this application. Focal plane imaging has been demonstrated with pulsed THz systems, as well. A commonly used method is to utilize a large electro-optic crystal, such as one made of ZnTe, and an expanded optical probe The THz image projected onto the crystal will spatially modulate the polarization of the probe beam, which can be detected using polarization optics and a CCD. Such a system requires a strong THz pulse (such as those generated using an amplified Ti:Sapphire laser), which would currently prohibit field applications. Additionally, the EO crystal should be extremely uniform and have very little strain, as crystal variation and stress have a strong effect on the crystal birefringence. While it is possible to meet these requirements, such a crystal can be prohibitively expensive, especially at large sizes.

5. Conclusion

THz imaging and spectroscopy are promising solutions to the problem of identifying evaluating damage to carbon fiber composite materials. While additional progress in THz technology will be required for some applications, currently existing tools are able to provide useful and quantifiable information regarding the extent and severity of damage. Such techniques could potentially increase safety and efficiency in the defense and aerospace industries.

6. Acknowledgements The authors thank Xu Xie, Hua Zhong, Michael Shur and Nezih Pala for technical assistance. This material is based upon work supported by the National Science Foundation under Grant No. 03333 14.

31

224

N. Karpowicz et al.

References 1. B. B. Hu and M. C. Nuss, “Imaging with terahertz waves,” Opt. Lett. 20, 1716 (1995). 2. T.S. Hartwick, D. T. Hodges, D. H. Barker, and F. B. Foote, “Far Infrared Imagery,” Appl. Opt. 15, 1919 (1976). 3. T. Kleine-Ostmann, P. Knobloch, M. Koch, S. Hoffmann, M. Breede, M. Hofmann, G. Hein, K. Pierz, M. Sperling and K. Donhuijsen, “Continuous-wave THz imaging,” Electronics Letters, 37 1461 (2001). 4. K. Siebert, H. Quast, R. Leonhardt, T. Loffler, M. Thomson, T. Bauer, and H. G. Roskos, “Continuous-wave all-optoelectronic terahertz imaging,” Appl. Phys. Lett. 80 3003 (2002). 5. 5 . A. Dobroiu, M. Yamashita, Y. N. Ohshima, Y. Morita, C. Otani, and K. Kawase, “Terahertz imaging system based on a backward-wave oscillator,” Applied Optics 43 5637 (2004). 6. N. Karpowicz, H. Zhong, C. Zhang, K.-I Lin, J.-S. Hwang, J. Xu and X.-C. Zhang, “Compact continuous-wave subterahertz system for inspection applications,” Appl. Phys. Lett. vol. 86, 054105,2005.

7. B. Ferguson and X.-C. Zhang, “Materials for terahertz science and technology,” Nature Materials 1,26 (2002). 8. R. A. Cheville and D. Grischkowsky, Appl. Phys. Lett. “Time domain terahertz impulse ranging studies,” 67, 1960 (1995). 9. D. M. Mittleman, S. Hunsche, L. Boivin, and M. C. Nuss, “T-ray tomography,” Opt. Lett. 22, 904 (1997). 10. T. Domey, W. W. Symes, R. G. Baraniuk and D. Mittleman, “Terahertz multistatic reflection imaging,”J. Opt. Soc. Am. A , 19, 1432 (2002). 11. D. M. Mittleman, M. Gupta, R. Neelamani, R. G. Baraniuk, J. V. Rudd, and M. Koch, “Recent advances in terahertz imaging,” Appl. Phys. B, 68, 1085 (1999). 12. K. McClatchey, M. T. Reiten, and R. A. Cheville, “Time resolved synthetic aperture terahertz impulse imaging,” Appl. Phys. Lett., 79,4485 (2001). 13. A. W. Lee and Q. Hu, “Real-time, continuous-wave terahertz imaging by use of a microbolometer focal-plane array,” Opt. Lett. 30,2563-2565 (2005). 14. Jiang Z. P. and Zhang X.-C., “Terahertz imaging via electrooptic effect,” ZEEE Trans. Microw. Theory Tech. 47,2644 (1999). 15. Jiang, Z. and Zhang, X.-C., “Single-shot spatiotemporal terahertz field imaging.” Opt. Lett. 23, 1114(1998).

32

International J ourna l of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 225-237 @ World Scientific Publishing C o m p an y

World Scientific www.worldscientific.com

AN ANALYSIS OF THE THZ FREQUENCY SIGNATURES IN THE CELLULAR COMPONENTS OF BIOLOGICAL AGENTS

ALEXEI BYKHOVSKI Dept. ofElectrica1 and Computer Engineering, University of Virginia 351 McCormick Rd, P.O. Box 400743, Charlottesville, VA 22904, USA ab4kBvirginia.edu

TATIANA GLOBUS Dept. of Electrical and Computer Engineering, University of Virginia 351 McCormick Rd, P.O. Box 400743, Charlottesville, VA 22904, USA tg9a@virginia. edu

TATYANA KHROMOVA Dept. of Electrical and Computer Engineering, University of Virginia 351 McCormick Rd, P.O. Box 400743, Charlottesville, VA 22904, USA tbk4bBvirginia.edu

BORIS GELMONT Dept. of Electrical and Computer Engineering, University of Virginia 351 McCormick Rd, P.O. Box 400743, Charlottesville. VA 22904, USA [email protected]

DWIGHT WOOLARD U S . Army Research Laboratov, Army Research OfJice Research Triangle Park, NC 27709, USA [email protected]

The development of an effective biological (bio) agent detection capability based upon terahertz (THz) frequency absorption spectra will require insight into how the constituent cellular components contribute to the overall THz signature. In this work, the specific contribution of ribonucleic acid (RNA) to THz spectra is analyzed in detail. Previously, it has only been possible to simulate partial fragments of the RNA (or DNA) structures due to the excessive computational demands. For the first time, the molecular structure of the entire transfer RNA (tRNA) molecule of E. coli was simulated and the associated THz signature was derived theoretically. The tRNA that binds amino acid tyrosine (tRNA,J was studied. Here, the molecular structure was optimized using the potential

33

226

A . Bykhovski et al. energy minimization and molecular dynamical (MD) simulations. Solvation effects (water molecules) were also included explicitly in the MD simulations. To verify that realistic molecular signatures were simulated, a parallel experimental study of tRNAs of E. coli was also conducted. Two very similar molecules, valine and tyrosine tRNA were investigated experimentally. Samples were prepared in the form of water solutions with the concentrations in the range 0.01-1 mg/ml. A strong correlation of the measured THz signatures associated with valine tRNA and tyrosine tRNA was observed. These findings are consistent with the structural similarity of the two tRNAs. The calculated THz signature of the tyrosine tRNA of E. coli reproduces many features of our measured spectra, and, therefore, provides valuable new insights into bio-agent detection.

Keywords: THz, absorption, transfer RNA, E. coli

1. Introduction

Different cell components such as chromosomal deoxyribonucleic acids (DNAs), ribonucleic acids (RNAs), proteins and the bacterial cell wall all possess the potential for contributing to the cellular THz signature and might be essential for the detection of bacterial cells and spores. It was shown previously that whole cell bacterial IR spectra can be reproduced by combining proteidnucleic acid components with appropriate weights.’ Therefore, contributions from separate molecular components should provide us with an insight into the features of the whole cell bacterial THz spectra. Recently, we studied experimentally and theoretically THz transmissionlabsorption spectra of Escherichia coli cells, Bacillus subtilis (BG) cells/spores and their chromosome DNAs, and demonstrated that there is a correlation between chromosomal DNA signature and a corresponding entire spore/cell signature.’ Both the optical activity of a bacterial component at THz frequencies and its concentration inside bacterial cell influence the magnitude of its spectral contribution and its relevance to the biological warfare THz fingerprinting. It was demonstrated both experimentally and theoretically that the genetic material is highly optically active in the THz range.3-6 Therefore, RNAs’ contributions into the THz signature of bacterial cells might be very significant because of their high weight ratio (- 2O%).’ In the present paper, we describe our theoretical and experimental results on THz absorptionltransmission spectra of transfer RNAs of Escherichia coli. In biological warfare agents and their simulants, the genetic material is not dry rather it is solvated in a water solution (cytoplasm) inside a cell. Recently, we developed two classes of nucleic acid models taking into account water either implicitly or explicitly. Energy minimizations and normal mode analysis as well as molecular dynamics and a quasi-harmonic analysis were used to calculate atomic vibrations of solvated DNAs. These models were tested using a double stranded DNA fragment poly(dAT)-poly(dTA) (i.e. the DNA sequence is ATATATAT...) as one of components in a simulation procedure for modeling THz signature of genetic materials. To adjust the model parameters, we used our previous experimental data on THz signature of rather long poly(dAT)-poly(dTA) polymer chains (over 1000 base pairs). The calculated THz spectra correlated well with our measurements, however, only our explicit water model has accurately predicted the position of a significant peak at 14 cm-’. Also, we developed a novel approach to simulate THz signatures of real bacterial chromosomal DNAs taking into account their base pair distributions. For our initial THz absorption modeling, we

34

T H z Frequency Signatures in the Cellular Components of Biological Agents

227

built a 20 base pair Bacillus subtilis model which has approximately the same C/G/A/T ratios and base pair distributions as in the Bacillus subtilis 168 strain. Also, we built a 20 base pair E.-coli model with a sequence which has approximately the same CIGIAIT ratios and base pair distributions as the CFT073 strain of E. coli. An initial 20 base pair Bacillus anthracis DNA model was developed as well. Sodium ions were included in our model. To simplify the treatment, the effect of a solvent (water) was taken into account implicitly using our approach developed for the poly(dAT)-poly(dTA) polymer. The validity of our bacterial DNA models was verified using our experimental data. In this work, we apply the developed models to calculate a THz signature of an entire transfer RNA taking into account the explicit solvation by water molecules inside a bacterial cell. Also, measured transmission spectra of valine and tyrosine tRNAs of E. coli are presented. We compare our theoretical and experimental spectra obtained for tyrosine tRNA (tRNA,) and analyze their spectral features. 2. Theory

Transfer RNAs are low molecular weight form of RNA which serves as an amino acid acceptor. Transfer RNA forms an aminoacyl-tRNA ester bound to a ribosome and serves as the donor of aminoacyl residues in the ensuing reactions in which peptide chain is elongated. There is a substantial evidence that the molecular structure of tRNA has a specific 3-dimensional conformation. Many of its bases are paired and while tRNA denatures on heating, it quickly resumes its native state on cooling. Moreover, it has been demonstrated that the native cloverleaf conformation is essential for their biological activity. Although the tRNAs for different amino acids have quite different base sequences, all tRNAs of known sequence are capable of existing in the same cloverleaf conformation. In this conformations chains are arranged to maximize intrachain basepairing. X-ray diffraction methods helped to deduce the 3-dimensional atomic structure of tRNAs. On the other hand, their conformation may be close to the structure of ribosome RNAs (rRNA). For example, 5s RNA, a species of rRNA with a known structure, is similar in conformation to tRNA. Ribosomal RNAs constitute up to 213 of the weight of ribosomes. Together tRNAs and rRNAs components make about 20% of a dry cell material. Since tRNAs are comparatively small molecules and the atomic coordinates are known for some of them, it is possible to simulate the spectra of the entire molecule and compare the results with the experimental data of the same molecule. Such an analysis might also give an insight into THz spectral features of rRNA as well, because of the above mentioned structural similarity between tRNA and rRNA. 2.1. Model description Our initial model for a bacterial transfer RNA (tRNA), model I, did not include water molecules explicitly. The covalent bond energy, covalent angle energy, proper and improper torsions, non-bonded interactions including electrostatic, and Van der Waals interactions were taken into account in model I using AMBER 99 force field. This model proved to be inferior and did not explain our experimental results. Our basic model for a tRNA, model 11, includes a molecule surrounded by water of cytoplasm inside a bacteria (see Fig. 1) considered at normal environmental conditions (atmospheric pressure and room temperature).

35

228

A . Bukhovski et al.

Figure 1. Biological molecule surrounded by water molecules

In this work, the computer simulation of the entire tyrosine transfer RNA (tRNA,,,) molecule was performed (74 bases, about 40000 atoms including water) and its THz signature was calculated. The initial atomic coordinates of tRNA,, were taken from the Protein Data Bank.' The molecular structure was optimized using the potential energy minimization and molecular dynamical (MD) simulations. The effect of a liquid content of a bacterial cell was taken into account explicitly via the simulation of water molecules using TIP3P water model.', l o This water model represents a rigid water monomer with three interaction sites. Coulomb and Lennard-Jones potentials are taken into account in TIP3P model. A TIP3P water monomer geometry is represented by 0 - H distance of 0.09572 nm and HOH angle 104.52 deg. Pre-equilibrated box of water provided in Amber was used to build an initial atomic coordinates for the simulated system. An initial tRNA minimization was done using the conjugate gradient algorithm. Then a number of MD simulation runs for equilibration and temperature stabilization was performed. Using atomic trajectories from our room-temperature MD simulations, we calculated tyrosine tRNA THz spectra in a quasi harmonic approximation. 2.2. Simulation results

Figure 2 shows snapshots of the simulated tRNA,,

in XZ plane and YZ plane,

correspondingly. Solid circles represent top 6 atoms with biggest contributions into the oscillator strength for the normal mode with v = 23.2 cm-' and the light polarized along x. Three of these atoms are P (phosphorus), 2 are 0 (oxygen) and one is N (nitrogen). We tested both a simplified model (model I) and model 11. The resulting structure for a simplified model was optimized using the conjugate-gradient algorithm for the potential energy minimization until a convergence in atomic forces was reached.

36

T H z Frequency Signatures in the Cellular Components of Biological Agents

229

The positions of absorptions peaks on the frequency scale were calculated. Figure 3 compares results for tRNA absorption calculated using model I and model 11. Model I1 proved to be superior since it allowed a better fit to the experimental peak positions (see the comparison below in Section 5).

Y

X

Fig. 2. Snapshots of the simulated tRNA. Circles represent atoms. XZ plane (left) Y Z plane (right). Dots show 9 atoms with largest contributions into oscillator strength for the normal mode with v = 20.0 cm-I: 5 oxygen (including 2 phosphate group oxygen, 01P and 02P, 2 base oxygen and 1 sugar oxygen), 2 base carbon, 1 phosphorus and 1 base nitrogen.

We conclude that the effect of liquid cellular content might be even more important for adequate THz spectra modeling of tRNA compared to DNA due to a single-strand nature of all bacterial tRNAs. Even though a bacterial single-strand tRNA has fragments with complementary base composition bonded to each other, the rest of the molecule has no hydrogen bonds between bases because bases do not match.

37

230

A . Bykhovski et al.

10

15

20

25

30

Frequency, cm-' Fig. 3. The average absorption coefficient for model I (lower curve) and model I1 (upper curve). The tRNA for tyrosine. Model I - the minimization without water, Model I1 MD with water. ~

It appears that such an arrangement makes tRNAs more exposed to interactions with water molecules surrounding it. Double-strand bacterial DNA has stronger hydrogen bonds than partially bonded tRNA so a DNA structure dynamics might be less affected by its ambience. Only the explicit water model seems to adequately represent an effect of water in cell on tRNA dynamics. The calculated THz spectra of tRNA in 5 - 30 cm-' range are presented in Fig. 4 for different light polarizations. The molecule is oriented approximately as in Fig. 2. According to our simulation results, multiple absorption peaks exist in this range, in particular, around 10.5 cm-', 14 cm-', 15 cm-', 20 cm-' and 23 cm-'.

0 ' 10

15

20

25

30

Wave Number, cm-' Fig. 4. The calculated THz spectrum of tRNA,, (MD, quasi harmonic approximation). Light polarized along X (dashed line). Light polarized along Y (dotted line). Light polarized along Z (solid line).

38

T H z Frequency Signatures in the Cellular Components of Biological Agents 231

3. Materials and methods As substrates, we used polyvinylidene chloride (PVDC) films (Saran FilmBSaran, the Dow Chemical Company). The details of techniques for the sample preparation and the THz spectroscopic characterization are described in Ref.". Each sample has been prepared as a cell assembled of two substrate (or window) thin films and a polyethylene spacer placed between them. We prefer to use very thin cell windows (film substrates) to avoid interference (etalon) effects in these cell components. Because of larger wavelengths of radiation in the THz compared to the far-IR range and in order to achieve a spectral resolution of 0.25 cm-' that is required for resolving resonance features with line width of 0.5 cm-', interference effects can become a serious problem, reducing the accuracy and reproducibility of measurement results if not properly eliminated. The thickness of the spacer was 12 pm, and the inner diameter was 18 mm. Up to 10 p1 of solution has been placed on the supporting film inside the spacer, covered with the second film and pressed down to achieve even distribution of the solution inside the cell. Since saran film melts at only 80 OC, cells made with this material were sealed by pressing with a preheated to 100 OC copper cylinders for no longer than -2-3 sec on the both sides of a spacer in order to prevent evaporation and leakage of the solution from the cell samples. The sealed samples held solutions without changes in transmission for at least 24 hours. Due to the surface tension, both substrate materials kept thin layers of solution between two films even without sealing for some shorter period of time. THz spectra were recorded on a IFS 66 V Fourier-transform infrared spectrometer from Bruker Instruments (Bremen, Germany) equipped with a Si bolomoter (Infrared Laboratories Inc., USA). Typically, 32 interferograms were averaged to reduce noise. Under these settings, the recording of a single spectrum takes only 42 sec. The noise of a Si bolometer operating at 1.8K is extremely low, and with the optical aperture of 12 mm the signal to noise ratio is >500. The reproducibility of the instrument in transmission mode is better than 0.2 to 0.3 % under these conditions. A cooled filter with a cut-off frequency of 35 cm-' was used inside the bolometer cryostat. The sample was placed in the standard Bruker sample holder in the focus of the beam in the sample compartment. The whole instrument, except for the sample compartment, was evacuated to 10 mbar. Measurements in air are possible since there are almost no absorption lines from water vapors or oxygen in the 10 - 25 cm-' spectral range. A notable exception is a relatively weak water absorption band at around 18.6 cm-'. To separate spectral features of biological molecules from the contributions made by water and substrate materials, transmission spectra of pure water between two substrate films were recorded. Spectra of biomolecules in the aqueous solution were calculated against the transmission of cells with matching amounts of pure water between the corresponding substrates. Since in all spectra of this study the concentration of RNA was very low (0.01-0.1%) and since the absorption of biomolecules is less than that of water, we believe that the absolute level of transmission of the cells with biological materials is determined mainly by the contribution from water and substrate. Therefore, the

39

232

A . Bykhovski et al.

transmission of the RNA was obtained by simply dividing the measured spectrum of the dilute RNA sample by the spectrum of water with the same or next higher level of transmission. 4. Experimental results

4.1. THz characterization of E-coli ‘s transfer RNA diluted solutions. We have prepared samples and measured spectra of valine transfer RNA and tyrosine tRNA in the form of very diluted solutions sealed between substrates made from polyethylene (Saran film, 12 pm thick). As was mentioned above, the results are obtained after dividing measured spectra by measured transmission of pure water contained between the same substrate films. Since it is very difficult to reproduce exactly the same amount of solution or water, the error of several percent in absolute value of resulting spectra is possible. For this reason, we first of all compare the resonance peak positions. Figure 5 demonstrates a very good consistency in results for several concentrations. Positions of resonance features are well reproduced. The intensity of peaks is weakly dependent on the material concentration. This result might be partially attributed to the error in the absolute value discussed above. From the Fig. 6 it is clear that tyrosine and valine tRNAs have very similar signatures, which is not surprising since they belong to the same class of small transfer RNA. 1.oo

0.98

0.92

0.90 10

12

14

16

18

20

22

24

Frequency, cm-’

Figure 5. Spectra of tyrosine RNA solutions with 3 different concentrations between two Saran films Averaged result is also shown.

40

T H z Frequency Signatures in the Cellular Components of Biological Agents

233

0.99

0.98

0.97

0.96

0.95

0.94

0.93 10

12

14

16

18

20

22

24

Frequency cni’

Figure 6. Averaged results for the spectra of tyrosine (bottom) and valine (top) RNA water solutions between Saran films.

5. Experiment v theory Since the entire molecule was simulated, a comparison with experiments became more straightforward than for the bacterial DNAs. The experimental absorption signature was extracted from the measured transmission spectra of tyrosine tRNA of E. coli. The extraction of the absorption coefficient and estimation of the refractive index of the biological material was done as described in our paper.I2 In Figure 7, the comparison is made between the results of THz signature simulations (model I - MD with the explicit water) for a non-polarized light, obtained as an average over polarized light results, and extracted experimental spectrum of tyrosine tRNA. Most of the experimental peaks of absorption in 10-25 cm-’ range are matched by our simulation (see Fig. 7 and Table 1). In particular, our calculation (model 11) correctly predicts strong absorption features around 10.6-10.8, 13.7-14, 14.9-15, 15.6, 16.2-16.3, 20, and 23-23.2 cm-’. We conclude that the effect of liquid cellular content might be even more important for adequate THz spectra modeling of tRNA due to a single-strand nature of all bacterial tRNAs. Even though a bacterial single-strand tRNA has fragments with complementary base composition bonded to each other, the rest of the molecule has no hydrogen bonds between bases because bases do not match. It appears that such an arrangement makes tRNAs more exposed to interactions with water molecules surrounding it. Only explicit water simulations seem to adequately represent the effect of water on tRNA dynamics. On the other hand, a double -strand bacterial DNA has stronger hydrogen bonds than partially bonded tRNA, so a DNA structure dynamics might be less affected by its ambience. The theoretical and experimental absolute values of the absorption coefficient

41

234

A . Bykhovski et a1

were not compared. Hence, the combination of resonance peak frequencies gives the most reliable information on the THz signature of biological species. 40 I

!

--5

.-

.. . ... ... .. ... .

.. . ... .. . ... ...

.. ... ..

...

..

..

30

0

. ... ..

.. .. . ... . ... .. ... . ... ... ..

..

.

.

s

0

.. .. ... ..

... ....

... ....

... ....

... ....

... ....

... ....

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

..

I

10

12

14

la

16

20

..

.. I

22

24

Frequency, cm-' Fig. 7. The average absorption coefficient: experiment for saran (dotted line) and theory, explicit water model, (solid line).

Table 1. E.coli tRNA: theoretical and experimental (saran) peak positions, cm-'. Theory

Experiment

10.6 11.4 11.9 13.0

10.8 12.8 13.2 14.0 15.0 15.6 16.3 17.2

13.7 15.0 15.6 16.2 17.2 17.6 18.0 19.3 20.0 21.2 21.7 23.2

18.1 19.3 20.0 21.0 22.2 23.0 23.7

24.0 24.6

24.2

42

T H z Frequency Signatures in the Cellular Components of Biological Agents

235

Table 2. Simulated tRNA and DNA peak positions of E.coli, cm-’ tRNA, Theory (present work)

DNA, Theory (Ref. 2,

10.6 11.4 11.9

10.2 10.9 12.0 12.6 13.5 13.8 15.2 15.9

13.0 13.7 15.0 15.6 16.2 17.2 17.6

16.8 17.9 18.3 18.9 19.2 19.8 20.8

19.1 19.3 20.0 21.2 21.7 23.2 24.0

22.0 22.3 22.8 24.0

24.6

24.8

We also compared predicted tRNA and DNA absorption peak positions of E.coli in the 10-25 cm-’ range (see Table 2 and Fig. 8).

40

I

0

10

12

14

16

18

20

22

24

Wave Number. cm-l

Fig. 8. The calculated spectra of tRNA,, (solid line), present work, and DNA (dashed line) of E. coli from Ref. ’.

It should be noted that DNA and tRNA have different atomic structures; therefore differences in their absorption spectra are to be expected. In particular, the Thymine base exists only in DNA. It is replaced by a Uracil base in all RNAs. Also, DNA and tRNA have very different conformations. In a DNA, all bases are paired, which means that

43

236

A . Bykhovski et al.

bases in a base pair interact with each other via a hydrogen bond (A-T and C-G). This interaction leads to a double strand helical DNA conformation. On the other hand, many but not all bases are paired in tRNAs. As a result, tRNA exists in a cloverleaf conformation (see Fig. 2). Hence, DNA and tRNA can have different normal modes and different resonance frequencies for the light absorption. Indeed, the differences in peak structures of simulated tRNA and DNA spectra of E.coli are noticeable, especially around 10-11 and 20-23 cm-' (see Fig. 8). The simulated peak positions are correlated at 14 cm-I, 15 cm-' and 19 cm-'. Peaks at 14 cm-' and 15 cm-l are quite common for different nucleic acids.' Correlations between the THz spectra of tRNA,,, and DNA of E. coli were also observed experimentally. Therefore, nucleic acids have some common features in THz range; however their spectral fingerprints are distinguishable. Since combined DNA and RNAs make almost a quarter of the dry bacterial weight of E. coli, nucleic acids provide the important contribution into the cellular spectrum. The obtained results point to the importance of water-RNA interactions for RNA dynamics and absorption spectrum, and, therefore, for the THz signature prediction of bio warfare agents in the sub THz range.

6. Acknowledgments This work was supported by the US DOD under Contract No W911NF-05-1-0033 and DAAD19-00-1-0402, and in part by U S . Army NGIC Contract # DASCO1-01-C0009.

References 1. Ye. L. Belokour, G. I. Dovbeshko, G. S. Litvinov, Experimental and computer modeling of vibrational spectra of heterogenous complexes of biological macromolecules, J. Molecular Structure, 267, 61-66, (1992). 2. Alexei Bykhovski, Xiaowei Li, Tatiana Globus, Tatyana Khromova, Boris Gelmont, Dwight Woolard, Alan C. Samuels, and James 0. Jensen, THz absorption signature detection of genetic material of E. coli and B. subtilis, SHE, 5995, 198-207 (2005). 3. M. Bykhovskaia, B. Gelmont, T. Globus, D. L. Woolard, A. C . Samuels, T. Ha-Duong and K. Zakrzewska, Prediction of DNA far-IR absorption spectra based on normal mode analysis, Theor. Chem. Accounts, 106,22-21 (2001). 4. T. Globus, M. Bykhovskaia, B. Gelmont and D. L. Woolard, Far-infrared phonon modes of selected RNA molecules, In Instrumentation for Air Pollution and Global Atmospheric Monitoring, James 0. Jensen, Robert L. Spellicy, Editors, Proceedings of SPIE, 4574, 119-128 (2002). 5. .T. Globus, D. Woolard, M. Bykhovskaia, B. Gelmont, L. Werbos and A. Samuels, THzFrequency spectroscopic sensing of DNA and Related Biological Materials, International Journal of High Speed Electronics and Systems. 13(4), 903-936 (2003). 6. T. Globus, M. Bykhovskaia, D. L. Woolard, and B. Gelmont, Sub-millimeter wave absorption spectra of artificial RNA molecules, J. Phys. D: Appl. Phys., 36, 1314-1322 (2003). 7. A. L. Lehninger, Biochemistry, 2"d edition, 1981. 8. http:llwww.pdb.org. 9. D.A. Case, D.A. Pearlman, J.W. Caldwell, T.E. Cheatham 111, J. Wang, W.S. Ross, C.L. Simmerling, T.A. Darden, K.M. Merz, R.V. Stanton, A.L. Cheng, J.J. Vincent, M. Crowley, V. Tsui, H. Gohlke, R.J. Radmer, Y . Duan, J. Pitera, I. Massova, G.L. Seibel, U.C. Singh, P.K.

44

T H z Frequency Signatures in the Cellular Components of Biological Agents 237

Weiner and P.A. Kollman, AMBER 8, University of California, San Francisco. httu://amber.scriuus.edu/ (2004). 10. W.L. Jorgensen, J. Chandrasekhar, J. Madura & M.L. Klein, Comparison of Simple Potential Functions for Simulating Liquid Water., J. Chem. Phys., 79,926-935 (1983). 11. Tatiana Globus, Tatyana Khromova ,Boris Gelmont , Dwight Woolard , and Lukas K. Tamm, Terahertz characterization of diluted DNA solutions, Proceedings of SPIE Photonics West 2006 (Biomedical Vibrational Spectroscopy 111: Advances in Research and Industry), Biomedical Optics Symposium, San Jose, California, 6093, 7, (2006).

12. T. Globus, D. Woolard et al., THz-Frequency Spectroscopic sensing of DNA and related biological materials, Int. J. of High Speed Electronics and Systems, 13, 903-936 (2003).

45

This page intentionally left blank

International Journal of High Speed Electronics and Systems Vol. 17, NO. 2 (2007) 239-249 @ World Scientific Publishing Company

World Scientific www.woridscientific.com

STANDOFF SENSING AND IMAGING OF EXPLOSIVE RELATED CHEMICAL AND BIO-CHEMICAL MATERIALS USING THz-TDS Hua Zhong Centerfor THz Research, Rensseluer Polytechnic Institute, 110 8IhSK Troy, NY 12180 USA [email protected] Albert Redo-Sanchez Center for THz Research. Troy, NY I 2 1 90 redosa@rpi. edu Xi-Cheng Zhang Centerfor THz Research. Troy, NY 12190 [email protected] We report the sensing and imaging of explosive related chemical and bio-chemical materials by using terahertz time domain spectroscopy (THz-TDS) at standoff distance. The 0.82 THz absorption peak of RDX is observed at a distance up to 30 m away from the emitter and receiver. Multiple absorption features of FOX, 2,4-DNT and Glutamic Acid are identified by using a large scale 2-D imaging system. These results support the feasibility of using THz-TDS technique in remote sensing and detection of chemical materials. Keywords: terahertz, standoff sensing, spectroscopic imaging

1. Introduction

Standoff detection refers to sensing chemical and biological materials when the sensor is physically separated from the target. Although many techniques and methodologies are available for explosive screening when close proximity is possible, long standoff distance (several meters) sensing of explosive material remains as a bottleneck. The candidates for standoff detection include X-ray imaging, thermal neutron activation, microwave imaging, thermal radiation detection, and optical chemical sensors [ 13. Nevertheless, X-rays and thermal neutrons are potentially harmful to humans and suffer from severe loss at long distance. Microwave imaging and thermal radiation cannot deal with the traces of interested chemicals; neither do they provide enough spatial resolution. Optical spectroscopy has been envisioned as a promising group of methodology by providing non-contact and in-situ visibility to signatures of chemical explosives arising from energy transitions between modes. Recently, sensing of explosives at 50 m away by using Raman spectroscopy has been reported [2].However, a significant drawback of optical technique is their poor vision for threats concealed under covers.

47

240

H. Zhong, A . Redo-Sanchez F4 X.-C. Zhang

Terahertz (THz) imaging and sensing technologies have stood out in the past decade as a bridging technique between microwave and infrared sensing modalities. The THz band, which refers to the spectrum region between .0.1 THz to 10 THz, covers the fingerprints of many chemical explosives such as RDX, 2,4-DNT, PETN, and TNT [3-61. THz waves are transparent to many materials that are opaque to optical beam, such as foam, plastic, wood, and fabric. Because of its low photon energy, THz radiation is nonionizing and non-invasive to biological tissue [7]. Spatial resolution provided by THz radiation is substantially higher than microwave due to its shorter wavelength. 2. Principle

At normal incidence, the complex reflection coefficient r" is given by the relation between the reflected beam E2 and the incident beam El. It can be derived through the complex refractive index of the material by [S]:

-

y = - -E -2

El

-

6-1 ti+l

-

(n-l)+iK (U+l)+iK

=fieib

(1)

If the phase shift 4 between El and E2 and the reflectance R are known, then the refractive index n and the extinction coefficient K can be calculated as:

n=

I+ R 1+ R - 2&cos4

K=

2& sin 4

(2)

1+R-2&cos4

The absorption coefficient a is then: 4ZV K a=---

c

(3)

Where v is the frequency and c is the speed of light. Owing to the coherent detection of THz-TDS, both phase and amplitude of each waveform are readily available. Therefore the extinction coefficient (and absorption coefficient) of the sample material can be calculated according to (1)-(3).

2.1. Standoff sensing of R D X The experimental set-up is designed in such a way that the measurement is mainly affected by the atmospheric attenuation, which means that the collection of the beam has to be optimized. The schematic diagram is illustrated in Fig. 1. THz pulses are generated using a Spectra Physics Mai Tai laser with 800-nm center wavelength, 80-fs pulse duration, 80-MHz repetition rate, and 800-mW average power. The THz radiation is produced by a p-type InAs crystal via photo-Dember effect. A 4-inch focal length parabolic mirror collects and collimates the THz pulses. The THz pulses travel in room air for 30 m and incident onto the sample's surface at a small angle (less than 10'). The specularly reflected THz pulses travels back to the sensor crystal for

48

Imaging of Explosive Related Materials Usang THz-TDS 241

another 30 m. The reflected THz pulses are finally collected by a large-size polyethylene lens and is detected by a ZnTe crystal [9]. Sample

Fig. 1. Experimental setup of standoff THz detection

2.1.1. Signature of RDX at close proximity Fig. 2 compares the absorption spectra of RDX measured in transmission mode in a nitrogen-purged cell (courtesy of Yunqing Chen, Haibo Liu and Jian Chen) in reflection mode at 17% and at 64% relative humidity (RH) levels. Reflection spectra match well with transmission spectrum at low F W level (17%). At a high RH level (64%), only the peak centered at 0.82 THz is visible.

h

7

8

250

=1: 200

9

4

.3

8

150

8

100

.!Z E-

50

$

13

Q

0

-50 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Frequency (THz)

Fig. 2 . The absorption spectra of RDX measured at 0.7 m in reflection geometry at different RH levels are compared with the spectrum measured in transmission geometry. Shaded areas represent the absorption lines around 0.82, 1.05, 1.33 and 1.55 THz. Reflection spectra match well with transmission spectrum at low RH level.

49

242

H . Zhong, A . Redo-Sanchez & X.-C. Zhang

2.1.2. Sensing of RDXat standoffdistance The absorption spectra of RDX and polyethylene (which has no absorption features within the THz range) measured at various distances are calculated and compared in Fig. 3. Each spectrum is plotted at the same scale but shifted for clarity. The absorption peak of RDX centered at 0.82 THz is still visible at 30 m. Besides that, the water vapor absorption at longer propagation distance causes the following issues: (1) The peaks are gradually broadened as traveling distance increased. This broadening is mainly due to increased phase and amplitude fluctuations at longer distances. Because during the calculation of absorption curve, each spectrum is averaged for multiple times, the increased spectral fluctuation smears out the absorption peak. (2) The spikes around water lines (0.38, 0.44, 0.56, 0.75 and 0.99 THz) are more prominent at longer distances. They are mainly contributed by the fluctuation and saturation of water vapor absorption, which can be alleviated by averaging if there is no long-term humidity variation in the atmosphere. Despite the spikes, the absorption peak of RDX is not strongly affected, which indicates the possibility to extend the sensing distance even further, at lower relative humidity levels.

'E 0

W

+

600 400

0

0.2

0.6

0.4

0.8

1.o

Frequency (THz) Fig. 3. The absorption spectrum of RDX compared with a polyethylene pellet at 2.5, 10, 20 and 30 m. The shaded area indicates the absorption peak round 0.82 THz. The spikes around 0.38, 0.44, 0.56, 0.75 and 0.99 THz are due to fluctuations of water vapor.

50

Imaging of Explosive Related Materials Using T H z - T D S

243

3. Spectroscopic focal-plane imaging of explosive materials at standoff distance The technique of THz focal-plane imaging was first developed in late 90s. Rather than focusing the THz pulse on the sample, quasi-plane wave illumination is used and both THz and probe beams are expanded. The THz beam is modified by the target (either in transmission or reflection) and overlaps with the probe beam in a large sensor crystal. Each point on the probe beam wave front is modulated by the local THz field, which carries the information of the target. The wave front is then captured by a CCD camera. By taking the 2D image at once and only scanning the time delay, the data acquisition speed is dramatically improved. is affected by sample In THz reflective focal-plane imaging, the phase shift surface roughness and sensor crystal non-homogeneity, leading to erroneous results when being applied to (2) [lo]. However, calculations show that the first order derivative of Y respect to the frequency (-dddv)can also provide the positions of absorption frequencies. A numerical approach can prove that the peak position deviations are within 0.02 THz [ 111. It is noteworthy that the property only holds to weakly polarized organic material, which has moderate absorption compared to dispersion in THz range, such as ERCs, biochemicals, etc. [11,12,13]. The schematic layout of the setup is illustrated in Fig. 4. The laser used is a Spectra Physics Hurricane with 1 KHz repetition rate, 100 fs pulse duration, 700 mW output power, and 795 nm center wavelength. The THz beam is generated with a 2.5 mm thick ZnTe crystal and is expanded to a diameter of 25 mm. The incident angle of THz beam onto the sample's surface is 15". A polyethylene lens with 200 mm focal length is used to image the sample target onto a large piece of ZnTe sensor crystal (50 mm by 50 mm by 5 mm). Both the distance of image and distance of object are 0.4 m (2f-20. The probe beam has an expanded diameter of 25 mm and co-propagates with THz beam through the sensor crystal. A Princeton Instrument CCD camera is used to capture the image of the probe beam. The spatial resolution of the imaging system is 2 mm. To ensure the best imaging quality, images of each sample were taken separately but concatenated together as a whole image. Probe beam

'0

Fig. 4. Experimental-set up of THz reflective spectroscopic focal-plane imaging system.

51

244

H. Zhong, A . Redo-Sanchez 63 X.-C. Zhang

The imaging targets are 2,4-DNT, Theophylline, RDX, Glutamic Acid and glass. All samples except the glass are prepared as compressed pellets with a thickness of -1.5 mm and a diameter of 13 mm. However, the RDX material is extremely brittle and only a small piece is used as imaging target. Glass does not have any absorption features within THz region and is used as a contrast sample. Fig. 5 shows the extinction coefficients of 2,4-DNT, Theophylline, RDX and Glutamic Acid measured with the same sample pellets, in transmission geometry by a standard THz-TDS. Each curve is plotted at the same scale. All labeled absorption peaks have been confirmed by other measurements [36,141. It is noteworthy that the absorption peaks of RDX at 0.82 THz ( K -0.3) and 1.33 THz (K -0.1) [5], 2,4-DNT at 1.08 THz ( K -0.2), Theophylline at 0.96 THz ( K -0.06) and Glutamic Acid at 1.21 THz ( K -0.1) are the most prominent within the spectral window between 0.4 THz to 1.6 THz. 2.0 k

s 0

Ew 8c .-+ 0

,I1.04 .Y+ I\A

1.33 1.55

1.5

f

0.96 Theophylline 1.02,4-DNT T 8

Glutamic Acid 0.0 r 0.4

I

t

RDX

Y

.d

0.82 OX2

1.6

T

1.4

4 ,I

0.8

1.2

1.6

II

Frequency (THz) Fig. 5. The extinction coefficient K of RDX, Theophylline, 2,4-DNT, and Glutamic Acid measured by transmission THz-TDS in nitrogen purged cell. Data is shifted for clarity. Arrows indicate the absorption peaks of each target.

3.1. Imaging results 3.1.1. Optical and THz images Fig. 6 is the optical picture of all five imaging samples and Fig. 7 shows the THz image obtained by taking the peak amplitude of the THz pulse at each pixel. Samples can not be identified from the THz image. 3.1.2. Spectroscopic image Fig. 8 shows the -dr/dv curve of five pixels randomly chosen from the image of each sample. Three absorption peaks of RDX, one of 2,4-DNT and one of Glutamic Acid are

52

Imaging of Explosive Related Materials Usang THz- TDS 245

well ~ ~ e n tHowever, ~ ~ e ~there . is no significant absorption features to extraet from the ~ ~ ~ ~ p ~ y spectrum. ll~ne's

Fig. 6. Optical picture of 2,4-DNT, Theopylline, RDX, Glutamic Acid and glass sample target.

Fig. 7. THz peak ~ ~ pimage ~ ofi 2,4-DNT, ~ ~ Theophylline e RDX, Glutamic Acid and glass. Samples can not be identified from this image.

I

0.80

i .

____I

1

I

I .2

0.8

I .6

Frequency (T&) Fig. 8. -& / dv o f RDX, ~heophyl~ine, 2,4-DNT, Glutamic Acid and glass from five randomly selected pixels on each sample image. The arrows indicate the absorption peaks. Three absorption peaks of RDX, one of 2,4DNT and one o f Glutamic Acid are located by comparing with Fig. 5 . No significant absorption feature on the curve of Tbeophyllinecan be identified,

53

246

H. Zhong, A. Redo-Sa,nchez i 3 X.-C. Zhung

In order to identify each sample from their -dr / dv spectra, the peak areas around 0.82 THz, 0.96 THz, 1.08 THz and 1.21 THz are integrated for each pixel with a width of 0. X 5 THz Analytically, the value of each pixel is cakulated as:

Where v2 and vI indicate the range of the integration. The pixels with peak within the window appear brighter and the ones without appear darker. The b a c k ~ r o value ~ d of the image i s set to be 0. The results are illustrated in Fig. 9, which show that at 0.82 TI-Fz, 1.08 TfTz and 1.21 TBlz, the samples that have the co~espond~ng a b ~ o ~ t i peak o n at each frequency appear to be the brightest, No significant cluttered d is ~ b u tio nat 0.96 THz (the ~ b s o ~ ~peak i o nof Theophylline) is addressed. The failure to identify T h e o p h y ~ ~isi ~due e to the fact that i ts absorption peak at 0.94 THz i s too weak to be resolved under the current imaging dynamic range [ 111.

Fig. 9. The images o f the sample targets formed by integrating the peak area around (a) 0.82 "Hz; (b) 0.96 THz; (c) 1.08 THz and (d) 1.21 TNz, with the width being 0.15 THz. Except for the image at 0.96 THz, which is supposed to be the absorption peak location of Theophylline, all three other samples can be identified at the images o f their corresponding absorption peak frequencies.

54

Imaging of Explosive Related Materials Using THz- TDS

247

3.1.3. Contrast of the image The contrast of the spectroscopic image is defined as [15]:

Where Ipeak is the mean value of the “bright” area and Iflooris the mean value of the background. Both values can be extracted from the histogram of the image. To fairly compare all four images, the pixel values of each image are normalized linearly to range from 1 to 10. Fig. 10 shows an example of the histogram of the spectroscopic image at 1.2 1 THz (Fig. 9 (d)). The pixel value on X-axis is normalized linearly to range from 1 to 10. The value of Iflooris the peak position of the normal distribution on the left and Ipeak is the one on the right. Both are marked with dashed lines. Table 1 lists the contrast of all four images in Fig. 9. It is noticed that the contrast value of the image at 0.96 THz is 0 as a result of a single normal of the histogram, which means that there is not any sample identified. The image at 0.82 THz has the highest contrast, which is also understood because the extinction coefficient K of RDX is the highest among them all. Therefore the calculations of image contrast provide a measure to evaluate the identification result.

2

-.-

v

4000

v1

a

3000

k

0

I

e,

2000-

2E

1000-

13

0-

2

6

4

8

10

Pixel value (I) Fig. 10. Histogram of the spectroscopic image at 1.21 THz on Fig. 9 (d). The pixel value is normalized from 1 to 10. The position of the peak on the left I,,,,,, represents the pixel value of the background, whereas the position of the peak on the right Zpeok indicates the pixel value of the foreground, which is the sample being identified.

55

248

H . Zhong, A . Redo-Sanchez 63 X.-C. Zhang

Table 1. Contrast of each image on Fig. 9. The image at 0.82 THz has highest contrast because the extinction coefficient K of RDX is measured to be the highest among the four. The image at 0.96 THz has zero contrast resulting from the single normal distribution on its histogram. The failure to identify any absorption feature at 0.96 THz image is due to insufficient dynamic range of the imaging system. Image @ frequency

Contrast

0.82 THz

0.96 THz

1.08 THz

1.2 1 THz

50%

0

20%

17%

4. Conclusion

To conclude, the results demonstrate that THz-TDS is a promising technology for standoff sensing o f explosive materials. The absorption peak o f RDX at 0.82 THz is detected at a standoff distance u p to 30 m, and multiple 2-D spectroscopic features of 2,4DNT, Theophylline, RDX, and Glutamic Acid are detected b y THz focal-plane imaging system at a distance up to 0.4 m. Acknowledgements This work is supported by Army Research Office and the National Science Foundation. We greatly acknowledge the generous help offered by Mr. Gary Young, Dr. Yuqing Chen, Dr. Xin He, Mr. Jian Chen, Dr. Glenn Bastiaans and Dr. Charles Davis.

References 1. J. E. Parmeter, “The challenge of standoff explosives detection”, Proceedings. IEEE 38th Annual 2004 International Carnahan Conference on Security Technology ,355-358 (2004). 2. J. C. Carter, S. M. Angel, M. Lawrence-Snyder, J. Scaffidi, R. E. Whipple and J. G. Reynolds, “Standoff Detection of High Explosive Materials at 50 Meters in Ambient Light Conditions Using a Small Raman Instrument”, Applied Spectroscopy, 59, 769-775 (2005). 3. M. C. Kemp, P. F. Taday, B. E. Cole, J. A. Cluff, A. J. Fitzgerald, and W. R. Tribe, “Security applications of terahertz technology”, Proc. SPIE 5070,44-52 (2003). 4. Y. Chen, H. Liu, Y. Deng, D. Veksler, M. Shur, X. -C. Zhang, D. Schauki, M. J. Fitch, and R. Osiander, “Spectroscopic characterization of explosives in the far infrared region”, Proc. SPIE 5411, 1-8 (2004). 5. K. Yamamoto, M. Yamaguchi, F. Miyamaru, M. Tani, M. Hangyo, T. Ikeda, A. Matsushita, K. Koide, M. Tatsuno and Y. Minami, “Noninvasive inspection of C-4 explosive in mails by terahertz time-domain spectroscopy”, Japanese J. of Appl. Phys., 43, n 3B, L414-417 (2004). 6. F. Huang, B. Schulkin, H. Altan; J. F. Federici, D. Gary, R. Barat, D. Zimdars, M. Chen and D. B. Tanner, “Terahertz study of 1,3,5-trinitro-s-triazine by time-domain and Fourier transform infrared spectroscopy”, Appl. Phys. Lett., 85, n 23, 5535-5537 (2004). 7. R. H. Clothier and N. Bourne, “Effects of THz exposure on human primary keratinocyte differentiation and viability”, J. Biol. Phys., 29, 179-185 (2003). 8. J. Jackson, Classical Electrodynamics, Wiley & Sons, (New York, 1975). 9. Q. Wu, M. Litz, and X.-C. Zhang, “Broadband detection capability of ZnTe electro-optic field detectors”, Appl. Phys. Lett., 68,2924-2926 (1996).

56

Imaging of Explosive Related Materials Using THt-TDS

249

10. E.M. Vartiainen, Y. Ino, R. Shimano, M. Kuwata-Gonokami, Y.P. Svirko, and K.-E. Peiponen, “Numerical phase correction method for terahertz time-domain reflection spectroscopy”, J. Appl. Phys., 96 (8), 4171-4175 (2004). 11. H. Zhong, “Terahertz wave reflective sensing and imaging”, Ph.D. thesis (2006). 12. F. Wooten, Opticalproperties of solids, Academic (New York, 1972). 13. K. Yamamoto, A. Masui and H. Ishida, “Kramers-Kronig analysis of infrared reflection spectra with perpendicular polarization”, Appl. Opt., 33, 6285-6293 (1994). 14. H. Liu, “Terahertz spectroscopy for chemical and biological sensing applications”, Ph.D. thesis (2006). 15. E. Hecht, Optics, Addison Wesley Longman, (1998).

57

This page intentionally left blank

International J ourna l of High Speed Electronics an d Systems Vol. 17, NO. 2 (2007) 251-260 @ World Scientific Publishing C o m p an y

World Scientific w w w worldscientific.com

FINGERPRINTING INSULINS IN THE SPECTRAL REGION FROM MID-IR TO THZ RENBO SONG and YUJIE J. DING* Department of Electrical and Computer Engineering, Lehigh University, Bethlehem. PA 18015, U.S.A *[email protected] YULIYA B. ZOTOVA

ArkLight, P. 0. Box 2, Center Valley, PA 18034, U.S.A

Motivated by the possibility of identifying and detecting certain biochemical species using Fouricr transform infrared spectroscopy (FTIR), we have investigated porcine, bovine, lispro, and human insulins. We have successfully observed and identified all the transition peaks for the four types of insulins in the frequency domains from mid-IR to THz. In the mid-IR region, ten transition peaks have been observed for all four insulins. Although these four insulins are made from amino acids which have either slightly different sequences or slight variations, the ten transition frequencics are virtually indistinguishable. However, for protamine sulfate some of the transition frequencies in the same mid-IR region are either red-shifted or blue-shifted relative to the corresponding ones for the four insulins. Furthermore, the strengths for several peaks of protamine sulfate are significantly reduced whereas the strength for only one transition peak is significantly enhanced, compared with the corresponding ones for the insulins. In the far-IR to THz transition region, the shapes and locations of the two transitions are quite different between the insulin species and protamine sulfate. In the THz region we have observed a linear dependence of the absorption coefficients on frequency for each of the four insulins and protamine sulfate.

Keywords: Vibrational transitions; Vibrational modes; Spectroscopy; Fourier Transforms Infrared Spectroscopy; far infrared; THz; Optical diagnostics for medicine; Boson Peak.

1. Introduction

Insulin is a polypeptide hormone which regulates carbohydrate metabolism. It has been used medically in some forms of diabetes mellitus. Since insulin can be oligomerized into a less soluble hexamer state, the clinical efficiency is reduced by delayed absorption rate. To overcome such a deficiency, insulin lispro (Humalog, rDNA origin) was introduced and developed. Insulin lispro is an analog to human insulin. It has become a rapid-acting

59

252

R. Song, Y. J . Ding &

Y.B. Zotova

and parenteral agent for lowering blood glucose. Chemically, it is created when the amino acids at positions 28 and 29 on the insulin B-chain are reversed, i.e. the Lys(B28), Pro(B29) human insulin analog. Besides insulin lispro, insulins from animals such as cows and pigs (i.e. bovine insulin and porcine insulin) are similar to human insulin in terms of their functions. As a matter of fact, they have been used for the treatment of diabetes mellitus since 1922. Between bovine and porcine insulins, porcine has a composition and sequence of the amino acids which are closer to that for human insulin. Indeed, the primary differences between the sequences of the amino acids for porcine and bovine insulins and human one are reflected by three and one amino acid, respectively. In Table 1, we have summarized the detailed variations of the amino acids used to form four insulin species. Table 1. Variations of Sequences of Amino for Four Different Insulins. Species

A8

A10

B28

B29

B30

Human

Thr

Ile

Pro

LYS

Thr

Lispro

Thr

Ile

LYS

Pro

Thr

Bovine

Ala

Val

Pro

Lys

Ala

Porcine

Thr

Ile

Pro

LYS

Ala

Nowadays, 4 million Americans rely on insulin to live. Since April 1, 2006 Eli Lilly has stopped the manufacturing of the animal insulins extracted from pancreatic tissue of cows and pigs. Since some patients have adverse reaction to the insulin lispro, it may be important for us to develop a simple spectroscopic technique to differentiate different types of the insulins. As the very first step of our attempt, we would like to fully understand the vibrational transitions for all these insulin species and to eventually differentiate them based on the spectroscopic signatures. In order to differentiate the four insulins from other types of the proteins, the protamine sulfate from salmon sperm was chosen and measured by us. Protamine sulfate is a purified mixture of simple protein principles obtained from the sperm or testes of suitable species of fish. It has the property of neutralizing heparin by reversing the anticoagulant effects of heparin by binding to it. It has long been demonstrated in the past that Fourier transform infrared spectroscopy (FTIR) is extremely valuable for mapping out the vibrational modes in biochemical samples. Although Raman spectroscopy is a powerful tool for fingerprinting amide I and I11 bands, Raman signal is rather weak. Typically, it is of Rayleigh line in terms of its intensity. Furthermore, when the frequency of radiation is sufficiently low, i.e. less than 100 cm-', this technique is incapable of discriminating Rayleigh line from Raman bands. Raman spectroscopy could be sometimes destructive. On the other hand, FTIR is nondestructive and complementary to Raman technique. It can be effective in the lowfrequency end of the spectrum (i.e. less than 100 cm-'). Moreover, it can be used to gain additional information on the vibrational modes by studying the Raman-inactive amide B,

60

Fingerprinting Insulins in the Spectral Region

253

amide IV, V, and VII bands. In the past, there have been some studies on insulin based on FTIR.'-3 However, there are limited data available in the range of 5 500 cm-' and >, 3500 cm-'. Furthermore, none of the previous studies were made directly on microcrystalline insulin. Indeed, for the most studies made in the past the range of the frequencies was limited to 1580-1700 cm-',' 600-3600 cm-1,21500-1800 cm-', 3200-3400 cm-', and 45005500 ~ m - 'Besides .~ FTIR, Raman spectroscopy was also used to study insulin samples in the range of 500-1500 cm-l and 500-1800 cm-I. It has become known that the absorption for insulins is quite high in the frequency range of 500-3600 cm". Therefore, it has been quite challenging for anyone to obtain pure free-standing pellets suitable for making transmission measurements. To avoid such a problem, we would like to find other spectral range, i.e. 3600-7000 cm-' in which the absorption coefficients are relatively lower. Therefore, it would be easier for us to prepare the free-standing pellets. Furthermore, the water absorption plays a less important role in this range. In the past, FTIR has also been used to investigate different types of the biological molecules. Using FTIR, Beetz and Ascarell? measured poly(I).poly(C), various nitrogenous bases, and nucleosides in the region 25-250 cm-l. In addition to reporting the hydration dependence of a 45 cm-' mode, they observed features in the region of -300500 cm-' that were attributed to the ribose ring vibrations. These results were consistent with the calculation^.^ Later on, Powell et al.' investigated far-infrared vibrational modes in polynucleotides on vacuum-dried, free-standing, unoriented films. Recently, Globus, Woolard et al. obtained one of the most stimulating results by revealing the phonon modes in DNA's.~ The result obtained in Ref. 9 represents one of the most comprehensive and authoritative investigations in the field, which is supported by their theory." In this paper, we present our results on the measurements and analysis of the relative absorption spectra on four insulins, i.e. human insulin, insulin lispro, bovine insulin, and porcine insulin microcrystalline samples, and protamine sulfate from salmon sperm in significantly expanded ranges of the electromagnetic spectrum using FTIR. For the first time, we have identified ten new characteristic transition peaks for insulins eight of which are in the region of 3600-7000 cm-' and two of both biochemical species are in the FarIWTHz Region (50-450 cm-'). Based on the transitions observed in the above two spectral ranges, we can easily differentiate each of the insulins from protamine sulfate. Further more, we have observed a linear dependence of absorption coefficients on frequency in the THz range (20-105 cm-I) for both types of proteins, which in our opinion is due to the presence and contribution of the boson peaks.

2. Sample Preparation and Experimental Setup

Human insulin and insulin lispro were obtained from Eli Lily, bovine insulin and porcine insulin were from Sigma-Aldrich, and protamine sulfate was from ICN Biomedicals. All of the four insulin crystals and protamine sulfate were in the powder form. For each of

61

254

R. Song, Y.J . Ding B Y. B. Zotova

the insulin species and protamine sulfate, a pair of the pellets with different thicknesses, but the same diameter of 5 mm, was produced under the same pressure. In order to make the thicknesses for each pair different, 5 mg and 10 mg powders were used to produce the pellets for the FTIR measurement in the mid-IR region (3600-7000 cm-') and THz domain (i.e. 20-105 cm-I) whereas 5 mg and 7 mg powders were used for the far-IlUTHz transition range (i.e. 50-450 cm-'). The infrared transmission spectra of these insulin and protamine sulfate samples were measured using a bolometer in the far-infraredTHz and THz ranges and HgCdTe (MCT) detector in the mid-IR range. The FTIR used in our measurements is a standard model of IFS 66viS manufactured by Bruker Optics. Each Sample was mounted inside the vacuum sample compartment with a pressure of 4-5 mbar to significantly reduce the background absorption caused by the water vapor present in the air. Each transmission spectrum was obtained after taking the average over 256 interferograms in the single-sided acquisition mode. In order to achieve the highest signal-to-noise ratios, the spectral resolutions were set to 1 cm-', 0.5 cm-', and 0.5 cm-l for the mid-IR, far-IRiTHz, and THz ranges, respectively. For each type of the biochemical species studied in this paper, the absorption coefficients were deduced from the measurements of the transmission spectra made on the pair of the pellet samples: m

where dl and d2are the thicknesses for each pair of the samples and T I and T2 are the corresponding transmittances for the same pair. Using Eq. (l),the absorption coefficients are not affected by Fresnel refraction at the front and back aidsample interfaces. Therefore, we can eliminate the contributions originated from the frequency-dependent indices of refraction. If the thicknesses are precisely known, Eq. (1) can be used to determine the absolute values of the absorption coefficients.

3. Results and Discussions In this section, we summarize our results obtained on the five biochemical species and provide in-depth discussions about them. We have divided this section into three subsections as follows. 3.1. Absorption spectra in Mid-Infrared region This region is defined with the frequencies of the electromagnetic radiation in the range of 3600-7000 cm-' (the wavelength in the range of 1.428-2.778 pm). Following our measurements, we have plotted the absorption spectra in such a frequency range in Fig. 1 for the five biochemical species. From Fig. 1, one can identify all ten transition peaks for each species with their peak frequencies summarized in Table 2. Since the four insulins and protamine sulfate are similar to the peptide group in terms of the formation of amino

62

Fingerprinting Insulins in the Spectral Region

255

acids, we assigned all these ten transitions based on the previous assignments of the similar transitions studied for the peptide group." Specifically, the first four peaks summarized in table 2 correspond to the bands originated from the combination of CH stretching near 2900 cm-' and the CH deformation in range of 1300-1460 cm-l. Therefore, the frequencies for the combination of the vibrational modes would cover the range of 4200-4360 cm-'. In this regard, it is obvious that the frequencies for the first two transitions in Table 2 are somewhat lower than those for the standard peptide group. We have noticed that the oscillator strengths for the second and third transition peaks of protamine sulfate are significantly reduced compared with those for the four insulins. The fourth transition peak for protamine sulfate is blue-shifted from the corresponding ones for the four insulin species by an amount of 27.9-39.4 cm-'. The fifth transition peak is originated from the combination band of twice of C=O stretching at 1650 cm-l and the peptide-group mode near 1250 cm-'.(i.e. the corresponding frequency of 4550 cm-I). After comparing the transition frequencies for the five species, we have realized that the four insulin species have essentially the same frequencies whereas the protamine sulfate has the frequency which is significantly red-shifted by an amount of 75.2-86.3 cm-'. In addition, the peak for protamine sulfate is barely visible in the spectrum. The combination of NH stretching mode near 3280 cm-' and peptide-group mode near 1550 cm-' yielded the sixth peak. Once again, the oscillator strength for this peak for protamine sulfate is much smaller than that for the four insulin species, making this peak barely visible. Moreover, this peak is red-shifted from the corresponding ones for the insulin species by an amount as large as 32.3 cm-I. The seventh transition peak in the spectra is the result of the combination of OH stretching near 3400 cm-' and deformation near 1645 cm-'. The oscillator strength for this combination of the vibartional modes is significantly enhanced for protamine sulfate, making it one of the most pronounced transitions in the spectrum. Furthermore, the transition frequency for protamine sulfate is red-shifted from those for the insulin species by an amount of as large as 38 cm-l. One can see from Fig. 1 that the eighth, ninth, and tenth transition peaks have lower strengths, making the accurate determinations of the transition frequencies the most difficult. The eighth and ninth transitions were assigned by us to the combination of CH symmetric stretching near 2925 cm-' and antisymmetric stretching at 2853 cm-l whereas the tenth peak was due to the overtone of NH stretching near 3280 cm-'. We have noticed that the strengths for these three peaks are reduced for protamine sulfate, especially for the eighth and ninth peaks, compared with those for the four insulin species. Since these peaks are much weaker than the rest of the transition peaks, it is not possible for us to deduce any meaningful shifts of the transition frequencies between protamine sulfate and four insulin species. Out of the ten absorption peaks for the four insulin species, identified by us in this spectral range, the only peaks at 4602 cm-' and 4840 cm-l were previously observed on the insulin ~ o l u t i o n to , ~ the best of our knowledge. On the other hand, none of the ten peaks were previously measured on protamine sulfate.

63

256

R. Song, Y.J . Ding B Y.B. Zotowa

Frequency (crn-I) Fig. 1. Absorption spectra for bovine insulin, porcine insulin, insulin lispro, human insulin, and protamine sulfate from salmon sperm, in the mid-infrared region. Dotted lines correspond to the locations of the transition peaks for the four insulin species. Five arrows mark the locations of the five transition peaks for protamine sulfate with their frequencies different from those for the four insulin species.

Table 2. Transition frequencies for ten transition peaks are given in cm-' and labeled by characteristic vibrational modes for five biochemical species. CH

C=O

NH

OH

CH

NH

It is worth noting that the absorption spectra of the four insulin species are remarkably similar to one another. The differences between the transition frequencies for the corresponding peaks reflected in Table 2 are just the errors of our measurements. Therefore, the variations of the amino acids for the four insulins summarized in Table 1 have negligible effects on their transition frequencies in the range of 3600-7000 cm-I. Although the spectral resolution for our measurements of the transmission spectra in this frequency range was set to 1 cm-I, the convolution among the linewidth for each transition and the noise levels of the mercury lamp and MCT detector placed a lower limit on the accuracy for the measured transition frequencies. According to Table 2, such

64

Fingerprinting I n s d i n s in the Spectral Region 257

an accuracy can be as good as 0.9 cm-' (the second peak) or as bad as 30.4 cm-' (the tenth peak).

3.2 Absorption spectrum in Fur-IWTHz transition region In this subsection, we report our results obtained on the five biochemical species in the spectral range between the far-IR and THz domain, i.e. 50-450 cm-' (1.5-13.5 THz). The spectra for the five species are plotted in Fig. 2. In this range, for the first time to the best of our knowledge, we have observed two transition peaks for each biochemical species. Within the experimental errors the frequencies for each of the two corresponding peaks are more or less the same for all four insulin species, i.e. 132 cm-' and 327 cm-' (the first and second peaks) respectively. On the other hand, for protamine sulfate these two peaks are blue-shifted relative to those for the insulin species. The amount of the blue-shift is about 42 cm-' and 13 cm-' for the first and second peaks, respectively. In addition, the oscillator strength for the peak at the frequency of 174 cm-l is much higher whereas that at 340 cm-' is relatively lower for protamine sulfate compared with that for the four insulins.

Frequency (an-') Fig.2. Absorption spectra for porcine insulin, bovine insulin, human insulin, insulin lispro, and protamine sulfate in the spectral range between the far-IR and THz domain. Dotted lines were drawn to demonstrate that the corresponding transition frequencies for the four insulin species were more or less the same. Two arrows marked the locations of the two transition peaks for protamine sulfate.

By comparing Fig. 1 with Fig. 2, the linewidths for the transition peaks in the farIWTHz transition range are comparable to that in the mid-infrared region. It is worth

65

258

R. Song, Y. J . Ding & Y. B. Zotova .

mentioning that we only used one pellet of protamine sulfate made from 3 mg powder to measure the transmittance. This is due to the fact that the absorption coefficients for protamine sulfate are much higher than that for the insulin species in this spectral range. As a result, a thinner pellet is quite difficult to make. After measuring the transmittance, the absorption coefficients were obtained simply by taking the natural log of the transmittances and then removing the negative sign. Such a method should be still reliable if the refractive index changes little across each transition peak. Following the previous results on p~lyglycine,'~,'~ we assigned the two peaks at 132 cm-' and 327 cm-' to the combination of C-N twist and N-H out-of-plane angle bend and the combination of C-C-N angle deformation, N-C-C angle deformation and N-C stretching, respectively. At this point, we are not sure about the origin of the transition at frequency of 174 cm-' for protamine sulfate. On the other hand, the second peak at the frequency of 340 cm-l for protamine sulfate should be also due to the combination of CC-N angle deformation, N-C-C angle deformation and N-C stretching. 3.3 Absorption spectrum in THz region In this subsection, we summarize our results obtained on the five biochemical species in the THz region, i.e. 20-105 cm-' (0.6-3.15 THz). The absorption spectra for the five biochemical species in this spectral region are shown in Fig. 3. Based on our results presented in Fig. 3, all the curves can be well fitted by linear dependences, i.e. the absorption coefficient increases linearly with frequency. To pinpoint the origin of the linear dependences, let us consider harmonic oscillators contributing to the absorption process. According to Debye model, the absorption coefficient is proportional to the square of the frequency, i.e. a = a,$ if v < < v, where vis the frequency of radiation, v, is the resonant frequency of the oscillators, and a. is a constant. Under such a condition, a/?- vs. v should be a constant. However, based on Fig. 3, a cc v, which implies that there is a vibrational density of states (VDOS) which is additional to that only determined by the simple Debye model. Such an excessive VDOS can also result in a boson peak. With the presence of the boson peak, the specific heat is actually larger than that predicted by the standard Debye model. In the past, some pioneering work was already carried out on the boson peaks.14-16For example, through Raman scattering, a boson peak was observed in amorphous polymers.15Since our insulin and protamine sulfate samples are similar to polymers in terms of structures, random interactions between individual protein chains may contribute to VDOS. We believe that such additional VDOS could be responsible for the linear dependences of the absorption coefficient on the frequency of

the THz radiation. However, due to the lower output powers of our mercury lamp in the low-frequency end, we could only observe the rising edge of the boson peak as the frequency is decreased. Obviously, additional work is necessary.

66

Fingerprinting Insulins in the Spectral Region

259

I

1

I

frequency (cm-1)

Fig.3. Absorption spectra for porcine insulin, bovine insulin, insulin lispro, human insulin, and protarnine sulfate in the THz region. Dashed lines correspond to linear fits to the measured absorption coefficients.

4. Conclusion We have successfully measured and identified twelve transition peaks in the absorption spectra for five biochemical species, i.e. bovine insulin, porcine insulin, human insulin, insulin lispro, and protamine sulfate from salmon sperm. We have demonstrated that the transition peaks are more or less the same for the four insulin species. However, the differences between the corresponding transition peaks of the insulins and protamine sulfate such as the transition strengths and frequencies are significant enough such that it may be possible for us to eventually use the spectral signatures to differentiate different biochemical species. Moreover, we have also observed the linear dependences of absorption coefficient on THz frequency. We believe that the linear dependences may be linked to the existence of the boson peaks both of which are originated from the random interactions between individual protein chains. Acknowledgement This work has been supported by U.S. Army Research Office. References 1. L. Nielsen, S. Frokjaer, J. F. Carpenter and J. Brange, Studies of the structure of insulin fibrils by Fourier transform infrared (FTIR) spectroscopy and electron microscopy, J. Pharm. Sci. 90,29-37 (2001).

67

260

R. Song, Y. J . Ding B Y. B. Zotova

2. M. Beer, G. B. B. M. Sutherland, K. N. Tanner and D. L. Wood, Infrared spectra and structure ofproteins, Proc. Roy. SOC.A 249, 147-172 (1959). 3. E. J. Ambrose and A. Elliot, Infrared spectroscopic studies of globular protein structure, Proc. ROY.SOC.A 208,75-90 (1951). 4. V. P. Drachev, M. D. Thoreson, E. N. Khaliullin, V. J. Davisson and V. M. Shalaev, Surfaceenhanced Raman difference between human insulin and insulin lispro detected with adaptive Nanostructures, J. Phys. Chem. B 108, 18046-18052 (2004). 5. C. Ortiz, D. Zhang, Y. Xie, V. J. Davisson and D. Ben-Amotz, Identification of insulin variants using Raman spectroscopy, Anal. Biochem. 332,245-252 (2004). 6. C. P. Beetz, Jr. and G. Ascarelli, Far-infrared absorption of nucleotides and poly(I).poly(C) RNA, Biopolymers 21, 1569-1586 (1982); The low-frequency vibrational modes of an RNA: Poly(I).poly(C), Biopolymers 15, 2299-2301 (1976); The low frequency vibrations of pyrimidine and purine bases, Spectrochimica Acta Part A 36, 299-313 (1980); The low frequency vibrations of the nucleosides: uridine, cytidine and inosine. Determination of vibrations associated with the ribose ring, Spectrochimica Acta Part A 36, 525-534 (1980). 7. J. M. Eyster and E. W. Prohofsky, Lattice vibrational modes of poly (rU).poly(rA): a coupled single-helical approach, Biopolymers 13,2527-2543 (1 974). 8. J. W. Powell, G. S. Edwards, L. Genzel, F. Kremer, A. Wittlin, W. Kubasek and W. Peticolas, Investigation of far-infrared vibrational modes in polynucleotides, Phys. Rev.A 35, 3929-3939 (1987). 9. T. R. Globus, D. L. Woolard, B. L. Gelmont, A. C. Samuels, B. L. Gelmont, J. Hesler, T. W. Crowe and M. Bykhovskaia, Submillimeter-wave Fourier transform spectroscopy of biological macromolecules, J. Appl. Phys. 91, 6105-6113 (2002); D. L. Woolard, T. R. Globus, M. Bykhovskaia, A. C. Samuels, D. Cookmeyer, J. L. Hesler, T. W. Crowe, J. 0. Jensen, J. L. Jensen and W. R. Loerop, Submillimeter-wave phonon modes in DNA macromolecules, Phys. Rev.E 65, 05 1903/1-11 (2002). 10. M. Bykhovskaia, B. Gelmont, T. Globus, D. L. Woolard, A. C. Samuels, T. H. Duong and K. Zakrzewska, Prediction of DNA far-IR absorption spectra based on normal mode analysis, Theor. Chem. Acc. 106,22-27 (2001). 11. K. T. Hecht and D. L. Wood, The Near Infra-Red Spectrum of Peptide Group, Pro. Roy. SOC. LondonA 235, 174-188 (1956). 12. K. Yamamoto, K. Tominaga, H. Sasakawa, A. Tamura, H. Murakami, H. Ohtake and N. Sarukura., Far-infrared absorption measurements of polypeptides and cytochrome c by THz Jpn. 75, 1083-1092 (2002). Radiation, Bull. Chem. SOC. 13. B. Fanconi, Low-Frequency Vibrational spectra of some homopolypeptides in the solid state, Biopolymers 12,2759-2776 (1973). 14. T. S. Grigera, V. Martin-Mayor, G. Parisi and P. Verrocchio, “Phonon interpretation of the ‘boson peak‘ in supercooled liquids”, Nature 422,289-292 (2003). 15. T. Achibat, A. Boukenter, E. Duval, G. Lorentz and S. Etienne, Low-frequency Raman scattering and structure of amorphous polymers: Stretching effect, J. Chem. Phys. 95, 29492954 (1 99 1). 16. S. Kojima, V. N. Novikov and M. Kodama, Fast relaxation, boson peak, and anharmonicity in Li20-B203glasses, J. Chem. Phys. 113, 6344-6350 (2000).

68

International Journal of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 261-270 @ World Scientific Publishing Company

World Scientific www.worldscientific.com

AMBIENT AIR USED AS THE NONLINEAR MEDIA FOR THZ WAVE GENERATION Xu Xie, Jianming Dai, Masashi Yamaguchi and X.-C. Zhang Center for THz Research, Rensselaer Polytechnic Institute, 1I 0 8Ihst., Troy, NY 12180, USA zhangxc@rpi. edu We report the first systematic study of broadband THz wave generation with ambient air as the nonlinear media. Generation of pulsed THz waves with mixing the fundamental and the second harmonic beams in air has been previously demonstrated by Cook et a/. and Hartmut et a1 while different groups obtained different results. To verify the proposed mechanism of strong THz wave generation, we measured dependence of generated THz field on the polarization, amplitude and phase of the individually controlled two beams. Our results confirm that four-wave-mixing rectification is the major mechanism this phenomenon, and the amplitude and polarity of generated THz wave can be controlled by the relative phase of the beams. This work is significant by providing the feasibility of THz wave generation and detection with a standoff distance greater than 50 meters. Keywords: Terahertz; air plasma

1. Introduction Recent advances in nonlinear optics and semiconductor physics have boosted THz science and technology applied in various fields including material characterization, nondestructive evaluation, imaging and testing of pharmaceuticals I . However, despite increasing demand for THz wave standoff imaging and sensing in recent years, the development of THz field applications grows relatively slowly. The major issue encumbering attempts to use THz science and technology outside of labs is the intense water vapor absorption in the atmosphere. Compared to visible light, which has 0.01 dB/km attanuation in ambient air, attanuation of THz waves is four orders higher. Using air as nonlinear medium to generate THz wave has attracted more attention recently because of its potential applications for long distance THz wave sensing and imaging. Rather than sending the THz wave from a remote source, this method enables THz wave generated close to the imaging or sensing target. As a result, the strong water vapor absorption of THz waves in atmosphere is avoided, replaced by that of the optical laser beam. Hamster et al. first observed THz wave generation from the laser focus by focusing a intense laser beam (peak power 10I2 W) into air and the mechanism is assigned to the ponderomotive force which separates electrons and ions in the ionized air to form a net dipole In the mid-90’s Cook et al. first demonstrated that the THz wave generation efficiency can be significantly improved when focusing the optical fundamental wave together with its second-harmonic (SH) wave into air Recently a THz field greater 23

’.

‘.

69

262

X . Xze et al

than 100 kV/cm has been reported by using a similar experimental arrangement with shorter laser pulses and focusing lens with short focal length ’. Power dependence measurements have been performed to verify the proposed mechanism while different groups obtained different results. All previous experimental designs utilized a thin BBO crystal (typically type I). By focusing an intense laser beam (w)from an amplifier laser system through the BBO crystal, coherent SH wave ( 2 4 can be generated and its intensity and polarization can be adjusted by changing the azimuthal angle of the BBO crystal relative to the laser beam polarization. Because the propagation direction of the SH wave follows its fundamental beam, focuses of both beams will spatially overlap each other. This perfect overlap results in a strong THz wave generation from the ionized air at the focus. As proposed by those groups, four-wave-mixing (FWM) of the fundamental and the SH wave in the ionized air is the mechanism of THz wave generation. However, as proposed third order nonlinear process, the symmetry of this phenomenon abides by the third-order nonlinear susceptibility tensor i3’ while the SH wave generation from a BBO crystal is f ’ symmetric. These two different symmetries must trade off each other to get a higher THz wave generation. As a result, simply changing the laser power and measuring the THz wave intensity may cause the inconsistency of THz wave strength as a function of laser power 4, 5 , In order to explore the mechanism of this phenomenon, we performed the first systematic study of THz wave generation in air with individually controlled polarization, phase, and amplitude of the fundamental and the SH beams. Our measurement confirms that the four-wave-mixing (FWM) in the air plasma is the major mechanism of THz wave generation and the efficiency can be controlled by the phase difference between the two beams. We also prove that the optimal efficiency of the THz wave generation is achieved when all the waves (w,2w, and THz waves) are of the same polarization which corresponds to the nonlinear tensor component i’?, in the FWM process.

’.

2. Experimental setup

Experimental setups for THz wave generation in air reported by previous publications are schematically illustrated in Figure 1. Figure l(a) shows setup of single optical beam excitation experiment as reported by Hamster et al. ’. Laser beam with peak power exceeding 10’’ W is focused to ionize air and plasma is formed at the laser focus. THz wave generation in this setup is assigned to the ponderomotive force driving electrons and ions to form a net dipole ’. Though intense THz wave generation was observed, this method has low THz wave generation efficiency considering the input strong laser field. Figure l(b) is the common setup for THz wave generation with mixing the fundamental and the SH beams through FWM process. The SH beam is generated by focusing an intense laser beam through a BBO crystal (the thickness varies from 0.1 mm to 2 mm) which is the widely used crystal for SH wave generation and then both of the SH beam

70

Ambient Air for T H z Wave Generation 263

and residual fundamental beam are focused in air and plasma is also formed at the focus to generate THz wave.

Fig. 1 Schematic illustrations of the experimental set ups for THz wave generation in air: (a) A single optical beam excitation in which the THz wave generation is assigned to the ponderomotive force separating electrons and ions to form a net dipole. (b) The common setup of THz wave generation with mixing of the fundamental and the SH waves. An intense laser beam is focused through a thin BBO crystal and the generated SH wave will focused together with the residual fundamental beam. At the focus air plasma is formed and THz wave is generated.

Our design of the individually controlled phase, polarization and amplitude of each is realized by using two dichroic mirrors as shown in Figure 2. Collimated laser beam propagates though a 3 mm thick BBO crystal and the generated SH beam is separated from the residual fundamental beam by a long-wave-pass dichroic mirror. Then these two beams propagate along different optical paths. In each beam path there are a half wave plate, an attenuator and a time delay line driven by a piezoelectric stage with a resolution of 13.6 n d s t e p (90 attosecondstep) to change polarization, amplitude and phase of each beam. After the individual control these two beams are combined again by a second dichroic mirror and the combined beam is focused by a fused quartz lens with 200 mm focal length. Noticed that there are multiple relative delays in the setup design, by scanning the relative time delay between the THz wave and probe beam, THz waveform can be obtained as the basic concept of the THz time domain spectroscopy method. Then if the time delay between the THz wave and probe beam is fixed at the peak (maximum signal) and the relative time delay z between the fundamental and the SH beams is scanned by changing their relative optical path length AZ, measured intensity of THz

71

264

X.Xae et al.

wave will be modulated as the interference pattern between the wand 2w beams since the FWM theory predicts that generated THz field is proportional to the product of the input fields. The inset shows how the intensity of generated THz wave is modulated by the relative delay zbetween the two input beams. The emitted THz wave was measured by the standard time-resolved electro-optic (EO) sampling method with a 3-mm thick oriented ZnTe crystal as the THz wave sensor

’.

Detector

Fig. 2 Experimental setup of individually controlled fundamental and SH wave for THz wave generation. Phase, amplitude and polarization of both beams can be controlled by the optics in each beam’s path. The insert illustrate how relative phase delay between the two beams affects intensity of generated THz wave.

In our experiment the Laser source is a Ti-Sapphire amplified laser system (SpectraPhysics Hurricane i) with a central wavelength of 800 nm, 120 fs pulse duration, 860 pJ pulse energy and 1 kHz repetition rate.

3. Measurement and experimental results We first repeated THz wave generation experiment following the setup illustrated in Figure 1. Plots of three THz waveforms generated in the air by different methods or different wavelengths are shown in Figure 3. The bottom and middle curves are obtained from a single beam (fundamental and SH, respectively) excitation by using the method shown in Figure l(a), and the top curve is obtained by using the method of Figure l(b). One can see the mixing of the fundamental and SH beams can generate stronger THz wave compared with a single beam excitation. In addition, a single SH beam can generate stronger THz wave compared to that generated by the fundamental beam if the amplitude of THz wave is linearly scaled by laser power. Considering that the mechanism of THz

72

Ambient Air for T H z Wave Generation 265

wave generation by a single beam excitation is assigned to the ponderomotive force which separates electrons and ions in air plasma to form a net dipole, we explain this phenomenon by the smaller focal spot of short wavelength laser beams, which has higher power density at the focus and leads a lower air break down threshold and easier to ionize air lo.

:I800 nm+'400 nm

- 592 pJ

Fig. 3 THz waveforms obtained by (a) mixing the fundamental and the SH beams in air, (b) and (c) by focusing a single beam in air (fundamental or SH), respectively. The detection is standard electro-optic sampling method with a 3-mm thick ZnTe crystal as the THz wave sensor.

With use of the individually controlled fundamental and SH wave, we explored the nature of FWM for THz wave generation in air plasma. A type-I BBO crystal has the maximum second harmonic wave conversion efficiency when the polarization of the input fundamental beam perpendicular to the eaxis, and in this geometry the polarization of generated SH beam is parallel to the BBO's e-axis. In other words, the SH beam and the fundamental beam will have their polarization perpendicular to each other " when maximum SH wave is generated. In our experiment, the azimuthal angle of the BBO crystal is fixed at the position where the maximum SH beam is generated and the residual fundamental beam keeps linear polarized. Meanwhile, the EO sampling geometry for THz wave detection is polarization sensitive. By rotating the azimuthal angle of the sensor ZnTe crystal by 90 degrees relative to the probe beam polarization only p- or s-polarization component of the THz wave is detected. In this way, we measured the polarization dependence of generated THz wave of both of the input fundamental and SH beams. Figure 4 plots eight waveforms with different combinations of the polarizations of the THz wave, fundamental and SH waves. As mentioned above, the polarizations of the

73

266

X . X i e et al.

input laser beam and the SH beam are changed intentionally and the polarization of the THz wave is measured. The four letters in the figure corresponds to the polarizations of THz wave, 2 w , w and w wave respectively in order to relate the polarization dependence of this phenomenon with the symmetry of the third order nonlinear susceptibility tensor of the media. Detailed discussion follows. In our assumption x corresponds to ppolarization and y corresponds to s-polarization, so the four plots of the left trace are equivalent to those of the right trace. One can see when both the fundamental and the SH beams are p-polarized, the emitted THz wave has its maximum amplitude also with ppolarization which corresponds to tensor component f)==, as shown in the top (xxxx) in Figure 4. A perfect isotropic media predicts zero component of f)x,yu while our measurements show its amplitude is of the same order of ,$'?x,, . In addition, one notice that the last plot (xyyy) which corresponds to ,$? ' , is not consistent with the zero susceptibility tensor component of an isotropic media, which predicts that both spolarized w and 2w beams should not generate a p-polarized THz wave. This inconsistency may come from the spatial asymmetry of laser induced plasma.

Fig. 4 Interference pattern of rectified field measured by THz pulse peak amplitude from different combinations of polarization of the w and 2w beams. The four letters in j3) subscripts represent the polarization of THz wave,

the 2 w wave and the two of

w waves, respectively, in which x corresponds to p-polarization and y corresponds

to s-polarization. The four plots in the lefi trace are equivalent to those in the left. An isotropic media forbids both q x x and xyyy components in the third nonlinear susceptibility tensor.

In Figure 5 the coherent nature of THz wave generation of this method is shown.

For illustration purpose, Figure 5(a) is a 4-fs time window of the interference pattern

74

Ambient A i r f o r T H z Wave Generation 267

from the xxxx combination in Figure 4. The solid curve in the figure is a fit of an optical oscillation with 400 nm wavelength (SH wave wavelength) which shows the amplitude of generated THz wave is modulated as the interference of the fundamental and the SH waves. As a result, one can expect that the THz field polarity can also be controlled by changing the phase between fundamental and the SH waves. We used the piezoelectric stage to change the relative path by 200 nm (corresponding to half wavelength of the SH wave or 667 attosecond phase shift between the fundamental and the SH beams for v, = T C ) and then scanned the probe beam, far field THz field waveforms with opposite polarities are obtained, as shown in Figure 5(b). From the figure one can see the THz field intensity is extremely sensitive to the phase; the noise of the peak signal is from the phase fluctuation.

i

Fig. 5 (a) The generated THz wave as interference between the wand 2 0 beams with a fine temporal resolution.

~ . rThe ) . relative phase between the w and 2 0 beams is The solid curve is a fit by cos(kz,A.l)= c o ~ ( w ~ ~ (b) changed by n (Al = 200 nm, T = 667 as) and the time delay between the THz wave and probe beam is scanned,

THz waveforms with opposite polarities are obtained.

Since our experimental design permits us to adjust the power of the fundamental and the SH beams individually, direct measurements of the emitted THz field amplitude versus each beam’s energy were performed, as shown in Figure 6(a) and 6(b). The process was to change the power of one beam with setting that of the other beam as a constant and scan their relative delay to obtain the interference pattern. Here, the y-axis THz field is the maximum amplitude picked from the interference data. Our experimental result shows good fit of generated THz wave amplitude linearly proportional to the power of the fimdamental beam and to the square root of the power of the SH beam, as shown in the figure by the solid curves.

75

268

X . X i e et al.

.

1.2 1.o

0.8 a , . U .-3 0.6 -

a

0.6

0.4 0.2

0

20 40 60 80 100 120 140 160 180 200

0.0

10

20

30

40

‘400nm

50

60

70

80

(PJ)

Fig. 6 Dependence of THz filed amplitude on the power of the (a) fundamental and (b) the SH beams. The solid curves are the linear fit and the square-root fit, respectively. Once the plasma is created, the THz wave signal follows the equationETHZ0~ ~

(

&1, ~

1

COS(V)

.

There is a turning point around 150 pJ (combined pulse energy of w and 2w beams) in Figure 6(a). This confirms that in this method THz wave generation is related to the air ionization threshold. By considering the combination of different wavelength laser beams, we estimated the power density is 1 . 5 ~ 1 0W/cm2 ’~ at the laser focus assuming the focal spot with 30 pm in diameter; this number is consistent with previous reports

’.

4. Theoretical discussion Observations in our experiments can be explained by the FWM THz rectification. Different from the second order nonlinear rectification process which has been widely used for THz wave generation from a Zinc-blende structure semiconductor material, in the third order FWM THz rectification process, there are three input beams needed and their frequencies add to nearly zero (THz frequency). When a pulsed laser is used, the nonlinear response is driven by the envelope of the input fields. The output rectified wave is the THz wave. Mathematically, this third-order nonlinear process is related with i”( R : 2w + R, -w ,-w), where R is the frequency of emitted THz wave. Predicted by the four-wave-mixing theory, the THz field has the form as ‘, (1) ET&) cc E2& - z)E;(t) E i ( t ) Here zdescribes the interference between the fundamental and the SH waves and this interference can

x(3)

In previous publications this phase factor is only determined by the different dispersion

In onlydetermined by the the different different dispersion dispersion In previous previous publications publications this this phase phase factore factor isisonly determined by

of the fundamental and the SH waves in air or inserted dispersive materials. We let the fundamental and the SH beams propagate along different paths before they are

76

Ambient Air f o r THz Wave Generation 269

recombined, as shown in Figure 2. In this way the relative phase can be arbitrarily controlled and has the form p = kz,Al = 0.1400~. And when p is changed by TI which corresponds to half of the SH wave wavelength as the optical path difference, the sign of ETHzis opposite, which is the origin of the polarity control of the generated THz wave. When describing the THz field as a function of optical beam power, Equation (2) can also be written as &Hr OC I, cos(d (3) This gives the power dependence of amplitude of generated THz wave on laser power. In Previous publications the relationship of the generated SH wave and the fundamental wave is considered as E2wK I,orETHz K 1; C O S ( ~ ), but this relationship is actually complicated when the SH wave generation efficiency is low. The symmetry of the third order nonlinear tensor i3’ determines the polarization dependence of this process. If the nonlinear media is spatially isotropic, there are three independent components in the third order susceptibility tensor: f ) x x x x , i3),x, and p)x,,, with the four subscripts corresponding respectively to polarizations of Q, 201, w, w beams; while the component x(3)xyyyis zero. Our observations revealed a small amplitude of f)x,, and relatively big fIv,,. Also air nonlinearity cannot afford such a strong THz wave generation. Considering the consistency of THz wave generation threshold with the air ionization threshold, this is explained by the enhancement of nonlinearity in air plasma and this plasma media is not spatially isotropic. Meanwhile, if i3) of ambient air is used to calculate THz wave amplitude with the 1 . 5l ~ O I 4 W / c d laser power intensity 11, the value of THz field is about four orders smaller than our measurement. This behavior reveals the laser induced plasma with a greatly enhanced f ’is the nonlinear media in which the THz wave is generated. This explains the turning point in Figure 6(a). Once plasma is generated as laser filed exceeds the air ionization threshold, THz wave generation efficiency will be improved as plasma has enhanced third order nonlinearity.

Jr2w

x(~)

5. Conclusion

In conclusion, we demonstrated that the FWM rectification in the laser induced plasma is the main mechanism of the THz wave generation in the air plasma through the use of individual control of the w and 2w beams. The polarity and the strength of the emitted THz field are completely controlled by the relative phase between the w and 2w waves. The measured THz field amplitude is proportional to the pulse energy of the w beam and the square root of the pulse energy of the 2w beam once the total optical pulse energy is beyond the plasma formation threshold. The optimal efficiency of the THz field generation is achieved when all the waves (w, 213, and THz waves) possess the same polarization in the FWM process. Finally, it is worth pointing out that if we switch the order of the THz field and the second harmonic field in the third order susceptibility in the four-wave-mixing optical process, it should be possible to measure the THz wave by using the air as a nonlinear sensor. In fact, the inverse process of FWM in the central

77

270

X . X i e et al.

symmetric materials, such as silicon and liquids, for the detection of free space THz pulse has been demonstrated 12, 1 3 . We gratefully acknowledge Dr. Jingzhou Xu and Tao Yuan for useful discussion. This work is supported by Army Research Office and the National Science Foundation. References 1 2 3 4 5 6 7

8 9 10 11 12 13

“Material for Terahertz Science and Technology”, B. Ferguson and X.-C. Zhang, Nature Material 1,26 (2002). H. Hamster, A. Sullivan, S. Gordon, W. White, and R.W. Falcone, Phys. Rev. Lett. 71,2725 (1993). H. Hamster, A. Sullivan, S. Gordon, and R.W. Falcone. Phys. Rev. E 49, 671 (1994). D.J. Cook and R.M. Hochstrasser, Opt. Lett. 25, 1210 (2000). M. Kress, T. Loffler, S. Eden, M. Thomson, and H.G. Roskos, Opt. Lett. 29, 1120 (2004). T. Loffler, M. Kress, M. Thomson and H.G. Roskos, ACTA PHYSICA POLONICA A 107,99 (2005) W. P. Leemans, C.G. R. Geddes, J. Faure, Cs. Toth, J. van Tilborg, C.B. Schroeder, E. Esarey, G. Fubiani, D. Auerbach, B. Marcelis, M.A. Carnahan, R.A. Kaindl, J. Byrd, and M.C. Martin, Phys. Rev. Lett. 91,074802 (2003). T. Bartel, P. Gaal, K. Reimann, M. Woerner, and T. Elsaesser, Opt. Lett. 30, 2805 (2005). Q. Wu, M. Litz and X.-C. Zhang, Appl. Phys. Lett. 68,2924 (1996). Tran X. Phuoc, Opt. Comm. 175 419 (2000). R.W. Boyd, “Nonlinear optics”, Boston : Academic Press, 1992. A. Nahata and T.F. Heinz, Opt. Lett. 23 67 (1998). D.J. Cook, J.X. Chen, E.A. Morlino and R.M. Hochstrasser, Chem. Phys. Lett. 309, 221 (1999).

78

International Journal of High Speed Electronics and Systems Vol. 17, NO. 2 (2007) 271-281 @ World Scientific Publishing Company

World Scientific www.worldscientific.com

TIME DOMAIN TERAHERTZ IMAGING OF THREATS IN LUGGAGE AND PERSONNEL DAVID ZIMDARS

Picornetrix, LLC., 2925 Boardwalk Dr. Ann Arbor, Michigan 481 04, United States of America dzimdars@picometrix. corn JEFFREY WHITE

Picometrix, LLC., 2925 Boardwalk Dr. Ann Arbor, Michigan 481 04, United States of America [email protected] G . STUK

Picometrix, LLC., 2925 Boardwalk Dr. Ann Arbor, Michigan 481 04, United States ofAmerica [email protected] G. SUCHA

Picometrix, LLC., 2925 Boardwalk Dr. Ann Arbor, Michigan 48104, United States of America [email protected] G. FICHTER

Picometrix, LLC., 2925 Boardwalk Dr. Ann Arbor, Michigan 481 04, United States of America &hter@picometrix. com S. L. WILLIAMSON

Picometrix. LLC., 2925 Boardwalk Dv. Ann Arbor, Michigan 481 04, United States of America [email protected]

We demonstrate a large area time domain terahertz (THz) imaging system capable of scanning 1 meter square area in less than 20-100 minutes for several security applications.. The detection of concealed explosives; metallic and non-metallic weapons (such as ceramic, plastic or composite guns and hives); and flammables in luggage, packages and personnel has been demonstrated. Transmission mode images of luggage containing threat items are discussed. Reflection mode images of luggage and personnel are discussed. Time domain THz images can be analyzed for 3 dimensional and volumetric information. Time domain THz images have advantages over coherent narrow band imaging methods, with freedom from interference artifacts and with greater ability to discard irrelevant or intervening reflections through time discrimination.

Keywords: terahertz, imaging, biomedical, reflection, security, spectroscopy

79

272

D. Zamdars et a1

1. Introduction

The use terahertz-time domain imaging’3’2334’5 (also popularly referred to as “T-Ray” imaging) to see through visually opaque materials and to spectroscopically interrogate samples is an exciting and actively investigated area. Terahertz inspection of personnel6, the detection of concealed explosives’, biological agents, chemical weapons, flammables, metallic and non-metallic weapons, and other potentially dangerous items are applications which show great promise in the arena of homeland security. Terahertz imaging has shown great potential in several biomedical areas such as burn imaging’, detection of skin cancer8, and pharmaceutical tablet imaging’. Commercially available terahertz reflection imaging is being adopted for non-destructive evaluation (NDE) applications in aerospace and other government and industrial settings. For example, NASA is currently employing terahertz reflection NDE to examine the space shuttle external tank sprayed on foam insulation (SOFI) for voids and disbonds. The development of each of these application has been limited by slow imaging speed (10s of minutes or hours), small scan areas ( 0.8 eV. Specifically, the new I-RTD (see Fig. 2) has an Inl.,Ga,As (x = 0.52) space-collector region grown upon an n+ InP substrate. This leads to bandgaps of 0.804 eV for the Inl.,Ga,As layer and 1.42 eV for the InP substrate at room temperature and prevents any significant optical absorption to 1.55 laser pulses within these regions. Since the lattice mismatch between InGaAs (5.8477A) and the substrate InP (5.8687A) is very negligible at 0.36%, the InGaAs layer can be assumed to be unstrained for the studies presented here.

Nd=3x10" cm~'

n+ InGaAs (-1urn)

Nd,=l x I O " cm~'

UndoDed InGaAs (10nm)

Well: Undoped lnGaA

I

Undoped InGaAs (10nm)

I. I

Barrier :Undoped GaSbAs

N~,=I~IO'' cm" Nd=3x10" crn.3

I

n+ InGaAs (-lpm)

n+ InP substrate

I

Fig. 2. Schematic cross section of I-RTD diode based upon ternary semiconductor material systems.

150

Optically- Triggered I - R T D Hybrid T H t Oscillator Desagn

343

The composite materials for the remainder of the I-RTD structure were selected4 to provide for: (a) a band-gap to VB electron optical-excitation that is greater than - 0.8 eV; and (b) a minimal residual gap between the second VB barrier and the spacer-region. Here, GaSb,As,., was utilized for the barrier regions and since it is not lattice-matched to the In,.,Ga,As (x = 0.52) space-collector region, and limits were placed on the amount of allowed strain (i.e., < 2.5 %) to insure that high quality hetero-interfaces are achievable. For MBE grown GaSbyAsl,/ Inl.,GaxAs interfaces it is possible to apply Vegard’s law to prescribe the strain induced in the GaSbyAsl, layer. When GaSbyAsI., is pseudomorphically grown on the chosen In,.,Ga,As (x = 0.52; lattice constant = 5.8477) material along the (001) direction, the GaSbyAsl, layers can be expected to undergo compressive biaxial strain, and as was shown previously4, the perturbations to the individual band-edge positions due to strain effects (hydrostatic and shear) can be described by simple relations derived from the Model Solid Theory.6 These models were used to show that it is most advantageous to utilize a composition of y = 0.77 which

,,

Table 2: Ino&ao 52As/GaSbo Aso 21 (units of energy: eV)

E,,

Eg2

E,Y

AUC

AUV

E

0.804

0.668

0.114

0.555

0.69

-2.44%

6,

2.24%

6Ec 0.183

~EHH ~ELH 0.076

-0.104

results in compressive strain (i.e. E < 0 ) in the GaSbAs layer and the smallest E, (-0.1 eV) while maintaining an acceptable strain-limit (i.e., E = -2.44 %) for good quality interfaces. Table 2 summarizes the resulting band-structure characteristics for new I-RTD design in terms of the band diagram given in Fig. 1, where the conduction band offset is defined as, AUC = E,, -E,,, and the valence band offset is defined as, AUV = E,, - E,, . As the main objective of this design is to enable VB-well charging (and subsequent OT-discharging) of the type illustrated in Fig. 3, the most noteworthy observations include: (i) all allowed optically-induced VB-to-CB transitions > 0.8 eV; (ii) the residual gap, E, ,and the energy difference between the VB-well state, E , , and the unoccupied states in the collector is reduced for E~ the new design which E, enhances interband tunneling; (iii) a deeper VB-well A U v which allows for greater holecharging and modulation capability; and (iv) larger ‘t CB-barriers AUc which allows for larger dynamic-range in the oscillator current density Fig. 3. The physical processes for (1) CB electron transport, (ii) Interband tunneling and (iii) Optical discharging. and power.

151

-

1

344

D. Woolard et al.

2.2. I-RTD Electron Transport Physics

This new I-RTD hetero-structure was selected because the VB of the GaSbo.77AS^.^^ barrier lies very close to the CB of the Ino.48Gao.52As and thus a strong interaction between the VB-well and CB-collector is expected (see (ii) in Fig. 3). Hence, multi-band electron transport effects will be important and they are described in a similar fashion to earlier works"2 using the six-band Kane model, except now the strain-induced effects in the GaSbo.77 As0.23 layer are incorporated using the classical Bir and Pikus approach.728 This approach, which incorporates the strain-induced shifts in the bandstructure potential energy using the coordinate-space transformations, 7 = 7 + .F. 7 , where 7 is the deformed coordinate and is the strain tensor with elements sa, and {a,p} is all permutations of { x , y , z } , is summarized in the sub-section below. Details on the applications of these models4, and a full description of their derivation', can be found in companion publications to this research effort. 2.2.1. Pikus-Bir Hamiltonian When this transformation is applied the Pikus-Bir form of our previous multi-band decoupled-Hamiltonian becomes

',

I .

1 T--P(1-2&)kz

where P,

=SE,

and we

have

, -P,

0

+ Q, = SE,,

introduced the

and -P, -Q, terms

Khh =

=SE,

-[-A2

are the strain corrections, (yl' + y 2 ' ) k 2+ 3y2' k z 2 ] and

2m0

K/h =--[3y2 A*

'

k, 2 -(y, ' + y 2 ' ) k 2 ] - s k 2from earlier derivations?

2m0

m0

2.2.2. Conduction Band Transport Model Equation (1) may be applied to derive a revised set of differential equations for the multiband wave-function projections within the strained GaSbAs barrier layers4and these may be combined with the unstrained wave-function relations and boundary conditions derived earlier in', l o to numerically solve for the wave-functions throughout the I-RTD structure. The resulting values for the amplitude of transmitted wave, t,, , can then be used as before2to derive the CB current density using,

where Af

=f(E)-

f ( E + eV,) andf(E) is Fermi function in the emitter region

152

Optically-Triggered I - R T D Hybrid T H z Oscillator Design

345

2.2.3. Intevband Tunneling Model Differential equations describing VB electron dynamics and the interband tunneling in the strained semiconductor GaSbAs layers can be also be derived from Eq. (I), and when combined with the unstrained wave-function relations and boundary conditions 1,lO . . , it is possible to derive the bound-state VB-well state, E , , and the developed earlier wave-functions associated with electron escape from the VB-well via interband tunneling into the spacer-collector region. The hole-charge density and tunneling current are related through the dynamic equation, dQ/dt = JR = J R , + JR- , where the tunneling current can be written as4

where AcR+is the wave-function coefficient for propagation into the collector and the net current is integrated over the range from the maximum in-plane wave-vector, k,, , for empty VB states (i.e. occupied by holes) to the maximum in-plane wave-vector, k,,, , that corresponds to empty states in the collector (- Fermi energy). 2.2.4. Optical Discharging Model During the optical discharging process, the variation of hole-charge is described by, is the optical discharging current which is dQ/dt = JR + J , , where J , proportional to the sheet hole-charge density. Here, the frequency of the optical-triggered oscillation can be set, and easily tuned, by the repetition rate of optical pulses. A model for optical excitation of VB holes to the free-states of the emitter3was modified to reflect the influences of strain, and for light ( F = F , cos(u, t ) ) polarized along the VB-well

=-a

interfaces ( 2 direction) y is given by

where E ,

= E, - h o p

is the escape energy (see Fig. 3) and Mhr

= ~ z u h * z yis,

the

transition matrix for holes excited from the confined HH state ah to the free HH state I,V~ (with

momentum hkhf = d2mh(ev0 - E E ) ) by light of intensity I

= ( c / 8 ~ ) n , 2 F. i

3. I-RTD Oscillator Simulation Forward biasing of the I-RTD will initiate interband tunneling that charges the VB-well region and that leads to a time-dependent modulation of the total device current The models of the previous section were applied to show (see Fig. 4) that this is both a robust and rapid process - e.g., 2.7 x 105A/cm2change in CB current over 2.5 ps due to VB well charging. In the oscillator scheme proposed, the frequency is set by an ultra-short laser pulse ( - 0 . 3 ~ ~that ) removes the stored hole-charge either at a pre-set threshold current or by a free-running pulsed process. Since this oscillation mechanism does not require the I-

153

D. Woolard et al.

346

RTD to be biased at the negative 0 010 0 50 differential resistance (NDR), lowfrequency stability is not a design $ 0 008 0 40 _* /problem of the OT-I-RTD 0 $ 0 006 -030 oscillator and this advantage leads E to larger device areas, currents, and a 0 004 ~ 0 2 0x m output power. Simulation results 6 were generated4 for an I 0 002 0 10 .c Ino.48Gao.52As/GaSbo.77Aso,23 OT-I/' , , OOOO/' RTD oscillator with a well width of 0 00 6.5nm and barrier widths of 2 nm. The emitter and collector contact was doped to Nd= 3 x 10" cm-3and the device area is chosen to be A=100 pm2. The dc biasing was V, =0.85 and the fundamental mode was set at a phase-angle o f p = x . The optical intensity of the il=1.55ym (0.8 eV) laser pulses was set to 3.5 x lo7 W/cm2. The results for 500 GHz oscillation cycles for an impedance match of 0.43+j1.12 R and room temperature (300K) operation are plotted in Fig. 5. The first-harmonic output power density is 1.45 x lo4 W/cm2. Hence, the total output power is 14.5 mW for the assumed diode area of 100 ym2, if a short-pulse 1.55pm laser with 500 GHz repetition rate were available.

-

-~

~

-

5

N

#/

m

-

~

7

'

'

'

'

'

'

'

'

'

'

'

'

'

'

'

r

70

............

07

,"terband

60

06

z -

05

"7

4 +

3

5

04

X

03

3

L

02 01

00

50 40

30 20

-0 1 -0 2 2 ~ 1 0 ' ~

3 ~ 1 0 ' ~

4 ~ 1 0 ' ~

5 ~ 1 0 ' ~

6x10

t (s)

10 2 ~ 1 0 ' ~

3 ~ 1 0 ' ~

4 ~ 1 0 ' ~

5 ~ 1 0 ' ~

6x10"

t (s)

Fig. 5. I-RTD current oscillations for 100 FmZdiode area.

Fig. 6. Holes in VB-well as a function of time

Presently, the state-of-the-art in 1.55 ym pulsed lasers possess repetition-rates of -50-150 GHz, which means this OT-I-RTD oscillator design should be able to produce (100/500) x 15 mW = 3 mW at 500 GHz, which is significantly more than has been achieved using all solid-state devices at room temperature. However, this approach is immediately amenable to further improvements that will scale with laser technology. For example, the time evolution of the holes in the GaSbAs barrier layer, @Ifor /.,this OTI-RTD design are plotted in Fig. 6. These results indicate that a 1.55 ym laser of 3.5 X lo7 W/cm2 intensity is not capable of completely discharging the VB-well (i.e., -30% more dynamic range remains) so much more power could be generated if higher intensity lasers were available. Hence, as laser technology advances, power will scale with both repetition-rate and intensity, with the only limit being power handling (thermal heating). The sections that follow will address laser integration and heating effects.

154

Optically-Triggered I - R T D Hybrid T H z Oscillator Design

347

4. I-RTD Optical-Triggering Strategy The I-RTD is composed of a Type-11-offset hetero-barrier structure having Ino,48Gao.5~As in the 6.5-nm-thick quantum well and cladding layers, and GaSbo.77Aso.z3 in the 2 nm thick barriers. The role of the laser pump is to photo-excite holes from the GaSbo.77A~o.23 barrier on the collector side of the structure. To do this without generating excess photocarriers in the InGaAs cladding regions, it is important that the laser photon energy hv be less than the Ino,48Gao.52Asbandgap, which is a0.80 eV at 300 K. This is accomplished with h -1.55 pm - a wavelength where there is a wealth of laser technology including the erbium-doped fiber components. The analysis below will show how this technology can be used to provide laser intensities up to 3 . 5 ~ 1 0W/cm2 ~ at pulse widths of -0.3 ps and repetition frequencies greater than 100 GHz. Like many quantumwell optoelectronic devices before it, the I-RTD has a polarization coupling rule that requires the incident electric field to be polarized perpendicular to the plane of the wells. One of the betterknown optoelectronic devices prior to the IRTD was the quantum-well inter-subband photoconductor (QWIP). In our experience, the simplest coupling strategy for device research is the lapped substrate facet shown in Fig. Fig. 7. Optical coupling configuration for I-RTD. 7.11-14 A chip of the device material under investigation is diced so that a 210-pm diameter device is located near one edge. The chip is then mounted on an angled jig and that edge is lapped to an angle, 6 , where S - 45" . A laser beam at il = 1.55 pm is coupled through the angled facet using the fiber-to-free-space coupler shown in Fig. 7. The objective lens in the coupler is positioned to create a focus inside the angled facet, and a component E j cos(6) of the incident electric field Ei has the correct polarization for coupling to the I-RTD. The beam propagates into the substrate with much refraction because of the high index n = 3.3 of the InP substrate. For the same reason, a higher intensity can be obtained by focusing the laser beam into the semiconductor rather than free space. These effects can be quantified through the following analysis.

4.1. Spatial and Polarization Effects One may start by assuming that the beam in Fig. 7 can be modeled as a linearly polarized TEMooGaussian mode having an electric field component of: l 5

155

348 D. Woolard et al.

where 7 = v o / n is the intrinsic impedance of the semiconductor, q0 (= 377 0) is the intrinsic impedance of free space, and n is the rehactive index. The quantity w is the “spot size”, i.e., the value of Y where the E-field magnitude is down by I/e compared to the maximum on-axis value. The locus of points where Y = w forms the Gaussian-beam profile - a useful visualization tool and what is shown in Fig. 1. The quantity wo is the minimum spot size that is assumed to occur inside the semiconductor just below the IRTD mesa. The quantity I , is the maximum on-axis intensity, which occurs in the plane of the minimum spot size. In most experiments the intensity “profile”, Z ( Y ) can be deduced by knowing the power and the approximate spot diameter 2w (easily measured with a set of calibrated pinholes). The intensity is related to the beam power through integration

I(

2n w

W

P = p . d s = I p ( r , B ) v d ~ d Q = 2 n Z , e x p - 2 r 2 / w 2 ) r d r = Z 0 ( n w 2 / 2 ) (6) S

0

0 0

The minimum spot size achievable with a Gaussian beam is a matter of some debate and always depends on the quality of the focusing optics (e.g., the degree of astigmatism). But even in non-ideal cases, it is generally agreed that a Gaussian beam can be focused to a diameter comparable to the “diffraction” limit for circular apertures, d = 2.44 f A , where f is thef-number of the optics and A is the wavelength in the medium. Substitution into (2) with w, = d/2 yields a diffraction-limited, maximum intensity I,

= 2P/(1.22. n . f

.A)’ = 2 n 2 Pl(1.22. n.f .A 0 f

(7)

Finally this maximum intensity in the semiconductor must be adjusted for two electromagnetic effects: (1) propagation at a non-normal direction relative to the aperture of absorption (the IRTD device), and (2) polarization not ideal for exciting the device transition. The first effect is calculated from the following (Lambert) theorem: the power passing through any aperture Ais just $ 0 2 , where ? is the Poynting vector S = Z k and k is the propagation unit (wave) vector. From the drawing in Fig. 1,

? 2 = S A cos(6) .

The second effect is calculated by noting that since the desired Efield component is proportional to sin(S), the intensity with the correct polarization should vary as sin2(6). Adding these effects to (7), one obtains the “available” intensity for I-RTD quantum-well absorption as:

4.2. Temporal Effects Because the desired goal of the optical excitation is to discharge one of VB quantum wells at a rate approaching 1 THz, a good candidate is a 1.55-micron fiber laser increased in pulse-repetition frequency by bit-rate multiplication. A block diagram of the approach is shown in Fig. 8. The exciter is a 40G fiber laser - now a common commercial component for optical-fiber telecommunications that produces an output wavelength

156

Optically-Triggered I - R T D Hybrid T H z Oscillator Design

349

around 1.55 micron '.PRFo to free-space and an average output m + u EDFA p l e r power of roughly 20 mW (in low-cost V units). The output I*' Stage 2nd stage 3rd stage consists of a sequence of pulses at a Fig. 8. Technique for 320 GHz pulse stream at = 1.55 micron wavelength. repetition frequency (PRF) between about 38 and 42 GHz, and pulse width of about 1 ps or slightly less. Bitrate multiplication consists of a sequence of doubling stages as shown in Fig. 8. In each stage the pulse train is divided into two equal-power portions, one of which is delayed with respect to the other by approximately half the pulse-period of the input train. The two portions are then recombined in fiber, producing an output at twice the input PRF, but with significantly reduced power because of insertion losses in the fiber divider and combiner. For the three sequential stages as shown in Fig. 8, there is a total multiplication of 23 = 8 times. Hence, a 40G laser input produces a 40x8 = 320 GHz PRF at the output. A typical 8x BRM has just under 15 dB of insertion loss.I6Fortunately, the 1S5-micron region is bestowed with a broadband amplifier - the EDFA - which readily provides about 30 dB of gain and which boosts the 320G pulse train up to the 1 W level or higher. It is straightforward to carry out a time-domain analysis for the approach in Fig. 8 and to combine the results with Eq. 8 to estimate the total available peak power and intensity from the source. We represent each pulse from the 40G source laser by a well-known expression for instantaneous power from mode-locked laser theory: P(t) = Po sech2[u(t - t,)]where Po is the (temporal) peak power. Because the pulse width is much less than the pulse period, or IPRF, a good approximation is

Po,, = PRF .Upulsewhere Upulse=

J P(t)dt = 2P0/ u .

Analysis shows that the full-

-02

width at half-maximum of the sech2 function is t ,

= 1.762/a

. Hence we can write the

expression Po = 0.88 Pave/(PW.tp ) = 0.88Pave / D where D is the duty cycle. For the

320G source output, we have t,

- l.0ps and D = 0.32 , so that for Pa,

= 1W

(limited by

Quantum-well IRTD structure

1

Power ~JV]

I0

Fig. 9. Maximum intensity at center of focused beam generated by the bit-rate-multiplier scheme of Fig. 8

Fig. 10. Rugged optical coupllng technique based on single-mode, polarization-preserving optical fiber

157

350

D. Woolard et al.

the EDFA), we find Po a 2.8 W . Fig. 9 plots the maximum intensity according to (8) versus the peak beam power P = Po for various values of the opticalf-number assuming /lo= 1.55 pm, 6= 45", and n = 3.3 (InP). In other words, the vertical axis in Fig. 8 is the intensity at the center of the TEMoo Gaussian measured at the time when the incident pulse is at its maximum power. Because of the large increase in average power made possible by the EDFA, the maximum intensity in the quantum well region lies in the range 107-to-108W/cm2 assuming the beamf-# is between 1.0 and 2.0. So it is quite possible to generate the intensities, I > 3.5~10'W/cm2, needed to completely discharge the I-RTD oscillator. 4.3. Fiber Optical Coupling The free-space coupling technique shown in Fig. 7 is probably not amenable to the field implementation that will ultimately face components integrated into THz systems. A more rugged approach is the fiber-coupling technique shown in Fig. 10. The 9-pmdiameter core of a single-mode, polarization-preserving optical fiber is bonded to the angled facet using an industry-standard epoxy. The beam from the fiber propagates into the substrate with little divergence because of the high refractive index of the InP substrate ( n 5 3.3). The spot diameter at the base of the device mesa will be -7 micron, so that with 6= 45" a majority of the power will couple into its 11-pm-diameter (100 sq micron area). 5. I-RTD Thermal Analysis

Thermal performance is one of the most important yet often neglected aspects of ultrafast device design. In the I-RTD one can expect two significant thermal mechanisms: (1) Joule heating from the electrical dissipation, and (2) optical heating from the welldischarging laser pulses. In this section we develop a -dsimple model for these and use them to predict the junction temperature TJ of the I-RTD when operating Rc at 500 GHz. This is a common metric for the Rs thermal performance of all ?7 semiconductor devices and in the present context of the (4 (b) mesa-isolated I-RTD device Fig. 11. (a) Mesa geometry, and (b) equivalent thermal model, of I-RTD shown in Fig. 11 (a), TJ is just the average temperature at the top of the mesa.

?T 3

5.1. Heat Generation Mechanisms To predict the Joule heating, it is important to note that to operate at frequencies approaching 1 THz, I-RTDs must be fabricated with high peak current density - a well-

158

Optically- Triggered I-RTD Hybrid T H z Oscillator Design

351

known fact with normal RTD oscillators. The peak current density in the baseline I-RTD structure is 2 . 7 ~ 1 A/cm2 0~ - a value close to that found in the fastest RTDs made from the InGaAdAlAs and InAs/AlSb materials systems." When undergoing relaxation oscillations whose pulse-repetition frequency is comparable to the peak-to-valley switching time, the oscillations become quasi-sinusoidal so that a good estimate of the average current through the device is I,, = Zp/2 . I 8 Similarly, since the relaxation oscillations dynamic loadline generally loops around the peak point, the average voltage is = . Hence the Joule heating is PAVE= ZAvEVAvE = ZpVp/2. As peak

vA, v,

voltage generally does not depend on area, a scale-dependent form is PAvE

= J,

A V, /2 .

To evaluate the laser heating, one must first consider the recombination mechanism for the holes in GaSbAs quantum well after they are excited into the extended valence band by the laser pulse. If confined to the quantum well, the photo-excited holes would probably recombine with an electron, yielding a radiative annihilation that would nearly conserve total energy and not generate significant heat. But in the large bias electric field of the I-RTD as shown in Fig. 3, the photo-excited hole will escape from the quantum well and drift toward the (electron) emitter side. By emitting phonons, the holes will relax down to the valence band edge in the InGaAs emitter. This drift and relaxation process should occur on the picosecond time scale, far too fast for cross-gap annihilation to occur. Once the holes are fully relaxed to the InGaAs valence band edge, annihilation will likely occur by (non-radiative) Auger recombination in the n-type InGaAs material. The time scale for Auger recombination can be much faster than the cross-gap radiative lifetime (-1 ns) in the InGaAs. Given this recombination mechanism, one can assume that all of the holes excited out of the barrier quantum well will ultimately produce heat The heat generated by the photoexcited holes Q h is then the product Q h = hvopAlq, where op is the peak hole sheet density, hv is the pump photon energy, A is the device active area, and q,is the hole lifetime. The maximum hole-sheet charge density has already been computed from the dynamic simulations of the baseline structure, as plotted in Fig. 4: op-7x10" holes/cm2, and Th = 2.0 ps. Hence, Q h =: 4.2x104.A W. Combining this with the Joule heating term, we have a total heat Q,, = V , . J ~ . A / ~ + ~ V . ~ ; A / Z ~ (9) 5.2. Thermal Resistance and Junction Temperature

The simplest way of fabricating the I-RTD device is by mesa isolation using wet chemical or reactive-ion etching. This leads to the geometry seen in cross-section in Fig. 11 (a). The heat generated in the active region must be dissipated by columnar flow down the mesa and then lateral spreading in the bottom InGaAs epilayer and the InP substrate. Heat flow in this type of geometry is well understood and can be represented by an equivalent thermal circuit containing a network of thermal resistances and capacitances, and in the steady state has the form shown in Fig. 11 (b). The columnar heat flow is represented by a thermal resistance component, Rth,C = t / d , where t is the height of the column, K is the thermal conductivity, and A is the mesa cross-sectional area. The spreading heat flow is represented by a resistance, Rth,S = (nweJ1, where

159

352

D. Woolard et al.

is the equivalent radius.” This assumes that the heat spreads from a circle at the bottom of the mesa into a half space of thermal conductivity K - not a bad assumption when the mesa is near the center of the substrate chip and is much smaller laterally than the chip. For our baseline mesa device req = (100 pm2/x)1’2 = 5.6 pm. Given these thermal resistance terms, the steady-statejunction temperature is

This function is plotted in Fig. 12 as a function of device area and has been parameterized for various values (i.e., 0.03, 0.06, 0.1, & 0.2) of the InGaAs offset region, t. The ambient temperature is TA= 25 C. Consistent with the above simulations, we assume V, = 0.85 V and Jp= 2 . 7 ~ 1 0A/cm2. ~ Note how quickly Tj varies with t, consistent with the fact that the Ino.48Ga 0.52A~proposed for the active region has rather poor thermal 200 conductivity, K , = 0.05 W/cm-K, which is a stark contrast to the InP lm substrate thermal conductivity of K S = n 0.68 W/cm-K. It is important to loo mention that the thermal model just ,f constructed is an optimistic 5o _ ~ ~ ~ ~ ~ approximation as it ignores three important effects regarding the heat 0 transport: (i) the thermal conductivity 0 50 I00 150 200 typically drops with temperature as K Mesa Area ISa _ .Micron1 = K300 (T/300)-a where a 5/4 Fig. 12. Junction temperature versus InGaAs thickness t. (Bloch’s thermal law); (ii) phonons are reflected at the InGaAdInP hetero-interface; and (iii) some InGaAs must lie below the interface, as shown in Fig. 11 (a) to provide for a low-resistance bottom ohmic contact (i.e., as most epi-ready InP substrates are semi-insulting and can not be used for contacts). Hence, if the spreading starts in the low-KInGaAs rather than the InP, the total thermal resistance will increase and so will TJ. Fortunately, the InGaAs epi-layer can be made very thin (i.e., only leaving - 30-60 nm InP transition region) and according to the results of Fig. 12 this will allow a significant margin for maintaining safe junction temperatures below 150 C. Furthermore, if secondary heating effects were to become an issue, standard TE cooling+technology can be very effective for their treatment.

-

-

-

-

6. Conclusions This paper has presented a complete design and implementation strategy for a new optically-triggered (OT) interband resonant-tunneling-diode (I-RTD) device that has potential for generating terahertz (THz) frequency oscillations and achieving enhanced output power levels under pulsed operation. This OT-I-RTD oscillator utilizes novel nanoscale mechanisms to enable a methodology for exceeding the output power levels of all other solid-state sources at ambient (300 C) and TE cooled temperatures (-73 C). A good current example of the utilization of multi-stage TE cooling is in Judson Technology InAs detectors.

160

~

Optically-Triggered I-RTD Hybrid T H z Oscillator Design

353

References 1. B. Gelmont, D. Woolard, W. Zhang and T. Globus, “Electron Transport within Resonant Tunneling Diodes with Staggered-Bandgap Heterostructures,” Sol. St. Electr., 46, 1513 (2002). 2. D. Woolard, W. Zhang, and B. Gelmont, “A Novel Interband-Rcsonant-Tunneling-Diode (IRTD) Based High-Frequency Oscillator,” Sol. St. Electr., 9,257 (2005). 3. W. Zhang, D. Woolard, E. Brown and B. Gelmont, “A Novel I-RTD Based Optically-Pulsed Hybrid Device for Generaing THz oscillations,” Pro. SPIE Symp. Opt. East, 5995,59950s (2005). 4. D. Woolard, W. Zhang, E. Brown, B. Gelmont and R. Trew, “An Optically-Triggered I-RTD Hybrid Device for Continuous-Wave Generation of THz Oscillations,” Pro. SPIE Symp.on Defense & Securitj, 6212, 621207 (2006). 5. H. Sakaki, L. L. Chang, R. Ludeke, C-A. Chang, G. A. Sai-Halasz and L. Esaki, “In,.,Ga,AsGaSb,,As, Heterojunctions by Molecular Beam Epitaxy,” Appl. Phys. Lett. 31, 21 1 (1977). 6. Van de Walle, “Band Lineups and Deformation Potentials in the Model Solid Theory,” Phys. Rev. B, 39, 1871 (1989). 7. G. L. Bir and G. E. Pikus, “Symmetry and Strain-induced Effects in Semiconductors, ” (Wiley, New York, 1974). 8. S. L. Chuang, “Physics of Optoelectronic Devices,” (Wiley, New York, 1995). 9. W. Zhang, D. Woolard, E. Brown, B. Gelmont and R. Trew, “Design and Optimization of an IRTD Hybrid THz Oscillator based upon Inl.,Ga,As/GaSbYAsl., Heterostructure Systems,” accepted to the IJHSES (2006). 10. D. L. Woolard, et.al, “Advanced Theory of Instability in Tunneling Nunostructure, ” Chap. 8 in “Terahertz Sensing Technology,” Vol.11 (World Scientific, Singapore, 2003). 11. E.R. Brown, K.A. McIntosh, F. W. Smith, and M.J. Manfra, “Coherent detection in a GaAdAlGaAs MQW structure,” Appl. Phys. Lett., 62, 1513 (1993). 12. H. C. Liu, J. Li, E. R. Brown, K. A. McIntosh, K. B. Nichols, andM. J. Manfra, “Quantum-well intersubband heterodyne infrared detection up to 82 GHz,” Appl. Phys. Lett., 67, 1594 (1995). 13. H. C. Liu, G. E. Jenkins, E. R. Brown, K. A. McIntosh, K. B. Nichols, and M. J. Manfra, “Optical heterodyne detection and microwave rectification up to 26 GHz using qnantum-well infrared photodetectors,” IEEE Electron Dev. Lett., 167, 253 (1995). 14. E.R. Brown and K.A. McIntosh, “III-V Quantum- Well Structures for High-speed Electronics”, in Advances in Research and Development , Thin Films, Vol. 23, ed. by M. Francombe and J. Vossen (Academic Press, New York, 1997). 15. A. Yariv, “Quantum Electronics,” 2”dEd (Wiley, New York, 1975). 16. See, for example, the Calmar Optics model BRM-T-16, www.calmaropt.com. 17. E. R. Brown, “High-speed resonant-tunneling diodes,” pp. 306-347 in Heterostructure and Quantum Devices, ed. by N. G. Einspruch and W. R. Frensley, (Academic, Orlando, 1994). 18. E. R. Brown, C. D. Parker, S. Verghese, M. W. Geis, and J. F. Harvey, “Resonant-tunneling transmission-line relaxation oscillator,”Appl. Phys. Lett., vol. 70 (21), pp. 2787-2789 (1997). 19. E.R. Brown “THz Generation by Photomixing in Ultrafast Photoconductors” in Terahertz Sensing Technology Vol. I (World Scientific, Singapore, 2003).

161

This page intentionally left blank

International Journal of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 355-365 @ World Scientific Publishing Company

World Scientific www.worldscientific.com

TERAHERTZ PHONON-POLARITON IMAGING FOR THE APPLICATION OF CHEMICAL DETECTION Masashi Yamaguchi, Minfeng Wang, and Pablo Suarez

Department ofphysics, Applied Physics and Astronomy, Rensselaer Polytechnic Institute, I I 0 Eighth Street, Troy,NY 12180,USA [email protected]

A combination of Terahertz (THz) polariton spectroscopy and polariton imaging technique for the application to chemical sensing is presented. We use phonon-polaritons, a coupled oscillation of the lattice vibration and radiation field, as an intense radiation source for THz spectroscopy. The propagation process of the polaritons generated in one of the two LiNbOJ transducer crystals through the sample sandwiched between the crystals is visualized using a polariton imaging technique. Partially reflected polaritons at the transducer-sample interface and polaritons partially transmitted through the sample are visualized simultaneously in a single frame of an image. The temporal profile of reflected and transmitted phonon-polaritons can be obtained without scanning the delay time between the pump and probe femtosecond laser pulses unlike THz time-domain spectroscopy which requires point-by-point acquisition of the temporal pulse profile using conventional pump-probe scheme. The results suggest possible application of this technique to the chemical sensing with fast acquisition rate. The technique has been successfully applied to the measurement of liquid and solid samples, and simultaneous measurement of multiple samples has also been achieved. Keywords: Phonon Polantons; Imaging; Terahertz Spectroscopy; Chemical Detection.

1. Introduction Using the relatively unexplored terahertz (THz) portion of the electromagnetic spectrum has recently been a topic of commercial and scientific interest in various areas.'-' The use of THz spectroscopy chemical sensing technologies is of particular interest for security and defense applications. Many chemical and biological agents have distinct low frequency vibrational and rotational modes whose frequency falls into the THz frequency range. It has been demonstrated that these absorption peaks can be used as fingerprints of particular molecules. Time domain terahertz spectroscopy (TDS) using femtosecond laser pulses is a typical experimental tool for sensing chemicaihiological agents.'92 In this type of experiments, THz radiation propagates through free space and the radiation transmitted through or reflected from the sample is detected. Simultaneous determination of amplitude and phase of the THz wave before and after transmission through the sample makes it possible to determine full dielectric response of materials.

163

356

M . Yamaguchi, M . Wang t3 P. Suarez

An alternative approach using phonon-polaritons, an admixture of lattice vibration and electromagnetic waves, has been successfully demonstrated in recent year^.'^"^,'^ In polariton-based THz spectroscopy, optically excited polaritons are used as an intense radiation source for the THz spectroscopic measurements. A family of methods including generation, detection, and control of polaritons has been developed.I6 Polariton-based THz spectroscopy is advantageous for the integration of THz chemical sensing devices because of its compact size of experimental set-ups.” For further improvement of polariton-based THz spectroscopy for the application of chemical sensing, we applied polariton imaging technique to the detection scheme of polariton-based spectroscopy to increase the acquisition rate. Fast acquisition is obvious necessity for the practical use of THz spectroscopy in real world applications. As we will show later in this article, by combining these two techniques it is possible to retrieve a THz pulse profile, which contains all necessary information for THz spectroscopy from a single frame of a polariton image, while conventional pump-probe scheme requires scanning the delay between pump and probe and detecting the THz pulse profile pointby-point. Using our method, the acquisition rate can be increased up to lo3 times. In the following section, we will outline the characteristics of polariton-based spectroscopy, and then describe the experimental results of polariton-based THz spectroscopy combined with polariton imaging technique.

2. Phonon-polaritons and Polariton-based Terahertz spectroscopy When terahertz radiation propagates through a polar crystal, it behaves in a slightly different manner from its free space propagation. In the case of electromagnetic radiation propagation through a polar crystal, the electric field of the electromagnetic wave induces the ionic motions or polar transverse optical phonons. The induced polar transverse optical phonons produce the dipole moments as they oscillate, and these oscillations of the dipole moments can emit electromagnetic waves at the oscillation frequency. The electromagnetic wave inside the polar crystal is not separable from the polar optical phonons. These linearly coupled oscillations of the lattice vibration and electromagnetic waves are called phonon-polaritons (hereafter simply “polaritons”). These coupling effects are prominent when the frequency of the electromagnetic wave is close to the polar optical phonon frequency. Usually the frequency falls into the THz frequency range. The coupling of the THz radiation and optical phonons changes drastically the dispersion relation as well as group velocities. 17 The first report on THz spectroscopy using polariton was made in 1984 by Auston et.al.’* Although the authors did not use the polariton terminology, they generated THz radiation in a lithium tantalite crystal and demonstrated the reflection spectroscopy from super lattice of semiconductor. They described the THz generation in lithium tantalite as Cherenkov radiation through optical rectification process. Later, polariton spectroscopy in transmission geometry using transient grating geometry was demonstrated by Crimmins et.al.” Paxton et.al.I2 used broad band excitation and an interferometric detection scheme for polariton spectroscopy, and applied it to the study of ferroelectric relaxor phase transition of KTN( KTal.,Nb,03) as a practical terahertz spectroscopic method. They described the THz wave generation process as impulsive

164

Terahertz Phonon-Polariton Imaging for Chemical Detection 357

stimulated Raman process.20Wahlstrand et.aL21 showed that the existing two models of THz wave generation in lithium tantalite crystal, Cherenkov radiation and impulsive stimulate scattering, can be described in unified form. Polariton based approach has significant advantages over traditional THz spectroscopy based on free space propagation of THz radiation. These advantages include: Neither THz optics, nor THz detector required. Compactness of the physical size of the imaging cell. Spatial resolution is determined by the optical probe wavelength. Controllable THz waveform through pulse shaping techniques. Reference and sample signal for the absorption experiments measured in a single experiment unlike traditional free space based THz spectroscopy. In this section, we will describe the novel polariton spectrometer detected with polariton imaging technique. The use of polariton imaging technique for the detection of polaritons has notable advantage of fast acquisition rate, as well as simultaneous measurement of reflected and transmitted polaritons from samples, which increases the accuracy of the determination of dielectric response of the samples. 2.1. Polariton based spectroscopy with polariton imaging detection

The “Spectrometer” consists of two transducer crystals and samples placed between these two crystals with physical contact. The schematics of the experimental configuration is shown in Fig. 1. Optically generated polaritons in one of the polar transducer crystals travel towards the sample, and at the edge of the transducer crystal polaritons are partially reflected from the surface of a sample placed adjacent to the transducer crystal, while the rest is transmitted into the sample material. A polariton propagating into a non-polar sample material becomes an electromagnetic wave without coupling to a particular phonon mode in the sample, and is subject to the damping and phase shift as it propagates in the sample due to dielectric properties of the sample material. As soon the electromagnetic wave reaches the other LiTa03 transducer crystal, it will couple to the A1 optical phonon mode and form phonon polaritons again. In this study, the propagating polaritons were detected using the polariton imaging technique24. In our imaging experiments, the diameter of the probe pulses is chosen to be large enough to cover the entire sample, and the images are detected by a CCD camera. THz polaritons induce the change of phase of these probe pulses when they reach the probe positions. The relative phase change of the probe pulses can be detected as a change of the interference pattern of these two pulses. Real space images of the propagating polaritons are recorded at different times by the probe pulses that are variably delayed relative to the pump pulses. The samples can be in liquid or solid form.

165

358

M . Yamaguchi, M . Wung B P. Suarez

2.2. Polariton generation Ultrafast laser pulses were used to excite phonon-polaritons through the Sample Impulsive Stimulated Raman Scattering (ISRS) process.20 The polariton formation can be verified by measuring the dispersion relation of the polaritons where the effect of the coupling between the excited optical phonons and the terahertz radiation appears clearly. It has been reported that Al optical phonons in LiTa03 generated through ISRS satisfy the dispersion relation of phononpo1aritons2'. We must note that there exist alternative models that describe generation of phonon-polaritons by the difference frequency generation or optical prbbe rectificationz2. In addition, when the phase velocity of the pump pulse is greater than Fig.1 Schematics of polariton-based THz the group velocity of THz radiation, the spectrometer. Phonon-polaritons are generated in emission can also be interpreted as one of the LiTaO, transducer crystals by line focused femtosecond laser pulses. Excited Cherenkov radiation''. polarions propagate towards the sample and are In order to excite phonon-polariton partially reflected from and partially transmitted by ISRS, the mode needs to be infrared through the sample. Delayed probe pulse with active in addition to being Raman active large diameter covering the entire crystals is used to image the polaritons. since the phonon mode needs to couple with THz radiation field to form phononpolaritons. This situation is realized only in crystals without inversion symmetry. In this study, L i m o 3 crystals were chosen for the transducer crystal and the lowest-lying Al mode was used for the excitation. Barker et.al. reported that the most of the oscillator strength in the low-frequency region is concentrated in this modez3,hence the strength of excited phonon-polaritons related to this branch has the strongest electric field. Transducer crystals need to satisfy the abovementioned criteria, however, sample materials need not satisfy those criteria, and are not limited to polar materials". Polaritons are generated in one of the transducer crystals by femtosecond laser pulses. Pump beam is line focused in order to excite a plane wave of polaritons rather than spherical wave (In Fig. 1, the pump pulse is represented by a vertical line). Generated polaritons propagate in a direction slightly tilted from the incidence plane of optical light. The tilted angle is about 19 degrees for LiTa03 crystals, as will be described in more detail below. In our experimental configuration, polaritons are generated in polar crystals such as LiTa03 and LiNb03 using femtosecond laser pulses, and the excited polaritons propagate toward the sample crystal.

2.3. Polariton propagation and forward angle Excited polaritons propagate in the direction tilted from the direction of optical pump pulses. The propagation direction of polaritons is determined by the ratio of the

166

Terahertz Phonon-Polariton Imaging for Chemical Detection 359

speed of the pump pulses and polaritons. An intuitive explanation for this is as follows. The optical pump pulse at a certain instance can be considered as a point source of polaritons. The pump pulse excites a series of spherical polariton wavelets continuously as it propagates, and the excited polaritons start propagating away from the excited positions as spherical wavelets centered on these positions. The resulting wavefront, which is a superposition of these wavelets, is tilted from the propagation direction of the pump pulse20. In the case of LiNbO3 crystal, THz polaritons propagate 24 degrees away from the face of the transducer crystal. The effect of this tilted propagation causes the difference of the apparent propagation distance on the imaging data and real propagation distance. This effect has been considered in the quantitative analysis of polariton spectroscopy, otherwise it can cause inaccurate determination of the real and imaginary parts of dielectric function12. The polariton causes change of dielectric constant at the location through the Pockels effect, and this change is proportional to the amplitude of the polariton. The probe pulses experience the change of their phase due to the induced change of the refractive index. With the use of phase sensitive detection of the transmitted probe light, time dependent spatial profile of the terahertz polaritons can be monitored. LiNbOj [ Since the group velocity of Optical probe pulse $>-. ^ Wave front of polaritons is comparable to that of the THz polaritons at t=t| probe pulses, the phase matching condition between the probe pulses and polaritons is Wave front of important ". Figure 2 shows the top view of polaritons '' Optical Probe pulse ,-fc~~+ the transducer crystal and also the phase '• at t=t >t at t=t matched propagation of optical probe pulses and polaritons. Optical probe pulses Fig.2 Phase matching condition of optical probe propagate perpendicular to the surface of pulse and polaritons. The top view of crystals is the transducer crystals, while polaritons shown, and the propagation of optical probe and propagate in the tilted direction (24 degrees

u

THz polaritons is indicated.

u^boj

for

Suppose the optical probe

pulse and the wavefront of polaritons overlap at the position marked with the circular symbol at t=tj. If the angle between the probe pulses and polaritons is properly chosen, as the probe pulse propagates in vertical direction the wavefront of polaritons propagates in the direction tilted from vertical, as shown in Fig. 2. The wavefront of polaritons always overlaps with the probe pulse as it propagates through the crystal. The probe pulse accumulates a phase shift as it sees the tilted wavefront of polaritons continuously while propagating through the crystal. This phase matching condition can be automatically satisfied when the probe pulses have the same wavelength as pump pulses and propagate in the same direction.20 When the wavelengths of the pump and probe are different, the propagation direction of the pump and probe pulses needs to be tilted relative to one another. The signal intensity increases as the crystal thickness increases until the thickness reaches the size of the Rayleigh range of the focusing lens.

167

360

M . Yamaguchi, M . Wang B P. Suarez

2.4. Polariton detection and polariton imaging

Polaritons in transducer crystals can be monitored through different physical mechanisms such as diffraction from polar it on^'^, induced phase, and polarization change of the probe pulses. All of these techniques can provide both amplitude and phase information of polaritons simultaneously, as in THz TDS (Time Domain Spectroscopy). In this work, we employed the polariton imaging technique to obtain two-dimensional image of the spatial amplitude patterns of the polaritons in transducer crystals.24 The polariton imaging is based on the phase contrast imaging technique24. The dominant effect of THz polaritons on the propagation of optical probe pulse is the phase shift due to the polariton induced local refractive index change due to the Pockels effect. The electric field of the polariton changes local refractive index and the change is proportional to the electric field. Hence, the spatial amplitude pattern of polaritons is replicated to the spatial pattern of the phase shift of transmitted probe pulse. Phase contrast imaging is used to convert the phase pattern into the intensity pattern so that a CCD camera can record spatial patterns of the electric field of polaritons. Talbot imaging technique has been used to convert the spatial phase patterns into amplitude patterns. Talbot self-imaging was extensively reviewed by Patroski for the case of the periodic spatial amplitude and phase patternsz5. It has been shown that there is a contribution to the intensity from the phase shift in the planes away from the focal plane. Practically, Talbot imaging can be realized by simply displacing the CCD camera from the image plane of the optical system. The drawback of this method is that optimal image and optimal contrast do not coincide and the image typically becomes blurry. This drawback can be avoided by applying another phase contrast imaging technique such as interferometric imaging.

3. Experimental Delayed probe pulse with latqe diameter

Nonlinearcrystal lor SHG

Ti-sapphire laser Reptation rate 1 KHz Pulse width 1 mJlPulre Wave length 800nm

Fig.3 Experimental setup for the polariton spectroscopy detected with the Talbot imaging technique. The imaging cell consists of the sample and the transducer LiNb03 crystals. A sample is placed between the two transducer crystals.

Figure 3 gives experimental setup for the polariton imaging. Tisapphire laser output from a regenerative amplifier is split into the probe and pump pulses. The pump pulse (800 nm) is line focused into LiNb03 crystals using cylindrical lens to excite plane wave polaritons. Probe pulses are variably delayed along a computer controlled translation stage, and then frequency is doubled by P-barium borate crystal (BBO). Probe pulses with a large diameter spot size of 15-25 mm were used to obtain the image of the entire

crystals. The position of the CCD camera is placed slightly out of the image plane to visualize phase modulation of probe pulses as an amplitude pattern through Talbot

168

Teruhertz Phonon-Po~ur~to7~ ~ q ~ ~ fgo r~ Chemical n g Detection. 361

imaging24.The iise of different frequency for the pump and probe pulses allows to filter out unwanted scattered strong pump pulses and improve the image contrast By c h a ~ g i ~ g the delay time of the probe pulse, a time resolved “movie” of polariton propa~ationcan be recorded. Sample materials are physically sandwiched between the two LiNbOs t r a n s ~ u ~ e r crystals. For liquid samples, they were filled in a thin plastic bag sand placed between the transducer crystals.

4. Results 4.1. Polariton imaging in LiNbO crystal and polartion spatial profile Polariton imaging allows determining the spatial amplitude pattern of the polaritons in a

Fig.4 Polariton images in LiNbQ3 at delay time of (a) t=5.8 ps and (b) F27.6 ps ,and the poladon spatial profiles at (c) e 5 . 8 ps and id) F27.6 ps.

single frame of the image. The intensity of the transmitted probe pulses is ~ ~ d u ~ by a ~ e d the phase shift due to the polaritom. Figures 4(a) and (b) show examples of the polariton

169

362

M. Y a . ~ ~ ~ g u cM h a. ,Wung &' P. Sua.rez

i ~ a g ~ data n g in Li%03. The two images are taken at a different time delay between the pump and probe pulses. The counter propagating two polariton pulses are shorn in the figures. The wavefronts of the polaritons appear as a vertical line. This is because the polaritons were excited by a line source. Polaritons with different waveforms also can be excited, which will expand the cap~bil~ties of polariton spectroscopic ~ m a g i ~The g . probe pulses acquire phase mo~ulationwhile they propagate through the sample, md the phase ~ e in d Fig.4. The mod~lat~on is converted into the intensity modula~ionof ~ ~ s m i light spatial profile o f the THz polaritons plotted in Fig. 4(c) and (d) are o b ~ i n e dfrom the same image data in Fig. 4(a) and (b), respectively. A single frame of the image data contains the i n f o ~ a ~ i oofnthe spatial polariton profile.

4.2. THz ~ o ~~ ~~ ~e ~~~ et~ ~#~ewith ~~s ~t ~e o~o ~~ ~y ~ E ~ t Q o ~ ~ ~ The use of p o ~ a r ~ t ~imaging on technique for the detection of THz polarition s ~ e c ~ s isc advantageous ~ p ~ due to the fast a ~ ~ u i s rate, ~ ~ ~asowell n as the p o s s ~ obf ~ ~ ~ simultaneous measurement of reflected and transmitted polaritons From the sample. Figure 5 presents polariton image data taken in the three crystal ~ ~ n ~ ~ u r a i t Q n shown in Fig. 1 . In this experiment, sapphire crystal was used as a sample and was sandwiched between two LiNbQ3 transducer crystals. The thickness o f the sapphire plate was li mm. The contacting surfaces of these crystals were optically flat, and the three crystals were clamped together to provide physical contact with one another. W e n the surfaces are not in contact, polaritons will be totally reflected at the air-l.iNbQ3 interface because the ~ n c ~ ~angle e n t to the surface will be beyond the critical angle due to the forward angle. The "bad" contact is relatively easy to find out practically. ~~~~~~~

ritsns tz5.8 ps

L

I

~~~~

Horizontal position

(W

(4

Fig. 5. a. Polariton images at different time delay after pump pulse. Polaritons in LiNbO3 crystal are clearly seen. Static dark spots and lines in the images are physical scratches on the crystal surface. b. Polariton amplitude YS. horizontal position in crystals. Polariton amplitude i s proportional to the intensity (or grayscafe) of the images in Fig5 (a). The peak shapes represent spatial profile of THz polariton amplitude at the time delay.

P 7Q

Terahertz Phonon-Poluriton Imaging for CFhemical Detection 363

Figure 6 shows the polaritons generated in the EiNb03 crystal on the rightdefects on hand side of the figure propagating towards the lefthand side and being patially reflected at sapphire~EiNb~3 interface, with the rest of the into wave propagating sapphire. At the interface of 9' the- sapphire plate on the ~ r ~p ~ i a~r i ~ o~Reflected ~ s ~ polaritom ~ ~ t t ~ ~ other side, polaritom are partially transmi~~ed to Fig.6 Polanton imaging of sapphire (A12O3) between two transducer ClptalS. LiNbcB3 on the left and partially reflected back to sapphire The image of polaritons i s shown at the instance when the partially ~ a n s m i t t e ~ pol~ssitonsare in LiNbQl on the left-hand side, and the reflected polaritons are in Limo3 on the r ~ ~ h t side - ~ofthe ~ d figure, as indicated by the arrows. The horizontal arrows in Fig. 6 indicate the distance from the L i ~ Q ~ - s a p p h ~interface re to the respective wavefronts of reflected and transmitted polaritons. The difference of the m o w Lengths is due to the difference of refractive indexes of sapphire and LiMQ3, which determines the ~ ~ ~ a velocity. ~ t o nThe a ~ p l ~ t u d of e s the reflected and ~ a n s m i t ~ polaitons ed are related to the damping in LiNbQ3, sapphire and the reflectivity at surfaces. Figure 6 indicates that a single image of ~ o l a ~ t o nats a certain instance contains information sf both refractive index and damp in^. Fourier analysis of the spatial profile of reflected and ~ a n s polaritons ~ ~ ~ further e ~ provides frequency dependent complex dielectric fmction. ~ i ~ u l ~ a n e~o euass u r e ~ eof n ttransmission and reflection further increases the r e l i a b i ~ i ~ of the d e ~ e ~ ~ n ~oft i the o n dielectric response h c t i o n . The spectroscopic imaging ensures s i ~ i ~ c ~faster n t ~acquisition y time. The entire spatial pattern of polariton pulse i s obtained in a single frame of imaging m e a s u r ~ ~ e nwhich t , normally takes ~ i~ lis e c o n d s to seconds. Only two fkames of the image (spatial profile of plaritons before and after they propagate through the sample) are required to obtain fiall dielectric response in the Tmnsrntttd R+Lrle*d imaging meas~e~e~~s. ,/. Simultaneous spectroscopic Physical

I

-

< .

4 ~

l

~

~

~

~

(

~

H

,

~

Fig.7 Polmiton spectroscopicimaging of liquid sample (hexane) placed m a plastic bag. Transmitted and reflected polaritons are shown.

171

~

i

n

mcaswmerits of iriultiple samples habe also k e n denionstratrd, where sapphire and 1.iNbQ3 plaies were stacked w-tically and sand\vichcd btwteen two LiNbO, transducer crystals. The results of thew ~measurements ~ s t ~ ~ are p ~ e s e n ~ ~ ~

in detail elsewhere. The capability of sssimu~taneous ~ e a s u r e ~ e n t of s ~ u ~ t samples is important for

~

364

M. Yamaguchi, M . Wang & P. Suarez

chemical and biological sensing applications, as well as the capability of measuring liquid samples and solutions.2 Figure 7 shows an example of the measurement of a liquid sample, where hexane was placed in a plastic bag between LiNb03 transducer crystals. Both transmitted and reflected polaritons are shown in the image in Fig.7, as indicated by vertical lines. A single frame of the polariton image contains all information necessary for linear THz spectroscopy, and together with fast acquision rate this provides significant advantage over conventional THz time-domain spectroscopy. With these experiments we have demonstrated the advantages of the polariton imaging technique and showed that it can be applied successfully to the measurement of both solid and liquid samples, and multiple samples as well. References 1.

2. 3. 4. 5.

D. L.Woolard, W. R. Loerop, and M. Shur (Ed.), Selected T0pic.s in Electronics and Systems, “Terahertz sensing technology”, Vol. I& V01.2 ,World Scientific, New Jersey, (2004). B. Ferguson, X.-C. Zhang, Materials for terahertz science and technology, Nature Materials, 1 , 2 6 - 33, (2002). C. Zandonella, Terahertz imaging: T-ray specs, Nature, 424, 721 - 722, (2003). K. Wang, D. M. Mittleman, Metal wires for terahertz wave guiding, Nature, 432, 376 - 379, (2004). R. Huber, F. Tauser, A. Brodschelm, M. Bichler, G. Abstreiter, A. Leitenstorfer, How manyparticle interactions develop after ultrafast excitation of an electron-hole plasma, Nature, 414, 286 289, (2001). H. Koehler, A. Tredicucci, F. Beltram, H. Beere, E. H. Linfield, A. Davies, R. Giles, A. David, R. Iotti, and Rossi, Fausto, Terahertz semiconductor-hetcrostructure laser, Nuture,417, 156 -159, (2002). G. L. Carr, M. C. Martin, W. R. McKinney, K. Jordan, G. R. Neil, G. P. Williams, Highpower terahertz radiation from relativistic electrons, Nature, 420, 153 - 156, (2002). R. A. Kaindl, M. A. Carnahan, D. Hagele, R. Lovenich, D. S. Chemla, Ultrafast terahertz probes of transient conducting and insulating phases in an electron-hole gas, Nature, 423, 734 - 738, (2003). L.Young, V. V. Prabhu, E. W. Prohofsky, Calculation of far-infrared absorption in polymer DNA, Phys.Rev.A, 39, 3173-3180, (1989). D. L. Woolard, T. Koscica, D. L. Rhodes, H. L. Cui, R. A. Pastore, J. 0. Jensen, J. L. Jensen, W. R. Loerop, R. H. Jacobsen, D. Mittleman, and M. C. Nuss, Millimeter Wave-induced Vibrational Modes in DNA as a Possible Alternative to Animal Tests to Probe for Carcinogenic Mutations, J. Appl. Toxico., 17, 243-246, (1997). M. Walther, B. Fischer, M. Schall, H. Helm and P. Uhd Jepsen, Far-infrared vibrational spectra of all-trans, 9-cis and 13-cis retinal measured by THz timedomain spectroscopy, Ckem. Phys. Lett., 332,389-395, (2000). B. J. Paxton, M. Yamaguchi, and K. A. Nelson, Phonon-polariton based THz spectroscopy, Ultrafast Phenomena XIV, T. Kobayashi, T. Okada, T. Kobayashi, K.A. Nelson, and S. De Silvestri , eds., pp. 236-238, Springer-Verlag, Berlin, (2005). B. Ferguson, X.-C. Zhang, Materials for terahertz science and technology, Nature Materials 1,26-33 (2002). R. M. Koehl, S. Adachi, and K. A. Nelson, Real-space polariton wave packet imaging, J. Chem.Phys., 110, I3 17-1320, (1999). N. S. Stoyanov, D. W. Ward, T. Feurer, and K. A. Nelson, Terahertz polariton propagation in patterned materials, Nature Materials, 1, 95-98, (2002). T. Feurer, J.C. Vaughan, and K.A. Nelson, Spatiotemporal coherent control of lattice vibrational waves, Science 299, 374-377 (2003). For example, Polaritons, Ed. E.Burstein, and F.Demartin, (Pergamon Press, New York, 1974). ~

6.

7. 8. 9. 10.

11.

12.

13. 14. 15. 16.

17.

172

Terahertz Phonon-Polariton Imaging f o r Chemical Detection 365

18. D.H.Auston, K.P.Cheung, J.A.Valdmanis, and D.A.Kleinman, Cherenkov Radiation from Femtosecond Optical Pulses in Electro-Optic Media, Phys, Rev.Lett., 15, 1555-1558( 1984). 19. T. F. Crimmins, M. J. Gleason, D. W. Ward, and K. A. Nelson , A simple terahertz spectrometer, Ultrafast PhenomenaXII, T. Elsaesser, S. Mukamel, M. M. Murnane, and N. F. Scherer, eds. pp., 221-223 (Springer-Verlag ser. Chem. Phys., V. 66,2001). 20. T. F. Crimmins, N. S. Stoyanov, and K. A. Nelson, Heterodyned impulsive stimulated Raman scattering of phonon-polaritons in LiTa03and L i m o 3 , J.Chern.Phys., 117,2882-2896, (2002). 21. J. K. Wahlstrand and R. Merlin, Cherenkov radiation emitted by ultrafast laser pulses and the generation of coherent polaritons, Phy.Rev.B., 68, 054301(2003). 22. M. Bass, P. A. Franken, J. F. Ward, and G. Weinreich, Optical rectification, Phys.Rev.Lett., 9, 446-448, (1962). 23. A. S. J. Barker, A. A. Ballman, and J. A. Ditzenberger, Infrared Study of the Lattice Vibrations in LiTa03, Phys.Rev.B, 2,4233-4239, (1970). 24. R. M. Koehl, S. Adachi, and K. A. Nelson, Real-space polariton wave packet imaging, J. Chem.Phys., 110, 1317-1320, (1999). 25. K. Patorski, The self-imaging phenomenon and its applications, in Progress in Optics, E. Wolf, ed., ), XXVII, 1-108 (North-Holland, Amsterdam, 1989).

173

This page intentionally left blank

International Journal of High Speed Electronics and Systems Vol. 17, NO. 2 (2007) 367-382 @ World Scientific Publishing Company

\e

World Scientific www.worldscientific.com

NEW TECHNIQUE TO SUPPRESS SIDELOBE CLUTTER IN PERIMETER SECURITY SYSTEMS

G. W. WEBB Institute f o r Pure and Applied Physical Sciences University of California Sun Diego, 9500 Gilrnan Drive, La Jolla, CA 92093-0360, USA [email protected]

I. V. MININ Novosibirsk State Technical University, Russia 20, K.Marksa Street, Novosibirsk, 630092, Russia iapp@sibmail. ru

0.V. MININ Novosibirsk State Technical University, Russia 20, K.Marksa Street, Novosibirsk, 630092, Russia

[email protected]

False echoes degrade the operation of radar and imaging antennas.'The false returns or clutter arise from antenna sidelobes and raise the threshold of detection in perimeter security systems. Accordingly there is great interest in reducing the sidelobes in present day millimeter wave and future terahertz antennas. Here we describe a new technique to suppress antenna sidelobe returns. The technique exploits our ability to distinguish between the phase of desired signals arriving in the antenna main beam and the phase of undesired clutter signals coming from the sidelobes. We demonstrate that through modulation of phase and taking the Fourier transform of the received signal, we can preferentially suppress the clutter return relative to the desired main beam signal. Suppression of average clutter return of over 25 dB is found. Index words - radar antennas, antenna theory, phase modulation, clutter suppression

I. Introduction Clutter is defined as unwanted "echoes" or false returns in the operation of a radar system'. These false echoes complicate the operation of a radar system and reduce its performance. Examples of such systems are perimeter security systems, meteorological radars, air traffic control radars, and tracking radars of many types. A valid radar return from a target involves energy transmitted in the main beam of the radar antenna and received back as a valid echo in the main beam. Fig. 1 shows a schematic of different types of returns, both wanted and unwanted.

r"d 1

multipath

single path

SL-SL

Fig. 1. Schematic transmitheceive antenna with reflected radiation: a) single path mainlobe to mainlobe b) single path sidelobe to sidelobe, c) multipath mainlobe to sidelobe. All paths are reciprocal. Multipath involves more than one reflection.

175

368

G. W . Webb, I. V. Mznzn & 0 . V . Minin

The false echoes arise from transmissions or returns at angles outside of the main beam (mainlobe) due to a variety of processes. One type of false return process arises from radiation transmitted out of a radar sidelobe that is reflected from an object, terrain, etc. and returns to the radar through the sidelobe. Here we refer to any response from an angle outside of the antenna mainlobe as a sidelobe. Another process, multipath scattering, involves radiation transmitted by the main beam which is reflected from the target to the ground then received by the radar's sidelobe. This process produces a ghost image on the radar screen far in angle from the real target but which moves with the real target. The common feature of these sidelobe returns is that the radar displays them as clutter or a false target images in the beam direction. The standard technique for reducing a radar antenna's sensitivity to clutter is to reduce sidelobe level by carefully shaping the feed illumination pattern of the antenna2. We have discovered a new technique that greatly suppresses signals received from out of the main beam. It can be applied in addition to standard techniques. Our technique has grown out of our work on a new type of reference phase in some types of antennas and our application of this technique to greatly reduce multipath fading in communication antenna systems. 11. Reference Phase And Multipath Fading

Communication antennas are important examples where fading reduction is desired. In general, a microwave communications antenna receives a signal from a transmitter that is a sum of the desired line-of-sight (LOS) signal and of non-line-of-sight (NLOS) signals. The NLOS signals arise from reflections off structures, terrain, etc. and from diffraction off of obstacles. Importantly, the NLOS signals can be coherent with the LOS signal. Under real world circumstances the LOS and NLOS signals can be received with nearly equal amplitude and nearly 180" out of phase producing destructive interference known as multipath fading3. We have developed a new technique to combat multipath fading in communication antennas which we apply here to radar applications. This technique is based on the existence of a free parameter in the design of Fresnel zone plate antennas, shown schematically in Fig. 2. Historically, zone plate antennas have been designed with a specific choice for this parameter, which can be taken as a type of phase reference.

phase = (rl

360"

+ 1-2 - R) --h

@ ref

Fig. 2. Schematic ray through antenna aperture between source S and detection point P. R is the shortest = distance between S and P when S and P are co-axial with the aperture. Historically, the implicit choice of 0" for R has been made.

176

Suppressing Sidelobe C l u t t e r in P e r i m e t e r S e c u r i t y S y s t e m s

369

We identified two methods of interpreting the parameter, either in terms of a reference radius4 or equivalently a reference phase5, as shown in Fig. 2. In a somewhat different but related context, a phase offset was introduced in the optimization of continuous-relief diffractive optical elemend. Here, for simplicity, we treat the free parameter as a reference phase in the calculation of ray phase in zone plate antennas. Previously reference phase in zone plate antennas was implicitly assumed to be zero, which we will refer to as the standard value. With non-standard choices of reference phase the actual configuration of the zone plate is changed as indicated in Fig. 3.

a/3

0

200

400

600

R, rel.unit Fig. 3. Phase profile vs. radius, modulo 360", for three different values of B,,f. The bottom curve is for the standard zone plate c o n s t r ~ c t i o n ~ ~ ~ .

We have established both experimentally at 39 GHz and theoretically that shifts the LOS beam phase with a positive linear slope through 360" as variation of shown in Fig. 4 with only slight changes in antenna

Main Beam Phase vs Reference Phase

0 beam

0

60

120

180 240

300 360

0ref

Fig. 4. Main beam phase ebeam at the receiver of the FZP antenna vs. phase reference BrCf.Note that the variation of Abealn is 360" for a 360" variation of BrCf',The phase of the carrier has been suppressed.

177

370

G. W. Webb, I. V. Minin 63 0. V. Minin

Of particular importance is that varying the reference phase does not shift the phase of NLOS signals which enter the antenna from outside the LOS main beam by an equivalent amount. For example, the phase of the 20" off axis NLOS signal does not change by an equivalent amount in Fig. 5 or can vary with a negative slope at very large NLOS angles. The difference in phase sensitivity of LOS and general NLOS signals to erefis exploited to eliminate MPF. Sidelobe Phase ONLOS vs Reference Phase Ores 0

-40

-80 ONLOS

-120

-160

-200

0

60

120

180

240

300

360

Ores

from 20" sidelobe at the receiver of the FZP antenna of Fig. 2 vs. phase Fig. 5. Example of signal phase reference OrCf.Note that the total variation of ONLOSis only about 40" for a full 360" variation of Or+

The difference in phase sensitivity is important under conditions where there is significant multipath fading. We showed earlier that under conditions of destructive interference between LOS and NLOS signals the reference phase could be adjusted to remove the conditions for destructive interference in a communication a n t e ~ m a ~Here '~. we first consider static changes in reference phase and then modulation of the reference phase in time to discriminate between the main beam (LOS) signal of a radar antenna and the out of main beam (NLOS) clutter signal. Iii. Mean Sidelobe Level With Static ORef First we calculate the radiation pattern for FZP antennas with different values of reference phase in the Fresnel-Kirchhoff scalar amplitude formalism:

u

P

[cos(n, '1)

=--

178

-

cos(n, r2)ldA

Suppressing Sidelobe C l u t t e r in P e r i m e t e r S e c u r i t y S y s t e m s

371

Up is the amplitude of the optical disturbance at detection point P due to radiation of wave vector k emitted from a source and passing through antenna aperture A, as shown in Fig. 2. The integral extends over the aperture which contains a Fresnel Zone Plate of the opaque zone type. The zone plate is designed with specified value of the variable Oref. The phasor e-iWtand wave vector k, describes the radiation and the real part of Up is taken as the signal at P. Fig. 6 shows the antenna patterns for four different values of Oref. Note that as Oref is varied the position of sidelobe relative maxima and nulls change with angle.

One-way Patterns for Oref

0

5

10

15

20

25

30

35

40

45

Angle (deg)

Fig. 6 . One-way antenna patterns for Bf, = 0" (solid line), 90' (dotted), 180" (dashed), 270" (dash dot). One way transmission of radiation through the antenna is considered.

We display the main beam power and the mean sidelobe power level averaged from 5" to 45" for twelve different values of grefin increments of 30" in Fig. 7. Note that when the reference phase is varied over 360" the main beam power varies only about 0.2 dB and that the mean sidelobe level is improved by about 2 dB over the standard reference phase value of Oref = 0".

179

372

G. W. Webb, I. V. Minin & 0.V.Minin

Mainlobe and Mean Sidelobe Level vs €Iref 0.5

0 0

-0.5

0

0 0 0

B)

0

-1

0 0

0 . .

a

0 0 m . . . . ~ C

0

-1.5

0

0

0

0

0

0 0

O

-2

-

.

0

0 0

-2.5

0

90

.

180

270

360

@ref

Fig. 7. Mainlobe power (solid line) and mean sidelobe (MSL, dotted line) level as a function of B,f. 8,f is varied in 30" increments from 0" to 330". Mainlobe power of 0 dB is for Bref = 0". MSL is average from 5" to 45 'in 1" angular steps. MSL 0 dB power level is for = 0".

IV. One-way Mean Sidelobe Level With Modulated €IRef

Narrow band detection at 00 As an alternative to selective static change in Oref, it can be given a continuous dynamic variation or modulation. We consider a modulation of the reference phase having a linear "ramp" time variation:

where Oref undergoes a change of 27t in time T. The ramp is equivalent, modulo 360°, to a saw tooth ramp of period T and angular frequency 00:

180

Suppressing Sidelobe Clutter in Perimeter Security Systems

373

T is assumed to be such that oo) and k (to), they are updated using a gradient descent algorithm8.

where e is the update step size. It determines how much of an effect is the magnitude and phase error have on the new values for the complex refractive index. (A reasonable value is e = 0.01). The algorithm updates the complex index of refraction functions until the total error in (12) is no longer monotonically decreasing. The implemented algorithm includes the three steps outlined above to generate a complex index of refraction function for a variety of thicknesses in question. Fig. 3.4 shows the final error obtained from the described gradient descent algorithm at different thicknesses for a silicon sample. The shape of the plot is typical. The decaying exponential trend for small thicknesses is due to the exponential bias by (9) for the model. The growing exponential trend for large thicknesses is an artifact of the update size in the gradient descent algorithm. A smaller update coefficient (e) would maintain the error curve near zero but dramatically increase the computational load. Under simulation, a global minimum is observed, however, experimentally obtained data, at best, provides the deepest local minimum. If the material under investigation has a relatively small real index of refraction, the concave error surface surrounding this minimum is extremely narrow. We need to determine which thickness and complex refractive index pair are the predicted properties of our sample. Instead of using the total error, we introduce a total variation of degree one9. D[m\ = \n[m - 1] - n[m]\ + \k[m - 1] - k[m\, TV = ZD[m]

(16)

where the complex index of reflection is n (a>) = n(K>) — jk((o) and the sum ranges between 250 GHZ and 2 THz. The total variation measures the smoothness of the refractive index at the minimum identified by the gradient descent algorithm for each thickness. The deepest local minimum for TV is more easily determined than the deepest local minimum for the total error since the TV error surface provides a very broad concave region around the deepest local minimum. For most samples, we do not expect that the recorded complex index of refraction varies dramatically from one frequency sample to another, since the sampled frequency step size is relatively small. (Our temporal window width is approximately 50 ps with a sample rate of 5 fs which gives a frequency sampling of Af = 20 GHz).

231

424

A. Sokolnikov

Although the index may have strong variations with frequency, the majority of solid materials do not have special features compared to Af'°.

total error (arbitrary units) 10

-

thickness, mm

Fig. 3.4 Total error between the modal and measured signals for a 0.51 mm thick sample of silicon plotted over a range of thicknesses.

At each thickness, we use the recorded complex index of refraction at the final total error to calculate the total variation. As shown in Fig. 3.5, the complex index of refraction shows a marked reduction in oscillations at the proper thickness. The amount of rippling of n (co) and k (o>) also decreases as / increases. By identifying the thickness at which the deepest local minimum for total variations occurs, the algorithm identifies the proper thickness. As we approach it, the oscillations in the complex index of refraction decrease. The general trend is a decrease in amplitude as the thickness increases. This leads to the use of the deepest local minimum and the total variation of degree on in (16) and (17). Due to the large range of thicknesses to be considered, we apply the three steps outlined above to the three different thickness ranges and stepping distances. The first pass uses a coarse stepping distance over the full range of thicknesses identified in (13). The next two passes use finer stepping distances over a limited range identified from the previous pass. The deepest local minimum of the total variation in each pass determines

232

Remote Identijication of Foreign Subjects 425

the center point for the next finer pass. In a limited number of situations, the final pass did not contain a minimum, therefore, we modified the total variation metric: TV2 = C1D[m]- D[m = 111

7

(17)

real indices of refraction

1= 0.40 mm 4.2

I = 0.45 mm

3.6

1 = 0.55 mm 1= 0.60 mm

I

I

0

0.4

,

0.6

I

0.8

I

1.0

1

I

I

1.2

1.4

1.6

I

1.8

I

2.0

b frequency, THz

Figure 3.5 Final real indices of refraction obtained for different thicknesses.

This modified total variation takes the absolute difference between adjacent points of (16). Using (17), we are able to amplify the variations in smoothness.

4. Results

We examined several high index materials and investigated the general limits of the described approach which requires that the primary transmission pulse and two multiples are identified in signal. Materials with low refractive indices may not have enough multiples above the signal-to-noise (SNR) limit. The results for silicon wafers, GaAs and LiNbO3 (ordinary axis) are presented. Let us first consider the data for the silicon. Previous THz-TDS studies used silicon as a sample material since it has essentially zero absorption and a flat spectral response (i.e., (i.e., ii(o) = 3.42 - jO)." Fig. 3.4 shows that the total error plotted over the initial coarse sampling of thickness for a high resisting silicon wafer sample (p > lo4 0.cm). Measurements made with calipers gave the thickness of approximately 0.51 +/- 0.01 111111. The deepest local minimum identified in Fig. 3.4 is approximately 30 pm off the measured value. The predicted thickness error is one tenth of the coherence length of the terahertz pulse.

233

426

A . Sokolnikov

T

index of refraction

Expected real

3

I

0

Predicted real

0.5

1

1.5

2

b frequency, THz

Fig. 3.6 Real and imaginary index of refraction for the predicted thickness identified in Fig. 3.4. Complex index of refraction for Si is 3.418-j0 (from a number of recent publications).

Fig. 3.6 contains the real and imaginary indices of refraction at the thickness 0.54 nm identified from the TV2 metric in (17). The solid lines indicate the predicted values. The line at 3.42 is the expected real refractive index [ 111. The predicted and expected values for the imaging components are superimposed. The real index is slightly low, since the predicted thickness is rather high, both the real and imaginary predicted values are independent of frequency as expected. The primary source of error is due to the alignment of the sample for which the THz beam should be at a normal. The second sample, semi-insulating GaAs, has the following expected complex index of refraction”.

where II is frequency in THz. Due to the conventional manufacturing process, the complex index of refraction of GaAs varies noticeably from sample to sample, thus (18) is a good approximation. The predicted thickness for GaAs was 0.38 mm compared to the measured thickness of 0.41 +/-0.01 111111. Fig. 3.8 shows the comparison between the predicted and measured refractive index values. Note that for GaAs variations of the index are more pronounced. A small variation in the predicted thickness (Fig. 3.8) corresponds to a slight difference in the real index. Third, LiNb03 (ordinary axis) is examined. The results are given in Fig. 3.9. The dashed line represents the values for L i m o 3taken from literature.

234

Remote Identification of Foreign Subjects 427

15

t

total variation (arbitrary units)

I 0

b

I

0.4

0.6

0.8

1.O

thickness, mm

Fig. 3.7 Total variation (TV) measure for GaAs is shown with a 0.01 mm step.

A simulation is used to determine the limit of producing the parameters indicated above. To validate our approach at low values of the refractive index, we created a simulation system to yield the input and output time domain wave forms. A Rayleigh distribution models give a faster current rise and slower current fall in a photoconductive switch. We modified the distribution to smoothen the region around the origin to make it differentiable. The derivative of the modified Rayleigh distribution is used as a model for a single-cycle THz pulse. From the SNR measurements of the THz system, we can expect a ratio of 1,500:l using a relatively large number of average wave forms (> 500). Uisng simulations and a captured noise signature, we determined the minimum real index resolved by the algorithm. Fig. 3.10 shows the range of parameters for which the method is applicable. The solid area represents the passing region. The area below is signal - to - noise limited.

235

428

A . Sokolnikov

index of refraction Predicted 1 = 0.38 mm 4.2

Real with I = 0.41

3.4

Expected value

.

frequency, THz

Figure 3.8 Refractive index values for the predicted and measured thickness

.+-

7.0

m

&

!2

expected values

Q-

6.8

6.6

6.2

I

0.25

0.4

0.6

0.8

1.o

-

1.1

Figure 3.9 Predicted real refractive index and the real index generalized from the literature for a sample of LiNbOi (ordinary axis)’’.

236

Remote Identajicataon of Foreign Subjects

I,"-

429

Lr

m

4 ._

l

i 2 ; I

I I

I I I

I

..

1 1

I

I

3

I

I I I I

0 I I

I I

6 I

0.4

: 11

I

1 I

1 1

1

1 1 1 I

I I I

1 1 1

w

i

I

I

I

I

1 1

1 1

I 1 I I

1 1 1 1

.

.

I

1

1 I I I I

1 1 1 1 1

1

1

6 i I I

1 1

I I 1 I I

1 1 1 1 1

I I I I I I I I

1 1 1 1 1 1 1 1

I

I I

1 1

I

I I I

1 1 1

I I I

1 1 1

.

.

i Q

I I

1 I

i

i

i

n

I I I I I I

I I I I I I I I

I

I I

mm 0.6

1.o

0.8

1.2

1.4

Fig. 3.10 Limits of the described method are shown for a simulated material at a signal - to

-

noise ratio of

1,500: 1. The circles stand for the test cases that passed for each thickness given by the vertical dashed lines. At each point, the real refractive index was constant vs. frequency while the imaginary component was zero.

5. Conclusions The password tablet, thus, can be made out of GaAs by conventional techniques of lithography. The accounted for discrepancy of about 0.3 mm means that the tablet is not technologically expensive since the modern day lithographical accuracy substantially exceeds 0.1 mm limit. On the other hand, GaAs is wide spread and comparatively inexpensive. LiNb03 seems to be more of a challenge because of the lack of the GaAs's advantages, however, it is a likely choice. The low index materials have much more limited capabilities of becoming suitable for the password tablet in particular because of the signal-to-noise ratio. There is also a dependence of measurements on the initial angle at which the radiation beam falls on the sample. At the same time the precision of thickness determination can be even higher than with the high-index materials. As an example, a Teflon sample yielded less than 0.01 mm error. At present only the interferometric approach can give a good resolution. A noninterferometric technique can only hardly resolve a sample.

237

430 A . Sokolnikov

A spectroscopic imaging method was used to determine simultaneously and independently the complex index of refraction and thickness of a sample. A usage of multiple wave forms allows obtaining the same parameters for both high and low index materials. Thickness measurements yielded an error which is much smaller than the coherence length of the used THz source.

References 1. A. Sokolnikov, “Resonance Amplification of the Probing Signals in Optical Coherence Tomography (OCT)”, SPIE Transactions, (2005) 2. L Duvillaret, F. Caret, and J. Coutaz, “A Reliable Method for Extraction of Material Parameters in Terahertz Time-Domain Spectroscopy”, IEEE J. Sel. Top. Quant. Elec., 2, 739 - 746, (1996) 3. L Duvillaret, F. Caret, and J. Coutaz, “Highly Precise Determination of Both Optical Constants and Sample Thickness in Terahertz Time-Domain Spectroscopy”, Appl. Opt., 38,409 - 41 5, (1999) 4. T. Domey, J. Johnson, D. Mittleman, R. Baraniuk, “Imaging with Terahertz Pulses”. 5. A. Sokolnikov, “Adaptive Nonintrusive Terahertz Identification”, SPIE Proceedings, Paper #6201-63,2006 6. E. Hecht, Optics, 2”ded. Reading, MA. Addison-Wesley, (1987) 7. P.E. Ciddar, “Refractive index of air: new equations for the visible and near infrared”, Appl. Opt., 35, 1566-1573, (1996) 8. S. Haykin, Adaptive Filter Theory, Englewood Cliffs, NJ, Prentice Hall 9. J.E. Odegard and C.S. Bums, “Discrete finite variation: a new measure of smoothness for the design of wavelet basis.” Proc. Of ICA SSP, 1467 - 1470, (1996) 10. M. Nuss and J. Orenstein, “Terahertz Time-Domain Spectroscopy (THz - TDS) in Millimeter and Submillimeter Wave Spectroscopy of Solids”, ed. G. Gmner, Heidelberg, Germany, SpringerVerlag, (1998) 11. D. Grischowsky, S. Keiding, M. VanExter, and C. Fattinger, “Far-infrared time-domain spectroscopy with terahertz beams of dielectrics and semiconductors”, Opt. SOC.Am B, 7,2006 2015, (1990) 12. E.D. Palik, Handbook ofoptical Constants of Solids, Academic Press, (1985)

238

International Journal of High Speed Electronics a n d Systems Vol. 17, NO. 2 (2007) 431-443 @ World Scientific Publishing Company

World Scientific www.worldscientific.com

TERAHERTZ INTERFEROMETRIC AND SYNTHETIC APERTURE IMAGING ALEXANDER M SINYUKOV Department of Physics, New Jersey Institute of Technology, Newark, New Jersey 07102 [email protected] APARAJITA BANDYOPADHYAY', AMARTYA SENGUPTA', ROBERT B BARAT', DALE E GARY', ZOI-HELEN1 MICHALOPOULOU3,DAVID ZIMDARS4, AND JOHN F FEDERICI' 'Department of Physics 'Otto York Department of Chemical Engineering 3Department of Mathematical Sciences New Jersey Institute of Technology, Newark, New Jersey 07102 4Picometrix,Inc, 2925 Boardwalk, Ann Arbor, MI 48113

Experimental results of homodyne terahertz interferometric 1-D and 2-D imaging are presented. The reconstructed images of a point source are in a good agreement with theoretical predictions. The performance of an N element detector array is imitated by only one detector placed at N positions. Continuous waves at 0.25-0.3 THz are used to detect a metal object behind a barrier. 1-D images of a C-4 sample have been obtained at several terahertz frequencies. Focusing issues of 2-D imaging have been demonstrated. The terahertz interferometric imaging method can be used in defense and security applications to detect concealed weapons, explosives as well as chemical and biological agents. Keywords: Terahertz, near-field, interferometric, image.

1. Introduction

Terahertz (THz) waves (wavelength range -0.3 - 3mm) show promise for defense and security applications' since they propagate easily through most nonpolar, nonmetallic material such as packaging, clothing, shoes, book bags and other barriers. Explosives, chemical and biological agents have characteristic absorption lines in THz frequency rangezx3,which can be used to identify these materials. Moreover, THz waves bring minimal or no risk to human health making it an attractive imaging modality for routine screening of people or animals.

239

432

A . M . Sinyukov et al.

At short distances, atmospheric attenuation and scattering of THz waves is low. Therefore, several methods can be used for close range THz imaging (