Terahertz frequency detection and identification of materials and objects: [proceedings of the NATO Advanced Research Workshop on Terahertz Frequency Detection and Identification of Materials and Objects, Spiez, Switzerland, 7 - 11 July 2006]

Terahertz Frequency Detection and Identification of Materials and Objects

NATO Science for Peace and Security Series This Series presents the results of scientific meetings supported under the NATO Programme: Science for Peace and Security (SPS). The NATO SPS Programme supports meetings in the following Key Priority areas: (1) Defence Against Terrorism; (2) Countering other Threats to Security and (3) NATO, Partner and Mediterranean Dialogue Country Priorities. The types of meeting supported are generally "Advanced Study Institutes" and "Advanced Research Workshops". The NATO SPS Series collects together the results of these meetings. The meetings are coorganized by scientists from NATO countries and scientists from NATO's "Partner" or "Mediterranean Dialogue" countries. The observations and recommendations made at the meetings, as well as the contents of the volumes in the Series, reflect those of participants and contributors only; they should not necessarily be regarded as reflecting NATO views or policy. Advanced Study Institutes (ASI) are high-level tutorial courses intended to convey the latest developments in a subject to an advanced-level audience Advanced Research Workshops (ARW) are expert meetings where an intense but informal exchange of views at the frontiers of a subject aims at identifying directions for future action Following a transformation of the programme in 2006 the Series has been re-named and re-organised. Recent volumes on topics not related to security, which result from meetings supported under the programme earlier, may be found in the NATO Science Series. The Series is published by IOS Press, Amsterdam, and Springer, Dordrecht, in conjunction with the NATO Public Diplomacy Division. Sub-Series A. B. C. D. E.

Chemistry and Biology Physics and Biophysics Environmental Security Information and Communication Security Human and Societal Dynamics

http://www.nato.int/science http://www.springer.com http://www.iospress.nl

Series B: Physics and Biophysics

Springer Springer Springer IOS Press IOS Press

Terahertz Frequency Detection and Identification of Materials and Objects edited by

R.E. Miles University of Leeds, U.K.

X.-C. Zhang Rensselaer Polytechnic, Troy, U.S.A.

H. Eisele University of Leeds, U.K. and

A. Krotkus Semiconductor Physics Institute, Vilnius, Lithuania

Published in cooperation with NATO Public Diplomacy Division

Proceedings of the NATO Advanced Research Workshop on Terahertz Frequency Detection and Identification of Materials and Objects Spiez, Switzerland 7 – 11 July 2006

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN 978-1-4020-6502- 6 (PB) ISBN 978-1-4020-6501-9 (HB) ISBN 978-1-4020-6503-3 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands.

www.springer.com

Printed on acid-free paper

All Rights Reserved © 2007 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

CONTENTS

Organising Committee

ix

Sponsors

x

Preface xi Theme 1: Devices Terahertz Emission from Semiconductors Excited by Ultrafast Laser Pulses A. Krotkus, R. Adomavičius, and V. L. Malevich

3

Terahertz Generation by Multiplication Jan Stake, Tomas Bryllert, T. Arezoo Emadi, and Josip Vukusic

17

Towards Superlattice Terahertz Amplifiers and Lasers Alvydas Lisauskas, Ernst Mohler, Hartmut G. Roskos, and Nataliya V. Demarina

31

Tailoring the Emission of Terahertz Quantum Cascade Lasers Richard Green, Lukas Mahler, Cosimo Mauro, Tonia Losco, Ji-Hua Xu, Alessandro Tredicucci, Fabio Beltram, Harvey Beere, and David Ritchie

41

Guided Propagation of Terahertz Pulses on Metal Wires Kanglin Wang and Daniel M. Mittleman

55

Superlattice and Other Negative-Differential-Resistance Devices: Current Status Heribert Eisele

69

Theme 2: Interactions with Materials Molecular and Organic Interactions A. G. Davies and E. H. Linfield

v

91

vi

CONTENTS

Terahertz Beam Interactions with Amorphous Materials Mira Naftaly and Robert E. Miles Development of Tagless Biosensors for Detecting the Presence of Pathogens Jing–Yin Chen, Joseph R. Knab, Shuji Ye, Yunfen He, and Andrea G. Markelz

107

123

Theme 3: Detection and Sensing Improvements to Electronic Techniques for Terahertz Spectroscopic Detection Daniel W. van der Weide, Alan D. Bettermann, Min K. Choi, and John Grade Terahertz Time-Domain Spectroscopy of Crystalline and Aqueous Systems Peter Uhd Jepsen, Hannes Merbold, Zhengxin Li, Xiaoyu Xing, and Stewart Clark Continuous-Wave Terahertz Photomixer Systems for Real-World Applications Ian S. Gregory, Hideaki Page, and Lee Spencer

137

147

167

Theme 4: Systems for Security Systems Requirements for A Multi-Channel Terahertz Contraband Scanner William S. Truscott

187

Challenges to Terahertz Counter-Terrorism and Security-Related Applications Howard Cummins

205

Terahertz Detection of Illegal Objects Roger Appleby, Peter R. Coward, and Gordon N. Sinclair

225

Terahertz Rays to Detect Drugs of Abuse Kodo Kawase, Adrian Dobroiu, Masatsugu Yamashita, Yoshiaki Sasaki, and Chiko Otani

241

CONTENTS

Terahertz Spectroscopy for Explosive, Pharmaceutical, and Biological Sensing Applications Hai-Bo Liu and Xi-Cheng Zhang Terahertz Communications: A 2020 vision Martin Koch

vii

251

325

Theme 5: Overview Applied Terahertz Science: The Technology of the Future, and Always Will Be? Martyn Chamberlain

341

List of speakers

353

List of participants

355

Group photograph

357

Index

359

ORGANISING COMMITTEE

Professor X.-C. Zhang (Rensselaer Polytechnic, USA)

Co-director

Professor A. Krotkus (Semiconductor Physics Institute, Vilnius, Lithuania)

Co-director

Professor R. E. Miles (University of Leeds, UK)

Secretary

Dr. H. Eisele (University of Leeds, UK)

Treasurer

ix

We wish to thank the following for their financial contributions which have contributed greatly to the success of the workshop.

NATO US Army International Technology Centre – Atlantic, Research Division US Air Force Office of Scientific Research Teranova (European Union Integrated Project)

x

PREFACE This volume contains an account of the third NATO-sponsored meeting on terahertz science and technology. The first, held in 1996, was an Advanced Study Institute (ASI) on “New Directions in Terahertz Technology” which in turn was followed in 2000 by an Advanced Research Workshop (ARW) entitled “Terahertz Sources and Systems”. In these earlier meetings, the difficulties involved in generating THz frequency signals formed a background to most of the presentations and discussions – with frequent mentions of the 0.3 to 3 THz “Terahertz Gap”. This gap is still with us but, with advances in technology, has been reduced to the region between about 1.0 THz and 2.5 THz. In the 2000 ASI, the lack of a complete set of THz electronic devices, in particular an amplifier, was identified as an obstacle to progress. While this is still the case, it became clear from this ASI that THz amplifiers based on semiconductor superlattice structures are very close to being realized. The purpose of this workshop was to explore the potential of THz techniques in the detection of illegal substances, especially for security purposes. The workshop set out to identify the state of the art, to define the scope of the problem and the areas where THz frequency radiation could contribute. Extensive studies have shown that drugs of abuse and explosives can be identified from their THz spectra and prototype systems are already in place in Japan to scan mail online as it passes through the automated systems. From this and other work reported at the ASI it is clear that THz technology has a part to play in security screening but developments are still required to improve the delivery of THz signals to the area of concern and its subsequent detection. One possibility is that THz techniques may not be the first line of detection but could be used for more detailed identification of, say, suspicious packages. Stand-off detection, as is required in the identification of suicide bombers, is a more difficult issue but the delegates concluded that detection at 30 m is possible but effective systems will only result from close cooperation between the THz generation and signal-processing communities. We learned at this meeting that THz radiation can be generated at a distance using air as the transducer. However, this necessitated the use of a powerful laser beam which could not be employed in a public place such as an airport, but could perhaps be implemented in other situations. We trust that the readers of this book will gain a flavour of the excitement of the meeting. However, one thing that is difficult to convey here is the wonderful way in which we were looked after by our hosts Peter and Margit Schlatter for whom nothing was too much trouble.

xi

Theme 1 DEVICES

TERAHERTZ EMISSION FROM SEMICONDUCTORS EXCITED BY ULTRAFAST LASER PULSES

A. KROTKUS* AND R. ADOMAVIČIUS Semiconductor Physics Institute, A. Gostauto 11, 01108, Vilnius, Lithuania V. L. MALEVICH Institute of Physics, National Academy of Sciences of Belarus, F. Skorina Ave. 68, 220072, Minsk, Belarus

Abstract. Various physical mechanisms leading to terahertz (THz) emission from semiconductor surfaces illuminated by femtosecond laser pulses are analyzed. Results obtained on different materials are described and relative efficiency of these materials as THz emitters is compared.

Keywords: semiconductors, InAs, THz emission, THz emission spectroscopy

1. Introduction Currently, there are two basic approaches for generating THz radiation beams which utilize ultrashort laser pulses: photoconductive technique based on high-speed photodetectors integrated with wideband-radiating antennae and bare semiconductor surfaces illuminated by femtosecond laser beams. In the case of photoconductive antennae that are used both for THz pulse generation and their detection, the main problem remaining is the availability of the proper semiconductor material that should be photosensitive at the laser wavelength and should have other distinctive properties like high resistivity and sub-picosecond carrier lifetimes. For the most popular in THz pulse applications femtosecond laser system – mode-locked

______ *

To whom correspondence should be addressed: Arunas Krotkus, Semiconductor Physics Institute, A. Gostauto 11, 01108, Vilnius, Lithuania; e-mail: [email protected]

3 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 3–16. © 2007 Springer.

4

A. KROTKUS ET AL.

Ti:sapphire laser (laser wavelength of ~800 nm), the material of choice is low-temperature molecular-beam-epitaxy grown GaAs (LTG GaAs) layers [1]; however, the search for the most suitable semiconductor that could be used in THz antennae activated by femtosecond lasers generating at longer wavelengths from 1 to 1.5 µm is still not finished. Ion-implanted InGaAs [2, 3] and LTG GaBiAs [4] are among the possible candidates for ultrafast photodetectors for this spectral range. In the second case, when THz pulses are generated by illuminating semiconductor surfaces with ultrashort laser pulses all mentioned above restrictions on the material properties are less important. The best THz emitters are narrow-gap semiconductors like InAs [5], which have optical absorption edge corresponding to the mid-infrared spectral region. Stronger or weaker THz pulses are emitted from the majority of weakly or moderately doped semiconductors with rather long carrier lifetimes. Besides of the universal occurrence of this effect, emission from the semiconductor surfaces provide much wider and better shaped THz beams than those generated by photoconductive antennae, which could be preferable for some specific applications like THz imaging. In this chapter, we will review the physical mechanisms leading to THz emission from the semiconductor surfaces and will compare the characteristics of this effect in various materials. 2. Physical Mechanisms 2.1. INSTANTANEOUS POLARIZATION EFFECTS

The lowest order nonlinear optical response of a non-centrosymmetric crystal is caused by the second-order susceptibility χ2 that leads to the sum and difference frequency generation. In the case when the optical beam contains nearly the same frequencies (which is typical for femtosecond laser spectra) and interacts nonlinearly with the crystal, the difference frequency is in the dc range and the induced polarization is referred as optical rectification (OR) with the induced charge displacement following the optical pulse envelope. When a built-in dc field Es is present at the semiconductor surface, transient THz polarization can also be induced due to the third-order nonlinear susceptibility χ3; THz pulse magnitude generated due to this electrical-field-induced optical rectification (EFIOR) effect will be proportional to the effective second-order susceptibility χ2 χ 2eff = χ 3 E s .

TERAHERTZ EMISSION FROM SEMICONDUCTORS

5

As a fingerprint of nonlinear optical THz pulse generation mechanisms is usually considered the dependence of the emitted THz pulse amplitude on the orientation of the optical field with regard to the crystallographic axes. It has been pointed out recently [6] that these so-called azimuthal angle dependences of the THz emission efficiency from zinc-blende semiconductors like GaAs measured at different crystallographic planes can also help in distinguishing OR and EFIOR mechanisms. For (111) plane both effects give similar cos 3ϕ – type azimuthal dependences, whereas for (100) plane THz signal is constant for EFIOR mechanism and depends on the azimuthal angle ϕ proportional to cos 2ϕ – for OR mechanism. THz generation due to nonlinear optical interaction is significantly modified when the energies of the laser quanta become larger than the fundamental energy bandgap of the semiconductor Eg [7]. It has been shown by Sipe and Shkrebtii in [8] that laser-induced nonlinear polarization in this case contains an additional to OR contribution arising from the spatial shift of the charge during excitation leading to the “shift” current. 2.2. PHOTOCURRENT SURGE EFFECTS

2.2.1. Surface field effect If there is a static electric field at the surface of semiconductor illuminated by an ultrashort laser pulse, the free carriers created by this pulse are driven by the static field and accelerate along the field direction, forming a transient photocurrent. The rise time of the photocurrent pulse is comparable to the optical pulse duration; its decay time corresponds to the shorter of two characteristic times: carrier lifetime or their transit across the field region time. This fast changing photocurrent leads to the radiation of the electromagnetic waves with amplitude proportional to the first time derivative of the current, in the far field. Built-in electric fields in the majority of semiconductor surfaces appear due to the band bending imposed by the surface potential different from the energy position of the Fermi level in the bulk of the material. The sign of this field is usually opposite for n-type and p-type crystals, which leads also to the opposite polarities of THz pulses generated from the surfaces of these crystals. 2.2.2. Photo-Dember effect Transient photocurrent surge can appear at the semiconductor surface illuminated by ultrafast laser pulse even in the absence of the surface

6

A. KROTKUS ET AL.

electric field. This situation is typical for narrow-gap semiconductors, which are relatively good THz emitters despite of the small band bending at their surfaces. Electrons and holes excited at the surface are predominantly moving perpendicular to it, towards the bulk of the material. Electron diffusion usually is faster than hole diffusion, therefore a dynamic photovoltage can develop at the illuminated surface that can lead to the THz emission. This effect is especially strong in semiconductors with a narrow energy bandgap, where the electrons are excited with quite large excess energies εex that can exceed 1 eV for the excitation by Ti:sapphire laser quanta ( hν ≈ 1.55 eV). In the majority III–V and IV–VI narrow-gap materials characteristic time of the electron-LO phonon scattering is around 200 fs, the majority of these scattering processes lead only to small changes in electron momentum, therefore, during the first few hundred femtoseconds after the photoexcitation excitation, the electron movement is purely ballistic rather than diffusive. 2.2.3. Optical orientation For III–V semiconductors the ballistic photocurrent can depend on the crystallographic direction and can have a component parallel to the surface, which could explain the azimuthal angle dependencies of the THz emission without involving nonlinear optical interactions. Valence band of III–V semiconductors consists of two subbands of heavy and light holes at the center of the Brillouin zone. Due to the optical selection rules, the momenta of electrons excited from the heavy-hole band by linearly polarized light will mainly lie in the plane perpendicular to the direction of the light electrical field (Figure 1). On the other hand, the momentum distribution for electrons excited from the light-hole band will be stretched out along the direction of the optical field polarization. When the light is falling into air-semiconductor interface at incline angle and is absorbed at a thin surface layer, the electrons moving to the right in Figure 1 will be diffusively scattered by the surface and a lateral photocurrent component will arise. This lateral photocurrent effect has been discovered in GaAs crystals in [9]. In this experiment as well as in the observation of the polarized hot-electron photoluminescence [10], cryogenic temperatures and photon energies leading to the photoelectron energies smaller than the optical phonon energy were used for reducing the electron-scattering rate. However, when the semiconductor is excited by a short light pulse, such lateral surface photocurrent component will be present even at the room temperature also for larger energy quanta during the first few hundreds of femtoseconds when the electrons are still moving ballistically and could

TERAHERTZ EMISSION FROM SEMICONDUCTORS

7

lead to the THz radiation that will be efficiently outcoupled from the semiconductor. Real iso-energetic surfaces in the valence bands of III–V semiconductors are not spherical but warped as a result of their cubic symmetry; therefore, amplitudes of the lateral and perpendicular to the surface photocurrent components will depend on the orientation of the polarization vector relative to the crystallographic axes.

Figure 1. Illustration of the surface photocurrent appearing due to the optical alignment of the electrons and their diffusive scattering at the surface.

2.3. COLLECTIVE RESPONSE EFFECTS

When THz emission from semiconductor surfaces by ultrashort laser pulses is caused by coherent movement of photoexcited carriers, its intensity and density depends on the basic semiconductor material parameters themselves, especially on its doping level. The fast change of the electric field at the vicinity of the surface can initiate plasma oscillation of the extrinsic carriers that will amplify the THz wave [11]. Pulsed excitation of the semiconductor can also lead to the coherent generation of infrared-active lattice vibrations that would produce oscillations of macroscopic dielectric polarization and emission of electromagnetic waves at the phonon frequency. Such an effect has been observed in Te [12] and several other semiconductors.

8

A. KROTKUS ET AL.

3. Discussion of the Results Obtained on Various Semiconductor Materials 3.1. INDIUM ARSENIDE

THz generation from InAs was investigated more often than from any other narrow-gap semiconductors, because these crystals are most efficient emitters, especially when a strong magnetic field is applied in parallel to their surfaces [5]. This enhanced magnetic field influence (magnetic fields of ~1 T can lead to an increase of the emitted THz power by a factor of 100) had focused the attention of researchers on the photo-Dember effect, although some authors have also pointed out the role of other phenomena like bulk OR [13], magneto-plasma waves [14], or coupled plasmon-phonon modes [15] in THz radiation from InAs surface.

Figure 2. Dependence of the emitted THz field amplitude on the InAs crystal doping level.

Possible influences of the surface electric field to the THz emission were usually written off due to a narrow-bandgap in InAs and potentially smallband bending at the surface of this material. This is not always justifiable because it is known that surface potential in InAs is fixed at fairly high

TERAHERTZ EMISSION FROM SEMICONDUCTORS

9

(0.18–0.2 eV [16]) values above the conduction band minimum and in a p-doped crystal surface depletion layer can be sufficiently wide and strong. It has been found in [17] that p-type InAs is a better THz emitter than n-type InAs (Figure 2), which was explained by a possible contribution of the EFIOR effect. A similar conclusion was drawn in [6] from the analysis of the symmetry of the azimuthal dependences of THz radiation measured by illuminating different crystal planes of InAs and also from the fact that the surfaces of InAs illuminated by P-polarized optical beams radiate both P and S-polarized THz signals, which is typical to nonlinear optical rather than to the current surge mechanisms of THz generation. Dependences of the THz-radiation efficiency from femtosecond laser illuminated InAs surfaces on the photon quantum energy were measured in [18]. The results of such measurements are presented in Figure 3. THz field amplitude increases with the increasing photon energy, reaches maximum at hν = 1.6 eV, and then decreases. Such a shape of the spectral dependence for THz emission could be expected in the case when free electron contribution to this effect is dominating. When the quantum energy is large

Figure 3. Dependences of the THz emission efficiency in InAs on the femtosecond laser quantum energy.

10

A. KROTKUS ET AL.

enough, the electrons are excited high in the conduction band, where they are efficiently scattered to the subsidiary L valleys with a low mobility. Intervalley separation in the conduction band of InAs determined from the experiments presented on Figure 3 ∆εΓL = 1.08 eV coincides with the previous estimations of this parameter. However, this contradicts with the previously made conclusion on the prevalence of the EFIOR effect that should not be influenced by the electron scattering rate variation. Moreover, it can be seen from Figure 3 that the azimuthal angle anisotropy of THz amplitude at large photon energies is also decreasing. In part, these contradictory experimental facts could be explained by the optical alignment effect, however, the calculation of this contribution using commonly accepted valence band parameters of InAs had shown that the anisotropy of this band can lead to a much weaker azimuthal angle dependence of the radiated THz signal as it is observed experimentally. The situation becomes more transparent, when one remembers that photoexcited electrons can themselves create at the crystal surface a strong electrical field. Electrons are excited by Ti:sapphire laser quanta at excess energies of ~1 eV and move towards the bulk with the velocity greater than 1.5108 cm/s. Their movement is being stopped by an electrical field that is appearing due to the electron and hole separation. Figure 4 shows the results of numerical Monte Carlo simulation of the temporal and spatial dynamics of the surface potential in InAs after its photoexcitation by 150-fs duration laser pulse. It can be seen

Figure 4. Spatio-temporal dynamics of the surface potential in InAs excited by femtosecond laser pulse.

TERAHERTZ EMISSION FROM SEMICONDUCTORS

11

from Figure 4 that electrical fields as large as 2·105 V/cm are developing during the laser illumination over distances reaching ~0.5 µm. These fields are much stronger than any static, built-in electrical field in InAs, thus they could induce an intense nonlinear optical interaction with the exciting laser pulse, and cause THz generation due to the EFIOR effect. As a test of this mechanism of THz generation from InAs surface an additional pump-and-probe experiment was performed. The laser beam was split into two parts, one of which (the probe beam) was incident on the sample’s surface at 45° angle and generated THz pulses monitored by the detector. The second (the pump beam) part of the laser beam was parallel to the surface normal; it was focused to a rather large spot (diameter of ~0.5 mm) in order to avoid possible interference at the detector of THz pulses generated by both laser beams. Carrier densities of approximately 4·1017 cm−3 and 1018 cm−3 were excited by the probe and pump beams, respecttively. S-polarized component of the THz signal at its maximum was measured as a function of the time delay between the probe and pump pulses. Figure 5 shows the results of the pump-and-probe experiment. THz emission increases at times close to the overlap of both optical pulses and becomes weaker when the pump pulse arrives to the sample before the probe pulse.

Figure 5. Results of the pump-and-probe THz generation experiment on InAs.

12

A. KROTKUS ET AL.

The effect of carriers excited by the pump pulse on the THz signal generated by the probe pulse is twofold. First of all, electron ballistic movement and the charge carrier separation add to the surface electric field and leads to an increase in the generated THz signal. Secondly, when the pump pulse arrives at the surface at much earlier time moment than the probe pulse, carriers excited by the pump pulse have already cooled down and contribute to the relaxation of the surface electrical field and to the reduction of THz generation. The shape of the experimental trace presented on Figure 5 evidences the presence of both effects. The peak of THz emission at time close to the zero delay is due to the enhancement of the EFIOR contribution in the field induced by the pump pulse and the reducetion of this emission at longer delays is, most probably, caused by the screening of the surface electrical field by the photoexcited carriers. The timescale of both contributions coincides with the characteristic times of the surface field nucleation and decay obtained by numerical calculations (Figure 4) and discussed above. 3.2. OTHER NARROW-GAP SEMICONDUCTORS

When the ballistic photocurrent surge effect is dominates the THz emission from semiconductor surfaces, the emission becomes stronger with increasing electron excess energy. This is illustrated in Figure 6 by the THz pulse amplitudes measured at the same experimental conditions on different semiconducting compounds from the InAs/GaAs system. Throughout this system, the energy bandgap varies from 0.35 eV for InAs to 1.4 eV for GaAs, and the electron-excess energies – from 1.05 eV to 0.15 eV (when Ti:sapphire laser pulses are used for the illumination of the samples). Therefore, semiconductors with a narrower bandgap should be more efficient THz emitters. InSb has a band structure similar to that of InAs, but with two times smaller energy bandgap, however, THz power radiated from this semiconductor surface illuminated by femtosecond Ti:sapphire laser pulses is ~100 times weaker. It has been suggested that this reduction of THz emission efficiency is caused by the intervalley scattering of the photoexcited electrons. Spectral dependence of THz radiation from InSb supports this conclusion [18]; intervalley energy separation in the conduction band of InSb evaluated from this dependence is equal to ∆εΓL = 0.53 eV. It should be pointed out that for longer laser wavelengths of around 1.5 µm, InSb is preferred over InAs as a THz-emitting material.

TERAHERTZ EMISSION FROM SEMICONDUCTORS

13

Figure 6. Amplitudes of THz pulses emitted from the surface of various compounds from the InAs/GaAs system when excited by femtosecond Ti:sapphire laser pulses.

CdxHg1−xTe crystals are also narrow-gap semiconductor compounds widely used in infrared devices. THz emission from these compound semiconductor surfaces illuminated by Ti:sapphire laser pulses was investigated in [19]. The samples investigated were CdxHg1−xTe crystals with x = 0, 0.2, and 0.3. HgTe sample was single-crystalline, whereas thesamples of remaining two alloy compositions were epitaxial layers grown on CdTe substrates. Figure 7 shows the time-domain waveforms measured on three CdxHg1−xTe samples with different alloy composition. As it can be seen from that figure, the largest amplitude has THz transient emitted by the sample with x = 0.2. In contrary to our expectations, HgTe, for which the excess energy of photoexcited electrons should be the largest, radiates the transient with the smallest amplitude.

14

A. KROTKUS ET AL.

Figure 7. THz field transients measured for the samples from different CdxHg1−xTe alloys.

The band structure of CdxHg1−xTe alloy system changes drastically when altering the alloy composition. On substituting Cd for the heavy Hg atom the energy spectrum of the CdxHg1−xTe alloys exhibits a transition from a semimetal behavior in HgTe to a semiconductor behavior in CdTe. The bandgap changes almost linearly with x and becomes equal to εg = 0.35 eV for x = 0.3. Spin-orbital split band Γ7 energetic position relative to the heavy-hole band is changing with the alloy composition slower than the bandgap and, for the alloy compositions investigated in our work, is in the range of ∆0 = 1−1.2 eV. Optical pulses generated by Ti:sapphire laser (hν = 1.51 eV) can excite electrons from all three: light-hole, heavy-hole, and spin-orbital split valence bands. Excess energy of electrons and THz transient magnitude excited by first two of those transitions will decrease with x, which is inconsistent with the experimental observations. Moreover, transitions from the heavy-hole band will excite electrons to very high energies in the conduction band, where possible scattering to the subsidiary, larger effective mass L-valleys (positioned at 1.2–1.4 eV in CdTe/HgTe system [20]) will reduce the photo-Dember voltage and the radiated THz signal.

TERAHERTZ EMISSION FROM SEMICONDUCTORS

15

Therefore, we suppose that most important are the transitions from the spin-orbital split valence band to Γ6 conduction band valley. 3.3. GERMANIUM

As compared to compound semiconductors, the data on THz emission from the surfaces of elementary semiconductors, such as silicon or germanium, are very scarce. For example, Ge, which is a better THz emitter as Si, excited by Ti:sapphire laser pulses with the duration of ~100 fs radiates THz pulses with an amplitude almost one order of magnitude smaller than InAs [21]. However, some peculiarities of the energy band structure of this material raise a hope that it can find some particular applications in optoelectronic THz range devices. Ge is an indirect bandgap material with the lowest conduction band valleys positioned at L points of the Brillouin zone. Energy bandgap of Ge is rather narrow ( ε g = 0.74 eV), therefore Ti:sapphire laser quanta excite electrons by direct transitions into Γ valley lying 0,12 eV above the Lvalleys. Photoexcited electrons are then rather fast (~50 fs) scattered from the Γ valley to L and X-valleys, where their mobility is quite low. Therefore, in combination with laser pulses longer than 50 fs, Ge emits THz radiation rather poorly – the magnitudes of THz pulses radiated from a Ge surface illuminated by 150-fs duration Ti:sapphire laser pulses are ~3% of that radiated by p-type InAs crystal. However, when shorter laser pulses are used, the effect of the high mobility G valley electrons becomes more important and THz pulse amplitude can increase. Experiments performed with 20-fs duration pulses have shown that relative amplitude of THz pulses emitted from Ge surface increases by more than 4 times as compared with 150-fs pulse excitation. 4. Conclusions Semiconductor surfaces illuminated by femtosecond laser pulses can rather efficiently radiate THz pulses. THz radiation from semiconductor surfaces can be caused by a number of different physical mechanisms involving both bound and free electrons. Most efficient emitters are group III–V narrow-gap semiconducting compounds such as InAs. Relative intensity of THz pulses radiated from laser excited semiconductor surfaces depends on the details of the electron band structure of the material and its doping level as well on the wavelength and the duration of the laser pulses.

16

A. KROTKUS ET AL.

References 1. A. Krotkus and J-L. Coutaz, Semicond. Sci. Technol., 20, S142 (2005). 2. C. Carmody, H. H. Tan, C. Jagadish, A. Gaarder, and S. Marcinkevičius, Appl. Phys. Lett., 82, 3913 (2003). 3. N. Chimot, J. Mangeney, L. Joulaud, H. Bernas, K. Blary, and J. F. Lampin, Appl. Phys. Lett., 87, 193510 (2005). 4. K. Bertulis, A. Krotkus, G. Aleksejenko, V. Pačebutas, R. Adomavičius, G. Molis, and S. Marcinkevičius, Appl. Phys. Lett., 88, 201112 (2006). 5. N. Sarukura, H. Ohtake, S. Izumida, and Z. Liu, J. Appl. Phys., 84, 654 (1998). 6. M. Reid, I. V. Cravetchi, and R. Fedosejevs, Phys. Rev., B 72, 035201 (2005). 7. X.-C. Zhang, Y. Jin, K. Yang, and L. J. Schowalter, Phys. Rev. Lett., 69, 2302 (1992). 8. J. E. Sipe and A. I. Shkrebtii, Phys. Rev., B 61, 5337 (2000). 9. V. L. Alperovich, V. I. Belinicher, V. N. Novikov, and A. S. Terekhov, JETP Pisma, 31, 581 (1980). 10. V. I. Zemskii, B. P. Zakharchenia, and D. N. Mirlin, JETP Pisma, 24, 96 (1976). 11. R. Kersting, K. Ulterrainer, G. Strasser, H. F. Kaufmann, and E. Gornik, Phys. Rev. Lett., 79, 3038 (1997). 12. T. Dekorsy, H. Auer, C. Waschke, H. J. Bakker, H. G. Roskos, H. Kurz, V. Wagner, and P. Grosse, Phys. Rev. Lett., 74, 738 (1995). 13. P. Gu, M. Tani, S. Kono, K. Sakai, and X.-C. Zhang, J. Appl. Phys., 91, 5533 (2002). 14. J. N. Heyman, P. Neocleous, D. Hebert, P. A. Crowell, T. Mueller, and K. Unterrainer, Phys. Rev., B, 64, 085202 (2001). 15. M. P. Hasselbeck, D. Stalnaker, L. A. Schlie, T. J. Rotter, A. Stintz, and M. SheikBahae, Phys. Rev. B, 65, 233203 (2002). 16. C. Affentaugscheggl and H. H. Wieder, Semicond. Sci. Technol., 16, 708 (2001). 17. R. Adomavičius, A. Urbanowicz, G. Molis, A. Krotkus, and E. Šatkovskis, Appl. Phys. Lett., 85, 2463 (2004). 18. R. Adomavičius, G. Molis, A. Krotkus, and V. Sirutkaitis, Appl. Phys. Lett., 87, 261101 (2005). 19. A. Krotkus, R. Adomavičius, G. Molis, A. Urbanowicz, and H. Eusebe, J. Appl. Phys., 96, 4006 (2004). 20. R. Dornhaus, G. Nimtz, and B. Schlicht, Narrow-gap Semiconductors (Springer, Berlin, 1983). 21. A. Urbanowicz, R. Adomavičius, A. Krotkus, and V. L. Malevich, Semicond. Sci. Technol., 20, 1010 (2005).

TERAHERTZ GENERATION BY MULTIPLICATION JAN STAKE*, TOMAS BRYLLERT, T. AREZOO EMADI, AND JOSIP VUKUSIC Department of Microtechnology and Nanoscience, Chalmers University of Technology, SE-412 96 Göteborg, Sweden

Abstract. We report on the status of symmetric varactor diode multipliers for signal generation in the terahertz frequency range. The progress and basic principles of heterostructure barrier varactor (HBV) diodes are presented. Furthermore, the design methodology and electro-thermal simulation results of high-power HBV multipliers for signal generation in the millimeter and submillimeter wave region are also presented. Finally, a state-of-the-art HBV tripler with an output power of 0.2 Watt at 113 GHz is presented.

Keywords: frequency multiplier, heterostructure barrier varactor (HBV), varactor diode, submillimeter wave generation, terahertz sources

1. Introduction Terahertz technology, in this chapter defined as technology using frequentcies from about 100 GHz to several THz, is gradually maturing and already used in many applications.1 The main applications today are within science, particularly radio astronomy.2 Applications under development or still at a planning stage are high-speed wireless networks, short-range high-resolution radar sensors, medical and biological imaging, high-speed inter-satellite communication, earth environment monitoring, military applications, and surveillance systems, notably security systems. Most such applications crucially depend on the availability of reasonably inexpensive, lightweight and compact sources and detectors. However, as the frequency approaches 1–2 THz, the output power for both electronic and photonic signal sources drops rapidly. This is commonly referred to as the THz gap. Traditional

______ * To whom correspondence should be addressed. Jan Stake, Department of Microtechnology and Nanoscience, Chalmers University of Technology, SE-41296 Göteborg, Sweden; e-mail:jan.stake@ chalmers.se

17 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 17–30. © 2007 Springer.

18

J. STAKE ET AL.

fundamental signal sources are backward-wave oscillators (BWO) that can generate a few milliwatts at 1.5 THz, and optically pumped lasers for spot frequencies between 0.5 and at least 5 THz with output power in the mW range. However, these sources are heavy, bulky, and very expensive. The quantum cascade laser (QCL) is a strong candidate for signal generation above ∼2 THz.3 The major problem is that the present technology requires cryogenic cooling at long wavelengths, which makes QCLs awkward to use. Another approach is photomixing4, which has shown promising results between 100 GHz and 1 THz, but the output power has only been in the µW range in the best case, apart from a recent demonstration of 20 mW at 100 GHz.5,6 Furthermore, Gunn devices are widely used as efficient and low noise sources in the millimeter wave region. State-of-the-art InP Gunn oscillators deliver an output power of about 5 mW around 300 GHz.7 The most promising approach to deliver reasonable power level in the terahertz frequency range, 0.2–2 THz, with a solid-state source is to use a frequency multiplier. Traditionally, this has been accomplished with a nonlinear capacitor such as a reverse-biased Schottky Diode (SD).8–10 However, the power-handling capability of the SD is limited by the device area, which has to be very small at high-operating frequencies. From a circuit point of view, higher harmonic multiplication factors (>×3) become increasingly difficult to implement and, therefore, high-order multipliers are usually realized as a chain of several low-order SD varactor multipliers.11 In order to provide compact solutions and sufficient power levels at terahertz frequencies for future heterodyne imaging systems there is a strong need to develop broadband, highly efficient, multipliers with a high multiplication factor (>3) and improved power-handling capability. The invention of the heterostructure barrier varactor (HBV) diode, at Chalmers,12 offers a very promising alternative to the SD varactor for signal generation in the submillimeter wavelength region. Since the HBV has a symmetric capacitance-voltage (C-V) characteristic, it operates unbiased and will only generate odd harmonics of the low-frequency signal, which simplifies the design of high-order multipliers (×3, ×5). Another important advantage with the HBV diode compared to the SD varactor is that several barriers can be epitaxially stacked, which increases the power-handling capability considerably. At short millimeter wavelengths, corresponding to a frequency of 250 GHz, 10 mW, and at least 10% efficiency have been demonstrated13–16 with HBVs. At high power levels, the effect of selfheating17 must be considered and the thermal resistance of the diode chip geometry should be minimized in order to keep the operating temperature reasonable.16,18,19

TERAHERTZ GENERATION BY MULTIPLICATION

19

Because of the apparent advantage of symmetric varactors for multiplication, i.e. exclusion of biasing and idler circuitry, another scheme has recently been proposed.20 By Anti-serially connecting two Schottky varactor (ASV) diodes similar I-V and C-V characteristics as for the HBV diode are obtained. Recently, published results for a tripler circuit employing ASVs grown on AlGaAs/GaAs showed an impressive conversion efficiency of 22% at 228 GHz.20 This approach has therefore emerged as very interesting alternative for high-frequency multiplication. This chapter focuses on the principles, progress and status of HBV multipliers. 2. Varactor Frequency Multipliers 2.1. BASIC CONCEPTS

A varactor is a nonlinear reactive device useful for harmonic generation, parametric amplification, mixing, detection, and voltage-variable tuning.21 Varactors normally exhibit a voltage-dependent capacitance and can be fabricated from a variety of semiconductor materials22. A common varactor is the reverse biased Schottky diode. Advantages of varactors are low loss and low noise. The maximum frequency of operation and performance is mainly limited by a parasitic series resistance, see Figure 1.

Figure 1. Equivalent circuit for a pure varactor.23

An important and extensively used figure-of-merit for varactors is the dynamic cutoff frequency, fc, which is defined as

fc =

S max − S min 2πRs

(1)

where Smax and Smin are the maximum and minimum elastances during a pump cycle, respectively, and Rs is the series resistance. The starting point for a varactor design24,25 is, hence, to maximize the elastance swing, Smax – Smin, and minimize any losses, Rs. For semiconductor varactors, the maximum elastance swing is limited by at least one of the following conditions:

J. STAKE ET AL.

20

• • • •

Depletion (modulation) layer punch-through, Large electron conduction from impact ionization, forward conduction current, and self-heating effects, Current saturation.26 The saturated electron velocity in the material determines the maximum length an electron can travel during a quarter of a pump-cycle, Large temperature rise due to dissipated power.

Increasing the pump power beyond any of the above conditions will result in reduced performance and probably introduce extra noise. In general, to analyze a varactor multiplier at high frequencies and drive levels, the conduction current through the device and the effect of device temperature should be included in the model.27

Figure 2. Comparison of HBV (left) and ASV (right) topologies for planar, symmetric diodes. The insets show the corresponding diode orientations.

During the early years of varactor diode development, the diodes were often mounted in a whisker contacted fashion, with waveguides serving as input/output feeds.12 During the last years, however, the majority of the designs are made in a planar28 topology and also fully integrated structures.9 Figure 2 shows a comparison of the planar HBV and ASV topologies. The inset shows the diode orientations for the two cases. 3. Heterostructure Barrier Varactor Multipliers 3.1. GROWTH AND DESIGN OF THE MATERIAL STRUCTURE

An HBV is a symmetric device composed of a high bandgap semiconductor (barrier) that is surrounded by moderately doped low band-gap semiconductors (modulation layers), see Table 1. In this work, the modulation and barrier layers are comprised of In0,53Ga0,47As and In0,52Al0,48As respectively. We also use a pseudomorphic

TERAHERTZ GENERATION BY MULTIPLICATION

21

(3 nm) AlAs layer in the centre of the barrier in order to increase the effective potential barrier, resulting in a very low-leakage current. In Figure 3 (TEM), the AlAs and the In0,52Al0,48As layers of the electron barrier are clearly shown. For this material system, an experimental study shows that the optimum total barrier thickness range is between 10–14 nm in order to minimize any conduction current.29 TABLE 1. A typical two-barrier InGaAs/InAlAs/AlAs HBV layer structure (Chalmers MBE1197)

No.

Layer

Material

15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

Contact Modulation Spacer Barrier Barrier Barrier Spacer Modulation Spacer Barrier Barrier Barrier Spacer Modulation Contact Substrate

InxGa1−xAs In0.53Ga0.47As In0.53Ga0.47As In0.52Al0.48As AlAs In0.52Al0.48As In0.53Ga0.47As In0.53Ga0.47As In0.53Ga0.47As In0.52Al0.48As AlAs In0.52Al0.48As In0.53Ga0.47As In0.53Ga0.47As In0.53Ga0.47As InP

Thickness [nm] 260 250 5 3,5 3 3,5 5 250 5 3,5 3 3,5 5 250 500

Doping [cm–3] ∼1019 1×1017 undoped undoped undoped undoped undoped 1×1017 undoped undoped undoped undoped undoped 1×1017 ∼1019 SI

Figure 3. TEM picture of the barrier region (Chalmers MBE1197).

22

J. STAKE ET AL.

Optimum epi-layer structures for high-power applications have been designed and studied. These multi-barrier structures were grown using the molecular beam epitaxy (EPI 930) at the Nanofabrication laboratory, Chalmers. The optimum doping concentration and layer thicknesses are calculated by optimisation of the dynamic cutoff frequency.25,30 The main goal is a low series resistance and a large elastance (1/C) modulation swing and there is a trade-off with respect to the doping concentration in the modulation layer, pump frequency, and maximum voltage. The barrier structure is designed to minimize conduction current through the structure. Calibration of the growth conditions is crucial since the HBV structure contains relatively thick epi-layers (1–5 µm), heterostructures, as well as strained layers. Top quality of the epi-growth is needed, translating into high electron mobility, high level of symmetry and excellent barrierblocking characteristics. All epi-materials are analyzed during growth and characterized after growth using TEM, x-ray diffractometer (lattice strain) and by fabricating large test diodes (A ≈ 3,000 µm2) for I–V and C-V characterization. Measurements of a three-barrier HBV material (Chalmers MBE995) are shown in Figure 4. The Chalmers MBE995 has a modulation layer thickness of l = 370 nm, buffer layer thickness = 0.5 µm, barrier thickness of b = 15 nm, doping concentration of Nd = 1 × 1017 cm–3.

Figure 4. C-V and I-V measurements for a three-barrier HBV material, Chalmers MBE995.

TERAHERTZ GENERATION BY MULTIPLICATION

23

3.2. HBV DEVICE MODEL

The parallel plate capacitor model, where the plate separation should be replaced with the sum of the barrier thickness, b, the spacer layer thickness, s, and the length of the depleted region, w, is normally an adequate description of the differential capacitance. The depletion length is bias dependent and the layer structure is symmetric, therefore the elastance is an even function of applied voltage and is given by:

S=

1 N⎛ b s w⎞ = ⎜ + + ⎟ C A ⎝ εb εd εd ⎠

2εd Vd w= qN d

(2)

where Vd is the voltage across the depleted region, Nd is the doping concentration in the modulation layers, b is the barrier thickness, s is the undoped spacer layer thickness, A is the device area, εb and εd are the dielectric constants in the barrier material and modulation layers, respectively. The maximum capacitance or the minimum elastance, Smin, occurs at zero bias. However, due to screening effects, the minimum elastance, Smin, must include the extrinsic Debye length, LD, as:-

Smin = LD ≡

1 N ⎛ b 2s 2L ⎞ = ⎜ + + D⎟ εd ⎠ Cmax A ⎝ εb εd

εd kT

(3)

q2N d

To achieve a high Cmax/Cmin ratio, the screening length can be minimized with a sheet doping, Ns, at the spacer/depletion layer interface. The minimum capacitance, Cmin, is normally obtained for punch-through condition, i.e. w = l, or when the breakdown voltage, Vmax, is reached. An accurate quasi-empirical expression for the C-V characteristic of homogeneously doped HBVs has been derived by Dillner et al.31 The voltage across the nonlinear capacitor is expressed as a function of its charge as: Q ⎛ ⎛ ⎛ ⎞⎞⎞ − bQ sQ Q2 4kT ⎜ 2L D AqN d ⎟⎟⎟ (4) V (Q,T) = N ⎜ +2 + Sign(Q)⎜ + 1− e 2 ⎜ ⎟ ⎜ ⎟ ⎜ εb A ⎟ q ⎝ εd A 2qN d εd A ⎠⎠⎠ ⎝ ⎝

J. STAKE ET AL.

24

where T is the device temperature, q is the elementary charge, and Q is the charge stored in the HBV. The series resistance Rs associated with a varactor diode is a quantity, which summarizes the resistive losses that characterize various layers and connections forming the device. The main contributions to the series resistance are the ohmic contact resistance, the resistance in the modulation layers, and the spreading resistance. Given the voltage-charge relationship in (4), the series resistance can be expressed as25

⎛

⎛ − Q ⎞⎞ 2L D AqN d ⎜ ⎟ (5) R(Q,T) = Rs − + 2LD e −1 ⎟ ⎜ ⎟⎟ A ⎜ qN d A ⎝ ⎠⎠ ⎝ where Rs is the zero-bias series resistance and ρd is the resistivity of the

ρd N ⎜ Q

modulation layer,

ρd =

1 . qN d µe (T, N d )

(6)

The conduction current is a mixture of thermionic emission and tunneling through the barrier and hence difficult to model with a physical analytical expression. However, an empirical model based on the Sinh(x) function can often be used to describe the antisymmetrical I–V characteristic of the HBV. A complete electro-thermal model (Chalmers HBV Model) can be found in reference by Ingvarson et al.27 3.3. HBV MULTIPLIER CIRCUITS

3.3.1. Status of HBV multiplier research An HBV quintupler (×5) with a state-of-the-art conversion efficiency of 11% has been demonstrated at 100 GHz.32 HBV triplers (×3) have been shown to provide 200 mW at 114 GHz33 and also recently a state-of-the-art efficiency of more than 21% at an output frequency of 102 GHz has been achieved.34 At short millimeter wavelengths, corresponding to a frequency of 200–300 GHz, 10 mW and at least 10% efficiency have been demonstrated.13–16 In terms of output power, the best results have been achieved using a filter circuit on AlN instead of quartz (see Figure 5). The AlN substrate provides a better heat-sink for the flip-chip mounted diode. According to detailed analysis, there is room for substantial improvements in terms of output power in the millimeter wavelength range 27,35 with a proper electrical and thermal design. Our predictions suggests an output power of >0.5 W at 100 GHz, 50–100 mW around 200 GHz, 10–20 mW around 500 GHz and ∼1mW at 1 THz for high-power single HBV diode

TERAHERTZ GENERATION BY MULTIPLICATION

25

multipliers. Such power levels have not been demonstrated using Schottky diode (SD) multipliers. This is especially true when compared with SD triplers.36 More important, the overall conversion efficiency, reliability, cost and weight are improved with less number of cascaded modules for a complete terahertz source.

Figure 5. Reported output power for HBV multipliers (September 2006). The red line shows the predicted capability of the HBV technology (single device).

3.3.2. A high-power millimeter wave tripler In this section, we describe the development of a high-power millimetre wave tripler. The design flow can be summarized in a number of steps: • • • •

The diode geometry (area, number of barriers, layer structure, etc.) is chosen based on parameters like input power and operating frequency. The optimum diode-embedding impedances are simulated using harmonic balance techniques, using the electro-thermal HBV device model developed at Chalmers. 27 Embedding circuits, including waveguide probes are designed using a combination of ideal circuit simulations (Agilent ADS) and 3D EM simulations (Ansoft HFSS). The final circuit response is simulated using the nonlinear device model and extracted S-parameters from the completed embedding circuitry including waveguide feeds.

26

J. STAKE ET AL.

Figure 6. SEM picture of a 500-µm2 HBV. This 4-mesa, 12-barrier device is fabricated from Chalmers MBE1038 epitaxial material.

Figure 7. 113-GHz high power tripler. AlN microstrip circuit mounted in the waveguide block.

In the current design a 3-barrier epi-material is used, and the diode consists of four series-connected mesas in order to increase the powerhandling capability even further, see Figure 6. A major limitation for highpower multipliers is the self-heating of the devices; this results in lower efficiency and device breakdown. To improve the thermal properties, the microstrip circuit is fabricated on an AlN substrate – AlN is a material with excellent thermal conductivity (~175 W/mK). Both the HBV chip and the AlN substrate were lapped down to 40/100 µm, respectively. The hybrid circuit was then mounted in the waveguide block with a WR-22/WR-10 input and output waveguides respectively. Figure 7 shows the hybrid circuit

TERAHERTZ GENERATION BY MULTIPLICATION

27

mounted in one half of the waveguide block. Note the fixed tuned back shorts, which are part of the matching circuitry. Figure 8 shows the output power and flange-to-flange conversion efficiency versus available input power. The tripler exhibits a maximum output power of 195 mW at 113 GHz. The performance can be improved further with an improved heatsink, e.g. monolithic HBV circuit, and further optimized matching circuit.

Figure 8. Output power and conversion efficiency for a 3×37.7 GHz frequency multiplier. The measurement result is from a 700-µm2, 12-barrier HBV.

4. Discussion The output power around 100 GHz for single HBV diode triplers is comparable to state-of-the-art Schottky doublers.8,37 This statement is true despite large investments during many years in the development of SD and related multiplier circuits. Obviously, concerning the HBV technology large improvements can be gained by optimizing the devices and the circuits. Furthermore, HBV multipliers with output frequencies above 500 GHz have not yet been reported.

28

J. STAKE ET AL.

Acknowledgments This work is supported by the European Space Agency (“HBV Devices and MMICs”, No.18164/04/NL/LvH), the Swedish Foundation for Strategic Research (HSEP), and the Swedish Defence Research Agency (FOI). We would like to thank Mahdad Sadeghi at the Nanofabrication laboratory, Chalmers for providing MBE grown HBV materials.

References P. H. Siegel, Terahertz technology in biology and medicine, IEEE Trans. Microw. Theory Tech., 52, 10, 2438–2447 (2004). 2. J. Grahn, P. Starski, J. Stake, and S. Cherednichenko, Novel devices for (sub)millimeterwave space applications, IEICE Trans. Electron., E89-C, 7, 891–897 (2006). 3. R. Köhler, Terahertz semiconductor-heterostructure laser, Nature, 156–159 (May 9, 2002). 4. H. Ito, S. Kodama, Y. Muramoto, T. Furuta, T. Nagatsuma, and T. Ishibashi, High-speed and high-output InP-InGaAs unitraveling-carrier photodiodes, IEEE J. Sel. Top. Quantum Electron., 10, 4, 709–727 (2004). 5. H. Ito, T. Ito, Y. Muramoto, T. Furuta, and T. Ishibashi, Rectangular waveguide output unitraveling-carrier photodiode module for high-power photonic millimeter-wave generation in the F-band, J. Lightwave Technol., 21, 12, 3456–3462 (2003). 6. F. Nakajima, T. Furuta, and H. Ito, High-power continuous-terahertz-wave generation using resonant-antenna-integrated uni-travelling-carrier photodiode, Electron. Lett., 40, 20, 1297–1299 (2004). 7. H. Eisele and R. Kamoua, High-performance oscillators and power combiners with InP Gunn devices at 260–330 GHz, Microw. Wireless Components Lett., IEEE [see also IEEE Microwave Guided Wave Lett.] 16, 5, 284–286 (2006). 8. J. Ward, E. Schlecht, G. Chattopadhyay, A. Maestrini, J. Gill, F. Maiwald, H. Javadi, and I. Mehdi, Capability of THz sources based on Schottky diode frequency multiplier chains, in IEEE International Microwave Symposium, vol. 3, pp. 1587–1590 (2004). 9. I. Mehdi, G. Chattopadhyay, E. Schlecht, J. Ward, J. Gill, F. Maiwald, and A. Maestrini, Terahertz multiplier circuits, in IMS 2006 (2006). 10. T. W. Crowe, D. W. Porterfield, J. L. Hesler, W. L. Bishop, and D. S. Kurtz, Integrated terahertz transmitters and receivers, 1 edn., SPIE-International Society for Optical Engineering Providence, RI, USA, pp. 1–9 (2003). 11. G. Chattopadhyay, E. Schlecht, J. Ward, J. Gill, H. Javadi, F. Maiwald, and I. Mehdi, An all-solid-state broad-band frequency multiplier chain at 1500 GHz, IEEE Trans. Microw. Theory Techn., 52, 5, 1538–1547 (2004). 12. E. L. Kollberg and A. Rydberg, Quantum-barrier-varactor diode for high efficiency millimeter-wave multipliers, Electron. Lett. 25, 1696–1697 (1989). 1

TERAHERTZ GENERATION BY MULTIPLICATION

29

13. X. Mélique, J. Carbonell, R. Havart, P. Mounaix, O. Vanbésien, and D. Lippens, InGaAs/InAlAs/AlAs heterostructure barrier varactors for harmonic multiplication, IEEE Microw. Guided Wave Lett., 8, 7, 254–256 (1998). 14. R. Meola, J. Freyer, and M. Claassen, Improved frequency tripler with integrated singlebarrier varactor, Electron. Lett., 36, 9, 803–804 (2000). 15. L. Dillner, W. Strupinski, S. Hollung, C. Mann, J. Stake, M. Beardsley, and E. L. Kollberg, Frequency multiplier measurements on heterostructure barrier varactors on a copper substrate, Electron Device Lett., 21, 5, 206–208 (2000). 16. Q. Xiao, J. L. Hesler, T. W. Crowe, R. M. Weikle, and Y. Duan, High efficiency heterostructure-barrier-varactor frequency triplers using AlN substrates, in IMS 2005 Long Beach, CA, (2005). 17. J. Stake, L. Dillner, S. H. Jones, C. M. Mann, J. Thornton, J. R. Jones, W. L. Bishop, and E. L. Kollberg, Effects of self-heating on planar heterostructure barrier varactor diodes, IEEE Trans. Electron Devices 45, 11, 2298–2303 (1998). 18. M. Ingvarson, B. Alderman, A. Ø. Olsen, J. Vukusic, and J. Stake, Thermal constraint for heterostructure barrier varactor diodes, IEEE Electron Device Lett., 25, 11, 713–715 (2004). 19. T. A. Emadi, J. Vukusic, M. Ingvarson, A. Ø. Olsen, T. Bryllert, E. L. Kollberg, and J. Stake, High Power HBV Multipliers for F- and G- Band Applications, IRMMW 2004/THz 2004 M. Thumm, and W. Wiesbeck, Karlsruhe, Germany, pp. 319–320 (2004). 20. M. Krach, J. Freyer, and M. Claassen, Power generation at millimetre-wave frequencies using GaAs/GaAlAs triplers, Physica Status Solidi C: Conferences 1, 8, 2160–2182 (2004). 21. P. Penfield and R. P. Rafuse, Varactor Applications (MIT Press, Cambridge, 1962). 22. S. M. Sze, Physics of Semiconductor Devices, 2nd edn. (Wiley, Singapore, 1981). 23. A. Uhlir, The potential of semiconductor diodes in high frequency communications, Proc. IRE 46, 1099–1115, (1958). 24. C. C. H. Tang, An exact analysis of varactor frequency multiplier, IEEE Trans. Microw. Theory Tech. April, 210–212, (1966). 25. J. Stake, S. H. Jones, L. Dillner, S. Hollung, and E. L. Kollberg, Heterostructure barrier varactor design, IEEE Trans. Microw. Theory Techn., 48, 4, Part 2 (2000). 26. E. L. Kollberg, T. J. Tolmunen, M. A. Frerking, and J. R. East, Current saturation in submillimeter wave varactors, IEEE Trans. Microw. Theory Tech., 40, 5, 831–838, (1992). 27. M. Ingvarson, J. Vukusic, A. Ø. Olsen, T. A. Emadi, and J. Stake, An electro-thermal HBV model, in IMS 2005, Long Beach, CA, pp. 1151–1153 (2005). 28. W. L. Bishop, K. McKiney, R. J. Mattauch, T. W. Crowe, and G. Green, A novel whiskerless Schottky diode for millimeter and submillimeter wave applications, in 1987 IEEE MTT-S International Microwave Symposium Digest IEEE, Las Vegas, NV, pp. 607–610 (1987). 29. A. T. Emadi, T. Bryllert, M. Sadeghi, J. Vukusic, and J. Stake, Optimum barrier thickness study for the InGaAs/InAlAs/AlAs heterostructure barrier varactor diodes, Submitted to Appl. Phys. Lett. (2006).

30

J. STAKE ET AL.

30. M. Ingvarson, A. Ø. Olsen, and J. Stake, Design and analysis of 500 GHz heterostructure barrier varactor quintuplers, in 14th International Symposium on Space TeraHertz Technology, Tucson, AZ, (2003). 31. L. Dillner, J. Stake, and E. L. Kollberg, Modeling of the heterostructure barrier varactor diode, in International Semiconductor Device Research Symposium (ISDRS), Charlottesville, pp. 179–182 (1997). 32. T. Bryllert, A. Ø. Olsen, J. Vukusic, T. A. Emadi, M. Ingvarson, J. Stake, and D. Lippens, 11% efficiency 100 GHz InP-based heterostructure barrier varactor quintupler, Electron. Lett., 41, 3, 30 (2005). 33. T. Bryllert, J. Vukusic, T. A. Emadi, and J. Stake, A high-power frequency tripler for 100 GHz, in The Joint 31st International Conference on Infrared and Millimeter Waves and 14th International Conference on Terahertz Electronics, Shanghai, China, pp. 30 (2006). 34. J. Vukusic, B. Alderman, T. A. Emadi, M. Sadeghi, A. Ø. Olsen, T. Bryllert, and J. Stake, HBV tripler with 21% efficiency at 102 GHz, Electron. Lett., 42, 6, 355–356 (2006). 35. T. A. Emadi, J. Vukusic, M. Ingvarson, M. Sadeghi, T. Bryllert, A. Ø. Olsen, and J. Stake, Design, fabrication and characterisation of high power HBV diodes, in International Symposium on Space Terahertz Technology, J. Stake and H. Merkel, Eds., Chalmers, Göteborg, Sweden, pp. 285–290 (2005). 36. T. W. Crowe, T. C. Grein, R. Zimmermann, and P. Zimmermann, Progress toward solidstate local oscillators at 1 THz, IEEE Microw. Guided Wave Lett., 6, 5, 207–208 (1996). 37. D. Porterfield, J. Hesler, T. Crowe, W, Bishop, and D. Woolard, Integrated terahertz transmit/receive modules, Proceedings of the 33rd European Microwave Conference, vol. 3, pp. 1319–1322 (2003).

TOWARDS SUPERLATTICE TERAHERTZ AMPLIFIERS AND LASERS ALVYDAS LISAUSKAS†, ERNST MOHLER, AND HARTMUT G. ROSKOS Physikalisches Institut, Johann Wolfgang Goethe-Universita¨t, Max-von-Laue-Str. 1, D-60438 Frankfurt am Main, Germany NATALIYA V. DEMARINA Electronics Department, Radiophysics Faculty, Nizhny Novgorod State University, Gagarin Avenue 23, Nizhny Novgorod 603950, Russia

Abstract. We describe our work towards THz sources which employ the “Bloch gain ” , a stimulated-emission mechanism which has been predicted as early as 1971 to exist for semiconductor superlattices but which researchers – in spite of much recent work – have not yet been able to take advantage of for the implementation of THz amplifiers and lasers. From a basic-physics point-of-view, the interest in Bloch gain arises from its dynamical, second-order character, involving simultaneous scattering of an electron and emission of a THz photon. This aspect has the practically important implication that the temperature dependence of the gain is determined to a large degree by the optical-phonon energy scale and not that of the photon energy, with the consequence that there is a rather slow roll-off of the gain with temperature. This feature together with the rather high-gain values which are calculated to be comparable with those of THz quantum cascade lasers at low temperature, fosters the hope that a Bloch THz laser could be the first semiconductor-based THz laser operating at room temperature. Keywords: semiconductor superlattice, terahertz radiation, terahertz lasing

1. Introduction The lack of compact and tunable sources of terahertz (THz) radiation together with the exciting physical properties of semiconductor superlattices have stimulated the experimental search for a novel laser gain mechanism often termed as Bloch gain. Shortly after the proposal of semiconductor superlattices,1 it was predicted that electrons oscillating at the Bloch frequency (ωB ) in the presence of dissipative scattering should provide gain at frequencies 2 †

To whom correspondence should be addressed, Physikalisches Institut, Johann Wolf¨ Max-von-Laue-Str. 1, D-60438 Frankfurt am Main, Germany, e-mail gang Goethe-Universitat, address: (lisauskas @ physik.uni-frankfurt.de) 31 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 31–40. © 2007 Springer.

32

A. LISAUSKAS ET AL.

ω < ωB . Recently, a series of theoretical investigations either within the semiclassical picture 3−5 or based on quantum-mechanical approaches 6 have provided support for this surprising prediction. A practical interest arises from the fact that the gain of this second-order process is predicted to be tens per centimeter 6at low temperatures thus favorably competing with the gain of THz quantum cascade lasers which, however, exhibit population inversion . 7 − 9 Equally important, theory predicts that the decline of the gain with temperature is significantly slower in the case of the Bloch gain. As we will discuss in more detail later, the gain in optimized superlattice structures should be large enough to permit laser operation at room temperature. 2. The Challenge of Finding Experimental Evidence for Bloch Gain Despite the sizeable number of theoretical investigations, no conclusive experimental proof for inversionless gain has ben produced yet. The key practical challenge is the tendency of superlattices to break away from the homogeneous-field situation which has been the idealized basis of nearly all theoretical treatments of the phenomenon. As this tendency is a result of the same nonlinear current-voltage characteristics associated with a negative differential velocity which also brings forth the Bloch gain itself, and as it becomes more severe as one tries to optimize the structure and the doping concentration of the superlattice with respect to the gain value, one deals indeed with a profound challenge. Researchers have been able to take advantage of the field instability and implemented novel Gunn-type superlattice oscillators for the sub-millimeter wavelength regime.10 − 12 While they were able to reach frequencies as high as 147 GHz directly and 300 GHz with an intrinsic tripling process 13 , the hope to simultaneously gain access to higher frequencies by Bloch gain proved futile. In fact, it has been shown theoretically that Bloch gain is quenched in a Gunn-type operation of a superlattice oscillator 14 . In a different approach, it was attempted to obtain evidence for the existence of Bloch gain by all-optical time-resolved experiments in combination with data evaluation in the framework of semiclassical theory 15, 16 . The authors measured the THz transients emitted from undoped GaAs/AlGaAs superlattices after excitation of exciton wavepackets with femtosecond laser pulses. Analyzing the Fourier spectra of the coherent THz pulses in terms of a frequency-depenent conductivity, the authors found the real part of this quantity to be negative and its spectra to exhibit a shape as predicted by the semiclassical transport theory. This was then interpreted as clear evidence for Bloch gain. In the meantime, we have pointed out that this approach is incorrect . 17,18 The correct quantity because the conductivity has been defined wrongly

TOWARDS SUPERLATTICE TH z AMPLIFIERS AND LASERS

33

to investigate would have been the small-signal conductivity at the operation bias voltage above the critical voltage (defined by the maximum of the current/voltage curve). This conductivity is, however, not directly accessible in this kind of THz-emission experiments. The quantity studied, in contrast, is a large-signal response function involving a linearization across the entire highly nonlinear current/voltage curve of the superlattice from zero bias up to the operation point. This reponse function cannot be compared with the (small-signal) conductivity from the semiclassical theory as has been done , 15,16 and it is hence not correct to infer Bloch gain by such a comparison. The closest any experiment has come to prove the existence of Bloch gain until now, is a THz-beam amplification experiment performed by Allen et al. at the THz-free-electron-laser facility at UCSB.19 The amplification medium was a waveguide structure containing pillars of electrically pumped stacks of 34 short-length InAs/AlSb superlattices. Each of the latter consisted of only 15 superlattice periods and subsequent superlattices were separated by 100 nm-thick highly doped field-pinning layers. With n+ -InAs contact layers on top and bottom of the stack, each pillar had a thickness of about 10 µm. When the superlattices were biased within the correct voltage range, the structure showed an increased THz transmittance as expected if amplification of the radiation from the free-electron laser in the pillars sets in. Self-starting laser action, however, has not been reported until now. With the stacking of short superlattice segments, the authors seem to have found an effective way to prevent the formation of field domains. In the design, they applied Kroemer’s rule to keep the length L of each superlattice segment below the domain-formation length: nL < 70 Fc /e, with n being the total carrier density in the superlattice,  and 0 denoting the relative dielectric permittivity of the superlattice and the vacuum permittivity, respectively, Fc representing the critical electric field of the superlattice, and e the elementary charge. While their design does not strictly fullfil Kroemer’s condition, it seems to be sufficiently close to keep field inhomogeneities insignificant. 3. Field Dynamics in a Superlattice In order to understand the field dynamics in superlattices better, we have investigated undoped superlattices and traced the spatial distribution of optically excited charge carriers along the superlattice growth direction as a function of time. This is possible, within limits, by pump-probe photocurrent spectroscopy at low temperatures because of the following reason. The carrier densities of interest for an amplifier or laser employing Bloch-gain are low in the 1015− 1016 cm−3 range, just as in the case of the quantum cascade

34

A. LISAUSKAS ET AL.

laser and for similar reason: to keep space-charge effects and THz-radiation losses limited. As a consequence, the sharp excitonic absorption lines are hardly broadened yet, and allow, with the strong dependence of their peak wavelength on the local electric field, to monitor the temporal evolution of the field, albeit in an integral manner, averaged over the growth direction of the superlattice and weighed there with the distribution of the charge carriers generated by the probe laser radiation. We have then applied the semiclassical transport theory to model and fit the measured data. The calculated data and the measured ones agree well with each other during the time scale of the sweep out of the electrons (nonnegligible deviations show up for longer time scales dominated by sweep-out of the holes, and seem to result from approximations of the model calculations which do not take into account the full complexity of the valence band with its warping and population by heavy and light holes which can also change their character by scattering 20 ). As any practical device would work with electrons and not holes, we seem to capture the relevant part of the field dynamics. Figure 1 shows data of model calculations which were performed to explain experimental results measured with an intrinsic GaAs/Al 0.3 Ga0.7 As superlattice. It consisted of 35 periods of 6.7 nm-wide wells and 1.7 nm-wide barriers and was grown by molecular beam epitaxy. With a width of the first electron miniband (∆) of 33 meV, it is in the intermediate coupling regime. Bloch-gain lasers and amplifiers would be realized with more strongly coupled systems, which would exhibit a considerably faster carrier transport and a correspondingly faster field dynamics. While it would be more difficult to temporally resolve the dynamics with this type of photocurrent spectroscopy, we expect that the results for the electron dynamics would be similarly well explained by the semiclassical model. The spatiotempoal evolution of electron density and electric field plotted in Figure 1 paints the following picture. First, the electrons sweep out of the 350 nm-thick superlattice on a time scale of only 10 ps. It is an important result of our study that experiment and semiclassical transport theory with plausible device parameters are in agreement with each other here which raises the level of confidence with respect to other predictions of the theory. Second, the drift of the charge carriers rapidly leads to a field pattern which can be identified as the onset of domain formation, with a region with a strong field gradient at the left side of the superlattice and a low-field, low-gradient region on the right side where the electrons leave the superlattice. At the carrier density chosen for this example, the domain formation is not complete before the electron density decreases too much. In other words, no region in the superlattice experiences full-field screening yet. This is consistent with an evaluation of Kroemer’s relation which predicts that the domain formation length should

TOWARDS SUPERLATTICE THz AMPLIFIERS AND LASERS

35

Figure 1. Results of simulations with parameters adjusted to obtain good agreement with the experimental data. Left panel: Depth profile of the electron density as a function of time in steps of 2 ps. Laser pulse generates electron-hole pairs in the undoped superlattice at t = 0 ps. Right panel: Corresponding depth profile of the electrical field. Model parameters: µe = 1,200 cm2 /Vs, µh = 100 cm2 /Vs, initial carrier density ne = 3.3 · 1015 cm−3 and starting field of 22.8 kV/cm.

slightly exceed the length of the superlattice for the chosen parameters. This result strengthens the confidence in the applicability of Kroemer’s formula. We point out that Bloch gain is fairly robust against field inhomogeneities. This is a consequence of the broadband nature of the gain which spans the frequency range from zero up to nearly the circular Bloch frequency ωB given by ~ωB = eFd, d being the spatial period of the superlattice and F the local field. If F decreases somewhere in a superlattice, then the spectral range over which gain is provided shrinks here, while the gain at low frequencies is enhanced. One must, however, strictly avoid full-domain formation because the gain vanishes in regions where the field drops to zero. 4. Estimates of Gain Values and Their Temperature Dependence Having corroborated that short-period superlattices indeed represent a practical solution of the domain-formation problem, we now estimate absolute values of the gain at specific target frequencies and as a function of temperature. We base these estimates on the following analytical expression for the small-signal gain given by 6 within the framework of a semiclassical single-relaxation-time model: 21 α(ω) =

h 1 − ω2B τ2 − iωτ i τ e2 d2 ∆ I1 (∆/2kB T ) n × Re . 0 nr c 2~2 I0 (∆/2kB T ) 1 + ω2B τ2 ω2B τ2 + (1 − iωτ)2

(1)

Here, kB is Boltzmann’s constant, nr the refractive index, and τ the scattering time constant of the dissipative processes. I1 (∆/2kB T )/I0 (∆/2kB T ) denotes

36

A. LISAUSKAS ET AL.

Gain, cm-1

100

dd==8.4 8.4nm nm n = 5x101515cm-3-3 n = 5x10 cm t = 200 fs t = 200 fs

1 THz

80

2 THz

60

3 THz

DD = =80 80meV meV

40 2 THz

1 THz = 36 meV D =D36 meV

20

3 THz

0 0

50

100 150 200 Temperature, K

250

300

Figure 2. Temperature dependence of Bloch gain calculated with Willenberg formula for two superlattices of different width ∆ of the first electron miniband. At each data point, the electrical field F across the superlattice has been varied in order to maximize the gain value.

the ratio of Bessel functions of orders one and zero, and describes the temperature dependence of the gain for a nondegenerate electron gas. The electric field strength F enters through the Bloch frequency ωB . While Monte Carlo simulations 22 have clearly identified the emission of longitudinal optical (LO) phonons as the dissipative process which more than any other determines the strength of the Bloch gain, this relaxation process does not enter explicitly into the Willenberg formula. The only parameter which reflects any specific scattering mechanism is the relaxation time τ. For our calculations, we select τ to be 0.2 ps, a nearly temperatureindependent value which is typical for optical-phonon emission in situations when the electrons have enough kinetic energy to emit an LO phonon of energy 36 meV. Figure 2 displays the Bloch gain calculated for two superlattices of different electron miniband width, but for the same superlattice period d = 8.4 nm (hence the ratio of the width of the wells to the width of the barriers changes as ∆ is varied). We assume a fixed electron density of n = 5 · 1015 cm−3 . The temperature dependence of the gain is plotted for three target frequencies, 1 THz, 2 THz, and 3 THz. In the simulations, the electrical field has been treated as an optimization parameter for the Bloch gain and hence, varies even from data point to data point. For any temperature and radiation frequency, the gain value of the superlattice with intermediate coupling strength (∆ = 36 meV) is always lower

TOWARDS SUPERLATTICE THz AMPLIFIERS AND LASERS

37

Figure 3. Literature data of the theoretical thermal dependence of the gain in THz quantum cascade lasers. Left panel: Data taken from Williams et al. 2003b, for a radiation frequency of 3.4 THz; the horizontal lines indicate typical losses of metal – metal waveguides and of single-plasmon waveguides. Right panel: Data taken from Callebaut and Hu, 2005, for a radiation frequency of 3.2 THz.

than that of the superlattice with strong coupling (∆ = 80 meV). For stronger coupling, the temperature dependence is also less pronounced. Another interesting aspect is the frequency dependence of the gain. Independently of the coupling strength, lower frequencies find a larger gain. It is worth mentioning that this behavior is opposite to that found in the case of quantum cascade lasers .9 The two decisive questions which decide over the fate of the Bloch gain on the long run, are the following: (i) Is the gain at any temperature large enough to be of practical interest, and (ii) is it large enough at 300K to potentially allow for laser operation at room temperature. In order to give at least a preliminary answer to this question – i.e. one which is only based on the calculations with the Willenberg formula we compare the data of Figure 2 for the Bloch gain in superlattices with gain data of state-of-the-art THz quantum cascade lasers. Figure 3 displays calculated data for the latter from two source .23,24 The left panel displays gain values for a laser operating at 3.4 THz which do not exceed 35 cm−1 at any temperature. The panel also displays typical waveguide losses encountered in such lasers (which are optimized for low losses): 18 –19 cm−1 for the metal – metal waveguide and 30 –31 cm−1 for the single-plasmon waveguide. Given the typical temperature dependence of the gain in THz quantum cascade lasers, the gain equals the losses at 80 – 90K, respectively 160K in the two cases. That the gain values of this panel are not singular in the literature is proven by the data shown in the right panel of Figure 3. Here, two theoretical approaches are compared with each other, a semiclassical one and one based on density-matrix

38

A. LISAUSKAS ET AL.

Monte Carlo simulations. While the semiclassical calculations predict gain values and a temperature dependence of the gain very similar to the data in the left panel, the more complex density-matrix calculation predict even higher gain up to 53 cm−1 at low temperatures. If we now compare these data with those of Figure 2 for a radiation frequency of 3 THz, we find that the superlattice with intermediate coupling strength exhibits somewhat lower gain values that the quantum cascade lasers. The gain of the strongly coupled superlattice, on the other hand, exceeds that of the quantum cascade lasers for all temperatures. The decrease of the gain with rising temperature is more gradual than in the case of the quantum cascade lasers. At room temperature, the gain still amounts to 35 cm−1 . This answers the first question: The gain compares well with that of the established lasers which implies that it is of practical interest. This holds even more so with respect to the second question: the gain of the strongly coupled superlattice is predicted to be sufficiently high to exceed the waveguide losses even at room temperature. Considering such promising predictions, especially with respect to the temperature dependence of the gain, one has to recall that the Willenberg formula does not treat the relevant scattering process of LO-phonon emission explicitly. Nevertheless, more sophisticated semiclassical theory based on single-particle Monte Carlo simulation of the electron motion in the presence of scattering at optical and acoustic phonons22 also corroborates high values of THz gain. It was shown that for a large amplitude of an applied THz field a superlattice displays larger amplification coefficient at the frequency slightly below the Bloch frequency, i.e. amplification coefficient increases with increasing frequency. It also grows larger with a miniband width. For the superlattice with a miniband width of about 90 meV it reaches about 40 cm−1 for frequency close to 3 THz, the electron density 5 · 1015 cm−3 and lattice temperature of 4K. This value is comparable with the gain number for a quantum cascade laser (Figure 3). It decreases only by 30% for lattice temperature of 300 K. The latter apparently implies that a semiconductor superlattice is a promising candidate for a terahertz emitter operating at room temperature. Such exciting results urge for further more detailed theoretical studies (and certainly experiments) in order to explore the validity of the predictions. 5. Summary We have explored some aspects of the Bloch gain of semiconductor superlattices with respect to the question whether it may provide a viable alternative to the realization of lasers for the terahertz frequency range. Common wis-

TOWARDS SUPERLATTICE THz AMPLIFIERS AND LASERS

39

dom leads one to assume that the second-order character of the processes involved limits the attainable gain values. Quite in contrast, estimates of the gain values for optimized superlattice structures let it appear competitive with the first-order gain of terahertz quantum cascade lasers. Furthermore, the temperature dependence is predicted to be more gradual promising laser operation even at room temperature. A third advantage of the second-order nature of the gain (not discussed in detail here) could be its extremely large bandwidth which promises tunability of a laser over a huge terahertz frequency range, provided this large gain bandwidth does not hinder the self-starting of a laser. With all its interesting and auspicious properties, it is striking that the experimental evidence even for the mere existence of the Bloch gain is still extremely scarce. This has to do on one hand with the considerable technological effort needed to explore it, but on the other hand also reflects the fundamental challenges imposed by this strongly nonlinear system. The most obvious one, the tendency to form field domains, however, now appears to get under control by measures such as the implementation of stacked short-period superlattices with a thickness below the finite domain-formation length. The stage seems set for significant experimental progress in the next few years. Acknowledgments Funding by DFG and NATO is acknowledged. We are grateful to K. K¨ohler for providing us top-quality superlattice samples over the many years of cooperation. One of us (NVD) thanks the Russian Agency of Education and CRDF (grant BF5M01) and RFBR for support.

References 1. L. Esaki and R. Tsu, Superlattice and negative conductivity in semiconductors, IBM J. Res. Dev. 14, 61–65 (1970). 2. S. A. Ktitorov, G. S. Simin, and Y. Sindalovski, Bragg reflections and high-frequency conductivity of an electronic solid-state plasma, Sov. Phys. Solid State 13, 1872 (1972). 3. A. A. Ignatov, K. F. Renk, and E. P. Dodin, Esaki-tsu superlattice oscillator – josephson-like dynamics of carriers, Phys. Rev. Lett. 70, 1996 –1999 (1993). 4. E. Schomburg, N. V. Demarina, and K. F. Renk, Amplification of a terahertz field in a semiconductor superlattice via phase-locked k-space bunches of bloch oscillating electrons, Phys. Rev. B 67, 155302 (2003). 5. D. A. Ryndyk, N. V. Demarina, J. Keller, and E. Schomburg, Superlattice with hot electron injection: An approach to a bloch oscillator, Phys. Rev. B 67, 033305 (2003). 6. H. Willenberg, G. H. Dohler, and J. Faist, Intersubband gain in a bloch oscillator and quantum cascade laser, Phys. Rev. B 67, 085315 (2003).

40

A. LISAUSKAS ET AL.

7. M. Rochat, L. Ajili, H. Willenberg, J. Faist, H. Beere, G. Davies, E. Linfield, and D. Ritchie, Low-threshold terahertz quantum-cascade lasers, Appl. Phys. Lett. 81, 1381–1383 (2002). 8. B. S. Williams, H. Callebaut, S. K. Q. Hu, and J. L. Reno, 3.4-thz quantum cascade laser based on longitudinal-optical-phonon scattering for depopulation, Appl. Phys. Lett. 82, 1015–1017 (2003). 9. R. Sachs and H. G. Roskos, Mode calculations for a terahertz quantum cascade laser, Opt. Express 12, 2062–2069 (2004). 10. J. Kastrup, R. Hey, K. H. Ploog, H. T. Grahn, L. L. Bonilla, M. Kindelan, M. Moscoso, and A. Wacker, J. Galan, Electrically tunable ghz oscillations in doped gaas-alas superlattices, Phys. Rev. B 55, 2476–2488 (1997). 11. E. Schomburg, S. Brandl, K. Hofbeck, T. Blomeier, J. Grenzer, A. A. Ignatov, K. F. Renk, D. G. Pavel’ev, Y. Koschurinov, V. Ustinov, A. Zhukov, A. Kovsch, S. Ivanov, and P. S. Kop’ev, Generation of millimeter waves with a gaas/alas superlattice oscillator, Appl. Phys. Lett. 72, 1498–1500 (1998). 12. E. Schomburg, R. Scheuerer, S. Brandl, K. F. Renk, D. G. Pavel’ev, Y. Koschurinov, V. Ustinov, A. Zhukov, A. Kovsh, and P. Kop’ev, Ingaas/inalas superlattice oscillator at 147 ghz, Electron. Lett. 35, 1491–1492 (1999). 13. K. F. Renk, B. I. Stahl, A. Rogl, T. Janzen, D. G. Pavelev, Y. I. Koshurinov, V. Ustinov, and A. Zhukov, Subterahertz superlattice parametric oscillator, Phys. Rev. Lett. 95, 126801 (2005). 14. N. V. Demarina, to be published. 15. Y. Shimada, K. Hirakawa, M. Odnoblioudov, and K. A. Chao, Terahertz conductivity and possible bloch gain in semiconductor superlattices, Phys. Rev. Lett. 90, 046806 (2003). 16. N. Sekine and K. Hirakawa, Dispersive terahertz gain of a nonclassical oscillator: Bloch oscillation in semiconductor superlattices, Phys. Rev. Lett. 94, 057408 (2005). 17. A. Lisauskas, N. V. Demarina, E. Mohler, and H. G. Roskos, Comment on dispersive terahertz gain of a nonclassical oscillator: Bloch oscillation in semiconductor superlattices, condmat/0605651. 18. A. Lisauskas, N. V. Demarina, E. Mohler, and H. G. Roskos, Bloch gain and single-pulse terahertz emission experiments, to be published. 19. P. G. Savvidis, B. Kolasa, G. Lee, and S. J. Allen, Resonant crossover of terahertz loss to the gain of a bloch oscillating inas/alsb superlattice, Phys. Rev. Lett. 92, 196802 (2004). 20. A. Lisauskas, N. V. Demarina, C. Bl¨oser, R. Sachs, H. G. Roskos, A. Juozapavicius, ¨ G. Valusis, and K. Kohler, Time-resolved photocurrent spectroscopy of optically excited superlattices and the prospects for bloch gain, Proceedings SPIE International Society for Optical Engineering 6118, 311 (2006). 21. A. A. Ignatov, and Y. A. Romanov, Nonlinear electromagnetic properties of semiconductors with a superlattice, Physica Status Solidi (b) 73, 327–333 (1976). 22. N. V. Demarina, and K. F. Renk, Bloch gain for terahertz radiation in semiconductor superlattices of different miniband widths mediated by acoustic and optical phonons, Phys. Rev. B 71, 035341 (2005). 23. B. S. Williams, S. Kumar, H. Callebaut, and Q. Hu, J. L. Reno, Terahertz quantum-cascade laser operating up to 137 k, Appl. Phys. Lett. 83, 5142–5144 (2003). 24. H. Callebaut, and Q. Hu, Importance of coherence for electron transport in terahertz quantum cascade lasers, J. Appl. Phys. 98, 104505 (2005).

TAILORING THE EMISSION OF TERAHERTZ QUANTUM CASCADE LASERS

RICHARD GREEN, LUKAS MAHLER, COSIMO MAURO, TONIA LOSCO, JI-HUA XU, ALESSANDRO TREDICUCCI,* AND FABIO BELTRAM NEST CNR-INFM and Scuola Normale Superiore, Piazza dei Cavalieri 7, I-56126 Pisa, Italy HARVEY BEERE AND DAVID RITCHIE, Cavendish Laboratory, University of Cambridge, Madingley Rd, Cambridge CB3 0HE, UK

Abstract. For THz quantum cascade lasers to prove useful for applications, certain requirements for their spectral performance will have to be met. Here, we focus on the provision of single mode operation. Distributed feedback devices lasing on a single longitudinal mode are reported using both first and second order gratings. We also report the operation of terahertz master oscillator power amplifier structures with the potential to increase the output power which is available in a single mode.

Keywords: terahertz, quantum cascade laser, intersubband

1. Introduction Quantum cascade lasers (QCLs) operating at terahertz frequencies were first demonstrated in 2002.1 Since then, there been great improvements in their performance, with operation in pulsed mode up to 164-K,2 high-output powers,3 and an operating wavelength range extending from 1.9 THz4 to 4.8 THz5 without the use of an external magnetic field.

______ * To whom correspondence should be addressed. Alessandro Tredicucci, NEST CNR-INFM and Scuola Normale Superiore, Piazza dei Cavalieri 7, I-56126 Pisa, Italy. e-mail: [email protected]

41 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 41–54. © 2007 Springer.

42

R. GREEN ET AL.

QCLs differ from conventional semiconductor lasers in several ways. In a conventional semiconductor laser, the emitted light results from the recombination of an electron within the conduction band of the device with a hole in the valence band. Thus, the wavelength of the emitted light is determined by the bandgap of the semiconductor material, together with a relatively small additional energy originating from the quantum confinement present in quantum well and quantum dot lasers. In contrast, in a QCL, the emitted light arises from radiative transitions of electrons between subbands in a system of coupled quantum wells. Hence, the emission energy can be controlled over a wide range by careful choice of the well and barrier thicknesses. Because of the intersubband nature of the transition, an electron will remain in the conduction band even after undergoing a radiative transition. By cascading many identical active regions, it is possible to achieve internal quantum efficiencies of more than one, with each electron responsible for the emission of more than one photon into the lasing mode. The various potential applications for coherent THz light each have different performance requirements. For the use of QCLs in gas spectroscopy, emission on a single longitudinal mode greatly simplifies the analysis of the resulting data.6 In addition, since the absorption lines being studied tend to be narrow, at least at low pressures, some small amount of tunability can ensure sufficient spectral overlap between the laser and the absorption line. High-emission powers are advantageous, since the signal to noise ratio achievable increases linearly with power for experiments involving direct transmission, and as the square of the emission power for photoacoustic spectroscopy. The use of QCLs as local oscillators for astronomical experiments presents very similar performance requirements spectrally, although less stringent in terms of output power. In contrast, for potential applications in free space optical communications the detailed mode structure of the emission line is less critical, as long as the emission line is located in a spectral region with low atmospheric absorption. More important here is the beam profile which must allow convenient coupling of the beam into an optical system and the ability to modulate the output at high rates. To provide direct modulation of the devices, low threshold currents are desirable, together with small device dimensions and convenient packaging to reduce the parasitic capacitance. There are of course some aspects of performance which are advantageous for all applications. Higher powers tend to increase signal to noise ratios, and simplify detection schemes and optical alignment, while increased operating

TAILORING THE EMISSION OF THz QCLs

43

temperatures will make the devices simpler to use while reducing the running costs and complication of the setup. In this chapter, we focus first on the spectral purity of THz QCLs. After a discussion of the different techniques by which single mode operation can be obtained, first- and second-order distributed feedback QCLs will be presented, based on novel, surface plasmon gratings. Preliminary results from master oscillator power amplifier structures based on THz QCLs will then be reported, followed by a discussion of possible methods for tuning the emission line of these devices. 2. Single Mode Emission from THz QCLs THz QCLs are usually processed into ridge waveguide devices, with end mirrors formed by cleaved facets, to form a Fabry-Perot (FP) cavity. Cavity lengths are typically in the range of about 1–4 mm, meaning that many modes of the cavity exist within the material gain peak. The consequences of this can be seen in Figure 1, which shows the evolution of the lasing spectrum of a typical FP device as the injection current is increased. The bottom panel of the figure shows the spectrum measured at 450-mA injection current, approximately 10% above threshold. The laser output is concentrated completely within a single longitudinal mode. At 500-mA current, while the spectrum is still largely single mode, a small second mode is emerging on the long wavelength side, with an intensity about 5% that of the main peak. By 550 mA the emission spectrum is strongly multimode, and this situation continues until the maximum output power of the device is reached, at about 750 mA. We see that single-mode operation can be achieved from a FP device, close to threshold. However, the output power available into this single mode is severely limited by the appearance of additional longitudinal modes at increased current levels. To provide true single mode operation over a wider range of conditions a different means of optical feedback needs to be provided. Ways of doing this include the use of coupled cavity lasers,7 an external cavity,8 and the fabrication of frequency selective mirrors,9 but the simplest and most common is the use of a distributed feedback (DFB) grating. This consists of a periodic perturbation introduced into the waveguide, designed to have a period equal to half the lasing wavelength within the material.

44

R. GREEN ET AL.

Figure 1. Emission spectrum of a typical FP THz QCL, measured at 7K, for a range of injection currents in pulsed mode. The cavity length is 1.44 mm, with a ridge width of 300 µm. Measured threshold current was 405 mA, corresponding to a threshold current density of 95 A cm−2.

2.1. DFB DEVICES BASED ON ETCHED GRATINGS

Before studying the different methods of introducing a periodic perturbation into the laser waveguide, it is worth considering the structure of the waveguide. Two types of waveguides are used for THz QCLs; the surface plasmon (SP) waveguide,1 which is the one considered here, and the double metal waveguide.10 In the SP case, the waveguide mode consists of a combination of the SP mode arising from the top metallic contact layer evaporated onto the top of the laser ridge, and the SP from a buried heavily doped layer positioned below the stack of active regions. There is also a top contact layer, consisting of about 200 nm of heavily doped GaAs just below the top metallization; this facilitates the provision of electrical contacts to the structure.

TAILORING THE EMISSION OF THz QCLs

45

In Reference 11, a grating was etched into this top contact layer before the evaporation of the Cr/Au top metal. Figure 2 shows the calculated optical mode profiles in three situations; with the complete 200-nm top contact layer, with the thickness reduced to 50 nm and with the layer completely removed. All three cases are considered with Cr/Au top metallization. The most prominent difference between the three curves is in the amount of overlap between the optical mode and the active region. This results in a modulation of the modal gain of gth∆Γ = 1.3 cm−1 for a typical waveguide structure emitting at 68 µm (4.4 THz). However it should also be noted that free carrier absorption within the contact layer is responsible for some of the optical losses within the waveguide, and so there is a loss modulation of ∆αw = 1.5 cm−1. These two mechanisms tend to cancel out as parts of the waveguide with the top contact layer removed experience lower losses, but also a reduced modal gain. Overall, this means that the imaginary part of the coupling constant, κ, for a DFB grating fabricated along these lines is very small, and κ is predominantly index coupled, arising from the variation in the real part of the refractive index of the two sections. It is estimated that κ ~2.5 cm−1.

Figure 2. (a) Calculated waveguide mode profile in three situations: with the top contact layer left intact (solid line), etched so that only 50 nm remains (dashed line), and completely removed (dotted line). (b) Solid lines show spectra measured close to the maximum optical power from etched grating DFB devices, 2.8-mm long and with three different grating periods. The dotted lines show the spectra measured from devices where an additional loss modulation had been introduced using an additional annealed contact step. The 9.2-µm grating period device was 3.2-mm long, and the 9.4-µm device 4.4 mm. All laser ridges were 150-µm wide.

46

R. GREEN ET AL.

Single mode operation was not observed from these devices. The dotted lines in Figure 2(b) show laser spectra measured from three devices fabricated with this type of DFB, but using slightly different grating periods. For the two longer grating periods, the measured spectra appear broadly similar to a FP laser spectrum, but with the presence of a stopband, corresponding to the suppression of two FP modes. This is typical of an undercoupled DFB device where the product of the coupling constant and the cavity length, κ·L, is much less than one, while there is also strong facet reflectivity. The position of this stopband tunes with grating period at a rate in good agreement with that expected based on the effective refractive index calculated for the device. The absence of a stopband in the spectrum measured from the device with the shortest grating period is attributed to an excessive detuning between the material gain peak and the grating resonance. The solid lines in Figure 2(b) show laser spectra from similar devices, fabricated using a slightly different approach. A grating was etched into the top surface of the ridge as before, but a AuGe/Au annealed contact was deposited onto the unetched parts of the grating. This type of contact results in an increase in the optical losses in these regions, and a complex coupled grating is obtained, where both real and imaginary parts of the coupling constant are significant. Single mode operation was observed from these devices, with the laser mode within the previously observed stop gap. 2.2. DFB DEVICES BASED ON SURFACE PLASMON GRATINGS

2.2.1. First-order gratings We have already noted that the optical mode in the waveguide arises from the SP associated with the Cr/Au top metallization; in regions where this has been removed no waveguide mode is present. To understand the operation of these SP DFB gratings, let us consider the situation of a normal ridge waveguide, containing one small region where a slit has been opened in the top metal. For a photon incident on the slit three processes can take place: it can be reflected, it can tunnel through this region without a guided mode, or it can be scattered up through the slit and lost to the laser mode. By fabricating a periodic array of these slits a DFB grating can be produced.12,13 The above picture is not quite accurate, however. At these long wavelengths, there is also a SP mode associated with the layer of heavily doped GaAs used to facilitate electrical contacts to the device, and even in the regions where the metal has been removed there will still be a waveguide

TAILORING THE EMISSION OF THz QCLs

47

mode due to this. True SP gratings can instead be easily obtained by etching away this heavily doped layer in the regions where the metal has been removed. The optimum slit width for the grating is the result of a trade off between the stronger coupling that arises from a wider slit and the increased scattering losses that also occur for wide slits. This is illustrated in Figure 3(b), which shows the calculated grating reflectivity for three different slit widths. The optimum width is found to be 2 µm, which corresponds to a duty cycle of ~10%.

Figure 3. Calculated reflectivity of a first-order surface plasmon grating for three different slit widths.

Figure 4(a) shows emission spectra from three SP DFB devices, emitting around 2.55 THz. Two of these devices (with grating periods of 16.6 and 16.4 µm) showed single mode operation, The multimode operation observed from the 16.2-µm grating period device is attributed to an excessive detuning between the grating mode and the material gain peak. The observed rate of tuning of the emission line with grating period is in good agreement with the expected value using the calculated modal refractive index. Figure 4(b) shows a similar spectrum, measured from the 16.4-µm grating period device, displayed on a logarithmic scale to show the high spectral purity obtained, with a side mode suppression ration (SMSR) >20 dB. For the majority of injection currents measured, single mode behaviour is observed, with a good side mode suppression ration of up to 20 dB. However, for very high injection currents, close to the roll-off in the LI curve, a second mode becomes visible, as shown in Figure 4(c). This is attributed to the presence of a higher-order transverse mode, which can be

48

R. GREEN ET AL.

prevented by reducing the width of the top metallization relative to the ridge. Devices fabricated in this way showed single mode emission for all investigated currents and temperatures.

Figure 4. (a) Spectra measured from first-order surface plasmon DFB lasers with three different grating periods. (b) Spectrum measured from a surface plasmon DFB structure with grating period 16.4 µm, shown on a log scale. (c) Spectrum measured from a 16.5-µm DFB structure, close to the maximum optical power. The smaller peak is attributed to the presence of a higher order transverse mode.

2.2.2. Second-order gratings Because of the intersubband nature of QCLs, the radiative transition will not couple to an optical field propagating in the growth direction, and so surface-emitting devices such as the VCSEL cannot be fabricated.14 However, it is possible for light emission from the surface of a device to be based on a scattering mechanism, either in a photonic crystal structure,15 or using a second order DFB grating.16,17 Second-order DFB structures have been produced based on the SP grating concept. The reflectivity of second-order gratings with different slit widths was simulated using a finite element technique. The results are reported in Figure 5(a), showing the optimum slit width to be the same as in the first-order case. Because the period of a second-order grating is twice that of a first order, this now corresponds to a duty cycle of 5%. The primary reason for the reduced reflectivity of the second-order gratings

TAILORING THE EMISSION OF THz QCLs

49

compared to the first order is the lower number of grating periods per unit length. Additionally, there will be higher scattering losses, due to the intentional vertical emission.

Figure 5. (a) Reflectivity of second-order surface plasmon gratings with different slit widths, compared with that of a first-order grating with 2-µm slits. (b) Emission spectra measured from two second-order DFB devices, with different grating periods. The measured light was emitted in the vertical direction. (c) Emission spectrum of a second-order DFB device, with 36.2-µm grating period, shown on a log scale. (d) Vertically emitted power from a second-order DFB QCL measured as a function of angle in the directions parallel to (dotted line) and transverse to (solid line) the laser ridge.

Vertical emission spectra measured from two second-order DFB lasers with different grating periods are shown in Figure 5(b). For these devices, the grating did not extend over the full length of the ridge; a 1.5-mm long grating section was fabricated in the centre of a 2-mm long laser ridge. To avoid any possibility of the bond wires affecting the grating properties, these were attached at either end of the ridge, in the regions of unpatterned top metallization. The tuning of the emission wavelength with grating period is again in good agreement with the expected value. This, together with the presence of vertical emission confirms that the grating is the dominant source of optical feedback in the devices.

50

R. GREEN ET AL.

For all devices measured, lasing was observed on a single longitudinal mode. However, for some devices a second transverse mode was observed under higher injection currents. This is not surprising for these large ridge widths (250–300 µm), since the waveguide losses for the first and second order transverse modes will be almost equal. Figure 5(c) shows the emission measured from a device emitting on a single transverse mode plotted on a log scale. Again, a very high spectral purity is observed, with a SMSR better than 20 dB. The vertically emitted power was measured as a function of the angle, both along and perpendicular to the direction of the waveguide, as shown in Figure 5(c). As expected the angular divergence in the direction along the ridge is considerably reduced, owing to the increased size of the emitting aperture. Second-order DFB lasers, and particular TM-polarized devices using metallic gratings tend to lase on an asymmetric longitudinal mode, resulting in a double-lobed far-field pattern. There are two possible reasons why this is not observed here. The first is due to insufficient angular resolution of the setup; a simple calculation based on reference 18 suggests that the separation between the two lobes should be less than 1°. It is also possible that the more complex structure of this sample, with facet reflections of arbitrary phase and the ridge sections without grating may generate phase shifts resulting in a more symmetric lasing mode. 2.3. MASTER OSCILLATOR POWER AMPLIFIERS

So far, we have seen that, using distributed optical feedback, it is possible to produce THz QCLs which show good single mode operation. However, the amount of power which it is possible to emit in a single mode is limited by the appearance of higher-order transverse modes at high currents. Although it is possible to avoid these by using narrow ridges to selectively increase the losses in the higher transverse modes, this has the effect of reducing the available output power from the device, and so is not an ideal solution. For lasers emitting at much shorter wavelengths, it has been possible to circumvent this problem by using a master oscillator power amplifier (MOPA) structure. This is a two-section device, with a narrow laser ridge acting as the master oscillator, emitting laser light in a single mode. This emitted light is then amplified in a flared second section of the device. Feedback is usually suppressed from the amplifier section, to ensure that it only has the effect of amplifying the light, without having any influence on its spectral properties. Thus, in an ideal MOPA, the threshold current of the laser section will be independent of the current level in the amplifier section.

TAILORING THE EMISSION OF THz QCLs

51

MOPA structures have been fabricated based on THz QCLs. A schematic diagram of the device is shown in the lower part of Figure 6. The two sections are separated by a 20-µm wide slot. The structure was defined using a single etch step, using inductively coupled plasma dry etching based on a Cl2/Ar/HBr chemistry, and a Cr etch mask. The upper part of Figure 6 shows an SEM image of the slot between the two sections of the device. After this processing step, narrow AuGe/Au annealed contacts were deposited at the edge of the device, followed by a Cr/Au top contact layer. Common side contacts were used for the two sections, resulting in a three terminal device. The sample was then thinned, cleaved, and indium soldered onto copper submounts for testing.

Figure 6. Scanning electron micrograph of part of the THz MOPA structure. The laser section (top) is separated from the amplifier section by a 20-µm slot. The lower part of the figure shows a schematic diagram of the structure, not to scale.

Figure 7 shows experimental data measured from a device with a 2.5mm long laser section, and a 780-µm amplifier section. In Figure 7(a) the threshold current of the laser section is shown as a function of the amplifier current. Initially, with cleaved facets on both ends of the device, a very strong dependence was seen. This was attributed to light reflected from the amplifier facet being coupled back into the laser section. With no current passing through the amplifier, this section will have a net absorption, and so the amount of light coupled back after passing through the amplifier will be small. However, as the amplifier current is increased towards transparency, this increases, effectively reducing the mirror losses of the laser section and so reducing the threshold current. Although the amplifier section did not lase on its own, it is likely that, for high current levels through both sections, the device was in fact acting as a coupled cavity laser. At very high amplifier currents, we are operating beyond the current carrying capability of the miniband within the active region, and the device

52

R. GREEN ET AL.

no longer shows gain. The threshold current of the laser section then returns to the same value as for an unbiased amplifier. In order to eliminate this problem, the amplifier facet was polished at an angle of 9°, as shown in the schematic diagram of Figure 6. The result of this can be seen in Figure 7(a), where the change in threshold with amplifier current is now very small. Light intensity-current curves were measured at a selection of amplifier currents by sweeping the laser current. These are shown in Figure 7(c). The highest amplification achieved was ~2, limited by the short length of the amplifier section. We estimate the maximum achievable amplification to be ~20, for an amplifier length ~3 mm; gain saturation is expected to prevent further improvement by using even longer amplifiers.

Figure 7. (a) Threshold current of the laser section, plotted against amplifier current for two devices with a cleaved amplifier facet (solid line) and an amplifier facet polished at 9° (dotted line). The threshold currents have been normalized to the value obtained with an unbiased amplifier section. (b) Power output measured from the MOPA structure at different amplifier currents, while keeping the laser current constant close to the maximum power. (c) Plot of the output light intensity against laser current measured at various different amplifier currents.

TAILORING THE EMISSION OF THz QCLs

53

The maximum output power measured in the course of the LI measurements is plotted against the amplifier current in Figure 7(b). We can see that this increases up to a current density of ~200 Α cm−2. This current density corresponds to the roll-off in the material gain, as can be seen in Figure 7(c). The reduction in power observed for very high amplifier currents is attributed to heating effects. 3. Conclusions and Acknowledgements THz QCLs emitting on a single mode have been demonstrated using both etched gratings and SP gratings, based on arrays of thin slits in the top metallization of the laser ridge. SP DFB lasers have also been fabricated based on second-order gratings, and vertical emission characterized. Finally, THz master oscillator power amplifiers have been fabricated and characterized. It is predicted that for an amplifier length of 3 mm these can show amplifications of up to 20. This work was funded by the European Commission through the IP project Teranova, the PASR project Terasec and the Marie Curie RTN POISE. Partial financial support from Physical Sciences Inc. is also gratefully acknowledged.

References 1. R. Köhler, A. Tredicucci, F. Beltram, H. E. Beere, E. H. Linfield, A. G. Davies, D. A. Ritchie, R. C. Iotti, and F. Rossi, Terahertz semiconductor heterostructure laser, Nature 417, 156 (2002). 2. B. S. Williams, S. Kumar, Q. Hu, and J. L. Reno, Operation of terahertz quantum cascade lasers at 164 K in pulsed mode and at 117 K in continuous-wave mode, Opt. Express 13, 3331 (2005). 3. B. S. Williams, S. Kumar, Q. Hu, and J. L. Reno, High power terahertz quantum cascade lasers, Electron. Lett., 42, 89 (2006). 4. B. S. Williams, S. Kumar, Q. Hu, and J. L. Reno, 1.9 THz quantum-cascade lasers with one-well injector, Appl. Phys. Lett., 88, 121123 (2006). 5. A. Tredicucci, L. Mahler, T. Losco, J. Xu, C. Mauro, R. Köhler, H. E. Beere, D. A. Ritchie, and E. H. Linfield, in Novel In-Plane Semiconductor Lasers IV, C. Mermelstein and D. P. Bour, Eds., Proceedings SPIE 5738, 146–158 (2005). 6. A. A. Kosterev, R. F. Curl, F. K. Tittel, C. Gmachl, F. Capasso, D. L. Sivco, J. N. Baillargeon, A. L. Hutchinson, and A. Y. Cho, Methane concentration and isotopic composition measurements with a mid-infrared quantum-cascade laser, Opt. Lett., 24, 1762 (1999).

54

R. GREEN ET AL.

7. L. Hvozdara, A. Lugstein, S. Gianordoli, W. Schrenk, G. Strasser, K. Unterrainer, E. Bertagnolli, and E. Gornik, Self-aligned coupled cavity GaAs/AlGaAs midinrared quantum –cascade laser, Appl. Phys. Lett., 77, 1077 (2000). 8. G. Totschnig, F. Winter, V. Pustogov, J. Faist, and A. Müller, Mid-infrared externalcavity quantum-cascade laser, Opt. Lett., 27, 1788 (2002). 9. L. A. Dunbar, V. Moreau, R. Ferrini, R. Houdré, L. Sirigu, G. Scalari, M. Giovannini, M. Hoyler, and J. Faist, Design, fabrication and optical characterisation of quantum cascade lasers at terahertz frequencies using photonic crystal reflectors, Opt. Express 13, 8960 (2005). 10. B. S. Williams, S. Kumar, H. Callebaut, Q. Hu, and J. L. Reno, Terahertz quantumcascade laser at λ≈100 µm using metal waveguide for mode confinement, Appl. Phys. Lett., 83, 2124 (2003). 11. L. Mahler, R. Köhler, A. Tredicucci, F. Beltram, H. E. Beere, E. H. Linfield, and D. A. Ritchie, Single-mode operation of terahertz quantum cascade lasers with distributed feedback resonators, Appl. Phys. Lett., 84, 5446 (2004). 12. J. C. Weeber, Y. Lacroute, A. Dereux, E. Devaux, T. Ebbesen, C. Girard, M. U. González, and A. L. Baudrion, Near-field characterization of Bragg mirrors engraved in surface plasmon waveguides, Phys. Rev., B 70, 235406 (2004). 13. L. Mahler, A. Tredicucci, R. Köhler, F. Beltram, H. E. Beere, E. H. Linfield, and D. A. Ritchie, High-performance operation of single-mode terahertz quantum cascade lasers with metallic gratings, Appl. Phys. Lett., 87, 181101 (2005). 14. M. Helm, in Intersubband Transitions in Quantum Wells: Physics and Device Applications I, H. C. Liu, F. Capasso, Eds. (Academic Press, San Diego, 2000). 15. R. Colombelli, K. Srinivasan, M. Troccoli, O. Painter, C. F. Gmachl, D. M. Tennant, A. M. Sergent, D. L. Sivco, A. Y. Cho, and F. Capasso, Quantum cascade surface emitting photonic crystal laser, Science 302, 1374 (2003). 16. C. Pflügl, M. Austerer, W. Schrenk, S. Golka, G. Strasser, R. P. Green, L. R. Wilson, J. W. Cockburn, A. B. Krysa, and J. S. Roberts, Single mode surface emitting quantum cascade lasers, Appl. Phys. Lett., 86, 211102 (2005). 17. O. Demichel, L. Mahler, T. Losco, C. Mauro, R. Green, J. Xu, A. Tredicucci, F. Beltram, H. E. Beere, D. A. Ritchie, and V. Tamošiūnas, Opt. Express, 14, 5335 (2006). 18. N. Finger, W. Schrenk, and E. Gornik, Analysis of TM-polarized DFB laser structures with metal surface gratings, IEEE J. Quantum Electron., 36, 780 (2000).

GUIDED PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES KANGLIN WANG AND DANIEL M. MITTLEMAN* Rice University, Department of Electrical and Computer Engineering, MS 366, Houston, TX 77251-1892, USA

Abstract. We demonstrate a new waveguiding structure for terahertz (THz) radiation, in which broadband THz pulses are confined and guided along a bare metal wire. This waveguide exhibits close to the lowest attenuation of any waveguide for broadband THz pulses reported so far. It also supports propagation of broadband radiation with negligible group-velocity dispersion, making it especially suitable for use in pulsed terahertz sensing and diagnostic systems. In addition, the structural simplicity lends itself naturally to the facile manipulation of the guided pulses, including coupling, directing, and beam splitting. These results can be described in terms of a model developed by Sommerfeld, for waves propagating along the surface of a cylindrical conductor.

Keywords: waveguide, surface plasmon polariton, time-domain spectroscopy

1. Introduction Rapid advances in laser technology have enabled various techniques for the generation and detection of electromagnetic radiation in the terahertz region (spanning from ~100 GHz to ~10 THz).1 As a result, numerous uses of terahertz radiation have been explored, including trace gas detection,2 medical diagnosis,3,4 security screening,5 and defect analysis in complex materials such as space shuttle tiles.6 Many of these studies have relied on terahertz time-domain spectroscopy, a technique for generating sub-picosecond pulses with spectral content spanning much of the THz band.7–10

______

* To whom correspondence should be addressed: Daniel M Mittleman, Rice University, Department of Electrical and Computer Engineering MS366, Houston, TX77251-1892, USA

55 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 55–68. © 2007 Springer.

56

K. WANG AND D.M. MITTLEMAN

However, progress has been limited by the overwhelming reliance on free-space transport of the terahertz beam, using bulk optical components. In many real-world situations, the sample or region to be studied may not be readily accessible to a line-of-sight beam. Common devices such as optical fiber-based sensors or medical endoscopes rely on the guided wave delivery of light to the remote-sensing location. In order to extend this paradigm to THz applications, the development of optimized guided wave devices is required. Furthermore, the development of practical THz waveguides will dramatically expand the application of THz-TDS in areas such as gas sensing and nanometer thin-film measurements.11,12 The development of THz waveguides has been hindered by the material properties and the application requirements in this spectral range. On the one hand, the characteristics of materials at THz frequencies make it extremely difficult to build a fiber to guide THz beams over a long distance. The most transparent materials for this range are crystalline (e.g. high-resistivity silicon), and thus are costly, fragile, and challenging to form into specific geometries for waveguide configurations. Other materials, such as low-loss polymers or glasses, are more malleable but exhibit prohibitively high-absorption losses for propagation distances of more than a few centimeters. For this reason, THz waveguides generally must rely on propagation in air, rather than via dielectric confinement as in an optical fiber. On the other hand, many THz applications rely on the use of broadband pulses for time-domain analysis and spectroscopic applications. To avoid pulse reshaping during propagation, the significant additional constraint of low dispersion is also required. But for many conventional metal waveguides (e.g. metal tubes), pulse reshaping in propagation is difficult to avoid, due to the extreme dispersion near the waveguide cutoff frequencies. Furthermore, finite conductivity of metals can lead to considerable losses in the wave propagation. Great efforts have been devoted to finding useful THz waveguides within the past few years, and various guides with quasi-optical coupling have been demonstrated. Most of these THz waveguides have been based on conventional guiding structures, such as metal tubes,11,13,14 plastic ribbons,15 or dielectric fibers.16 There have also been reports on the application of the latest technology of photonic crystal fibers to THz radiation.17,18 In all of these cases, the utility for transport of THz pulses is limited by group velocity dispersion of the guided waves. The most promising studies have reported dispersionless propagation in parallel metal plate waveguides.19–21 In this case the reported attenuation (~80 dB/m) is limited by the finite conductivity of the metal components.

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

57

In this paper we show how a metal waveguide with very simple geometry, namely a bare wire, can be used to guide broadband THz pulses with outstanding performance, including low-loss and negligible group velocity dispersion. The guided propagation of THz pulses on a metal wire shows similar behavior to the cylindrical surface guided waves first described by Sommerfeld.22,23 The structural simplicity of the wire waveguide presents great advantages in the manipulation of guided THz radiation.24,25 2. Characterization of the THz Wire Waveguide The propagation of THz radiation along bare metal wires was first observed in the demonstration of apertureless near-field scanning optical microscopy (NSOM) using THz-TDS.26,27 In order to directly observe and characterize the guided THz propagation on metal wires, we use a different configuration, employing fiber-coupled terahertz photoconductive antennas.28 This permits us to change the incident position (the start point of the propagation) and the detection position of the THz pulses, to observe the propagation distance dependance and the spatial profile of the guided mode.24 Scanning Optical Delay Line

Femtosecond Laser

Fiber Coupler

THz transmitter

x Input Coupler

Metal wire waveguide

y

z

Fiber Coupler

THz receiver Guided wave

Movable Stage

Movable Stage

Figure 1. Experimental setup for the direct characterization of the THz wire waveguide.

A schematic of the experimental setup is shown in Figure 1. The horizontally polarized THz pulses are focused onto the waveguide, a stainless steel wire. Another stainless steel wire is placed at the focal spot, oriented perpendicular to the waveguide (the y direction in Figure 1). This second wire serves as an input coupler. Scattering of the input THz radiation at the

58

K. WANG AND D.M. MITTLEMAN

intersection structure helps to excite the radially polarized mode which can propagate along the waveguide. Both the waveguide and the coupler are 0.9 mm in diameter, and the separation between them is 0.5 mm. The receiver is placed at the end of the waveguide and is oriented to detect only the vertically polarized component of the electric field in order to eliminate the possibility of detecting directly scattered radiation which would interfere with the detection of the guided mode. The incident THz beam is modulated by a chopper in front of the transmitter and a lock-in amplifier is used for detecting the induced photocurrent in the receiver. The THz transmitter, the focusing lenses, and the coupler are all mounted on a movable stage so that the incident position along the waveguide can be controlled. The THz receiver is mounted on a three-axis stage for detection at various positions with respect to the end of the waveguide. 2.1. SPATIAL PROFILE

Electric field (arb. units)

As the first step to characterize the propagating mode on the wire waveguide, we measure the spatial profile of the electric field around the waveguide by vertically scanning the THz receiver at the end of the waveguide. Figure 2 shows typical time-domain electric field waveforms, for two different receiver

0

5

10

15

20

25

30

Delay (ps) Figure 2. Time-domain electric field waveforms detected with the receiver 3 mm above (top) and 3 mm below (bottom) the waveguide. The polarity reversal shows the radial nature of the guided mode.

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

59

positions located 3 mm above and 3 mm below the wire waveguide. These waves are vertically (y) polarized, perpendicular to the horizontally (x) polarized input beam. The polarity reversal as the detector scans across the wire clearly shows the radial nature of the guided mode. The peak-to-peak amplitude of the waveform decays with increasing distance from the wire surface, approximately as 1/r where r is the radial distance from the axis of the wire.25 The observed behavior can be understood using Sommerfeld’s description of an electromagnetic wave propagating along the surface of a cylindrical conductor, a so-called Sommerfeld wire wave. In this case, it has been shown that the important propagating solution is an axially symmetric TM wave. Outside the metal, the variation of the radial electric field component (the dominant component) is described by a Hankel function, H1(1)(γr), where γ is defined in terms of the propagation constant k of the field outside the wire according to γ 2 = ω2 c 2 − k 2. For a perfectly conducting wire, γ = 0 and the field propagates with a velocity determined solely by the external medium (in our case, air).22 For large but finite conductivity, γ is small and the approximate form for the Hankel function can be used, appropriate for small argument:

H1( ) ( x ) ≈ −2i π x . 1

(1)

Thus, a Sommerfeld wire wave also exhibits 1/r decay, within a distance r0 << 1 γ of the wire surface. The Sommerfeld description can be used to estimate the distance that the wave extends from the metal surface, for a metal of finite conductivity. To do so, one must determine γ by solving the transcendental equation which results from the boundary conditions at the wire surface.23 For a stainless steel wire with a conductivity of 1.39 × 106 mho/m, we compute that 1 γ ≈ 7.8 mm . Thus, at a frequency of 0.3 THz, half of the power is transmitted through an area extending roughly 1.6 mm from the axis of the wire. This value is in good agreement with the measured behavior. 2.2. PROPAGATION CHARACTERISTICS

The propagation characteristics of the guided mode are studied by moving the incident position of the THz beam along the waveguide. In this way, we can obtain the time-domain waveforms as a function of propagation distance. There is no evident change in the temporal shape of the waveforms for propagation up to 24 cm, the limit of our optical delay line in these measurements. This shows that the propagation is largely dispersionless. As in the

60

K. WANG AND D.M. MITTLEMAN

NSOM experiment, we determine the broadband group velocity of the propagation mode by analyzing the dependence of the relative time delay of the waveforms on the propagation distance. A least-squares linear fit to these data yields the group velocity vg = (2.995 ± 0.001) × 108 m/s. This is to be expected, given that the Sommerfeld surface wave model predicts a group velocity deviating from c by less than one part in 104, for our experimental situation.

24 20 16 12 8 6 2

3 4

-10 -5 0 5 10 Receiver Displacement (mm)

Amplitude (arb. units)

Amplitude (arb. units)

Propagation distance (cm):

3 2 1 0 0 10 20 30 Propagation Distance (cm)

Figure 3. (left) The amplitude of the THz pluses as a function of the vertical displacement of the receiver, measured at different propagation distances. (right) The maximum peak-topeak amplitude of the THz pulses detected at each propagation distance (solid squares). Measurements with the receiver moved away from the end of the waveguide are also made (hollow squares), showing a sharp drop of the pulse amplitude as the radiation propagates off of the end of the waveguide into air.

In order to study the evolution of the guided mode as it propagates, we compare the spatial profile of the guided mode detected at different propagation distances. These are each obtained by moving the detector across the end of the waveguide. The results are shown in Figure 3. It is immediately clear that the electric field is more closely confined to the surface of the wire for the shortest propagation distances. Subsequently, the guided mode spreads laterally, especially during the first several centimeters of propagation. For each propagation distance, we extract the waveform with the maximum peak-to-peak amplitude. Except for the few shortest propagation distances, these are obtained at a fixed receiver offset of roughly 3 mm. These amplitudes are plotted as a function of propagation distance in Figure 3 (right). The amplitude attenuation coefficient α of the wire waveguide can be extracted from these data, simply by fitting the dependence of the pulse amplitude E on the propagation distance x to:

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

61

E ( x ) = E 0 e −α x

(2)

The value we obtain, α = 0.03 cm−1, is among the lowest of any waveguide for broadband THz pulses reported to date. The open squares in Figure 3 show the rapid decrease in the amplitude of the wave after it propagates off of the end of the waveguide. We note that our measurements do not reflect the losses associated with the coupling of the linearly polarized free-space THz beam to the guided mode. In the experiment described here, less than 1% of the power is coupled to the radially polarized waveguide mode from the free-space incident beam. More effective mode-matching is needed to improve the input coupling.29 This will be discussed further in Section 2.3. 2.3. A NEW MECHANISM FOR GROUP VELOCITY DISPERSION

The electromagnetic mode propagating on a metal wire is essentially equivalent to a surface plasmon polariton (SPP), albeit with a frequency far below that of the bulk plasmon frequency of the metal. Such interface excitations are well known in the case of flat metal surfaces. For SPPs on a planar metal surface, the dispersion curve asymptotically approaches the light line k = ω c as the frequency decreases, and accordingly, the phase velocity and group velocity gradually increase and approach c, the speed of light in air.30 Similar behavior has also been observed for SPPs propagating on Au and Ag nanowires at visible and near-infrared frequencies31–33 (see Figure 4). Time-domain waveforms of THz SPP pulses after propagating 20 cm on aluminum wires with diameters are shown. The vertical dashed line is shown for reference, to emphasize the unidirectional increase in the travel time with decreasing wire diameter. In order to investigate the details of the dispersion diagram for Sommerfeld waves, we study the propagation on very thin wires. To investigate the effect of surface curvature, we measure the guided propagation of broadband THz pulses on aluminum wires with diameters ranging from 2.4 mm down to 18 µm. The experimental setup for the THz-TDS measurement is similar to that shown in Figure 1 above. The metal wire is supported and stretched by a tightly fitting Teflon slab close to the distal end. The broadband single-cycle pulses of free space THz radiation are coupled to surface waves on the metal wire by scattering at a small (~400 µm) gap defined by the surface of the wire and a copper blade oriented perpendicular to the

K. WANG AND D.M. MITTLEMAN

62

0.5

Amplitude (arb. units)

Wire diameter: 2388 µ m

0.0

813 µ m -0.5

51 µ m 18 µ m -1.0 770

780

790

800

Delay (ps) Figure 4 . Time-domain waveforms of THz SPP pulses after propagating 20 cm on aluminum wires with diameters as shown. The vertical dashed line is shown for reference, to emphasize the unidirectional increase in the travel time with decreasing wire diameter.

wire.34 After a propagation distance of 20 cm, the electric field of the guided wave is detected by a fiber-coupled photoconductive antenna located 3 mm off the axis of the metal wire. We compare a series of measurements using wires of different diameters, while all other components stay fixed. The time-domain electric field waveforms detected on different Al wires are shown in Figure 4. For clarity, only four typical waveforms are shown here. The detected THz pulses maintain the single-cycle shape, showing that the SPP propagation is largely nondispersive. That is, the dispersion relation is largely linear within the bandwidth of the detected radiation (from 30 GHz up to about 500 GHz). However, we observe an increase in the transit time as the diameter of the wire decreases. Increased pulse reshaping becomes evident when the diameter is below 200 µm. This indicates that the dispersion relation is no longer linear when the diameter of the metal wire waveguide becomes sufficiently small. Detailed information concerning the propagation of the SPPs can be extracted from the Fourier transforms of the measured time-domain waveforms. From Figure 4, it is clear that the amplitude of the SPP decreases with decreasing wire size. This is due both to a decrease in the input coupling

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

63

efficiency as well as to increasing propagation losses on smaller wires. Of more interest is the spectral phase, from which we obtain the phase velocity vp = L/τp, where L is the propagation distance and τp is the phase time delay. This delay is related to the measured spectral phase by:

τ p (ω ) = τ 0 +

∆φ (ω )

(3)

ω

where τ0 is a reference phase time delay and ∆φ(ω) is the phase difference between the detected waveform and the reference waveform. For large wire diameters (e.g. larger than 1 mm), the SPP dispersion in the THz range approaches that of a flat metal surface. The propagation is essentially dispersionless, and the velocity approaches the speed of light in air. Therefore, we use the THz waveform measured on the largest diameter wire (2.388 mm) as the reference pulse with τ0 = L/c. Here, c is the speed of light in air ( c = c 0 / n air , c0 = 2.9979 × 108 m/s, nair = 1.000268).35 In

1.000

vp / c

0.999

0.998

813 mm 51 mm 18 mm

0.997

0.996

0.995 0.0

0.1

0.2

0.3

0.4

0.5

Frequency (THz) Figure 5. Experimentally determined phase velocity of SPP’s propagating on Al wires of different diameters (with the 2,388-µm wire used as a reference). The calculated phase velocities of these three wires (also referenced to the largest wire) are shown by the thick solid line, the thin solid line and the dashed line respectively. The calculation reproduces the striking result that the velocity decreases with decreasing frequency.

64

K. WANG AND D.M. MITTLEMAN

Figure 5 we plot the experimentally determined phase velocity vp(ω) for three different wire diameters. At all frequencies, the phase velocity decreases as the wire size decreases. This accounts for the increasing delay observed in Figure 4. These data also show that, rather than approaching the speed of light c, the phase velocity deviates increasingly from c as the frequency decreases. This means that the low-frequency components arrive later than the peak of the pulse rather than earlier. This is distinct from the behavior of SPPs on a planar metal surface, which show the opposite trend albeit with a much smaller magnitude.36 To understand the unusual dispersive behavior of SPPs on metal wires in the THz frequency range, we calculate the dispersion relation of SPPs over the whole frequency range from THz to visible light, for both a planar Al surface and for Al wires with various diameters. The dispersion relation of the Sommerfeld mode can be obtained by numerically solving the transcendental equation.22

k Al2

µ Al γ Al

2 J 1 (γ Al a) k air H 1(1) (γ air a) = J 0 (γ Al a) µ air γ air H 0(1) (γ air a)

(4)

Here a is the radius of the wire, µ is the magnetic permeability, k is the propagation constant in a homogeneous medium, γ is defined as γ2 = k2−h2 and h is the propagation constant of the surface mode. J0 and J1 are Bessel functions; H0(1) and H1(1) are Hankel functions. We model the properties of aluminum using the Drude model. To make the calculation valid over a very broad frequency range, we do not use the low-frequency approximation kAl=(ωµAlσAl)1/2exp(−iπ/4) or the high-frequency approximation ε(ω)=1−ωp2/ω2 as in previous work.23,37,38 The only approximation used in our calculation is J1/J0 = −i, which is valid because the radius of the wires used here are large compared to the skin depth ( γ Al a >> 1 ). At visible and infrared frequencies, the dispersive behavior of an SPP on an Al wire of 25-µm diameter is almost indistinguishable from that of a planar Al surface. But the low-frequency portion of the dispersion curve approaches the light line in a different way from that of a planar surface, which leads to the unique dispersive behavior observed in the THz frequency range. For comparison with the experimental data, we calculate the phase velocity of the SPPs on Al wires of several different diameters. The results are shown as solid curves in Figure 5. The calculated phase velocity agrees qualitatively with the observed results, reproducing in particular the notable feature of decreasing vp with decreasing frequency.36 It is clear from our data and from the calculations that the observed dispersive behavior only appears at THz and lower frequencies. This indicates

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

65

that for frequencies much lower than the surface plasmon frequency, the resonant interaction between the electromagnetic wave and the plasma oscillation is no longer the dominant mechanism for determining the properties of surface waves. Instead, the electromagnetic properties of the metal play an increasingly important role due to the larger skin depth at lower frequencies. This effect is enhanced for SPPs propagating on wires, due to the geometry of the metal surface. Since Sommerfeld waves are single-mode azimuthally symmetric TM waves, the electric field components inside the metal at a given point along the length of the wire are in phase. So, due to the curved nature of the surface, these evanescent field components can constructively interfere inside the metal. As a result, more power is transmitted inside the metal for SPPs on metal wires as compared to SPPs on planar metal surfaces. This enhanced skin effect is more significant for smaller wire diameters, since increased surface curvature leads to a larger overlap of the evanescent waves penetrating into the metal. It is also more significant at lower frequencies, due to the larger skin depth. This unusual dispersive behavior indicates the decreased plasmonic nature of SPPs in the THz frequency range. 3. Conclusions We have demonstrated a new type of THz waveguide with low loss, negligible group velocity dispersion and structural simplicity. This waveguide enables many new THz-sensing applications. It is now possible to direct the THz pulse inside of containers or around corners, where line-ofsight optics is not practical. Besides the waveguide described above, we have also tried many other metal wires as THz waveguides. The materials for these guides include steel, aluminum, copper, zinc, and nichrome. The wire diameter of these guides ranges from 0.5 mm to 6.4 mm. There is no strong difference in the performance of these waveguides, showing that THz pulses can be launched along any thin metal rod structures. In situations where the guided mode could be perturbed by other structures close to the waveguide, we could add a section of outer metallic shield to form a coaxial waveguide, as long as the additional ohmic losses can be tolerated. With a Y-splitter structure used to separate the output wave from the input wave, and a small mirror attached at the end of the waveguide as a 90° output director, we have successfully demonstrated a THz endoscope, by detecting THz pulses reflected from the bottom and the side wall inside a container.24,25 Further improvement could be made by combining an endoscope with an imaging system. This can be accomplished by scanning

66

K. WANG AND D.M. MITTLEMAN

the endoscope along the surface of the detected region, or alternatively, scanning or rotating the sample to obtain an internal THz image. One challenge for this goal is the low power transmitted by the endoscope which strongly limits the data acquisition rate as well as the dynamic range. With optimization of the mode of the input beam and the coupling geometry, the power launched into the endoscope probe can be greatly increased.29,39 It is also interesting to note that this waveguide naturally generates a radially polarized mode. So with a focusing lens mounted at the distal end of the endoscope, a higher resolution can be obtained than in the normal THz imaging system, due to the sub-diffraction-limited focusing of radially polarized beams.40,41 Furthermore, since the radially polarized mode is an ideal input field for a coaxial near-field probe42–44 or an apertureless nearfield optical antenna,26 nanometer-resolved endoscopic THz imaging may be possible.45 This will pave the way for a wide range of new applications for terahertz sensing and imaging. Acknowledgments This work has been funded in part by the R. A. Welch Foundation, the National Science Foundation, and by Advanced Micro Devices.

References 1. D. Mittleman, Ed., Sensing with Terahertz Radiation (Springer, Heidelberg, 2002). 2. R. H. Jacobsen, D. M. Mittleman, and M. C. Nuss, Chemical recognition of gases and gas mixtures with terahertz waves, Opt. Lett., 21, 2011–2013 (1996). 3. R. M. Woodward, V. P. Wallace, D. D. Arnone, E. H. Linfield, and M. Pepper, Terahertz pulsed imaging of skin cancer in the time and frequency domain, J. Biol. Phys., 29, 257–261 (2003). 4. D. Crawley, C. Longbottom, V. P. Wallace, B. Cole, D. D. Arnone, and M. Pepper, Three-dimensional terahertz pulse imaging of dental tissue, J. Biomed. Opt., 8, 303–307 (2003). 5. K. Kawase, Y. Ogawa, and Y. Watanabe, Non-destructive terahertz imaging of illicit drugs using spectral fingerprints, Opt. Express 11, 2549–2554 (2003). 6. S. Wang and X.-C. Zhang, Pulsed terahertz tomography, J. Phys. D, 37, R1–R36 (2004). 7. P. R. Smith, D. H. Auston, and M. C. Nuss, Subpicosecond photoconducting dipole antennas, IEEE J. Quant. Elec., 24, 255–260 (1988). 8. M. van Exter and D. Grischkowsky, Characterization of an optoelectronic terahertz beam system, IEEE Trans. Microw. Th. Tech. 38, 1684–1691 (1990).

PROPAGATION OF TERAHERTZ PULSES ON METAL WIRES

67

9. P. U. Jepsen, R. H. Jacobsen, and S. R. Keiding, Generation and detection of terahertz pulses from biased semiconductor antennas, J. Opt. Soc. Am. B, 13, 11, 2424–2436 (1996). 10. D. M. Mittleman, R. H. Jacobsen, and M. C. Nuss, T-ray imaging, IEEE J. Sel. Top. Quant. Elec., 2, 679–692 (1996). 11. G. Gallot, S. P. Jamison, R. W. McGowan, and D. Grischkowsky, Terahertz waveguides, J. Opt. Soc. Am. B, 17, 851–863 (2000). 12. J. Zhang and D. Grischkowsky, Waveguide terahertz time-domain spectroscopy of nanometer water layers, Opt. Lett., 29, 1617–1619 (2004). 13. R. W. McGowan, G. Gallot, and D. Grischkowsky, Propagation of ultrawideband short pulses of THz radiation through submillimeter-diameter circular waveguides, Opt. Lett., 24, 1431–1433 (1999). 14. J. A. Harrington, R. George, P. Pedersen, and E. Mueller, Hollow polycarbonate waveguides with inner Cu coatings for delivery of terahertz radiation, Opt. Express 12, 5263–5268 (2004). 15. R. Mendis and D. Grischkowsky, Plastic ribbon THz waveguides, J. Appl. Phys., 88, 4449–4451 (2000). 16. S. P. Jamison, R. W. McGown, and D. Grischkowsky, Single-mode waveguide propagation and reshaping of sub-ps terahertz pulses in sapphire fiber, Appl. Phys. Lett., 76, 1987–1989 (2000). 17. H. Han, H. Park, M. Cho, and J. Kim, Terahertz pulse propagation in a plastic photonic crystal fiber, Appl. Phys. Lett., 80, 2634–2636 (2002). 18. M. Goto, A. Quema, H. Takahashi, S. Ono, and N. Sarukura, Teflon photonic crystal fiber as terahertz waveguide, Jpn. J. Appl. Phys., 43, L317–L319 (2004). 19. R. Mendis and D. Grischkowsky, Undistorted guided-wave propagation of subpicosecond terahertz pulses, Opt. Lett., 26, 846–848 (2001). 20. R. Mendis and D. Grischkowsky, THz interconnect with low loss and low group velocity dispersion, IEEE Microw. Wireless Comp. Lett., 11, 444–446 (2001). 21. S. Coleman and D. Grischkowsky, A THz transverse electromagnetic mode twodimensional interconnect layer incorporating quasi-optics, Appl. Phys. Lett., 83, 3656– 3658 (2003). 22. J. A. Stratton, Electromagnetic Theory (McGraw-Hill, New York, 1941). 23. G. Goubau, Surface waves and their application to transmission lines, J. Appl. Phys., 21, 1119–1128 (1950). 24. K. Wang and D. M. Mittleman, Metal wires for terahertz wave guiding, Nature 432, 376–379 (2004). 25. K. Wang and D. Mittleman, Guided propagation of terahertz pulses on metal wires, J. Opt. Soc. Am. B, 22, 2001–2008 (2005). 26. K. Wang, A. Barkan, and D. M. Mittleman, Propagation effects in apertureless near-field optical antennas, Appl. Phys. Lett., 84, 305–307 (2004). 27. K. Wang, D. M. Mittleman, N. C. J. van der Valk, and P. C. M. Planken, Antenna effects in terahertz apertureless near-field optical microscopy, Appl. Phys. Lett., 85, 14, 2715–2717 (2004). 28. J. V. Rudd, D. Zimdars, and M. Warmuth, Compact fiber-pigtailed terahertz imaging system, Proc. SPIE 3934, 27–35 (2000). 29. J. Deibel, M. Escarra, and D. M. Mittleman, Photoconductive terahertz antenna with radial symmetry, Elec. Lett., 41, 9–10 (2005).

68

K. WANG AND D.M. MITTLEMAN

30. H. Raether, Plasmons on Smooth and Rough Surfaces and on Gratings (Springer, Berlin, 1988). 31. G. Schider, J. R. Krenn, A. Hohenau, H. Ditlbacher, A. Leitner, F. R. Aussenegg, W. L. Schaich, I. Puscasu, B. Monacelli, and G. Boreman, Plasmon dispersion relation of Au and Ag nanowires, Phys. Rev. B, 68, 155427 (2003). 32. J.-C. Weeber, Y. Lacroute, A. Dereux, E. Devaux, T. Ebbesen, C. Girard, M. U. González, and A.-L. Baudrion, Near-field characterization of Bragg mirrors engraved in surface plasmon waveguides, Phys. Rev. B, 70, 235406 (2004). 33. J. K. Lim, K. Imura, T. Nagahara, S. K. Kim, and H. Okamoto, Imaging and dispersion relations of surface plasmon modes in silver nanorods by near-field spectroscopy, Chem. Phys. Lett., 412, 41 (2005). 34. J. Saxler, J. Gómez Rivas, C. Janke, H. P. M. Pellemans, P. Haring Bolivar, and H. Kurz, Time-domain measurements of surface plasmon polaritons in the terahertz frequency range, Phys. Rev. B, 69, 155427 (2004). 35. D. R. Lide, CRC Handbook of Chemistry and Physics, 85th edn. (CRC Press, Boca Raton, 2004). 36. K. Wang and D. Mittleman, Dispersion of surface plasmon polaritons on metal wires in the terahertz frequency range, Phys. Rev. Lett., 96, 157401 (2006). 37. T.-I. Jeon, J. Zhang, and D. Grischkowsky, THz Sommerfeld wave propagation on a single metal wire, Appl. Phys. Lett., 86, 161904 (2005). 38. M. J. King and J. C. Wiltse, Surface-wave propagation on coated or uncoated metal wires at millimeter wavelengths, IRE Trans. Antennas Propag., 10, 246 (1962). 39. J. Deibel, K. Wang, M. Escarra, and D. Mittleman, Enhanced coupling of terahertz radiation to cylindrical wire waveguides, Opt. Express 14, 279–290 (2006). 40. S. Quabis, R. Dorn, M. Eberler, O. Glöckl, and G. Leuchs, Focusing light to a tighter spot, Opt. Comm., 179, 1–6 (2000). 41. R. Dorn, S. Quabis, and G. Leuchs, Sharper focus for a radially polarized light beam, Phys. Rev. Lett., 91, 233901 (2003). 42. U. C. Fischer and M. Zapletal, The concept of a coaxial tip as a probe for scanning near field optical microscopy and steps towards a realisation, Ultramicrosc., 42, 393–398 (1992). 43. F. Keilmann, FIR microscopy, Infrared Phys. Technol., 36, 217–224 (1995). 44. C. W. McCutchen, Transmission line probes for scanning photon-tunneling microscopy, Scanning: J. Scan. Microscopies 17, 15–17 (1995). 45. H.-T. Chen, R. Kersting, and G. C. Cho, Terahertz imaging with nanometer resolution, Appl. Phys. Lett., 83, 3009–3011 (2003).

SUPERLATTICE AND OTHER NEGATIVE-DIFFERENTIALRESISTANCE DEVICES: CURRENT STATUS

HERIBERT EISELE* Institute of Microwaves and Photonics, University of Leeds, Leeds, LS2 9JT, UK

Abstract. The paper discusses and compares the concepts, performance potential, and most recent experimental results of both classical and novel active two-terminal devices for low-noise RF power generation at submillimeter-wave frequencies up to 1 THz. These devices use transit-time, transferred-electron, and quantum-mechanical effects (or a combination of them) to create a negative differential resistance at the frequency of interest. Examples of state-of-the-art results are output power levels of more than 70 mW at 62 GHz and more than 10 mW at 132 GHz from GaAs/AlGaAs superlattice electronic devices; more than 9 mW at 280 GHz, 3.7 mW at 300 GHz, 1.6 mW at 329 GHz, and more than 40 µW at 422 GHz from InP Gunn devices; and more than 140 µW at 355 GHz from a GaAs tunnel-injection transit-time diode.

Keywords: Gunn devices, millimeter-wave devices, millimeter-wave generation, millimeter-wave oscillators, negative differential resistance, oscillator noise, phase noise, power combining, submillimeter-wave devices, submillimeter-wave generation, submillimeter-wave oscillators, superlattice, transferred-electron effect, transit-time diodes, tunneling.

1. Introduction Sources of coherent continuous-wave (CW) radiation are a key component in many emerging systems applications at submillimeter-wave frequencies. Wide availability, compactness, reliability, and low power consumption are

______

* To whom correspondence should be addressed: Heribert Eisele, School of Electronic and Electrical Engineering, University of Leeds, Leeds, LS2 9JT, UK, e-mail: [email protected]

69 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 69–88. © 2007 Springer.

70

H. EISELE

much sought-after features of such fundamental sources and they call for an all-solid-state solution1. The first microwave oscillator ever with a semiconductor device, a tunnel diode, exploited the principle of combining a passive resonant circuit with an electronic device that exhibits a negative differential resistance (NDR) around the resonant frequency of the circuit2–4 and thus overcomes the losses in the circuit. Since then, this basic principle has been extended successfully well into the frequency range of submillimeter waves. Tremendous progress in the upper frequency limits5–11 and RF output power levels12,13 of three-terminal devices in amplifiers has been demonstrated within the last decade and has resulted in a rapidly expanding use of such amplifiers in systems applications up to at least W-band (75–110 GHz). In addition, the availability of high-performance monolithic millimeter-wave integrated circuits (MMICs) fostered the development of sufficiently broadband frequency multiplier chains with GaAs Schottky varactor diodes up to at least 2.5 THz14,15. Despite these major technological advances, there have been few practical demonstrations of fundamental oscillators with three-terminal devices above 100 GHz. Furthermore, only low RF power levels of less than 1 mW were reported16–19. In contrast, two representatives of active two-terminal devices, i.e. Gunn or transferred-electron devices (TEDs) and impact avalanche transit-time (IMPATT) diodes, are operated at higher oscillation frequencies and, in oscillators, generate higher RF output power levels than three-terminal devices. Together with electronic vacuum tubes such as backward-wave oscillators (BWOs), these two-terminal devices continue to play an important role in systems applications at high millimeter-wave and submillimeter-wave frequencies. This paper discusses the concepts and compares the performance potential of active two-terminal devices that meet the following criteria: they either exhibit a strong potential of reaching submillimeter-wave frequencies up to at least 1 THz as evidenced by detailed device analysis or simulations or already generated substantial amounts of CW RF output power with excellent noise performance at frequencies above 300 GHz. 2. Technology Issues The dc-to-RF conversion efficiencies of electronic RF sources generally decrease with frequency and, at high millimeter-wave frequencies, rarely exceed 10%3,4. As a consequence, most of the dc input power is dissipated in the devices as heat, and proper heat management becomes as crucial a technological issue as appropriate low-loss circuits in order to extract the

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

71

highest RF power levels from these devices. Selective etching technologies3,4,20 yield substrateless devices with low thermal resistances, particularly on diamond heat sinks, but also help minimize the skin-effect resistance as much as possible. These technologies enabled the recent experimental results above 100 GHz with state-of-the-art performance from Si IMPATT diodes21, GaAs IMPATT diodes20,22, GaAs tunnel-injection transit-time (TUNNETT) diodes20,23–26, and InP Gunn devices20,27–31. The RF properties of these devices are discussed in subsequent sections. The use of diamond heat sinks or heat spreaders becomes mandatory when the highest RF power levels at millimeter- and submillimeter-wave frequencies need to be generated with these devices3,4,20. 3. Fundamental Device Limitations and Potential Solutions Waveguide circuits, commonly used with active two-terminal devices above 60 GHz2,3,31, are limited in how small a positive imaginary part Xres they offer to the active device near the circuit resonance2,28,29. To fulfill the oscillation conditions3,4 the negative imaginary part Xdev of the device, which arises mostly from the “cold” capacitance Ccold, must be matched (Xres + Xdev = 0). As Ccold of the device scales with device area A, the oscillation conditions impose an upper limit on A. Immaterial whether effects from the carrier transit time τ = ldev/vs are exploited to create the NDR or are detrimental (as, for example, in resonant tunneling diodes and three terminal devices), the material-dependent carrier drift velocity vs forces the device length ldev to decrease with frequency. As Ccold = εrε0A/ldev has an upper limit from the aforementioned oscillation conditions, A strongly decreases with frequency. Avalanche breakdown imposes an upper limit, Ebr, on the electric fields in any electronic device. Therefore, the decrease in ldev, Vbias ∝ Ebr × ldev, and A with operating frequency is responsible for the well-known fundamental limits in the power generation capabilities of all electronic devices. Nonetheless, at least two potential solutions exist to overcome the limitations, for example, in active two-terminal devices, and, most importantly, initial promising results have already been obtained experimentally. Most active two-terminal devices have strongly nonlinear properties, and these properties suggest their use as self-oscillating frequency multipliers. This extraction of power at higher harmonic frequencies requires the oscillation condition only to be met at the fundamental frequency and the transit time τ in the device to be chosen in relation to the (longer) RF period of the fundamental frequency. This means that the limits on Ccold and consequently A are less restrictive. In addition, somewhat higher contact resistances ρc are tolerated.

72

H. EISELE

This principle has been known for decades and has been exploited successfully, in particular, with transferred-electron devices3,28–31. Recently, it has been shown to work equally well with several other types of active twoterminal devices and some examples with resonant tunneling diodes (RTDs), IMPATT diodes, and TUNNETT diodes are given in subsequent sections. Semiconductor material systems with high values of Ebr and vs as well as low values of the dielectric constant εrε0 are particularly well suited to overcome the aforementioned limitations and one example of a novel device structure is discussed in Section 6. Superlattices allow achieving values of vs that are different from the bulk semiconductor materials composing them. Therefore, they are another method to overcome the aforementioned limitations, and initial experimental results as described in Section 5 are highly promising. 4. Resonant Tunneling Diodes Resonant tunneling through discrete energy levels of a so-called quantum well was first described in 197432. Major advances in growth techniques, such as molecular beam epitaxy (MBE) or metalorganic chemical vapor deposition (MOCVD), had to occur in the 1980s before device structures suitable for oscillators could be grown and evaluated. As can be seen from Figure 1, RTDs in the InAs/AlSb material system yielded the highest oscillation frequency of any fundamental solid-state RF source33,34, but RF power levels of, for example, 3 µW at 360 GHz and 0.3 µW at 712 GHz are low34. This makes it very difficult to employ standard phase-locking techniques either for frequency stabilization or in systems applications where the exact oscillation frequency must be known or be tied to a very accurate reference frequency. Recently, higher RF power levels of 23 µW at the fundamental frequency of 342 GHz and 0.6 µW at the third-harmonic frequency of 1.02 THz were measured with RTDs in the GaInAs/AlAs material system, but the RTDs were operated in a “quasi-CW” mode with a pulse length of 0.3 ms and a repetition rate of 300 Hz35. The NDR that is present in RTDs from dc to the highest oscillation frequencies causes such RF sources to be prone to severe instabilities or bias oscillations, which ultimately limit their RF power generating capabilities. Monolithic integration36 and power combining36,37 was shown to alleviate some of these problems. Although an example of an RTD-based local oscillator (LO) for a receiver system was demonstrated38, problems associated with RTDs have so far prevented their use in systems applications. Present measurement equipment offers only limited sensitivity, which degrades significantly at submillimeter-wave frequencies. Therefore, the low RF power levels from

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

73

RTDs also severely restrict how accurately important oscillator characteristics, such as free-running oscillator spectra and oscillation frequency shift vs. temperature or dc bias, can be determined. A record dc-to-RF conversion efficiency of 50% at a microwave frequency of 2 GHz for an RF output power of 20 mW39 was reported from a quantum well injection transit-time (QWITT) diode, which, in this particular case, was an RTD with a very short adjacent drift region. In contrast and as shown in Figure 1, dc-to-RF conversion efficiencies of RTDs at millimeter- and submillimeter-wave frequencies fall well below 1% and may indicate inferior impedance matching and/or significant losses from series resistances in ohmic contacts and RF circuit.

Figure 1. Published state-of-the art results from RTDs and SLEDs in the 30–1,000-GHz frequency range. Numbers next to the symbols denote dc-to-RF conversion efficiencies in percent, where applicable.

5. Superlattice Electronic Devices Recently than RTDs, superlattice electronic devices (SLEDs) in the GaAs/AlAs and InGaAs/InAlAs material systems were demonstrated as millimeter-wave oscillators40–45. SLEDs exploit the Bloch effect, which occurs

74

H. EISELE

when the barrier layers that separate the quantum wells of the superlattice are thinner than the de Broglie wavelength of electrons in the wells. Instead of discrete energy levels for electrons, as exploited in the aforementioned RTDs, energy minibands form. Electrons are energetically confined within these minibands. In the case of no scattering, they undergo oscillatory motion, the so-called Bloch oscillations, when a dc bias is applied across the superlattice. With ever-present phonon scattering, electrons lose some of their energy as they travel, and then the Bloch effect causes a region of negative differential mobility (NDM) in the average electron drift velocity46. SLEDs exploit this NDM in a mode of operation similar to that of Gunn devices (which are described in Section 7). The NDM, however, arises from a very different physical process, the Bloch effect, with much shorter relevant time constants than those of the transferred electron effect in Gunn devices44. Therefore, SLEDs are expected to reach much higher oscillation frequencies47 than, for example, InP Gunn devices. In addition to the results from RTDs, Figure 1 also summarizes the published results from SLEDs40–45. These SLEDs were typically evaluated as quasi-planar devices, first in standard rectangular metallic waveguides40,41 and then in a configuration with contact pads for a standard coplanar waveguide ground-signal-ground probe to be connected to a spectrum analyzer42–44. Recently, they were mounted and tested in waveguide cavities micromachined in the photosensitive SU-8 epoxy45. SLEDs are not associated with the same severe restrictions3,4,20 that bias instabilities impose on the device areas A of RTDs and, as shown in Figure 1, their RF output power levels and dc-to-RF conversion efficiencies tend to be higher than those of RTDs. The result with the so far best dc-to-RF conversion efficiency of 5% at 64.4 GHz was obtained in the aforementioned micromachined waveguide cavity. These promising results, however, are thought to be limited by heat dissipation issues. As can be seen from Figure 2, first attempts at adopting the same advanced fabrication technologies as for state-of-the-art GaAs TUNNETT diodes3,4,20 resulted in two orders of magnitude higher RF output power levels from GaAs/AlAs SLEDs on integral heat sinks when they were operated in the fundamental mode around 65 GHz. Two similar standard superlattices were employed, one with a nominal n-type doping of 1×1017 cm−3 and 120 periods of 12 monolayers of GaAs and 3 monolayers of AlAs (wafer 1)48, the other with a nominal n-type doping of 1.4×1017 cm−3 and 100 periods of 14 monolayers of GaAs and 3 monolayers of AlAs (wafer 2).

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

75

Figure 2. RF output power levels and corresponding dc-to-RF conversion efficiencies (in percent) of GaAs/AlAs SLEDs on integral heat sinks operating in the fundamental mode around 65 GHz (◄: wafer 1, ►: wafer 2, see text).

Figure 3. Spectrum of a free-running oscillator with a superlattice electronic device in a second-harmonic mode, center frequency: 132.64 GHz, vertical scale: 10 dB/div., horizontal scale: 2 MHz/div., resolution bandwidth: 200 kHz, video bandwidth: 10 kHz.

Devices from the same wafer 1 (and also on integral heat sinks) but operated in a second-harmonic mode yielded RF output power levels of more than 10 mW around 130 GHz with corresponding dc-to-RF conversion efficiencies of up to 0.9%. Figure 3 shows an example of the clean spectra that are achieved using SLEDs in free-running oscillators, in this case operating in a second-harmonic mode at 132.6 GHz. Strong signals were also detected in the waveguide circuit at even higher harmonics,

76

H. EISELE

for example, the third and fourth harmonic 48, and this observation is in line with previous results 40–45. 6. Transit-Time Diodes Si IMPATT diodes were the first semiconductor devices to generate RF power above 300 GHz and, as shown in Figure 4 (and in conjunction with Figures 1 and 8), yielded the highest RF power levels from any solid-state fundamental RF source up to 300 GHz. Exemplary RF power levels (and corresponding dc-to-RF conversion efficiencies) of 50 mW (<1%) at 245 GHz51, 12 mW (<0.5%) at 255 GHz51, 7.5 mW (0.35%) at 285 GHz52, 1.2mW (<0.05%) at 301 GHz53, and 0.2 mW at 361 GHz52 were measured at very high operating junction temperatures of typically more than 300 C with waveguide circuits at room temperature54. Higher RF power levels and oscillation frequencies were achieved by cooling the heat sink of the diode and the waveguide circuit to 77K (liquid nitrogen) and 4.5 mW (0.13%) at 295 GHz, 2.2 mW (0.047%) at 412 GHz were reported55. Operation in a harmonic mode has been assumed at least for the IMPATT diodes with some of the highest frequencies56, but the modes of operation of submillimeter-wave Si IMPATT diodes have never been studied in greater detail. Recent results from full-band Monte Carlo simulations indicate that the full potential of Si-based mixed tunneling and impact ionization transittime (MITATT) and Si IMPATT diodes around 300 GHz has not yet been exploited60. Tunneling is known as a fast and “quiet” carrier injection mechanism and makes the tunnel-injection transit-time (TUNNETT) diode another prime candidate for a submillimeter-wave source. Its CW operation has been demonstrated already up to 706 GHz23–26,29,48,57–59 (with RF output powers of 0.6 nW and 0.2 nW at 655 GHz and 706 GHz, respectively) and efficient heat management with diamond heat sinks has been employed exclusively. The aforementioned extraction of power from GaAs TUNNETT diodes at higher harmonics yielded state-of-the-art RF power levels exceeding, for example, 10 mW, 9 mW, and 4 mW at second-harmonic frequencies of 202 GHz, 210 GHz, and 235 GHz, respectively24,25,29 and more than 0.14 mW in a third-harmonic mode at 355 GHz26. These results are also included in Figure 4. With a DC power consumption of less than 0.8 W, the 355-GHz oscillator can be operated from a battery. Figure 3 shows its clean spectrum, which agrees well with the excellent noise performance of TUNNETT diodes from the same epitaxial material in the fundamental mode20,23 and a second-harmonic mode24,25,29. In contrast, the

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

77

Figure 4. State-of-the art results from Si and GaAs transit-time diodes under CW operation in the 30–400-GHz frequency range. Numbers next to the symbols denote dc-to-RF conversion efficiencies in percent.

noise performance of Si IMPATT diodes severely limits for what systems applications they are considered and used20 although they are still the most powerful solid-state sources at high millimeter-wave frequencies of 100– 300 GHz3,4,20. As mentioned in Section 3, the carrier drift velocity vs, breakdown electric field Ebr, and dielectric constant εrε0 are the most relevant semiconductor material parameters that determine the ultimate performance and frequency limits in transit-time devices. These parameters, but also the thermal conductivity are more favorable in GaN than in GaAs4,20. However, the doping levels required for interband tunneling in GaN pn-junctions are very difficult to achieve with current epitaxial growth methods.

78

H. EISELE

Figure 5. Spectrum of a free-running TUNNETT diode oscillator in a third-harmonic mode, center frequency: 354.802 GHz, vertical scale: 10 dB/div, horizontal scale: 500 kHz/div, resolution bandwidth: 100 kHz, video bandwidth: 5 kHz.

A novel transit-time device structure that utilizes the unique tunneling properties of GaN/AlGaN heterostructures was proposed recently61. These tunneling properties arise from large piezoelectric and spontaneous polarization charges at the GaN/AlGaN heterojuntion interface62,63. Initial results from an energy-momentum simulation program indicate the strong potential of such a device as a low-noise millimeter- and submillimeterwave oscillator with RF output power levels of, for example, more than 35 mW around 300 GHz and more than 11 mW around 450 GHz from two device designs with similar tunnel-injection regions, but drift regions of different lengths. RF power generation over wide frequency range was observed with all device designs so far. More than 70% of the maximum power was available between 250 GHz and 370 GHz with a design for 300 GHz as shown in Figure 6 and between 350 GHz and 480 GHz with a design for 440 GHz as shown in Figure 7. Power generation over such wide frequency ranges makes this novel device structure well suited for a tunable oscillator if an appropriate RF circuit with a wide tuning range is used. Bias voltages and dc input power levels are low enough for portable systems applications. In particular, bias voltage of less than 8 V or smaller and a total power consumption of less than 1.3 W allow devices with either design to be operated from a battery.

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

79

7. Transferred-Electron Devices Like the aforementioned TUNNETT diodes, transferred-electron (Gunn) devices have been known and studied as low-noise oscillators for more than 40 years, from microwaves to the highest millimeter-wave frequentcies. The transferred-electron effect in these devices depends only on bulk semiconductor material properties, i.e. the presence of at least two suitable energy minima in the conduction band, and more than ten semiconductor materials in the III–V and II–VI groups are known to exhibit such characteristics2,3,31. However, only devices in the GaAs and InP material systems have been exploited commercially for systems applications. Despite the rapid progress in three-terminal devices, Gunn devices continue to play a critical role in transmitter applications or as low-noise local oscillators (LO) at millimeter-wave and submillimeter-wave frequencies, especially if vacuum tubes or bulky gas lasers need to be avoided.

Figure 6. Predicted RF performance of an AlGaN/GaN heterostructure transit-time device designed for 300 GHz. ● ● : RF output power, ■■: dc-to-RF conversion efficiency for ρc = 5 × 10−7 Ω•cm2; -●--●-: RF output power, -■--■-: dc-to-RF conversion efficiency for ρc = 1 × 10−6 Ω•cm2; VDC = 8 V, JDC = 100 kA•cm−2, RL = 1 Ω, Tj < 500K.

Favorable material parameters of InP, for example, include a higher thermal conductivity, a higher effective transit velocity, and, most importantly, much shorter energy relaxation, acceleration, and deceleration times3,4,20,31. Relaxation times impose fundamental physical limits on the maximum operating frequencies of Gunn devices with short active regions,

80

H. EISELE

Figure 7. Predicted RF performance of an AlGaN/GaN heterostructure transit-time device designed for 450 GHz. ● ● : RF output power, ■■: dc-to-RF conversion efficiency for ρc = 3 × 10−7 Ω•cm2; -●--●-: RF output power, -■--■-: dc-to-RF conversion efficiency for ρc = 5 × 10−7 Ω•cm2; VDC = 7.5 V, JDC = 190 kA•cm−2, RL = 1 Ω, Tj < 500K.

and these frequency limits have long been thought to be in the range of 100– 130 GHz in GaAs and 200–230 GHz in InP. The difference in the fundamental frequency limits is evident from Figure 8 where a significant roll-off in the performance of GaAs Gunn devices occurs at around 80 GHz, whereas it sets in at approximately twice this frequency for InP Gunn devices. The inherent material properties permit InP Gunn devices to reach higher operating frequencies and generate higher RF power levels3,4,20,31. Therefore, this section mainly concentrates on the properties and characteristics of millimeter- and submillimeter-wave oscillators with InP Gunn devices. As shown in Figure 8, the highest RF output power levels from any Gunn device above 80 GHz were achieved with devices that had a threelayer structure (n+n–n+) as well as a graded doping profile (3LGP) in the active region and were mounted on diamond heat sinks20,27,29,31. Operation in the fundamental mode was observed up to 165 GHz and operation in a second-harmonic or a third-harmonic mode was identified as a way to overcome the inherent physical frequency limits in Gunn devices. Operation in a third-harmonic mode yielded the so-far highest operating frequency with a Gunn device of 426 GHz48 and RF output power levels of more than 40 µW at 412 GHz and 422 GHz49.

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

81

RF power levels of, for example, 34 mW at 193 GHz, 26 mW at 199 GHz, and more than 1 mW at 315 GHz were measured with 3LGP structures that were optimized not for operation in this second-harmonic mode, but for operation in the fundamental mode28–31. Higher secondharmonic frequencies of up to 345 GHz were observed with devices that had a 3LGP structure optimized for harmonic power extraction48,50,64. RF power levels 3.7 mW and 1.6 mW were measured with calorimetric power meters at 297 GHz and 329 GHz, respectively50. With bias voltages of less than 6 V and dc input power levels of less than 1.2 W and 0.9 W, respectively, these two oscillators are good examples that such sources can be operated easily from a battery. As can be seen from Figures 1, 4, and 8, the results above 290 GHz make InP Gunn devices the most powerful fundamental RF sources that are operated at room temperature and contain a single electronic semiconductor device per source. Figure 9 compares the measured performance of frequency multipliers using GaAs Schottky-barrier varactor diodes or InP-based heterojunctionbarrier varactor (HBV) diodes with that of select active two-terminal devices, such as RTDs, GaAs TUNNETT diodes, and InP Gunn devices. The performance of InP Gunn devices compares favorably with the undoubtedly impressive performance of frequency multipliers14,15, in particular when one takes into account that (1) the (power-combined) RF driver sources in multipliers require far more dc input power than a single InP Gunn device64; (2) typically two, four, or eight effectively power-combined varactor diodes per multiplier stage are required for safe RF input power handling14,15. Power combining of second-harmonic mode InP Gunn devices was achieved with an in-line waveguide circuit and, for example, a combined RF output power of 6.1 mW with a corresponding combining efficiency of 75% was measured at 285 GHz50. Recently, a combined RF output power of more than 9 mW with a corresponding combining efficiency of more than 100% was measured after the spacing between the two oscillator cavities had been adjusted with thin WR-6 waveguide washers. The spectra of free-running oscillators with harmonic-mode InP Gunn devices are very clean up to the highest frequencies ever measured 28,30,48,50. They correctly reflect the excellent noise performance as seen with InP Gunn devices operating in the fundamental mode3,27–31. Figure 10 with the spectrum of a second-harmonic mode oscillator at 327 GHz serves as an example.

82

H. EISELE

Figure 8. State-of-the art results from GaAs and InP Gunn devices under CW operation in the 30–400-GHz frequency range. Numbers next to the symbols denote dc-to-RF conversion efficiencies in percent.

The RF power levels from InP Gunn devices in the fundamental mode and a second-harmonic mode actually exceed early performance predictions. The results of more recent simulations48,64 are included in Figure 9 and they indicate that further increases in RF power levels are expected from more optimized device structures and circuits. These simulations employed a Monte Carlo-based harmonic balance (MCHB) technique that had shown good agreement between performance predictions and measured results65. This MCHB technique is based on the Monte Carlo code that was used in the design of structures that yielded record performance in the fundamental-mode above 100 GHz66. The simulations results for devices operating in a second-harmonic mode are summarized in Figure 11. These results indicate that InP Gunn devices have the strong potential of generating useful RF output power levels of several mW at least up to 450–500 GHz48,64. They also show what performance advantage is gained from using a device structure that has an active region with a doping notch at the cathode and a flat doping profile to

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

83

Figure 9. Published state-of-the-art results from frequency multipliers with GaAs Schottkybarrier varactors or InP-based heterojunction-barrier varactors (HBV) in the 100–1,200-GHz frequency range in comparison with published state-of-the-art results from GaAs TUNNETT diodes, InP Gunn devices, and RTDs above 200 GHz and performance predictions for InP Gunn devices (– – –) and AlGaN/GaN heterostructure transit-time devices (——). Frequency multipliers and fundamental oscillators were all operated at room temperature except where noted.

Figure 10. Spectrum of a free-running oscillator with an InP Gunn device in a secondharmonic mode, center frequency: 326.94 GHz, vertical scale: 10 dB/div., horizontal scale: 0.5 MHz/div., resolution bandwidth: 100 kHz, video bandwidth: 3 kHz).

H. EISELE

84

the anode side. However, more theoretical and experimental work in the areas of device and RF circuit optimization is required to exploit this potential to the full extent and to establish the ultimate performance limits of InP Gunn devices and their causes64. 8. Conclusion Figure 9 not only summarizes the published experimental results from frequency multipliers and select active two-terminal devices, but also includes the performance predictions for AlGaN/GaN heterostructure transit-time devices and InP Gunn devices from Figures 6, 7, and 11, respectively. This comparison emphasizes the strong potential of InP Gunn devices to generate useful power levels up to at least 500 GHz. In addition, SLEDs and the novel AlGaN/GaN heterostructure transit-time devices are the most promising approaches to realizing fundamental solid-state sources with RF output power up to 1 THz. More detailed comparisons of experimental results and performance predictions are necessary to refine the simulation methods and designs for high-performance InP Gunn devices and AlGaN/GaN heterostructure transit-time devices at submillimeter-wave frequencies. These advances in design tools, but also significant advances in material choices for superlattices, in material growth for GaN/AlGaN heterostructures, and in fabriccation technologies for GaN-based devices are necessary before the full potential of these active two-terminal devices can be harnessed. 100

graded structure notch structure rel. improvement

50

90 80 70

40

60 30

50 40

20

30 20

10

Relative improvement [%]

RF output power [mW]

60

10 0 200

250

300

350

400

450

500

Frequency [GHz] Figure 11. Performance improvement from doping profile optimization in second-harmonic InP Gunn devices.

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

85

Acknowledgment The author kindly acknowledges contributions to this work from many people, in particular, Ian Farrer, George I. Haddad, Ridha Kamoua, Madhusudan Singh, and Yuh-Renn Wu.

References 1. P. Siegel, Terahertz technology, IEEE Trans. Microw. Theory Tech., 50, 3, 910–928 (March 2002). 2. S. M. Sze, Physics of Semiconductor Devices, 2nd Edn. (Wiley, New York, 1981). 3. H. Eisele and G. I. Haddad, Active Microwave Diodes, in Modern Semiconductor Devices, S. M. Sze, Ed., Ch. 6, pp. 343–407 (Wiley, New York, 1997). 4. G. I. Haddad, J. East, and H. Eisele, Two-terminal active devices for terahertz sources, Terahertz Sensing Technology, D. L. Woolard, W. R. Leorop, and M. S. Shur, Eds., vol. I: Electronic Devices and Advanced Systems Technology, pp. 45–77 (World Scientific Publishing, Hong Kong, 2003). 5. H. Wang, R. Lai, Y. L. Kok, T. W. Huang, M. V. Aust, Y. C. Chen, P. H. Siegel, T. Gaier, R. J. Dengler, and B. R. Allen, A 155-GHz monolithic low-noise amplifier, IEEE Trans. Microw. Theory Tech., 46, 11, 1660–1666 (Nov. 1998). 6. B. Agarwal, A. E. Schmitz, J. J. Brown, M. Matloubian, M. G. Case, M. Le, M. Lui, and M. J. W. Rodwell, 112-GHz, 157-GHz, and 180-GHz InP HEMT traveling-wave amplifiers, IEEE Trans. Microw. Theory Tech., 46, 12, 2553–2559 (Dec. 1998). 7. S. Weinreb, P. C. Chao, and W. Copp, Full-waveguide band, 90 to 140 GHz, MMIC amplifier module, IEEE MTT-S Int. Microwave Symposium Digest, Denver, CO, pp. 1279–1280 (June 17–21, 1997). 8. R. Lai, M. Barsky, T. Huang, M. Sholley, H. Wang, Y. L. Kok, D. C. Streit, T. Block, P. H. Liu, T. Gaier, and L. Samoska, An InP HEMT MMIC LNA with 7.2-dB gain at 190 GHz, IEEE Microw. Guided Wave Lett., 8, 11, 393–395 (Nov. 1998). 9. Q. Lee, B. Agarwal, D. Mensa, R. Pullela, J. Guthrie, L. Samoska, and M. J. W. Rodwell, A > 400 GHz fmax transferred-substrate heterojunction bipolar transistor IC technology, IEEE Electron Device Lett., 19, 3, pp. 77–79 (March 1998). 10. D. Mensa, Q. Lee, J. Guthrie, S. Jaganathan, and M. J. W. Rodwell, Transferredsubstrate HBTs with 254 GHz fτ, Electron. Lett., 35, 7, 288–290 (April 1999). 11. R. T. Webster and A. F. M. Anwar, Noise in metamorphic AlGaAsSb/InGaAs/ AlGaAsSb HEMTs, Solid-State Electron., 48, 10–11, 2007–2011 (Nov. 2004). 12. Y. C. Chen, D. L. Ingram, R. Lai, M. Barsky, R. Grunbacher, T. Block, H. C. Yen, and D. C. Streit, A 95-GHz InP HEMT MMIC amplifier with 427-mW power output, IEEE Microw. Guided Wave Lett., 8, 11, 399–401 (Nov. 1998). 13. H. Wang, L. Samoska, T. Gaier, A. Peralta, H. H. Liao, Y. C. Chen, M. Nishimoto, and R. Lai, Monolithic power amplifiers covering 70–113 GHz, Radio Frequency Integrated Circuits Symposium Digest IEEE RFIC 2000, pp. 39–42 (June 11–13, 2000). 14. T. W. Crowe, D. W. Porterfield, J. L. Hesler, W. L. Bishop, and D. S. J. Kurtz, Integrated terahertz transmitters and receivers, Conference Chemical and Biological Standoff Detection, Providence, RI, Proceedings. SPIE – International Society for Optical Engineering vol. 5268, pp. 1–9 (2003). 15. J. Ward, E. Schlecht, G. Chattopadhyay, A. Maestrini, J. Gill, F. Maiwald, H. Javadi, and I. Mehdi, Capability of THz sources based on Schottky diode frequency multiplier

86

16. 17.

18. 19.

20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34.

H. EISELE chains, International Microwave Symposium Digest IEEE MTT-S 2004, pp. 1587–1590 (June 6–11, 2004). Y. Kwon, D. Pavlidis, T. L. Brock, and D.C. Streit, A D-band monolithic fundamental oscillator using InP-based HEMT’s, IEEE Trans. Microw. Theory Tech., 41, 12, 2336– 2344 (Dec. 1993). S. E. Rosenbaum, B. K. Kormanyos, L. M. Jelloian, M. Matloubian, A. S. Brown, L. E. Larson, L. D. Nguyen, M. A. Thompson, L. P. B. Katehi, and G. M. Rebeiz, “155- and 213-GHz AlInAs/GaInAs/InP HEMT MMIC oscillators”, IEEE Trans. Microw. Theory Tech., 43, 4, 927–933 (April 1995). K. Uchida, H. Matsuura, T. Yakihara, S. Kobayashi, S. Oka, T. Fujita, and A. Miura, A series of InGaP/InGaAs HBT oscillators up to D-band, IEEE Trans. Microw. Theory Tech., 49, 5, 858–865 (May 2001). S. Kudszus, W. H. Haydl, A. Tessmann, W. Bronner, and M. Schlechtweg, Push-push oscillators for 94 and 140 GHz applications using standard pseudomorphic GaAs HEMTs, International Microwave Symposium Digest IEEE MTT-S 2001, pp. 1571– 1574 (May 20–25, 2001). H. Eisele and G. I. Haddad, Two-terminal millimeter-wave sources, IEEE Trans. Microw. Theory Tech., 46, 6, 739–746 (June 1998). M. Wollitzer, J. Buechler, F. Schäffler, and J.-F. Luy, D-band Si IMPATT diodes with 300 mW CW output power at 140 GHz, Electron. Lett., 32, 2, 122–123 (Jan. 1996). M. Tschernitz and J. Freyer, 140 GHz GaAs double-Read IMPATT diodes, Electron. Lett., 31, 7, 582–583 (March 1995). H. Eisele and G. I. Haddad, Enhanced performance in GaAs TUNNETT diode oscillators above 100 GHz through diamond heat sinking and power combining, IEEE Trans. Microw. Theory Tech., 42, 12, 2498–2503 (Dec. 1994). H. Eisele, Efficient second-harmonic power extraction from GaAs TUNNETT diodes above 200 GHz, Electron. Lett., 34, 13, 1324–1326 (June 1998). H. Eisele, Efficient second-harmonic power extraction from GaAs TUNNETT diodes above 200 GHz, Electron. Lett., 34, 5, 1531 (July 1998). H. Eisele, 355-GHz oscillator with GaAs TUNNETT diode, Electron. Lett., 41, 6, 329– 331 (March 2005). H. Eisele and G. I. Haddad, High-performance InP Gunn devices for fundamental-mode operation in D-band (110–170 GHz), IEEE Microw. Guided Wave Lett., 5, 11, 385–387 (Nov. 1995). H. Eisele, Second-harmonic power extraction from InP Gunn devices with more than 1 mW in the 260–320 GHz frequency range, Electron. Lett., 34, 25, 2412–2413 (Dec. 1998). H. Eisele, A. Rydberg, and G. I. Haddad, Recent advances in the performance of InP Gunn devices and GaAs TUNNETT diodes for the 100–300-GHz frequency range and above, IEEE Trans. Microw. Theory Tech., 48, 4, 626–631 (April 2000). H. Eisele, High performance InP Gunn devices with 34 mW at 193 GHz, Electron. Lett., 38, 16, 923–924 (Aug. 2002). H. Eisele, Gunn or transferred-electron devices, in Encyclopedia of RF and Microwave Engineering, K. Chang, Ed., vol. 2, pp. 1857–1884 (Wiley, New York, 2005). L. L. Chang, L. Esaki, and R. Tsu, Resonant tunneling in semiconductor double-barriers, Appl. Phys. Lett., 24, 12, 593–595 (1974). E. R. Brown, C. D. Parker, A. R. Calawa, M. J. Manfra, C. L. Chen, L. J. Mahoney, and W. D. Goodhue, High-frequency resonant-tunneling oscillators, Microw. Opt. Tech. Lett., 4, 1, 19–23 (Jan. 1991). E. R. Brown, J. R. Söderström, C. D. Parker, L. J. Mahoney, K. M. Molvar, and T. C. McGill, Oscillations up to 712 GHz in InAs/AlSb resonant tunneling diodes, Appl. Phys. Lett., 58, 20, 2291–2293 (May 1991).

FUNDAMENTAL SOLID-STATE TERAHERTZ SOURCES

87

35. N. Orihashi, S. Suzuki, and M. Asada, One THz harmonic oscillation of resonant tunneling diodes, Appl. Phys. Lett., 87, 23, 233501-1–3 (Nov. 2005). 36. M. Reddy, S. C. Martin, A. C. Molnar, R. E. Muller, R. P. Smith, P. H. Siegel, M. J. Mondry, M. J. W. Rodwell, H. Kroemer, and S. J. Allen, Monolithic Schottky-collector resonant tunnel diode oscillator arrays to 650 GHz, IEEE Electron Device Lett., 18, 5, 218–221 (May 1997). 37. D. P. Steenson, R. E. Miles, R. D. Pollard, J. M. Chamberlain, and M. Henini, Demonstration of power combining at W-band from GaAs/AlAs resonant tunneling diodes, Proceedings of the 5th International Symposium Space Terahertz Technology, Ann Arbor, MI, pp. 756–767 (May 10–12, 1994). 38. R. Blundell, D. C. Papa, E. R. Brown, and C. D. Parker, Resonant tunneling diode as an alternative LO for SIS receiver applications, Electron. Lett., 29, 3, 288–290 (Feb. 1993). 39. S. Javalagi, V. Reddy, K. Gullapalli, and D. Neikirk, High efficiency microwave diode oscillators, Electron. Lett., 28, 18, 1699–1701 (Aug. 1992). 40. E. Schomburg, J. Grenzer, K. Hofbeck, T. Blomeier, S. Winnerl, S. Brandl, A. A. Ignatov, K. F. Renk, D. G. Pavel’ev, Yu. Koschurinov, V. Ustinov, A. Zhukov, A. Kovsch, S. Ivanov, and P. S. Kop’ev, Millimeter wave generation with a quasi planar superlattice electronic device, Solid-State Electron., 42, 7–8, 1495–1498 (Aug. 1998). 41. E. Schomburg, R. Scheuerer, S. Brandl, K. F. Renk, D. G. Pavel’ev, Yu. Koschurinov, V. Ustinov, A. Zhukov, A. Kovsh, and P. S. Kop’ev, An InGaAs/InAlAs superlattice oscillator at 147 GHz, Electron. Lett., 35, 17, 1491–1492 (Aug. 1999). 42. E. Schomburg, S. Brandl, K. Hofbeck, T. Blomeier, J. Grenzer, A. A. Ignatov, K. F. Renk, D. G. Pavel’ev, Yu. Koschurinov, V. Ustinov, A. Zhukov, A. Kovsch, S. Ivanov, and P. S. Kop’ev, Generation of millimeter waves with a GaAs/AlAs superlattice oscillator, Appl. Phys. Lett., 72, 12, 1498–1500 (March 1998). 43. S. Brandl, E. Schomburg, R. Scheurer, K. Hofbeck, J. Grenzer, K. F. Renk, D. G. Pavel’ev, Yu. Koschurinov, A. Zhukov, A. Kovsch, V. Ustinov, S. Ivanov, P. S. Kop’ev, Millimeter wave generation by a self-sustained current oscillation in an InGaAs/InAlAs superlattice, Appl. Phys. Lett., 73, 21, 21, 3117–3119 (Nov. 1998). 44. E. Schomburg, M. Henini, J. M. Chamberlain, P. Steenson, S. Brandl, K. Hofbeck, K. F. Renk, and W. Wegscheider, Self-sustained current oscillation above 100 GHz in a GaAs/AlAs superlattice, Appl. Phys. Lett., 74, 15, 2179–2181 (April 1999). 45. M. Häußler, E. Schomburg, J.-M. Batke, F. Klappenberger, A. Weber, H. Appel, K. F. Renk, H. Hummel, B. Ströbl, D. G. Pavel’ev, and Yu. Koschurinov, Millimetre-wave generation with semiconductor superlattice mounted in cavity fabricated by UVphotolithography and galvanoforming, Electron. Lett., 39, 10, 784–785 (May 2003). 46. L. Esaki and R. Tsu, Superlattice and negative differential conductivity in semiconductors, IBM J. Res. Dev., 14, 1, 61–65 (1970). 47. R. Scheuerer, E. Schomburg, K. F. Renk, A. Wacker, and E. Schöll, Feasibility of a semiconductor superlattice oscillator based on quenched domains for the generation of submillimeter waves, Appl. Phys. Lett., 81, 8, 1515–1517 (Aug. 2002). 48. H. Eisele, M. Naftaly, and R. Kamoua, Generation of submillimeter-wave radiation with GaAs TUNNETT diodes and InP Gunn devices in a second or higher harmonic mode, Int. J. Infrared Millimeter Waves, 26, 1, 1–14 (Jan. 2005). 49. H. Eisele, InP Gunn devices for 400–425 GHz, Electron. Lett., 42, 6, 358–359 (March 2006). 50. H. Eisele and R. Kamoua, High-performance oscillators and power combiners with InP Gunn devices at 260–330 GHz, IEEE Microw. Wireless Comp. Lett., 16, 5, 284–286 (May 2006). 51. K. Chang, W. F. Thrower, and G. M. Hayashibara, Millimeter-wave silicon IMPATT sources and combiners for the 110–260-GHz range, IEEE Trans. Microw. Theory Tech., 29, 12, 1278–1284 (Dec. 1981).

88

H. EISELE

52. M. Ino, T Ishibashi, and M. Ohmori, C. W. oscillation with p+-p-n+ silicon IMPATT diodes in 200 GHz and 300 GHz bands, Electron. Lett., 12, 6, 148–149 (March 1976). 53. T. Ishibashi and M. Ohmori, 200-GHz 50-mW CW oscillation with silicon SDR IMPATT diodes, IEEE Trans. Microw. Theory Tech., 24, 11, 858–859 (Nov. 1976). 54. N. B. Kramer and R. A. Johnson, Generating power at mm-wave frequencies, Microwaves RF, 23, 5, 243–249 (May 1984). 55. T. Ishibashi, M. Ino, T. Makimura, and M. Ohmori, Liquid-nitrogen-cooled submillimeter-wave silicon IMPATT diodes, Electron. Lett., 13, 10, 299–300 (May 1977). 56. M. Ohmori, T. Ishibashi, and S. Ono, Dependency of the highest harmonic oscillation frequency on junction diameter of IMPATT diodes, IEEE Trans. Electron Devices, 24, 12, 1323–1329 (Dec. 1977). 57. P. Płotka, J-i. Nishizawa, T. Kurabayashi, and H. Makabe, 240–325-GHz GaAs CW fundamental-mode TUNNETT diodes fabricated with molecular layer epitaxy, IEEE Trans. Electron Devices, 50, 4, 867–873 (April 2003). 58. J-i. Nishizawa, P. Płotka, H. Makabe, and T. Kurabayashi, 290–393 GHz CW fundamental-mode oscillation from GaAs TUNNETT diode, Electron. Lett., 41, 7, 80– 81 (March 2005). 59. J-i. Nishizawa, P. Płotka, H. Makabe, and T. Kurabayashi, GaAs TUNNETT diodes oscillating at 430–655 GHz in CW fundamental mode, IEEE Microw. Wireless Comp. Lett., 15, 9, 597–599 (Sept. 2005). 60. X. Bi, J. R. East, U. Ravaioli, and G.I. Haddad, Analysis and design of terahertz transittime diode, Proceedings of the 16th International Symposium Space Terahertz Technology, Gothenburg, Sweden (May 2–4, 2005) [CD ROM]. 61. H. Eisele, M. Singh, Y.-R. Wu, J. Singh, and G. I. Haddad, AlGaN/GaN heterostructure transit-time devices: a novel device concept for submillimeter-wave sources, Proceedings of the 16th Internatinal Symposium Space Terahertz Technology, Gothenburg, Sweden (May 2–4, 2005) [CD ROM]. 62. M. Singh, Y. Zhang, J. Singh, and U. Mishra, Examination of tunnel junctions in the AlGaN/GaN system: consequences of polarization charge, Appl. Phys. Lett., 77, 12, 1867–1869 (Sept. 2000). 63. M. Singh, J. Singh, and U. Mishra, Current-voltage characteristics of polar heterostructure junctions, J. Appl. Phys., 91, 5, 2989–2993 (March 2002). 64. H. Eisele and R. Kamoua, Submillimeter-wave InP Gunn devices, IEEE Trans. Microw. Theory Techn., 52, 10, 2371–2378 (Oct. 2004). 65. R. Kamoua, Monte Carlo-based harmonic balance technique for the simulation of highfrequency TED oscillators, IEEE Trans. Microw. Theory Tech., 46, 1376–1381 (Oct. 1998). 6 6 . R. Kamoua, H. Eisele, and G. I. Haddad, D-band (110–170 GHz) InP Gunn devices, Solid-State Electron., 36, 1547–1555 (Nov. 1993).

Theme 2 INTERACTIONS WITH MATERIALS

MOLECULAR AND ORGANIC INTERACTIONS A. G. DAVIES* AND E. H. LINFIELD School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, UK

Abstract. We review a selection of recent technological advances in terahertz frequency time-domain spectroscopy. We discuss the coherent generation of ultra-broadband terahertz radiation using a biased and asymmetrically excited low-temperature-grown GaAs photoconductive (PC) emitter. Using a backward collection method, terahertz radiation with frequency components over 30 THz can be collected, the highest observed for PC emitters. We outline two detection schemes, electro-optic (EO) detection using a ZnTe crystal, and PC-detection using a low-temperaturegrown GaAs PC receiver. The use of the PC receiver provides the timedomain spectroscopy system with a smooth spectral distribution between 0.3 and 7.5 THz, ideal for spectroscopic applications. We illustrate the technological developments with examples of transmission spectroscopy of polycrystalline organic materials. Specifically, we review measurements of the vibrational spectra of polycrystalline purine and adenine over the temperature range 4−290K. A number of well-resolved absorption peaks are observed, which are interpreted as originating from intermolecular vibrational modes mediated by hydrogen bonds. As the temperature is reduced, the observed absorption bands resolve into narrower peaks and some shift towards higher frequencies, which can be explained by the anharmonicity of the vibrational potentials. An empirical expression is given to describe this frequency shift.

Keywords: terahertz time-domain spectroscopy

______ *

To whom correspondence should be addressed. A. G. Davies, Institute of Microwaves and Photonics, School of Electronic Engineering, University of Leeds, Leeds LS2 9JT, UK; e-mail: [email protected]

91 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 91–106. © 2007 Springer.

92

A.G. DAVIES AND E.H. LINFIELD

1. Introduction Since the first use of femtosecond lasers to generate coherent terahertz (THz) frequency radiation,1 there has been a drive to develop higher power sources and systems of broader bandwidth.1–3 The motivation is that THz timedomain spectroscopy has been demonstrated to give much better performance in the far-infrared (IR) range than other spectroscopic techniques such as Fourier Transform IR spectroscopy,2,4 making it possible to study the temperature dependence of low-frequency vibrational spectra below 3 THz (100 cm–1),5 accessing new information about the structure and vibrational dynamics of solids. Indeed, such low-frequency vibrations play a crucial role in the functions of proteins and nucleic acids, for example, since they give rise to significant atomic rearrangements and conformational fluctuations.6–8 Hydrogen-bonded networks are particularly important in this context since they are more easily broken than covalent bonds and facilitate structural changes.9 As such, there is significant interest in the extension of spectroscopic investigations to the far-IR, which may give information about intermolecular vibrational modes mediated by hydrogen bonds. A number of techniques have been explored to improve the power and bandwidth of THz time-domain spectroscopy systems, including bulk electro-optic (EO) rectification,10,11 surface field generation,12,13 and ultrafast switching of photoconductive (PC) emitters.14–18 Of these different methods, PC emitters have proved to be the most efficient technique for converting visible/near-IR pulses to THz radiation,17,18 and have been widely used for THz spectroscopy and imaging. In this technique, electron– hole pairs are generated in a semiconductor crystal using an above-band gap femtosecond pulse, and these photoexcited carriers are then accelerated by an applied electric field. The physical separation of the electrons and holes forms a macroscopic space-charge field oriented opposite to the biasing field, and thus, the externally applied field is screened. This fast temporal change in electric field produces a transient current, which generates a pulse of electromagnetic radiation in the THz-frequency range. Theoretical simulations using this electric field screening model suggest that sub-100-fs electrical pulses can be obtained.19 However, until recently, in experiments, 200-fs electric pulses and 350-fs free-space THz pulses were amongst the shortest pulses realized for GaAs emitters, giving a useful bandwidth of about 4 THz. In Section 2, we discuss a procedure that enables THz radiation with frequency components over 30 THz to be obtained from a biased and asymmetrically excited GaAs PC emitter with a ZnTe EO detector. In Section 3, we then describe the use of a GaAs PC receiver as an alternative

MOLECULAR AND ORGANIC INTERACTIONS

93

detection scheme, which provides the time-domain spectroscopy system with a smooth spectral distribution between 0.3 and 7.5 THz, ideal for spectroscopic applications. In Sections 4 and 5, we illustrate the use of this technology for the study of organic polycrystalline materials such as cytidine, purine, and adenine, with temperature dependent spectroscopic measurements presented in Section 5. 2. Ultra-Broadbandwidth Terahertz Time-Domain Spectroscopy Figure 1 shows schematically the experimental arrangement for coherent generation and detection of ultra-broadbandwidth THz radiation. The emitter comprises two vacuum-evaporated NiCr/Au electrodes separated by a 0.4 mm gap (Figure 1(c)) fabricated on a 1-µm-thick low-temperaturegrown (LT) GaAs layer, which was deposited on a 0.53-mm-thick undoped

Figure 1. (a) Arrangement for coherent THz generation and detection. (b) The “backwards” emitted radiation is collected, leading to the significantly enhanced bandwidth. (c) Schematic of the electrodes used in the biased GaAs PC emitter. The white spot indicates the position of the asymmetric laser excitation. (From Y.-C. shen et al.21)

Figure 2. (a) The temporal THz wave form and, (b) its corresponding Fourier transform amplitude spectrum (upper trace, solid line), together with spectrum measured in the presence of PTFE sample (lower trace, dotted line). (From Y.-C. Sher et al.21)

GaAs substrate. LT-GaAs was chosen to give a short photocarrier recombination lifetime (0.4 ps in this case), together with both high resistivity and high carrier mobility.20 A bias voltage of ±120 V, modulated

94

A.G. DAVIES AND E.H. LINFIELD

at 31 kHz, was applied across the emitter. A pulsed Ti:sapphire laser of 300 mW average power (12-fs pulse width, 800-nm centre wavelength, and 76MHz repetition rate) was focused to a 40-µm spot diameter on the edge of one of the two NiCr/Au electrodes of the LT-GaAs emitter, to generate THz pulses. In contrast to previous experiments, where the THz radiation was collected forwards (that is, after being transmitted through the GaAs substrate), the THz radiation was collected backwards (in the direction of the reflected pump laser beam, see Figure 1(b)). As a result, the absorption and dispersion of the THz pulses in the GaAs substrate are minimized.21 The emitted THz pulses were collimated and focused onto the sample by a pair of parabolic mirrors. The transmitted THz pulse (or in an alternative arrangement, this could be the reflected pulse) was then collected and focused using another pair of parabolic mirrors onto a 20-µm-thick (100) ZnTe crystal glued onto a 1-mm-thick wedged ZnTe crystal for EOdetection.22,23 In all measurements, the variable delay stage, which provides the time delay between the THz pulse and the probe pulse, was scanned over a distance of 2 mm, providing a spectral resolution of 75 GHz (2.5 cm–1). Using a lock-in detection scheme referenced to the frequency of the bias across the GaAs emitter, a noise level equivalent to ∆I/I = 8 x 10–9 Hz–1/2 was observed. With this noise level, peak signals obtained in these experiments are as much as a factor of 1,400 above the noise floor. The apparatus shown in Figure 1 was enclosed in a vacuum-tight box, which was purged with dry nitrogen gas to reduce the effects of water vapour absorption. All measurements were performed at room temperature. Figure 2(a) shows a typical resulting temporal THz waveform, with Figure 2(b) showing the corresponding frequency spectrum. The first main positive and negative peaks of the pulse (Figure 2(a)) have full-width-athalf-maxima of 40 and 35 fs, respectively, representing the shortest THz pulses reported for PC emitters. The spectral dips at 5.2 and 8.0 THz (Figure 2(b)) are caused by absorptions in the ZnTe detector (transverseoptical (TO) phonon energy: 22 meV ≡ 5.3 THz) and the GaAs emitter (TO phonon energy: 33 meV ≡ 8.0 THz),24 respectively. Note that frequency components over 30 THz are observed. These represent the highest frequency components reported from a biased PC emitter, and are a direct consequence of the use of the reflection geometry. Figure 2(b) also shows (lower trace) the amplitude spectrum of the THz signal measured after transmission through a polytetrafluroethylene (PTFE) sample (Thermo Co., UK). A number of spectral features are observed, in particular two strong dips at 15.1 and 19.2 THz, caused by absorption owing to CF2 scissoring and wagging vibrations, respectively. In addition, a weak

MOLECULAR AND ORGANIC INTERACTIONS

95

absorption feature at 16.7 THz (CF2 twisting vibration) and a shoulder at 18.8 THz (CF2 wagging vibration) are also visible. Above 20 THz, the measured spectrum became noisy, although absorption features around 21.2, 23.7, and 32.3 THz are still visible. These values agree well with the published data,25 demonstrating the useful bandwidth of this THz spectrometer. An alternative approach for broadband THz generation is optical rectification of femtosecond Ti:sapphire laser pulses at the surface of a nonlinear optic crystal or semiconductor. The spectrum of such a source is also broad, extending in principle to over 30 THz.10 However; owing to lack of phase matching it exhibits only low average power. Higher average power can be obtained by phase-matched difference frequency mixing in nonlinear crystals such as GaSe. The THz radiation generated is tunable from the far-IR to the mid-IR by changing the crystal orientation to achieve phase matching for a given frequency component.11,26

Figure 3. (a) Temporal THz waveform and, (b) corresponding amplitude spectrum. Arrows mark spectral features corresponding to GaAs TO- and LO-phonons. Inset of (a) shows the schematic geometry of the LT-GaAs PC emitter (antenna gap of 400 µm, biased at ±120V at 11 kHz). Inset of (b) shows the schematic geometry of LTGaAs PC receiver (bowtie antenna with 8 µm gap). (From Y.-C. Shen et al.28 )

Figure 4. The Fourier-transform amplitude spectrum of the THz radiation measured after transmitting through a pure polyethylene reference pellet (upper trace), and a maltose/ polyethylene sample pellet (lower trace). Arrows indicate maltose vibrational modes. The Fouriertransforms were performed using the first derivatives of the measured signals, to minimize possible spectral artifacts caused by the offset of the measured signals. (From M. Tani et al.28 )

96

A.G. DAVIES AND E.H. LINFIELD

We also note that, in comparison with published results on GaAs p-i-n vertical diodes,23 where frequency components as high as 60 THz have been observed, the LT-GaAs emitter provides two to three orders of magnitude higher THz power (estimated from the published ∆I/I value).23,27 (It should be noted, however, that higher THz power does not necessary mean higher conversion efficiency because the GaAs p-i-n diodes use less pump laser power.) 3. Generation and Detection of Ultra-Broadband Terahertz Radiation Using Photoconductive Emitters and Receivers In Section 2, we discussed the generation of ultra-broadband THz emission from LT-GaAs PC antennae, using a 20-µm-thick ZnTe crystal as the THz sensor. The measured THz signal was complicated by phonon resonances in the ZnTe detector, and so to address this, we describe here a system combining a PC antenna emitter and a PC antenna receiver for the generation and field-resolved detection of ultra-broadband THz radiation.28 The measured THz signal has a smooth spectral distribution from 0.3 to 7.5 THz, ideal for spectroscopic applications. The general experimental arrangement is as described in Section 2. The Ti:sapphire laser output is again split into two parts: a 400-mW beam is focused onto the surface of a biased LT-GaAs PC emitter for THz generation, but now a 30-mW beam serves as the probe beam to gate a PC receiver antenna for THz detection. For both PC emitter and receiver, the NiCr/Au electrodes are vacuum-evaporated on a 1.0-µm-thick LT-GaAs layer grown at 250°C on a 0.53-mm-thick semi-insulating GaAs substrate. Ex situ post-growth annealing of the LT-GaAs allowed individual optimization of the carrier lifetimes and resistivities for the PC receiver and emitter.29 The laser pulse width incident on the PC emitter and receiver is estimated to be 20-fs. THz detection was achieved using the geometry shown in Figure 3(b), with once again absorption and dispersion in the GaAs substrate minimized. The variable delay stage was scanned over a distance of 2 mm, providing a spectral resolution of 75 GHz (2.5 cm–1). Figure 3(a) shows a typical temporal THz waveform generated from the PC emitter and detected with the PC receiver, with Figure 3(b) showing the corresponding frequency spectrum. In the frequency-domain, the spectrum of the THz radiation spreads continuously up to 15 THz, except for a spectral dip at 8.0 THz and a distinct peak at 8.7 THz. These two spectral features are explained as originating from longitudinal-optical (LO) and TO phonon resonances of the GaAs PC antennae. In the time-domain, the

MOLECULAR AND ORGANIC INTERACTIONS

97

phonon oscillations can be seen immediately after the initial transient, and extend over 12 ps. These oscillations were fitted with an exponentially decaying sine wave of frequency ν,

ETHz (t ) = C1e − t /τ sin( 2πνt − C2 ) ,

(1)

where C1 and C2 are constants, with good agreement (not shown) found between the fitted and the measured THz signal.28 The best-fit results give an oscillating frequency ν ≡ 8.7 THz and a decay time τ ≈ 3.0 ps, confirming that these oscillations originate from the GaAs LO phonon resonance (νLO ≡ 8.76 THz). Ultra-broadband PC-detection was first demonstrated by Kono et al.30 using optical rectification of a femtosecond laser pulse in a 100-µm-thick ZnTe crystal for THz generation (i.e. EO-generation and PC-detection). The amplitude of the THz electric field measured in the work described here is about two orders of magnitude larger (four orders of magnitude larger in THz power) than that reported by Kono et al. owing to the larger power available from the PC emitter. Therefore the combination of PC emitter and PC receiver antennae proves an ideal THz system for practical spectroscopy applications. As an example, Figure 4 shows the vibrational spectrum of maltose measured with such a system. Samples were prepared by mixing finely milled maltose polycrystalline powder (Sigma-Aldrich Co.) with polyethylene powder (Sigma-Aldrich Co.) in a mass ratio of 1:10, and then compressing the mixture to form a pellet of thickness 1.3 mm. The low-frequency vibrational modes of maltose were previously measured with a conventional THz time-domain spectrometer, but with a frequency range limited to 0.3–3.0 THz.31 In contrast, here 14 vibrational modes are resolved over a much wider frequency range (0.3–7.3 THz), demonstrating the extended bandwidth of this THz spectroscopy system. Note that these observed vibrational modes correspond to both intermolecular and intramolecular interactions.32 The THz signals presented in Section 2 covered a broader (over 30 THz) spectral range. The reduced spectral coverage here is mainly a result of the decreased sensitivity of the PC-detection antenna at high frequencies. For EO-detection, the EO crystal detects the polarization change of the probe beam induced by the THz electric field in the sensor crystal. Therefore the detected signal is simply proportional to the THz electric field. For PC-detection of THz radiation, the PC antenna works as an integrating detector. The photocurrent from the antenna, J(t), is proportional to the time integration of the product of the incident THz electric field, ETHz(t), and the photo-generated carrier density, N(t), in the PC antenna:

98

A.G. DAVIES AND E.H. LINFIELD ∞

J (t ) = eµ ∫ ETHz (t ' ) N (t '−t )dt ' , −∞

(2)

where e is the electron charge and µ is the electron mobility of the LTGaAs. Therefore, for an ideal PC antenna with ultrashort carrier lifetime, J (ω ) ∝ ETHz (ω ) ; whilst for a PC antenna with a long carrier lifetime (for example, a semi-insulating GaAs PC antenna), J (ω ) ∝ ETHz (ω ) / ω . For ultrabroadband THz detection using a PC antenna with a finite carrier lifetime, the sensitivity at higher frequencies is thus expected to decrease (we also note that the structure of the PC receiver antenna may affect its spectral response, and that Kono et al. used a dipole-type antenna, different from the bowtie-type PC antenna used here). Nevertheless, PC-detection still provides a better signal-to-noise-ratio (dynamic range) than EO-detection, up to frequencies in excess of 8 THz. 4. Cytidine: An Exemplar PolyCrystalline Organic System To demonstrate further the capability of broadband THz spectroscopy for probing both intramolecular and intermolecular vibrations, the THz transmission spectrum of polycrystalline cytidine was measured using the PC-emission/EO-detection system discussed in Section 2.21 Cytidine is one of four nucleosides found in RNA and DNA, and polycrystalline samples contain an extensive network of intermolecular hydrogen bonds.33,34 Samples were prepared by forming a thin layer of finely milled cytidine powder (Sigma-Aldrich EC No. 2006109, 99% purity) between two transparent 9-µm-thick polyethylene films. Nineteen vibrational modes were observed in the frequency range 1–20 THz (Figure 5). Amongst these, five (12.1, 13.0, 16.6, 17.9, 18.6 THz) have been observed previously using Fourier Transform IR spectroscopy.35 The THz measurements, however, provide a broader spectral coverage, and reveal more vibrational modes. These include so-called external or lattice modes involving the motion of molecules moving relative to each other in the unit cell of the cytidine crystal. The accurate calculation and assignment of such vibrational modes is difficult and requires the inclusion of non-covalent long-range weak forces including hydrogen bonds. 33,36 Indeed, although these forces play a pivotal rule in biomolecular structure and function, they are poorly understood. The rich spectral features of the THz spectrum may provide important information to understand better the nature of such forces.

MOLECULAR AND ORGANIC INTERACTIONS

99

Figure 5. Transmission spectrum of a cytidine sample. Shaded areas centered at 5.1 and 8.0 THz contain singularities caused by the TO phonons in the ZnTe and GaAs crystals, respectively. Arrows indicate cytidine vibrational modes. (From Y.-C. Shen et al.21)

Raman spectroscopy has also resolved 14 vibrational modes of cytidine below 20 THz.37 However, the THz and Raman spectral data are, in general, complementary. Depending on the nature of the vibration, which is determined by the symmetry of the molecule, vibrations may be active or forbidden in the IR (THz) or Raman spectra. Neutron inelastic scattering (NIS) spectroscopy, on the other hand, is in principle sensitive to all vibrational motions, and therefore should be able to reveal all vibrational modes of cytidine. However, only 12 vibrational modes below 20 THz were resolved in NIS measurements.33,37 This is partly a result of the limited spectral resolution (~1 THz ≡ 33 cm–1) achieved in the measurements. In addition, NIS spectroscopy is most sensitive to vibrational modes involving hydrogen displacements, owing to the high cross section of the hydrogen atoms in the NIS process. As a result, vibrational modes arising from heavy atom (C, N, O, and P) vibrations, which can be probed by optical spectroscopy (Raman scattering and IR absorption), may not be resolved in NIS measurements. 5. Temperature-Dependent Low-Frequency Vibrational Spectra of Purine and Adenine In this section, we present temperature-dependent measurements of the farIR vibrational spectra of purine and adenine, using THz time-domain spectroscopy.38 The evolution of the absorption features is continuously mapped from 4 to 290K, and an empirical expression is presented describing the observed frequency shifts as a function of temperature.

100

A.G. DAVIES AND E.H. LINFIELD

Samples were prepared by mixing purine or adenine polycrystalline powder with polyethylene powder in a mass ratio of 1:10 and then compressing the mixture into a ~1.3-mm-thick pellet in a copper ring of 8-mm diameter. The copper ring ensures adequate thermal contact whilst allowing the THz beam to pass through the sample. The sample was fixed via the copper ring into the cold finger of a cryostat equipped with Mylar windows (MicrostatHe, Oxford Instruments) for THz spectroscopy measurements. A calibrated rhodium–iron resistance thermometer mounted adjacent to the sample on the cold finger was used to monitor the sample temperature. All chemicals were purchased from Sigma-Aldrich and used without further purification. Adenine (Lot 96H040625) and purine (Lot 27114-079) have molecular weights of 135.10 and 120.12, respectively. Both had a purity of 99%. X-ray diffraction measurements (Philips X-Ray Diffractometer PW1050)

Figure 6. Absorption spectra of (a) purine and (b) adenine at 4K and 295K. The 4K spectra are vertically offset. (From Y.-C. Shen et al. 38 )

Figure 7. (a) The evolution of the vibrational modes of purine in the temperature range 4–290K. A total of 120 spectra were measured. For clarity, only spectral data in the 1.0–2.3 THz range are shown. (b) Absorption spectra of purine at 4, 54, 105, 153, 204, 253, and 295K (from top to bottom). The spectra are vertically offset and the dotted lines are guides to the eye. (From Y.-C. Shen et al. 39 )

MOLECULAR AND ORGANIC INTERACTIONS

101

revealed that both the adenine and purine samples were polycrystalline. The polyethylene (Lot 17410AO-061) had a spectrophotometric grade purity and was amorphous. It was almost transparent in the frequency range 0.2–3.0 THz.39,40 In addition, the grain size of these powder samples was much smaller than the THz wavelength, minimizing the effects of Mie scattering. The THz time-domain spectroscopy apparatus was as discussed in Section 2, but in this case, a biased semi-insulating GaAs PC emitter was used for THz generation and a 1-mm-thick ZnTe crystal for EO detection. The variable delay stage was scanned over a distance of 10 mm, providing a spectral resolution of 15 GHz (0.5 cm–1). Figure 6(a) shows the measured absorption spectra of purine at room temperature and at 4K. A number of well-resolved absorption peaks are seen in the far-IR region between 0.2 and 3.0 THz. In addition, the baseline increases with frequency. As the temperature is reduced from room temperature to 4K, all of the observed absorption bands resolve into narrower peaks and most shift towards higher frequencies. In order to see if these effects are specific to purine, the spectra of samples of four nucleic bases (adenine, guanine, cytosine, and thymine), prepared in an identical manner, were examined. In each case, a different series of spectral features was observed, but in general, as with purine, the frequencies and intensities of the features increased as the temperature was lowered. As an example, the spectra of adenine are shown in Figure 6(b). The small-amplitude, ripple-like oscillations in the spectra result from multiple THz reflections in the sample and the detection crystal. By subtracting the weak features arising from reflections out of the raw time-domain data prior to the Fourier transform, these oscillations are minimized. In order to study the temperature dependence of the vibrational spectra quantitatively, 120 spectra of purine were recorded over the temperature range 4–290 K during a 2-h warming-up process (Figure 7). Figure 8 shows the frequency variation of the purine vibrational mode centred at 1.68 THz (at 4K) as a function of temperature. The peak stays at a fixed frequency until 80K, whereupon it progressively shifts to lower frequency approximately linearly with temperature. Since these measurements were made on polycrystalline samples, rather than non-interacting molecules in, for example, the gas phase, one would expect to see extensive intermolecular as well as intramolecular vibrations in this frequency range owing to the hydrogen-bonded networks. The importance of the intermolecular bonding was confirmed by molecular modeling based on density functional theory (DFT) calculations.38 DFT

102

A.G. DAVIES AND E.H. LINFIELD

calculations performed on a single adenine molecule and an adenine molecular dimer indicate that all vibrational modes in the frequency range investigated experimentally are a result of intermolecular interactions. Furthermore, DFT calculations on larger molecular clusters (tetramers etc.) suggest that these vibrational modes are non-localized and are of a collective (phonon-like) nature.

Figure 8. Temperature dependence of the resonance frequency centred at 1.68 THz (at 4K). The open circles are experimental data and the solid line is calculated using the empirical expression given in Eq. (3). The inset shows the best-fitting parameters in units of THz, K, and 10–4 THz/K for ⌡0, TC and A, respectively. (From Y.-C. Shen et al. 38 )

If all the contributions to the frequency shift are mediated by phonons, one might expect that the temperature dependence of the vibrational modes will follow a Bose-Einstein distribution.41 The frequency shift for each of the observed absorption features was fitted by:

υ (T ) = υ 0 − ATC (eT

c

/T

−1) ,

(3)

where υ0 is the centre frequency of the vibration mode at 0 K and A is a constant. TC is a characteristic temperature related to the energy of the mode. Figure 8 shows both the experimental and best-fit results of the purine peak centred at 1.68 THz at 4K. The agreement is reasonably good over the entire temperature range, and the fitting parameters are shown in the inset. The temperature-induced shifts for the purine peaks at 1.90 THz and 2.35 THz (measured at 4K) have been measured and fitted in the same way, and the best-fit parameters are also listed in Figure 8 inset. It is important to note that the characteristic temperatures (TC ) for the 1.08, 1.90, and 2.35 THz peaks do not scale with their frequency. In addition, the centre frequencies of two specific purine vibrational modes (1.11 THz and 1.43 THz) remain

MOLECULAR AND ORGANIC INTERACTIONS

103

almost unchanged with temperature, although the intensities of all vibrational modes increase as the temperature is reduced. These observations indicate that a number of mechanisms are likely to be involved in interpreting the temperature dependence of the vibrational spectra of these samples in this low-frequency range. For low-frequency vibrational modes, the ground-state vibrational levels are significantly depopulated at room temperature (290K ≡ 6 THz). In addition, the vibrational potentials are anharmonic, and therefore the spacing between adjacent energy levels decreases for higher energy vibrational states.8 This anharmonicity might explain the temperature dependence of the frequency shift observed in these experiments. At higher temperatures, more molecules are in higher vibrational excited states, and hence the decreased energy spacing results in the overall absorption envelope for the ensemble being shifted to lower frequency. As the temperature is lowered and more molecules populate the ground state, the average frequency of the absorption envelope would tend to increase. In addition, the distribution of the states should be narrower at lower temperature, which agrees well with the experimental results, since the absorption features sharpen as the temperature is reduced. The different behaviour seen at 1.11 and 1.43 THz possibly results from the complex nature of the hydrogen-bonded solids studied here. For example, the unit-cell volume of hydrogen-bonded solids may either increase or decrease with temperature.42 In addition, the average hydrogenbond strength may decrease with increasing temperature,43,44 as can be inferred from the observation of the red shift of the hydrogen-bond (O– H…O) stretching frequency in the far-IR and low-frequency Raman spectra of water as the temperature is increased.45,46 These effects change the shape of the potential energy curves and hence the vibrational energy levels in the hydrogen-bonded system, and this could lead to either an increase or decrease in peak position with changing temperature, or even leave it unaltered. 6. Conclusion We have discussed the generation of broadband THz radiation from an LTGaAs PC emitter, and the use of this powerful source for ultra-broadband THz spectroscopy. This extended frequency range is immediately useful for time-resolved THz spectroscopy, and provides important information on the vibrational modes arising from both intramolecular and intermolecular interactions. We also note that the time-resolved nature of this system

104

A.G. DAVIES AND E.H. LINFIELD

provides a unique tool for pump-probe studies of the dynamical properties of materials in the mid-IR and far-IR frequency ranges. We have also reviewed the use of LT-GaAs antennae for both generating and detecting ultra-broadband THz radiation. The results show that such a PC-generation/PC-detection scheme leads to a smooth spectral distribution up to 8 THz and provides better signal-to-noise-ratio compared with both EO-generation/PC-detection and PC-generation/EO-detection schemes. It thus makes an ideal system for THz time-domain spectroscopy in the frequency range 0.3–7.5 THz for a number of applications. Finally, we discussed temperature-dependent measurements on the farIR vibrational spectra of hydrogen-bonded purine and adenine, by THz time-domain spectroscopy. A number of vibrational modes were found to shift to higher intensities and frequencies as the temperature was reduced, and these were modelled by an empirical formula. Theoretical research will be needed to interpret fully the behaviour observed. Acknowledgements This work was supported by the EPSRC (UK), the Research Councils UK (Basic Technology Programme), the Royal Society, the EC Framework V programme “TeraNova”, Toshiba Research Europe Ltd., and the Association of Commonwealth Universities. We are very grateful to P. C. Upadhya, Y.–C. Shen, I. S. Gregory, C. Baker, H. E. Beere, W. R. Tribe, and M. J. Evans for helpful discussions.

References 1. M. C. Nuss and J. Orenstein, in Millimeter and Submillimeter Wave Spectroscopy of Solids, G. Grüner, Ed. (Springer, Berlin, 1998), and references therein. 2. B. Ferguson and X.-C. Zhang, Nat. Mater., 1, 26 (2002). 3. M. C. Beard, G. M. Turner, and C. A. Schmuttenmaer, J. Phys. Chem. B, 106, 7146 (2002). 4. P. Y. Han, M. Tani, M. Usami, S. Kono, R. Kersting, and X.-C. Zhang, J. App. Phys., 89, 2357 (2001). 5. B. Fischer, M. Walther, and P. Uhd Jepsen, Proceedings of 2002 IEEE Tenth International Conference on Terahertz Electronics, J. M. Chamberlain, A. G. Davies, P. Harrison, E. H. Linfield, R. E. Miles, and S. Withington, Eds. (IEEE, Piscataway, USA), vol. 74 (2002). 6. D. M. Leitner, Phys. Rev. Lett., 87, 188102 (2001). 7. L. Y. Zhu, J. T. Sage, and P. M. Champion, Science, 266, 629 (1994).

MOLECULAR AND ORGANIC INTERACTIONS

105

8. A. Xie, Q. He, L. Miller, B. Sclavi, and M. R. Chance, Biopolymers, 49, 591 (1999). 9. R. H. Garrett and C. M. Grisham, Biochemistry Chapter 1 (Saunders College Publishing, Fort Worth, 1999). 10. Q. Wu and X.-C. Zhang, Appl. Phys. Lett., 71, 1285 (1997). 11. R. A. Kaindl, F. Eickemeyer, M. Woerner, and T. Elsaesser, Appl. Phys. Lett., 75, 1060 (1990). 12. T. Dekorsy, H. Auer, H. J. Bakker, H. G. Roskos, and H. Kurz, Phys. Rev. B, 53, 4005 (1996). 13. M. B. Johnston, D. M. Whittaker, A. Corchia, A. G. Davies, and E. H. Linfield, Phys. Rev. B, 65, 165301 (2002). 14. D. Krokel, D. Grischkowsky, and M. B. Ketchen, Appl. Phys. Lett., 54, 1046 (1989). 15. U. D. Keil and D. R. Dykaar, Appl. Phys. Lett., 61, 1504 (1992). 16. N. Katzenellenbogen and D. Grischkowsky, Appl. Phys. Lett., 58, 222 (1991). 17. I. Brener, D. Dykaar, A. Frommer, L. N. Pfeiffer, J. Lopata, J. Wynn, K. West, and M. C. Nuss, Opt. Lett., 21, 1924 (1996). 18. G. Zhao, R. N. Schouten, N. van der Valk, W. Th. Wenckebach, and P. C. M. Planken, Rev. Sci. Instrum., 73, 1715 (2002); 47, 3699 (2002). 19. E. Sano and T. Shibata, Appl. Phys. Lett., 55, 2748 (1989). 20. K. Zhang and D. L. Miller, J. Electron. Mater., 22, 1433 (1993). 21. Y.–C. Shen, P. Upadhya, E. H. Linfield, H. E. Beere, and A. G. Davies, Appl. Phys. Lett., 83, 3117 (2003). 22. Q. Wu and X.-C. Zhang, Appl. Phys. Lett., 70, 1784 (1997). 23. A. Leitenstorfer, S. Hunsche, J. Shah, M. C. Nuss, and W. H. Knox, Appl. Phys. Lett., 74, 1516 (1999). 24. M. Tani, R. Fukasawa, H. Abe, K. Sakai, and S. Nakashima, J. Appl. Phys., 83, 2473 (1998). 25. A. Gruger, A. Regis, T. Schmatko, and P. Colomban, Vib. Spectrosc., 26, 215 (2001). 26. R. Huber, A. Brodschelm, F. Tauser, and A. Leitenstorfer, Appl. Phys. Lett., 76, 3191 (2000). 27. A. Leitenstorfer, S. Hunsche, J. Shah, M. C. Nuss, and W. H. Knox, Phys. Rev. Lett., 82, 5140 (1999); Phys. Rev. B, 61, 16642 (2000). 28. Y.-C. Shen, P. C. Upadhya, H. E. Beere, E. H. Linfield, A. G. Davies, I. S. Gregory, C. Baker, W. R. Tribe, and M. J. Evans, Appl. Phys. Lett., 85, 164 (2004). 29. I. S. Gregory, C. Baker, W. R. Tribe, M. J. Evans, H. E. Beere, E. H. Linfield, A. G. Davies, and M. Missous, Appl. Phys. Lett., 83, 4199 (2003). 30. S. Kono, M. Tani, and K. Sakai, Appl. Phys. Lett., 79, 898 (2001); Appl. Phys. Lett., 77, 4040 (2000). 31. P. C. Upadhya, Y.-C. Shen, A. G. Davies, and E. H. Linfield, Vib. Spectrosc., 35, 139 (2004). 32. M. Dauchez, P. Lagant, P. Derremaux, G. Vergoten, M. Sekkal, and B. Sombret, Spectrochim. Acta 50A, 105 (1994). 33. M. Plazanet, N. Fukushima, and M. R. Johnson, Chem. Phys., 280, 53 (2002). 34. B. Fischer, M. Malther, and P. Uhd Jepsen, Phys. Med. Biol., 47, 3807 (2002). 35. M. Rozenberg, C. Jung, and G. Shoham, Phys. Chem. Chem. Phys., 5, 1533 (2003). 36. N. Leulliot, M. Ghomi, H. Jobic, O. Bouloussa, V. Baumruk, and C. Coulombeau, J. Phys. Chem. B, 103, 10934 (1999). 37. M. Mathlouthi, A. M. Seuvre, and J. L. Koenig, Carbohydr. Res., 146, 1 (1986).

106

A.G. DAVIES AND E.H. LINFIELD

38. Y.-C. Shen, P. Upadhya, A. G. Davies, and E. H. Linfield, Appl. Phys. Lett., 82, 2350 (2003). 39. M. Walther, B. Fischer, M. Schall, H. Helm, and P. Uhd Jepsen, Chem. Phys. Lett., 332, 389 (2000). 40. M. Walther, P. Plochocka, B. Fischer, H. Helm, and P. Uhd Jepsen, Biopolymers, 67, 310 (2002). 41. L. Vina, S. Logothetides, and M. Cardona, Phys. Rev. B, 30, 1979 (1984). 42. M. Rozenberg, A. Loewenschuss, and Y. Marcus, Chem. Phys., 2, 2699 (2000). 43. S. Woutersen, U. Emmerichs, H.-K. Nienhuys, and H. J. Bakker, Phys. Rev. Lett., 81, 1106 (1998). 44. P. Acharya and J. Chattopadhyaya, J. Org. Chem., 67, 1852 (2002). 45. H. R. Zelsmann, J. Mol. Struct., 350, 95 (1995). 46. K. Mizoguchi, Y. Hori, and Y. Tominaga, J. Chem. Phys., 97, 1961 (1992).

TERAHERTZ BEAM INTERACTIONS WITH AMORPHOUS MATERIALS MIRA NAFTALY AND ROBERT E. MILES* Institute of Microwaves and Photonics, University of Leeds, Leeds, LS2 9JT, UK

Abstract. Terahertz (THz) time-domain spectroscopy (TDS) is used to study two types of amorphous materials: glasses and polymers. The theory of far-infrared (IR) absorption in amorphous materials is used to analyse the results, and to understand the differences in THz absorption among the sample materials. A family of related borosilicate glasses has been examined along with silica glass, and their THz absorption coefficients and refractive indices are compared. Two chalcogenide glasses are also studied. Three types of polymer plates have been examined, and their THz transmission properties are compared with those of glasses. Polymerisation in SU8 films has been studied by exposing samples to UV for different lengths of time and comparing their THz transmission properties.

Keywords: terahertz spectra, amorphous materials, glasses, polymers, paper

1. Introduction In the preceding chapter, Davies and Linfield describe how materials such as explosives and drugs of abuse can be detected and identified from their THz spectra, where the distinct features observed are a consequence of the molecular structure of the material in question. When in transit, these suspect substances are likely to be found in some form of container made of an amorphous material such as glass, plastic, or paper. The THz spectrum of the amorphous enveloping material is quite different to that of the illegal substance but, in most cases, the spectrum will be obtained by looking through the container. It is therefore important to determine the effect of the

______ *

To whom correspondence should be addressed: R E Miles, Institute of Microwaves and Photonics, School of Electronic and Electrical Engineering, University of Leeds, Leeds LS2 9JT, UK, e-mail: [email protected] 107 R. E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 107–122. © 2007 Springer.

108

M. NAFTALY AND R.E. MILES

container on the telltale THz spectrum. For this reason, TDS using THz radiation [1–3] is currently being employed to characterise amorphous materials. In this chapter we examine in detail two types of container materials: glasses and polymers. These are compared with paper, which has been discussed in more detail elsewhere [4]. Far-IR transmission in both glasses and plastics has in the past been studied using FTS [4–10]; however, THz TDS has significant advantages which generate better quality data, and therefore enable a more detailed analysis to be carried out. The most important of these is that THz TDS yields both the absorption coefficient and the refractive index of the material, unlike FTS, which measures only absorption. Furthermore, TDS has a very high dynamic range: typically >106 in power, compared to ~103 for FTS [11], thus allowing transmission studies to be carried out on materials with relatively high absorption coefficients. Far-IR transmission in amorphous materials, such glasses and polymers, is of particular interest, because absorption is mediated by phonon processes, and therefore is related to a number of material properties. A model developed by Schlömann [13] and Strom in the 1970s [5, 6] explains far-IR absorption in amorphous materials in terms of disorder-induced coupling of radiation into the acoustic phonon modes of the material. The model, which applies at frequencies (ν) such that 2πν >> VD/l, where VD is the Debye velocity of sound and l is the average correlation length in the material, states that the frequency-dependent absorption coefficient α(ν) can be described by the power-law relation:

n(ν ) α (ν ) = K ν β ,

(1)

where n(ν) is the frequency-dependent refractive index, and the exponent for glassy materials is β ≈ 2. The coefficient K is determined by a number of material properties, and is given by: K =

4π 2 q 2 N κ 2 , ρ c VD3

(2)

where N is the density of charge fluctuations of amplitude q, ρ is the mass density, c is the speed of light, and κ = (n2+2)/3 is the local field correction. Notably, Eq. (2) indicates that K increases linearly with the density of charge fluctuations, and is therefore a measure of disorder in the material.

INTERACTIONS WITH AMORPHOUS MATERIALS

109

2. Experimental The THz TDS system uses a 60-fs Ti-sapphire laser (Spectra-Physics Tsunami) in a conventional configuration [14] for free-space THz generation and detection. The average power of the laser is 1.2 W, of which approximately 1.1 W is used for THz generation from a biased GaAs emitter. Electro-optic (EO) detection with balanced photodiodes is employed to observe the THz signal. The dynamic range of the THz spectroscopy system is around 2,000 in amplitude, and the usable bandwidth is from 100 GHz to 3 THz. The glass samples were commercial optical flats (Schott Glass, from UQG Optics) with the thicknesses between 1 and 2 mm. One of the glasses was fused silica; the rest were different compositions in the BK7 borosilicate family. The reason for choosing this range of glasses was to investigate a family of similar compositions and to relate the differences in THz transmission to the glass properties; and further, to compare THz transmission in these multicomponent glasses with that in pure silica. Although the glass compositions are proprietary and therefore not available, the glass properties are provided in data sheets. The range of study was further extended by looking at two chalcogenide glasses (provided by P. Mason of Qinetiq): arsenic sulphide (As2S3) and GAST (Ge20As10Se30Te30). These glass samples were also produced as optical flats of 1.1-mm thickness, similar to the commercial samples described earlier. Polymer plates of high-density polyethylene (HDPE), TPX (TPX is a 4-methylpentene-1-based polyolefin manufactured by Mitsui Chemicals, Inc.), polytetrafluoroethane (PTFE), polystyrene, Delrin (Delrin, or polyoxymetylene, is an acetal resin manufactured by DuPont), polycarbonate, and Perspex (Perspex is a trade name for polymethyl methacrylate) were obtained from commercial suppliers. SU8 is a three-component epoxy resin widely used as an ultra-thick negative photoresist [15]. The resin is deposited by spin-coating onto a substrate, and is then cured by exposure to UV light, causing formation of cross-linkages. SU8 samples were prepared by spin-coating onto a polystyrene substrate. Polystyrene was chosen because it has good THz transparency [9, 10] and is available as optically smooth, flat plates of uniform 4-mm thickness. The spin-coated SU8 layers had varying thicknesses of between 0.6and 1.2 mm, owing to the difficulty of controlling the coating process. One sample was left uncured, while the others were UV-cured for varying lengths of time using standard semiconductor photolithography exposure equipment. The THz transmission properties of all samples were then measured.

110

M. NAFTALY AND R.E. MILES

The THz transmission spectra and phase shifts are obtained from the measured time-domain data of THz electric field using a standard FFT application (Origin 7.5). Absorption coefficients (α) and refractive indices (n) of the samples are then calculated using the equations:-

α (ν ) = −

2 d sample

n(ν ) = 1 +

⎡

ln ⎢

E

sample

⎣T E

⎤ ⎥ (ν ) ⎦

(ν )

reference

c [φsample (ν ) − φreference (ν )] 2πν d sample

(3)

(4)

2

⎛n −1 ⎞ T = 1 − ⎜ average ⎟ , ⎜n ⎟ ⎝ average + 1 ⎠

(5)

where E(ν) and φ(ν) are the amplitude and phase of the THz field at the frequencyν, and dsample is the sample thickness; T is the fraction of power transmitted through the air–material interface. 3. Results and Discussion 3.1. GLASSES

Figures 1 and 2 show the absorption coefficients and refractive indices of the borosilicate glasses studied, labelled by their trade name designations. The curves for silica are also shown for comparison. The measured frequency range differs among the glasses because the dynamic range of the TDS system limits the accessible frequencies [16]. Both the dynamic range and the absorption coefficient are strongly frequency-dependent [16], thus more strongly absorbing glasses have a narrower data range. Oscillations seen at low frequencies are an artefact arising from secondary reflection signals in thin samples. In Figures 1 and 2 clear differences are observed for the different glass types. As seen in Figure 1, THz absorption is much stronger in borosilicate glasses than in silica. In simple terms, the increased THz absorption can be understood as arising from the presence in multicomponent glasses of ionic network modifiers [17], especially alkali oxides, which increase the microscopic polarisability of the glass. In addition, referring to Eq. (1), stronger

INTERACTIONS WITH AMORPHOUS MATERIALS

111

absorption can be explained by the higher degree of disorder in multicomponent glasses. In Figure 2, only weak dispersion is observed in the THz refractive indices of all glass samples, and the values are above those observed in the visible range (Figure 5). The weak dispersion may be explained by the fact that THz frequencies lie far away from the bandgap or resonances of these materials. The higher refractive indices at THz compared with the visible are consistent with the general trend of increasing dielectric constant at lower frequencies [18]. In order to determine the absorption parameters K and β, the product of absorption coefficient and refractive index was fitted to the allometric function of the form given by Eq. (1). An example of such a fit for silica glass is shown in Figure 3. The good fit between the data and the model, along with the large number of points in the measured range, demonstrates the advantages of THz TDS in obtaining low-noise data, as compared with the previous FTS techniques [5–8]. The calculated parameters K and β are listed for all glasses in Table 1. The values of K and β obtained for silica are similar to those quoted by Strom [5, 6]. In considering Figures 1 and 2, it is notable that THz absorption is higher in those glasses which also have a high THz refractive index. Since all these glasses belong to a single borosilicate family of glass compositions, it is meaningful to examine this trend further. In order to do this, Figure 4 plots the absorption parameter K vs. the refractive index at 0.5 THz (chosen because it is close to the mean value over the THz bandwidth measured). It is seen in Figure 4 that K does indeed increase with the refractive index.

Absorption coefficient (mm-1)

8 SF15 NSF10 2K7 BK7 B270 SK10 SiO2

7 6 5 4 3 2 1 0 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

THz

Figure 1. THz absorption coefficients of borosilicate glasses and silica.

M. NAFTALY AND R.E. MILES

112

3.4

SF15 NSF10 2K7 BK7 B270 SK10 SiO2

3.2

Refractive index

3.0 2.8 2.6 2.4 2.2 2.0 1.8 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

THz

Figure 2. THz refractive indices of borosilicate glasses and silica.

RI x absorption coefficient (mm-1)

2.5

SiO2

2.0

1.5

1.0

0.5

0.0 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 3. The product of THz absorption coefficient and refractive index in silica showing the fit to Eq. (1).

A simple argument based on Eq. (2) suggests that K should increase as the fourth power of the refractive index, owing to the local field correction factor. For this reason, a fit to a 4th-degree polynomial is shown by a dotted line in Figure 4. However, other factors in Eq. (2) will also affect K, which would explain the poor fit between the data and the polynomial model. In particular, K may be expected to increase with the number of dangling bonds in the glass structure, i.e. increased q2N. The refractive index should also increase with the number of ionic dangling bonds owing to their weaker bonding strength.

INTERACTIONS WITH AMORPHOUS MATERIALS

113

Furthermore, it may be expected that the refractive index at all frequentcies should change with the glass composition at a similar rate. In other words, all glasses in a compositional family should exhibit the same dispersion, resulting in a linear relationship between the refractive index values at THz and in the visible. This indeed is shown in Figure 5, which plots the refractive index at 0.5 THz against that at 589 nm (the sodium D line, nD). 70 60

2

K (s /mm)

50 40 30 20 10

SiO2

0 1.8

2.0

2.2

2.4

2.6

2.8

3.0

3.2

Refractive index @ 0.5 THz

Figure 4. The calculated parameter K vs. the refractive index at 0.5 THz for the borosilicate glasses and silica. 3.4

Refractive index @ 0.5 THz

3.2 3.0 2.8 2.6 2.4 2.2 2.0 1.8 1.40

SiO2

1.45

1.50

1.55

1.60

1.65

1.70

1.75

1.80

Refractive index @ 589 nm

Figure 5. The refractive index at 0.5 THz vs. that at 589 nm for the borosilicate glasses and silica.

It is also of interest to consider the relationship between the parameter K and another glass property. The glass transition temperature Tg is the temperature at which glass undergoes a phase transition from the liquid to

M. NAFTALY AND R.E. MILES

114

the glassy state. The glass transition occurs at higher temperatures for the strongly bonded, more covalent glasses with a longer-range order [19]. Tg tends to decrease in weakly bonded, more ionic glasses with more disordered networks and numerous dangling bonds. According to Eq. (2) earlier, this is also true of the parameter K. To explore this relationship, Figure 6 plots K vs. Tg for the borosilicate glasses and silica. It is seen that, as expected, K decreases sharply as Tg increases. Since Tg is an important parameter in characterising many aspects of glass behaviour, this relationship with THz absorption is of particular significance. Another area of interest is to apply THz TDS to the study of “exotic” glasses, such as the chalcogenide-based materials, and to compare them with silica. The THz absorption coefficients and refractive indices of two such glasses, As2S3 and GAST, are shown in Figures 7 and 8. Chalcogenide glasses have a structure which is different from that of silica, and their bonding is weaker. As2S3 is, like silica, a binary glass; while GAST is a multicomponent glass. The values of the refractive index and β in As2S3 glass are similar to those reported by Onari et al. [7]; however, the value of absorption at 1 THz measured here is approximately 40% higher. This may be due to the differences in the fabrication process of the glass samples. 70 60

2

K (s /mm)

50 40 30 20 SiO2

10 0 400

500

600

700

800

900 1000 1100 1200

Glass transition temperature Tg °C

Figure 6. The calculated parameter K vs. the glass transition temperature Tg for the borosilicate glasses and silica.

It is seen in Figure 7 that the binary As2S3 glass has weaker absorption than the multicomponent GAST glass. However, these two glasses have different components and structures, and therefore are not directly comparable. It is also seen that both chalcogenide glasses have weaker absorption

INTERACTIONS WITH AMORPHOUS MATERIALS

115

than the borosilicates, which may be explained by the absence of ionically bonded compounds in their compositions. Nevertheless, As2S3 has much stronger absorption than SiO2, which may be ascribed to its weaker bonding and more disordered structure [7]. With regard to the refractive indices in Figure 8, it is seen that, as with the borosilicate glasses, the dispersion is very low and the glasses with the higher refractive index also exhibit stronger absorption. This supports the explanation that, other factors being similar, the parameter K increases approximately as the square of the refractive index. 10

Absorption coefficient (mm-1)

SiO2

8

As2O3 GAST

6

4

2

0 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 7. THz absorption coefficients of the chalcogenide glasses and silica.

Refractive index

3.8 3.6

SiO2

3.4

As2O3

3.2

GAST

3.0 2.8 2.6 2.4 2.2 2.0 1.8 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 8. THz refractive indices of the chalcogenide glasses and silica.

M. NAFTALY AND R.E. MILES

116 3.2. POLYMERS

THz TDS is particularly suited to the study of polymers, owing to their good transparency at THz frequencies. We looked at THz transmission in seven types of polymer plates: HDPE, TPX, PTFE, polystyrene, Delrin, polycarbonate, and Perspex. Of these, HDPE, TPX, PTFE, and polystyrene are all highly transparent to THz beams, and as such are useful materials for substrates, cells and fillers employed in THz spectroscopy. Figure 9 shows the absorption coefficients of the polymer sheets studied; while Figure 10 shows their refractive indices. Of the polymers studied, HDPE, PTFE, and Delrin have a milky white appearance in the visible, while the rest are transparent and colourless. It is seen that, as expected, transparency at THz is not correlated with that in the visible. The values of absorption and refractive indices of HDPE, TPX, and PTFE are similar to those reported by Birch et al. [10]; however, the values for polystyrene are significantly different, which may be attributable to a different composition or additives in the material. The product of refractive index and absorption coefficient was fitted to Eq. (1) for all samples, and the values of K and β are listed in Table 1. It is of interest that the value of β is below that of glasses (~2) for TPX, polystyrene, and polycarbonate, indicating their non-glassy structure. It is also of interest that in the case of these polymers, absorption strength is not correlated with the refractive index, unlike the case in glasses.

-1

Absorption coefficient (mm )

2.0

1.5

HDPE TPX PTFE polystyrene Delrin polycarbonate Perspex

1.0

0.5

0.0 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 9. THz absorption coefficients of polymer sheets.

Considering the refractive indices in Figure 10, it is seen that Delrin, polycarbonate, and Perspex, all have significant dispersion; whilst the other

INTERACTIONS WITH AMORPHOUS MATERIALS

117

polymers have weak dispersion similar to glasses. The three polymers with substantial dispersion are also those with the strongest THz absorption. Moreover, as discussed in the following section below, SU8 is also characterised by both strong dispersion and strong absorption. This indicates that there may be a relationship between THz dispersion and absorption in polymers. 2.2 2.1 HDPE TPX PTFE polystyrene Delrin polycarbonate Perspex

Refractive index

2.0 1.9 1.8 1.7 1.6 1.5 1.4 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 10. THz refractive indices of polymer sheets. TABLE 1. Materials and the calculated values of K and β parameters Type Glass

Polymer

SU8

Material SiO2 SF15 NSF10 2K7 BK7 B270 SK10 As2S3 GAST HDPE TPX PTFE Polystyrene Delrin Polycarbonate Perspex Unpolymerised SU8 Polymerised SU8

K, s2/mm 0.4 41 66 18 20 22 29 7.0 12 0.02 0.1 0.2 0.3 1.5 1.9 2.3 4.0 2.6

β 2.0 2.7 3.1 2.9 2.2 2.8 3.7 1.9 1.7 1.9 1.3 1.7 1.0 1.7 1.3 1.9 1.0 0.9

118

M. NAFTALY AND R.E. MILES

3.3. POLYMERISATION IN SU8

Fully polymerised films of SU8 have been studied using THz TDS [15]; the interest in this material being due to its fairly low loss at THz frequencies and to its widespread use in microelectronic fabrication. Here we examine the differences in THz transmission between polymerised and unpolymerised SU8 and also look at the rate of the polymerisation process. Figures 11 and 12 show the absorption coefficients and refractive indices of unpolymerised and polymerised SU8 layers which have been exposed to UV for 3 min. It is seen that both absorption and refractive index are higher in unpolymerised material. As with glasses, this may be attributed to the presence of dangling bonds in unpolymerised material; these bonds disappear when cross-linkages are formed during polymerisation. The large difference in the values of absorption and refractive index between polymerised and unpolymerised material indicates that THz TDS may be of use in determining the degree of polymerisation in SU8. Two more features are of interest. The product of the absorption coefficient and refractive index for SU8 is nearly linear with frequency (β≈1, see Table 1). This is similar to some of the polymers discussed earlier, but unlike glasses. In addition, the refractive index exhibits a strong negative dispersion, again similarly to some polymers, but unlike glasses. Since the SU8 samples had varying thicknesses, it was not possible to compare directly THz transmission in samples exposed to UV for different lengths of time. Therefore in order to analyse the transmission data, an approximate model was employed, which may be termed the “polymerisation front” model. According to this description, during UV exposure the light is mostly absorbed in a thin layer of SU8. As this layer is polymerised, it becomes transparent to UV; thereby allowing the layer underneath to be exposed. Thus, a polymerisation front moves downwards through the SU8 film at a constant rate, until the whole of the film is fully polymerised. During this process, the SU8 film contains a time-varying proportion of polymerised and unpolymerised material. The measured refractive index and absorption coefficient of partially polymerised films are then given by:

d n(t ) = d p n p + du nu d α (t ) = d pα p + duα u d = d p + du

(6)

INTERACTIONS WITH AMORPHOUS MATERIALS

119

5

-1

Absorption coefficient (mm )

unpolymerised 4

3

2

polymerised

1

0 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 11. THz absorption coefficients of polymerised and unpolymerised SU8 layers. 1.95 unpolymerised Refractive index

1.90

1.85

1.80 polymerised 1.75

1.70 0.0

0.5

1.0

1.5

2.0

2.5

THz

Figure 12. THz refractive indices of polymerised and unpolymerised SU8 layers.

In Eq. (6), d is the thickness of the SU8 film, n is the refractive index, α is the absorption coefficient, and the subscripts denote polymerised and unpolymerised material. The parameters n(t) and α(t) are measured in samples that were UV-exposed for different lengths of time t. In Eqs. (6), the only unknown quantities are dp and du, which therefore can be calculated from two independent sets of data, that for n(t) and α(t). The result is shown in Figure 13, where the polymerisation depth dp is plotted against the UV-exposure time. The approximately linear increase of dp with time confirms the validity of the “polymerization front” description of the process.

M. NAFTALY AND R.E. MILES

120

1.0

Polymersisation depth (mm)

calculated from absorption coefficient calculated from refractive index

0.8

0.6

0.4

0.2

0.0 0

10

20

30

40

50

60

70

Time (s)

Figure 13. Polymerisation depth in SU8 as a function of UV-exposure time, calculated from refractive index and absorption data. 3.4. PAPER

As was shown earlier, the absorption spectra of amorphous materials such as glass and polymers have no outstanding features in the THz region. Therefore, if such materials should be employed as containers for illegal substances, then any features observed in the absorption spectrum of the package may be confidently ascribed to its contents. However, in contrast, paper materials have distinct absorption peaks which, moreover, vary with the type of paper. Figure 14 below shows examples of THz absorption in several types of paper. For more details see [4]. These spectra show that care must be taken when attempting to identify a suspect material contained in a paper or cardboard package.

Absorption coefficient (mm-1)

5

4

types of paper: pure cellulose, n=1.40 letterhead, n=1.49 glossy, n=1.60 thin card, n=1.46

3

2

1

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Frequency (THz)

Figure 14. THz absorption coefficients of different types of paper.

INTERACTIONS WITH AMORPHOUS MATERIALS

121

4. Conclusions THz TDS can be used to determine absorption coefficients and refractive indices in glasses, polymers, and papers. The data analysed for glasses and polymers is found to be in agreement with the theory of far-IR absorption in amorphous materials. (In the case of glasses, significant correlations were found with glass properties.) The absorption properties therefore behave in a predictable manner depending on the material and it is therefore possible to allow for the effect of absorption in the container when scanning for illegal substances. The effect of polymerisation on the THz spectra of SU8 films exposed to UV was also explored. Spectra for paper and cardboard show detail which must be accounted for in routine scanning for illegal substances. Acknowledgements This work was supported by the Research Councils (UK) Basic Technology Research Programme. Thanks also to Nik Dimitrakopoulos for preparing the SU8 samples and Dr. Paul Mason of Qinetic for supplying the exotic glasses.

References [1] [2] [3] [4]

[5]

[6]

P. Y. Han and X.-C. Zhang, Free-space coherent broadband terahertz time-domain spectroscopy, Meas. Sci. Technol., 13, 1747–1756 (2001). M. C. Beard, G. M. Turner, and C. A. Schmuttenmaer, Terahertz spectroscopy, J. Phys. Chem B, 106, 7146–7159 (2002). M. Hangyo, T. Nagashima, and S. Nashima, Spectroscopy by pulsed terahertz radiation, Meas. Sci. Technol., 13, 1727–1738 (2002). M. Naftaly, A. P. Foulds, R .E. Miles, and A. G. Davies, Terahertz transmission spectroscopy of nonpolar materials and relationship with composition and properties, Int J. Infrared and Millimeter Waves, 26, pp. 55–64 (2005). U. Strom, J. R. Hendrickson, R. J. Wagner, and P. C. Taylor, Disorder-induced far infrared absorption in amorphous materials, Solid State Commun., 15, 1871–1875 (1974). U. Strom and P. C. Taylor, Temperature and frequency dependences of the far-infrared and microwave optical absorption in amourphous materials, Phys. Rev. B, 16, 5512– 5522 (1977).

122 [7]

[8]

[9] [10] [11] [12]

[13] [14] [15]

[16] [17]

[18] [19]

M. NAFTALY AND R.E. MILES S. Onari, K. Matsuishi, and T. Arai, Far-infrared absorption spectra and the spatial fluctuation of charges on amorphous As-S and As-Se systems, J. Non-Crystal. Solids, 86, 22–32 (1986). S. A. FitzGerald, J. A. Campbell, and A. J. Sievers, Two-level systems and excitedstate transitions in fluorite mixed crystals and silica glass, Phys. Rev. Lett., 73, 3105– 3108. (1994). J. R. Birch, The far-infrared optical constants of polypropylene, PTFE and polystyrene, Infrared Phys., 33, 33–38 (1992). J. R. Birch, J. D. Dromey, and J. Lesurf, The optical constants of some common lowloss polymers between 4 and 40 cm−1, Infrared Phys., 21, 225–228 (1981). G. W. Chantry et al., Far infrared and millimeter-wave absorption spectra of some lowloss polymers, Chem. Phys. Lett., 10, 473–477 (1971). P. Y. Han, M. Tani, M. Usami, S. Kono, R. Kersting, and X.-C. Zhang, A direct comparison between terahertz time-domain spectroscopy and far-infrared Fourier transform spectroscopy, J. Appl. Phys., 89, 2357–2359 (2001). E. Schlömann, Dielectric losses in ionic crystals with disordered charge distributions, Phys. Rev., 135, 2A, pp. A413–A419 (1964). M. Naftaly and R. E. Miles: Terahertz time-domain spectroscopy: A new tool for the study of glasses in the far infrared, J. Non-Crystalline Solids, 351, 3341–3346 (2005). S. Arscott, F. Garet, P. Mounaix, L. Duvillaret, J.-L. Coutaz, and D. Lippens, Terahertz time-domain spectroscopy of films fabricated from SU-8, Electron. Lett., 35, 243–244 (1999). P. U. Jepsen and B. M. Fischer, Dynamic range in terahertz time-domain transmission and reflection spectroscopy, Opt. Lett., 30, 29–31 (2005). M. J. Jackson and B. Mills, Thermal expansion of alumino-alkalisilicate and aluminoborosilicate glasses – comparison of empirical models, J. Mater. Sci. Lett., 16, 1264– 1266 (1997). S. O. Kasap, Principles of electronic materials and devices, Chapter 7, 2nd edn. (McGraw-Hill, 2002). D. R. Uhlmann, Glass formation, J. Non-Crystal. Solids, 25, 43–85 (1977).

DEVELOPMENT OF TAGLESS BIOSENSORS FOR DETECTING THE PRESENCE OF PATHOGENS JING–YIN CHEN*, JOSEPH R. KNAB, SHUJI YE, YUNFEN HE, AND ANDREA G. MARKELZ State University Of New York at Buffalo, Department of Physics, 239 Fronczak Hall, Buffalo, NY, USA 14260

Abstract. The vibrational modes corresponding to protein tertiary structural motion lay in the far-infrared or terahertz (THz) frequency range. These collective large-scale motions depend on global structure and thus will necessarily be perturbed by ligand-binding events. We discuss the use of THz dielectric spectroscopy to measure these vibrational modes and the sensitivity of the technique to changes in protein conformation, oxidation state and environment. A challenge of applying this sensitivity as a spectroscopic assay for ligand binding is the sensitivity of the technique to both bulk water and water bound to the protein. This sensitivity can entirely obscure the signal from the protein or protein–ligand complex itself, thus necessitating sophisticated sample preparation making the technique impractical for industrial applications. We discuss methods to overcome this background and demonstrate how THz spectroscopy can be used to quickly assay protein binding for proteomics and pharmaceutical research.

Keywords: biosensors, THz spectroscopy, far-infrared, lysozyme, tri-N-acetyl-Dglucosamine, biomolecular sensing, ligand binding

1. Introduction A key component to pharmaceutical development is the determination of protein–protein and protein–ligand binding. Rapid optical assays to determine this binding could have a significant impact, however not all binding events will result in changes to UV/Vis absorbance or fluorescence since binding

______

* To whom correspondence should be sent: J.Y. Chen, State University of New York at Buffalo, Department of Physics, 239 Fronczak Hall, Buffalo, NY, USA 14260; e-mail: [email protected]

123 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 123–134. © 2007 Springer.

124

J.-Y. CHEN ET AL.

may occur at a site remote from the optically active region. In the case of mid-infrared vibrational spectroscopy, while binding at any given site will locally change the MIR vibrational modes involving on the order of 5–10 atoms, this local change may have little effect on the dense overlapping modes in the amide I and II regions. Global collective vibrational modes which involve large-scale structural motions on the other hand would be affected by any binding event within the molecule. These vibrational modes associated with protein tertiary structure lie in the far-infrared or THz frequency range (0.03–6 THz, 1–200 cm–1). Methods of measuring these low-frequency modes include inelastic neutron scattering1 (INS) and THz dielectric spectroscopy.2–5 Recently there have been several reports of sensing biomolecular binding using these techniques. For example, INS measurements demonstrated this binding sensitivity in the case of binding of methotrexate with dihydrofolate reductase which indicates an increase in the vibrational density of states at these frequencies.1 INS however is not a technique that is amenable to rapid parallel assays for drug discovery. Over the last 10 years, there has been a significant increase in spectroscopic research in the THz frequency range with applications from package monitoring to chemical sensing.6 This is due in large part to significant development in THz dielectric spectroscopy tools. Recently, Zhang et al. reported the sensitivity of THz dielectric response to biotin–avidin binding5 using a sophisticated high sensitivity measuring system and thin film samples. An overarching impediment to THz biosensing has been sample preparation. There are several concerns for biosensing and material characterization in general in this frequency regime. Among these concerns are multiple reflection artifacts and water background. Multiple reflection artifacts occur due to constructive/destructive interference effects within samples with thicknesses on the order of the probing wavelength. The structure of these artifacts while normally periodic, can due to multiple effects be falsely identified as spectroscopic features or inhibit proper determination of dielectric response. Multiple reflection effects can be avoided by (a) embedding the biomolecule in a matrix and preparing a sufficient thickness to reduce etalon for high resolution studies, (b) reducing resolution and having sufficient film nonuniformity or (c) numerical removal of multiple reflection effects. The water background is not so easily dealt with. Bulk water has a dielectric relaxation response with the relaxation time ~ 8.3 ps7 resulting in a peak in the absorbance at 1 THz. The absorption coefficient8 for bulk water at 1 THz is 200 cm–1. Our measurements3,4 of protein films suggest that a fully hydrated protein without bulk water present has an absorption

DEVELOPMENT OF TAGLESS BIOSENSORS

125

coefficient at 1 THz ~ 100 cm–1. In general, it is found that the transmission through a protein solution is in fact greater than the transmission through bulk water due to volume exclusion of the bulk water by the protein and the lower absorption coefficient of the protein itself. If the precise volume of the hydrated protein was known, one could in theory still extract the protein contribution and perform binding measurements in solution; however, this volume is not trivial to determine. There is not a clear distinction4 between water directly bound to the protein, the first hydration shell and the water immediately adjacent to this first hydration shell. Thus the precise definition of the hydrated protein volume is not known. Furthermore, this volume will likely change with binding, which can result in either a more compact or a more extended structure. Thus, it would appear that the number of unknowns is sufficient to eliminate using THz measurements of solutions as a method of binding determination. One method to remove concerns for water effects is to remove the water from the samples such as by using pressed pellets made from lyophilized powders or using dried thin films. Sample preparation for lyophilized powders is straight forward, but not easily scalable. In general, it involves lyophilization of the biomolecule and the bound biomolecular complex from solution, mixing of the powder into a THz transparent material, often polyethylene powder, and then pressing pellets of sufficient thickness to remove concern of multiple reflection effects. While effective, this method does not allow for significant parallel processing and rapid determination of protein–ligand interaction. Thin film preparation requires careful drying procedures from starting solutions to ensure sample uniformity. These films are often fragile, and the THz measurement is highly dependent on the relative humidity of the experimental apparatus, since the protein films have a high affinity for adsorbed water. Here, we consider a straight forward solution to the bulk-water problem, which is to simply freeze out the rotational relaxation modes of bulk water that give rise to the large absorbance at THz frequencies. Previously, Zhang et al. as well as others9 have shown that the absorption coefficient and index of refraction of water take a precipitous drop below freezing. The hydrogen-bonding network in the ice phase strongly inhibits water dipole response to the THz field. On the other hand, the global vibrations of the protein in the THz frequency remain active. Recently, we have exploited this effect to examine the so-called glass transition in proteins at 200K; a significant increase in far-infrared response was observed at 200K, and this response was easily distinguished from bulk water. Thus hydrated protein dynamics can be examined without concern for the bulk-water background.

126

J.-Y. CHEN ET AL.

Here, we use this result to see if one can then sense protein–ligand binding in triacetylglucosamine (3NAG)/hens egg white lysozyme (HEWL) mixtures as measured by THz time domain spectroscopy (THz-TDS). The measurements are performed below 273K, and, for the first time, demonstrate that the complex refractive index of 3NAG–HEWL is smaller than free HEWL. 2. Methodology Hen egg white lysozyme (HEWL) lyophilized power (Sigma Aldrich L6876) is dissolved in trizma buffer (pH 7.0, 0.05 M) and the final concentration of the lysozyme solution is 200 mg/ml. Tri-N-Acetyl-Dglucosamine (N, N′, N′′-Triacetylchitotriose; 3NAG) lyophilized power is purchased from Toronto Research Chemicals, Inc. (T735000). The HEWL+3NAG-binding solution is made by dissolving 3NAG powder into the 200 mg/ml lysozyme solution and the molar ratio of lysozyme to 3NAG is 1:1. The solution sample cell for the transmission measurements is composed of two brass plates sandwiching two quartz windows (Spectrocell Inc.) with a ~250 µm spacer.10 The top plate of the cell has two identical apertures respectively for the sample and reference. About 14 µl of the protein solution is pipetted to fill the bottom aperture and the upper aperture is left empty as a reference. The entire measurement was performed for two different sets of the lysozyme and HEWL+3NAG-binding solutions and results were found to be excellent agreement. The ligand binding for the solutions was verified by the fluorescence measurements before THz transmission measurements. The fluorescence measurements are taken by using the SLM8100 spectrofluorimeter. The excitation wavelength for both types of solutions is 280 nm. We found the fluorescence peak of lysozyme solutions is about 343 nm and that of HEWL+3NAG-binding solutions is 332 nm as expected for bound and unbound HEWL.11,12 The solution sample cell is mounted inside a continuous-flow cryostat (Cryo Industries). A silicon diode temperature sensor is located on the sample stick adjacent to the solution cell. Liquid nitrogen is used as the cryogen and the fluctuation at each temperature is lower than 0.01K. Each temperature measurement is taken at least 10 min after the temperature is stable. At each temperature, measurements were repeated more than three times in order to make sure that the sample reached the thermal equilibrium.

DEVELOPMENT OF TAGLESS BIOSENSORS

127

Terahertz time domain spectroscopy (THz-TDS) was used to determine the THz dielectric response. The THz electric field is generated by a hertzian dipole antenna and detected by the electro-optical detection.13,14 The bandwidth for this study is good from 0.2 to 1.6 THz at room temperature and from 0.2 to 2.0 THz at low temperatures. A Ti-sapphire laser (65 fs, 82 MHz) is used to generate and detect THz electric field. The entire THz spectroscopy system is continuously purged with dry nitrogen gas in a plexiglass box prior to and during the measurements to avoid the atmospheric water absorption. Each measurement consists of measuring the relative transmission through the reference and sample by toggling between the two apertures of the sample cell. THz-TDS measures the complex field transmission. We normalize our measurements to the reference transmission yielding a net field transmission with transmitted field amplitude, |t|, and the phase, φ: t=

E sample Ereference

= t e iφ = e ikd ( ns −1) e

−

αd 2

,

(1)

where k is the frequency in wavenumbers (cm–1), d is the thickness of sample, ns is the refractive index of sample, and α is the absorption coefficient. 3. Data and Result The temperature-dependent THz transmission for HEWL and trizma buffer solutions is shown in Figure 1 for several frequencies. The temperature dependence of the transmission of HEWL solution is relatively flat for T < 200K, and then decreases more rapidly above 200K. Above 273K, the transmission for both protein and buffer solutions dramatically drops due to the strong absorbance for bulk liquid water. For the protein solutions, the transition observed at 200K corresponds to the dynamical or “glass” transition15 that has been observed in neutron inelastic scattering and crystal x-ray diffraction measurements of the atomic root mean square displacement. A more thorough discussion of measurements of the glass transition at THz frequencies can be found elsewhere.10,16,17 Here we focus on the dramatic increase in transmission below freezing. One may intuitively expect that this large increase is due to the hydrogen bond network in the ice phase eliminating the individual water molecule dipoles from following the oscillating field. We note that the temperature where the abrupt increase in absorbance occurs of several protein solutions is different from the

J.-Y. CHEN ET AL.

128

freezing point of pure buffer solution or pure water, 273K. Typically the transmission increase occurs closer to 280K, as shown in Figure 1 for HEWL. This result suggests that the high concentration protein solution enforces sufficient structure on the system to suppress rotational response above the bulk-water melting temperature. Further discussion of this effect will appear elsewhere.16,17 While the water contribution is clearly suppressed below freezing, this does not guarantee that one can easily extract the protein contribution. Comparison measurements with pure buffer are shown in Figure 2, along with HEWL bound with 3NAG as a function of frequency for 295K and 270K. As seen in the figure, for 295K the transmission of the protein solutions is higher than the pure buffer, due to the volume displacement of the bulk water, whereas at 270K the absorbance for the protein solutions is higher than that of the bulk buffer particularly at the higher frequencies. Thus the relative index and absorption coefficients for protein solutions can be determined for as a function of concentration without concern for volume displacement.

1.2 1.0

|t|

0.8 0.6 0.4 0.2

Lysozyme solution 0.52 THz 0.84 THz 1.17 THz 1.31 THz

Trizma Buffer 0.51 THz 0.83 THz 1.17 THz 1.32 THz

0.0 100

150 200 Temperature (K)

250

300

Figure 1. Transmission, |t| of the 200 mg/ml lysozyme and trizma buffer solution as a function of temperature at several frequencies. The transmission of the lysozyme solution shows a dramatic increase at 280K at each frequency. Below 280K, the transmission of HEWL solution increases with the decreasing temperature. In comparison with the lysozyme solution, |t| of the trizma buffer has a big jump at 273K. Besides, |t| of buffer also increases with the decrease in temperature.

DEVELOPMENT OF TAGLESS BIOSENSORS

129

We now turn to the question of whether the THz dielectric response can distinguish between HEWL and 3NAG+HEWL. The native state and binding of HEWL and HEWL+3NAG solutions were verified by the fluorescence measurements. For each sample solution, the transmission at 295K is much less than that at ~270K. For both temperatures shown, the transmission decreases with increasing frequency for the entire frequency range and oscillates with a period ~0.8 THz. The oscillations are most dramatic at 270K and correspond to multiple reflections within the sample cell. At room temperature, it is difficult to distinguish any difference between the three solutions as the interplay between the excluded volume and the high absorbance of bulk water obscures the protein and bound protein contribution. In contrast, for T<273K there is a clear difference. The phase and period of the oscillations change for the different samples indicating differences in the real part of the index, whereas the transmission on average appears greatest for the pure buffer, less for 3NAG–HEWL and smallest for HEWL. However, the use of the raw transmission data is not

1.2 1.0

|t|

0.8 0.6

HEWL+NAG binding solution 295K HEWL+NAG binding solution 270K HEWL solution 295K HEWL solution 270K Trizma Buffer solution 295K Trizma Buffer solution 268K

0.4 0.2 0.0 0.0

0.5

1.0

1.5

2.0

Frequency (THz) Figure 2. Transmission, |t| of the 200 mg/ml lysozyme (solid symbols), HEWL+3NAG binding (open symbols), and buffer (solid line and dashed line) solutions as a function of frequency at 295K and at ∼270K. At room temperature, the THz absorption of bulk water is high and the transmission through sample is larger due to volume exclusion of water by protein. Below 273K, ice absorbs less and the absorbance of the protein sample is larger than pure solvent. See text for further discussion.

J.-Y. CHEN ET AL.

130

satisfactory, especially in light that it oscillates to values somewhat greater than 1. This greater than 1 transmission is due to referencing to the empty cell which has Fresnel reflection losses from the inside surfaces. The buffer, HEWL and 3NAG–HEWL solutions apparently have significant index matching with the quartz windows, significantly reducing this loss. In order to quantitatively compare the dielectric response between bound and unbound HEWL, we need to extract the complex dielectric response. In order to extract the complex dielectric response from the data, we must account for multiple reflection effects and Fresnel loss at the interfaces. In Figure 2, the oscillations in |t| arise from multiple reflections within the sample cell (multiple reflections due to the 2-mm thick quartz windows are below the measurement resolution of 0.7 cm–1). Further the transmission of the field through the protein solution is higher than that through the air gap, due to index matching of the protein solution with the quartz windows. Our measured index for the quartz windows is parameterized in the 0.1–2.5 THz range as nw = 1.92 – 0.21exp(–5ν0.79), where ν is the frequency in THz. We extract the complex index using an expression for the field transmission including multiple reflection effects:

t=

Esample Ereference

=e

iko d [( ns −1) + iκ s ]

tws tsw 1 − raw2 ei 2 ko d . twa taw 1 − rsw2 ei 2 ko d ( ns +iκ s )

(2)

The first two terms on the right hand side are the usual propagation term through a sample with thickness d and the Fresnel transmission term. The third term arises from multiple reflections within the cell, often referred to as the Fabry Perot term. The Fresnel transmission and reflection coefficients, tij and rij, for the ith to jth interfaces and have the form

tij =

2 Ni Ni + N j

rij =

Ni − N j Ni + N j

,

(3)

where Ni is the complex index related to the dielectric response ε through N i = ε i = ni + iκ i . In Eq. (2) subscript s stands for the sample and subscript w is for window, e.g. tws is the Fresnel transmission from the window to sample. In the low-loss regime, we neglect the imaginary part of the index in the Fresnel terms. Eq. (2) is fit to the magnitude and phase of the measured transmission for both the protein solution and pure buffer. The values of κ and n are extracted and shown in Figures 3 and 4, respectively.

DEVELOPMENT OF TAGLESS BIOSENSORS

131

2.6 2.4

Index

2.2 2.0 1.8

HEWL+NAG binding solution 295K HEWL+NAG binding solution 270K HEWL solution 295K HEWL solution 270K Trizma Buffer solution 295K Trizma Buffer solution 268K

1.6 1.4 0.0

0.5

1.0

1.5

2.0

Frequency (THz) Figure 3. Refractive index of the 200 mg/ml lysozyme (solid markers), HEWL+3NAG binding (open markers), and buffer (solid and dashed lines) solutions as a function of frequency at 295K and at ∼270K. The 293K buffer data are truncated due to the limited dynamic range of our transmission measurements, which cannot reliably measure transmissions less than 1% above 1.5 THz.

In Figure 3, the refractive index of the lysozyme, HEWL+3NAG binding and buffer solutions as a function of frequency are shown for 295K and 270K. For each solution, the refractive index at room temperature is larger than that at low temperature for the entire frequency. The index at 295K decreases with increasing frequency for all three solutions. Furthermore, the index at ~270K is frequency independent in the range between 0.5 and 2.0 THz for all three solutions. Compared to the index of the HEWL solution, that of the 3NAG+HEWL-binding solution is smaller for the entire frequency (0.2–2.0 THz) for both temperatures. The imaginary part of the index, κ of the HEWL and 3NAG+HEWL binding solutions both at 295K and at 270K as a function of frequency are shown in Figure 4. The imaginary part of the index of both HEWL and 3NAG+HEWL solutions at 270K increase almost linearly with the increasing frequency up to 2 THz, indicative of glass-like behavior. For each protein solution, κ at room temperature is much larger than that at T < 273K because of the significant absorption of bulk water. At both 295K and 270K, κ of the lysozyme solution is larger than that of the 3NAG+HEWL-binding solution for

J.-Y. CHEN ET AL.

132 0.12

HEWL+NAG binding solution 295K HEWL+NAG binding solution 270K HEWL solution 295K HEWL solution 270K

0.10

0.8

0.06 0.6

κ (295 Κ)

κ (270 Κ)

0.08

1.0

0.04 0.4

0.02 0.00 0.0

0.5

1.0

1.5

2.0

Frequency (THz) Figure 4. Imaginary part of the index, κ of lysozyme (line with markers) and HEWL+3NAG-binding (dashed and dotted lines) solutions as a function of frequency at 295K and 270K. The vertical axis of data for both protein solutions taken at 295K is on the right and that of data taken at 270K is on the left.

the entire frequency (0.2–2.0 THz). It is interesting to note that the change with binding is seen at room temperature and is consistent with that below room temperature. 4. Conclusions The THz dielectric response of the lysozyme and HEWL+3NAG binding solutions was measured as a function of frequency at room temperature and at 270K. The THz transmission of the lysozyme solution at 295K cannot be distinguished from that of the HEWL+3NAG-binding solution due to the large contribution from the solvent. After the solvent was frozen, the protein dominates the THz spectrum. The complex refractive indices of the HEWL+3NAG-binding solution is found to be smaller than those of lysozyme solution over the 0.2- to 1.8-THz range. The decrease in the complex refractive indices suggests that the density of modes at low frequencies is blue shifted or the dipole coupling decreases with binding. A blue shift of the collective vibrational modes implies a stiffening of the structure, or loss of flexibility with binding which is inconsistent with average Debye Waller factors from x-ray measurements, <x2 > = 25 Å2 for HEWL+3NAG from PDB 1HEW and <x2 > = 22 Å2 for HEWL from PDB

DEVELOPMENT OF TAGLESS BIOSENSORS

133

1BWH. These results support a change in the dipole coupling with binding. Neutron inelastic measurements as well as calculations would help in the resolution of the cause of the change in the THz dielectric response. The results demonstrate the sensitivity of THz dielectric response to protein–ligand binding. The method of THz dielectric spectroscopy with cooled solutions is straight forward and amenable to rapid assay using plates with arrays of solution wells. Acknowledgments This work was supported by the American Chemical Society (PRF 39554AC6) and the National Science Foundation (NSF CAREER PHY-0349256, NSF IGERT DGE0114330, and NSF REU DMR-0243833). We would like to thank Dr. Bright in the Chemistry department in University at Buffalo for helping us to do the fluorescence measurements. We also would like to acknowledge M/A-Com, a unit of Tyco Electronics, for providing us with the free GaAs wafers for generating antennas.

References 1. B. Erika, B. Torsten, O. Martin, L. Ruep, D. Roy, F. John, and C. S. Jeremy, Direct determination of vibrational density of states change on ligand binding to a protein. Phys. Rev. Lett., 93, 2:028103 (2004). 2. M. Brucherseifer, M. Nagel, P. H. Bolivar, H. Kurz, A. Bosserhoff, and R. Buttner, Label-free probing of the binding state of DNA by time-domain terahertz sensing. Appl. Phys. Lett, 77, 24, 4049–4051 (2000). 3. J. Y. Chen, J. R. Knab, J. Cerne and A. G. Markelz, Large oxidation dependence observed in terahertz dielectric response for cytochrome c. Phys. Rev. E (Statistical, Nonlinear, and Soft Matter Physics), 72, 4:040901 (2005). 4. J. R. Knab, J- Y. Chen, and A. G. Markelz, Hydration dependence of conformational dielectric relaxation of lysozyme. accepted by Biophys.l J. (2005). 5. A. Menikh, S. P. Mickan, H. Liu, R. MacColl, and X. C. Zhang, Label-free amplified bioaffinity detection using terahertz wave technology. Biosens Bioelectron, 20, 3:658(2004). 6. D. L. Woolard, W. R. Loerop, and M. S. Shur, Eds. Terahertz Sensing Technology. Volume 2 (World Scientific Publishing Co. ; 2003). 7. N. Nandi and B. Bagchi, Dielectric relaxation of biological water. J. Phys. Chem. B, 101, 50, 10954–10961 (1997). 8. J. T. Kindt and C. A. Schmuttenmaer, Far-infrared absorption spectra of water, ammonia, and chloroform calculated from instantaneous normal mode theory. J. Chem. Phy., 106, 11, 4389–4400 (1997). 9. Z. Chun, L. Kwang-Su, X. C. Zhang, W. Xing, and Y. R. Shen, Optical constants of ice Ih crystal at terahertz frequencies. Appl. Phys. Lett., 79, 4, 491–493 (2001).

134

J.-Y. CHEN ET AL.

10. J. R. Knab, J-Y Chen, R. Kao, and A. G. Markelz, Observation of protein dynamical transition in terahertz dielectric response. To be published. 11. S. S. Lehrer and G. D. Fasman, Fluorescence of lysozyme and lysozyme substrate complexes. Separation of tryptophan contributions by fluorescence difference methods. J. Biol. Chem., 242, 20, 4644–4651 (1967). 12. Y. Amo and I. Karube, Dielectric measurements of lysozyme and tri-N-acetyl-Dglucosamine association at radio and microwave frequencies. Biosens Bioelectron, 12, 9–10 :953(1997). 13. Y. Cai, I. Brener, J. Lopata, J. Wynn, L. Pfeiffer, J. B. Stark, Q. Wu, X. C. Zhang, and J. F. Federici, Coherent terahertz radiation detection: Direct comparison between freespace electro-optic sampling and antenna detection. Appl. Phys. Lett., 73, 4, 444–446 (1998). 14. D. Grischkowsky and N. Katzenellenbogen, Femtosecond Pulses of Terahertz Radiation: Physics and Applications. Salt Lake City, Utah 1991. 15. W. Doster, S. Cusack, and W. Petry, Dynamical transition of myoglobin revealed by inelastic neutron scattering. Nature, 337, 6209:754 (1989). 16. J. R. Knab and A. G. Markelz, Temperature dependence of lysozyme solution in terahertz dielectric response. To be published. 17. J-Y. Chen, J. R. Knab, and A. G. Markelz, Dynamical transition of terahertz dielectric response in oxidation states of cytochrome c. To be published.

Theme 3 DETECTION AND SENSING

IMPROVEMENTS TO ELECTRONIC TECHNIQUES FOR TERAHERTZ SPECTROSCOPIC DETECTION DANIEL W. VAN DER WEIDE,* ALAN D. BETTERMANN, AND MIN K. CHOI Department of Electrical & Computer Engineering, University of Wisconsin, 1415 Engineering Dr. Madison WI, 53706 USA JOHN GRADE Tera-X, LLC 8551 Research Way, Suite 175, Madison, WI 53562 USA

Abstract. Spectroscopic imaging with terahertz (THz) or submillimeterwave (SMM) sources holds great promise for both defense and dual-use applications, such as for detection of chemical/biological weapons (CBW), concealed explosives, and other weapons (particularly nonmetallic varieties), and even through-the-wall imaging. To perform spectroscopy with active illumination of the target, either multiple or tunable continuous-wave (CW) sources or broadband pulsed sources are needed; passive illumination (e.g. using the cold sky) is limited to outdoor settings. While using incoherent (or intentionally decohered) illumination, either from the sky, a noise source, or a frequency-modulated CW source helps to reduce the interference caused by standing-wave phenomena (analogous to laser speckle), all such approaches have severe limitations in that they cannot perform accurate ranging, they are limited to a narrow range of frequencies, or are relatively weak. They are also all fundamentally limited to incoherent detection, which has limited signal-to-noise ratio (SNR) performance, lacking the advantages of heterodyne downconversion and detection. Using pulsed, broadband, coherent THz or SMM sources, and detectors is ideal for spectroscopic imaging and detection. Keywords: nonlinear transmission line, coplanar waveguide, coherent detection

______ *To whom correspondence should be addressed: Daniel van der Weide, Department of Electrical & Computer Engineering, University of Wisconsin, 1415 Engineering Dr. Madison WI, 53706 USA

137 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 137–145. © 2007 Springer.

138

D.W. VAN DER WEIDE ET AL.

1. THz Free-Space Propagation The boundaries of the THz frequency range and the technologies associated with it have almost as many definitions as the number of laboratories pursuing THz studies. What unifies most researchers in THz technology is their work with quasi-optical propagation. Although important work has been done to probe and map THz frequencies on planar waveguides and circuits, the overwhelming majority of publications in this field discuss free-space propagation. This naturally involves both antennas borrowed from scaled-down microwave concepts and optics adapted from infrared and visible light. Propagation over distances greater than ~1 m is envisioned to be necessary to achieve standoff detection of concealed threats. Scientific curiosity has motivated a substantial amount of narrowband or CW work in the THz regime for space research and cosmology, where atmospheric absorption cannot interfere with observations. Back on Earth, however, pressure-broadened absorption lines of water vapor place severe limitations on the distance many THz waves can propagate. Although sending THz frequencies through gasses and measuring selective absorption has been useful for spectroscopy of narrow absorption lines,1 atmospheric absorption due to water vapor is largely a parasitic effect in THz sensing, imaging, and communications unless it is used for calibration. Water vapor absorption at 50% relative humidity at frequencies above 0.5 THz rises to between 0.1 and 10 dB/m in broad peaks that severely limit the distance of detectable propagation, effectively rendering the atmosphere black. This in turn limits stand-off distances when using >0.5 THz frequencies for sensing and imaging the dielectric contrast of remote targets. It also challenges communications systems attempting to use these frequencies for long distances. Conversely, THz communications within rooms or buildings will naturally be quite secure, and THz sensors will exhibit little interference with other systems or with each other. Liquid water in living tissue limits depths of THz penetration to millimeters, limiting THz medical imaging to applications involving skin conditions and teeth. One approach to gaining greater stand-off detection ability is to develop higher-power sources. Both solid-state two-terminal sources like Gunn or IMPATT diodes and vacuum electronic devices like klystrons, magnetrons, and backward-wave oscillators have continued their steady march toward higher power and higher frequencies within the microwave regime over the past several decades. One of the most prominent gaps in the THz toolkit,

ELECTRONIC THz SPECTROSCOPIC DETECTION

139

however, is the lack of suitable broadband and high-power amplifiers (to say nothing of low-noise amplifiers). While numerous source concepts, both broadband and narrow-band tunable, are under current investigation, the quest to find the analogy of a broadband laser’s gain medium in the THz regime has been challenging. The problem is one of fundamental physics: it gets increasingly difficult with high frequency to move carriers in semiconductor crystals, while from the visible working towards the THz, inverting carrier populations becomes harder as level separations approach a few times kT. One promising approach warranting research is to achieve a powerful THz vacuum electronic amplifier, such as traveling-wave tubes (TWTs), both for narrowband and wideband signals.2 Large standoff radar applications would benefit from the high power available from THz-regime gyrotrons, but important research challenges would include finding ways to significantly reduce their size and weight. Thus atmospheric absorption can make it largely impractical to do spectroscopy or imaging above 0.5 THz (what we will call the “upper THz”) beyond a meter or so of distance between emitter and detector. While this limits the range of applications to mostly laboratory studies and close-in spectroscopic imaging for industrial purposes, the majority of THz systems and results to date have been achieved in this upper THz regime. 2. Application of Pulsed Electronic Spectrometer Even with atmospheric absorption, there are windows of transmission to potentially enable large standoff detection of concealed threats. By illuminating a sample with a pico- or sub-picosecond electrical stimulus and coherently detecting its response, we can obtain reflection and transmission spectra of several substances, and compare them as a starting point for distinguishing biological, chemical, and energetic threats. We use an allelectronic THz spectrometer with phase-locked microwave sources to drive GaAs nonlinear transmission lines, enabling measurement of both broadband spectra and single lines with high precision. Our use of stable and frequency-offset synthesizers gives a user-selectable beat frequency that is low enough for processing by conventional 100-kHz instrumentation. We have also implemented a frequency-offset generator (single-sideband modulator) that can enable the system to be produced using integrated circuits.3 Increasingly sophisticated bioweapons and explosives require increasingly sophisticated detection technologies. Nonmetallic threats (e.g.

140

D.W. VAN DER WEIDE ET AL.

explosives) have motivated a multipronged approach to detection, including residue sniffing and computerized tomography. These techniques, however, suffer from invasiveness, slowness, unfamiliarity to the public, and significant potential for false negatives and positives. Broadband radiation of this range is normally difficult to achieve but has great potential in for screening; many compounds show specific absorption and dispersion in the 1–500 GHz range. These absorptions can arise both from intrinsic chemical responses (such as long-chain polymer vibrational or librational modes) and from the aperiodic crystalline-air matrix that could “trap” short-wavelength radiation. Because of the difficulty in deconvolving these effects from those of standing waves, it is still challenging to implement this technology in the field. Ways around these issues include variable rate pulsing to time-gate the received pulse and removed the effects of multiple reflections. Ultrawideband (UWB), carrier-free, impulse, or baseband radar has been rapidly gaining popularity in applications where complex and elusive targets are the norm. While UWB radar works up to about 10 GHz, our circuits generate extremely short electrical impulses (~1 ps) with correspondingly mm-wave spectra. Because we employ coherent detection, we reject noise outside the frequencies of interest. These systems can be deployed for spectroscopic imaging, as shown in Figure 1. With the appropriate optics, output power, and receiver sensitivity, the potential exists for a new type of security screening tool. Until now these approaches, whether optoelectronic or purely electronic, have been severely limited by standing-wave effects that arise from their coherence: interference between multiple targets that are positioned at integer multiples of λ/2, where λ is the wavelength of interest. Because coherent pulsed systems send out trains of narrow pulses, the standard approach to mitigating standing-wave effects is to time-gate the generator and detector, sending out a pulse, then waiting for it to reflect from the target and return to the detector before emitting another pulse. Since light travels at ~30 cm/ns, a target located at 1.5 m would impose a ~10 ns repetition rate (allowing for round-trip time), and the return-pulse time of arrival would be indistinguishable from a target at 3 m, giving rise to ambiguity. Thus with 100-MHz pulse rates (typical of mode-locked Ti:sapphire lasers), the unambiguous range is limited to less than 2 m unless time-gating is used. Time-gating, however, reduces average power, increasing the time to acquire an image. Were mode-locked lasers or solidstate nonlinear transmission line (NLTL) sources operating at 10 GHz (i.e. 100 ps) to be employed, their unambiguous range would be 100× shorter.

ELECTRONIC THz SPECTROSCOPIC DETECTION

141

Modulation of the pulse stream can be used to reduce ambiguity with pulsed systems. Using this spectrometer in our laboratory, we have measured broadband reflections from nonmetallic targets that indicate specific spectroscopic signatures associated with threats such as spore-forming bacteria and hazardous chemical.4 Since this system is based on integrated circuits, covers an unprecedented broad range of nonionizing, safe and low-power frequencies (10 to >500 GHz), and can distinguish among a large variety of concealed threats on personnel,5 it could prove useful as a field-deployable early-warning system for detecting hidden explosives or for release of biological agents.6

Figure 1. Electronic THz spectroscopic imaging system showing the optical arrangement using off-axis paraboloidal mirrors for focusing and collimating the THz beams.

3. THz Optics Since most applications for THz technology use free-space propagation, managing THz “light” is critical to success, especially because both the power of THz sources is low, even though the dynamic range of THz systems can peak at 60 dB. Far-field imaging and sensing constitute the

142

D.W. VAN DER WEIDE ET AL.

majority of current THz activity, though researchers have also employed near-field techniques such as using sharp, conductive probes as near-field antennas to localize THz energy, in some cases to submicrometer extents. It is also possible to generate and detect near-field THz light with subwavelength-sized crystals and excitation beams. Far-field optics in the visible spectrum is much larger than the wavelength of light; this is not true in most THz systems, and the situation is further complicated by the >10:1 range of wavelengths common to pulsed THz technology. Transducing, guiding, and focusing this broad range of wavelengths with high efficiency down to the diffraction limit is difficult. Metallic losses and modal dispersion in the source and detector substrates tend to attenuate the higher frequencies, while the limited spatial extent of reflective optics permits diffraction of longer wavelengths. New techniques such as photonic bandgap structures and ultrabroadband antennas are being incorporated into THz systems to improve optical management. By controlling the dimensions of the THz beam (i.e. the antenna pattern), THz imaging and sensing systems will realize considerable gain, even with their current power limitations. This in turn will create new opportunities for greater standoff distances in remote sensing. 4. NLTL Circuit Improvements We have demonstrated the possibility of using electronic THz system for a standoff detection.4 In that experiment, detection and classification of different powders was performed within 1-m distance between the sample and the transmitter and receiver. Used in longer-distance standoff detection, the signal strength would have to be higher. Pumping the NLTL with higher power could cause problems such as excessive heat and device failure. As one way of increasing the output power from the NLTL, we focused on improving the power efficiency of the system. Most of its input power is lost through NLTL due to various loss mechanisms. One of the major losses is the substrate loss. Therefore, if we make the length of the NLTL shorter, we could gain more power. However, if the line length decreases, frequency conversion from low to high frequency might be limited. One way to reduce the line length with the same frequencyconversion efficiency is increasing the impedance of the transmission lines, so the gap between varactors becomes closer. There are two ways of increasing the impedance of CPWs. One is to fabricate the signal line narrower, and the other is to fabricate the gap between signal line and ground to be wider. However, narrowing the signal

ELECTRONIC THz SPECTROSCOPIC DETECTION

143

line increases ohmic loss, and widening the gap between signal line and ground does not effectively increase impedance after some point. Some researchers improve the performance of CPW with air-bridge type structures, and they reported much lower losses as well as the increase of impedance of the line.7–10

Figure 2. SEM picture of CPW with mesa etching. It is clear that the signal line formed airbridge type line with varactor isolation.

Established NLTL fabrication processes isolate varactors from each other with ion implantation. This ion implantation requires several steps such as mask metal deposition and removal. This step also reduces the device size with lateral ion migration (straggle) when the substrate is bombarded, requiring precise characterization of the process. Alternatively, GaAs dry etching followed by mesa etching can define the device area more precisely than ion implantation. This process also has an advantage of removing substrate under the CPW resulting in an air-bridge-like structure, so the varactor isolation and transmission line impedance increase can be achieved in one step. Figure 2 shows an oblique SEM of the mesa etched CPW. We used a network analyzer to measure the effective dielectric constant of the CPW before and after mesa etching (Figure 3). The effective dielectric constant after mesa etching was reduced by more than a factor of two, and the corresponding impedance of the CPW was about 71 ohms before the mesa etching, increasing to about 120 ohms after etching. Assuming the inductance of the line does not change by much, and the capacitance of the line changes mostly, this increase of the impedance allows more closely spaced varactors in NLTL. This can result the reducing the total length of the NLTL while improving the power conversion efficiency.

144

D.W. VAN DER WEIDE ET AL.

Figure 3. Measured effective dielectric constant of the 140 µm CPW before (line) and after (dot) mesa etching. Mesa etching almost doubled the impedance of the line.

5. Conclusion With advances in coherent generation and amplification, it will be possible to develop arrays of these sources and detectors in an imaging system to offer a new screening technology. Several labs have now shown that after collecting data under different scenarios and building a database, it may be possible to screen out biological or chemical threats in real time. It is thus attractive to create a smaller size, lower cost spectrometer with better standoff detection ability using electronic techniques. With improvements to NLTLs there is potential for integration directly into existing security portals, all resulting from an integrated circuit approach that uses no moving parts. Acknowledgment This work was funded by the US Army Research Office, the Air Force Office of Scientific Research, DARPA, and the Office of Naval Research.

ELECTRONIC THz SPECTROSCOPIC DETECTION

145

References 1. D.W van der Weide, J. Murakowski, and F. Keilmann, Gas-absorption spectroscopy with electronic terahertz techniques, IEEE Trans. Microw Theory Tech., 48(4, pt2), 740– 743, (2000). 2. J.H. Booske, et al., Terahertz-regime, micro-VEDs: evaluation of micromachined TWT conceptual designs. in Pulsed Power Plasma Science (IEEE Las Vegas, NV, 2001). 3. P. Akkaraekthalin, S. Kee, and D.W. van der Weide, Distributed broadband frequency translator and its use in a 1–3 GHz coherent reflectometer, IEEE Trans. Microw Theory Tech., 46(12), 2244–2250 (1998). 4. M.K. Choi, A.D Bettermann, and D.W. van der Weide, Biological and chemical sensing with electronic THz techniques, Proceedings of SPIE, 5268, pp. 27–35 (2004). 5. D.W. van der Weide, Wideband terahertz sensing and spectroscopy with electronic sources and detectors, in Terahertz sources and systems (Kluwer Academic Publishers: Dordrecht, The Netherlands, 2001), pp.301–314. 6. M.K. Choi, et al., Broadband 10–300 GHz stimulus-response sensing for chemical and biological entities, Physics in Med. and Bio., 47, 3777–3787 (2002). 7. H.S. Lee, D.H. Shin, Y.H. Chun, S.C Kim, B.O. Lim, T.J. Baek, S.K. Kim, H.C. Park, and J.K. Rhee, Design and characterisation of micromachined transmission line with dielectric post for millimetre-wave applications, IEEE Electronics Lett., 39, 25, 1827– 1828 (2003). 8. V. Milanovic, M. Ozgur, D.C. DeGroot, J.A. Jargon, M. Gaitan, and M.E. Zaghloul, Characterization of broad-band transmission for coplanar waveguides on CMOS silicon substrates, IEEE Trans. Microw Theory Tech., 46, 5, 632–640 (1998). 9. V. Milanovic, M. Gaitan, E.D. Bowen, and M.E. Zaghloul, Micromachined microwave transmission lines in CMOS technology, IEEE Trans. Microw Theory Tech., 45, 5, 630– 635 (1997). 10. F. Schnieder, R. Doerner, and W. Heinrich, High-impedance coplanar waveguides with low attenuation, IEEE Microwave and Guided Wave Lett., 6, 3, 117–119 (1996).

TERAHERTZ TIME-DOMAIN SPECTROSCOPY OF CRYSTALLINE AND AQUEOUS SYSTEMS PETER UHD JEPSEN,† HANNES MERBOLD, ZHENGXIN LI, AND XIAOYU XING COM•DTU, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark STEWART CLARK Department of Physics, Durham University, Durham DH1 3LE, United Kingdom

Abstract. There is a fundamental difference between the dielectric spectra of crystalline systems as well as amorphous and liquid systems in the THz range. Here we discuss recent theoretical progress on the calculation of the lowest vibrational modes in crystalline compounds. With density-functional perturbation theory it is possible to simulate, in a quantitative manner, the THz vibrational spectrum of hydrogen-bonded molecular crystals. In contrast to crystalline systems, aqueous systems have a featureless dielectric spectrum in the low-THz range. We will show that it is possible to use reflection THz spectroscopy to measure the alcohol concentration in aqueous solutions in a manner that is independent of the contents of other ingredients such as sugar, yeast, and organic particles. Keywords: THz time-domain spectroscopy, crystalline compounds, aqueous systems, alcohol concentration measurements

1. Introduction In the recent years there has been a tremendous activity in the field of basic and applied THz-frequency research. A sizable fraction of this effort has been focused on the exploitation of the fact that most organic molecules in the solid state have a rich and distinct dielectric spectrum in the THz region 0.3– 5 THz. It has turned out that the vibrational modes found in this particular region of the electromagnetic spectrum are highly characteristic not only for the molecule, but also for its environment. The space group symmetry of

†

To whom correspondence should be addressed, COM•DTU, Technical University of Denmark, DK-2800 Kongens Lyngby, Denmark, e-mail address: (jepsen @ com.dtu.dk)

147 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 147–165. © 2007 Springer.

148

P.U. JEPSEN ET AL.

the unit cell as well as the content of co-crystallized solvent molecules (e.g. water) has been shown to be the main factors forming the THz vibrational spectrum of the solid. The richly structured dielectric spectra often observed in poly- and singlecrystal materials, including powders, are due to combinations of phonon- and intramolecular modes of the crystallites or single crystals. On the other hand, in amorphous condensed-phase systems the existence of spectrally localized features is extremely rare because of the strong coupling between the random environment and intramolecular modes.1 This study showed that the longrange order of the environment of the molecules is one of the dominating factors in the shaping of the dielectric spectrum of the molecules. In this work we present a generally applicable ab initio simulation method that is capable of predicting the position and intensity, as well as identifying the normal modes of vibrational spectra in the THz region. The applicability of the method is demonstrated with results of the simulation of vibrational modes of the hydrogen-bonded molecular crystal sucrose. The experimentally determined absorption spectrum of polycrystalline sucrose has been reported earlier by the Freiburg THz Laboratories.1 In spite of the lack of spectral features in the THz range of the dielectric function of amorphous materials and aqueous solutions, THz spectroscopy turns out to be very useful for identification also of ingredients in solution, and also to determine their concentration with high accuracy. We will show how THz reflection spectroscopy can be used to determine the alcohol concentration with high accuracy in a wide range of beverages, independent of the presence of microparticles, carbon dioxide, and other ingredients. 2. THz Time-domain Spectroscopy Standard transmission THz time-domain spectroscopy (TH z - TDS) has been used for the experimental determination of the dielectric function of the crystalline systems studied in Section 3 as well as for the reflection spectroscopy work reported in Section 4 and has been described in previous publications .2,1, 3 This method uses femtosecond excitation and gate pulses in two synchronized pulse trains from the same femtosecond oscillator to generate and to detect ultrashort bursts of far-infrared radiation (the THz pulses). The experimental setup is illustrated schematically in Figure 1. Each of the excitation pulses drives an ultrafast current in a photoconductive switch. The rapid acceleration dynamics of the photogenerated charges leads to emission of a short pulse of electromagnetic radiation. This pulse is transmitted through the sample, and detected in another photoconductive switch which is gated by a second replica of the femtosecond pulse. By gradually changing the

ELECTRONIC THz SPECTROSCOPIC DETECTION

Figure 1.

149

Experimental setup for THz time-domain spectroscopy.

arrival time of the gate pulse with respect to the THz pulse while recording the induced photocurrent in the detector we can measure the temporal profile of the THz pulse with subpicosecond time resolution. In the transmission experiments used for the recording of crystalline spectra in Section 3 the sample is placed in a closed-cycle helium cryostat equipped with 6 mm thick polymer windows, transparent to THz radiation. The spectra presented here are recorded at 10K which make a good comparison to the zero temperature ab initio results. In the reflection measurements described in Section 4 the THz beam is focused to a frequency-independent spot in the reflection plane. In contrast to the transmission measurements, here the reflected trace contains both the reference and the sample signals. 3. Modelling of THz Spectra of Crystalline Compounds The reliable prediction of the precise position and strength of the peaks in the THz-frequency spectrum of crystalline compounds, as well as the assignment of these modes to specific molecular motion, has remained an unmet challenge until now for all but the simplest systems. Exceptions are systems of high symmetry and with a small number of atoms in the unit cell of the crystal. The phonon spectra of many inorganic semiconductors are well understood, silicon, germanium, and gallium arsenide being well-known examples. Similarly, the phonon spectra of dielectrics with simple crystal structures can be calculated, with polyethylene and diamond as classic examples. When the number of atoms in the unit cell increases and the interaction between the atoms in the crystal becomes weaker than in covalent or ionic bonded crystals the prediction of the phonon spectrum becomes an extreme theoretical challenge. This is due to the long-range, weak interactions and the large number of ions involved in the description.

150

P.U. JEPSEN ET AL.

The description of solid-state, crystalline compounds must take the periodic arrangement of atoms in the crystal into account . 4 – 6 For a few, particularly simple crystalline systems, the periodic crystal structure can be approximated by linear strings of the molecules, and density-functional theory (DFT) applied to isolated clusters of molecules can under special circumstances successfully simulate the general appearance of THz absorption spectra . 3 However, in the general case the full crystal structure must be taken into account in order to obtain a realistic description of the lowest vibrational modes in a molecular crystal. This is the fundamental reason why calculations on isolated molecules or small units of molecules are quite successful in reproducing the mid-infrared vibrational spectra of molecules even in the condensed phase, but fail to predict the position and intensity of low-frequency modes, typically below 5 –10 THz .7, 8 Therefore calculations on isolated molecules should be interpreted with extraordinary care when discussing vibrational modes in solid-state materials in the THz region. Recently, Korter et al. performed solid-state simulations, using the software packages DMol, CHARMM, and CPMD, predicting the THz-frequency vibrational spectra of the high explosive HMX 9 and the amino acids serine and cysteine, 10 taking the periodic boundary conditions of the crystal structure into account. This approach has led to reasonable overall agreement between experimental absorption spectra and simulated ones. 3.1. DENSITY-FUNCTIONAL PERTURBATION THEORY

The main goal of the calculations are to predict the low frequency vibrational modes of molecular crystals. To obtain accurate frequencies and spectroscopic intensities of low-frequency modes is extremely computationally demanding and very high convergence criteria are required throughout the calculations. In other work in which the low-frequency THz modes of molecular materials have been calculated, there has been a severe limitation in that non-periodic molecular clusters have been used where up to eight molecules were used. Although those calculations are adequate for describing the intramolecular modes, the long-range interactions of the crystal are completely ignored. To accurately predict the low-frequency modes, which are essentially intermolecular in nature, the full periodic structure must be considered. In addition to this, the long-range dipolar interaction induced by zone-center phonon modes need be considered which does not form part of the cluster calculations. For this we must turn to perturbative approaches which includes the long wavelength limit electric field induced by some phonon modes. Again, this is essential not only for obtaining accurate frequencies but also their spectroscopic intensities.

ELECTRONIC THz SPECTROSCOPIC DETECTION

151

The calculations are based on the plane-wave density functional method within the generalized gradient approximation as implemented in the Castep code .11,12 Norm-conserving pseudopotentials in the Kleinman-Bylander 13 form are used to describe the electron – ion interactions. The valence electron wavefunctions are expanded in a plane-wave basis set to a kinetic energy cutoff of 1200 eV which converges total energies to better than 0.1 meV/atom. Brillouin zone integrations are performed using a k-point set that converges the energies to an equivalent accuracy. Electronic minimizations are performed using a preconditioned conjugate gradient scheme 14 and are converged to machine accuracy (approximately 10−13 eV/atom). Geometry optimizations are also performed using a conjugate gradients scheme; accurate geometries were found to be essential in order to obtain reliable values for the low frequencies of the molecular crystals. It is important to note that obtaining frequencies in the THz region is a very demanding task from a computational point of view. In molecular crystals there is a wide range of bonding strengths and therefore to obtain the low-frequency modes accurately, all the self-consistent and the perturbative calculations along with the geometries of the system considered must be converged to much tighter tolerances than is usual in standard plane-wave pseudopotential calculations. In general we have found that the total energy of the system (in terms of k-point sampling, total energy convergence, etc) must be converged to almost machine precision to finally obtain accurate low-frequency modes. Once accurate geometric and electronic structures are obtained, we perform density-functional perturbation calculations based on the formalism of Gonze, et al. 4, 5 using the Castep code . 11, 12, 6 The zone-centre phonon modes are calculated and also the materials’ dielectric properties, bulk polarizability and Born effective charges. From this we are able to compute the spectroscopic intensities of the modes and compare directly with experiment. 3.2. MODELLING RESULTS AND DISCUSSION

We now present the results of the THz experimental frequencies and intensities and compare to the theoretical values. The experimentally recorded absorption spectrum of sucrose is shown as the full line in Figure 2, scaled with the molar concentrations of the sample. The frequencies and intensities of the corresponding calculated modes are shown as vertical bars in the same graph. The overall intensities have been scaled by the same factor to fit within this representation, but the relative line strengths have not been adjusted. The intensities of the measured absorption lines are reliable since it was ensured that the spectrometer was operated below its saturation .15 Hence not only the position but also the relative strengths of the absorption lines can be

152

Figure 2. sucrose.

P.U. JEPSEN ET AL.

Experimental (solid line ) and calculated (vertical bars) vibrational spectrum of

compared with the simulated values. This allows a very stringent test of the simulation results. We find a very good agreement between theory and experiment. Basically the position as well as the intensity of each observed absorption band in the frequency region below 4 THz is reproduced by the DFPT simulation. The simulation predicts only position and intensity of the normal modes of the crystal. Hence the observed line widths cannot be compared with simulations. The simulation results indicate, however, that for instance the broad absorption band between 3.2 and 3.8 THz is probably composed of several vibrational modes. There are a few modes in the experimentally determined spectrum that is not reproduced by the simulation – specifically the weak features at 2.3, 2.55, and 2.6 THz. This may indicate that these observed modes are associated with combination bands which are not accounted for in the simulation. The close agreement between experiment and theory allows us, for the first time, to assign specific normal modes to the observed vibrational frequencies. As an example of such an assignment, Figure 3 shows a graphical rendering of the normal mode at 1.9 THz. The direction and relative amplitude of the motion of each atom is indicated with arrows. The dimensions of the optimized unit cell is indicated by the box structure, and the optimized atom placement within the unit cell is shown. To indicate the intermolecular network we show the molecules in the neighboring unit cells, indicated by transparent colors. The intermolecular hydrogen bonds are illustrated with dark gray connections, and the intramolecular covalent bonds are shown as light gray connections. The figure clearly confirms

ELECTRONIC THz SPECTROSCOPIC DETECTION

153

Figure 3. Motion of the atoms in the sucrose crystal associated with the vibrational frequency of 1.9 THz. The surrounding unit cells of the crystal are indicated in half-transparent colors. Hydrogen bonds are shown in dark gray color while the intramolecular covalent bonds are shown in a light gray color.

that sucrose is held together by very strong intermolecular network of hydrogen bonds. The normal-mode motion indicated in the figure allows us to draw an important and, in our opinion, quite general conclusion about the nature of the low-frequency modes of molecular crystals. A pure intermolecular mode would lead to motion of the atoms of each molecule to have the same amplitude and identical or at least highly aligned direction. However, inspection of Figure 3 shows that this is not the case. Both direction and amplitude of the motion of each atom in the molecules is only lightly correlated to that of the other atoms. Hence the predicted normal-mode motion is neither a pure intermolecular, phonon-like motion nor a pure intramolecular vibrational mode. In contrast there is a strong and complicated coupling between the intra- and intermolecular motion, involving both hydrogen bonds and covalent bonds. This illustrates very clearly that a sharp distinction between inter- and intramolecular modes in the THz range is not possible in this case. Inspection of the other low-frequency modes of the sucrose crystal as well as simulation results on other molecular crystals indicates clearly that such a distinction is not possible in molecular crystals, even for the lowest-frequency modes.

154

P.U. JEPSEN ET AL.

4. Disordered Systems: Determination of Alcohol Concentrations The clear spectral features observed in molecular crystals disappear when the system is transferred to a phase with no long-range order, such as amorphous or liquid materials. In this section we will demonstrate that in spite of the lack of distinct spectral features in the THz region, it is still possible to use this frequency range for quantitative spectroscopic work. We will illustrate this with a discussion of a method for the precise determination of the content of alcohol in commercial beverages such as beer, wine, and spirits. Alcohol concentration measurements are currently performed with specialized spectroscopic equipment, employing transmission measurements. The precise calculation of the alcohol content then relies on a chemometric analysis of the measured transmission spectrum in the vicinity of a characteristic vibrational band of the alcohol molecule in the near- or mid-infrared region. Therefore the precision of the measurement is highly sensitive to the thickness of the sample chamber as well as the preparation of the sample material. Furthermore, because of the limited dynamic range of FTIR spectrometers, the range in which the measurements are not influenced too severely either by noise or saturation of the transmission is limited, and commercial instruments based on optical methods are calibrated and optimized for certain beverage types, i.e. specific alcohol ranges. We have developed a method for the determination of alcohol content in a wide range of alcoholic beverages that is insensitive to the content of for instance yeast, sugar, and other organic residues as well as carbonation of the beverage. The method is based on a reflection measurement, and hence it eliminates the need for special sample preparation and it is well suited for integrated, continuous monitoring of the beverage during production. 4.1. SELF-REFERENCED REFLECTION THz -TDS

For the measurements described here we use a method that we call selfreferenced reflection THz-TDS. In the following we will describe the principles of this measurement method. Earlier Thrane et al.16 and Rønne et al.17,18 used similar methods for the determination of the temperature dependence of the dielectric function of pure water. The fundamental geometry of a self-referenced reflection THz-TDS experiment is illustrated in Figure 4. The following analysis is described in the frequency domain although the raw data are recorded in the time domain and consist of a single trace containing both the reference reflection Ere f (t) and the sample reflection signal E sam (t). The data analysis starts with a separation of the two signals in the time domain, and then they are Fourier transformed to the frequency domain individually. These two signals are named Ere f (ω) and E sam (ω), where ω = 2πν is the frequency measured in radians per second.

ELECTRONIC THz SPECTROSCOPIC DETECTION

^

3 2 m e d iu m

n = n + ik n d

f

a ir

n

S i

S i

n 1

in p u t, E 0 ( w )

155

a ir

q s a m p le , E re fe re n c e , E

re f

s a m

(w )

(w )

Figure 4. Geometry and notation used in the description of a reflection THz-TDS experiment.

It is assumed that the input signal is incident on the front surface of the window at an angle θ. Consequently the incidence angle φ on the second interface is (1) sin θ = nS i sin φ . The effective single-pass propagation distance de f f inside the window material is nS i dS i dS i = q . (2) de f f = cos φ n2S i − sin2 θ The basic quantity needed in the analysis is the ratio between the sample signal and the reference signal. This ratio can be expressed as E sam (ω) t12 rˆ23 t21 = exp(2inS i ωde f f /c) . Ere f (ω) r12

(3)

The transmission and reflection coefficients are dependent on the incidence angle, and the reflection coefficient rˆ23 is complex if the material behind the window is absorbing. This is the normal case in the spectroscopic measurement. For purposes of calibration of the spectrometer and characterization of the window material itself the measurement can also be carried out without sample material behind the window. In that case the reflection coefficient rˆ23 is a real-valued number. The transmission coefficients t12 and t21 as well as the reflection coefficient r12 are always real-valued. Since the radiation must be coupled into the window material only angles below the total internal reflection angle are considered in this analysis.

156

P.U. JEPSEN ET AL.

The various transmission and reflection coefficients are 2 cos θ 2 cos θ sin φ , = t12 = q sin(θ + φ) cos θ + n2S i − sin2 θ q 2 n2S i − sin2 θ 2 cos φ sin θ = t21 = q , sin(θ + φ) 2 2 cos θ + nS i − sin θ q cos θ − n2S i − sin2 θ r12 = , q cos θ + n2S i − sin2 θ q p n2S i − sin2 θ − 1 − sin2 θ , r23,air = q p n2S i − sin2 θ + 1 − sin2 θ q p n2S i − sin2 θ − nˆ 2 − sin2 θ . rˆ23,sample = q p 2 2 2 nS i − sin θ + nˆ 2 − sin θ

(4)

(5)

(6)

(7)

(8)

4.2. CHARACTERIZATION OF THE WINDOW MATERIAL AND BEAM DISPLACEMENT

Normally the window material has known dielectric properties. High-purity silicon (Si) has an index of refraction of 3.4244 at room temperature17 and neglible absorption. The high index of refraction is ideal for the purpose because it results in a reflection coefficient at normal incidence of – 0.54. However, in the case of an unknown window material it is, at least in principle, possible to use the analysis described here also to characterize the window material itself. At normal incidence with a collimated THz beam the solution is known.17 The known dielectric properties of the window material can also be used to solve an important problem in the experimental geometry. As can be seen in Figure 4 the reference beam and the sample beam are displaced with respect to each other after the window reflections. This is difficult to take into account in a strictly formal manner, but nevertheless the effect of the displacement can be described in a more empirical way, as discussed in.18 With no sample attached to the back surface of the window the ratio of the sample and reference signals is t12 r23,air t21 E sam re f re f = Am exp(i∆m ) = exp(2inS i de f f ω/c)A exp(i∆) . Ere f r12

(9)

ELECTRONIC THz SPECTROSCOPIC DETECTION

157

Here the complex correction factor A exp(i∆) summarizes the complicated effects of the displacement of the sample beam with respect to the reference beam. The measured amplitude ratio and phase difference between the referre f ence and the sample beams in this reference measurement are denoted Am re f and ∆m . The idea behind this is to adjust the amplitude and phase of the sample signal in such a way that the analysis results in the correct index of the window material. In the subsequent analysis of the signal from the spectroscopic measurement with a sample attached to the back side of the window this correction factor is assumed to be the same as without sample attached to the window. Some tedious algebraic manipulation leads to expressions for the amplitude A and phase ∆ of the complex correction factor,

re f

A = Am · re f

1 re f = Am t12 t21

∆ = ∆m −

q 2  2 2 cos θ + nS i − sin θ · , q 4 cos θ n2S i − sin2 θ

2nS i de f f ω −π. c

(10)

(11)

These values can be stored for use in the analysis of spectroscopic measurements using the same experimental geometry (i.e. the same incidence angle, window thickness, and index of refraction). This indicates that it is re f re f required to measure Am and ∆m every time the THz beam path is aligned, or the position of the reflection unit in the THz beam path is altered. If the amplitude and phase of the actual spectroscopic measurement with new a sample attached to the window back surface are denoted Anew m and ∆m , respectively, then the quantities used in the subsequent analysis should be Anew Anew m = mre f · t12 t21 , A Am new = ∆m − ∆ .

A?m =

(12)

∆?m

(13)

4.3. EXTRACTION OF THE DIELECTRIC FUNCTION OF THE SAMPLE

Based on the results of the previous section it is now assumed that primary experimental data are available, and that the correction factor taking the displacement of the sample beam with respect to the reference beam has been determined. If this is the case then Eq. (3) can be written as A?m exp(i∆?m ) = rˆ23,sample

t12 t21 exp(2inS i de f f ω/c) . r12

(14)

P.U. JEPSEN ET AL.

158

Since the complex reflection coefficient rˆ23 contains the dielectric function of the sample material, the following relation is useful, rˆ23,sample = A?m exp(i∆?m ) ·

r12 exp(−2inS i de f f ω/c) . t12 t21

(15)

Reinserting the expression for A?m and ∆?m results in an expression containing only the raw experimental data from the new measurement and from the reference measurement without the sample attached to the back surface of the window, rˆ23,sample =

Anew m re f Am

   re f · r23,air exp i ∆new m − ∆m

q p n2S i − sin2 θ − 1 − sin2 θ · q = p n2S i − sin2 θ + 1 − sin2 θ    re f × exp i ∆new . m − ∆m Anew m re f Am

(16)

(17) (18)

Using Eq. (8) to find the expression for the complex index of refraction nˆ of the sample results in q 2 1 − rˆ23,sample n2S i + 4ˆr23,sample sin2 θ . (19) nˆ = 1 + rˆ23,sample

Figure 5. (a) Overview of time-domain data from the reflection unit showing the referenceand sample reflections and (b) a detailed view of the modifications of the sample reflections due to samples with varying water content.

ELECTRONIC THz SPECTROSCOPIC DETECTION

159

The left panel of Figure 5 shows a temporal trace recorded with a silicon window of thickness 2.005 mm. The first pulse (the reference pulse) is the part of the THz pulse reflected from the lower air/window interface as shown in Figure 4, and the second pulse (the sample pulse), arriving at the detector 46 ps later, is the part of the THz pulse reflected from the window/sample interface. The right panel of Figure 5 illustrates the effect on the sample signal when liquid solutions with varying water content are placed in contact with the silicon window. In this case the solution is a dioxane –water mixture since dioxane is miscible with water and dioxane has a very low absorption compared to water in the THz range. We observe a phase advance of the sample signal with increasing water content. This phase shift of the reflection coefficient is a direct signature of the increasing absorption coefficient of the sample material. We also observe a decrease of the amplitude of the sample signal with increasing water content. This is due to an increasing refractive index of the sample solution with increasing water content. Hence the index contrast between the silicon window and the sample material, and therefore also the magnitude of the reflection coefficient, is reduced. In Figure 6 the room-temperature absorption coefficient and the index of refraction of pure water and of pure ethanol is shown in the frequency range 0.1–1.2 THz, determined by the method described above.

Figure 6. (a) Absorption coefficient and (b) index of refraction of pure water and pure ethanol, recorded at room temperature.

160

P.U. JEPSEN ET AL.

Both the absorption coefficient and the index of refraction of water is significantly higher than that of ethanol. This is due to a stronger intermolecular hydrogen bond network in water compared to that in ethanol. Both spectra are in agreement with reports in the literature.17,19 The strong difference in the dielectric properties of water and ethanol allows determination of the alcohol contents in water– ethanol mixtures. As will be demonstrated, the alcohol content of commercial beverages can also be determined with good precision with this method. We observe no distinct spectral features in the absorption and refractive index spectra of water and ethanol in the low-THz range, and this is also the case for water– ethanol mixtures. The spectral shape changes gradually from the pure water spectrum to the pure ethanol spectrum as the alcohol fraction of a water– ethanol mixture increases. For this reason it should be possible to estimate the alcohol concentration in such a mixture by its absorption and index of refraction at a single frequency. Alternatively the average values over the useful bandwidth of the spectrum or values at several frequencies can be used for the estimation. In order to verify this we prepared 21 reference mixtures with varying water– ethanol fractions from pure water to pure ethanol in 5%-steps, measured by weight. We then characterized their absorption coefficients and indices of refraction in the THz range. The solid lines in Figure 7 show the absorption coefficient and the index of refraction as function of the ethanol concentration of the mixtures, averaged over the frequency range 0.2–1.2 THz. The average index of refraction, as well as the average absorption coefficient, decreases with increasing ethanol content of the sample, and the variation is quite strong. Because of this strong correlation between the alcohol content and the optical properties of the reference liquids, we also measured the corresponding THz properties of a wide range of commercial alcoholic beverages, as listed in Table 1. The data points shown in Figure 7 are the average values of the absorption coefficient and the index of refraction of the different alcoholic beverages and liquors investigated in this study. The error bars shown together with the data points are the standard deviation between five subsequent measurements on the same sample without changing the experimental conditions. The deviation between different samples of the same beverages is somewhat larger than indicated in this figure. As in the case of pure ethanol –water mixtures we also observe a strong correlation between the alcohol content and the THz dielectric properties. In this study we confirm this correlation, and note that it is therefore possible to estimate the alcohol concentration based on the measured THz properties. A detailed analysis will be presented in a forthcoming publication. Here we

ELECTRONIC THz SPECTROSCOPIC DETECTION

161

Figure 7. (a) Absorption coefficient and (b) index of refraction of pure ethanol –water mixtures of varying ethanol content. The data points shown in gray are measurements of commercial beverages and alcohols.

note that the accuracy of the method relies critically on the reference data set. A chemometric analysis indicates that an accuracy of half of the interval between the reference mixtures is easily possible. In this case the interval was 5% by weight, and we obtain an absolute accuracy of 2–3%. A more closely spaced reference data set with 1%-intervals should lead to an absolute precision significantly below 1%, approaching that of state-of-the-art commercial equipment for alcohol-concentration measurements in liquids. The alcohol concentrations shown in Table 1 are as given on the label of the bottles and were not verified by other methods. The concentrations may therefore be slightly higher than the actual concentrations since the seal of some of the bottles were broken several months prior to the measurements. This may be the reason why some of the data points fall above the reference curves. We notice a better agreement between the reference curve of the index of refraction and the index of refraction of the commercial beverages and liquors than those of the measured absorption coefficient. This may indicate that the refractive index is a more stable indicator of the alcohol content than the absorption coefficient. Several of the liquors in Table 1 have a significant amount of sugar added. Supplementary measurements on sugar solutions show that the index of refraction is insensitive to the sugar content

162

P.U. JEPSEN ET AL.

TABLE 1. Overview of alcoholic beverages and liquors characterized by THz spectroscopy, with indication of the alcohol strength in volume fraction and weight fraction Beverage

.

Organic beer (Denmark) Strong dark beer (Belgium) Sake (Japan) Souchu (Japan) Gin (United Kingdom) Williams Pear Schnaps (Germany) Obstwasser (Germany) Pernod (France) Gin (United Kingdom) Red Absinthe (France) Green Absinthe (Bulgaria) Green Absinthe (Bulgaria)

Alcohol volume fraction % 5.8 10 15 25 37.5 40 40 45 47.5 55 72 85

alcohol weight fraction % 4.6 8.0 12.1 20.4 31.1 33.3 33.3 37.8 40.1 47.2 64.6 79.4

of a water solution whereas the absorption coefficient is significantly reduced by the addition of sugar to a water solution. Hence we conclude that the index of refraction is very important for an accurate determination of the alcohol concentration in mixtures containing water, ethanol, and sugar. We also note that the presence of other ingredients such as coloring agents, taste and odor components, carbon dioxide, and solid precipitates has minimal effect on the dielectric spectrum of the liquid. This is illustrated in Figure 8 where the absorption coefficient and index of refraction of a softdrink is compared to those of destilled water. The measurements were performed on the carbonated softdrink and then on the softdrink with the carbon dioxide removed by thorough shaking of the bottle. The THz optical properties of the softdrink are obviously not influenced by the significant number of gas bubbles present at the interface between the liquid and the silicon window of the spectrometer during the measurement. The figure illustrates that even the presence of significant amounts of sugar (approximately 12% by weight) has very little influence in the index of refraction of the liquid. 5. Conclusions We have demonstrated that the plane-wave DFPT method taking periodic boundary conditions into account is capable of simulating the THz vibrational spectra of different hydrogen-bonded crystals with convincing accuracy.

ELECTRONIC THz SPECTROSCOPIC DETECTION

163

Figure 8. (a) absorption coefficient and (b) index of refraction of carbonated and decarbonated softdrink and destilled water.

The detailed information available with the simulation method presented here about the solid-state THz vibrational modes of molecules will enable researchers to obtain new and important insight into the weak, delocalized forces that hold hydrogen-bonded crystals together. The interplay between weak and strong intermolecular forces can lead to unexpected behavior of vibrational modes. An example of this is the unusual blue-shift with increasing temperature of the frequency of the lowest vibrational modes in sucrose .1 With the identification of the specific vibrational modes responsible for the blue-shifting absorption line, it should be possible in the future to perform molecular dynamics simulations taking the temperature of the crystal into account, and investigate the temperature-dependent interplay between weak and strong intermolecular forces. Similar studies could also further aid the understanding of THz anharmonicity in biologically relevant molecules, such as biotin .20 In the second part of the paper we have demonstrated that even disordered systems, where the long-range order is absent, contains useful and relevant information in the THz range. THz reflection spectroscopy can be used as a stable and precise method for contact-free, optical determination of the alcohol content in commercial beverages and liquors. The method is insensitive to the other ingredients such as sugars, colors, carbon dioxide, and organic precipitates such as yeast. We expect this method to enable studies of a wide range of biological materials where the interaction between water and biological matter plays a role.

164

P.U. JEPSEN ET AL.

Acknowledgements We acknowledge partial financial support from the EU project TeraNova and from the Danish Research Agency. We also acknowledge valuable discussions with Cecilie Rønne about reflection spectroscopy and the dielectric properties of water. References 1. M. Walther, B. M. Fischer, and P. U. Jepsen, Noncovalent intermolecular forces in polycrystalline and amorphous saccharides in the far infrared, Chem. Phys., 288, 261–268 (2003). 2 . M. Walther, P. Plochocka, B. Fischer, H. Helm, and P. U. Jepsen, Collective vibrational modes in biological molecules investigated by terahertz time-domain spectroscopy, Biopolymers (Biospectroscopy), 67, 310–313 (2002). 3. B. M. Fischer, M. Hoffmann, H. Helm, G. Modjesch, and P. Jepsen, Chemical recognition in terahertz time-domain spectroscopy and imaging, Semicond. Sci. Technol., 20, S246–S253 (2005). 4. X. Gonze, First-principles responses of solids to atomic displacements and homogeneous electric fields: Implementation of a conjugate-gradient algorithm, Phys. Rev. B, 55, 10337– 10354 (1997). 5. X. Gonze and C. Lee, Dynamical matrices, born effective charges, dielectric permittivity tensors, and interatomic force constants from density-functional perturbation theory, Phys. Rev. B, 55, 10355–10368 (1997). 6. K. Refson and P. R. T. S. J. Clark, Variational density-functional perturbation theory for dielectrics and lattice dynamics, Phys. Rev. B, 73, 155114 (2006). 7. M. Takahashi, Y. Ishikawa, J. Nishizawa, and H. Ito, Low-frequency vibrational modes of riboflavin and related compounds, Chem. Phys. Lett, 401, 475–482 (2004). 8. Y. Chen, H. Liu, Y. Deng, D. Schauki, M. J. Fitch, R. Osiander, C. Dodson, J. B. Spicer, M. Shur, and X. C. Zhang, THz spectroscopic investigation of 2,4-dinitrotuluene, Chem. Phys. Lett., 400, 357–361 (2004). 9. D. G. Allis, D. A. Prokhorova, and T. M. Korter, Solid-state modeling of the terahertz spectrum of the high explosive HMX, J. Phys. Chem. A, 110, 1951–1959 (2006). 10. T. M. Korter, R. Balu, M. B. Campbell, M. C. Beard, S. K. Gregurick, and E. J. Heilweil, Terahertz spectroscopy of solid serine and cysteine, Chem. Phys. Lett., 418, 65–70 (2006). 11. M. D. Segall, P. J. D. Lindan, M. J. Probert, C. J. Pickard, P. J. Hasnip, S. J. Clark, and M. C. Payne, First-principles simulation: Ideas, illustrations and the CASTEP code, J. Phys.: Condens. Matter, 14, 2717–2744 (2002). 12. S. J. Clark, M. D. Segall, C. J. Pickard, P. J. Hasnip, M. J. Probert, K. Refson, and M. C. Payne, First principles methods using CASTEP, Zeitschrift F¨ur Kristallographie, 220, 567–570 (2005). 13. L. Kleinman and D. M. Bylander, Efficacious form for model pseudopotentials, Phys. Rev. Lett., 48, 1425–1428 (1982).

ELECTRONIC THz SPECTROSCOPIC DETECTION

165

14. M. C. Payne, M. P. Teter, D. C. Allan, T. A. Arias, and J. D. Joannopoulos, Iterative minimization techniques for ab initio total-energy calculations: Molecular dynamics and conjugate gradients, Rev. Mod. Phys., 64, 1045–1097 (1992). 15. B. M. Fischer and P. U. Jepsen, Dynamic range in terahertz time-domain transmission and reflection spectroscopy, Opt. Lett., 30, 29–31 (2005). 16. L. Thrane, R. H. Jacobsen, P. U. Jepson, and S. R. Keiding, THz reflection spectroscopy of liquid water, Chem. Phys. Lett., 240, 330–333 (1995). 17. C. Rønne, L. Thrane, P. O. Åstrand, A. Wallquist, K. V. Mikkelson, and S. R. Keiding, Investigation of the temperature dependence of dielectric relaxation in liquid water by THz reflection spectroscopy and molecular dynamics simulation, J. Chem. Phys., 107, 5319 (1997). 18. C. Rønne, Intermolecular liquid dynamics studied by THz-spectroscopy, Ph.D. thesis, Aarhus University (2000). 19. J. T. Kindt and C. A. Schmuttenmaer, Far-infrared dielectric properties of polar liquids probed by femtosecond terahertz pulse spectroscopy, J. Phys. Chem., 100, 10373–10379 (1996). 20. T. M. Korter and D. F. Plusquellic, Continuous-wave terahertz spectroscopy of biotin: Vibrational anharmonicity in the far-infrared, Chem. Phys. Lett., 385, 45–51 (2004).

CONTINUOUS-WAVE TERAHERTZ PHOTOMIXER SYSTEMS FOR REAL-WORLD APPLICATIONS IAN S. GREGORY*, HIDEAKI PAGE, AND LEE SPENCER TeraView Ltd., Platinum Building, St. John’s Innovation Park, Cambridge, CB4 0WS, UK

Abstract. The terahertz (THz) region is beginning to be exploited for many “real world” applications. The development of pulsed photoconductive THz generation and detection has already yielded a range of successful products, and further research into both applications and complementary THz technologies promises much more. In security screening, THz radiation has potential to image through clothing to detect concealed objects. In medical imaging, THz shows promise in both diagnosis roles and as a surgical tool. There are also many hitherto unexploited applications in both analytical spectroscopy and gas phase sensing. However, the demonstration of performance and functionality is only the first step in a product development process, which must also address commercial and physical constraints. Continuous-wave (cw) THz systems based on photomixer technology are attracting increasing interest, and appear to fulfill many of these criteria. In this chapter, an overview of this technique and practical implementation considerations for “real world” applications are discussed, with demonstrations of our cw technology and routes to future commercial development.

Keywords: continuous-wave, medical, photomixer, security, spectroscopy, terahertz, TeraView

______

* To whom correspondence should be addressed. Ian Gregory, TeraView Ltd., Platinum Building, St. John’s Innovation Park, Cambridge, CB4 0WS, UK; e-mail: [email protected]

167 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 167–184. © 2007 Springer.

168

I.S. GREGORY ET AL.

1. Introduction The last two decades have seen considerable advances in the status of THz technology. A multitude of source concepts have been demonstrated, including optical semiconductor switches,1 quantum cascade lasers,2 gas lasers,3 vacuum tubes (including backward wave oscillators,4 and free electron lasers5), electronic diodes and multipliers,6 and nonlinear optical down-conversion.7 To date, only a few of these technologies have convincingly begun to be exploited commercially. In many application areas, including pharmaceutical and chemical analysis, security, medical and nondestructive testing, many more demands are placed on a system than are required from a laboratory bench demonstrator in a research environment. Such constraints may be loosely compiled into three groups: performance, physical, and commercial. Performance constraints may be regarded as the fundamental ability of the technology to meet capability requirements, and are usually considered during “proof-ofconcept” work. The format of the specification will depend largely on the application: in security screening, for example, performance constraints will be summarized in terms of detection rates, false alarm rates, throughput, and the degree of automation. Physical constraints describe the practicality of using the technique on location, typically considering: size, weight, footprint, start up and shut-down times, reliability, servicing, maintenance and consumables requirements, and operating temperature, humidity, and vibration envelopes. Commercial constraints cover the system cost, and maintenance and operating costs (including personnel and power requirements). Rapid data acquisition, high detection rates, and low false alarm rates clearly place high demands on current technologies, given that many remain immature. In general, a policy of no consumables is also mandatory. In practice this means no cryogenically cooled quantum cascade lasers or bolometers for the potential source and detector options. Constraints often imposed on size, weight or footprint also prohibit free electron lasers and many vacuum tube sources, leaving photoconductive devices, electronic devices, and parametric oscillators amongst the candidates for immediate product development. For several years now, TeraView Ltd. (UK) has commercially supplied complete instrumentation in the form of imager and spectrometer systems, primarily for research use and for solutions in the pharmaceutical industry. All of the systems are based on a proven core system built around a femtosecond laser, using photoconductive switches for both emission and time-gated detection. In the tablet imaging application, the unique functionality arises

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

169

principally from the ability of time-domain pulses to discriminate between the various layers which comprise the coating of pills, using the time-offlight modality to measure each layer thickness and integrity.8 The spectroscopic function allows various polymorphs of active ingredients to be distinguished and compositions evaluated9: while the importance with respect to differing responses in the body is obvious – this is also of tremendous value from a patenting perspective. Not only is it possible for a pharmaceutical company to patent individual polymorphs of a substance, but the THz image data also allows internal structures to be patented, adding to the intellectual property. The pharmaceutical application is an example of an industry in which THz offers a unique solution, and thus the price points of the current technology may be tolerated. However, this is not true of many other potential application areas, particularly for high volume requirements. For example, the security industry is particularly cost-sensitive and individual instruments for threat sensing or detection are unlikely to be viable unless the purchase cost can be brought comfortably below the US $100,000 level, with similar levels present in the medical sector. This threshold would appear to be largely independent of the exact functionality of such a system, and life cycle costs must also be low. This cost sensitivity may be partly attributed to the expectation that THz technology would serve to add value to existing techniques (such as x-ray, ultrasound, or millimetre-wave imaging), rather than replace them entirely. The concept of “sensor fusion” would also allow THz components and functionality to be added to mature solutions incrementally, as the technology is developed. Established pulsed THz technology is proving to be a versatile and powerful research tool. In practice, however, the single greatest restriction to cost reduction is the requirement for a femtosecond laser. A further problem is the issue of size: not only is the Ti:sapphire laser cavity relatively bulky, but the requirement for an external power supply, separate pump-laser and closed-cycle chiller compounds the problem. For some applications, it would be preferable to remove the requirement for the femtosecond laser altogether. The possibility to use semiconductor diode lasers could immediately solve all of these problems; hence the interest in systems based on cw photomixer sources. The production cost of a single diode may be as little as 1% of that for a complete femtosecond laser. The reduction in size is a further significant factor, since the internal cavity itself is only a few hundred microns long. Thus it would seem that cw photomixer sources and detectors, driven by compact diode lasers, could be highly appropriate for many “real world” applications.

170

I.S. GREGORY ET AL.

However, these apparent advantages do come at the expense of the maturity of the technology. While femtosecond switches were demonstrated for both THz emission and detection in the 1980s, analogous results for cw photomixers were not presented until over a decade later. In 1993, a detailed proposal for such a photomixer design was published by Brown et al.10 and preliminary experiments used a dual-mode cw Ti:sapphire laser to produce narrowband radiation at 200 GHz. In 1995, the same group went on to demonstrate a photomixer that, for the first time, convincingly emitted in the THz region.11 This utilized a second cw Ti:sapphire laser, with both lasers independently tunable. The reported output power was found to be independent of frequency up to about 300 GHz and of order 1 µW, before rolling off at 12 dB per octave. A simpler arrangement based on low-cost diode lasers was demonstrated by Matsuura et al. in 1997.12 In both cases, the detection scheme based on a silicon composite bolometer was sufficiently sensitive to allow the output to be measured beyond 3 THz. It was not until 1998 that a photomixer was shown also to function as a coherent detector of cw-THz radiation. Verghese et al. demonstrated homodyne detection for the first time, using two cw Ti:sapphire lasers to illuminate both the emitter and receiver photomixers.13 The homodyne signal amplitude was constant at frequencies up to 600 GHz, before rolling off to give a signal-to-noise ratio (SNR) of just 3:1 at 2 THz. More recent work at TeraView has substantially improved upon this performance.14,15 Since this preliminary work proved the concept, further research has opened the door to both substantial performance improvements, and to examine how the technology might be implemented in a variety of application areas. At TeraView and elsewhere, the feasibility for applications in security, medical, and analytical instrumentation has been outlined, supported by funding from many interested parties. Through this work, the 10year maturity deficit to the pulsed work has been substantially reduced, despite the fact that cw technology generally receives less interest in academic circles, where engineering issues are much less important than the ability to demonstrate novel physics! 2. Overview of CW-THz Photomixer Systems 2.1. TECHNICAL SCOPE

The physical mechanism behind the generation of THz radiation in a cw device is very similar to that harnessed by existing TeraView pulsed THz technology. A near-infrared laser beam is used to modulate the conductance

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

171

of a biased photoconductive switch over THz timescales. Electrical oscillations are produced at THz frequencies, which are transmitted into free space via an antenna and silicon lens. Unlike pulsed THz systems, which use an ultrafast optical pulse to induce the THz transient, photomixing utilizes the beat frequency between two cw lasers. In this scheme, two monochromatic semiconductor diode lasers are detuned to allow their difference frequency to fall within the THz region. The beams are mixed in a photoconductive switch, forcing the conductance to oscillate sinusoidally at the THz difference frequency, and THz radiation is emitted, as illustrated in Figure 1.

Figure 1. Schematic diagram of the photomixer emitter. The carrier density in the LT-GaAs oscillates at the optical beam frequency, inducing a THz signal in the antenna, emitted in phase with the incident optical beat.

Like the optical stimulus, the THz radiation is emitted as a cw with a nanosecond coherence time, rather than a single sub-picosecond pulse. The corresponding instantaneous bandwidth is therefore very narrow, typically much less than 1 GHz. As with pulsed THz systems, detection is possible in the reverse scheme using the laser sources to modulate (gate) the sensitivity of the photoconductive detector in a similar manner. In this case, the receiver is essentially acting as a homodyne mixer, with the measured voltage given by: ∞

Vout (t ′) ∝ ∫ ETHz cos(ωt ) cos[ω (t + t ′) + ϕ ]dt , −∞

(1)

where ETHz is the amplitude of the incident THz electric field, and a variable time delay in the receiver optical beam, t’, may be introduced to recover the phase information. φ is a relative phase parameter, which is zero for the special case when the total optical path lengths of the emitter and receiver beams are equal.

172

I.S. GREGORY ET AL.

Although the role of the antenna (namely to couple the electrical oscillation in the electrodes to a free space electromagnetic wave) is identical for both pulsed and cw devices, there are different optimization conditions to consider for the semiconductor. One issue is the finite time allowed for photo-excited carriers in the switch to recombine. In the pulsed system, an 80-MHz repetition rate (associated with the femtosecond laser) allows a 12-ns interval for the switch to “reset” between pulses. In the case of cw, there is no such interval, and so the semiconductor must respond on much shorter timescales (of order picoseconds). A second issue is the very different impedance presented by the switch itself. For a pulsed and cw laser of equivalent average power, the instantaneous peak power will vary by several orders of magnitude. For a typical pulsed duty cycle of order 10,4 the peak power can easily reach many hundreds of watts during the actual pulse generation. For a cw laser, however, the peak power is precisely twice the average power, perhaps a few tens of milliwatts. This results in a corresponding difference in the instantaneous impedance of the excited switch. The cw antenna designs must account for this discrepancy, to preserve the impedance match between the THz circuit and the antenna. 2.2. CW VS. PULSED PHOTOCONDUCTIVE SYSTEMS

Comparison with the performance of a pulsed photoconductive system generally shows (perhaps a little unfairly given the aforementioned offset in maturity) a deficit in both bandwidth and SNR. It must therefore be accepted that the technical details of the system will determine the applications of the technology: indeed there are a number of experimental advantages inherent in the cw system specification. The issue of spectral resolution is an excellent example. The bandwidth of pulsed THz output is necessarily broad, since radiation must be simultaneously generated over a large frequency range to achieve the required maximum THz frequency. The upper limit of the bandwidth is determined principally by the pulse duration of the femtosecond optical pulse. While this permits broadband spectroscopy, the resolution is determined by the extent of the measurement in the time-domain, with typical limits of order 1 GHz (0.03 cm−1). In cw, however, the only frequency components permitted are the difference frequencies, defined only by the cross-correlation of the two lasing linewidths of the cw sources, which are generally very narrow. Spectra with resolutions of order 5 MHz are already achievable,16 with potential for reduction to the 100 kHz regime. At 1 THz, this could represent a resolution better than 1 part in 10.7 Thus, for applications such as low

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

173

pressure gas-phase spectroscopy, cw-THz sources should be far preferable to their pulsed counterparts. The narrow linewidth also implies a much higher spectral power density as compared to a pulsed THz system. In principle, this is expected to increase significantly the SNR for spectroscopy or imaging where observable features occur in a narrow frequency range. In addition, the quasimonochromatic nature of cw-THz systems allows all of the components of the system, (including materials, electrodes, antennae, waveguides, and detection electronics) to be optimized at one frequency. This removes much of the element of compromise, which is necessary in broadband systems. A further potential advantage of cw-THz homodyne detection, compared to the analogous optical gating technique, is the larger depth of field possible in terms of sample assessment. This is because the requirement for the THz and gating pulses to arrive simultaneously is relaxed. For the complete recovery of frequency, phase, and amplitude information, the collected interferogram requires a time delay equivalent only to one complete wavelength (although information on spectral bandwidth will be lost). This is increased substantially for the case of the pulsed detection scheme, if no loss of information is to occur. In addition, the narrow frequency range present in cw-THz beams ensures that the radiation is much more spectrally homogeneous than its pulsed counterpart, reducing unwanted beam artifacts resulting from aberration and dispersion. 2.3. COMMERCIAL SCOPE

These expected technical advantages and disadvantages aside, it is not anticipated that cw-THz technology would compete directly with pulsed equipment. The applications addressable by each technology tend to be mutually exclusive, dictated by commercial constraints and functionality. For example, the complete systems and solutions offered by pulsed spectrometer and imaging instruments such as the current TeraView products are ideal platforms with which to validate new applications and identify specific requirements. It is very unlikely that any cw products based on photomixers would encroach on this territory, at least in the near future. The photomixer technology is much more likely to address specific, well-characterized problems, where it offers cost reduction, size reduction and where custom components can be optimized. There will always be high value applications where the added functionality of pulsed THz systems acts to justify the present cost. For example, the interpretation of reflected waveforms in the

174

I.S. GREGORY ET AL.

time domain is far simpler in the case of pulsed apparatus, especially when multiple interfaces or graded refractive indices are present. 3. System Engineering Considerations A particularly prominent point on the route from a laboratory curiosity to a commercial product is the construction of a laboratory demonstrator. This milestone essentially marks the cessation of research activities, and the beginning of the product development phase. A schematic of a demonstrator for a THz imaging system based on photomixing diode lasers is shown in Figure 2: this was constructed at TeraView for proof of concept measurements in security and medical applications. While it proves the functionality of the THz components, it also serves to highlight the engineering limitations of such a setup. The use of free-space laser beams and optics allows an experiment to be rapidly built, but the requirement to precisely align the components confines them to an optical bench, or at the very least, an inflexible casting. The use of a beam-splitter to combine the beams is particularly inelegant – providing one degree of freedom with

Figure 2. Schematic diagram of a THz reflection imaging system using a photomixing emitter and detector, as arranged on an optical bench. In this configuration, beams from two diode lasers are overlapped using a beam-splitter, and a delay induced using a mechanical rapid scan delay line, operated at up to 20 Hz.

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

175

which to produce two collinear beams, at least one of which must be subsequently aligned into an optical delay retro-reflector. The arrangement is relatively unstable with respect to small changes in alignment, which may be induced by vibration, or fluctuations in the ambient temperature or humidity. The preferred solution to these problems is the adoption of fibre optic coupling, as shown in Figure 3. In comparison to the free-space arrangement in Figure 2, the optical beams are transmitted by a single-mode fibre “end-to-end” from the diodes to the semiconductor devices. Each laser beam is independently aligned to a fibre-port integrated onto the laser baseplate. The beams are combined and then split 50:50 in a fused-fibre 2×2 coupler, with the outputs phase-modulated using a piezoelectric fibre modulator system. This uses a ceramic motor to alternately stretch and compress the fibre, changing the relative optical path lengths in analogy to the mechanical systems more typically employed. The total loss of this integrated fibre assembly is less than 3 dB, providing a time delay of approximately 15 ps at scan rates of up to 100 Hz. The principal advantage of this approach is the improved stability of the optical alignment, whilst removing the requirement for an optical bench. Optical components may be physically moved with respect to each other to facilitate integration into packaging or to comply with space constraints incurred by the application. The single-mode fibre also ensures 100% spatial-mode overlap between the two frequency components, improving the efficiency of the mixing at the semiconductor, as well as protecting the lasers against optical feedback. Not least, there is also a significant improvement from a laser safety viewpoint, since the removal of open beams declassifies the laser from class 3B or class 4, down to class 1, allowing much more flexibility in the packaging, and easier commercial certification.

diode 1

emitter

22 fibre coupler piezoelectric time delay loops

Silenses

collimated THz beams

diode 2 receiver

Figure 3. Schematic diagram of the fibre-coupled optical excitation system. The nearinfrared beams are interfaced to both the lasers and the THz devices using fibre connectors, and the frequency combination and time delay is integrated into the fibre assembly.

176

I.S. GREGORY ET AL.

Finally, the introduction of fibre-optic transmission does not have the detrimental effects on the THz generation that have been observed in some pulsed (femtosecond) systems. For cw beams, the peak powers are sufficiently low to avoid nonlinear artifacts, and dispersion is not a problem since the cw linewidths are very narrow. The termination of the fibre is integrated into the device “cartridge” that allows the output to be precisely aligned to a lens, matching the optical mode to the device active area, as is illustrated in Figure 4. The ability to use standard FC/PC or APC connectors to join fibers underlines the modular approach, allowing ease of interchange of devices (perhaps for different applications or frequency ranges). The enclosed device is robust with respect to handling and static discharge, and requires no translation stages to use.

Figure 4. Transverse cross-section of the fibre-coupled device interface. The precise alignment of the fibre ferrule, optical lens, THz device, and silicon lens may be factorylocked to produce a robust cartridge.

4. Application Examples 4.1. SECURITY APPLICATIONS

As previously outlined, the security industry requires low cost, high volume production. Provided that sufficient functionality can be demonstrated,17 cw photomixing technology may well be a suitable candidate. This section summarizes two proof-of-principle application examples appropriate to the detection of nonmetallic hard objects and plastic explosives. Figure 5 shows a cw-THz reflection image of a concealed ceramic knife blade captured at a single frequency using the photomixer imaging system in Figure 2. The image demonstrates that the single-frequency beam (here at 0.4 THz) is able to penetrate the barrier (twice) and produce a high-quality image, with much higher spatial resolution than typical millimeter-wave images.

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

177

Figure 5. Visible photograph (left) and cw-THz image (right) of a ceramic knife hidden behind a denim cloth (to simulate the detection of a weapon in a pocket). This image was taken at relatively close range (about 20 cm from source) and by sequentially raster scanning pixels over a period of a few minutes.

Figure 6 shows the use of a similar arrangement to identify a test sample of Semtex plastic explosive, amongst confusion materials of similar consistency and density. The detailed THz spectroscopic characterization of high explosive and confusion materials is described elsewhere18 – the purpose of this demonstration is to show that spectroscopic imaging is possible with a cw system tuned to a small number of discrete frequency values. The three samples were mounted on a quartz window, and then imaged through a cloth barrier at a variety of single frequency values. The image shown in Figure 6 was derived from functions of the difference between reflection images taken at 0.5 and 0.9 THz. Since RDX, an active constituent of Semtex, has a spectroscopic feature just below 0.9 THz, the difference image clearly shows an enhanced contrast for this sample, in comparison to the inert samples. Essentially, this technique also acts to

Figure 6. Images of a test sample containing Semtex plastic explosive (A), and example confusion materials gum (B) and blu-tack (C). The sample shown in the visible photograph (left) was imaged at two discrete cw-THz frequencies, and the composite difference image (with components at 0.5 and 0.9 THz) is shown to the right.

178

I.S. GREGORY ET AL.

normalize the image with respect to system fluctuations and attenuation through uncharacterized barriers. Thus, given a priori knowledge of the spectral features present in the substance detected, threshold parameters can be assigned to optimize the trade-off between a low false alarm rate and a high detection rate. The artifacts at the edges of the confusion materials occur because the lateral resolution of the two diffraction-limited images is offset on account of the frequency difference. Thus this particular example also highlights the added value of an image in addition to spectroscopic information. Although the extension of such techniques to stand-off applications at larger distances is not trivial, cw-THz systems do offer some advantages over the pulsed time-gated technology. For example, for a sufficiently narrow linewidth, the coherence length of the generated THz wave can be several tens of meters. This allows a collimated beam to “search” simultaneously over a large range of distances, without the requirement to “gate” the target distance to equalize the path length. The narrow linewidth also allows high spectral powers – particularly effective since all of the available power can be targeted at specific frequencies where the rangelimiting effects of water vapor absorption can be overcome. 4.2. MEDICAL IMAGING AND CANCER CHARACTERIZATION

Another important and previously characterized application that is expected to benefit from cost-reduced technology is medical imaging. Previous studies have shown that, under controlled conditions, THz radiation might be used to characterize cancerous tissue.19 Historically, interest has focused on skin cancers, since they are easily accessible and relatively well defined. However, other cancers may also be addressed by THz technology – many having much higher mortality rates than treated cancers of the skin. Since the high water content of the human body forms an effective barrier to THz radiation, one has to either use an endoscope to probe tumours on tissue– airspace interfaces inside the body cavities, or access the tumour during surgery. Clearly, there are significant and specialized engineering requirements in each case. Conservative breast surgery is one possible application of a medical THz probe to characterize tissue in situ.20 In this case, the purpose is not to diagnose cancer, but a surgical tool to discriminate between diseased and healthy tissue. A probe to give instant feedback to the surgeon would be invaluable – reducing the likelihood of either removing insufficient tissue (which would require the patient to return for further surgery) or removing far too much tissue (causing unnecessary disfigurement, and in extreme

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

179

cases, the loss of the breast). The majority of the characterization work has been carried out using femtosecond pulsed systems, using various parameters relating to the reflected pulse shapes to define the discrimination thresholds. A large tissue database, compiled by TeraView and partners, has shown that THz radiation can be used to correctly distinguish tissue types in up to 90% of cases, under controlled conditions and with carefully chosen references. However, cost reduction is generally seen as a necessary condition for any acceptance of THz technology in this area. When imaging using a single frequency, or combination of multiple discrete frequencies,21 the discrimination parameters more conveniently fall in the frequency (as opposed to time-) domain. Figure 7 shows characterization of healthy and diseased tissue using Fourier transformed timedomain data from a pulsed THz system. Each curve represents the average of many pixels of the respective tissue type. On a log-linear plot, the difference in gradient may form the basis of the tissue-type classification. Inset are images of one such excised sample used in the study. (A) shows a visible photograph of the sample, with a dotted curve overlaid to delineate the cancerous (light) and healthy (dark) regions. This same dotted curve is overlaid onto the THz image, (B), taken at a single frequency of 0.5 THz

Figure 7. The reflected THz power as a function of frequency, for both diseased (solid curve) and normal (dashed curve) tissue. Inset is an example of a tissue image: (A) is a visible photograph of the excised tissue, and (B) is a cw-THz image taken at 0.5 THz.

I.S. GREGORY ET AL.

180

using the cw photomixer system. The pixel correlation is convincing, but closer inspection reveals that the diseased area extends beyond the visible boundary in the THz image (white arrow), indicating the increased subsurface lateral extent of the tumor. 4.3. ANALYTICAL SPECTROSCOPY AND GAS SENSING

The final application example concerns the use of a tunable cw photomixer system to produce continuous spectra – vital for analytical spectroscopy and high resolution gas-sensing applications. The principle here is that although cw-THz emission and detection is instantaneously narrowband, the working frequency can be tuned across a frequency range of interest to acquire a spectrum sequentially in the frequency domain. This “sweeping” technique here utilizes distributed feedback (DFB) diode lasers, which have a wide mode-hop-free tuning range accessed by varying the temperature of one or both of the diode cavities.22 The changing cavity length causes a corresponding change in the optical lasing frequency, which directly changes the difference frequency between two such diodes.

1.5

1.0

Absorption (Offset for clarity)

Absorption coefficient (decadic)

2.0

Semtex PE-4 RDX PETN HMX TNT

0

1 2 3 Frequency (THz)

0.5

4

RDX

0.0 0.0

0.2

0.4

0.6

0.8

1.0

Beat frequency (THz)

Figure 8. Absorption spectrum of RDX plastic explosive, measured in a transmission geometry by sweeping the DFB difference frequency between 0.1 and 0.95 THz. The broad absorption band at 0.8 THz is evident, and is consistent with previous measurements made with TeraView’s pulsed spectrometer (inset).

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

181

Figure 8 shows the absorption spectrum of a pellet containing RDX plastic explosive, measured by cw photomixers using this temperaturetuning technique. The absorption was calculated by dividing the transmission measurement by a reference measurement taken in the absence of a sample. Comparison with TeraView’s original characterization work on explosives using a pulsed spectrometer18 (inset) yields a good agreement. Figure 9 shows spectra taken of water vapor in the laboratory atmosphere using the same system. 2.0 57.196 38.679

Absorption (decadic)

1.5

58.739

36.622

1.0

47.012

18.517 24.998

0.5

0.0 0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

1.8

2.0

2.2

2.4

Frequency (THz)

Figure 9. Absorption spectra of laboratory water vapor acquired using the cw temperature-sweeping technique. The arrows show direct correlation with known absorption lines, with their catalogued spectral positions quoted in units per centimeter.

The absorption lines are well known, and the wavenumber of principal catalogued features are marked on the plot, per centimeter. Good agreement is seen, both for the relative positions of the features, and also their magnitudes. The absolute position of the lines could be used to calibrate the frequency axis of the photomixer spectrometer. The resolution offered by the spectrum in Figure 9 is of order 1 GHz, which is slightly better than is routinely achieved using a pulsed FFT spectrometer, and limited by the choice of step size in the frequency sweep. With the cw spectrometer, the resolution limit of the “sweep” is solely defined by the individual laser linewidths, and is independent of the time-domain acquisition parameters. Figure 10 shows a subset of the same spectrum measured with a 200 MHz

I.S. GREGORY ET AL.

182

sampling interval, clearly revealing the feature at 36.622 cm−1 to be a doublet. This figure also demonstrates the ability of the cw swept system to interrogate only the required subset of the whole frequency range accessible by the instrument, cutting acquisition times by collecting only the required data. 1.0 38.679

Absorption (decadic)

0.8

36.622

0.6

0.4

0.2

0.0 1.04 1.06 1.08 1.10 1.12 1.14 1.16 1.18 1.20 1.22 1.24 Frequency (THz)

Figure 10. Water vapor spectrum acquired at an increased resolution of 200 MHz. The feature at 36.622 cm−1 is resolved into a doublet, with a satellite feature at higher energy.

5. Conclusions In this chapter, we have overviewed the principles and potential of cw-THz emission and detection using photomixers. It is shown that the technique fulfills many of the performance, physical and commercial constraints imposed by the applications under consideration, and the functionality is demonstrated in three example areas: security screening, medical surgical tools, and analytical gas spectroscopy. TeraView’s existing pulsed technology offers complete systems and solutions, and provides a research platform with which to evaluate applications, and there are many high value applications in which pulsed technology is a long-term solution. Therefore cw technology complements existing technology. In particular, we have emphasized that although the cw photomixer approach is very versatile in terms of addressable applications, the full performance can only be realized when the system is optimized exclusively

TERAHERTZ PHOTOMIXER SYSTEMS AND APPLICATIONS

183

for the application of choice. The modular approach to the component engineering outlined in this chapter lends itself to this optimization process, and allows component supply on a “plug and play” basis, facilitating development work with partners and customers. Further development of both hardware and software will allow increased sophistication, allowing the performance to be iteratively improved to the requirements of operational systems. Acknowledgments Aspects of this work were carried out under contract for the UK Government, the European Union (TeraNova), and the UK Department of Trade and Industry (project code TP/2/SC/6/S/10313), whose support is gratefully acknowledged.

References 1. P. R. Smith, D. H. Auston, and M. C. Nuss, Subpicosecond photoconducting dipole antennas, IEEE J. Quantum Electron., 24, 2, 255–260 (1988). 2. R. Köhler, A. Tredicucci, F. Beltram, H. E. Beere, E. H. Linfield, A. G. Davies, D. A. Ritchie, R. C. Iotti, and F. Rossi, Terahertz semiconductor-heterostructure laser, Nature, 417, 156–159 (2002). 3. Coherent 5100 Patrick Henry Drive, Santa Clara, CA 95054, USA. www.cohr.com. 4. 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, 30, 5637–5646 (2004). 5. G. L. Carr, M. C. Martin, W. R. McKinney, K. Jordan, G. R. Neil, and G. P. Williams, High power terahertz radiation from relativistic electrons, Nature, 420, 153–156 (2002). 6. G. Chattopadhyay, E. Schlecht, J. S. Ward, J. J. Gill, H. H. S. Javadi, F. Maiwald, and I. Medhi, An all-solid-state broad-band frequency multiplier chain at 1500 GHz, IEEE Trans. Microw Theory Tech., 52, 5, 1538–1547 (2004). 7. K. Imai and K. Kawase, A frequency-agile terahertz-wave parametric oscillator, Optics Express, 8, 13, 699–704 (2001). 8. A. J. Fitzgerald, B. E. Cole, and P. F. Taday, Nondestructive analysis of tablet coating thicknesses using terahertz pulsed imaging, J. Pharm Sci., 94, 1, 177–183 (2005). 9. C. J. Strachan, P. F. Taday, D. A. Newnham, K. C. Gordon, J. A. Zeitler, M. Pepper, and T. Rades, Using terahertz pulsed spectroscopy to quantify pharmaceutical polymorphism and crystallinity, J. Pharm Sci., 94, 4, 837–846 (2005). 10. E. R. Brown, F. W. Smith, and K. A. McIntosh, Coherent millimeter-wave generation by heterodyne conversion in low-temperature-grown GaAs photoconductors, J. Appl. Phys., 73, 3, 1480–1484 (1993).

184

I.S. GREGORY ET AL.

11. E. R. Brown, K. A. McIntosh, K. B. Nichols, and C. L. Dennis, Photomixing up to 3.8 THz in low-temperature-grown GaAs, Appl. Phys. Lett., 66, 3, 285–287 (1995). 12. S. Matsuura, M. Tani, and K. Sakai, Generation of coherent terahertz radiation by photomixing in dipole photoconductive antennas, Appl. Phys. Lett., 70, 5, 559–561 (1997). 13. S. Verghese, K. A. McIntosh, S. Calawa, W. F. Dinatale, E. K. Duerr, and K. A. Molvar, Generation and detection of coherent terahertz waves using two photomixers, Appl. Phys. Lett., 73, 26, 3824–3826 (1998). 14. I. S. Gregory, C. Baker, W. R. Tribe, I. V. Bradley, M. J. Evans, E. H. Linfield, A. G. Davies, and M. Missous, Optimization of photomixers and antennas for continuouswave terahertz emission, IEEE J. Quantum. Electron., 41, 5, 717–728 (2005). 15. I. S. Gregory, W. R. Tribe, C. Baker, B. E. Cole, M. J. Evans, L. Spencer, M. Pepper, and M. Missous, A continuous-wave terahertz system with a 60 dB dynamic range, Appl. Phys. Lett., 86, 204104 (2005). 16. G. Mouret, S. Matton, R. Bocquet, F. Hindle, E. Peytavit, J.F. Lampin, and D. Lippens, Far-infrared cw difference-frequency generation using vertically integrated and planar low temperature grown GaAs photomixers: application to H2S rotational spectrum up to 3 THz, Appl. Phys. B, 79, 6, 725–729 (2004). 17. C. Baker, W. R. Tribe, T. Lo, B. E. Cole, S. Chandler, and M. C. Kemp, People screening using terahertz technology, Proceedings Of SPIE, 5790, 1–10 (2005). 18. Y. C. Shen, T. Lo, P. F. Taday, B. E. Cole, W. R. Tribe, and M. C. Kemp, Detection and identification of explosives using terahertz pulsed spectroscopic imaging, Appl. Phys. Lett. 86(24), 241116 (2005). 19. E. Pickwell, A. J. Fitzgerald, B. E. Cole, P. F. Taday, R. J. Pye, T. Ha, M. Pepper, and V. P. Wallace, Simulating the response of terahertz radiation to basal cell carcinoma using ex vivo spectroscopy measurements, J. Biomed. Opt., 10 6, 064021 (2005). 20. A. J. Fitzgerald, V. P. Wallace, M. Jimenez-Linan, L. Bobrow, R. J. Pye, A. D. Purushotham, and D. D. Arnone, Terahertz pulsed imaging of human breast tumors, Radiology, 239, 2, 533–540 (2006). 21. I. S. Gregory, W. R. Tribe, M. J. Evans, T. D. Drysdale, D. R. S. Cumming, and M. Missous, Multi-channel homodyne detection of continuous-wave terahertz radiation, Appl. Phys. Lett., 87, 034106 (2005). 22. S. Kraft, A. Deninger, C. Truck, J. Fortagh, F. Lison, and C. Zimmermann, Rubidium spectroscopy at 778–780 nm with a distributed feedback laser diode, Laser Phys. Lett., 2 2, 71–76 (2005).

Theme 4 SYSTEMS FOR SECURITY

SYSTEMS REQUIREMENTS FOR A MULTI-CHANNEL TERAHERTZ CONTRABAND SCANNER WILLIAM S. TRUSCOTT* School of Electrical and Electronic Engineering, University of Manchester, PO Box 88, Manchester, M60 1QD, UK

Abstract. The direct use of terahertz (THz) spectral features in screening for explosives, drugs, and other contraband is limited since adjacent materials may obscure the signature of a suspect material. This paper describes a system that overcomes this problem by combining tomographic imaging with THz spectroscopy. This paper identifies a set of 30-GHz-wide windows in the atmospheric water vapour absorption up to 3 THz: data collected in these bands permit the reconstruction of an object with 1-cm resolution. Images of the distribution of innocuous and suspect material within the object are generated by an algorithm providing an optimised filter. The spectral characteristics of any region within the object may also be examined. Practical issues associated with implementing this system are discussed.

Keywords: submillimetre wave imaging, submillimetre wave spectroscopy, tomography, security, object detection, explosions, drugs of abuse, terrorism

1. Introduction This paper examines the problem of using the particular terahertz (THz) spectral characteristics of a substance that an authority wishes to identify, including both drugs of abuse and potential explosives, when it is concealed among other materials that obscure the identifying spectral features. A solution is proposed that combines tomographic techniques for object reconstruction with data-filtering techniques based on the knowledge of the

______ *

To whom correspondence should be addressed: W.S. Truscott, School of Electrical and Electronic Engineering, University of Manchester, PO Box 88, Manchester, M60 1QD, UK; e-mail: [email protected]

187 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 187–204. © 2007 Springer.

188

W.S. TRUSCOTT

THz responses of the suspect material and the ensemble of other materials likely to be encountered in that screening situation. The system provides the operator with images of the distribution of innocent and questionable materials using an algorithm to provide a filter that is optimised for probabilities of encountering different materials. The operator may also view the local THz spectrum of any region identified as suspect to provide further information on which appropriate action may be based. The paper presents both the concept of the system and conducts a highlevel systems analysis to demonstrate its feasibility. Among the issues examined are the power budget, which dictates that the receivers should be wide aperture cameras rather than low gain antennas, and the effects of atmospheric absorption of THz signals which are overcome by the use of selected frequencies at which absorption is sufficiently low. The paper also discusses the factors which will determine the most favourable technology to use in the generation and detection of the THz radiation. 2. Problem Definition Reflection and transmission spectra of many materials, including both explosives and psychoactive drugs, were among the early results published by the developers of THz technologies based on femtosecond pulsed lasers and related excitation techniques.1–3 While some features of these early results are now understood to be experimental artefacts, the primary conclusion of these studies, specifically that it is possible to identify an unmixed block of any one of these materials by its THz spectrum, is generally accepted. The problem in applying this understanding to the identification of these materials in real world situations stems primarily from the relatively broad spectral lines characteristic of solids at room temperature. One major and unavoidable source of this broadening is the anharmonicity of the effective potentials for the associated molecular motions. Because the quanta of THz energy are significantly smaller than thermal energies these vibrational modes are highly excited. The consequent breadth of the lines results in significant overlaps between the spectral features of materials of official interest and those of other materials which might commonly be encountered in screening situations. The observation, in reflectance or transmission, of the overall THz spectrum of a postal packet, baggage item, or human being, is unlikely to indicate with the required level of confidence the presence or absence of one of a list of suspect materials. A combination of the imaging and spectroscopic capabilities of THz radiation has the potential to give a solution to this

SYSTEM REQUIREMENTS FOR A SCANNER

189

challenge. This paper addresses many of the issues that will be of significance in the development of such a solution. 3. Systems Concept A two-stage identification process for a particular material is presented in this paper. The complexities of this process stem from the difficulty of using overall spectra to establish the presence or absence of a particular material. The first stage of this concept identifies regions of potential interest by a method that has parallels with other techniques for clustering such as Principal Component Analysis. The second stage is equivalent to a detailed spectroscopic survey of those selected areas. In the first stage of the concept, the system forms two 3D THz images of the collection of objects under scrutiny. A “background” image is constructed using conditions that minimise the contribution of the material of particular interest to the strength of the image. A second, “indicative”, image is constructed using conditions that maximise the contribution of that material in comparison with other materials that are likely to be encountered in that situation. The concept is to identify specific materials through their THz spectra therefore the background and indicative images will be formed using different combinations of image data obtained at a selected set of wavelength ranges. Regions that feature strongly in the indicative image in comparison with the background image are those most likely contain the material in question. In the second stage of the identification process, the THz spectra of these suspect regions will be presented to the operator in a form that allows the appropriate decision and subsequent action to be taken. 4. Reflection and Scattering of Electromagnetic Waves An electromagnetic wave will propagate uniformly through a medium with constant dielectric properties. Reflection and scattering will occur at any change in the dielectric properties, whether these are due to interfaces between materials or to fluctuations within a material. Certain solids, particularly crystalline materials, have absorption lines in their THz spectra; these are typically 200 GHz full width at half maximum at room temperature3. These lines are generally weak (∆εr"~ 0.1) and are associated through Kramers-Kronig relationships with dispersive changes in the real part of the permittivity. A consequence of these variations is that the reflectivity of a plane material–air interface will increase in the region of the resonance before falling to a slightly lower value at frequencies above

190

W.S. TRUSCOTT

the resonance. Amorphous and liquid materials have dielectric losses that increase smoothly with frequency although the real part of their permittivity is generally frequency independent. Reflections at interfaces with these substances will therefore show a weak monotonic increase with frequency. In many common materials the fluctuations in permittivity will be large because the material is constructed from an arrangement of subcomponents intermixed with air, for example, woven from threads which are themselves spun from fibres. Because these structures have submillimetre characteristic length scales, the amplitude of the scattered THz wave will be frequency dependent, normally with a monotonic increase in scattering with frequency. In quasi-bulk samples of these materials, such as folded clothing, multiple scattering will lead to diffuse reflection with a penetration depth that decreases with increasing frequency. To a first approximation, a single frequency THz beam incident on an object from a reasonably distant source will have a plane wavefront at the object. In a similar way any scattered THz radiation collected by a reasonably distant detector may also be approximated as the result of a plane wave travelling in the direction of the detector. If a number of dielectric fluctuations or discontinuities are within the object then each will contribute by scattering or reflection to the wave directed towards the detector. The amplitude at the detector will be given by the phasor sum of all the scattering amplitudes. In the absence of multiple scattering and absorption, both of which may need to be considered in a real THz measurement, the phase factor for any dielectric perturbation is determined by the sum of the perpendicular distances between that perturbation and each of two reference wave fronts, one normal to the direction of the source and the other normal to the direction of the detector. Simple geometry shows that this phase factor is equal to 4πdsin(θ)/λ, where d is the projection of the distance between the perturbation and the intersection of the two reference planes onto the bisector of the angle between the directions towards the source and detector, and θ is half the angle between the two reference wave fronts. This analysis is very similar to that for the Bragg reflection of x-rays. The above analysis indicates that an observation of the amplitude and phase of scattered THz radiation at a particular frequency and angle measures the Fourier transform of the scattering amplitudes projected onto the bisector of the scattering angle with an effective wavelength λ/2sin(θ). If the frequency of the THz radiation is varied in N steps of δf over a narrow range from f0 − Nδf/2 to f0 + Nδf/2, then the observed frequency dependence of the resultant scattering amplitude and phase may be inverse

SYSTEM REQUIREMENTS FOR A SCANNER

191

Fourier transformed to give the variation along the bisector of the projected scattering intensity. The spatial resolution is equal to c/[Nδfsin(θ)] and the range covered by the projection without aliasing is given by c/[δfsin(θ)]. As a specific example, if the frequency is stepped over 30 GHz in steps of 0.3 GHz then the transformed data will cover a minimum projected length of 1 m with a resolution of 1 cm. A series of measurements using P sources and Q detectors can give a set of projections of the overall dielectric properties of the material within the volume being examined along PQ different axes. When the reflected THz signal has a component due to multiple scattering, that part will generally be delayed with respect to one reflected at the surface of the material and the delay will show significant fluctuations over relatively narrow frequency ranges. Materials that show diffuse reflection will appear shifted away from the source and detector in any inverse transform of the frequency dependence of a received signal and will also be dispersed over a greater distance than expected from the depth at which the diffuse reflection occurs. 5. Tomography Tomography is the science of reconstructing 3D images of objects from a set of projections. In the most favourable examples, such as medical imaging, there will be at least as many data points in the observed set as in the reconstructed image. In many other examples of tomography, particularly in industrial applications, the knowledge of the probable form of the scattering objects is used to reconstruct an image with more pixels than elements in the original data-set. If the requirement for a particular system was to provide an image with 1% resolution in all three dimensions, for example, 1-cm resolution over a volume of 1 m3, then the reconstructed image would have 106 data points. Measurements made with a set of 8 sources and 16 detectors that gave 128 independent projections, each with 100 data points could be used to generate such a reconstruction, provided the number of independent scattering surfaces within the volume was reasonably small. 6. Major Considerations 6.1. FREQUENCY RANGE FOR IDENTIFICATION

Several papers within this volume, together with others in the general literature have noted that there is a broad pattern of agreement between

192

W.S. TRUSCOTT

different researchers on the principal THz spectral features in reflection or transmission of a number of explosives and psychoactive drugs. Studies on a range of materials such as saccharides4 and theophylline5 indicate that the spectral lines, which are presumed to be associated with k ≈ 0 transverse optical phonons, are dependent on the morphology and crystallinity of the material. These spectra show relatively wide absorption lines, with a full width at half maximum that can be as large as 0.3 THz. The frequency range typically covered by the published spectra is 0.3–3 THz; this is sufficient to allow the clear identification of the pure samples of the materials of interest. This range will be assumed in the development of the concept in the rest of this paper. The spectral range required by a real system is highly dependent on the context in which it will be used. A balance needs to be drawn between the increased certainty of identification resulting from an increment in the spectral range and the increased cost associated with this benefit. The systems engineer must also consider whether an equivalent gain can be achieved at a lower cost by combining the information obtained using THz imaging and spectroscopy with that obtained from other modalities. 6.2. ATMOSPHERIC ABSORPTION OF TERAHERTZ (THZ) RADIATION

The absorption of THz radiation by water molecules in the atmosphere is one of the greatest constraints in the use of THz imaging and spectroscopy in practical applications, particular in outdoor environments.6, 7 While the strongest absorption is over a narrow range of frequencies, typically 2 GHz, around the resonant frequency of each transition, the line shape is Lorentzian which implies that the influence of the strongest absorption lines can extend over many tens of GHz. Table 1 shows the frequency ranges of transmission windows at least 30 GHz broad above 500 GHz. These have been calculated assuming a particular mass of water vapour at atmospheric pressure and 27°C from the published line frequencies and intensities for the water molecule.8 They are intended to be representative of the amount of absorption that might be expected in a “typical” indoor material detection application. The windows shown in Table 1 are of great significance in any THz spectroscopic system operating in a relatively uncontrolled atmosphere. It is noteworthy that only one of the windows above 1 THz is more than 30GHz wide. The minimum transmission in all the other windows above this frequency will decrease rapidly if spectral measurements are made over a broader bandwidth. This implies that the maximum spatial resolution of an

SYSTEM REQUIREMENTS FOR A SCANNER

193

image constructed from data collected within a single window is 1 cm. If the system requirements were for a better spatial resolution, then it would be necessary to control the water vapour mass within the imaging path. Another issue associated with these windows is that there are some significant frequency gaps between certain windows; the largest of these is 270 GHz between the window centres at 1,560 and 1,830 GHz and there are three other gaps above this window with widths around 200 GHz. A material that had characteristic absorption features 200-GHz wide at 1,695, 2,200, 2,405, and 2,615 GHz would be extremely difficult to identify because only the wings of the lines would lie in the available windows. TABLE 1. Windows above 500 GHz in the transmission through the atmosphere in the presence of water vapour Minimum frequency GHz

477

585

627

790

923

1,010

Maximum frequency GHz

515

615

713

903

957

1,040

Maximum transmission

0.99+

0.99

0.99+

0.99+

0.99

0.97

Minimum transmission

0.99

0.97

0.99

0.99

0.98

0.96

Minimum frequency GHz

1,245

1,347

1,463

1,545

1,815

1,972

Maximum frequency GHz

1,275

1,377

1,530

1,575

1,845

2,002

Maximum transmission

0.96

0.97

0.98

0.96

0.90

0.96

Minimum transmission

0.90

0.94

0.97

0.85

0.86

0.94

Minimum frequency GHz

2,093

2,283

2,500

2,705

2,822

2,905

Maximum frequency GHz

2,124

2,313

2,530

2,735

2,852

2,935

Maximum transmission

0.92

0.85

0.91

0.79

0.84

0.82

Minimum transmission

0.90

0.69

0.87

0.73

0.81

0.76

6.3. POWER BUDGET

The tomographic process described above has similarities with radar imaging; the radar equation can therefore be used to find the ratio of the received power to the transmitted power:

W.S. TRUSCOTT

194

Pr Gt Gr λ2 σ . = (4π ) 3 R 4 Pt In the case of a system with antenna gains, Gt, Gr of 10 dB, operating at a wavelength, λ, of 200 µm and an object cross section, σ, of 1 cm2, this equation gives:

Pr 100.4 ×10−8. 10 − 4 − = = 2.5 ×10 15 (4π )3.81 Pt for an object distance of 3 m. This calculation shows that a radar system could not operate under these conditions without thermal damage to the object being imaged. The reason for this is the extremely small area required by a receiving antenna with a gain of 10 dB at a wavelength of 200 µm, approximately 1 mm2. A similar calculation applied to the eye would indicate that human vision was not feasible. A receiver with a reasonable aperture, for example, 50-mm diameter, would have, at these wavelengths, a very narrow beam, ~0.2, and a correspondingly high gain. In order to make use of a wide aperture while simultaneously detecting the radiation from across the whole object, an array of detectors is required in the focal plane of an objective that defines the entrance aperture to the detecting system. The appropriate equation to use in this case is that for a camera:

B=

IR 8f 2

for an illumination, I, of 10−3 Wm−2; a reflectance, R, of 0.1; and a system with an f number of 4; the image brightness, B, will be:

B=

10−3 × 0.1 = 8 ×10 −7 Wm−2. 8 × 42

This is a plausible power density for measurement by heterodyne methods. 7. Development of System Solution The considerations described above are sufficient to form an outline of the system that is proposed in this work. The object to be investigated is surrounded by an array of sources and detectors distributed over a range of azimuths and elevations so as to give data that corresponds to a sufficient set of independent projections to enable the spatial variation of the scattering

SYSTEM REQUIREMENTS FOR A SCANNER

195

amplitude across the object space to be reconstructed. The sources and detectors operate across the 30-GHz range of each of the set of frequency bands selected from Table 1 so as to give a clear distinction between the materials to be identified and their likely surroundings. The Fourier transforms of the received signal amplitude and phase relative to a standard calibration give projections of the scattering amplitudes onto axes determined by the location of the source and detector. These projections are then combined using appropriate tomographic inverse transforms to construct images related to the scattering amplitude at different locations in the object at the selected frequency band. Two images are generated by combining those from different frequency bands: the first, indicative image is a combination of the images filtered in such a way that regions made of the any innocuous material likely to be encountered in the application will have a minimal presence whereas regions that contain the suspect material have the greatest amplitude. The filter generating the second image is arranged so that it accepts regions with the spectral characteristics of all commonly encountered materials but rejects those containing a significant fraction of the suspect material. Regions composed of the latter should not be present in the second image in contrast to regions composed of any other likely material. Regions of the image space that show significantly more strongly in the first image than the second are most likely to be composed of the material in question. The scattering amplitude of that volume can be displayed for the set of frequency bands and compared with that of the material in question to establish the probability of the identification. 8. Dependence of Images on Frequency Band The proposed system operates over a wavelength range of at least a factor of six and possibly ten from a minimum of 100 µm to at least 600 µm and possibly 1 mm. In each band the resolution of the reconstructed image will be determined by the range of frequencies used; if this is the same for each band then the resolution of the projections will also be the same, for example, at 1 cm if a bandwidth of 30 GHz is employed. If a region has a structure on a scale of, for example, 300 µm then the images at the different frequency bands will be substantially different because of the change in the scattering behaviour of each 1-cm³-volume element as the wavelength changes from much smaller than the structure to much larger than the structure. Cotton thread is one material that has a structure on this scale: in the lower frequency bands clothing will act as a uniform material reflecting a portion of the radiation and allowing the rest to propagate without

196

W.S. TRUSCOTT

distorting the wavefront but in the highest frequency bands the same clothing will act as a diffuse scatterer changing the shape of the wavefront and resulting in a very different reconstructed image for the same region. The algorithms used to combine the images will have to accommodate this complexity; there is a necessity for a considerable amount of experimental work to be undertaken using a system prototype in order to obtain a sufficient understanding of this problem for it to be modelled and optimal imaging filters to be developed. 9. Sources and Detectors 9.1. SWEPT VS. PULSED ILLUMINATION

The system concept requires the amplitude and phase of the scattered signal to be recorded over a 30-GHz bandwidth with a frequency resolution of 0.3 GHz in each of the selected frequency bands. There are four principal methods of achieving this coverage; the first is to illuminate the object with a series of identical pulses, each with Fourier components covering the 30GHz band, implying a pulse length around 30 ps, with a minimum period between pulses of 3 ns. The second is to use a technique similar to orthogonal frequency division multiplexing (OFDM, a technique employed in terrestrial digital television) in which the illuminating source generates the necessary set of Fourier components which are then combined to give an output whose envelope has a time dependence that is determined by the phase relationships of the separate components. This envelope would repeat with a period, 3 ns, equal to the inverse of the frequency spacing, 0.3 GHz. The third method is for the illuminating source to generate a wave of constant amplitude whose frequency is swept smoothly across the 30-GHz band at a rate that allows the independent measurement of the received amplitudes at frequency steps of 0.3 GHz. In the final method a wave of constant amplitude is generated for illumination; the frequency of this is stepped discretely by 0.3-GHz intervals over the required 30-GHz band. The latter two methods are very similar; both have a minimum period for each sweep of 300 ns. Heterodyne receivers will be employed, these reduce the effective frontend bandwidth and hence the noise level. In the first two schemes the receiver will require an overall bandwidth of 30 GHz, which could be achieved by a set of parallel IF channels, each of which is digitised and transformed to obtain the received amplitude at each frequency point within the bandwidth. If the sequence of pulses has long-term phase coherence

SYSTEM REQUIREMENTS FOR A SCANNER

197

then the bandwidth at the final detector can be very narrow, as can be in the case of the quasi-OFDM system, since the individual Fourier components will be phase coherent over long time scales. The overall IF bandwidth would be much lower in the cases where the frequency is swept or stepped since the LO frequency can follow the source frequency. The narrowest IF bandwidth can be used when a stepped frequency is employed with a guard interval that allows for the variation in time for the scattered wave. 9.2. SOURCE AND DETECTION TECHNOLOGY

The selection of the most satisfactory scheme to generate and detect the THz waves requires a careful balance of many factors in the light of the available technologies. The need to generate THz waves with controlled spectral characteristics in a series of frequency bands favours generation and detection techniques in which the THz frequency is determined by the beating of two lasers, which are either fed through a modulator to achieve the envelope associated with the first two schemes, or can be wavelength modulated in the latter schemes. If the higher bandwidth receivers required for the first two schemes are both technological feasible and affordable, then it is likely that the scheme that gives the highest average radiated power will be the most effective. It is known that THz sources based on semiconductor materials with high trapping rates rapidly saturate which makes them unsuitable for CW generation. It is possible that moderate long pulses can be generated at regular intervals with a higher average power than any of the other schemes, in which case this is likely to be the optimum approach. Since both dielectric lenses and optically stimulated detectors can be employed over the range of THz frequencies from 0.3 to 3 THz, it is possible to use one camera at each receiver location across the set of frequency bands. If the focal plane array is structured to perform well at the highest frequencies, then the density of receiving elements will be much higher than the resolving power in the lowest bands and the receivers will be less well matched to the incoming waves. The inevitable frequency dependence of the receiver system adds further complexity to the image reconstruction algorithms, but with the benefit that the receivers will, because they have some imaging capability, provide many more independent data-sets than a single channel receiver would.

198

W.S. TRUSCOTT

9.3. CALIBRATION

The image generation mechanism in the system concept is to invert a large set of projections that have been obtained by inverse Fourier transforms of frequency-swept data. The relative positions in space of these projections are determined by the phases of the signals received at the different detectors. A phase calibration procedure is required in order to achieve the correct alignment of the projections in real space and enable the proper reconstruction of the original object. One method of calibration that will enable the phase and amplitude characteristics for each projection to be recorded for all the bands is to suspend, at the centre of the volume to be imaged, a stationary conducting metal sphere with a diameter at least ten times the longest wavelength. The phases observed in this calibration procedure should be corrected for the radius of the sphere in order to establish a zero phase corresponding to reflection at its centre. All reconstructions made with data that is measured relative to this zero phase will be referenced to this common central point. 9.4. OBJECT MOVEMENT

The inverse Fourier transform method of obtaining the projection of the scattering amplitude on a particular axis employs phase as well as amplitude information. A phase change of π/4 radians across the frequency band in the course of the collection of the data will be sufficient to invalidate the inversion process. This corresponds to a movement of 6 µm at the highest frequencies. The maximum velocity of travel for the object is therefore determined by the time taken to collect the data comprising a single projection. In the most favourable case, in which a single illuminating pulse gives a sufficient signal to noise ratio for a satisfactory projection to be recovered, the measurement time is 3 ns and a velocity as high as 1,000 ms−1 could be accepted. In the more restrictive case of a frequency sweep, for which the minimum measurement time is 300 ns, a velocity of 10 ms−1 would be acceptable. In a real system some degree of averaging will be required to achieve adequate signal to noise ratios, and this will impose more stringent limits on the maximum velocity that can be allowed while still obtaining a good set of projections. The actual limit depends on the method of averaging adopted; higher velocities can be accepted if the datasets are inverse transformed before averaging than would be the case if the averaging occurs while the absolute phase of the signal is still of significance. In any initial trial of the system concept the simplest approach

SYSTEM REQUIREMENTS FOR A SCANNER

199

to signal averaging is likely to be adopted which will restrict such systems to investigating stationary objects. 10. Weighting of Selected Bands The most challenging aspect of the proposed system is to establish the method used to generate the indicative and background images. This is a problem that has similarities with Principal Component Analysis and other clustering techniques. An essential feature of solutions to these problems is that the optimal choice of cluster parameters is a function of the scattering amplitudes in the different frequency bands of the material of interest and also of all the materials that are likely to be encountered in that application. Any place within the indicative image is, in essence, a low-dimensional projection of the overall spectral response at that place. The aim of the algorithm that generates the images should be to minimise the probability of confusion between the suspect material and all others. The frequency bands in which there is the greatest difference between the materials will have the greatest difference in weight between the indicative and background images, whereas frequency bands which show little distinction will have little weight. If a new innocuous material is now encountered fairly often in the application and this material is distinguished from the one of interest in a frequency band other than those used principally to distinguish the current range of materials, then that frequency band will have its weight in the indicative and background images changed. The outcome of scattering or reflection measurements made on a material in N different frequency bands can be plotted as a point in an Ndimensional coordinate space. Any proposal for a spectroscopic material identification system will be predicated on there being a significant distance between the point characteristic of the material to be identified and those corresponding to other materials that are likely to be encountered in that application. There are two confounding features in the problem of identifying materials over which the observer has no control: variable packing density and the intermixing of materials. In developing any algorithm to generate the images it is reasonable to make certain assumptions: first, that any material has scattering amplitudes in each of the bands at which observations are made that are, for a fixed density, reproducible within an error that can be determined experimentally. This set of scattering amplitudes and their probable variation can be represented as a small volume within the N-dimensional coordinate space. The second assumption is that the effect of reducing the density of the material by packing it less

W.S. TRUSCOTT

200

densely is to cause the location that characterises the probable scattering amplitude to follow a smooth trajectory within the coordinate space which, in the limit of low densities, will arrive at the origin. The third assumption is that the effect of mixing two materials in different ratios is to cause the location that represents the scattering amplitudes to make a smooth trajectory from the location characteristic of the first material to the location characteristic of the second as higher and higher proportions of the second are added to the first. These assumptions will not be valid in circumstances where the loose packing or intermixing is achieved by subdividing the bulk material into layers or lumps whose spacing approaches the wavelength at one or more the bands at which image data is collected. These subdivided materials will appear, in this context, as new materials, since they will have a very different variation of scattering amplitude with frequency to that of the undivided material.

ABC

Z V

Y

0.8 0.6 0.4 0.2 Σmat X

W Figure 1. The proposed method of selection by spectral characteristics for the indicative image illustrated for the case of measurements in five spectral bands. The hypervolume Σmat includes the range of spectral coordinates observed for commonly encountered materials at various densities. The region ABC corresponds to the spectral characteristics of undiluted suspect material. The line labelled with the scale 0.2–0.8 is drawn from ABC to Σmat passing through the measured data for a certain location in the object. The intensity of that location in the indicative image is given by the scale.

SYSTEM REQUIREMENTS FOR A SCANNER

201

If the scattering amplitudes of the different innocuous materials likely to be encountered in the application are measured in the various frequency bands, then a volume of the N-dimensional coordinate space can be constructed that surrounds the origin and is bounded by the regions characteristic of different materials in their unmixed and undiluted form. All points within this volume represent potential spectral responses for various densities and degrees of intermixing that may reasonably be encountered. Any observed spectral response that lies outside this hypervolume will indicate the presence of either an unexpected material or the suspect material. Figure 1 shows this concept for a projection of a 5D (V, W, X, Y, Z) space. The envelope of responses for likely materials and their mixtures at various densities is indicated by the shape labelled Σmat while the region of possible responses for the undiluted suspect material is labelled ABC. Clearly a measurement that shows that the scattering amplitude at some point within the object lies within Σmat is indicative of the presence of one of the previously identified probable materials whereas one that lies within the region ABC is likely to indicate the presence of the undiluted suspect material. The measurements in the different bands may indicate that the scattering at some place within the object is represented by a point lying between the region ABC and the hypervolume Σmat, for example, along the vector labelled with the scale 0.2–0.8. A plausible explanation for this spectral response is that the suspect material has been mixed with a common-or-garden material whose response lies within the hypervolume Σmat in a ratio that depends on the distance along the vector from ABC as indicated by the scale. This distance may be used to determine the intensity of the indicative image for that point within the object. This algorithm which gives the greatest intensity for regions close to the volume ABC and zero intensity for the regions within the hypervolume Σmat is able to deal with complex forms for the hypervolume enclosing all probable materials. The discussion in this paper has hitherto been based on screening for a single suspect material whose THz properties are described by the region ABC in Figure 1. It is clear that, if it was desired to screen for further suspect materials that were characterised by other regions in the hyperspace shown in Figure 1, then the same algorithm could be used to provide appropriate amplitudes for the contributions of a particular location within an object to further indicative images. The sole requirement for this extension is that the spectral features of each suspect material should map to regions outside the hypervolume Σmat. It is possible to combine a number of these indicative images using colour-mapping techniques to allow this screening system to search for several suspect materials at once with a

202

W.S. TRUSCOTT

facility for switching the image to more specific identification modes by changing the colour mapping. This ability to screen for multiple substances does not require any additions to the THz and electronic signal processing systems. The additional complexity for the data analysis is small and the techniques are well established. The proposed system could therefore be used flexibly within an international airport to screen both outgoing baggage for explosives and incoming baggage for drugs of abuse which could greatly increase its benefit to the community as a whole. 11. Conclusion This paper has presented a solution to the problem of the identification of a suspect material that is concealed among other materials which mask its particular THz spectral features. The paper has demonstrated that the combination of tomographic object reconstruction techniques with the measurement of the amplitude of the THz radiation from a small number of sources scattered towards a number of appropriately placed receivers will allow the spectral features of particular regions to be examined. By collecting data across a set of 30-GHz-wide bands these regions may be as small as 1 cm³. Appropriate frequencies for these bands have been identified through an examination of the water vapour absorption spectrum in the 500-GHz to 3-THz range. An algorithm for combining the object reconstructions obtained in different bands is proposed which is based on the knowledge of the spectral characteristics of the suspect material and other materials likely to be encountered in a particular screening situation. An important feature of this algorithm is that it is able to handle both mixtures of materials and also different packing densities for the materials. While the concept has the potential to be used in screening people in, for example, an airport portal, it is likely that the system will first be developed for screening packages and baggage that can be held stationary for the period of time required to obtain a good signal to noise ratio within the limitations of current source and receiver technology. The power budget has been examined leading to the conclusion that a form of THz camera technology will be required in order to give signal levels that are observable using heterodyne techniques. Four different techniques for generating the required source spectrum have been examined and their relative merits discussed. The choice of one of these methods will be largely dependent on the material characteristics of the source used. The proposed system has been demonstrated to be a solution to the screening problem that is feasible with current technology; furthermore, the

SYSTEM REQUIREMENTS FOR A SCANNER

203

current rate of advance in THz sources and detectors suggests that satisfactory images should soon be obtainable with reasonable throughputs for realistic high security screening applications. The principal requirement for this system to be implemented is the collection of a large amount of information on the THz-scattering properties of innocuous materials likely to be encountered in any proposed screening situation. Current technology is well adapted for the collection of this data, and there is a strong case that this rather mundane research should be initiated now so that the information is available as soon as the proposed screening systems become desirable in cost-benefit terms. Acknowledgements The author is grateful for the support for his work provided by the Research Councils UK Basic Technology Programme, “The Development of Terahertz Technology for Imaging and Spectroscopy in the Physical, Medical and Biological Sciences” and to NATO for the funding of this workshop.

References [1] V. Agrawal, T. Bork, S. Kee and D. W. Van Der Weide, Electronic THz reflection spectroscopy for detecting energetic materials, IEEE Sixth International Conference on Terahertz Electronics Proceedings, THZ 98, pp. 34–37 (1998). [2] S. H. Wang, B. Ferguson, C. Mannella, D. Abbott, and X-C Zhang, Powder detection using THz imaging technical digest, Summaries of Papers Presented at the Quantum Electronics and Laser Science Conference (IEEE Cat. No. 02CH37338) pt. 1, 44, vol. 1 (2002). [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, Proceedings of SPIE, vol. 5070, pp. 44–52 (2003). [4] M. Walther, B. M. Fischer, and P. U. Jepsen, Noncovalent intermolecular forces in polycrystalline and amorphous saccharides in the far infrared, Chem. Phys., 288, 2–3, 261–268 (March 2003). [5] P. C. Upadhya, K. L. Nguyen, Y. C. Shen, J. Obradovic, K. Fukushige, R. Griffiths, L. F. Gladden, A. G. Davies, and E. H. Linfield, Transformation-induced evolution of farinfrared vibrational modes in theophylline, Conference Digest of the 2004 Joint 29th International Conference on Infrared and Millimeter Waves and 12th International Conference on Terahertz Electronics (IEEE Cat. No. 04EX857), pp. 429–30 (2004). [6] M.van Exter, Ch. Fattinger, and D. Grischkowsky, Terahertz time-domain spectroscopy of water vapor, Opt. Lett., 14, 20, 1128–1130 (Oct. 1989).

204

W.S. TRUSCOTT

[7] K. A. McIntosh and S. Verghese, Demonstration of terahertz water-vapor spectroscopy with a photomixer transceiver, Proceedings of SPIE, vol. 3794, pp. 16–24 (1999). [8] H. M. Pickett, R. L. Poynter, E. A. Cohen, M. L. Delitsky, J. C. Pearson, H. S. P. Muller, and J. Quant, Submillimeter, millimeter, and microwave spectral line catalog, Spectrosc. & Rad. Transfer 60, 883–890 (1998). On-line access at http://spec.jpl.nasa.gov/

CHALLENGES TO TERAHERTZ COUNTER-TERRORISM AND SECURITY-RELATED APPLICATIONS HOWARD CUMMINS* HM Government Communications Centre Hanslope Park, Milton Keynes, Buckinghamshire, MK19 7BH,UK

Abstract. Since early 2002, HMGCC has been monitoring developments in Terahertz (THz) technology in government, academia, and industry, as applied to counterterrorism and wider security applications, both at home and abroad. In addition it has funded proof of concept studies in UK industry and research in UK universities (Leeds and Durham) to help form a view as to its likely performance. This paper summarizes the results of work that HMGCC and other UK Government Departments have undertaken to date and draws a positive conclusion as to the prospective operational performance of THz systems in the security domain. Suggestions as to the way the technology needs to be developed, to enable both cost and functionally effective systems to be produced that will satisfy operational needs, are made.

Keywords: counter-terrorism, security, imaging, technical security countermeasures, remote sensing

1. Introduction The ability to detect and identify objects hidden behind barriers is one of the core requirements for security inspection systems. To meet the threats found in today’s security environments (hidden weapons, body-worn explosives, etc.) a strong operational need has emerged from the security community

______ * To whom correspondence should be addressed. Howard Cummins, HM Government Communications Centre, Hanslope Park, Milton Keynes, Buckinghamshire, UK, MK19 7BH; e-mail: howardc@hmgcc. gsi.gov.uk

205 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 205–224. © 2007 Springer.

H. CUMMINS

206

worldwide for systems to detect these at longer ranges than achieved by traditional security portals. Terahertz (THz) technology offers a potential way forward however, in its current state of detailed scientific understanding and technological development, it is by no means certain that it will provide operationally effective solutions. 1.1.

DEFINING “THz”

There is some argument as to the strict definition of “THz.” It is accepted in academia that it covers the frequency range 300 GHz–10 THz. Industry tends to stretch this down in frequency to 100 GHz. Because the properties of materials change so rapidly over this frequency range it is in fact more useful, from a technological perspective, to subdivide this into high millimeter, submillimeter and above 1 THz (1–10 THz). 1.2.

WHY MIGHT THz BE USEFUL?

THz radiation has a number of properties that make it a good candidate technology for the following reasons: • • •

It is nonionizing and intrinsically safe. Many visually opaque materials, such as clothing and packaging materials, are transparent to THz radiation. It offers specificity in identifying materials as this is the lowest frequency at which inter and intra molecular resonances in solids occur at room temperature (gases start in the mm waveband).

THz sensing therefore has the potential to identify a threat material uniquely, a property which separates it from some of the lower frequency techniques, such as passive and active millimeter wave imaging. 1.3.

WHY IS THz STILL A PROBLEM?

From the technological perspective: •

Atmospheric attenuation due to molecular rotational and vibrational absorption is high, with only certain transparent atmospheric windows available across the millimeter and THz frequency range.

COUNTER-TERRORISM AND SECURITY

•

• •

207

Most importantly, the interaction of THz radiation with materials, knowledge of which is essential in predicting target phenomenology, is still very poorly understood. In some respects it is principally scattering. High-power frequency-agile sources of THz radiation have not yet been developed to technological maturity. The technology is expensive because it is so immature and specialized

2. Security Problems of Interest THz Spectrographic and Anomaly Detection Imaging offers solutions to counterterrorism and security-related problems. We have studied the following potential applications:

•

• • •

Threat Detection at Range – Suicide Bomber – Long Range – Weapons (Knives, Guns, etc. – may not be metallic) – Long & Short Range Threat Detection Close In – Mail Examination – Short Range Nonintrusive search to prevent removal of classified material – Short Range Technical Security Countermeasures (TSCM) – Detecting Tampering of Electronic Components – Short Range – Nondestructive Examination of Building Structures – Short Range

3. Short Range THz Pulse Imaging The Spatial Temporal parameters of THz Pulse Imaging (TPI) Scanners are typically • • •

A 3.5-THz half-cycle pulse is 43-µm long (in vacuo). At 3.5 THz the wavelength is 86 µm (in vacuo). A beam diameter > 100λ is necessary for geometrical optics approximations to apply which corresponds to 8.6 mm.

208

H. CUMMINS

These parameters result in an axial resolution of 20 µm and scanning resolution of 250 µm in a practical machine. 3.1.

IMAGING ELECTRONIC COMPONENTS

There is a need in the technical security arena to detect tampering in electronic components. THz imaging is attractive as it enables both singlesided and double-sided imaging of electronic component constructions. Also it is unlikely to damage sensitive electronic components as it is nonionising radiation. It has good penetration of the plastics/ceramics used in component/printed circuit board construction.

Figure 1. Top and bottom view of Intel 486 processor IC in ceramic package.

Figure 2. THz image of Intel 486 processor showing top surface, subsurface interconnects and silicon-die and bottom-level pins.

Time domain 2D imaging can be achieved by combining the normal b-scan representations obtained from a terahertz pulsed imager (TPI). We can plot the image that is formed by the reflected pulses at a particular

COUNTER-TERRORISM AND SECURITY

209

time/depth. In the following example, we use an Intel 486 chip as shown in Figure 1. The ceramic packaging is opaque to visible wavelengths but not to THz radiation – providing the possibility of hidden object detection. Figure 2 shows time domain TPI images of the 486 processor shown in Figure 1. These images are generated at three different depths, by selecting different time of flight points, and plotting the detected amplitude. 3.2. CHEMICAL-BIOLOGICAL MAIL THREAT

In response to many “white powder” incidents post 9/11, we have studied the potential application of THz imaging to the detection of threats arriving in the mail, in particular those arising from the inclusion of chemical and biological substances. As a case study we took the example of the Anthrax-contaminated envelopes received by a number of individuals in the USA post just after 9/11. The issues addressed were: • •

Discrimination Isolation (Machine/Personnel Contamination)

For the latter it is easy to exploit the low absorption of polythene across the THz band to isolate the suspected item from the environment, yet still allow effective probing by THz radiation. We took a simulated Anthrax envelope with identical text and folding as that sent to Tom Brakow at NBC. In this test the envelope was loaded with 1–2 g lycopodium powder (Club moss spores) as the Anthrax simulant. The results of the imaging experiments (conducted by Teraview) are shown in Figures 3, 4, and 5. It can be seen that the text in the threat letter can be read and potentially reconstructed without opening the envelope. The presence of powder can also be seen.

210

H. CUMMINS

Figure 3. Composite THz pulse image of a letter containing an anthrax powder simulant and a folded “Threat” letter.

Figure 4. Individual letters of words written in ball point pen in the “Threat” letter revealed through the envelope.

Figure 5. Contrast image of anthrax powder simulant in envelope.

Less encouraging has been the spectrographic detection of Anthrax simulant. For this we used bacillus globigii aka Bacillus subtilis var niger (Bg) 90% mixed with 10% AEROSIL R972 (dispersal agent). The THz absorption spectra of Anthrax stimulant, paper (envelope) and various “white powder” materials seen in hoax Anthrax envelopes are shown in Figures 6, 7, and 8.

COUNTER-TERRORISM AND SECURITY

211

Figure 6. THz absorption spectra of bacterial spores (bacillus subtilis var niger (Bg) 90% mixed with 10% AEROSIL R972), fuming silica (AEROSIL R972), and lycopodium powder on vacuum grease.

Unfortunately the very small peak in the “bacterial absorption” corresponds to the absorption peak in envelope paper, although this might just be for the particular envelope used.

Figure 7. THz absorption spectra of paper (envelope) and “white powder” materials seen in hoax anthrax envelopes.

The absorption spectra for icing sugar indicate that there is potentially good discrimination between organic chemical compounds. However, the polymerisation of these compounds that occurs when they form living structures wipes out the THz spectra, making discrimination impossible.

H. CUMMINS

212

Figure 8. Bulk THz absorption spectra of bacterial spores (bacillus subtilis var niger (Bg) 90% mixed with 10% AEROSIL R972), fuming silica (AEROSIL R972) and lycopodium powder.

3.3. IMAGING INTO BUILDING STRUCTURES

We have investigated the potential application of THz imaging as a Technical Security Countermeasures Search Tool for the detection of: • •

Buried Devices (Metallic and Nonmetallic) Improvised Explosive Devices

The conclusions is that the dynamic ranges available with the current state of source and detector technologies only really allow effective imaging to a depth of around a centimeter into common building materials. Table 1 shows the upper frequency limits for 1-cm penetration in common building materials. However, this may still be useful to detect disturbances to decorative finishes, although still not cost competitive with lower frequency radar, infrared and ultraviolet techniques. A dynamic range of 106 represents the current performance of commercial (Teraview) THz imagers

COUNTER-TERRORISM AND SECURITY

213

TABLE 1. Upper frequency limits for 1-cm penetration in common building materials Material Polystyrene UPVC Brick Semi-engineering brick Ceramic tile MDF Oak Pine Plasterboard Fiberboard Reconstituted stone Lightweight aerated block Leca block

3.4.

Frequency Limits (GHz) for 1-cm penetration Dynamic range = 106 Dynamic Range = 108 >2 THz >2THz 760 925 439 527 542 660 586 454 571 278 381 480 710 644 770 >556 205 292 130 180 <150

CONCLUSION ON SHORT-RANGE APPLICATIONS

Early work HMGCC funded with Teraview suggests a potential for shortrange applications. However the system cost is too high to gain acceptance by the security community (The current cost of £250K/system this is about 10 times higher than would be acceptable). Further work needs to be done to realise deployable systems for shortrange imaging applications. This includes as: • •

Signal-processing techniques to separate out the spectra of layered and more complex structures1 Development of source and detector technology to reduce system costs

4. Threat Detection at Range “THz” technology is seen as providing solutions to the detection of: (i) (ii)

Suicide Bombers Concealed Weapons

H. CUMMINS

214 4.1.

BACKGROUND

Electromagnetic remote sensing technology falls into two categories: (i) Passive (Radiometer) where the radiation measured is provided by the target itself or reflected/scattered toward the receiver from sources in the environment. (ii) Active (Radar) where the target is illuminated from an artificial source that may or not be collocated with the receiver. The technological readiness of THz imaging is best assessed by taking a systems approach to both of the above categories in a particular application. 4.2.

SYSTEMS

System calculations have been developed and refined for millimeter wave imaging systems over the last 40 years and these may be readily used to predict the performance of systems operating at THz frequencies. What is lacking is the basic measurement data describing the target phenomenology over the frequency ranges of interest. The factors affecting the performance of passive (Radiometer) and active (Radar) systems2 are shown in Figure 9.

Scene of Target Atmosphere • • •

Target Background & Surroundings Countermeasures

Interfering Sources • • •

Sensor • • •

Absorption Scattering Thermal Emission

• •

Radar Radiometer

Natural thermal sources Artificial Sources Jammers

Figure 9. Factors affecting system performance analysis.

There are common parameters affecting the performance of both passive Radiometer and active Radar systems and these are explored below.

COUNTER-TERRORISM AND SECURITY

215

Figure 10. Typical atmospheric absorption coefficient at room temperature.

4.3. ATMOSPHERIC ABSORPTION

The absorption by the atmosphere has been extensively measured and modeled across the millimeter and THz frequency ranges, principally in support of missile seekers and military communications in the millimeter bands and by astronomers in the THz bands. Figure 10 shows the atmospheric absorption under typical sea level atmospheric conditions. There are subbands of lower (than average) absorption throughout the THz frequency range that could potentially be exploited. 4.4. TARGET PHENOMENOLOGY

To predict the performance of both active and passive THz systems we need an accurate model of the target physics. 4.4.1. Absorption due to barrier materials The dominant interaction between THz radiation and target materials is scattering. Work by Chamberlain et al. at Durham who are extensively modeling and measuring clothing3 suggests that:

H. CUMMINS

216

• • •

Scattering from clothing is essentially isotropic. Losses are primarily due to scattering. Reflection is low, around 1–2% (which is good from the active clutter point of view).

Denim Ladies Fleece

100% cotton 100 % Polyester

Suit Breast Pocket pure new all wool Ladies Jumper 100 % acrylic Shirt Breast Pocket 65% polyester 35% cotton Shirt Breast Pocket terylene & cotton Shirt

67 % polyester 33% cotton

absorption (arbitrary units)

5

4

3

2

1

0 0.0

0.5

1.0

1.5

2.0

frequency / THz

Figure 11. Absorption of clothing samples in the THz region.

From their preliminary results (Figure 11), using either a focused or unfocused beam for a single transit through a polyester shirt, transmission is about 50% measured at 1.5 THz or as a ratio of peak heights in the time domain. When the signal is reflected back through the material from a mirror placed behind the shirt, transmission seems to be about 70% (comparing peak–peak amplitudes only). Note this is using a photoconductive receiver in an unfocused geometry only, whereas single transmission measurements are from both photo-conductive and electro optic focused/ unfocused beams. Reflection from shirt surfaces is very low indeed. The data, even with processing, shows no real reflection (in the unfocused beam.) Tweed which is a much thicker and more complex structured material, exhibits 25% direct transmission, and the double transmission, with mirror reflection, seems to be 60% whilst about 5% is reflected off the surface.

COUNTER-TERRORISM AND SECURITY

217

In both the above materials, there seems to be higher transmission for the unfocused beam undergoing a double transmission compared to the single transmission. Taken together the high absorption of the atmosphere and barrier material in the THz band causes a successive low pass filter effect so we lose high- frequency fidelity (Table 2). TABLE 2. Atmospheric and clothing (fleece and shirt) attenuations in the THz low absorption atmospheric windows THz Window center frequency (THz)

Atmospheric base e extinction (m−1) 25°C 30% RH

Atmospheric absorption (dB) 25°C 30% RH

Clothing attenuation (dB) – fleece

Clothing attenuation (dB) – polyester cotton shirt

0.5

10−2

0.4

5

1.5

0.65

10−2

0.4

8

2

0.4

12

3

0.87

10

−1

1.02

10

4

20

4

1.29

10−1

4

25

5

4

30

7

4

40

8

1.38 1.52

4.4.2

−2

−1

10

−1

10

Target materials

The claim that remote stand-off spectroscopy is capable of detecting the components of a body-worn Improvised Explosive Device (IED) is still open to challenge. This is principally because of the high attenuation of some types of clothing material over the spectral range where explosive spectral features appear (see Figure 12). Also the way in which these spectra might vary with environmental conditions (temperature, humidity, etc.) and interference effects from the target structure is not well understood. The probability of being able to detect explosives with lower frequency spectral features, such as RDX and PE4 is higher due to clothing and atmospheric attenuation increasing with frequency above 1 THz. The reflection coefficient of RDX at its spectral feature at 750 GHz is 6% but is only 3% over the rest of the “THz” frequency range. This would give little contrast against the human body, which has a reflection coefficient of around 10% over this spectral range.

218

H. CUMMINS

The probability of obtaining spectroscopic identification of HMX (Home Made explosives) is even less, as the crystalline nature of HMX leads to specular (as opposed to diffuse) reflection. However, in the target scene this might show up as a strong anomaly. Detecting the metallic content (ball bearings, nuts and bolts, etc.) of IEDs is easier due to their high reflectivity. The reason to go to higher frequency in this case is increased resolution.

Figure 12. Explosives absorption in THz spectral range (Teraview).

4.5. FUNDAMENTAL LIMITS TO NOISE FLOORS IN THz RECEIVER SYSTEMS

We can use thermal (Johnson-Nyquist) noise up to kT/h (6 THz at 300K). The noise power Pn = kTB where k = Boltzmann’s constant, T = absolute temperature, B = bandwidth, h = Plank’s constant. The exact equation for the noise power is: Pn = hfB/(ehf/KT−1)

COUNTER-TERRORISM AND SECURITY 4.6.

219

RADIOMETERS

Radiometers have been well developed by the astronomy community over a number of years through well-funded EPSRC programmes. The system elements comprising radiometers are detectors and optical components. At their present stage of technological development, passive radiometers are confined to the millimeter waves (below 300 GHz). As there is no spectroscopic response of solids and liquids in this spectral range it enables anomaly detection only, through absorption and reflection of objects against the background and the thermal radiation of the body. Figure 13 shows the reflection coefficient of human skin derived from in vivo complex refractive index measurements4 in the THz frequency range.

Figure 13. Reflection coefficient of human skin.

Typical emissivity figures for the human body in this spectral range are shown in Table 3. TABLE 3. Emissivity of the human body Frequency 100 GHz 250 GHz 3 THz

Emissivity 0.5 0.8 0.99

What is not so understood is the target phenomenology in the security arena. At the high and submillimeter bands, etalon effects dominate. In this

220

H. CUMMINS

situation, one of the major sources of false alarms in passive radiometer systems is reflection of background radiation by clothing, especially outdoors where the cold sky radiation dominates. At THz frequencies, radiation from the background, (including the sky) is at ambient temperature, due to high atmospheric absorption/emission. In this case background radiation effects are less of a problem.

Figure 14. Zenith brightness temperature over millimeter wavebands.

Figure 15. Zenith brightness temperature in submillimetre waveband.

COUNTER-TERRORISM AND SECURITY

221

At high-millimeter wave frequencies, the thickness of clothing is often in the region of a half or quarter wavelength, leading to Fabry-Perot interference effects (equivalent to the oil film effect at visible light frequencies). This can give rise to glint, which is a particular problem when imaging outdoors due to the cold sky being reflected by the outer layer of clothing. This gives the appearance of colder objects on the body, leading to possible false alarms. Figures 14 and 15 show the sky brightness temperature over the millimeter and submillimeter bands.5 4.7.

RADAR

As well as insufficient understanding of target phenomenology, the lack of high power, frequency agile sources has hampered the development of this technique at THz frequencies. 5. Performance Case Studies 5.1.

RADIOMETER

The temperature sensitivity of a full power radiometer Ts is given by:

Ts = (Tantenna + Treceiver)/(BIF τintegration)½ where, Tantenna = Antenna Temperature, Treceiver = Receiver Temperature, BIF = IF Bandwidth, τintegration = Integration Time For an 800-GHz cryogenically cooled imager with a receiver noise temperature of 300K, and a 4-GHz IF bandwidth imaging 100×30 pixels at 3 fps, the sensitivity = 0.9K The radiometric temperature Tradiometric of a target with emissivity ε and reflectivity ρ at a thermodynamic temperature Tthermodynamic is given by: Tradiometric = εTthermodynamic + ρTbackground If the total loss between the target and the antenna is L and the target fills the field of view of the radiometer, its apparent temperature Tapparent is given by: Tapparent = Tradiometric/L The radiometric contrast, ∆T, between two targets at different apparent temperatures ∆T = Tapparent1 – Tapparent2

H. CUMMINS

222

5.1.1. Skin vs. Explosive–Radiometric Contrast A target skin to explosive (PE4) radiometric contrast is 2K. In this case we can tolerate a loss of 2.5 which would correspond to transmission through 10 m of atmosphere (30% humidity and 20°C) and a polyester shirt. 5.2. RADAR

As a case study of an active imaging radar, we can estimate the transmitter power required to detect a target underneath a fleece garment at a range of 10-m range (Table 4). We assume that the system is operating in an atmospheric window (of lower than average attenuation) in the THz band with the following parameters: •

• • • • •

60-cm-diameter optics – 2D2/λ = 1,920 m – Hence in radiative near field so just atmospheric attenuation as focused beam 10-m range of 30% humidity atmosphere at 20°C gives one-waypath loss of 0.9 IF bandwidth = 4.5 kHz for imaging 100×30 pixels at 3 fps sensitivity (highly idealised coherent detection is assumed as the local oscillator can be derived from “racked out” transmitter signal). Room temperature mixer noise temperature 1,300K (7.4 dB NF) Heterodyne receiver sensitivity = −130 dBm Illumination area = 100×30 cm at 10 m

TABLE 4. Illumination power for 10-m range – target concealed behind a fleece jacket

THz Window center frequency (THz)

Transmitter power (dBm) Transmitter power (W)

0.5

−19

12.5µ

0.65

−13

50µ

0.87

−5

316µ

1.02

11

13 m

1.29

21

126 m

1.38 1.52

31 51

1.26 126

COUNTER-TERRORISM AND SECURITY

• •

223

Power required/pixel for S/N of 100 = −110 dBm Total power required to be transmitted at target assuming 6% reflectance into hemisphere = −29 dBm

6. Conclusions • • • •

We are achieving a good level of understanding of system phenomenology in the THz frequency regime. The required levels of performance for security applications are available with the current THz systems. THz technology is not a “Silver Bullet” but is an effective adjunct to other systems, such as thermal (mid infrared). There is still much work to be done to engineer systems to the cost levels acceptable to the security industry.

7. The Way Forward/Challenges in the Security Domain In the author’s view: •

• •

THz imaging is a developing field. It is worth comparing its development history with that of thermal imaging, but with the major difference that it is subject to a civil rather than a military funding model. The time is ripe to apply techniques from other areas of radio, optical, and imaging engineering, e.g. Ultra Wide Band (UWB) radar signal processing. The system costs needs to be lowered through technology development to introduce sources and detectors which are more like conventional radio electronics.

Acknowledgments The author would like to thank the following for much of the work and useful discussions leading to the material presented here: Teraview, Thruvision, Foreign and Commonwealth Office, HOSDB, Professor Martyn Chamberlain, Peter Swift and John Fletcher, of the Department of Physics, University of Durham, Professors Bob Miles, Edmund Linfield and Giles Davies of the Institute of Microwaves and Photonics, School of Electrical and Electronic Engineering, University of Leeds.

H. CUMMINS

224

References 1. 2. 3.

4.

5.

J. R. Fletcher et al., Detection of separation of laminated materials by reflected THz pulses, Berlin Non Destructive Testing Conference (Sept. 2006) (To be published). C. R. Seashore, Missile Guidance, in Infrared and Millimeter Waves Chapter 3, vol. 4, K. J. Button and J. C. Wiltse., Eds. (Academic Press, 2003). ISBN 0-12-147704-5. J. R. Fletcher, G. P. Swift, De Chang Dai, J. A. Levitt, and J. M. Chamberlain, Propagation of terahertz radiation through random structures: a novel theoretical approach and experimental validation, J. Appl. Phys., 101, 013102 (2007). E. Pickwell, B. E. Cole, A. J. Fitzgerald, M. Pepper, and V. P. Wallace, In vivo study of human skin using pulsed terahertz radiation, Phys. Med. Biol., 49, 1595–1607 (2004). R. E. Hills, A. S. Webster, D. A. Alston, P. L. R. Morse, C. C. Zammit, D. H. Martin, D. P. Rice, and E. I. Robson, Absolute measurements of atmospheric emission and absorption in the range 100–1,000 GHz, Infrared Phys., 18, 819–825 (1978).

TERAHERTZ DETECTION OF ILLEGAL OBJECTS ROGER APPLEBY*, PETER R. COWARD, AND GORDON N. SINCLAIR QinetiQ, St Andrews Road, Malvern, Worcs, WR14 3PS

Abstract. This paper describes the main parameters – contrast, spatial resolution, and thermal sensitivity – which define the performance of any stand-off imaging system. The origin of the signature for both metal and dielectric objects hidden under clothing in the frequency range from 100 GHz to 500 GHz is discussed. At 100 GHz the signature is dominated by reflection whilst at 500 GHz it is dominated by emission. A 94-GHzpassive millimetre-wave imaging system has been designed and fabricated to image objects under clothing. This imager is based on a Schmidt camera folded using polarisation techniques.

Keywords: passive millimetre wave imaging, 100 GHz, 500 GHz

1. Introduction The benefit of imaging in the frequency range from 100 to 500 GHz is that materials such as clothing are transparent. This ability to see through clothing and detect hidden objects such as weapons and contraband means that this technology has the potential to be a useful tool in the fight against terrorism. There are other applications, such as the detection of illegal immigrants in soft-sided vehicles and the detection of biological and chemical hazards, but in this paper we will only consider detection of objects carried under people’s clothing at stand-off ranges in excess of 1 m. This is not a new application and was first reported1 in 1979 when the need for new personnel screening methods was recognised. This early work explored the possibility of using heterodyne receivers based on Schottky

______ *

To whom correspondence should be addressed: Roger Appleby, Qinetic, St. Andrews Road, Malvern, WorcsWR14 3PS, UK; e-mail:[email protected]

225 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 225–240. © 2007 Springer.

R. APPLEBY ET AL.

226

diode mixers at 100 and 300 GHz. In the years since 1979 there have been many breakthroughs in technology such as Monolithic Microwave Integrated Circuits (MMIC) at frequencies up to 220 GHz,2 the first real time imager3 and high- speed computers making possible sophisticated image processing. These developments have enabled the design and fabrication of imaging systems suitable for stand-off detection of hidden objects in real time. In Section 2, a review of the important parameters which define standoff system performance is given, and in Sections 3 and 4 system designs are discussed for people screening at 100 and 500 GHz. 2. Stand-off Imaging Systems The ability to detect an object using an imaging system is dependent on many parameters, and the imaging process is shown pictorially in Figure 1. Typically the transfer of information is a function of the parameters describing each of these elements: • • • • •

The radiometric properties of the scene and the environment Atmospheric attenuation between the scene and the optics Transmission coefficient and point spread function of the optics The sensitivity of the detector Processing electronics, display, and the interaction with the human visual system

Sophisticated computer models have been used to simulate the performance of imaging systems.4,5

Figure 1. Simulation of imaging Performance.

It is beyond the scope of this work to fully describe all the steps in the imaging process. Only the target signature, spatial resolution of the optics, and thermal sensitivity of the detector will be discussed here. This allows a

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

227

qualitative insight into the performance of imaging systems at 100 and 500 GHz. 2.1. TARGET SIGNATURE

The optical behaviour of materials can be described in terms of their transmission t, emissivity ε, and reflection r, which sum to 1 as shown in Eq. (1).

ε + r +t =1

(1)

A Far Infrared Polarising Michelson Interferometer has been used to determine the optical properties of clothing, skin, and explosives in the frequency range from 60 GHz to 1.2 THz. This instrument can be used to measure transmission, reflection, and refractive index. The clothing and explosive samples were 50 mm in diameter and were either positioned close to the detector for measurement of transmission or within the fixed arm of the interferometer to measure the refractive index. For reflection measurements of the skin, a mirror in the spectrometer was replaced by a sheet of expanded polystyrene against which an area of the body was held for the duration of the measurement. The angle of incidence was 30° and the polarisation was vertical. Some of the results are shown in Figure 2. In these plots the data below 60 GHz is unreliable due to a lack of energy from the light source in this region. 2.1.1. Clothing Samples of a cotton T-shirt, sweat shirt, denim, and a woollen jumper were measured. Figure 2 shows that the transmission of the T-shirt was 90% at 100 GHz falling to 80% at 500 GHz. The denim sample had a similar transmission at 100 GHz but fell off more quickly to 60% at 500 GHz. 2.1.2. Guns and knives Metallic objects are highly reflecting both at 100 and 500 GHz and have a reflectivity of 1. 2.1.3. Explosives The measured transmission spectra of explosives show a reduction in transmission from ~70% at 100 GHz to ~5% at 500 GHz. This measured loss in transmission could be due either to an increase in absorption or scattering. Whatever the mechanism, samples with a thickness above a few

R. APPLEBY ET AL.

228

mm were measured as virtually opaque above 500 GHz. These results are typical of measurements made on many different types of explosives. At frequencies below 200 GHz, where the material is essentially transparent and the surfaces are relatively flat, interference effects are present. Explosives can act like an anti-reflection coating to the skin and depending on the particular situation, can substantially reduce the reflection of millimetre-wave radiation from the body. There will also be interference effects in any air gap between the skin and the sample, which can increase the apparent reflectivity of the skin. This is a complex situation and the quality of the surfaces and the spacing between the different surfaces of skin and explosives are important.

(a)

(b)

(c) Figure 2. Transmission measurements of (a) T-shirt, (b) Explosives. Graph (c) contains reflectivity data of skin and water.

To illustrate this effect a simple model is used where the explosives are considered to be flat and in contact with skin. The properties of skin are

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

229

similar to water; see Section 2.1.4 and water is used here to model the skin. The dielectric properties of water are taken from Ulaby.6 The refractive index of a generic explosive is taken as n = 1.6 at 100 GHz and 1.6–0.02i at 500 GHz. The methodology for optical coatings is used for an explosives sample of thickness 5 mm and the reflectivity on water has been predicted and is shown in Figure 3.

Figure 3. Explosives on water.

At 100 GHz the signature of explosives on water is dominated by reflection having a mean of 0.3 with fringes of amplitude 0.3. If the bandwidth of the observing instrument is larger than the fringe separation, the observed contrast of explosives on water will tend towards the mean reflectivity of the fringe pattern. The reflectivity of explosives on skin at 100 GHz can then be approximated by taking the mean reflectivity of the explosives on water and multiplying by the ratio of the reflectivity of skin to water (0.35/0.45) as shown in Figure 2. At 100 GHz this will be 0.3 × 0.35/0.45 = 0.23. At 500 GHz the imaginary part of the refractive index is increased to account for the decrease in transmission shown in the figure. Whilst this might not be strictly correct if the loss in transmission is due to scattering, it will provide an estimate of the signature. The result is that at 500 GHz, the mean reflectivity is reduced to 0.055 with fringes of amplitude 0.03. The optical properties of the sample are now dominated by its emissivity or

R. APPLEBY ET AL.

230

scattering, and the reflectivity will be dominated by front surface reflection only. 2.1.4. Skin The reflectivity of skin can be shown to closely follow the reflectivity of water which is also plotted in Figure 2c. Furthermore it can be anticipated that the level of this reflectivity will vary as a function of blood flow in the skin. At 100 GHz, skin typically has a reflectivity of ~0.35 which falls to ~0.10 at 500 GHz. 2.1.5. Contrast model Having determined the optical properties of skin, explosives on skin and clothing a model can be used, to predict what the contrast of the target against the skin.

Figure 4. Contrast model for object on skin.

In Figure 4, the temperatures T adjacent to arrows are radiometric temperatures, as they would appear to a passive imaging system looking along the arrow. Temperatures To, Tp, TB are actual physical temperatures of skin, explosive material, and clothing respectively. The emissivity ε, reflectivity r, and transmission t are also subscripted to indicate the material. As radiation passes through the various layers, it is modified by their optical properties. The radiation from the environment TH passes through the clothing becoming TI. TI is related to TH by the optical properties of the clothing, emissivity εc, reflectivity rc, and transmission tc. The clothing has a physical temperature TB and will emit εcTB and will reflect radiation from the object rcTT, which will also contribute to TI. This process is described in

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

231

Eq. (2). The radiation then passes through the object and is reflected by the skin, back through the object and clothing and is determined by TN and can be observed by an imager. Contrast is then the difference between the radiation reflected from the skin when the object is present and that when the object is absent. Thus the parameters are related as described in the following model:

TI =εC TB + rC TT + tC TH

(2)

TM = ε M TP + rm TS + t m TI

(3)

TS = ε S TO + rS TM

(4)

TT = ε M TP + rM TI + t M TS

(5)

T N = ε C TB + rC TH + t C TT

(6)

The optical properties of the materials to be modelled are taken from measured data and are given in. Interference effects are accounted for using the fringe magnitude calculation techniques described in Section 2.1.3. A summary of the inputs to the model are given in Table 1. TABLE 1. Optical properties at 100 and 500 GHz

No object Explosive on skin Metal Skin Denim T-Shirt

100 GHz 0 0.74 0 0.65 0.09 0.04

Emissivity 500 GHz 0 0.95 0 0.91 0.49 0.2

Reflectivity 100 500 GHz GHz 0 0 0.26 0.05 1 1 0.35 0.09 0.01 0.01 0 0

Transmission 100 500 GHz GHz 1 1 0 0 0 0 0 0 0.9 0.5 0.96 0.8

2.1.5 (a) Environmental radiometric temperature The radiometric temperature of the surroundings (called the environmental radiometric temperature here) is dependent on the situation to be modelled.

R. APPLEBY ET AL.

232

1. Outdoor imaging at 100 GHz: The environmental radiometric temperature is modelled as 100K, i.e. approximately 200K colder than ambient temperatures. This represents the cold sky illumination that can be present in outdoor scenes in this waveband. 2. Indoor imaging at 100 GHz and 500 GHz: The cold sky illumination is not present, and the scene is illuminated by radiation from buildings and other items at near-ambient temperatures. Thus the environmental radiometric temperature is modelled as 300K for this scenario. 3. Outdoor imaging at 500 GHz: The sky radiometric temperature is warm due to high atmospheric attenuation. The environmental radiometric temperature is modelled as 300K for this scenario. 4. Artificial incoherent illumination (such as by hot emissive panels in a “chamber” arrangement) at 100 GHz or 500 GHz: These situations are modelled by an environmental radiometric temperature of 500K. 2.1.5 (b) Output of model The modelled apparent temperature of a gun, generic explosives, and skin are plotted against environmental radiometric temperature in Figure 5, for a situation where a T-shirt is present. The results are also tabulated at 100K, 300K, and 500K in Table 2. T shirt 500GHz 500

500 450 400 350 300 250 200 150 100

Gun Generic explosive No object

Apparent temperature

Apparent temperature

100GHz T Shirt

400

Gun

300

Generic explosive No object

200 100

0

100

200

300

400

Environment temperature

500

0

200

400

Environment temperature

Figure 5. Contrast predictions for generic explosives and a gun at 100 and 500 GHz.

2.1.5 (c) Results for 100 GHz At 100 GHz, the contrast of explosives to the skin is dominated by the differences in reflectivity. A 100-GHz system outdoors, where the environmental radiometric temperature is modelled as 100K, will observe a contrast of 18K between skin and explosives, and a contrast of 125K between skin and a metal gun. For a 100-GHz system indoors, the environmental radiometric temperature is 300K, and the modelled contrast for explosives is 2K and the

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

233

metal gun 4K. It should be noted that the contrast for explosives passes through a null at an environmental radiometric temperature of 280K and similarly for the metal gun at 310K. If an illuminating chamber (for example) was used to provide an environmental radiometric temperature of 500K, the contrast will be inverted and similar in magnitude to that seen for a 100-GHz system outdoors. 2.1.5 (d) Results for 500 GHz At 500 GHz the contrast of explosives is dominated by emissivity differences rather than by reflectivity differences as it was at 100 GHz. For indoor and outdoor situations, where the environmental radiometric temperature is modelled at 300K, the contrast for explosives is 3K and a metal gun is 6K. Environmental radiometric temperatures where contrast nulls occur in this case are 200K for explosives and 310K for the metal gun. If an illuminating chamber (for example) was used to provide an environmental radiometric temperature of 500K, the contrast between the gun and skin will be 111K but the contrast between explosives and skin 8K. In this scenario where the signature is dominated by differences in emissivity, small changes in physical temperature and temperature gradients can result in large changes in contrast. This simple model does not account for these temperature differences or gradients and in practice the contrast could be much greater. It is worth noting that active systems which use coherent sources can have equivalent temperatures in excess of 10,000K and can in principle exploit very small differences in reflectivity. TABLE 2. Contrast for metal gun and explosives at 100 and 500 GHz 100K

300K

500K

−125 –

−4 −6

115 111

18 –

−2 −3

−22 −8

Metal gun 100 GHz 500 GHz Generic explosives 100 GHz 500 GHz

234

R. APPLEBY ET AL.

2.2. SPATIAL RESOLUTION

The spatial resolution θ of an imaging system is determined by Rayleigh’s equation:

θ=

1.22λ D

(7)

where λ is the wavelength and D, the diameter of the aperture of the imaging system. From this equation we can see that the larger the aperture or the smaller the wavelength, the better the spatial resolution. For example, if we consider an imaging system with a 50-cm-diameter aperture operating at 100 GHz the spatial resolution will be 7 mrad and with the same aperture at 500 GHz this will be 1.5 mrad. This will equate to a spot sizes of 70 and 15 mm respectively at a range of 10 m. It follows that a system with a 50-cm-diameter aperture operating at 500 GHz will image a target with 15-mm features at a 10-m range. At 100 m this same system will only be able to image features which are 150 mm. 2.3. RECEIVER THERMAL SENSITIVITY

The receiver thermal sensitivity ∆T of a radiometric imaging system is given by the radiometer equation:

∆T =

T A + TS

βτ

(8)

where TA is the noise temperature of the antenna and Ts is the system noise temperature, β is the RF bandwidth and τ the integration time. To maximise the thermal sensitivity we need to minimise Ts the system noise and maximise the bandwidth and integration time. The development of low noise radiometers based either on heterodyne or direct detection architectures has led to system noise temperatures Ts of <1,000K. It is also not unusual to achieve an RF bandwidth of 10 GHz or more at 100 GHz. If we consider that a thermal sensitivity of 1K is adequate for many security applications we can simply calculate when viewing an object at room temperature with a lossless system that this can be achieved with an integration time of 0.17 ms. If we consider a frame rate of 15 Hz which is sufficient for most security applications, a frame will be acquired every 66 ms. We can therefore use one receiver to measure 66/0.17 = 388 pixels. Electro-optic scanning systems can be used to scan a single receiver over many pixels and this is illustrated by the imager

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

235

described in the Section 3 which uses a conical scan and a linear array of receivers. 2.4. SYSTEM THERMAL SENSITIVITY

A receiver thermal sensitivity ∆TR of 1K can be related to the system thermal sensitivity ∆Ts by accounting for the loss in the optics Lo and the atmospheric attenuation LA as shown in Eq. (8) assuming a room temperature operation. The system thermal sensitivity when equal to the contrast of the target as given in Table 2 will give a 50% probability of detection, provided the object has a size which is greater than the antenna beamwidth.

∆Ts = ∆TR Lo L A

(8)

The atmospheric loss in clear air over a 20-m range is ~0.01 dB at 100 GHz and 1 dB at 5,000 GHz. We can assume an optical loss of 3 dB which gives a total loss of 3.01 and 4 dB respectively. Therefore the required thermal sensitivity will be increased by a factor of 2.0 at 100 GHz and 2.5 at 500 GHz. A ∆Ts of 2.5K would be adequate to detect all the threats shown in Table 2. with a 50% probability. If however a higher probability of detection is required, e.g. 90%, it will be necessary to improve ∆TS and thus ∆TR by a factor of 3 from 1K to 0.3K to ensure detection of all objects at 300K. This can be achieved by increasing integration time or reducing the system noise. 3. Imaging at 100 GHz In security scanning outdoors at 100 GHz the contrast is large due to cold sky reflections and can be as high as 200K between skin and metal as shown in Figure 5. A design which maximises the spatial resolution and minimises the number of receivers can be advantageous with such a large contrast. Such an imager has been reported7 and it is shown in Figure 6.

236

R. APPLEBY ET AL.

Figure 6. 94-GHz Security Imager.

The imager is mounted on a pan and tilt head with a bore sighted TV camera. The TV camera is used to acquire the target and the passive millimetre wave imager provides a real time image of the person. A top level specification for the imager is given in Table 3. TABLE 3. Imager specification Parameter Entrance pupil (mm) Operating frequency (GHz) Field of view (V × H deg.) Frame frequency (Hz) Thermal sensitivity (K) Beamwidth (deg.)

Value 800 94 20 × 10 15 5 0.29

As shown in Figure 6, the radiation enters the system from the left passing through the curved front mirror which is covered by a polarising grid and is then linearly polarised. It then passes through a quarter wave plate where it is converted to circular polarisation and falls onto the scanning mirror. This offset mirror rotates at 4 rpm, scans the beam in object space and is responsible for producing the conical scan pattern. On reflection from the scanner the beam re-enters the quarter wave plate where it is converted from circular polarisation back to linear polarisation. The electric vector of the radiation is now parallel to the grid direction of the front mirror and is reflected and focussed onto the receivers via the secondary mirror.

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

237

The imager is shown looking from the front quarter in Figure 6b with its covers removed. The receiver block has also been removed and normally sits in the slot in the centre of the front grid. In this picture the front mirror and quarter wave plate can clearly be seen. The imager is diffraction limited and has a beamwidth of 0.29° (4.9 mrad) which will equate to a spot size of 49 mm at 10 m which is adequate for most security-related tasks. A linear array of 64 direct detection receivers based on indium phosphide MMICs8 is used to detect the millimetre wave image produced by the optics. The use of direct detection removes the need for a local oscillator and maximises thermal sensitivity. A single line of 94 GHz direct detection receivers spaced at 5 mm is accommodated by using a subreflector. It would be difficult to achieve this spacing with heterodyne receivers. This design is very economical in receivers when compared to systems which use multiple rows to overcome sampling problems.9 The noise temperature of a typical receiver is 800K, and a typical bandwidth is 20 GHz. If the receivers were to provide a thermal sensitivity of 1K, then (using Eq. (2)) the required integration time is 60 µs for an object at room temperature. As the imager has a frame rate of 15 Hz, a total integration time of 66 ms is available. This illustrates that a scanned imaging system is practical using such receivers, and in the discussed imager, each receiver is scanned in a circle in one frame time by the scanning mirror. If the optics were lossless, one receiver could scan of the order of 1,000 picture points and still retain the 1K sensitivity. However, each circle consists of approximately 200 picture points and the additional integration time is used to compensate somewhat for loss in the optics. This design trades thermal sensitivity for spatial resolution. At 100 GHz it is essential to maximise spatial resolution and good sampling of the imagery.7 This was achieved by decreasing edge taper and consequently sacrificing system transmission, which was reduced to ~20%. This impacts directly on the thermal sensitivity which is reduced to 5K. This is adequate for outdoor imaging as shown in Figure 5. Note that at 100 GHz atmospheric loss is negligible at ranges less than 100 m and can be ignored. 4. Imaging at 500 GHz No imaging systems at 500 GHz have been reported; however images from laboratory demonstrators using a coherent illumination source have been reported at 600 GHz.10 At 500 GHz the atmospheric attenuation is 50 dBkm−1 and will have a significant impact on the design of systems at this

238

R. APPLEBY ET AL.

frequency. Imaging using a passive or active system at 500 GHz is possible and is discussed briefly below. 4.1. PASSIVE IMAGING

It will not be feasible to manufacture an imager based on the optical design described in Section 3 for 94 GHz. This design makes use of expanded polystyrene which needs to be machined to an accuracy11 of 0.1 mm which is just within the tolerance of conventional milling machines. When scaled to 500 GHz an accuracy of 0.02 mm would be required which is beyond conventional machining. The front grid and quarter wave plate would also be much more difficult to realise at 500 GHz. An imager could instead be constructed using reflecting optics, more typically found in the infrared, which would easily achieve the necessary tolerances. It will also be possible to generate scanning systems based on rotating mirrors and prisms. Detector technology will be constrained to heterodyne receivers and bolometers as there are currently no MMIC LNAs available at 500 GHz and therefore direct detection using amplifiers is not possible. Heterodyne receivers can achieve noise temperatures of 1,000K–1,500K and to a first approximation are comparable to receivers at 100 GHz. Bolometers have been reported12 which can detect 500-GHz radiation but their sensitivity remains orders of magnitude less than heterodyne systems. In Section 2.1.3 the signature of a metal object and a generic explosive were derived. Due to the strong atmospheric absorption the sky temperature will remain close to ambient and will not be a source of contrast. The model data used to produce Figure 3 shows that at an ambient temperature of 300K there is a temperature difference of 6K between skin and metal and 3K between explosives and metal. As previously stated these signatures will be a function of emissivity rather than reflectivity and as such will be strongly influenced by the physical temperature of the objects and any temperature gradients within them. A well-designed imager will have sufficient thermal sensitivity to detect such contrasts. Receivers with a thermal sensitivity of better than 1K are achievable at 500 GHz with a frame time of 15 Hz but the ability to scan 200 pixels in one frame time, as was done at 94 GHz, will not be possible. This will give rise to systems with reduced fields of view, or with a large number of detectors and a tendency towards staring arrays.

TERAHERTZ DETECTION OF ILLEGAL OBJECTS

239

4.2. ACTIVE IMAGING

An active system can utilise reflecting components and heterodyne receivers in the same way that a passive system can. Here we will consider a narrow band illumination system without a coherent detection process. This makes the architecture more akin to a flashlight illuminator and not a RADAR. If we illuminate the target with a coherent source, the radiation temperature can be in excess of 10,000K. It is possible to produce such a source by multiplying up a Gunn Oscillator. Indium phosphide Gunn devices have been reported13 giving ~20 mW at 200 GHz and, if tripled, would produce a viable illuminator with a power of a 1 mW assuming a 5% efficiency in the multiplier. Vacuum tubes also offer an alternative source technology but tend to have bulky power supplies. Using the models described in Section 2.1.3 if such illumination techniques are employed, the gun will have an apparent temperature of ~5,000K and the explosives ~50K. However with such a bright coherent source we must also consider multipath, speckle, and other coherent artefacts. The ability to detect targets will then be determined by signal to clutter rather than signal to noise. The usefulness of such systems will only be determined through experimental investigation due to the complex nature of the signature formation. 5. Conclusions The transmission, reflectivity, and refractive index of clothing and explosives were measured using a Michelson interferometer. A simple model has been used to predict target signatures for explosives and metallic objects. This model showed that at 100 GHz the signature is dominated by differences in reflection whilst at 500 GHz it is dominated by differences in emissivity. At 100 GHz it is necessary to account for interference effects between the body and the explosives sample. A 94-GHz imaging system, capable of detecting objects hidden under clothing when the subject is outdoors, was described. It has a field of view of 20°×10° and produces imagery at 15-Hz frame rate. This system was designed to maximise spatial resolution. At 500 GHz both active and passive systems are feasible although the passive signature is greatly reduced compared to 100 GHz. Illumination technologies may be deployed to enhance the signature.

240

R. APPLEBY ET AL.

References 1. D. T. Hodges, E. E. Reber, F. B. Foote, and R. L. Schellenbaum, Safeguards applications of far infrared radiometric techniques for the detection of contraband, Nuclear Management, Winter, pp. 75–78 (1979). 2. A. Tessman, 220-GHz metamorphic HEMT amplifier MMICs for high-resolution imaging applications, IEEE J. Solid State Circuits, 40, 10, 2070–2075 (2005). 3. M. Shoucri, R. Davidheiser, B. Hauss, P. Lee, M. Mussetto, S. Young, and L. Yujiri, A passive millimeter wave camera for landing in low visibility conditions, AIAA/IEEE Digital Avionics Systems Conference, pp. 93–98 (1994). 4. E. Jacobs, R. G. Driggers, K. Krapels, F. C. De Lucia, and D. Petkie, Terahertz imaging performance model for concealed weapon identification, Proceedings of SPIE – The International Society for Optical Engineering, vol. 5619, pp. 98–107 (2004). 5. N. A. Salmon, Scene simulation for passive and active millimeter and sub-millimeter wave imaging for security scanning and medical applications, Proceedings of SPIE – The International Society for Optical Engineering, vol. 5619, pp. 129–135 (2004). 6. F. T. Ulaby, R. K. Moore, and A. K. Fung, Microwave Remote Sensing: Active and Passive – From Theory to Applications, vol. 3, ISBN: 0890061920 (Artech House, 1986). 7. R. N. Anderton, R.1 Appleby, J. E. Beale, P. R. Coward, and S. Price, Security scanning at 94GHz, Proceedings of SPIE – The International Society for Optical Engineering, vol. 6211, pp. 62110C.1–62110C.7 (Apr. 2006). 8. A. R. Barnes, P. D. Munday, R. Jennings, M. Black, R. Appleby, R. N. Anderton, G. N. Sinclair, and P.R. Coward, MMIC technology and its applications in mm-wave imaging systems, in 3rd ESA Workshop on mm-Wave Technology and Applications, pp. 543–547 (2003). 9. R. Appleby, R. N. Anderton, S. Price, N. A. Salmon, G. N. Sinclair, P. R. Coward, A. R. Barnes, P. D. Munday, M. Moore, A. H. Lettington, and D. A. Robertson, Mechanically scanned real time passive millimetre wave imaging at 94GHz, Proceedings of SPIE – The International Society for Optical Engineering, vol. 5077, pp. 1–6 (Apr. 2003). 10. E. L. Jacobs, S. Moyer, C. C. Franck, F. C. DeLucia, C. Casto, D. T. Peckie, S. R. Murrill, and C. E. Halford, Concealed weapon identification using terahertz sensors, Proceedings of SPIE – The International Society for Optical Engineering, vol. 6212, pp. 62120J.1–62120J.10 (Apr. 2006). 11. R. Appleby, R. N. Anderton, N. H. Thomson, and J. W. Jack, The design of a real time 94GHz passive millimetre wave imager for helicopter operations, Proceedings of SPIE – The International Society for Optical Engineering, vol. 5619, pp. 38–46 (Oct. 2004). 12. M. Jack, E. Gordon, G. Graham, H. Fetterman, I. Dunayevskiy, and R. Lombardo, Advances in bolometer-based passive imagers for homeland security and law enforcement, Proceedings of SPIE – The International Society for Optical Engineering, vol. 5778, pp. 981–988 (Apr. 2005). 13. H. Eisele, and R. Kamoua, Sub-millimetre wave InP Gunn devices, IEEE Transactions on Microwave Theory and Techniques, vol. 52, 10, pp. 2371–2378 (Oct. 2004).

TERAHERTZ RAYS TO DETECT DRUGS OF ABUSE KODO KAWASE*,1, 2 ADRIAN DOBROIU,2 MASATSUGU YAMASHITA,2 YOSHIAKI SASAKI,2 AND CHIKO OTANI2 1

Nagoya University, Furocho, Chikusa, Nagoya, 4648603, Japan 2 Riken, 2-1 Hirosawa, Wako, 3510198, Japan

Abstract. We have developed compact THz-wave parametric generators with different characteristics that operate at room temperature. One generates high energy and broadband THz waves, being suitable for detecting the transmission of absorptive or diffusive samples, and the other has a potential of wide tunability and narrow linewidth, useful for spectroscopic measurements. In our laboratory, THz waves continue to broaden their range of applications. We have developed a basic technology for THz imaging which allows detection and identification of drugs concealed in envelopes by introducing the component spatial pattern analysis.

Keywords: terahertz, parametric generation, injection seeding, drug detection spectroscopic imaging

1. Introduction One of the most valuable properties of the terahertz (THz) radiation is its ability to pass through a wide range of substances, thus making it possible to “see” through many materials such as paper and cardboard, fabrics, plastics, wood, ceramics, and so on. This property opens the way to nondestructive and noninvasive inspection of packages of many kinds, from mail envelopes in post offices to luggage and personal belongings in airports. It is essential to also note the existence of chemically specific absorption

______ *

To whom correspondence should be addressed. Kodo Kawase, Quantum Engineering Department, Nagoya University, Furocho, Chikusa, Nagoya, 4648603, Japan; e-mail: kawase@nuee. nagoya-u.ac.jp

241 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 241–250. © 2007 Springer.

242

K. KAWASE ET AL.

spectra in the THz range, reflecting molecular transitions and intermolecular bonds, especially in crystalline organic substances. This facilitates chemical identification, and brings about a whole area of spectroscopic detection, testing, and analysis techniques. THz spectroscopy can qualitatively and quantitatively characterize the chemical composition of mixtures containing a wide range of substances. These paired properties – penetration and chemical specificity – make up the unique feature of the THz radiation, not to be found in any other part of the electromagnetic spectrum. Compared to the inspection made possible by x-ray imaging, THz radiation promises some advantages such as highly reduced risk of irradiation, increased image contrast to differentiate between various soft materials, and the possibility of chemical identification. A THz photon is weaker than an x-ray photon by about six orders of magnitude and as such is believed to produce no detectable effect on living tissue. On the other side of the electromagnetic spectrum, the millimeter waves and the microwaves have the disadvantage that they are generally not chemically sensitive, very few substances having fingerprint spectra in this range. In addition, for imaging applications, their longer wavelength translates into poor image resolution. 2. THz-Wave Parametric Generators There are three basic types of parametric sources1: the THz-wave parametric generator (TPG), the THz-wave parametric oscillator (TPO), and the injection-seeded TPG (is-TPG). They are all compact and operate at room temperature, which makes them suitable as practical sources. In principle, both a narrow linewidth and a wide tunability are possible in is-TPG systems with single-longitudinal mode near-infrared lasers as pump sources and seeders. The principle of operation of the parametric sources is as follows. When an intense laser beam propagates through a nonlinear crystal, the photon and phonon transverse wave fields are coupled, and behave as new mixed photon-phonon states, called polaritons. The generation of the THz radiation results from the efficient parametric scattering of laser light via a polariton, that is, stimulated polariton scattering. The scattering process involves both second- and third-order nonlinear processes. Thus, strong interaction occurs among the pump beam, the idler beam, and the polariton (THz) waves. A primitive TPG is shown in Figure 1; it uses just a nonlinear crystal placed in the pump beam. The output of such a device contains a wide range of THz frequencies as there is no frequency selector in place. For an

THz RAYS TO DETECT DRUGS OF ABUSE

243

efficient extraction of the THz wave from inside the crystal, several techniques have been tried, and the one presented in the figure, making use of an array of small silicon prisms, seems to be most suitable. The prism array covers the whole lateral surface of the crystal, increasing the collection efficiency, and minimizing the diffraction effects.

Figure 1. A THz-wave parametric generator (TPG) with a silicon prism array placed on the lateral surface of the LiNbO3 to increase the output and to reduce the diffraction of the THz wave by maximizing the coupling area. Since this setup does not use any frequency selection device, the generated THz radiation contains a wide range of frequencies.

Figure 2. The TPO configuration.

Coherent tunable THz waves can be generated by realizing a resonant cavity for the idler wave. This is the basic configuration of a TPO, and consists of a Q-switched Nd:YAG laser, the nonlinear crystal, and a resonator, as shown in Figure 2. The idler wave is amplified in the resonator consisting of a pair of flat mirrors with a half-area HR coating. The mirrors and crystal are installed on a computer-controlled rotating stage for precise tuning. With the slight variation in the phase-matching condition, the wavelength of the THz wave is tuned between 330 and 100 µm (in frequency from 0.9 to 3 THz); the corresponding idler wavelength changes from 1.075 down to 1.067 µm. In the is-TPG, the THz spectrum specific to a TPG source is narrowed to the Fourier transform limit imposed by the pulse length by introducing the injection seeding for the idler wave. Figure 3 shows our experimental

244

K. KAWASE ET AL.

setup of the is-TPG. The purity of the THz-wave frequency was dramatically improved. Simultaneously, the output power obtained is several hundred times higher than that of a conventional TPG. It was possible to tune the THz wavelength using an external cavity laser diode as a tunable seeder. A wide tunability, from 115 to 460 µm (0.6 to 2.6 THz), was achieved by changing simultaneously the seed wavelength and the seed incident angle; whereas at first, the correlation between the seeder wavelength and the beam angle was realized by a separate computer control, subsequently an optical setup was designed, which eliminates the need of all moving parts. The beam is steered while maintaining a fixed point in the crystal by using a diffraction grating and a lens system.2 To demonstrate

Figure 3. Experimental setup of the is-TPG.

Figure 4. An example of the absorption spectrum measurement of low-pressure water vapor at 1.92 THz. Resolution of less than 100 MHz (0.003 cm–1) was clearly demonstrated.

the continuous tunability and the high resolution of the is-TPG, the absorption spectrum of low-pressure (<1 torr) water vapor was measured. Figure 4

THz RAYS TO DETECT DRUGS OF ABUSE

245

shows an example of measurements around 1.92THz, where two neighboring lines exist. Resolution of less than 100 MHz (0.003 cm –1) was clearly shown. The maximum THz-wave output of 1.3 nJ/pulse (peak power over 300 mW) was obtained with a single-mode pump beam of 34 mJ/pulse and a seed beam of 50 mW. In our previous studies, the maximum THz-wave output from a conventional TPG and a TPO was 1 and 190 pJ/pulse, respectively. 3. Applications Several possible imaging configurations can be imagined for the reflection3 and scattering geometry. One example is schematically presented in Figure 5. A BWO source was used, with the output power in the range of 1 mW and a frequency tunability from about 500 to 700 GHz.4 In another series of experiments we used a microwave-seeded multistage frequency multiplier which delivers about 70 µW of power in the range 600–665 GHz. The output beam is collected by an off-axis parabolic mirror and then focused on the sample by another such mirror. The reflected or scattered beam is collected from the same side of the sample by a lens and then focused onto a detector, as shown on the right-hand side of the same figure. The wave collection is achieved in such a way that, for rather flat samples, the beam

Figure 5. Schematic of the optical setup for imaging in reflection and scattering. The lefthand side of the figure shows the sample illumination system; the right-hand side, showing the collection of the reflected and scattered waves, is seen from a 90°-rotated direction; the dotted circle represents the parabolic mirror focusing the beam on the sample. For demonstration, the inset shows one of the THz images taken with this setup (a paper clip).

246

K. KAWASE ET AL.

reflected specularly does not arrive at the detector. Such an arrangement is useful for increasing the contrast, in a manner similar to the dark-field imaging technique, which has already been used in THz transmission imaging.5 Samples having surface details larger than the wavelength can be imaged using THz waves to reveal those details, even with the sample placed inside a visually opaque container, if this container is transparent in the THz range. Reflected THz waves are generally weak due both to a low reflectivity in most materials and to the fact that these waves are dispersed in a usually wide solid angle. Our first tests realized with a 1-mW beam and using a LiTaO3 pyroelectric detector only had a poor dynamic range. Afterwards, we repeated the same measurements with a DLATGS sensor, which turned out to have a much better sensitivity, according to our estimation about 3,000 times higher than the LiTaO3 sensor.

Figure 6. Powder visibility in scattering mode depends on grain size. (a) A series of sucrose powder samples placed in small polyethylene bags, all containing about the same quantity of powder and having equal size, were imaged with a BWO set at 617 GHz (λ = 486 µm); the grain size in micrometers is indicated under each image. (b) The scattering intensity is plotted against the grain size divided by the wavelength. Each black dot corresponds to one sample; the error bars show actually the grain size range.

THz RAYS TO DETECT DRUGS OF ABUSE

247

Samples that are made of powder do not reflect waves just at the surface, but the incoming waves are scattered from a more or less deeper layer, thus carrying information about the spectral absorption properties of the powder. In addition to providing information on the powder distribution one can thus obtain chemical information about a powder target through THz-transparent packaging, without recurring to transmission imaging. Scattering properties of powders depend on the grain size. Figure 6 shows a demonstration where a series of fructose powder samples of increasing grain sizes was imaged providing an idea of what kinds of powder scatter THz waves efficiently at a given wavelength. This test shows that when the grain size is approximately equal to the wavelength or up to a few times larger than that, the scattering efficiency is highest. The scattering mode was also tested in imaging a sample prepared from fructose powder arranged in the shape of letter R. As we are considering using this technique for noninvasive illegal drug detection in mail, we also checked if such imaging can be performed with the sample inside an envelope, which was simulated by covering the sample with a sheet of usual paper. The two cases show little or no difference, as it can be seen in Figure 7.

Figure 7. (a) Sucrose powder with grain size between 300 and 355 µm, arranged in the shape of letter R and imaged in scattering mode. The source was a microwave-seeded frequency multiplier at 650 GHz (wavelength 461 µm). (b) The same sample, imaged through a sheet of usual paper.

Chemical imaging combines spectroscopy, imaging, and particular data processing methods. A sample containing several chemicals is imaged at a number of frequencies and the data obtained allows the distribution of each chemical within the sample to be calculated. The mathematical development6 of such a calculation method and its application to illicit drug detection7 using a TPG imaging system have already been reported. Here we give another example, for which a THz-TDS (time domain spectroscope) system was used. In this experiment we chose three chemicals – codeine, cocaine, and sucrose – and measured their transmission spectra in

248

K. KAWASE ET AL.

the THz range. As seen in Figure 8, these substances have specific transmission properties, a key factor in their identification. The three chemicals were placed in small polyethylene bags which were in turn inserted in an envelope. The plastic bags overlap partially to check if this has any negative effects on the result. Figure 9 shows the subsequent steps of the process. Transmission images of the target are recorded at a number of frequencies. This number must be equal to or greater than the number of chemicals in the target; in our case sufficiently good results were obtained with eight frequencies.

Figure 8. Transmission spectra of codeine, cocaine, and sucrose from 0.5 to 2.5 THz.

The resulting images show differences in the absorption of each chemical as the frequency is changed, corresponding to their transmission spectra. Using the method called “component spatial pattern analysis,” the images are processed to calculate the distribution map of each chemical, that is, their concentration in each point of the target. In addition to the three substances expected to be found in the target, a fourth distribution map is calculated for the so-called frequency-independent component,8 which shows what part of the data does not change much with frequency, and helps removing nose and diffraction effects from the final result. The last image in Figure 9 is produced as a combination of the three distribution maps, by assigning a color to each chemical component. It proves that the overlapping of the samples does not produce any discernable effects. A predictable application of this technique is the noninvasive detection of illicit drugs in mail dispatching centers.

THz RAYS TO DETECT DRUGS OF ABUSE

249

Figure 9. Chemical imaging demonstration with a target made of three partially overlapped samples: codeine, cocaine, and sucrose. The upper eight images, marked with numbers representing THz frequencies, are the transmission images recorded with a THz-TDS system. These images together with the spectra in Figure 8 are processed, and the distribution map of each chemical is calculated as shown in the images labeled “Codeine,” “Cocaine,” and “Sucrose.” The image marked “F. i.” is a map of frequency-independent effects and is made up of diffraction artifacts and noise. The final image shows the quantity and distribution of each chemical within the target. From left to right: codeine, cocaine, sucrose.

Conclusion Properties of THz radiation were presented and details about several kinds of sources and detectors for this frequency range were given. While it is true that THz-wave applications are still making their first steps, the results shown here demonstrate that this field might change the way we think about homeland security, medical diagnosis, agricultural monitoring, and industrial quality testing9. This work was partially supported by the Asian Office of Aerospace Research and Development through Grant AOARD06-4014. References 1. K. Kawase, J. Shikata, and H. Ito, Terahertz wave parametric source, J. Phys. D: Appl. Phys., 35, R1–R14 (2002). 2. K. Imai, K. Kawase, H. Minamide, and H. Ito, Achromatically injection-seeded terahertz-wave parametric generator, Opt. Lett., 27, 2713–2715 (2002). 3. R. Woodward, B. Cole, V. Wallace, R. Pye, D. Arnone, E. Linfield, and M. Pepper, Terahertz pulse imaging in reflection geometry of skin cancer and skin tissue, Phys. Med. Biol., 47, 3853–3863 (2002).

250

K. KAWASE ET AL.

4. A. Dobroiu, M. Yamashita, Y. Ohshima, Y. Morita, C. Otani, and K. Kawase, Terahertz imaging system based on a backward wave oscillator, Appl. Opt., 43, 5637–5646 (2004). 5. T. Löffler, T. Bauer, K. Siebert, H. Roskos, A. Fitzgerald, and S. Czasch, Terahertz dark-field imaging of biomedical tissue, Opt. Exp., 9, 616–621 (2001). 6. Y. Watanabe, K. Kawase, T. Ikari, H. Ito, Y. Ishikawa, and H. Minamide, Component spatial pattern analysis of chemicals using terahertz spectroscopic imaging, Appl. Phys. Lett., 83, 800–802 (2003). 7. K. Kawase, Y. Ogawa, Y. Watanabe, and H. Inoue, Non-destructive terahertz imaging of illicit drugs using spectral fingerprints, Opt. Exp., 11, 2549–2554 (2003). 8. Y. Watanabe, K. Kawase, T. Ikari, H. Ito, Y. Ishikawa and H. Minamide, Spatial pattern separation of chemicals and frequency-independent components by terahertz spectroscopic imaging, Appl. Opt., 42, 5744–5748 (2003). 9. M. Yamashita, K. Kawase, C. Otani, T. Kiwa, and M. Tonouchi, Imaging of large-scale integrated circuits using laser-terahertz emission microscopy, Opt. Express, 13, 1, 115– 120 (2005).

TERAHERTZ SPECTROSCOPY FOR EXPLOSIVE, PHARMACEUTICAL, AND BIOLOGICAL SENSING APPLICATIONS HAI-BO LIU AND XI-CHENG ZHANG* Center for Terahertz Research, Rensselaer Polytechnic Institute, 1108th St., Troy, NY 12180, USA

Abstract. Terahertz (THz) radiation offers innovative sensing technologies that can provide information unavailable through other conventional electromagnetic techniques. With the advancement of THz technologies, THz sensing will impact a broad range of areas. This chapter focuses on the use of THz spectroscopy for sensing applications in three aspects: explosives detection, pharmaceutical identification, and biological characterization. A THz spectral database in the 0.1–20 THz range was established using both THz time-domain spectroscopy (THz-TDS) and Fourier transform farinfrared spectroscopy. The calculated spectra based on density functional theory and the experimental results are in good agreement in the 3–20 THz range, but not in the 0.1–3 THz range. It is also demonstrated that THz spectroscopy is capable of detecting and identifying the explosive Royal Demolition Explosive (RDX) in diffuse reflection geometry. THz-TDS was applied successfully for pharmaceutical identification, such as identifying hydrated and anhydrous drugs, probing the reaction kinetics of dehydrations, and solid-state reactions of pharmaceutical materials. It was found that most solid-state amino acids, purines, and other small biocompounds have THz absorption features in the 0.1–3 THz range. Solid-state proteins and bioactive protein microsuspensions in organic media also exhibit THz absorption features which may reflect their collective vibrational modes and could be used to probe their functional 3D conformation states. Additionally, owing to the high sensitivity of differential THz-TDS, it was successfully used to sense

______ *

To whom correspondence should be addressed. Xi-Cheng Zhang, Center for Terahertz Research, CII 9009, Rensselaer Polytechnic Institute, 1108th St., Troy, NY 12180, USA; e-mail: [email protected]

251 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 251–323. © 2007 Springer.

H.-B. LIU AND X.-C. ZHANG

252

the minute change of biological cell monolayers. These studies have pointed to new ways for using THz spectroscopy in pharmaceutical and biological sensing.

Keywords: terahertz spectroscopy, THz time-domain spectroscopy, sensing, explosive, density functional theory, diffuse reflection, pharmaceutical, pseudopolymorph, dehydration, solid-state reaction, biological, small biomolecule, protein, cell monolayer

1. Introduction Terahertz (THz) waves, or far-infrared waves, refer to the electromagnetic radiation in the frequency interval from 0.1 to 20 THz (There are several definitions for the THz range in THz community, e.g. 0.1–4 THz, 0.1–10 THz, or 0.1–20 THz. Here 0.1–20 THz is used). They occupy a large portion of the electromagnetic spectrum between the mid-infrared and microwave bands (in different units: 1 THz ↔ 1 ps ↔ 300 µm ↔ 33.3 cm−1 ↔ 4.1 meV ↔ 47.6 K). Throughout this chapter, the terms “THz” and “far-infrared” are used interchangeably. Electromagnetic spectroscopy has found numerous applications for chemical and biological sensing and screening. However, the spectroscopic study in the THz range was relatively limited due to the lack of powerful sources and efficient detectors in the past. In comparison with well-developed technologies and widespread applications in microwave, mid-infrared, visible, ultraviolet, and x-ray bands, fundamental research, advanced technological developments and applications in the THz band are still in infancy. During the last decade, various THz technologies have been undergoing rapid development and there has been an increasing interest in employing THz technologies for chemical and biological sensing. In brief, there exist four major motivations contributing to this interest: •

•

The THz spectral range is relatively unexplored. THz radiation may provide information unavailable through other conventional methods. New THz technologies and investigations will enable new chemical and biological sensing modalities in both fundamental scientific research and technological applications. In the THz range, many biological and chemical compounds exhibit characteristic absorptions and dispersions due to the vibrational transitions, mostly collective vibrational modes and intermolecular vibrational

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

•

•

253

modes (phonon modes). These THz vibrational modes provide fingerprint information for THz sensing and identification. THz waves show good penetration through numerous commonly used nonpolar dielectric materials that are opaque for visible or mid-infrared light, due to either a relative low scattering or the absence of THz vibrational modes in these materials. This enables THz technologies to inspect the insides of concealing barriers, such as paper, textile, plastic, wood, leather, and porcelain. THz waves have low photon energies (4 meV at 1 THz, one million times weaker than an x-ray photon) and will not cause harmful photoionization, therefore they are safe for both operators and targets (noninvasive and nondestructive).1, 2

In the last decades, especially since the advent of THz time-domain spectroscopy (THz-TDS), THz spectroscopy has found extensive applications in various fields, including astrophysics, atmospheric science, gas sensing, solid and liquid chemical compounds characterization and analysis, insulating, conducting, and semiconductor materials characterization, explosives detection, and security screening, pharmaceutics, biological, and biomedical study, etc. In recent years THz spectroscopy and imaging of explosives and related compounds (ERCs) have been investigated intensively with THz-TDS for potential defense and security applications. In 2003, it was firstly reported that there existed THz spectral features (0.1–3 THz) in a list of common energetic materials, such as 2, 4, 6-trinitrotoluene (TNT), hexahydro-1, 3, 5-trinitro-1, 3, 5-triazine (RDX, i.e. Royal Demolition Explosive), tetramethylene tetranitramine (HMX, i.e. High Melting Explosive), and pentaerythritol tetranitrate (PETN).3 Subsequent studies4, 5 validated the THz absorption of RDX. Most of the THz absorption features below 3 THz were assumed to arise from the intermolecular vibrational modes or phonon modes in solid-state explosive materials. Furthermore, detection of RDX by THz specular reflection spectroscopy and imaging has also been demonstrated recently.6 Compared with traditionally used inspection techniques such as x-ray imaging, THz imaging offer advantages including spectroscopic information, nonionizing irradiation, and higher image contrast for differentiating between various soft dielectric materials. These advantages stem from the fact that THz radiation, being more selective than x-ray, is more sensitive to the nature of these materials it passes through.7 THz-TDS was utilized in pharmaceutical research for the first time in 2002.8 It has proved to be a quick, simple, and versatile technique to identify

254

H.-B. LIU AND X.-C. ZHANG

and quantify the polymorphs of pharmaceutical materials.8–10 The different polymorphs or different crystallinity of a pharmaceutical substance exhibit different THz absorption features, presumably reflecting their intermolecular vibrational modes mediated by hydrogen bonds. Polymorphism and variations in the degree of crystallinity in a pharmaceutical substance may exhibit physicochemical differences that impact at therapeutic, manufacturing, commercial, and legal levels. In addition, it was demonstrated that THz-TDS was a useful technique for the nondestructive analysis of tablet coating thicknesses.11 These investigations imply that THz spectroscopy has promise to become a process analytical technology (PAT) proposed by US Food and Drug Administration (FDA), to enable the pharmaceutical industry to efficiently test quality during manufacturing process. It has also been found that many solid-sate biocompounds have THz absorption fingerprints, reflecting their intermolecular vibrational modes. The reported biocompounds include nucleotides,12–14 amino acids,15–19 vitamins,20 sugars,21,22 and other small biomolecules.23,24 Sensing these biocompounds is of significant interest in biological study or pharmaceutical research and industry. Low-frequency (THz) collective vibrational modes of proteins, DNA, and other large biomolecules may provide information about their 3D conformation states. For instance, the theoretical calculation indicate that some low-frequency modes are associated with collective motion of the tertiary subunits moving with respect to one another, or coherent movement of a portion of a structural subunit.25 It has been calculated that protein collective vibrational modes lie in the THz range for bovine pancreatic trypsin inhibitor (BPTI), a small globular protein.26,27 In the past, limited experimental investigations on the THz spectroscopy of proteins have been reported. Low-frequency spectra in the 0–4.5 THz range of α-chymotrypsin in solid form and lysozyme in aqueous solution were studied using Raman spectroscopy and some THz spectral features were found to be dependent upon the conformations of the protein molecules.28,29 Solid-state proteins including lysozyme, myoglobin, albumin, bovine serum albumin, and collagen were also investigated with FTIR and THz-TDS.15,30,31 The THz vibrations may play a crucial role in the biological functions of proteins since they give rise to significant atomic rearrangements and conformational fluctuations.32–34 In addition to the fundamental research in biology, THz waves have also found significant applications in biosensing and biomedical imaging. THz-TDS was used as a label-free probe to sense the binding state of DNA

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

255

via a planar resonator approach, potentially giving rise to a label-free genetic analysis system and future THz-based biochip technology.35 Due to its high sensitivity, the differential THz-TDS was applied for label-free bioaffinity (avidin–biotin binding) sensing and a comparable sensitivity as traditional technologies was achieved.36,37 Measurements on the complex dielectric properties of biological materials in the THz region date back at least to 1976.38 Several research groups have investigated excised and fixed tissue samples, either alcohol perfused,39 formalin fixed,40–43 or freeze dried and wax mounted,44 looking for inherent contrast to define unique modalities. Other groups have successfully applied THz pulsed imaging in reflection geometry for the study of human in vitro wet tissue45,46 and in vivo skin tissue (skin cancer diagnostics).47,48 However, the use of THz spectroscopy as a means for sensing is still immature. A large span of investigations and corresponding improvements of THz technologies are desired to enable widespread sensing modalities. We have applied THz spectroscopy in multidisciplinary investigations to explore feasibilities for both basic scientific research and technological applications. Some topics, including the THz spectroscopy of explosives and related materials, pharmaceutical substances, small biocompounds, and proteins, are continuative studies following research peers in the THz community. Some others are initiating investigations, such as identifying anhydrous and hydrated drugs via THz-TDS, studying dehydrations, and solid-state reactions of drug substances with THz-TDS, and sensing changes of cell monolayers with THz differential time-domain spectroscopy. These studies have opened new avenues for using THz spectroscopy in pharmaceutical and biological sensing. 2. THz Spectroscopy of Explosives and Related Compounds We have investigated the THz spectroscopy of a list of selected solid-state high ERCs. The explosive compounds usually contain nitro groups (NO2). Attaching three or four nitro groups to a compound leads to an extremely unstable situation. An explosion is the chemical reaction process of the substance transforming rapidly and violently into the gaseous state. Besides explosive materials, a number of explosive-related compounds were also investigated. These explosive-related compounds are the explosive byproducts or degradation products and are usually contained in explosive materials. For example, dinitrotoluene (DNT), dinitrobenzen (DNB), trinitrobenzen (TNB) are TNT’s by-products in manufacturing. The concentrations of these by-products depend on the manufacturing processes

256

H.-B. LIU AND X.-C. ZHANG

and the purification. In landmine leakage the DNT/TNT ratio is usually greater than one. 2-amino-4, 6-DNT and 4-amino-2, 6-DNT are degradation products of TNT. In soil, degradation rates yield half-lives of 40 and 100 days for 2, 4-DNT and TNT, respectively.49 Sensing these explosiverelated compounds will also contribute to explosives detection. There are various explosives detection methods, including bulk detection technologies, such as x-ray imaging, nuclear detection, infrared imaging, and trace detection technologies, such as mass spectroscopy, optical absorption, and fluorescence, anti-Stokes Raman scattering (CARS), and biological sensors.50 Limitations of some bulk detection methods are provided below in order to compare with THz technologies. x-rays have been used for many years to search for explosives and other contrabands in luggage and cargo containers. However, there are health concerns when people are exposed to them since x-rays are ionizing. Most of all, x-rays do not provide spectroscopic fingerprints of explosives. Nuclear detection suffers from a combination of health hazards and limitations in sensitivity for standoff detection. The disadvantage of infrared imaging is the lack of specificity for explosives. Since all these technologies have their own limitations, new advanced technologies are still highly desired due to the increasing security concerns. THz technologies have recently been recognized as new methods for the detection of bulk explosives due to their unique advantages, and may become complementary or competing techniques in the near future. In this chapter, THz technologies including THz-TDS and Fourier transfer farinfrared (FT-FIR) are tested as a potential sensing modality for explosives detection. 2.1. EXPERIMENTAL RESULTS (1.5–20 THz) AND COMPARISON WITH CALCULATIONS

2.1.1. Samples and methodology Acetone solutions of TNT, RDX, HMX and PETN and other solid-state ERCs (all with purity >99%) were purchased from AccuStandard, USA. Solid-state explosives were recrystallized from the solutions. The samples for FT-FIR measurements were prepared by casting films (with weights of 0.5–5 mg and thicknesses of 50–100 µm) from their solutions on polyethylene film substrates (Thermo Electron, Co.), which is almost transparent in the THz band. THz spectra of selected ERCs in the range of 2–20 THz were taken with a Bruker 66V/S FT-FIR spectrometer in transmission mode. A glowbar

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

257

source, a multilayer mylar T222 beam splitter, and a FIR-DTGS detector were used for the measurements. The spectra were recorded at 60 GHz (2 cm−1) resolution and with 16 coadded scans. The measurements were conducted under a vacuum with an atmospheric pressure of ~1 torr. The diffuse reflectance spectrum is extensively used in UV/visible and middle/near infrared bands for characterizing and analyzing powders and samples with rough surfaces. This spectroscopic technique has seldom been used in the THz region previously. For standoff THz sensing, the measurement in the diffuse reflection geometry will be more desired than the traditional transmission mode. This section also presents an investigation on the diffuse reflectance spectra of ERCs powders using FT-FIR spectrometer in combination with a diffuse reflectance accessory in the 2–20 THz range. The diffuse reflectance spectroscopy of RDX using THz-TDS coupled with a diffuse reflectance accessory will be discussed in the subsequent section.

Figure 1. Schematic diagram of the diffuse reflectance accessory.

THz diffuse reflectance spectra (2–20 THz) of ERCs powders were taken with a Bruker 66V/S FT-FIR spectrometer in combination with a Specac 19900 diffuse reflectance accessory (Specac, UK). A schematic diagram of the diffuse reflectance accessory is shown in Figure 1. In the diffuse reflectance accessory, the THz beam was focused on the sample by an off-axis ellipsoidal mirror. The incident angles are around 45°. About one quarter of all the diffusely reflected THz radiation was collected by the second off-axis ellipsoidal mirror. The unique configuration of this accessory deflected the specular reflection away from the collecting ellipsoidal

258

H.-B. LIU AND X.-C. ZHANG

mirror. The focal length of the two ellipsoidal mirrors was about 5 cm. The spectra were recorded at 120 GHz (4 cm−1) resolution and with 200 coadded scans. Each measurement took approximately 30 min. To eliminate the effect of water absorption, nitrogen was used to purge the system (with a relative humidity of ~0%). A number of ERCs were investigated under diffuse reflectance modes via FT-FIR and the spectra were presented in subsequent section, compared with calculation spectra and transmission spectra. The sample spectra are normalized to the spectrum of pure polyethylene powder (20–30 µm in diameter) and converted to a Kubelka– Munk51 plot by F(R∞) = (1− R∞)2/2R∞,

(1)

where R∞ is the diffuse reflectance of an optical thick sample. From the Kubelka–Munk theory it can be rewritten as F(R∞) = K/S,

(2)

where K is the absorption coefficient, and S is twice the scattering coefficient. 2.1.2. Comparison of calculation, transmission, and diffuse reflectance spectra The density functional theory (DFT) calculation has proven to be a reliable theoretical method to predict accurate vibrational frequencies for typical medium-size molecules and to interpret the measured infrared and Raman spectra. DFT calculations for RDX and PETN both predict structures essentially identical in geometry to those observed in the solid states.52,53 A DFT study of the structure and vibrations of TNT was also reported.54 In our study, the Gaussian 03 package55 based on DFT was employed for the THz spectral calculations of ERCs. To simplify the calculations, all calculations correspond to the absolute temperature zero (T = 0 K). The full geometric optimization and frequency computation of several ERCs were performed using the Gaussian 03 package with the Becke-3– Lee–Yang–Parr (B3LYP) functional56, 57 and 6−311+G** basis set.58,59 Minimum energy structures were found, which were confirmed by vibrational analyses. No negative frequencies were found. Six molecules (TNT, RDX, HMX, 2, 4-DNT, 1, 3-DNB, and 1, 3, 5TNB) were calculated. The calculated spectra were shown in Figure 2, together with the experimental results (transmission measurements and

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

(a) TNT

1.2

Calculation

0.3 Transmission 0.2 Diffuse reflection

0.1

Absorbance (a.u.)

Absorbance (a.u.)

0.4

0.0 2

4

6

0.8 Transmission

0.6 0.4

Diffuse reflection 0.2 0.0

8 10 12 14 16 18 20 Frequency (THz)

2

4

0.6 Calculation

0.4 Transmission

Absorbance (a.u.)

0.8

0.5

6

8 10 12 14 16 18 20 Frequency (THz)

(d) 2,4-DNT Calculation

0.3 Transmission 0.2 Diffuse reflection

0.1 0.0

2

4

6

8 10 12 14 16 18 20 Frequency (THz)

2

0.8

(e) 1,3-DNB Calculation

0.4 0.3

Transmission

0.2 Diffuse reflection

0.1

Absorbance (a.u.)

Absorbance (a.u.)

Calculation

0.4

0.0

Absorbance (a.u.)

(b) RDX

1.0

1.0 (c) HMX

0.2

259

4

6

8 10 12 14 16 18 20 Frequency (THz)

(f) 1,3, 5-TNB Calculation

0.6 0.4

Transmission

0.2

Diffuse reflection

0.0

0.0 2

4

6

8 10 12 14 16 18 20 Frequency (THz)

2

4

6

8 10 12 14 16 18 20 Frequency (THz)

Figure 2. Comparison of calculated spectra, transmission spectra, and diffuse reflectance spectra of (a) TNT, (b) RDX, (c) HMX (no diffuse reflectance spectrum), (d) 2, 4-DNT, (e) 1, 3-DNB, and (f) 1, 3, 5-TNB measured with FT-FIR. The spectra were vertically offset for clarity.

H.-B. LIU AND X.-C. ZHANG

260

Transmission

0.6 0.4

Diffuse reflection

0.2 0.0 2

4

6

8 10 12 14 16 18 20 Frequency (THz)

Absorbance (a.u.)

Absorbance (a.u.)

(a) 2-amino-4,6-DNT

0.8

(b) 4-amino-2,6-DNT

0.6

Transmission

0.8

0.4 Diffuse reflection 0.2 0.0 2

4

6

8 10 12 14 16 18 20 Frequency (THz)

Figure 3. Transmission spectra and diffuse reflectance spectra of (a) 2-amino-4, 6-DNT and (b) 4-amino-2, 6-DNT measured with FT-FIR. The spectra were vertically offset for clarity.

diffuse reflectance measurements using FT-FIR at room temperature) for comparison. In general, the calculated spectra agree with the measured spectra. The deviations between the calculated and measured spectra are attributed to the temperature dependence of the vibrational modes resulting from the anharmonicity of the vibrational potentials (the calculations correspond to the absolute temperature zero and the experiments were conducted at room temperature), packing force and the steric effect in the polycrystalline materials. Based on the calculated results of B3LYP/ 6−311+G** and with the aid from the visualization software in Gaussian View 3.09, the observed vibrational modes in the THz region were assigned. Figure 3 shows the absorption spectra obtained from transmission measurements and diffuse reflectance measurements for two additional ERCs. For the diffuse reflectance spectra, it becomes noisier as the frequency reaches below 4.5 THz. It was caused by a small amount of water vapor in the measurements and the low energy received by the detector at low frequencies under diffuse reflectance geometry. 2.2. EXPERIMENTAL RESULTS (0.1–3 THz) AND COMPARISON WITH CALCULATIONS

In the range of 0.1–3 THz, compared with FT-FIR, THz-TDS has advantages such as a higher SNR without using a liquid helium-cooled bolometer detector. The THz spectra of common ERCs were measured using THz-TDS in transmission mode. The samples for THz-TDS measurements were prepared as pellets. All the sample powders were gently ground using a mortar and pestle to reduce particle size (<50 µm) therefore minimize scattering. Some samples were used in pure form and some were prepared by mixing the sample powders

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

261

with polyethylene powder at different weight ratios, depending on their different THz attenuations and the available quantity of the pure samples (it is difficult and expensive to obtain explosive compounds). The pellet compressed from the polyethylene powder (<50 µm) is almost transparent in the THz range. The powder samples were compressed into pellets (~1 mm in thickness and 13 mm in diameter) using 4 tons of pressure with a hydraulic press. The measurements were conducted with THz-TDS system (bandwidth: 0.1–3 THz; resolution: 0.1 THz; zero-filling factor: 4) in a transmission mode under a nitrogen purge (relative humidity ~0%).

1.70 -1

~2.2

n

-1

α (cm )

1.96

0.82

60

60

1.72

1.04 ~1.5

40

1.68

40

1.74

30

1.72

2.8

2.82

HMX (35% w/w)

60 2.51

70

1.58

60

1.56

50

40

1.54

30

1.52

1.78

~1.2

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

1.68 2.8

1.66

1.54 PETN (50% w/w)

n

-1

α (cm )

50

1.60

0.4

0.74

2.00

2.84 1.52 2.17

1.50

40

1.48

30

1.46

n

1.2 1.6 2.0 2.4 Frequency (THz)

-1

0.8

0

α (cm )

70

0.4

1.64

1.70

~1.7

1.66

10 0

~2.2 1.76

20 20

1.78

TNT (pure)

50

n

RDX (35% w/w)

a (cm )

80

20

1.50

20

1.44

10

1.48

10

1.42

0

1.46

0

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

1.40

Figure 4. THz absorption and dispersion spectra of RDX, TNT, HMX, and PETN.

The absorption spectra and refractive indices of the samples RDX, TNT, HMX, and PETN are shown in Figure 4. The absorption spectra actually include the scattering backgrounds, resulting in increasing baselines over frequency (it is the same for other absorption spectra throughout the whole chapter). The absorption coefficients and dispersions of pure explosive materials are supposed to be larger than those of the mixture samples. The higher dispersions of explosives in the THz band are desired for the identifications of these explosives in reflection geometry. Figure 5 contains the absorption spectra of selected ERCs. Most of these

H.-B. LIU AND X.-C. ZHANG

262

low-frequency absorption features possibly result from intermolecular vibrations or phonon modes of polycrystalline ERCs. 2-amino-4,6-DNT (pure)

1.87

60

80 1.43

40

0

60

1.24

20 0.52

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

0

2.8

2.57

2,4-DNT (50% w/w)

60

-1

α (cm )

1.08

-1

α (cm )

2.64

40

0.96

40 1.36 0.45 0

80

0.4

0.4

1.2 1.6 2.0 2.4 Frequency (THz)

2,6-DNT (35% w/w)

1.10 1.2 1.6 2.0 2.4 Frequency (THz)

2.8

2.50

40

0.66 0.8

0.8

20

20

0

2.8

4-NT 1.20 1.86 (50% w/w) 1.36

80

60

0.4

0.8

1.35 1.56

1.2 1.6 2.0 2.4 Frequency (THz)

1,3-DNB (50% w/w)

2.8

2.37

60 -1

α (cm )

-1

α (cm )

4-animo-2,6-DNT (pure)

-1

α (cm )

-1

α (cm )

120

40 20 0

40 0.94 1.19 20

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

0

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

Figure 5. THz absorption spectra of selected ERCs. % w/w indicates the weight percentage of ERCs in polyethylene.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

Absorbance (a.u.)

0.4

263

2,4-DNT solution 6.5

0.3 0.2 2.54 0.1

Solid-state 2,4-DNT 0.0

2

4

6

8 10 12 14 Frequency (THz)

16

18

20

Figure 6. THz absorption spectra of 2, 4-DNT cast film (lower trace) and 6 × 10−3 mol/L 2, 4-DNT solution in toluene (upper trace). The low-frequency mode at 2.54 THz is from an intermolecular vibrational mode or phonon mode. The peak at 6.5 THz is related to toluene’s absorption.

However, only solid-sate crystalline or polycrystalline ERCs exhibit fingerprints in the 0.1–3 THz range. This was confirmed by experiments. The THz absorption spectra of the 2, 4-DNT solution and the 2, 4-DNT cast film are displayed in Figure 6. Most absorption peaks of solid-state 2, 4DNT exist in the spectrum of 2, 4-DNT solution, reflecting the intramolecular vibrational modes. The absorption peak at 2.54 THz, which disappears in the spectrum of 2, 4-DNT solution, possibly arises from an intermolecular vibrational or phonon mode. The absence of these lowfrequency absorption features was also confirmed for the other chemical compounds in an amorphous state. On the other hand, 2, 4-DNT and 2, 6DNT are isomers and suppose to have very similar intramolecular vibration modes. However they have different THz absorption features in the 0.1–3 THz range, indicating that these THz spectral features very likely represent the intermolecular vibrational or phonon modes in the solid-state materials. Therefore, explosives detection using these THz spectral fingerprints will not be feasible if explosives are in a liquid or amorphous form. Fortunately, most of the ERCs are polycrystalline materials. Some ERCs, such as 2, 4DNT, will keep the polycrystalline form even after it recovers from a melting state. The calculated spectra using Gaussian 03 (not shown) and the experimental results obtained via THz-TDS in the 0.1–3 THz range are not in good agreement because the DFT calculation only obtains the intramolecular vibrational modes. Calculating the intermolecular vibrational or phonon modes remains currently as a difficult problem due to the complexity

264

H.-B. LIU AND X.-C. ZHANG

involving the periodic density functional theory modeling. It will be a topic for our future investigation. According to our measurements, several ERCs do not contain any THz absorption features in the range of 0.1–3 THz, such as tetryl, 1, 3, 5-TNB, and 1, 4-DNB. An additional example is ammonium nitrate which was reported previously.60 However, they have fingerprints in the range of 3–20 THz. The detection of these ERCs requires special considerations since THz waves in the higher frequencies can not penetrate through many commonly used dielectric materials and their atmospheric attenuation is also different from those in the 0.1–3 THz range. The spectral fingerprints of ERCs in the range of 0.1–3 THz are of great significance for inspecting hidden explosives since THz waves in this range have high penetration through most commonly used dielectric materials, which are opaque for visible and infrared light or exhibit low contrast to x-rays. This makes THz spectroscopy or imaging uniquely applicable technologies for sensing hidden explosives behind some barrier dielectric materials. Particularly, the spectra fingerprints in the range below 1.5 THz are desired for the standoff explosives detection due to the relative low atmospheric attenuation in this range. 2.3. THz SPECTRAL DATABASE FOR EXPLOSIVES SENSING

Table 1 summarizes the absorption peak positions of solid-state ERCs in the 0.1–20 THz range, obtained with both the THz-TDS and FT-FIR spectrometers. Some of the data agree with the literature results61–63 and some others are new observations. The THz spectral database of ERCs is supposed to expand further with an increasing interest in explosives detection over the world. The THz fingerprints have a better specificity for sensing ERCs than the spectral features in the mid-infrared. The THz features contain very specific molecular information of ERCs, reflecting either the intermolecular vibrational models or the normal vibrational modes of the whole molecule, such as in-plane/out-plane bending and torsion of benzene ring. The spectral features in the mid-infrared range usually only reflect the vibration modes of single molecular bonds, which are also exhibited in many other similar chemical compounds (non-ERCs), resulting in difficulty in identification.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

265

TABLE 1. THz absorption peaks (THz) of ERCs ERCs

Measured absorption peak position (THz)

TNT

PETN

0.74, 1.2, 1.7, 2.20, 3.70, 4.45, 5.55, 8.27, 9.14, 9.78, 10.64, 11.01,13.86, 15.15, 16.95, 17.36, 19.17, 19.87 0.82, 1.04, 1.50, 1.96, 2.20, 3.08, 6.73, 10.34, 11.34, 12.33, 13.86, 14.54, 17.74, 18.12, 20.13 1.78, 2.51, 2.82, 3.42, 5.32, 6.20, 11.30, 12.00, 12.56, 12.96, 13.75, 14.55, 18.18, 18.51, 18.60, 19.40 2.0, 2.17, 2.84

Tetryl

5.97, 10.11, 11.28, 14.67, 16.14, 18.36

RDX HMX

2-amino-4, 6-DNT

0.96, 1.43, 1.87, 3.96, 5.07, 6.27, 8.49, 9.87, 10.77, 12.15, 13.44, 16.68

4-amino-2, 6-DNT

0.52, 1.24, 2.64, 3.96, 5.04, 5.82, 7.53, 9.30, 10.20, 11.13, 13.86, 14.97, 17.70

4-NT

1.20, 1.36, 1.86, 6.75, 8.85, 10.83, 14.04, 15.66, 18.51

1, 3, 5-TNB

4.17, 4.62, 10.05, 11.19, 13.80, 15.75, 19.05

1, 3-DNB

0.94, 1.19, 2.37, 10.56, 12.18, 15.33, 17.13

1, 4-DNB

3.24, 3.96, 5.55, 10.38, 12.45, 13.29, 15.21, 15.54

2, 4-DNT 2, 6-DNT

0.45, 0.66, 1.08, 1.38, 2.51, 4.96, 8.86, 10.58, 11.62, 12.81, 14.34, 15.74, 19.05, 20.00 1.10, 1.35, 1.56, 2.50, 5.61, 6.75, 9.78, 11.43, 13.32, 13.89, 15.39, 17.25

3, 5-dinitroaniline

0.96, 1.20, 3.18, 4.62, 5.04, 5.91, 7.44, 10.62, 10.98, 14.46, 16.41, 18.18

Many of these low-frequency phonon modes are also accessible to Raman spectroscopy. But the laser irradiation in Raman spectroscopy may induce a phase change, or initiate photochemical reactions in the samples. In addition, many vibrational modes are far-infrared active, but not Ramanactive. Particularly, Raman spectroscopy is not able to inspect targets behind barrier materials since the infrared or visible laser irradiation can not penetrate through most of dielectric materials. The absorption peaks in the THz band are generally many times lower than those in the mid-infrared range, resulting in a relatively low sensitivity for THz sensing. However, this is shadowed by the fact that THz waves have much higher penetration for many nonpolar dielectric materials than visible or infrared waves, which is critical to sense hidden explosives behind covers or inside packages. It has also proved that these THz fingerprints exhibit good repeatability according to our experiments and the literature. Figure 7 shows the absorption spectra of RDX obtained from four countries including UK61, Japan62, China, and USA (our lab).64 Five absorption peaks in the 0.1–3 THz range were well identified and are in good agreements using different samples and THz-TDS systems.

266

H.-B. LIU AND X.-C. ZHANG

Figure 7. Comparison of the absorption spectra of RDX from different countries.

If a real-world scenario was considered, both the atmospheric attenuation and a scattering effect would have significant effects on the spectroscopic results. Therefore, THz sensing in the transmission windows is desirable for practical applications. ERCs have many distinguishable THz fingerprints within the range of 10.7–20 THz where THz waves also have low atmospheric attenuation. However, THz waves in this range do not have high penetration (<1 mm) for dielectric materials compared with that in lower frequency ranges. Therefore the standoff detection in the range of 10.7–20 THz is limited to exposed explosives. To better employ the THz fingerprints of ERCs and mitigate atmospheric attenuation, further investigations need to be conducted. Sensing with narrow-linewidth CW THz waves is another promising direction due to their portability, low cost, and high power for single frequency. Particularly, the CW THz sensing frequency can be selected to be within the absorption or dispersion bands of ERCs and away from the water vapor absorption lines. An array of CW sources and detectors covering a range wherein ERCs have THz fingerprints and the atmosphere has low attenuation will be more applicable for standoff sensing.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

267

2.4. SENSING RDX WITH THz DIFFUSE REFLECTION SPECTROSCOPY64

When THz waves illuminate a flat sample surface, the reflected THz beam with a reflection angle the same as the incident angle can be theoretically treated using the Fresnel equations and it is called specular Fresnel reflection. In a real-world scenario, the targets are usually not flat and not aligned normal to the THz beam. Thus the direction of the specular reflection is hard to determine. It is more feasible to detect the diffusely reflected THz waves in real-world sensing applications. Figure 8 schematically shows the reflection geometry consisting of both specular Fresnel reflection and diffuse Fresnel reflections.51 This diffuse Fresnel reflection is different from what is described in previous section, which deals with powder samples instead of bulky samples discussed here. Since the sample surface is not optically flat for THz waves, the diffuse Fresnel reflections are caused by some tiny planes statistically distributed at all angles, as shown in Figure 8 Although the reflected THz waves from these small surfaces undergo specular reflection, the collected THz waves are seemingly the diffuse reflected parts with respect to the macroscopic sample surface. Therefore, we use the term “diffuse Fresnel reflection.” Since most of the sample surface is parallel with the macroscopic sample surface, reflection from that portion of the surface is dominant and is termed specular Fresnel reflection. However, studying the small amount of diffuse Fresnel reflection is more applicable than specular reflection for standoff detection of bulky explosives in the real world where target orientation with respect to the irradiation beam is randomly distributed.

Figure 8. Rough sample surface results in diffuse reflection. Specular reflection, indicated by the solid lines, has a reflection angle equal to the incident angle. Diffuse reflections, denoted by the dashed lines, have reflection angles that are independent of the incident angle.

268

H.-B. LIU AND X.-C. ZHANG

RDX is a highly explosive compound with both military and civilian applications. C-4, a plastic explosive containing 91% RDX in weight can be easily molded and is barely detectable by x-ray imaging. New noncontact and nondestructive techniques, such as THz spectroscopy, are desired to develop for effective detection of explosives like C-4. The objective of this section is to apply THz-TDS to investigate diffuse reflection spectroscopy of explosives using RDX as a model compound. 2.4.1. Samples and methodology Acetone solution of RDX (purity >99%) was purchased from AccuStandard, USA. Solid-state RDX was recrystallized from the solution. Two commonly used materials without THz absorption features, polyethylene and flour, were also tested to compare with RDX. Powdered materials of RDX, Teflon, polyethylene, and flour were ground to fine particles below 50 µm in diameter. The pure powder of each material was compressed into a pellet with a thickness of ~2 mm and a diameter of 13 mm. A Teflon pellet and a copper plate were used in the experiment as references. The surfaces of pellet samples and copper plate were not perfectly flat for THz waves. The approximate length of the surface variations was <50 µm for the pellet sample and <10 µm for the copper plate. As a result, small parts of THz waves can be diffusely reflected from the samples.

Figure 9. A schematic diagram of THz-TDS with a diffuse reflectance accessory.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

269

Figure 9 depicts the schematic diagram of a THz-TDS setup coupled with a diffuse reflectance accessory (Specac, UK). The emitted THz beam was collimated and focused by a pair of parabolic mirrors, then propagated through the diffuse reflectance accessory wherein sample was placed. In the diffuse reflectance accessory, the THz beam was focused on the sample by an off-axis ellipsoidal mirror and the beam size at the focal point is about 2 mm. The incident angles are around 45°. A quarter of all the diffusely reflected THz radiation was collected by the second off-axis ellipsoidal mirror. The unique configuration of this accessory deflected the specular reflection away from the collecting ellipsoidal mirror. The focal length of the two ellipsoidal mirrors was ~5 cm. About 1/10,000 of incident THz power was collected and detected in the diffuse reflection measurements. 2.4.2. Theoretical background THz-TDS can be used to measure both the phase and the amplitude of THz waves. Therefore, the complex refractive index of a sample can be obtained without using the Kramers–Kronig (K–K) transform.65 In reflection mode, since the phase of the THz pulse depends on the position of the reflected surface, an accurate phase measurement requires a reference reflector exactly in the same position as the sample. The reference reflector should have a very similar surface morphology as the sample as well. In real-world applications, this match is hard to realize. Thus for the samples that are not optically smooth for THz waves, the retrieval of phase information from the time-domain signal is difficult for reflection measurements. Although several techniques have been proposed to minimize this phase error,66–69 it still remains a bottleneck for THz-TDS in specular reflection geometry. Diffuse reflection is even more complicated than specular reflection. Retrieval of the phase has proved to be a difficult problem in this investigation. Considering this difficulty, we discarded phase information obtained in the reflection measurement. The K–K transform was used to obtain the absorption spectrum from the measured reflection spectrum. If the complex refractive index of the sample is N = n + ik, the amplitude reflectivity r and power reflectivity R are expressed as

r = Reiθ ,

(3)

where θ is the phase shift defined in terms of the Fresnel reflection coefficient. The K–K dispersion relationship is used to determine the spectrum of the complex refractive index from the specular reflection spectrum in many conventional spectroscopy technologies.70 R and θ in Eq. (3) are mutually correlated according to the K–K equation

270

H.-B. LIU AND X.-C. ZHANG

2ν 0

θ (ν 0 ) =

π

∞

P∫ 0

ln R(ν )

ν 2 −ν 02

dν ,

(4)

where P is related to the Cauchy principal value. Using this equation, the phase change at any frequency ν can be calculated when the power reflectivity R(ν) is obtained. Then the real and imaginary parts of the complex refractive index are calculated via the following formulas: n=

1− R 1 + R − 2 R cosθ

k=

2 R sin θ 1 + R − 2 R cos θ

(5) ,

(6)

The absorbance can be calculated as

α=

4πνk . c

(7)

The real part of the refractive index can be calculated in terms of Eq. (5). It reflects the similar spectral characteristics as the imaginary part of the refractive index based on K–K relationship. Since the absorption spectrum (imaginary part) is commonly used in sensing applications, the real part of the refractive index was not studied in this investigation. Strictly, the reflection spectrum needs to be measured at near-normal incidence to use the above equations. In this diffuse reflection measurements, incident angles have a range at around 45o (from ~35o to ~55o). To our best knowledge, there is no completely sound theory for retrieving the absorption spectrum from the diffuse reflection data obtained in this investigation. Utilizing the K–K transform and Fresnel equation to calculate the absorption spectrum in this case was an approximation. Although the absorption intensity obtained from Eqs. (6) and (7) has an arbitrary unit, the relative absorption spectrum provides sufficient fingerprint information for sensing explosives. 2.4.3. Experimental results and discussion The diffusely reflected THz pulses from the RDX and from the Teflon pellet surfaces were measured in a chamber purged with nitrogen to avoid the effect of water vapor absorptions in ambient air. The measured amplitude of diffusely reflected THz waves was about 1/100 (1/10,000 for the THz power) of that of the incident THz beam. Therefore, the dynamic range for the power spectrum of diffusely reflected THz waves was

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

271

approximately 20 dB (60 dB for nonreflection freepath measurements as mentioned previously). For each sample, we measured 10 different spots on the sample surface to reduce the surface effect. THz spectrum was obtained by applying the fast Fourier transforms (FFT) to each THz pulse waveform. The averaged reflected THz spectrum was calculated from the 10 spectra. The relative reflection spectrum of RDX can be calculated as R(ν)/R0(ν) , which is the ratio of the reflected THz power of RDX, R(ν), to that of Teflon, R0(ν), as plotted in Figure 10(a). The relative reflection spectra of polyethylene and flour are also presented for comparison. Because RDX is dispersive and absorptive in the THz band, its reflection spectrum differs from those of polyethylene and flour. After the K–K transform, the absorption spectra of RDX, polyethylene, and flour were acquired using Eqs. (4), (6), and (7), and are plotted in Figure 10(b). Figure 10(c) displays a direct comparison between the results of RDX from the transmission measurement and diffuse reflection measurement. The strong absorption peak at 0.82 THz and the three other relatively weak absorption peaks at 1.05, 1.35, and 1.55 THz can be well identified by the diffuse reflection measurement, and they agree closely with the transmission measurement. The transmission spectrum is in good accordance with the results previously reported.61, 62 There are some small spectral variations in the obtained absorption spectra of polyethylene and flour as shown in Figure 10(b). This finding differs from the result obtained from the transmission measurements of polyethylene and flour, which have no THz absorption features. We attribute the discrepancy to the surface difference between the sample (polyethylene or flour) and reference (Teflon). Because of the phase shift in the THz waves that occurs upon reflection, the mean-square electric fields present near an interface are dependent on the polarization of the incident radiation. This situation gives rise to a surface effect, often referred to as the surface selection rule, reflecting the information on the structural orientation of the sample surface.71 Changes in reflectivity due to these surface differences can result in a distortion in the band shape and a shift of the obtained absorption maximum compared with the result from a transmission measurement. However, predictions of band shapes are hindered by a complicated interplay of mean-square electric fields, angles of incidence, polarizations, and the reflectivities between different phases in a stratified medium. On the other hand, as we see from Figure 10(b), these distortions or spectral variations are relatively small compared with some intense absorption features in the RDX-absorption spectra.

H.-B. LIU AND X.-C. ZHANG

1.6

(a)

RDX Polyethylene Flour

1.4 1.2 1.0 0.8 0.6 0.4

0.6 Absorbance (a.u.)

Relative THz reflectance (a.u.)

272

0.5 0.4 0.3 0.2

0.2

0.1

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8

Frequency (THz)

Frequency (THz)

1.0

1.0 (c)

0.82 1.05

1.35 1.55

(d)

Transmission 0.6 0.4 0.2

Reflection

0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Frequency (THz)

Absorbance (a.u.)

0.8 Absorbance (a.u.)

RDX Polyethylene Flour

(b)

0.82 1.05

1.35 1.55

0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 Frequency (THz)

Figure 10. (a) Relative power reflection spectra of RDX, polyethylene, and flour, the diffuse spectrum of Teflon was used as reference; (b) Absorption spectra from the K–K transform of the reflection spectra; (c) The comparison between the absorption spectra from the transmission measurement (upper curve) and diffuse reflection measurement (bottom curve); (d) Absorption spectra of three different RDX samples. The dashed lines indicate the absorption peak positions. In (c) and (d), spectra are vertically shifted for clarity.

These results reveal that THz diffuse reflection spectroscopy can distinguish RDX from other materials that have no THz fingerprints, such as polyethylene and flour. In additional work, the measured phase was used to acquire the absorption spectrum using Eqs. (6) and (7) instead of using Eq. (4) to calculate the phase from the K–K relationship. However, because of the misplacement phase error in the diffuse reflection measurement, the absorption spectrum obtained (not shown in the paper) did not contain any RDX fingerprints. To test the reliability of the results shown in Figure 10(c), we measured three different RDX pellet samples that were prepared separately and

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

273

assumed to have small differences in surface morphology. The obtained absorption spectra are plotted in Figure 10(d). Although the three weak absorption peaks could not be repeated well because the surface effect described earlier, the strong one at 0.82 THz could be well identified in each sample. Detecting a spectral fingerprint at 0.82 THz of RDX is especially preferred on account of atmospheric attenuation. The atmospheric attenuation at 0.82 THz is relative low compared with that in the frequency above 1 THz, since 0.82 THz is away from the closest absorption lines of water vapor at 0.75 and 0.99 THz. The following results measured in the atmosphere (Figure 12) confirm this. In most of the reflection spectroscopic investigations, a metal surface or mirror is usually used as a reference. A Teflon pellet instead of a metal surface was used because of the similarity between the RDX and Teflon pellets, which were both compressed from powders. In order to compare the influences from the different references, we also chose the reflection spectrum of a metal surface as the reference to obtain the absorption spectrum of RDX. A copper plate was chosen in the investigation, and the measurement was also conducted in nitrogen. Figure 11 presents the comparison between the results of using Teflon and copper as references. Four absorption peaks still are observed in the case of using copper as reference, although there is some band shape change close to the absorption maximum at 0.82 THz. The band shape change can also be explained by the surface difference between the RDX pellet and copper plate. The RDX pellet surface is different from the copper plate surface, and more similar to the Teflon pellet surface. However, this band shape change does not affect the identification of the absorption fingerprints, as shown in Figure 11. Obtaining the same absorption fingerprints using different references is significant for practical applications. The result indicates that THz technology can be a feasible method for the real-world detection of RDX by using one or more surrounding materials (which have no THz features) as references. THz technologies combine two advantages for the detection of hidden explosives: explosives have THz fingerprints and many commonly used nonpolar dielectric materials have low absorption for THz waves. In order to demonstrate both capabilities, we investigated the THz diffuse reflection spectroscopy of RDX behind covering materials. The RDX samples were covered with four different barrier materials, paper (thickness: ~0.05 mm, white), polyethylene sheet (~0.1 mm, black), leather (~0.3 mm, yellow), and polyester cloth (~0.4 mm, green), which were all opaque for visible

H.-B. LIU AND X.-C. ZHANG

274

1.0 1.35 1.55

0.82 1.05 Absorbance (a.u.)

0.8 Copper as reference 0.6 0.4 0.2 0.0

Teflon as reference 0.4

0.6

0.8

1.0

1.2

1.4

1.6

Frequency (THz)

Figure 11. Absorption spectra of RDX from the K–K transform of the reflection spectra using Teflon and copper as the reference respectively. The dashed lines indicate the absorption peak positions.

light. The measurements were conducted in the atmosphere (with a relative humidity of ~20% at 25°C) instead of nitrogen and the whole path length of THz waves was ~110 cm. The absorption spectra obtained from the K–K transform are plotted in Figure 12. In the atmosphere, water vapor absorptions affect the measurements, especially in the range beyond 1 THz, since there are many water vapor absorption bands in the THz range, with absorption maximums at 0.56, 0.75, 0.99, 1.10, 1.11, 1.16, 1.21, 1.22, 1.41, 1.60, 1.66, and 1.71 THz.72,73 The strong water vapor absorptions reduce the SNR in these bands and lead to band distortions or artificial spikes in the absorption spectrum of RDX. Especially in the range close to 1.66 and 1.71 THz, the water vapor absorption almost attenuates the THz signal completely, resulting in the big artificial spikes shown in Figure 12(a). In addition, the barrier materials also lead to the distorted spectral band shapes and the shifts of absorption maximums. For the paper cover, because it is relatively thin and its surface is smooth, the obtained absorption spectrum is not distorted too much and in good agreement with the result without cover. For the other three covers, the rough surfaces and inhomogeneous thicknesses of the covers cause the distorted reflection spectra and therefore the distorted absorption spectra of RDX after the K–K transform. Because of both water vapor absorptions and covering effects, most of the weak absorption features of RDX can not be identified, and there are even artificial features. However, the absorption peak at about 0.82 THz is always observed behind all barriers in the atmosphere. This demonstrates the THz technique as an applicable tool for detecting hidden explosives, such as C-4 under clothing or inside package in diffuse reflection geometry.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

(a)

1.0

0.82 Water vapor absorption effect

0.5 No cover 0.4 0.3 0.2 0.1 0.0 0.4

Paper cover 0.6

0.8

1.0

1.2

1.4

Frequency (THz)

1.6

1.8

Absorbance (a.u.)

Absorbance (a.u.)

0.6

(b)

275

Polyethylene

0.8 0.6

Leather

0.4 0.82

0.2 0.0

0.4

0.6

Polyester

0.8 1.0 1.2 Frequency (THz)

1.4

Figure 12. (a) A comparison between the absorption spectra of RDX when RDX was bare and was covered with paper, both obtained from the diffuse reflection measurements; (b) Absorption spectra of RDX obtained from the diffuse reflection measurements under different covers. All the spectra were measured in the atmosphere with a relative humidity of ~20%.

3. THz Spectroscopy for Pharmaceutical Applications 3.1. IDENTIFICATION OF ANHYDROUS AND HYDRATED DRUGS

Identifying anhydrous and hydrated forms of pharmaceutical substances is of great importance in pharmaceutical science and industry. In pharmaceutical manufacturing, solid pharmaceutical materials may come in contact with water during processing, including crystallization, lyophilization, wet granulation, aqueous film-coating or spray-drying. In addition, they may be exposed to water during storage in the atmosphere or in a dosage form consisting of materials that contain water and are capable of transferring it to other ingredients.75 For some crystalline solids, hydrates are formed if water is the solvent of crystallization. The water molecules occupy certain positions in the crystalline lattice by forming hydrogen bonds or coordinate covalent bonds with the anhydrous drug molecules. Therefore the hydrated pharmaceutical materials have different crystalline forms from those of the anhydrates. Consequently, the physicochemical properties, such as density, solubility, melting point, thermal conductivity, and the physical and chemical stability of the hydrates may differ from those of the anhydrates. These physicochemical differences affect the bioavailability and product performance and could have impact at therapeutic, manufacturing, commercial, and legal levels.74,75 This effect is similar to polymorphism, which indicates the existence of more than one crystalline

276

H.-B. LIU AND X.-C. ZHANG

form of the same drug substance. The effect also happens to other solvent molecules beside water. The term pseudopolymorph has been applied to these hydrates or other solvates to distinguish them from polymorph.74 Various technologies have been used to identify anhydrous and hydrated pharmaceutical materials, such as XRPD, mid-infrared spectroscopy, Raman spectroscopy, thermal analytical methods, and solid-state NMR. However, each one has its own limitations. In this study, THz-TDS was tested as a tool for pseudopolymorph identification. The results from XRPD are used to support this investigation. THz-TDS has previously proved to be a quick, simple, and versatile technique to study the polymorphs of pharmaceutical materials.76–78

Figure 13. Molecular structures of (a) caffeine, (b) theophylline, (c) D-(+)-glucose, and (d) ampicillin.

Four drugs, including caffeine, theophylline, D-glucose, and ampicillin, were used as examples (Figure 13). Caffeine is a central nervous system stimulant, having the effect of warding off drowsiness and restoring alertness. Theophylline is used under a variety of brand names to prevent and treat wheezing, shortness of breath, and difficulty breathing caused by asthma, chronic bronchitis, emphysema, and other lung diseases. D-glucose is one of the most important carbohydrates which cells use as a source of energy and metabolic intermediate. It is a widely used substance in both food and pharmaceutical industries. Ampicillin is a penicillin-like antibiotic used to treat certain infections caused by bacteria such as pneumonia, bronchitis, and infections of ear, lung, skin, and urinary tract. The anhydrous β-phase caffeine, anhydrous theophylline (polymorphic form II), anhydrous D-(+)-glucose, D-(+)-glucose monohydrate, anhydrous ampicillin, ampicillin trihydrate (all with purity >99%) were purchased from Sigma-Aldrich and used without further purification. Caffeine hydrate

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

277

(each caffeine molecule has 4/5 water molecule79) and theophylline monohydrate were prepared by slow recrystallization from the water solution and stored in a sealed vessel at ~70% relative humidity in the presence of a NaCl saturated solution at room temperature. All the samples are polycrystalline materials. The sample powders were ground into fine particles (<50 µm) and then mixed with polyethylene powder (<30 µm) at different weight ratios. The caffeine and caffeine hydrate samples (both 160 mg) were pure; the theophylline and theophylline monohydrate samples both have a 33% w/w (50 mg/150 mg); the D-glucose and D-glucose monohydrate samples both have a 50% w/w (75 mg/150 mg); the ampicillin and ampicillin trihydrate samples both have a 33% w/w in polyethylene (50 mg/150 mg). The powder samples were compressed into pellets (~1–1.5 mm in thickness and 13 mm in diameter) with a hydraulic press under 4 tons pressure. The samples had enough thickness to avoid the etalon effect caused by multiple reflections of THz pulses. The measurements were conducted using the THz-TDS system (bandwidth: 0.1–3 THz; resolution: 0.1 THz; zero-filling factor: 4) in a transmission mode under a nitrogen purge (relative humidity ~0%). The XRPD spectra were obtained via a wide-angle x-ray powder diffractometer (Model XDS 2000, Scintag, USA). The samples were exposed to Cu Kα radiation (50 kV × 30 mA). The instrument was operated with a scan step size of 0.02 degree 2θ, and a scan rate of 4°min−1. All the measurements were conducted at room temperature (~25°C). Figure 14(a) shows the THz absorption spectra of anhydrous β-phase caffeine and caffeine hydrate at room temperature. The increasing baselines with frequency in the THz spectra result from the scattering of the samples (the same case for the subsequent results). The anhydrous caffeine has one major absorption peak at 1.26 THz, and a weak one at about 0.79 THz, while caffeine hydrate shows absorption peaks at 1.16, 1.49, and 2.26 THz. The spectral variations in the 2.5–3 THz range are not absorption peaks according to our repeated measurements, which result from the lower SNR and the higher attenuation of the sample in this band. The distinct THz absorption features in anhydrous and hydrated caffeine possibly reflect their different intermolecular vibrational modes or phonon modes. The theoretical calculation of these intermolecular vibrational modes or phonon modes currently remains as a difficult problem due to the complexity involving periodic density functional theory modeling. It will be a topic for our future investigation. The caffeine molecule possesses a hydrophilic center at the imidazole nitrogen atom, susceptible to hydrogen-bonding and the hydrate water molecule effloresces via a molecular escape tunnel

H.-B. LIU AND X.-C. ZHANG

278

through the crystallographic a-face of the crystalline lattice.80 Incorporating water molecules results in the formation of a new, distinct crystalline form with intermolecular vibration modes that are different from those of anhydrous caffeine. The XRPD patterns of anhydrous caffeine and hydrated caffeine, as illustrated in Figure 14(b), indicate the crystallinity difference between anhydrous and hydrated caffeine. The XRPD data are in good agreement with those previously reported.81 On the other hand, according to the literature, the mid-infrared and Raman spectra of anhydrous and monohydrated caffeine are similar.82 -1

(a)

3500 (b)

2.26

160

Intensity (CPS)

Absorption coefficient (cm )

200

120 Caffeine hydrate -1 (+ 40 cm ) 1.16 1.49 80 1.26

40 0

Caffeine

3000 Caffeine hydrate (+ 2000 CPS)

2500 2000 1500 1000 500

0.79

Caffeine

0

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

2.8

8

12 16 20 24 28 32 36 40 44 2θ (degree)

100 80 60

(a)

2.67

Theophylline monohydrate -1 (+ 40 cm ) 0.83

40

1.60 Theophylline

0.96

20 0

1.64

6000 Intensity (CPS)

-1

Absorption coefficient (cm )

Figure 14. (a) THz absorption spectra and (b) XRPD patterns of anhydrous caffeine and caffeine hydrate (both are pure samples). (b)

Theophylline monohydrate (+ 3000 CPS)

5000 4000 3000 2000

Theophylline

1000

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

2.8

0 8

12 16 20 24 28 32 36 40 44 2θ (degree)

Figure 15. (a) THz absorption spectra and (b) XRPD patterns of anhydrous theophylline and theophylline monohydrate (both have a 33% w/w in polyethylene).

The THz absorption spectra of anhydrous theophylline and theophylline monohydrate at room temperature also show evident differences, as plotted in Figure 15(a). Anhydrous theophylline exhibits prominent absorption peaks at 0.96 and 1.60 THz, while theophylline monohydrate has major

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

279

absorption peaks at 1.64 and 2.67 THz, and a small peak at 0.83 THz. Again, their different THz absorption features have been assigned to their different intermolecular vibrational modes. In the theophylline monohydrate, water molecules form infinite hydrogen-bonded chains that run through tunnels formed by surrounding theophylline molecules. The water chains are also cross-linked through hydrogen bonds by hydrogen-bonded theophylline dimers, and form a 2D hydrogen-bonded structure.83 Therefore hydration results in a different crystalline form from that of anhydrous theophylline. The XRPD spectra, as shown in Figure 15(b), indicate the evident crystallinity difference between anhydrous and monohydrated theophylline. The XRPD data are consistent with the literature.84,85 The mid-infrared spectra of anhydrous and monohydrated theophylline, as previously reported, do not exhibit remarkable differences.84,86 The THz absorption spectra of anhydrous D-glucose and D-glucose monohydrate at room temperature are illustrated in Figure 16(a). The THz absorption spectrum of anhydrous D-glucose is consistent with what previously reported.87,88 It shows distinct absorption features from that of Dglucose monohydrate. Anhydrous D-glucose has a major absorption peak at 1.44 THz, a broad absorption peak at ~2.6 THz, and two other weak absorption peaks at 1.29 and 2.10 THz. D-glucose monohydrate has remarkable absorption peaks at 1.82, 1.98, and 2.46 THz. The XRPD patterns illustrated in Figure 16(b) imply the crystallinity difference between the anhydrous and monohydrated D-glucose. Again, differences in the crystalline structure associated with the hydrogen-bonding and the incorporation of water molecules are probably responsible for the distinct absorption features of anhydrous and hydrated D-glucose.

(a)

1.82 1.98

2000

2.46

D-glucose monohydrate -1 80 (+ 60cm )

~2.6 1.44

40 D-glucose 0

0.4

0.8

2.10

1.29

1.2 1.6 2.0 2.4 Frequency (THz)

(b)

D-glucose monohydrate (+ 1000 CPS)

1600

120

2.8

Intensity (CPS)

-1

Absorption coefficient (cm )

160

1200 800 D-glucose

400 0 8

12 16 20 24 28 32 36 40 44 2θ (degree)

Figure 16. (a) THz absorption spectra and (b) XRPD patterns of anhydrous D-glucose and D-glucose monohydrate (both have a 50% w/w in polyethylene).

H.-B. LIU AND X.-C. ZHANG

-1

Absorption coefficient (cm )

80

(a)

1.98

60

Ampicillin trihydrate -1 1.67 (+ 30 cm ) 1.27 40 1.61

2.31 2.78 2.08

20 0

Ampicillin 0.70 0.4

0.8

2000

2.61

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

Intensity (CPS)

280

(b)

Ampicillin trihydrate (+ 1000 CPS)

1600 1200 800 Ampicillin

400 0 8

12 16 20 24 28 32 36 40 44 2θ (degree)

Figure 17. (a) THz absorption spectra and (b) XRPD patterns of anhydrous ampicillin and ampicillin trihydrate (both have a 33% w/w in polyethylene).

The THz absorption spectra of anhydrous ampicillin and ampicillin trihydrate at room temperature are shown in Figure 17(a). Anhydrous ampicillin has absorption peaks at 0.70, 1.61, 2.08, and 2.78 THz. Ampicillin trihydrate shows absorption peaks at 1.27, 1.67, 1.98, 2.31, and 2.61 THz. Figure 17(b) shows the XRPD spectra of anhydrous ampicillin and ampicillin trihydrate, indicating their distinct crystalline structures. The XRPD data are in good agreement with the literature.89 We have also attributed their distinct THz absorption features to the different hydrogen-bonding patterns and crystalline structure of anhydrous and hydrated ampicillin. In an additional investigation, using a Fourier transform far-infrared spectrometer, the absorption spectra of these anhydrous and hydrated samples in the 3–20 THz range were measured (data are not shown here). The absorption spectra of anhydrous samples are nearly identical to those of hydrated ones. Most of the absorption features in the range of 3–20 THz reflects the intramolecular vibration modes, especially the normal vibrational modes of the whole molecule, such as in-plane/out-plane bending and torsion of benzene ring.65 This implies that the hydrate water molecules do not affect the intramolecular vibrations of the anhydrous molecules in the 3–20 THz range, possibly due to the relatively weak hydrogen-bonding between them. However, the hydrate water molecules participate in the intermolecular vibrations and exhibit different intermolecular vibrational modes from those of anhydrous compounds. These results demonstrate that THz-TDS is a new advantageous technique for the pseudopolymorph identification and study, and has great potential to become a PAT in pharmaceutical production and quality control.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

281

3.2. SOLID-STATE REACTIONS OF DRUG SUBSTANCES

Understanding solid-state reactions and developing a theoretical basis for solid-state reactions is of great significance in chemical and biochemical fields, especially in pharmaceutical science and industry. Solid-state reactions include solid-state phase transformations (polymorphic transformation), reactions in which the solvent of crystallization is lost or gained, and various chemical reactions. Solid-state chemical reactions cover a broad range, such as solid-state oxidations, additions of gases to solids, solid-sate decompositions, solid-state photochemical reactions, solid-state thermal reactions.90 There is a need to understand solid-state reactions, especially for the solid-state reactions of pharmaceutical materials in terms of the molecular details of the reactions. Of particular interest is to determine the parameters of the molecular or crystal structure that can lead to retardation of the solid-sate reactions of drugs and thus render drugs more stable. In addition, solid-state reactions play an important role in the drug formulation and delivery, quality control, regulatory issues, and drug bioavailability. Up to now, many techniques have been used to characterize solid-state reactions, including XRPD, infrared spectroscopy, Raman spectroscopy, thermal analytical methods, and solid-state NMR. However, each of these techniques has its own disadvantages. Usually, two or more above techniques are combined to investigate a solid-state reaction process. XRPD is the most common method to study and monitor solid-state reactions based on distinguishing solid phases having different crystalline structures. But the use of x-rays, which are ionizing radiations, introduces more concern on the issue of personal safety and nondestructive measurements. And it requires rotating either the sample or x-ray source to obtain a diffraction pattern. Vibrational spectroscopy techniques have advantages over XRPD in that, ideally, samples can be monitored on line and the spectroscopic imaging of samples can be obtained in real time if desired. In the mid-infrared range, infrared spectroscopy, and Raman spectroscopy based upon the intramolecular vibrations of molecules are sensitive methods for the characterization and identification of different solid forms. Unfortunately, it is not always clear whether the spectral features are due to the differences of crystalline packing or molecular structure. Also, in some cases the infrared spectra of reactants and products are very similar. This is due to the fact that either there are only minor differences in crystal packing and conformation between the two crystal forms or perhaps that the many stretching and rotational vibrations in the molecule override the vibrational differences from different crystal packing.91

282

H.-B. LIU AND X.-C. ZHANG

Thermal analysis methods, such as differential scanning calorimetry and thermal gravimetric analysis, which involve heating the sample and measuring the change in some physical property (such as weight), are also important analytical tools for characterizing pharmaceutical solids. However, due to parallel processes, thermal analysis for mixed and complex systems in solid-state reaction is difficult.92 Low-frequency lattice vibrations, or phonon modes (10–150 cm−1, or 0.3– 4.5 THz), which correspond to librations and translations of the entire molecule, are accessible to Raman spectroscopy. Because the different crystal forms of a compound yield different phonon vibrational spectra, Raman spectroscopy can be used to characterize and identify the solid phases of substances. However, the measured relative intensities do not provide quantitative information on the concentrations of the various species in solid-state reactions.93 And the laser irradiation in Raman spectroscopy can induce a phase change, or initiate photochemical reactions in the samples. Solid-state NMR is useful to examine conformational and dynamic changes which occur in solid drugs during solid-state transformations and reactions. But establishing the experimental conditions and correctly assigning the signals in the spectrum can be a complex and time-consuming process.93 THz radiation in the range of 0.1–10 THz can induce low-frequency crystalline phonon vibrations, hydrogen-bonding stretches, and other normal vibrational modes of whole molecules. The transmitted or reflected THz spectrum of a sample will contain THz features reflecting these vibrational modes. These vibrational or stretching modes are related to certain molecular/crystal structures, which will change during the process of solidstate reactions. The quantitative analysis can be performed based on the Beer–Lambert law, a mathematical means of expressing how electromagnetic waves are absorbed by matter. Eq. (8) is for transmission mode and Eq. (9) is for reflection mode, respectively.

Aν = log(1 / Tν ) = c • εν • l ,

(8)

Aν = log(1 / Rν ) ∝ c • εν ,

(9)

where Aν is absorbance for frequency ν; Tν is transmittance; Rν is reflectance; c is molar concentration of the ingredient; εν is molar extinction (absorption) coefficient of the ingredient; l is the pathlength of the THz waves through the sample. Therefore, when a solid-state reaction is monitored or investigated by THz spectroscopy, the changes in the transmission or reflection THz

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

283

spectrum can be observed quantitatively upon changing the environmental conditions, such as temperature, atmosphere, or light illumination. A THz spectra database of initiate reactants, middle reactants, and final reactants for this solid-state reaction needs to be established previously. This database contains the THz spectra of each separate reactant, intermediate, and product, with their respective THz fingerprints indicated. In the data analysis, the THz spectra of reactants measured during the solid-state reaction are compared with those in the spectra database for the identification and quantitative analysis. This can be used to study or monitor the processes of solid-state reactions and obtain the reaction kinetics. Multivariate calibration methods, such as multiple linear regression (MLR), principal component regression (PCR), partial least squares (PLS), and neural networks (NN) can be used for a better quantitative analysis in THz spectroscopy. Multivariate calibration methods have advantage over classical univariate analysis because they are able to analyze multiple components simultaneously in the absence of distinct peaks attributed to single components. They also have improved their ability to detect interferents, and to quantify complex systems with unknown constituents.94 THz radiation will not cause any harmful photoionization for the sample and it is safe for human body. The coherent and time-gated detection method in THz-TDS is immune to the background THz noise such as those from the heating process.The THz-spectroscopic images (2D or even 3D) of samples can also be obtained in addition to the spectroscopic information. Furthermore, THz radiation can penetrate through many nonpolar dielectric materials and investigate targets behind covers or in containers. 3.2.1. Dehydration kinetics of D-glucose monohydrate The polycrystalline hydrates are important chemical or pharmaceutical compounds. In particular, pharmaceutical hydrates have been recognized as playing a significant role in drug formulation. The presence of the water molecules affects the intermolecular interactions and the crystalline disorder, thereby influencing the free energy, thermodynamic activity, solubility, dissolution rate, stability, and bioavailability of the pharmaceutical materials. Hence, studying the kinetics of both dehydration and hydration is exceedingly important. The anhydrous D-(+)-glucose and D-(+)-glucose monohydrate (both with purity >99.5 %) were purchased from Sigma-Aldrich and used without further purification. The samples were ground into fine particles (<50 µm) and then were compressed into pellets with polyethylene powder (<30 µm)

284

H.-B. LIU AND X.-C. ZHANG

at a 60% w/w (75 mg/125 mg). This time, the sample powder was not mixed with the polyethylene powder. Instead, the sample powder was placed in one side and polyethylene powder was placed in the other, then the sample was compressed into a pellet (~1–1.5 mm in thickness and 13 mm in diameter) with a hydraulic press using 4 tons pressure. This sample preparation ensures that the solid-state reaction happens in pure D-glucose monohydrate and the sample has an enough thickness to avoid multiple THz reflections (the etalon effect). The D-glucose monohydrate samples were placed in a heating chamber, wherein the temperature was controlled with an accuracy of ±0.5°C using a temperature controller (4,000 Series, Specac, UK). The measurements were conducted under a nitrogen purge (with a relative humidity of ~0%) to avoid the water vapor absorption of THz radiation. The obtained THz waveforms have a time delay of 10 picoseconds, corresponding to a spectral resolution of 100 GHz. After four times of zero-padding, FFT were applied on the THz waveforms and the THz spectra in the range of 0.1–3 THz were obtained. The achieved dynamic range of the THz power spectrum was about 60 dB. During the dehydration, each THz waveform acquisition took approximately 1 min. Figure 16(a) shows the THz absorption spectra of anhydrous D-glucose and D-glucose monohydrate at room temperature. The two absorption spectra have evident differences, which have been assumed to the different intermolecular vibrational modes of the anhydrous and hydrated D-glucose. Figure 18 plots the THz absorption spectra (in the 1.1–2.2 THz range for a clear view) of D-glucose monohydrate heated at 45°C for about 27 min, with the baselines corrected. When the D-glucose monohydrate was heated, the positions of absorption peaks shifted a little toward lower frequency compared with those at room temperature, as indicated in Figure 16(a) and 18. The shifts of different absorption peaks were not equal because of the irreproducibility of the spectra. These shifts are due to the temperature dependence of the vibrational modes resulting from the anharmonicity of the vibrational potentials (the same for the subsequent results).95,96 When the D-glucose monohydrate was heated, it began to lose its water and the anhydrous D-glucose structure formed. Therefore the absorption peaks at 1.80 and 1.96 THz (of D-glucose monohydrate) decreased due to the reduction of D-glucose monohydrate; and the absorption peaks at 1.28, 1.43, and 2.08 THz (of anhydrous D-glucose) increased due to the augment of anhydrous D-glucose. The increased absorption peaks reached to the maximum after a certain time, indicating the completion of dehydration. Since there is an overlap band between the

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

285

o

Heated at 45 C

30

-1

1.43

Absorption (cm )

absorption peaks at 2.08 and 1.96 THz, a quantitative analysis based on them is difficult. Therefore the absorption peaks centered at 1.28 and 1.43 were selected for quantitative analysis. The peak area (spectral area between 1.19 and 1.56 THz) was used to evaluate the quantity of anhydrous D-glucose in the dehydration process. The normalized dehydration fraction, or the normalized mole fraction of anhydrous D-glucose produced in the dehydration, α was obtained from the normalized absorption peak area between 1.19 and 1.56 THz. α was unity for the spectrum obtained when the dehydration totally completed (after being heated for ~27 min at 45°C).

20 2.08 1.28

1.4 1.6 1.8 2.0 Frequency (THz)

2.2

4.8 3.0

e (m

T im

12.4 8.5

in )

0 26.8 16.5

1.80 1.96 1.2

10

Figure 18. THz absorption spectra of D-glucose monohydrate heated at 45°C for about 27 min (the time between any two neighboring spectra is not equal). The absorption peaks at 1.80 and 1.96 THz deceased and the absorption peaks at 1.28, 1.43, and 2.08 THz increased over time due to the reduction of D-glucose monohydrate and increase of anhydrous D-glucose during the dehydration.

Figure 19(a) plots the normalized fraction of anhydrous D-glucose, α as a function of time during the dehydration at five different temperatures. The dehydration kinetics was investigated based upon various kinetics equations which have found the best application in solid-state reactions. Several kinetics equations or models have been attempted, including random nucleation and growth-controlled reactions (the Avrami–Erofe’ev equation), phase boundary-controlled reactions, diffusion-controlled reactions, power– law equations, and equations based on the order of reactions.97,98 In terms of the highest values of correlation coefficients, r, from the data fitting, the best

H.-B. LIU AND X.-C. ZHANG

286

fit was found with the 2D phase boundary-controlled reaction, known as the contracting area equation 97, 98

1 − (1 − α )1 / 2 = kt ,

(10)

where k is dehydration rate constant and t is time. Eq. (10) states a mechanism for the solid-state reactions that are controlled by the advancement of a 2D phase boundary from the outside of a crystal, inward at a constant velocity, i.e. a reaction proceeding from the surface of a circular disk, cylinder or rectangle, inward. Our result is consistent with what was obtained using other technologies. It was previously reported that the dehydration of carbamazepine dihydrate was 2D phase boundary controlled at water vapor pressures ≤5.1 torr while the Avrami-Erofeev kinetics (3D random nucleation) was followed at water vapor pressures ≤12 torr.99 In this study the dehydration was conducted in a nitrogen environment with a relative humidity of about 0% (water vapor pressures ≤5.1 torr), therefore following the 2D phase boundary-controlled reaction. Figure 19(b) shows the dehydration kinetics of D-glucose monohydrate evaluated according to the contracting area equation. The linear fits were plotted and the achieved correlation coefficients, r, were shown in Table 2. The high correlation coefficients achieved in our study imply that THz-TDS is an accurate tool to probe the dehydration kinetics. From the slopes of the linearly fitted lines lines in Figure 19(b), the dehydration rate constants k at different temperatures were derived, as shown in Table 2. 1.0 0.8

0.8

(b)

0.6

o

40 C o 42 C o 45 C o 48 C o 51 C

0.4 0.2 0.0

1/2

(a) 1-(1-α)

Fraction of production α

1.0

0.6 0.4 0.2 0.0

0

10

20

30 40 50 Time (min)

60

70

o

40 C o 42 C o 45 C o 48 C o 51 C

80

0

10

20

30 40 50 Time (min)

60

70

80

Figure 19. (a) α vs. time at five temperatures. α is obtained from the normalized absorption peak area between 1.19 and 1.56 THz; (b) Dehydration of D-glucose monohydrate evaluated according to the contracting area equation of the solid-state reaction. The scatter symbols are the experimental data and the solid lines are the linear fits. The experimental data exhibit good fits for the contracting area equation.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

287

TABLE 2. The correlation coefficients and the dehydration rate constants Temperature (°C)

Correlation coefficients r

40

0.9988

Dehydration rate constants k (min−1) 0.0129

42 45 48 51

0.9996 0.9991 0.9977 0.9963

0.0181 0.0374 0.0563 0.0889

The dehydration rates obtained in the temperature range of 40–51°C were used to generate the Arrhenius plot, which plots ln k vs. 1,000/T, as shown in Figure 20. The Arrhenius reaction rate equation is given by k (T ) = A exp( − E A / RT )

(11)

where A is an unknown constant, EA is the activation energy (kJ mol−1), R is the ideal gas constant (8.314 × 10−3 kJ mol−1 K−1), and T is the absolute temperature (in degrees Kelvin). From the linear fit (with a correlation coefficient of 0.9954), the calculated activation energy of the dehydration is about 149 kJ mol−1.

-2.0

-1

ln k (min )

-2.5 -3.0 -3.5 -4.0 -4.5 -5.0

3.08 3.10 3.12 3.14 3.16 3.18 3.20 1000/T (1/K)

Figure 20. The Arrehenius plot for the dehydration of D-glucose monohydrate. The solid circles are the experimental data and the solid line is the linear fit. The correlation coefficient for the linear fit is 0.9954.

The activation energy is the threshold energy for a chemical reaction to occur. The activation energy for the rupture of the hydrogen bonds is

H.-B. LIU AND X.-C. ZHANG

288

usually at the level of 10 kJ mol−1, which is much smaller than that of the dehydration of D-glucose hydrate. Therefore, the dehydration underwent a more complicated process than the rupture of the hydrogen-bonding between the D-glucose and water molecules. It includes breaking hydrate’s crystalline structure and forming new anhydrous crystalline structure. In our study, only the absorption peaks centered at 1.28 and 1.43 were selected for quantitative analysis. A more extensive analysis including other absorption peaks such as the ones at 1.80 and 1.96 THz can be used to characterize the reduction of D-glucose monohydrate. However, the overlap band between the absorption peaks centered at 2.08 and 1.96 THz makes it difficult to do quantitative analysis. Multivariate calibration methods are able to analyze multiple components (multiple absorption peaks) simultaneously in the absence of distinct peaks attributed to single components. They have advantage over classical univariate analysis. A multivariate analysis of the THz absorption spectra supposes to be more effective to characterize the dehydration process of D-glucose monohydrate. This work is ongoing. 3.2.2. Solid-state reaction of aminophylline monohydrate Figure 21 shows a complex pharmaceutical solid-state reaction upon heating: aminophylline monohydrate (I) decomposes to anhydrous theophylline (III) either directly or through an intermediate, anhydrous aminophylline or α-aminophylline (II). It is difficult to use thermal analytical techniques or mid-infrared spectroscopy to study this kind of complex solid-state reaction. XRPD was successfully utilized to study the kinetics of this solid-state reaction previously.100 In our investigation THz-TDS was demonstrated as a new method to study this reaction. Aminophylline monohydrate (I)

α-Aminophylline (II)

Anhydrous theophylline (III)

Figure 21. Solid-state decomposition of aminophylline monohydrate.

Figure 22(a) and (b) show the molecular structures of α-aminophylline and theophylline. Aminophylline is less potent and shorter-acting than theophylline. Its most common use is in bronchial asthma. The THz absorption spectra of aminophylline monohydrate, anhydrous theophylline, and

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

289

anhydrous aminophylline are plotted in Figure 22(c). Aminophylline monohydrate has an absorption peak at 1.25 THz. α-aminophylline has several absorption peaks, but only the one at ~2.3 THz shows up during the heating reaction of aminophylline monohydrate. Theophylline has two evident absorption peaks at 0.96 and 1.61 THz.

Figure 22. Molecular structures of (a) α-aminophylline and (b) theophylline; (c) The THz absorption spectra of aminophylline monohydrate, α-aminophylline, and theophylline (all with ~60% w/w in polyethylene).

Figure 23 illustrates the THz absorption spectra (in the 0.6–2.5 THz range for clarity) of aminophylline monohydrate heated at 75°C, with baselines corrected. When the aminophylline monohydrate sample was heated at 75°C, the positions of absorption peaks shifted a little toward lower frequency compared with those at room temperature, as indicated in Figure 22 and 23. Again, the shifts of different absorption peaks were not equal due to the limited resolution of THz-TDS (0.1 THz) and the irreproducibility of the spectra. These shifts result from the temperature dependence of the vibrational modes resulting from the anharmonicity of the vibrational potentials.95, 96 The absorption peaks at 0.94 and 1.64 THz

H.-B. LIU AND X.-C. ZHANG

290

are the fingerprints of theophylline and characterize the quantity of theophylline in the reaction. The peak at 0.94 THz enhanced due to the increase of anhydrous theophylline. The one at 1.64 increased, presumably indicating the augment of theophylline, however, with a peak position shifted to 1.55 THz. This possibly reflects a slightly different polycrystalline structure from that of theophylline. The absorption peak at 1.25 THz is the distinct fingerprint of aminophylline monohydrate and characterizes its quantity. It decreased during the reaction. The absorption peak at ~2.3 THz first increased then decreased during heating, which could indicate the quantity of the intermediate product, α-aminophylline.

30 20

I III 1.25 0.94

10

18

1.55 III 1.2 1.6 2.0 Frequency (THz)

e (m

12

in)

0 6

0.8

-1

40

II ~ 2.3

T im

III 1.64

Absorption (cm )

50

o

Heated at 75 C

24 2.4

Figure 23. THz absorption spectra of aminophylline monohydrate heated at 75°C for ~25 min (the time period between any two neighboring spectra is not equal).

The absorption peaks centered at 0.94, 1.26, and 2.3 THz were selected for quantitative analysis. The peak area of at 0.94 THz was used to evaluate the quantity of theophylline in the decomposition process. The peak area of at 1.25 THz was used to quantify the aminophylline monohydrate during the decomposition. And the peak area of at ~2.3 THz was used to quantify the α-aminophylline during the decomposition. The mole fractions of aminophylline monohydrate (αI), α-aminophylline (αII) and theophylline (αIII) during the reaction were obtained from the normalized absorption

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

291

peak area, respectively. The boundary conditions are: αI = 1 for t = 0 and α I = 0 for t = ∞; αII = 1 for t ~11 min (before that, the peak at ~2.3 was saturated hence no data were used), and α II = 0 for t = ∞; αIII = 0 for t = 0, and α III = 1 for t = ~24 min. The decomposition kinetics was investigated based upon various kinetics equations which have found the best application in solid-state reactions. Several kinetics equations or models have been attempted, including random nucleation and growth-controlled reactions (the Avrami-Erofe’ev equation), phase boundary-controlled reactions, diffusion-controlled reactions, power–law equations, and equations based on the order of reactions.97, 98 The experimental data exhibit the best fits for first order reactions: ln α = ±k t,

(12)

where ± indicates that the α is increasing (+) or decreasing (−) during the reaction. Figure 24 contains plots of the reaction of aminophylline monohydrate evaluated according to the first-order process of solid-state reaction. It is in good accordance with what reported in the literature.100 The experimental data exhibit good fits. High correlation coefficients were achieved and the reaction rate constants k were extracted from the linear fits, as shown in Figure 24. The study on the decomposition kinetics of aminophylline monohydrate at other temperatures is ongoing. THz-TDS offers a number of advantages over existing techniques for characterizing the kinetics of solid-state reactions. The high SNR of THzTDS makes it a capable spectroscopic technique for the quantitative analysis. The coherent and time-gated detection method in THz-TDS is immune to the background THz noise from the heating process. With THzTDS the amplitude and phase of each spectral component of the THz pulse can be determined. Therefore, a quantitative analysis of the absorptions can be readily conducted based upon reflection measurements (future work), which are more desirable in practical applications than transmission measurements. Moreover, THz waves are completely noninvasive/nondestructive for the study of solid-state reactions due to their low photon energies. Very importantly, THz-TDS extends electromagnetic spectral measurements to the THz range, in which spectroscopic studies of many solid-sate reactions were either nonexistent or scarce. THz-TDS is promising to provide information unavailable through other conventional methods like x-ray powder diffraction, mid-infrared spectroscopy, and Raman spectroscopy.

292

H.-B. LIU AND X.-C. ZHANG

(a) Aminophylline

0.0

monohydrate (I) peak at 1.26 THz -1 k = 0.0411 min r = 0.9964

ln ( αΙ )

-0.2 -0.4 -0.6 -0.8 -1.0 0

4

8

20

24

(b) α-Aminophylline (II)

0.0

peak at ~ 2.3 THz -1 k = 0.1588 min r = 0.9988

-0.5 ln ( αΙΙ )

12 16 Time (min)

-1.0 -1.5 -2.0 -2.5

ln ( αΙΙΙ )

12

16 20 Time (min)

0.0

(c) Theophylline (III)

-0.2

peat at 0.94 THz -1 k = 0.0518 min r = 0.9999

-0.4

24

-0.6 -0.8 -1.0 -1.2 0

4

8

12 16 Time (min)

20

24

Figure 24. The decomposition of aminophylline monohydrate evaluated according to the first-order process of solid-state reaction: lnα = kt, for (a) aminophylline monohydrate, (b) α-aminophylline, and (c) theophylline. The scatter symbols are the experimental data and the solid lines are the linear fits.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

293

4. THz Spectroscopy for Biological Applications 4.1. THz SPECTROSCOPY OF SMALL BIOMOLECULES

Sensing some small biocompounds such as amino acids, purines, purine derivatives is of significance for biomedical or pharmaceutical research and industry. Amino acids are the basic structural building units of proteins. They form short polymer chains called peptides or polypeptides which in turn form structures called proteins. As of today, over 100 amino acids have been found in nature. Among them 20 amino acids are encoded by the standard genetic code and are called proteinogenic or standard amino acids. Purines are biochemically significant as components of DNA and RNA, and are also found in a number of other important biomolecules, such as ATP, GTP, and coenzyme A. Some other purine derivatives, like caffeine and theophylline, are important biomolecules in daily life, as well as biomedical and pharmaceutical fields. In this study, a list of biocompounds including amino acids, purines, purine derivatives, and other small biomolecules have been investigated using THz-TDS in the range of 0.1–3 THz. All the samples were purchased from Sigma-Aldrich without further purification (purity >99%). Some samples were used in pure form and some others were prepared by mixing with polyethylene powder at different weight ratios, depending on their different THz attenuations. All the sample powders were gently ground using a pestle and mortar to reduce the particle size (<50 µm) which therefore minimize the THz wave scattering. The powder samples were compressed into pellets (~1 mm in thickness and 13 mm diameter) using 4 tons pressure with a hydraulic press. The measurements were conducted using a THz-TDS system (bandwidth: 0.1–3 THz; resolution: 0.1 THz; zero-filling factor: 4) in a transmission mode under a nitrogen purge. For the pure samples, the refractive indices, n, in addition to absorption coefficients, α, were obtained. Figure 25 shows the absorption coefficients and refractive indices of eight selected amino acids, which were all measured using pure samples. All of them exhibit distinct THz absorption features, which can be used as fingerprints for spectroscopic sensing. Many peaks have absorption coefficients beyond 50 cm−1 (after the subtraction of the baselines), which is advantageous for the sensitive THz identification. The baselines increase over frequency in the absorption spectra result from the scattering of the powder pellet samples. The dispersions exhibit in refractive index spectra confirm the absorption features, since the absorption spectrum is the first derivative curve of the dispersion spectrum according to the K–K relationship. The strong dispersions will also be significant for the identifications of these biocompounds in reflection geometry, which is more

H.-B. LIU AND X.-C. ZHANG

-1

α (cm )

n

-1

α (cm )

2.0

1.9

50

1.8

0.8

1.2 1.6 2.0 Frequency (THz)

1.9

60

1.8

40

1.7

20

1.6

1.7 2.8

2.4

n

80

0

0.4

2.0

α

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

160

160

2.1

1.9

40

1.8

0

1.7 2.0

n

80

80

1.8

40

1.7

0

0.8 1.2 1.6 Frequency (THz)

1.9

n

-1

n

α (cm )

2.0

-1

α (cm )

α

120

α

0.4

2.0 L-methionine

L-lysine 120

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

180

120

2.1

-1

n

-1

α (cm )

2.0

60

α (cm )

α

n

1.9

30

1.8

0

1.7

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

2.3

α

n

120

2.2

90

2.1

60

2.0

30

1.9

0

0.4

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

180

120

-1

n

80

150

2.1 2.0

60

1.8

30

1.7

40

1.8

20

1.7

0

0

1.6 2.8

1.2 1.6 2.0 Frequency (THz)

2.4

2.0 1.9

1.9

0.8

2.1

α

n

90

60

0.4

2.2 L-valine

n

-1

α (cm )

α

2.2

1.8

n

L-tryptophan

100

α (cm )

120

1.6

2.4 L-threonine

150

L-phenylalanine 90

1.5

n

α

n

100

0

L-leucine

100

n

L-isoleucine 150

2.1

120

2.1

200

n

294

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

Figure 25. Absorption coefficients and refractive indices of selected amino acids.

1.6

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

2.0 Glutamic acid (amino acid)

2.0

Absorbance (decadic)

Absorbance (decadic)

2.4

1.6 1.2 0.8 0.4 0.0

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

Absorbance (decadic)

Absorbance (decadic)

1.0 0.8 0.6 0.4 0.2 0.8

1.2 1.6 2.0 2.4 Frequency (THz)

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

Hypoxanthine (purine)

2.5 2.0 1.5 1.0 0.5 0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

1.2 1.6 2.0 2.4 Frequency (THz)

2.8

1.6 Allopurinol (purine)

1.2

Absorbance (decadic)

Absorbance (decadic)

0.4

2.8

0.8

0.4

0.8

1.2 1.6 2.0 2.4 Frequency (THz)

Thymine (purine)

1.2

0.8

0.4

0.0 0.4

2.8

0.4

0.8

2.4

2.0 Barbituric acid

Absorbance (decadic)

Absorbance (decadic)

0.8

0.0

0.4

1.6

1.6 1.2 0.8 0.4 0.0

1.2

3.0

Xanthine (purine)

1.2

0.0

Hydroxymethyl-amino acid 1.6

0.0

2.8

1.4

0.0

295

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

2.8

N-acetyl-aspartic acid (amino acid)

2.0 1.6 1.2 0.8 0.4 0.0

0.4

0.8

1.2 1.6 2.0 Frequency (THz)

2.4

Figure 26. Absorption spectra of some amino acids, purines, and other biomolecules.

2.8

296

H.-B. LIU AND X.-C. ZHANG

feasible in practical applications than in a transmission mode. Figure 26 shows the absorption spectra of some other selected amino acids, purines, and biomolecules, which were measured using mixtures with polyethylene powder. Most of the biocompunds also exhibit THz absorption peaks. Caffeine, theophylline, and theobromine are members of one group, which are called methylated xanthines, xanthine derivatives, or xanthines. These purine derivatives are important small biomolecules and pharmaceutical substances as well (detailed study about caffeine and theophylline is also described in the previous section). They have very similar molecular structures, as shown in Figure 27, however exhibit remarkably different THz absorption spectra, mostly due to their different intermolecular vibrational modes.

Figure 27. (a) Molecular structures of caffeine, theophylline, and theobromine; (b) THz absorption spectra of caffeine (pure), theophylline (33% w/w), and theobromine (33% w/w), with vertical offsets for clarity.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

297

Most of the selected small biocompounds have distinct THz absorption features except N-acetyl-aspartic acid, which indicates that THz-TDS is an effective technique for sensing these biocompounds in corresponding biochemical research and industry. For example, investigating the amount of free amino acids in dry seeds will be very helpful on nutritious study in agricultural fields. THz waves are able to penetrate through dry seeds easily therefore sense their inside materials nondestructively.101 The THz spectroscopic imaging (2D or even 3D) of seeds, especially “golden seeds” (e.g. some precious seeds after complicated genetic modifications) will be of significant interest for agricultural research and industry. THz-TDS can also be used in products screening and inspection in pharmaceutical manufacturing, since many of these biocompounds are pharmaceutical-related substances. 4.2. THz SPECTROSCOPY OF PROTEINS

4.2.1. THz spectroscopy of solid-state proteins Protein samples including egg white lysozyme, hemoglobin, α-chymotrypsin and bovine serum albumin (all with purity >98%) were purchased from Sigma-Aldrich without further purification. The samples were prepared by casting films (with weights of 0.5–1 mg and thicknesses of 50– 100 µm) from water solutions on polyethylene film substrates (Thermo Electron, Co.), which is almost free of THz absorption. In the range of 2–20 THz, the absorption spectra were measured using FT-FIR (resolution: 0.12 THz; zero-filling factor: 4). The measurements were conducted at room temperature and under a vacuum (~1 torr) to avoid water vapor absorption. The absorption spectra are shown in Figure 28. Lysozyme has a broad absorption peak at ~4.6 THz (153 cm−1) and three other minor absorption peaks at ~7 THz (233 cm−1), ~9.5 THz (317 cm−1), and 12.5 THz (417 cm−1), which are all in good agreement with the literature.15 Other three solid-state proteins including hemoglobin, α-chymotrypsin, and bovine serum albumin also have broad absorption features. These broad absorption peaks are possibly due to the large number of collective vibrational modes in these proteins. The assignments or theoretical calculations of these collective vibrational modes remain currently a very difficult problem due to the complexity of involving large numbers of atoms in such large molecules. The absorption spectrum of solid-state lysozyme is very different from that of lysozyme in solution as measured using Raman spectroscopy.29 Two factors possibly contribute to this. First, the conformational state of lysozyme molecules in solid form is different from that in an aqueous

H.-B. LIU AND X.-C. ZHANG

298

solution, which results in different THz collective vibrational modes. Second, Raman scattering (in THz band) and THz absorption are governed by different selection rules, therefore Raman spectroscopic features are different from the THz absorption features. 0.15

0.12 Lysozyme Absorbance (a.u.)

Absorbance (a.u.)

0.09 0.06 0.03 0.00

2

4

6

0.06 0.04 0.02 2

4

6

8 10 12 14 16 18 20 Frequency (THz)

0.16 α-Chymotrypsin

Bovine serum albumin Absorbance (a.u.)

0.10 Absorbance (a.u.)

0.08

0.00

8 10 12 14 16 18 20 Frequency (THz)

0.12

0.08 0.06 0.04 0.02 0.00

Hemoglobin

0.10

0.12

2

4

6

8 10 12 14 16 18 20 Frequency (THz)

0.12

0.08

0.04

0.00

2

4

6

8 10 12 14 16 18 20 Frequency (THz)

Figure 28. Absorption spectra of solid-state lysozyme, hemoglobin, α-chymotrypsin, and bovine serum albumin obtained via FT-FIR.

In addition, in the range of 0.2–2 THz, the absorption coefficient and refractive index of lysozyme (pellet sample in ~2.5 mm thickness prepared using 4 tons pressure with a hydraulic press) were measured using the THzTDS system (resolution: 0.1 THz; zero-filling factor: 4) in a transmission mode under a nitrogen purge (relative humidity ~0%), as illustrated in Figure 29. There are no absorption or dispersion features in the 0.2–2 THz range. The small variations in the bands close to 0.2 and 2 THz result from the relative lower SNR in those frequency range. Bacteriorhodopsin and myoglobin were also studied using THz-TDS previously.102,103 In both

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

299

cases, the THz absorbance also increases with frequency without strong narrow features over the range of 0.2–2 THz. The inhomogeneous broadening plus the intrinsically high spectral density of collective vibrational modes (as expected for big protein molecules of these sizes) possibly combine to obscure the absorption spectra in the region below 2 THz, making them featureless.

Figure 29. (a) Absorption spectrum and (b) refractive index of solid lysozyme obtained via THz-TDS.

4.2.2. THz Spectroscopy of protein microsuspensions in organic media The previous section presents the THz spectroscopy of proteins reflecting their collective vibrational modes in solid-state forms, with no bioactivities. Proteins, especially most of enzymes, usually need to be in aqueous solutions to be bioactive. Only the collective vibrational modes of bioactive proteins may provide information about their conformation states and therefore an understanding of protein functions or enzyme activities. Unfortunately, probing these bioactive proteins with THz spectroscopy is difficult due to the tremendous THz absorption of bulk water in aqueous solutions. On the other hand, there are some other bioactive protein systems in which bulk water can be avoided, such as hydrated protein powders, proteins in ice, and protein microsuspensions in organic solvents. In these environments, the THz dielectric properties of proteins are easier to measure than in aqueous solutions, but at the expense of investigating a system that is away from the in vivo condition of proteins. Protein microsuspensions in organic solvents are suitable systems for THz investigation, since biochemical people are able to suspend functional proteins (enzymes) in some nonpolar organic solvents, which have small

300

H.-B. LIU AND X.-C. ZHANG

dielectric constants (absorptions) in the THz band. In addition, there are also a large number of parameters to vary in organic solvents, including the shape of the protein molecule, the amount of bound water (different hydration), pH value, temperature, etc. Most importantly, the technological utility of enzymes can be enhanced greatly by using them in organic solvents rather than their natural aqueous reaction media although generally the catalytic activity displayed by enzymes in neat organic solvents is far lower than in water. Studies over the last 15 years have revealed not only that this change in solvent is feasible, but also that in such seemingly hostile environments enzymes can catalyze reactions impossible in water, become more stable, and exhibit new behavior such as “molecular memory.” Of particular importance has been the discovery that enzymatic selectivity, including substrate, stereo-, regio-, and chemoselectivity, can be markedly affected, and sometimes even inverted, by the solvent. Enzymecatalyzed reactions in organic solvents, and even in supercritical fluids and the gas phase, have found numerous potential applications, some of which are already commercialized.104 Therefore, a study of this kind of protein system with THz spectroscopy is of great significance. In our investigation, some preliminary studies on the THz spectroscopy of α-chymotrypsin in isooctane were conducted using both FT-FIR and THz-TDS. Firstly, the THz spectra of a list of organic solvents which are commonly used organic media for enzymes in biochemical research and industry were tested for their THz transmissions. These organic solvents include toluene, methanol, 1, 4-dioxane, acetone, tetrahydrofuran, hexane, and isooctane. The samples were measured in a sealed liquid cell with polyethylene windows using FT-FIR. The organic solvents with the best THz transmission (1.5–20 THz) are isooctane and hexane (both are nonpolar solvents). Isooctane was selected for experiments. The suspensions of enzymes within organic solvents normally result in a two-phase system as all water-soluble proteins possess a significant amount of strongly bound water (also called biological water), even when in an apparently dry state. Two configurations for a protein within an organic solvent are shown schematically in Figure 30(a) and (b). The configuration that protein encapsulated in the organic solvent with the assistance of the anionic surfactant, also called ion-paired form of the protein,105 was studied here. A typical surfactant, sodium bis(2-ethylhexyl) sulfosuccinate (AOT), illustrated in Figure 30(c) was used to extract protein from the aqueous phase to the isooctane phase. α-chymotrypsin, an important enzyme, was used as an example.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

301

About 10 mM α-chymotrypsin was dissolved in bis-tris propane (a biological buffer) solution (pH 7.8) with 6 mM CaCl2 added. About 15 mL of this solution was mixed with 2 mM AOT in 15 mL isooctane using magnetic stirring for 3 min. The two phases were then incubated for 24 h at room temperature followed by centrifugation at 6,000 rpm for 5 min to obtain clear phases. The organic (isooctane) phase was then recovered and assayed for α-chymotrypsin using absorbance at 280 nm (with an extinction coefficient of 40.4 M−1cm−1). The achieved concentration of α-chymotrypsin was 5 mg/mL. The samples with concentrations of 15 and 30 mg/mL were also obtained by evaporating the isooctane.105

Figure 30. Schematic diagram showing two configurations of a protein within an organic solvent. (a) Almost-anhydrous protein suspended in the organic solvent. The protein is surrounded by a thin bound water layer; (b) Protein encapsulated in the organic solvent with the assistance of the anionic surfactant. The surfactant, e.g. AOT, is only found at the bound water boundary; (c) Molecular structure of AOT.

In the range of 2–20 THz, the absorption spectra were obtained with FT-FIR (resolution: 0.12 THz; 16 scan; zero-filling factor: 4). The samples (100 µm in thickness) were measured in a sealed liquid cell with polyethylene windows at room temperature and under a vacuum (~1 torr) to avoid water vapor absorption. The AOT-isooctane solution with the same concentration as that in the α-chymotrypsin sample was measured as the reference spectrum. The absorption spectra are shown in Figure 31. The absorption spectra of the bioactive α-chymotrypsin (ion-paired form in isooctane) are different from that of solid-state α-chymotrypsin. There are some artifacts in the 11– 13 THz range due to the low SNR of the system in this range. By measuring three samples with different concentrations of α-chymotrypsin,

H.-B. LIU AND X.-C. ZHANG

302

the absorption features were clarified, as shadowed in Figure 31. The absorption peaks increase with concentration as expected. These absorption peaks may reflect the collective vibrational modes of the bioactive αchymotrypsin in the THz range, which could provide information about its functional 3D conformation states. However, the assignments of these collective vibrational modes are currently difficult due to the complexity of involving large numbers of atoms in protein molecules. 0.8

Absorbance (a.u.)

α-Chymotrypsin cast film

0.6

0.4

30 mg/mL

α-Chymotrypsin

in AOT-isooctane 15 mg/mL

0.2

5 mg/mL 0.0

2

4

6

8 10 12 14 Frequency (THz)

16

18

20

Figure 31. Absorption spectra of the bioactive α-chymotrypsin (ion-paired form in isooctane) with different concentrations, in comparison with the absorption spectrum of solid-state α-chymotrypsin (vertically offset by 0.55 for clarity). The shadow areas mark the absorption features of the bioactive α-chymotrypsin.

In addition, in the range of 0.2–2.8 THz, the absorption coefficient and refractive index of bioactive α-chymotrypsin microsuspensions in AOTisooctane were also measured using THz-TDS (resolution: 0.1 THz; zerofilling factor: 4) under a vacuum. The liquid cell, sample, and reference used in the measurements were the same as what was previously described. Figure 32 shows the absorption spectrum of the bioactive α-chymotrypsin microsuspensions in AOT-isooctane compared with that of an α-chymotrypsin pellet. In the absorption spectra, the increasing baselines with frequency result from the scattering of the suspension and powder sample. There are no absorption or dispersion features in solid-state α-chymotrypsin in the 0.2–2.4 THz range. However, the bioactive α-chymotrypsin exhibits a weak absorption peak in the 1.5–2.2 THz range for, possibly reflecting the collective vibrational mode of the bioactive α-chymotrypsin in this lowfrequency range.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

303

To vary the parameters, such as the amount of bound water (different hydration), pH value, and temperature, in these bioactive α-chymotrypsin microsuspensions then study the corresponding THz spectroscopy to probe the 3D conformation states of functional proteins will be a topic for future investigations.

Absorbance (a.u.)

2.0 α-Chymotrypsin pellet α-Chymotrypsin

1.5

(in AOT-isooctane)

1.0

0.5

0.0

0.4

0.8 1.2 1.6 Frequency (THz)

2.0

2.4

Figure 32. The absorption spectrum of bioactive α-chymotrypsin microsuspensions in AOT-isooctane, in comparison with that of solid-state α-chymotrypsin (powder pellet).

4.3. SENSING CHANGES OF CELL MONOLAYERS WITH THz DIFFERENTIAL TIME-DOMAIN SPECTROSCOPY106

Within the last three decades, numerous technologies have been developed for measuring the shape and behavior of individual cells and cell clusters in culture. For example, optical microscopes are the benchmark tools for the characterization of cells in the visible and infrared bands, while electric cell-substrate impedance sensing (ECIS) technology, developed by Applied BioPhysics Inc. (New York, USA), measures the subtle changes in AC impedance (100 Hz–100 kHz) of cell monolayers in real time.107 Timedomain dielectric spectroscopy is another well-developed technology used to characterize cell suspensions over a frequency range of 100 kHz–10 GHz.108 What all of these techniques share in common is that they are noninvasive and can be applied to living cells, which allows for measurements of how these cells respond to specific stimuli over time. But a wide range of cellular behaviors occur at a small-enough scale that they are not detectable using these devices. For example, the electrical impedance across a monolayer of cells is caused by very close association between the plasma membrane of neighboring cells, such that very small changes (on the scale of nanometer distances) can cause dramatic alterations

304

H.-B. LIU AND X.-C. ZHANG

in the paracellular permeability of these monolayers. This is especially true in the microvasculature, where minute changes in endothelial cell shape affect the permeability of capillaries, which in turn affects the transport of nutrients and cellular metabolites through tissues. Devices such as ECIS, which indirectly measure these morphological changes, are not capable of resolving the changes of individual cells within a monolayer. Likewise, even the best optical microscopes cannot resolve structures smaller than 200 nm. Thus, these methods can provide real-time measurements of cell behavior, but at the cost of poor resolving power. A potential solution to this problem is to sense living cells in the THz band (0.1–10 THz). Biomaterials inside cells, including amino acids, nucleic acids, proteins, and DNA have characteristic spectral features in the THz band.12, 15, 35 Historically, the THz band was relatively unexplored as a sensing tool in biological science, due to the absence of powerful sources and efficient detectors. THz-TDS, developed in the last decade, enables us to explore biosensing in the THz band. Compared with other THz technologies, THz-TDS holds a much higher SNR, ideal for biosensing and identification. THz differential time-domain spectroscopy (THz-DTDS) improves the SNR of THz-TDS by directly measuring the difference between the THz transmissions of a thin film and its blank substrate using a double lock-in amplifier system. It was originally used to characterize thin dielectric films with submicron thickness.109 It was then used for bioaffinity sensing due to its high sensitivity.36, 37 Since liquid water has large absorption coefficients in the THz band, it is supposed that THz-DTDS will be capable of characterizing very thin water layers. Liquid water has an absorption coefficient of ~200 cm–1 at 1 THz at room temperature110; therefore the THz amplitude change caused by 1 nm-difference of liquid water layer is approximately 10−5. The SNR of the THz-DTDS system can reach 20,000:1, which enables THz-DTDS to sense a 1 nm-thickness difference of liquid water. This is essential to sense live cells, as they consist of more than 70% (w/w) water. Given its potential to resolve small structural details, THz-DTDS was used to resolve small changes in monolayers of live cells. Specifically, vascular endothelial growth factor (VEGF) was used to induce a flattening and retraction of endothelial cells within a monolayer, and compared the resolving power of THz-DTDS with conventional phase contrast microscopy and ECIS.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

305

4.3.1. Samples and methodology Substrates (~2×3 cm) were cut from tissue culture-treated dishes (polystyrene, almost transparent for THz waves). Bovine lung microvessel endothelial cells (BLMVEC) were cultured for five days on polystyrene substrates in MCDB 131 complete medium containing antibiotics (VEC Technologies, USA) to 100% confluence at 37°C in a humidified atmosphere containing 5% CO2. The growth medium was renewed every other day. The cell confluence was assessed by phase contrast microscopy for the THz-DTDS experiments, and by a plateau in the transendothelial resistance for the ECIS experiments. Before treatment with VEGF, the cells were washed twice with phosphate-buffered saline and growth media was changed to a defined medium (Dulbecco’s modified Eagle Medium, supplemented with 1% wt/vol bovine serum albumin) without growth factors. VEGF was added to one group of samples at a final concentration of 125 ng/ml in the defined medium. The other group of samples was untreated. Phase contrast images and THzDTDS measurements were acquired after 3 and 24 h of treatment. The blank polystyrene substrates were tested with THz-DTDS for their smoothness. The differences between the right half and left half of the blank substrates introduce small THz differential signals at the level of the background noise. Half of the cell monolayer on a substrate was removed to obtain the cell samples for the THz-DTDS measurement, as indicated in Figure 33. Before the THz-DTDS measurements, the cell samples were extracted from the medium and placed in ambient air for about 1 min to dry the cell surface completely. To avoid the possible errors introduced by the incomplete drying, the cell samples were tested by THz transmission measurements before the THz-DTDS measurements. THz-DTDS measurements were usually made after the samples were dried in ambient air for about 1 min to ensure the surface water was completely evaporated. Figure 33 is a schematic of the THz-DTDS system. In this work, the THz emitter was an AC (10 kHz sinusoidal wave) driven GaAs antenna. The pump laser was a Tsunami femtosecond laser (Spectra-Physics, USA), with a pulse duration of ~100 fs, wavelength of 800 nm, and average power of 1.3 W. The generated THz pulses propagated through four off-axis parabolic mirrors and were focused on a 2 mm-thick <110> ZnTe crystal, wherein the probe beam detected the THz field via electro-optic sampling. The sample was placed between the second and third parabolic mirrors, with the THz beam focused on it. A galvanometer-controlled sample holder shook the sample with a frequency of 20 Hz and an amplitude of ~5 mm,

306

H.-B. LIU AND X.-C. ZHANG

causing the THz beam to alternately transmit through the cell monolayer (sample) and blank substrate (reference). The system measured the THz transmission difference between the sample and reference. Figure 33 illustrates a schematic of THz-DTDS measuring a cell monolayer sample. Each measurement took approximately 1 min. The THz pulse had a pulse width of ~2 ps, and a spectral bandwidth of 0.1–2 THz. THz-DTDS typically has a SNR five times better than conventional THz-TDS. In addition, THz-DTDS measures the difference between the sample and reference at a modulating frequency, e.g. 20Hz, which alleviats difficulties associated with temporally varying effects of water vapor absorption in ambient air.

Figure 33. A schematic diagram of the THz differential time-domain spectroscopy system with an illustration of a cell monolayer sample.

Bright-field images of the BLMVEC samples were obtained using a 10 × objective Nikon Eclipse TE2000 fluorescence inverted microscope equipped with a digital camera and SPOT image capturing software. ECIS 1600R (Applied BioPhysics Inc., USA) was used to measure the resistance of BLMVEC monolayers in the vertical direction, the same as the THz transmission direction in THz-DTDS measurement. ECIS arrays were prepared by adding a gelatin solution to each well (0.5 cm × 1 cm). The gelatin coats the well bottoms and provides a substrate, which the endothelial cells adhere to. Wells were washed three times using phosphate buffered saline then seeded with BLMVEC. The cells were allowed to grow to confluence for four days. The arrays were plugged into the instrument, and a baseline measurement of transendothelial resistance was obtained. Approximately 1 h later, the growth medium was exchanged for a

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

307

defined medium with and without VEGF. ECIS measurements were taken over a period of approximately 40 hours. 4.3.2. Results and discussion Figure 34(a) plots the THz differential waveform of BLMVEC monolayer and the transmitted THz waveform after blank polystyrene substrate (reference waveform). Figure 34(b) is the THz spectrum obtained by applying FFT on the reference waveform. The achieved dynamic range of the reference waveform exceeds 20,000, as indicated in Figure 34(a), enough to probe a thickness change in ~nm scale for liquid water or cell monolayers.

Figure 34. (a) THz differential waveform of BLMVEC (vertical scale expanded by a factor of 100) monolayer and the transmitted THz waveform after blank polystyrene substrate. The inset spectrum indicates the noise level of background; (b) THz spectrum of the THz waveform after FFT.

The THz differential waveforms of the blank substrate, the untreated cell sample and the VEGF-treated cell sample (for 3 hours) are shown in Figure 35(a). All the measurement results were obtained at room temperature in ambient air. The THz differential signal peaks of the cell monolayers were higher than that of the acellular (blank) substrates, indicating that we were able to identify the cell monolayers on the substrates with no difficulty. Importantly, the THz differential signal peak of VEGF-treated BLMVEC was 29% lower than that of the matched, untreated cells after only three hours of VEGF exposure (9.1 vs. 6.5 in arbitrary units, with a fluctuation of approximately 0.2). Optical phase contrast microscopy images revealed no noticeable morphological differences between treated and untreated samples, as shown in Figure 35(b) and (c), demonstrating that the cellular structural changes accounting for the different THz signals were too subtle to detect by conventional means.

308

H.-B. LIU AND X.-C. ZHANG

After one day of VEGF treatment, the difference between the peak differential signals of the treated and untreated BLMVEC increased to 43% (9 vs. 5.1), as shown in Figure 36(a). The one-day VEGF exposure resulted in a subtle but noticeable change in cell morphology, as shown by the phase contrast images in Figure 36(b) and (c). However, phase contrast microscopy lacks the resolving power to detect changes in the principal cellular structure targeted by VEGF: the tight junction (or zonula occludens). The tight junction is formed by close apposition of the lateral membranes of

Figure 35. (a) THz differential waveforms of blank substrate, untreated BLMVEC, and VEGF-treated BLMVEC (for 3 h); (b) and (c) Optical phase contrast microscopy images of BLMVEC before and after the treatment (for 3 h), respectively.

Figure 36. (a) THz differential waveforms of the blank substrate, untreated BLMVEC and VEGF-treated BLMVEC (for one day); (b) and (c) Optical phase contrast microscopy images of BLMVEC before and after the treatment (for one day), respectively.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

309

Figure 37. A schematic illustration used to determine the differential signal between the reference (blank substrate) and the cell monolayer on substrate.

adjacent endothelial cells, such that the intercellular spacing between cells is as small as 10–15 nm. The increases in permeability induced by VEGF therefore increase the size of this spacing. But this size is well below the practical threshold of even the best light microscopes (200 nm), so that even relatively large changes in the spacing are not detectable by conventional light microscopy. To directly observe this change in living cells is simply impossible using these microscopes. The decrease of the THz differential signal is most likely due to morphological changes in the cells after the VEGF treatment. A simplified model assumed for the THz wave transmission through a cell monolayer may help elucidate the mechanism of the THz differential signal decrease. Figure 37 illustrates a schematic diagram used to determine the THz differential signal between the reference (blank substrate) and the cell monolayer on substrate. In this model, the cell monolayer is represented as a thin film, with a thickness of d and a refractive index of n2(ω). ω is the angular frequency. E0(ω) is the incidental field, Eref(ω) is the reference field, and Ecell(ω) is the transmitted field after cell monolayer. n1 and n3 are the refractive indices of the air and the substrate, respectively. rij and tij represents the reflection and transmission coefficients of i→ j interface, respectively (i, j = 1, 2, and 3, i ≠ j). According to the Fresnel’s Law, the reference field is

E ref (ω ) = t13 exp(iωd / c ) E 0 (ω ) .

(13)

By considering the multireflection within the monolayer,111 the field Ecell(ω) can be written as

H.-B. LIU AND X.-C. ZHANG

310

Ecell (ω ) =

t12 t 23 exp[iδ (ω ) − α (ω )d ] E0 (ω ) , 1 − r21r23 exp[2iδ (ω ) − 2α (ω )d ]

(14)

where α(ω) is the amplitude absorption coefficient of the cell monolayer, and δ(ω) is the phase change of the field after the cell monolayer,

δ (ω ) = n2 (ω )ωd / c ,

(15)

c is the speed of the light in vacuum. For a thin sample, like a cell monolayer with a thickness approximating one micron, d « c/ω, therefore δ(ω) « 1. Under this approximation, we have

E diff (ω ) ≈ i

⎡ (n − n1 )(n 2 − n3 ) ⎤ E ref (ω )d ⎢n 2 − 1 + 2 ⎥ c (n1 + n3 ) ⎣ ⎦

ω

⎡ (n − n )(n − n ) ⎤ − Eref (ω )α (ω )d ⎢1 + 2 1 2 3 ⎥ , (n1 + n3 ) ⎦ ⎣

(16)

where Ediff(ω) = Eref(ω) – Ecell(ω). Therefore Ediff(ω) is proportional to the thickness of the monolayer sample d and is dependent on the complex THz dielectric property (refractive index n(ω) and absorption coefficient α(ω)) of the cell monolayers. The THz differential signal differences between the untreated cells and the treated cells for 3 h and those for one day are 29% and 43%, respectively. Since a 10–30% thickness change is within a normal range after the treatment for these cells, we assume that the thickness change of the BLMVEC monolayers is a main factor contributing to the THz differential signal differences upon the VEGF treatment. Based upon Eq. (16), the refractive index n(ω) and absorption coefficient α(ω) have a relatively smaller effect on the THz differential signal. n(ω) and α(ω) were hardly measured using THz-DTDS because the cell monolayers were not uniform films and their exact thickness values were not available. Therefore we could only give an approximate estimation. Assuming n(ω) = 2.2, α(ω) = 200 cm−1 (the same as liquid water), and d = 2 µm, a 20% decrease of n(ω) without a change of α(ω) will result in ~5% reduction of the THz differential signal. Another contribution was assumed from the increased spacing between cells after VEGF treatment, which enhanced the THz transmission across the cell monolayer and thereby decreased the THz differential signal. While conventional phase contrast light microscopy failed to detect these subtle morphological variations, they were readily detectable with THz-DTDS.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

311

ECIS was also used to confirm that VEGF induced the characteristic, precipitous drop in transendothelial resistance that accompanies BLMVEC cell flattening and retraction, under the same conditions that yielded the differences in THz-DTDS peaks. Because the plasma membrane of cells is a very poor conductor, electrical current flows primary through the intercellular spaces, such that transendothelial resistance increases as cells grow more confluent and the intercellular spaces become smaller. Once cells reach 100% confluence, the resistance is almost entirely dependent upon the structural integrity of tight junctions formed between neighboring cells. When this resistance is plotted against time, as shown in Figure 38(a), we

3000 Resistance (ohm)

2500 2000

Untreated

1500 1000

Treated

500

VEGF added

0

0

5

10 15 Time (hour)

20

25

(a)

Normalized Resistance

Untreated 0.8

0.8

0.6

0.6 0.4

Treated

0.4

0.2

0.2 0.0

0

5

10 15 Time (hour)

20

25

0.0

Normalized THz attenuation

1.0

1.0

(b)

Figure 38. (a) ECIS data vs. time for untreated and VEGF-treated BLMVEC. Time was set at zero when VEGF was added to treat the sample; (b) A comparison between the ECIS and THz-DTDS results. The resistance and THz attenuation are normalized to those of the untreated samples. The solid curves are ECIS data and the open circles are THz-DTDS data, with error bars indicated respectively.

H.-B. LIU AND X.-C. ZHANG

312

can detect even small changes in tight junction integrity, indicated by drops in resistance. Prior to the addition of VEGF, the AC resistance of the two BLMVEC monolayer samples were approximately equal. However, the resistance decreased immediately after addition of VEGF, then slowly “rebounded” as a result of the morphology changes affected by VEGF. The resistance difference between treated and untreated monolayers increased from 29% after 3 h to 50% after one day of VEGF exposure, which is remarkably close to the THz-DTDS results. Figure 38(b) presents a comparison between the normalized AC resistance (ECIS) and THzDTDS attenuation data over time, and they agree very well. The error bars in Figure 38(b) represent the standard deviations of three independent measurements for three groups of cell samples prepared under the same conditions. The standard deviations were 0.4–5% for ECIS data, and 2–4% for THz-DTDS data. Table 3 summarizes the comparison. The ECIS data confirm the morphology change of the cells detected with THz-DTDS. Further analysis of the ECIS and THz-DTDS results requires additional modeling of the ECIS output112 and is not covered here. TABLE 3. Comparison between ECIS and THz-DTDS results

AC resistance difference (ECIS) THz attenuation difference(THz-DTDS)

Frequency range 100 Hz–100 kHz 0.1–2 THz

Treated for 3h 29%

Treated for 24 h 50%

Standard error of the mean 0.4–5%

29%

43%

2–4%

THz-DTDS offers several advantages over existing techniques such as ECIS for sensing subtle changes in cultured cells. First, ECIS cannot reliably measure subtle variations in cell thickness or rearrangements in cell architecture (reflected by changes in dielectric properties) that do not affect intercell spacing, but these properties are readily detectable with THzDTDS. Second, THz photons have energies on the meV scale, which is 1/1,000 of the photon energy of visible light. Therefore THz waves are completely noninvasive for living cells. Third, ECIS requires plating cells on a customized substrate containing gold film electrodes, whereas THzDTDS can measure cell layers cultured on standard polystyrene or glass substrates. Finally, THz-DTDS extends the electromagnetic spectral measurement of cells to the THz range, a relatively unexplored space. THzDTDS is promising to provide THz spectral information unavailable through other conventional methods like optical phase contrast microscope and ECIS.

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

313

Some challenges remain to be solved for THz-DTDS to become an applicable cell sensing technology. Firstly, in transmission-based measurements, the cell samples must be extracted from the culturing medium to avoid the tremendous THz absorption of bulk water. Reflection-based THz spectroscopy may provide a solution to avoid this problem. Secondly, powerful THz sources and efficient THz detection methods need further improvements for more sensitive and real-time measurements. Finally, it is still challenging to employ the current THz-DTDS as an affordable biosensor, due to its cumbersome size and the high expense of femtosecond lasers. However, with the development of THz technology, THz spectroscopy integrated in chip with cell culturing may conceivably be implemented as a biosensor. 5. Conclusion and Prospects 5.1. DISCUSSIONS AND CONCLUSION

Our studies focus on the use of THz spectroscopy as a method for the detection, identification, and characterization of explosive, pharmaceutical, and biological materials, demonstrating that THz technologies are powerful tools for both fundamental scientific research and advanced technological applications. Although the atmosphere significantly attenuate THz waves, there exist THz transmission windows between water vapor absorption lines, allowing THz waves to propagate long distances (>10 m) to achieve standoff sensing. Most of the polycrystalline ERCs have THz spectral fingerprints in the 0.1–20 THz range, enabling THz technologies to detect and identify ERCs with a high specificity. Due to their excellent penetration through dielectric materials, and safety for both operators and targets, THz waves are well suited for the detection of hidden explosives. DFT calculations have been successfully employed to investigate the geometric structure and THz vibrational frequencies of some ERCs. And the calculated vibrational frequencies are in good agreement with the experimental data. Moreover, THz spectroscopy was successfully applied to detect and identify the explosive RDX in diffuse reflection geometry, which is significant for the standoff explosives detection in real-world scenarios. It is concluded that THz technologies are very promising for the detection of bulky explosives, and they will offer an opportunity for transformational advances in defense and security.

314

H.-B. LIU AND X.-C. ZHANG

The anhydrous and hydrated drugs were well identified via THz spectroscopy, presumably reflecting their different intermolecular vibrational modes due to differences in their crystallinity. The XRPD patterns confirmed the crystallinity differences between the unbonded and bonded drugs, as well as those of the anhydrous and hydrated drug substances. It is concluded that THz-TDS is an advantageous technique for pseudopolymorphic identification, and THz technologies have great potential to become a PAT in pharmaceutical production and quality control. THz-TDS was used to monitor the dehydration process of drug hydrates and a complex solid-state reaction of aminophylline monohydrate, in terms of the different THz absorption features during the solid-state reactions. It has successfully probed the dehydration kinetics of D-glucose hydrate and the kinetics of the solid-state reaction of aminophylline monohydrate. These investigations have opened new avenues for applying THz technologies in pharmaceutical science and industry. A number of biocompounds including amino acids, purines, purine derivatives, and other biomolecules exhibit THz absorption features in the 0.1–3 THz range, which are of significant interest in agricultural, pharmaceutical, and biomedical research and industry. THz spectroscopy of both solid-state proteins and protein microsuspensions in organic media was investigated and their THz absorption features possibly reflect their collective vibrational modes and could be used to probe their functional 3D conformation states. THz differential time-domain spectroscopy is capable of sensing minute changes of cell monolayers, pointing a new way for biosensing via THz technologies. THz technologies have proven to be powerful methods for diverse applications in the chemical, pharmaceutical, and biological fields. THz spectroscopy is well suited to complement conventional chemical and biological sensing and analysis technologies, such as x-ray diffraction, FTIR, nuclear magnetic resonance. 5.2. OBSTACLES AND CHALLENGES

However, there are still a number of obstacles and challenges in the use of THz spectroscopy and imaging for sensing applications. Some obstacles result from the intrinsic properties of THz waves. Some others are related to certain THz technologies. Water absorption and water vapor absorption of THz radiation may be the most restrictive problems. The tremendous water absorption (absorption coefficient of water is over 200 cm−1 at 1 THz at room temperature) has

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

315

greatly limited the applications of THz sensing and imaging in aqueous solutions or for wet samples, especially in biological and biomedical areas. Water vapor absorption affects the long distance (i.e. >100 m) propagation of THz waves in the atmosphere therefore restricts the standoff sensing distance. In THz sensing, most useful fingerprints arise from the phonon modes in the low-frequency range (<5 THz). But these phonon modes are generally many times or even one order weaker than the vibrational modes in the mid-infrared range, resulting in a relatively low sensitivity for THz sensing. In addition, for many chemical and biological compounds, especially big biomolecules, there are too many collective modes or intermolecular vibrational modes and it usually results in broad absorption bands or even featureless absorption spectra in the THz range, which reduces the specificity of THz sensing. The range of 0.1–3 THz is well suited for sensing hidden targets behind dielectric barrier materials. But for many chemical and biological compounds, the broad widths (usually >0.1 THz) of the absorption peaks and the overlaps of absorption bands also result in a relative low specificity compared with the sensing in mid-infrared and other bands. In many cases, THz sensing is similar to near-infrared sensing and multivariate analysis may be more effective in the screening and quantitative analysis. Understanding the intermolecular vibrational or phonon modes in the THz range is significant for both fundamental study and THz sensing. However, the theoretical calculations remain currently as a difficult problem due to the complexity involving the periodic density functional theory modeling and the computing of a large number of atoms in one crystalline cell (for example, there are four RDX molecules or 84 atoms in one crystalline cell). Several groups have tried but the investigations were unsuccessful. The resolution of conventional THz imaging is limited by the wavelength of THz waves (0.1 mm at 3 THz), which is not enough for some sensing and imaging application, especially in biomedical areas, for instance, biochip sensing or single cell-level characterizations. Powerful THz sources and sensitive THz detection methods desire further developments to enhance the SNR and acquisition speed of THz sensing and imaging, which are still much lower than those in mid-infrared and UV-visible bands. Different THz systems have their own limitations too. FT-FIR is limited for laboratory use due to the use of liquid helium in the bolometer. THzTDS systems measure the THz field therefore can be used to obtain the

316

H.-B. LIU AND X.-C. ZHANG

complex dielectric properties of materials. But this may limit its applications in real-world scenarios because the target does not always have a flat surface or homogeneous density. The target’s morphology information will be incorporated into the phase, resulting in an aberrant THz waveform and thereby artificial features in the THz spectrum. Additionally, although the broad band nature of THz-TDS makes it suitable for experimental investigations in the laboratory, it may prevent its applications for long distance standoff sensing due to the water vapor absorption. Moreover, there can be reflected peaks from the emitter, detector, and sample, due to the etalon effect and resulting in interference fringes in the THz spectrum. Finally, it is still challenging to employ the current THz-TDS as an affordable sensing technique, due to the high expense of femtosecond lasers (~$10,000 for femtosecond oscillators and ~$50,000 for femtosecond fiber lasers). 5.3. PROSPECTS FOR FUTURE WORK

The most recent developments in THz technology and new explorations of THz spectroscopy and imaging have advanced THz sensing applications in widespread areas. Progress critical to the successful implementation of these THz applications includes further improvements of THz technologies, an establishment of a more comprehensive THz spectral database for chemical and biological materials, advancements of applicable detection and identification THz systems in reflection modes, and developments of optimized data screening algorithms. Although THz-TDS has proven to be a useful technique in various studies and applications, its intrinsic characteristics may limit its application in real-world scenarios CW THz systems with narrowband THz radiation may evade these two major problems in real-world sensing applications. An array of CW sources and detectors covering a range wherein the target has THz fingerprints and the atmosphere has low attenuation will be more applicable for chemical and biological sensing in real-world scenarios. In most parts of our experiments, the measurements were conducted in transmission modes. However, for many applications, reflection measurements are more desirable. The detection and identification of RDX in a diffuse reflection mode described in our study has already demonstrated this possibility. However, widespread investigations on reflection measurements of explosive, chemical, and biological agents via either narrowband

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

317

or broadband THz waves need to be conducted to test and improve the sensing ability of THz technologies in reflection modes. In addition, advanced data classification and fusion algorithms will contribute considerably to the spectral analysis in THz sensing. Multivariate calibration methods, such as multiple linear regression (MLR), principal component regression (PCR), partial least squares (PLS) and neural networks (NN) can be used for identifying multispectral signatures, extracting signals from complex spectra, and achieving a better quantitative analysis. The existing obstacles and challenges do slow down the implement THz technologies for chemical and biological sensing in practical applications. However, THz science and technology has already shown its potential. Cooperated with the worldwide research activities, our THz sensing investtigations indicate THz technologies will play more and more important roles in fundamental scientific studies, technical applications, and even our daily lives.

References 1.

2. 3.

4.

5.

6.

7.

G. C. Walker, E. Berry, N. N. Zinovev, A. J. Fitzgerald, R. E. Miles, M. Chamberlain, and M. A. Smith, Terahertz imaging and international safety guidelines, in Medical Imaging 2002: Physics of Medical Imaging, L. E. Antonuk and M. J. Yaffe, Ed., Proceedings of SPIE, vol. 4682, pp. 683–690 (2002). E. Berry, Risk perception and safety issues, J. Biol. Phys., 29, 263–267 (2003). M. C. Kemp, P. F. Taday, B. E. Cole, J. A. Cluff, A. J. Fitzgerald, and W. R. Tribe, Security applications of terahertz technology, in Terahertz for Military and Security Applications, R. J. Hwu and D. L. Woolard, Ed., Proceedings of SPIE, vol. 5070, pp. 44–52 (2003). 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, Jpn. J. of Appl. Phys., 43, L414–L417 (2004). 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 timedomain and Fourier transform infrared spectroscopy, Appl. Phys. Lett., 85, 5535–5537 (2004). Y. C. Shen, T. Lo, P. F. Taday, B. E. Cole, W. R. Tribe, and M. C. Kemp, Detection and identification of explosives using terahertz pulsed spectroscopic imaging, Appl. Phys. Lett., 86, 241116 (2005). K. Kawase, Terahertz imaging for drug detection and large-scale integrated circuit inspection, Optics and Photonics News, 34–39 (Oct. 2004).

318 8.

9.

10.

11. 12.

13.

14.

15.

16. 17.

18.

19.

20. 21. 22.

23.

24.

H.-B. LIU AND X.-C. ZHANG P. F. Taday, I. V. Bradley, D. D. Arnone, and M. Pepper, Using terahertz pulse spectroscopy to study the crystalline structure of a drug: a case study of the polymorphs of ranitidine hydrochloride, J. Pharm. Sci., 92(4), 831–838 (2002). C. J. Strachan, T. Rades, D. A. Newnham, K. C. Gordon, M. Pepper, and P. F. Taday, Using terahertz pulsed spectroscopy to study crystallinity of pharmaceutical materials, Chem. Phys. Lett., 390, 20–24 (2004). C. J. Strachan, P. F. Taday, D. A. Newnham, K. C. Gordon, J. A. Zeitler, M. Pepper, and T. Rades, Using terahertz pulsed spectroscopy to quantify pharmaceutical polymorphism and crystallinity, J. Pharm. Sci., 94(4), 837– 846 (2005). A. J. Fitzgerald, B. E. Cole, P. F. Taday, Nondestructive analysis of tablet coating thicknesses using terahertz pulsed imaging, J. Pharm. Sci., 94, 177–183 (2005). B. M. Fischer, M. Walther, and P. Uhd Jepsen, Far-infrared vibrational modes of DNA components studied by terahertz time-domain spectroscopy, Phys. Med. Biol., 47, 3807–3814 (2002). Y. C. Shen, P. C. Upadhya, E. H. Linfield, and A. G. Davies, Temperature-dependent low-frequency vibrational spectra of purine and adenine, Appl. Phys. Lett., 82(14), 2350–2352 (2003). J. Nishizawa, T. Sasaki, K. Suto, T. Tanabe, K. Saito, T. Yamada, and T. Kimura, THz transmittance measurements of nucleobases and related molecules in the 0.4 to 5.8 THz region using a GaP THz wave generator, Opt. Comm., 246, 229–239 (2005). L. Genzel, L. Santo, and S. C. Shen, Far-infrared Spectroscopy of Biomolecules, in Spectroscopy of Biological Molecules, C. Sandory and T. Theophanides, Ed., pp. 609– 619 (D. Reidel, Boston, 1984). J. Bandekar, L. Genzel, F. Kremer, and L. Santo, The temperature dependence of the far-infrared spectra of L-alanine, Spectrochim. Acta A, 39A, 357–366 (1983). B. Yu, F. Zeng, Y. Yang, Q. Xing, A. Chechin, X. Xin, I. Zeylikovich, and R. R. Alfano, Torsional vibrational modes of tryptophan studied by terahertz time-domain spectroscopy, Biophys. J., 86, 1649–1654 (2004). T. M. Korter, R. Balu, M. B. Campbell, M. C. Beard, S. K. Gregurick, and E. J. Heilweil, Terahertz spectroscopy of solid serine and cysteine, Chem. Phys. Lett., 418, 65–70 (2006). M. Yamaguchi, F. Miyamaru, K. Yamamoto, M. Tani, and M. Hangyo, Terahertz absorption spectra of L-, D-, and DL-alanine and their application to determination of enantiometric composition, Appl. Phys. Lett., 86, 053903 (2005). T. M. Korter and D. F. Plusquellic, Continuous-wave terahertz spectroscopy of biotin: vibrational anharmonicity in the far-infrared, Chem. Phys. Lett., 385, 45–51 (2004). P. C. Upadhya, Y. C. Shen, A. G. Davies, and E. H. Linfield, Terahertz time-domain spectroscopy of glucose and uric acid, J. Biol. Phys., 29, 117–121 (2003). J. Nishizawa, K. Suto, T. Sasaki, T. Tanabe, and T. Kimura, Spectral measurement of terahertz vibrations of biomolecules using a GaP terahertz-wave generator with automatic scanning control, J. Phys. D: Appl. Phys., 36, 2958–2961 (2003). 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 time-domain spectroscopy, Chem. Phys. Lett., 332, 389–395 (2000). M. Walther, P. Plochocka, B. Fischer, H. Helm, and P. Uhd Jepsen, Collective vibrational modes in biological molecules investigated by terahertz time-domain spectroscopy, Biopolymers, 67, 310–313 (2002).

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

319

25. R. Nossal and H. Lecar, in Molecular and Cell Biophysics, 1st edn. (Addison-Wesley, Redwood City, CA, , 1991). 26. A. Roitberg, R. B. Gerber, R. Elber, and M. A. Ratner, Anharmonic wave functions of proteins: quantum self-consistent field calculations of BPTI, Science, 268, 1319–1322 (1995). 27. S. Hayward and N. Go, Collective variable description of native protein dynamics, Annu. Rev. Phys. Chem, 46, 223–250 (1995). 28. K. G. Brown, S. C. Erfurth, E. W. Small, and W. L. Peticolas, Conformationally dependent low-frequency motions of proteins by laser Raman spectroscopy, Proc. Nat. Acad. Sci. USA, 69(6), 1467–1469 (1972). 29. L. Genzel, F. Keilmann, T. P. Martin, G. Winterling, Y. Yacoby, and M. W. Makinen, Low-frequency Raman spectra of lysozyme, Biopolymers, 15, 219–225 (1976). 30. A. G. Markelz, A. Roitberg, and E. J. Heilweil, Pulsed terahertz spectroscopy of DNA, bovine serum albumin and collagen between 0.1 and 2.0 THz, Chem. Phys. Lett., 320, 42–48 (2000). 31. C. Zhang, E. Tarhan, A. K. Ramdas, A. M. Weiner, and S. M. Durbin, Broadened farInfrared absorption spectra for hydrated and dehydrated myoglobin, J. Phys. Chem. B, 108, 10077–10082 (2004). 32. D. M. Leitner, Vibrational energy transfer in helices, Phys. Rev. Lett., 87, 188102 (2001). 33. L. Y. Zhu, J. T. Sage, and P. M. Champion, Observation of coherent reaction dynamics in heme proteins, Science, 266, 629–632 (1994). 34. A. Xie, Q. He, L. Miller, B. Sclavi, and M. R. Chance, Low frequency vibrations of amino acid homopolymers observed by synchrotron far-IR absorption spectroscopy: excited state effects dominate the temperature dependence of the spectra, Biopolymers, 49, 591–603 (1999). 35. M. Nagel, P. H. Bolivar, M. Brucherseifer, H. Kurz, A. Bosserhoff, and R. Bu, THz technology for label-free genetic diagnostics, Appl. Phys. Lett., 80, 154–156 (2002). 36. S. P. Mickan, A. Menikh, H.-B. Liu, C. Mannella, R. MacColl, D. Abbott, J. Munch, and X.-C. Zhang, Label-free bioaffinity detection using terahertz technology, Phys. Med. Biol., 47, 3789–3795 (2002). 37. A. Menikh, S. P. Mickan, H.-B. Liu, R. MacColl, X.-C. Zhang, Label-free amplified bioaffinity detection using terahertz wave technology, Biosen. Bioelectron., 20, 658– 662 (2004). 38. T. S. Hartwick, D. T. Hodges, D. H. Barker, and F. B. Foote, Far infrared imagery, Appl. Opt., 15(8), 1919–1922, (1976). 39. P. H. Siegel, R. J. Dengler, E. Mueller, C. Readhead, and S. E. Fraser, Scanned Pixel Heterodyne Terahertz Imaging, in the British Royal Soc. Terahertz Gap Meeting (June 3, 2003). 40. K. J. Seibert, T. Loffler, H. Quast, M. Thomson, T. Bauer, R. Leonhardt, S. Czasch, and H. G. Roskos, All-optoelectronic continuous wave THz imaging for biomedical applications, Phys. Med. Biol., 47(21), 3743–3748 (2002). 41. A. J. Fitzgerald, E. Berry, N. N. Zinov’ev, S. Homer–Vanniasinkam, R. E. Miles, J. M. Chamberlain, and M. A. Smith, Catalogue of human tissue optical properties at terahertz frequencies, J. Biol. Phys., 29, 123–128 (2003). 42. T. Loffler, K. Siebert, S. Czasch, T. Bauer, and H. G. Roskos, Visualization and classification in biomedical terahertz pulsed imaging, Phys. Med. Biol., 47(21), 3847– 3852,(2002).

320

H.-B. LIU AND X.-C. ZHANG

43. T. Loffler, T. Bauer, K. J. Siebert, H. G. Roskos, A. Fitzgerald, and S. Czasch, Terahertz dark field imaging of biomedical tissue, Opt. Express, 9(12), 616–621(2001). 44. P. Knobloch, K. Schmalstieg, M. Koch, E. Rehberg, F. Vauti, and K. Donhuijsen, THz imaging of histo-pathological samples, Proceedings of SPIE, vol. 4434, pp. 239–245 (2001). 45. P. Wallace, D. Arnone, R. M. Woodward, and R. J. Pye, Biomedical applications of terahertz pulsed imaging, Proceedings of 2nd Joint EMBS/BMES Conference, Houston, TX, pp. 2333–2334 (2002). 46. R. M. Woodward, V. P. Wallace, R. J. Pye, B. E. Cole, D. D. Arnone, E. H. Linfield, and M. Pepper, Terahertz pulse imaging of ex vivo basal cell carcinoma, J. Investigative Dermatol., 120(1), 72–78 (2003). 47. B. E. Cole, R. Woodward, D. Crawley, V. P. Wallace, D. D. Arnone, and M. Pepper, Terahertz imaging and spectroscopy of human skin, in-vivo, Proceedings of SPIE, vol. 4276, pp. 1–10 (2001). 48. R. M. Woodward, B. E. Cole, V. P. Wallace, R. J. Pye, D. D. Arnone, E. H. Linfield, and M. Pepper, Terahertz pulse imaging in reflection geometry of human skin cancer and skin tissue, Phys. Med. Biol., 47, 3853–3863 (2002). 49. J. C. Oxley JC, J. L. Smith, E. Resende, and E. Pearce, Quantification and aging of the post-blast residue of TNT landmines, J. Forensic. Sci., 48(4), 742–753 (2003). 50. National Research Council, Existing and potential standoff explosives detection techniques (The National Academies Press, 2004); http://www.nap.edu/catalog/10998. html. 51. J. P. Blitz, Diffuse Reflectance Spectroscopy, in Modern Techniques in Applied Molecular Spectroscopy Chapter 5, F. M. Mirabella, Ed., (Wiley, 1998). 52. B. M. Rice and C. F. Chabalowski, Ab initio and nonlocal density functional study of 1, 3, 5-trinitro-s-triazine (RDX) conformers, J. Phys. Chem. A, 101(46), 8720–8726 (1997). 53. Y. A. Gruzdkov and Y. M. Gupta, Vibrational properties and structure of pentaerythritol tetranitrate, J. Phys. Chem. A, 105(25), 6197–6202 (2001). 54. J. Clarkson, W. E. Smith, D. N. Batchelder, D. A. Smith, and A. M. Coats, A theoretical study of the structure, internal rotations and vibrations of 2, 4, 6-trinitoluene, J. Mol. Struct., 648(3), 203–214 (2003). 55. M. J. Frisch, et al., GAUSSIAN 03 (Revision B.05), Gaussian, Pittsburgh, PA, (2003). 56. A. Becke, Density-functional thermochemistry. III. The role of exact exchange, J. Chem. Phys., 98(7), 5648–5652 (1993). 57. C. Lee, W. Yang, and R. G. Parr, Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density, Phys. Rev. B, 37, 785–789 (1988). 58. A. D. McLean and G. S. Chandler, Contracted Gaussian basis sets for molecular calculations. I. Second row atoms, Z = 11–18, J. Chem. Phys., 72, 5639–5648 (1980). 59. R. Krishnan, J. S. Binkley, R. Seeger, and J. A. Pople, Self-consistent molecular orbital methods. XX. A basis set for correlated wave functions, J. Chem. Phys., 72(1), 650– 654 (1980). 60. D. J. Cook, B. K. Decker, G. Maislin, and M. G. Allen, Through container THz sensing: applications for explosive screening, Proceedings of SPIE, vol. 5354, pp. 55– 62 (2004). 61. M. C. Kemp, P. F. Taday, B. E. Cole, J. A. Cluff, A. J. Fitzgerald, and W. R. Tribe, Security applications of terahertz technology, in Terahertz for Military and Security

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

62.

63. 64.

65.

66.

67. 68. 69.

70. 71. 72. 73.

74. 75.

76.

77.

78.

321

Applications, R. J. Hwu and D. L. Woolard, Ed., Proceedings of SPIE, vol. 5070, pp. 44–52 (2003). 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, Jpn. J. Appl. Phys., 43, L414–L417 (2004). Infrared Spectra (Sadtler Research Laboratories, Researchers, Editors & Publishers, Philadelphia, 1980). H.-B. Liu, Y. Chen, G. J. Bastiaans, X.-C. Zhang, Detection and identification of explosive RDX by THz diffuse reflection spectroscopy, Opt. Express, 14(1), 415–423 (2006). Y. Chen, H.-B. Liu, Y. Deng, D. Schauki, M. J. Fitch, R. Osiander, C. Dodson, J. B. Spicer, M. Shur, and X. C. Zhang, THz spectroscopic investigation of 2, 4dinitrotoluene, Chem. Phys. Lett., 400, 357–361 (2004). S. Nashima, O. Morikawa, K. Takata, and M. Hangyo, Measurement of optical properties of highly doped silicon by terahertz time domain reflection spectroscopy, Appl. Phys. Lett., 79, 3923–3925 (2001). M. Khazan, R. Meissner, and I. Wilke, Convertible transmission-reflection timedomain terahertz spectrometer, Rev. Sci. Instrum., 72, 3427–3430 (2001). A. Pashkin, M. Kempa, H. Nemec, F. Kadlec, and P. Kuzel, Phase-sensitive timedomain terahertz reflection spectroscopy, Rev. Sci. Instrum., 74, 4711–4717 (2003). 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). F. Stern, in Solid Sate Physics, F. Seitz and D. Turnbull, Ed., vol. 15, p. 299 (Academic Press, New York, 1963). F. M. Mirabella, in Modern Techniques in Applied Molecular Spectroscopy Chapter 5 (Wiley , 1998). M. van Exter, C. Fattinger, and D. Grischkowsky, Terahertz time-domain spectroscopy of water vapor, Opt. Lett., 14, 1128–1130 (1989). T. Yuan, H.-B. Liu, J. Xu, F. Al-Douseri, Y. Hu, and X.-C. Zhang, Terahertz timedomain spectroscopy of atmosphere with different humidity, in Terahertz for Military and Security Applications, R. J. Hwu and D. L. Woolard , Ed., Proceedings of SPIE, vol. 5070, pp. 28–37 (2003). K. R. Khankari and J. W. D. Grant, Pharmaceutical hydrates, Therm. Acta, 248, 61–79 (1995). S. R. Byrn, R. R. Pfeiffer, M. Ganey, C. Hoiberg, G. Poochikian, Pharmaceutical solids: a strategic approach to regulatory considerations, Pharmaceut. Res., 12(7), 945– 954 (1995). P. F. Taday, I. V. Bradley, D. D. Arnone, and M. Pepper, Using terahertz pulse spectroscopy to study the crystalline structure of a drug: a case study of the polymorphs of ranitidine hydrochloride, J. Pharm. Sci., 92(4), 831–838 (2002). C. J. Strachan, T. Rades, D. A. Newnham, K. C. Gordon, M. Pepper, and P. F. Taday, Using terahertz pulsed spectroscopy to study crystallinity of pharmaceutical materials, Chem. Phys. Lett., 390, 20–24 (2004). C. J. Strachan, P. F. Taday, D. A. Newnham, K. C. Gordon, J. A. Zeitler, M. Pepper, and T. Rades, Using terahertz pulsed spectroscopy to quantify pharmaceutical polymorphism and crystallinity, J. Pharm. Sci., 94(4), 837– 846 (2005).

322

H.-B. LIU AND X.-C. ZHANG

79. H. Bothe, H. K. Cammenga, Compostion, properties, stability and thermal dehydration of crystalline caffeine hydrate, Therm. Acta., 40, 29–39 (1980). 80. S. R. Byrn, and C.-T. Lin, The effect of crystal packing and defects on desolvation of hydate crystals of caffeine and L-(–)-1,4-cyclohexadiene-1-alanine, J. Am. Chem. Soc., 98, 4004–4005 (1976). 81. H. G. M. Edwards, E. Lawson, M. Matas, L. Shields, and P. York, Metamorphosis of caffeine hydrate and anhydrous caffeine, J. Chem. Soc., Perkin Trans., 2(10), 1985– 1990 (1997). 82. M. Matas, H. G. M. Edwards, E. E. Lawson, L. Shields L, and P. York, FT-Raman spectroscopic investigation of a pseudopolymorphic transition in caffeine hydrate, J. Mole. Struc., 440, 97–104 (1998). 83. C. Sun, D. Zhou, D. J. W. Grant, and J. V. G. Young, Theophylline monohydrate, Acta. Cryst., E58, O368–O370 (2002). 84. E. Suzuki, K. Shimomura, and K. Sekiguchi, Thermochemical study of theophylline and its hydrate, Chem. Pharm. Bull., 37, 493–497 (1989). 85. E. Suihko, J. Ketolainen, A. Poso, M. Ahlgren, J. Gynther, and P. Paronen, Dehydration of theophylline monohydrate a two step process, Int. J. Pharm., 158, 47– 55 (1997). 86. D. J. Sutor, The structures of the pyrimidines and purines. VI. The crystal structure of theophylline, Acta. Cryst., 11, 83–87 (1958). 87. J. Nishizawa, K. Suto, T. Sasaki, T. Tanabe, and T. Kimura, Spectral measurement of terahertz vibrations of biomolecules using a GaP terahertz-wave generator with automatic scanning control, J. Phys. D. Appl. Phys., 36, 2958–2961 (2003). 88. P. C. Upadhya, Y. C. Shen, A. G. Davies, and E. H. Linfield, Terahertz time-domain spectroscopy of glucose and uric acid, J. Biol. Phys., 29, 117–121 (2003). 89. C. L. Liu, S. M. Wu, T. Chang, M. L. Liu, H. J. Chiang, Analysis of a mixture containing ampicillin anhydrate and ampicillin trihydrate by thermogravimetry, oven heating, differential scanning calorimetry, and x-ray diffraction techniques, Anal. Chim. Acta., 517, 237–243 (2004). 90. S. R. Byrn, R. R. Pfeiffer, and J. G. Stowell, in Solid-State Chemistry of Drugs Chapter 1, 2nd edn. (SSCI, 1999). 91. S. R. Byrn, R. R. Pfeiffer, and J. G. Stowell, in Solid-State Chemistry of Drugs Chapter 8, 2nd edn. (SSCI, 1999). 92. G. Buckton and A. E. Beezer, The applications of microcalorimetry in the field of physical pharmacy, Intl. J. Pharm., 72(3), 181–192 (1991). 93. R. K. Khankari and D. J. W. Grant, Pharmaceutical hydrates, Therm. Acta, 248, 61–79 (1995). 94. K. R. Beebe, R. J. Pell, and M. B. Seasholtz, Chemometrics: A Practical Guide, p. 348 (Wiley , New York, 1998). 95. L.Vina, S. Logothetidis, and M. Cardona, Temperature dependence of the dielectric function of germanium, Phys. Rev. B, 30, 1979–1991 (1984). 96. Y. C. Shen, P. C. Upadhya, E. H. Linfield, and A. G. Davies, Temperature-dependent low-frequency vibrational spectra of purine and adenine, Appl. Phys. Lett., 82, 2350– 2352 (2003). 97. J. H. Sharp, G. W. Brindley, and B. N. Achar, Numerical data for some commonly used solid state reaction equations, J. Amer. Ceram. Soc., 49, 379–382 (1966). 98. S. R. Byrn, R. R. Pfeiffer, and J. G. Stowell, in Solid-State Chemistry of Drugs Chapter 21, 2nd edn. (SSCI, 1999).

TERAHERTZ SPECTROSCOPY FOR SENSING APPLICATIONS

323

99. J. Han and R. Suryanarayana, Influence of environmental conditions on the kinetics and mechanism of dehydration of carbamazepine dehydrate, Pharm. Develop. Tech., 3(4), 587–96 (1998). 100. S. K. Rastogi, M. Zakrzewski, and R. Suryanarayanan, Investigation of solid-state reactions using variable temperature X-ray powder diffractometry. II. aminophylline monohydrate, Pharm. Res., 19( 9), 1265–1273 (2002). 101. M. C. Beard, G. M. Turner, and C. A. Schmuttenmaer, Progress towards twodimensional biomedical imaging with THz spectroscopy, Phys. Med. Biol., 47, 3841– 3846 (2002). 102. C. Zhang, E. Tarhan, A. K. Ramdas, A. M. Weiner, and S. M. Durbin, Broadened farinfrared absorption spectra for hydrated and dehydrated myoglobin, J. Phys. Chem. B, 108, 10077–10082 (2004). 103. S. E. Whitmire, D. Wolpert, A. G. Markelz, J. R. Hillebrecht, J. Galan, and R. R. Birgey, Protein flexibility and conformational state: a comparison of collective vibrational modes of wild-type and D96N bacteriorhodopsin, Biophys. J., 85, 1269– 1277 (2003). 104. A. M. Klibanov, Improving enzymes by using them in organic solvent, Nature, 409, 241–246 (2001). 105. V. M. Paradkar and J. S. Dordick, Mechanism of extraction of chymotrypsin into isooctane at very low concentrations of Aerosol OT in the absence of reversed micelles, Biotechnol. Bioeng., 43, 529–540 (1994). 106. H.-B. Liu, G. Plopper, S. Earley, Y. Chen, B. Ferguson, and X.-C. Zhang, “Sensing the change of biological cell monolayers by THz differential time-domain spectroscopy,” in press, Biosensors and Bioelectronics (2006). 107. I. Giaever and C. Keese, A morphological biosensor for mammalian cells, Nature, 366, 591–592 (1993). 108. R. Lisin, B. Z. Ginzburg, M. Schlesinger, and Y. Feldman, Time domain dielectric spectroscopy study of human cells. I. Erythrocytes and ghosts, Bioch. et Biophy. Acta, 1280, 34–40 (1996). 109. Z. Jiang, M. Li, and X.-C. Zhang, Dielectric constant measurement of thin films by differential time-domain spectroscopy, Appl. Phy. Lett., 76, 3221–3223 (2000). 110. C. Ronne, L. Thrane, P. O. Astrand, A. Wallqvist, K. V. Mikkelsen, and S. R. Keiding, Investigation of the temperature dependence of dielectric relaxation in liquid water by THz reflection spectroscopy and molecular dynamics simulation, J. Chem. Phys., 107, 5319–5331 (1997). 111. M. Born and E. Wolf, Principles of Optics Chapter 13, 4nd edn. (Pergamon Press, 1970). 112. I. Giaever and C. Keese, Micromotion of mammalian cells measured electrically, Proceedings of the National Academy of Sciences of the USA, vol. 88, pp. 7896–7900 (1991).

TERAHERTZ COMMUNICATIONS: A 2020 VISION MARTIN KOCH* Terahertz Consulting, Gliesmaroder St. 27, 38106 Braunschweig, Germany

Abstract. We discuss basic considerations for potential short-range THz communication systems which may replace or supplement present WLAN systems in 10–15 years from now. On the basis of a few fundamental estimations we show that such a system will need a line-of-sight connection between receiver and emitter. To circumvent the blocking of the direct line-of-sight connection indoor THz communication systems will also have to rely on non-line-of-sight paths which involve reflections off the walls. The reflectivity of the walls can be enhanced by dielectric mirrors. This new scheme makes steerable high-gain antennas a necessity. Hence, a wireless THz communication system can not be a simple extension of the existing technology of today’s local area networks. Instead it involves completely new concepts and ideas that have not yet been worked upon.

Keywords: wireless terahertz communication systems, terahertz antennas, terahertz reflectors

1. Introduction The demand for bandwidth in wireless short-range communication systems has doubled every 18 months over the last 25 years [1]. Since there is no reason why this trend should come to a sudden end one can extrapolate that wireless data rates of 15 Gbps will be needed in 10 years from now. Some of the applications which will require this tremendous bandwidth can already be foreseen others will emerge as technology evolves.

______

* To whom correspondence should be addressed. Martin Koch, Terahertz Consulting, Gliesmaroder St.27, 38106 Braunschweig, Germany, www: http://www.terahertzconsulting.de, e-mail: nfo@terahertz consulting.de

325 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 325–338. © 2007 Springer.

326

M. KOCH

One vision is that people wirelessly download the content of an entire movie DVD in just a few seconds while standing in front of a public multimedia station. The movies could be downloaded onto a mobile personal terminal which acts as a universal, context-aware communication tool. These terminals could connect themselves to partner systems in order to create an integrated communications system. It is conceivable that people of all ages and in any possible situation will be equipped with such a terminal within the foreseeable future. Besides communication and entertainment the main purpose of such terminals will be a support in everyday life in various situations. They could, for instance, give early warnings if dangerous traffic situations are about to occur. They could sense health problems and, if necessary, call for immediate assistance. If the terminal is used in a hospital environment it may assist the wireless transmission of huge amounts of medical data, e.g. 3D tomography images. THz transmission links could also be used for military purposes. As we will discuss below, THz signals are not far-ranging as they are heavily attenuated by free-space damping. Furthermore, THz links will have to be directed. Because of both properties THz waves could provide secure links which would be ideal for a communication on the battlefield, e.g. between tanks or individual soldiers (see Figure 1).

Figure 1. Secure THz link on the battlefield.

The present short-range communication systems Bluetooth and wireless local area networks will not be able to supply the bandwidth needed in 10

TERAHERTZ COMMUNICATIONS: A 2020 VISION

327

years time. They use carrier frequencies of a few gigahertz only. The same is true for the currently emerging ultra wide band technology. Data rates for these systems are limited to data rates below 1 Gbps. Yet, stationary pointto-point systems that work at 60 GHz are already introduced. All of the above systems will satisfy our bandwidth needs for the next 10–15 years. However, future systems which provide data rates in excess of 10 Gbps will have to work at several ten or better a few hundred gigahertz. Consequently, carrier frequencies of short-range communication systems will rise. Yet, the use of bandwidth is strictly regulated in the USA up to frequencies of 300 GHz with a window from 275 to 300 GHz which is reserved for communications [2]. In Europe the allocation is the same but ends at 275 GHz. Hence, there is free communication bandwidth above 275 GHz. Systems that work at these high frequencies already fall into the terahertz (THz) range. Frequencies of a few hundred gigahertz could be provided by electronic sources. There is a variety of compact classical microwave sources which can generate frequencies up to several hundred gigahertz. These include resonant-tunneling diodes, transferred-electron devices, and transit-time diodes [3]. Recently Eisele and coworkers demonstrated RF powers of up to 3 mW for InP Gunn devices on diamond heat sinks operating in the second-harmonic mode in the frequency range between 250 and 330 GHz [4]. If this power level is not high enough it can be enhanced by a subsequent amplifier. However, even more practical could be to multiply a lower frequency Gunn diode in a multiplier chain. 1.1. WHY NOT INFRARED SYSTEMS OR QUANTUM CASCADE LASERS?

But why not turn directly to the near- or mid-infrared frequency range to build communication systems supporting data rates of several gigabits per second? This is a question which is often asked because situations we all experience in our daily life. For example the remote control of our TV set and wireless connections between a laptop computer and a printer work with infrared radiation. Yet, there are several reasons which prevent wireless IR links from providing such high data rates. The modulation scheme at wavelengths of 880, 1,310, and 1,550 nm is simply intensity modulation and the radiation is directly detected with a photodiode. This results in poor receiver sensitivity as compared to heterodyne detectors. Furthermore, the ambient light noise at these wavelengths is significant. Hence, the transmission of a symbol (a “0” or a “1”) needs some time to exceed the noise level. To transmit higher data rates (i.e. a symbol in a

328

M. KOCH

shorter time) one would have to increase the signal intensity. This is not possible because of the existing eye-safety limit [5]. Furthermore, this would also imply a higher power consumption with is a severe drawback for battery powered portable devices. Although wireless IR systems are more than 30 years old the highest data rates demonstrated so far are 155 Mbps [6], [7]. One publication addressing the possible use of IR radiation for short-range communication concludes “[T]he results prove that IR seems to be unable to replace RF for indoor LAN applications” [8]. The same is true for wavelengths of several micrometers which is the domain of quantum cascade lasers (QCLs). QCLs have come a long way since their first demonstration in 1994 [9]. Early QCLs needed cryogenic cooling, worked only in a pulsed mode and emitted only in the midinfrared. Since then tremendous progress has been made. The race has been towards continuous wave operation, higher temperatures, and longer wavelengths. Today mid-infrared QCLs work continuous wave and at temperatures that even exceed room temperature. These QCLs are ready for industrial sensing and imaging applications. Unfortunately, no direct midinfrared detectors exist which are sensitive, fast, and work at room temperature. Heterodyne detection is possible in principle but very difficult. Furthermore, the ambient light noise problem is even more severe here. According to Wien’s law the black body radiation curve for room temperature peaks at 17.64 THz, corresponding to a wavelength of 17 µm. Meanwhile there are already QCLs operating at a few THz. In 2002 Tredicucci and coworkers presented a QCL working at 4.4 THz [10]. Recently we have seen QCLs with working frequencies as low as 2.1 THz and a 3.2 THz QCL emitting continuous wave (CW) at 93K [11]. Despite these advancements it is currently believed that THz QCLs will neither show emission at frequencies much below 1.5 THz nor work CW at room temperature in the near future. Yet, there may be another way. Very recently Hoffmann and coworkers could show that a single interband laser diode being placed in a special external resonator and forced to emit on two-longitudinal modes simultaneously emits THz radiation via a nonlinear process called four-wave mixing [12]. The same scheme could be even more efficient for QCLs (including room temperature mid-infrared QCLs) for their larger dimensions. This, however, remains to be demonstrated. Nevertheless, it may be possible that two-mode QCLs will be used some day as short-range sources for indoor THz communication systems.

TERAHERTZ COMMUNICATIONS: A 2020 VISION

329

1.2. ATMOSPHERIC AND FREE-SPACE DAMPING

The carrier frequency of choice for the future THz communication systems could be 300 GHz. On the one hand, it is unregulated and five times higher than the highest frequency of 60 GHz used in wireless communications today. On the other hand, the atmospheric attenuation is still moderate. Figure 2 shows the results of atmospheric attenuation simulations performed with FASCOD3 for 1976 US standard atmosphere conditions in the frequency range from 0.1 to 1 THz. Obviously there are several strong water lines in the atmospheric damping spectrum at frequencies above 300 GHz. In these regions it would be difficult to maintain a sufficient link budget together with the wide bandwidth required by the high gigabits per second data rate. In addition to this atmospheric attenuation which arises mainly from rotational transitions of molecules one has free-space damping. Figure 3 shows the cumulative damping of free-space and gaseous attenuation as function of both link distance and frequency. Even at frequency windows of high atmospheric transmission the free-space damping which increases with frequency is severe. For instance, the free-space damping is 40 dB higher at 300 GHz as compared to 3 GHz. This limits the practical communication distance to several 10 m. Hence future ultra

Figure 2. Atmospheric attenuation in the range from 0 to 1 THz range. (Courtesy of R. Piesiewicz.)

330

M. KOCH

Figure 3. Cumulative free-space and gaseous attenuation in the range from 100 GHz to 1 THz range. (Courtesy of R. Piesiewicz.)

broadband THz wireless communication systems will be most likely limited to medium-link and short-link indoor applications. Outdoor scenarios are conceivable in places where adverse weather conditions are rare. 1.3. STATE-OF-THE-ART

Although in principle a communication system at 300 GHz could be built right away within a few months with existing microwave hardware it has, to the best of our knowledge, not been done at this point. The reason for this is probably the high price of microwave sources for 300 GHz. Such a system would be too bulky and too expensive to be sold as a commercial product. Hence, communication with THz waves is still in its infancy. Probably the first transmission with THz waves at several hundred GHz has been performed using a room temperature semiconductor THz modulator which is based on the depletion of a two-dimensional electron gas (2DEG) [13]. The upper part of Figure 4 shows a schematic of the modulators working principle. It is based on the well-established technology for producing high electron mobility transistors (HEMTs), in which a 2DEG is confined at a GaAs/AlxGa1-xAs heterointerface [14]. The electron density of the 2DEG can be controlled by the application of an external voltage to a gate electrode. In [13] and [14] the electrode was formed by a thin Cr layer with a diameter of 3 mm. If a voltage is applied the 2DEG is depleted below the gate electrode. Consequently, the structure absorbs and reflects less and has a higher transmission.

TERAHERTZ COMMUNICATIONS: A 2020 VISION

331

Figure 4. Top: Schematic of the working principle of a HEMT-like THz modulator. Bottom: Transmitted sine wave with a frequency of 5,130 Hz.

In the data transmission experiment [13] audio signals up to 25 kHz were imprinted onto a 75 MHz train of broadband THz pulses in a standard time-domain spectrometer (THz-TDS). This pulse train was used instead of CW carrier waves simply because the THz spectrometer was at hand. The modulator was placed in the intermediate focus of the spectrometer so that the communication channel had a length of 48 cm and ranged from the modulator to the receiver antenna. The lower part of Figure 4 shows a transmitted sine wave with a frequency of 5,130 Hz. The upper curve shows the voltage applied to the gate contact and the lower curve the detector signal. The sinus is clearly reproduced. The signal strength was sufficient to transmit pieces of music with a quality comparable of that of a phone call. Of course, this does not represent a serious approach towards a THz communication system since the transmittable data rate is limited by the RC time of the modulator. The RC time will always be quite high since the gate has to have a certain diameter to match the focus of the THz beam. Yet, the race is on towards short-range THz communication systems. A major step in this direction has recently been achieved at NTT. Nagatsuma and coworkers demonstrated a 120 GHz wireless link using photonic techniques for generation, modulation, and emission of millimeter-wave signals [15], [16], and an all electronic receiver based on InP-HEMT technology. They demonstrated a data rate of 10 Gbps and estimated a

332

M. KOCH

maximum transmission distance of 1.0 km [16]. Their demonstration experiment was performed in collaboration with broadcasting companies, where the last-mile transmission of multiple channel HDTV signals is urgently needed. 2. Basic Considerations In the following we will present some conceptual ideas for a short-range THz communication system. Even though the hardware does not exist at the moment we can already estimate some working parameters that the components will have to meet. For example, we can estimate the antenna gain required for a reliable THz link. Let us assume that the carrier frequency was 300 GHz, the bandwidth was 10 GHz, the modulation scheme was 2-PSK, the desired bit-error-rate was 10−6 and the power density emitted by the transmitter was 1 mW/cm2. Using these parameters and some very basic equations (see [17] for details, the essential formulas (e.g. Friis formula) are textbook knowledge) one can estimate that a gain of 31 dB will be needed per antenna to compensate for free-space damping. Hence, the THz emission has to be highly directed and the transmission relies on a line-of-sight link between emitter and receiver. This is an important conceptual difference to present day’s indoor wireless communication systems. The line-of-sight requirement leads to obvious problems that have to be dealt with: the link can be broken by moving people or other objects which block the direct line-of-sight path. This problem could be solved by alternate transmission routes mediated by indirect non-line-of-sight paths that involve reflections off the walls as illustrated in Figure 5. Using ray-tracing simulations it was recently shown that a reliable high bandwidth link can be maintained, if a THz communication system can rely on such non-line-of-sight “billiard” channels as a backup [18]. The results in [18] show that a THz link cannot rely only on the direct line-of-sight path alone as there is always the chance that it is blocked by somebody stepping into the beam path. However, this billiard scheme implies that the above mentioned high-gain antennas are so called “smart antennas” which allow for beam-steering. Presently there is a lot of research activity on steerable antennas from 1 to 60 GHz. Smart antennas for THz frequencies have not been demonstrated yet. 3. THz Antennas At THz frequencies several antenna types have been employed including the pyramidal horn cavity with dipole, the corner reflector array, the bow-

TERAHERTZ COMMUNICATIONS: A 2020 VISION

333

Figure 5. Top: If an unbroken line-of-sight link between emitter and receiver exists it is used for transmission. Bottom: If the link is broken by an obstacle a non-line-of-sight “ billiard” channel is used instead.

tie dipole, planar antenna structures on dielectric lenses, and waveguide feed horns. Especially the last two antenna types are of interest for a potential THz communication system. While waveguide feed horns offer excellent performance and low losses, planar antenna structures offer greater potential for integration with other planar devices. Recently first simulations using ADS Momentum have been performed to investigate the performance of microstrip patch antenna arrays at 0.3 THz [17], [19]. At these high frequencies dielectric and conductor losses are expected to be very high. The simulations aimed to quantify these losses and the antenna gain achievable with patch antenna arrays. Losses due to surface roughness were not included. In a first step a 4×4 antenna

M. KOCH

334

Figure 6. 4 × 4 patch antenna array with feeding network.

array of patch antennas fed by a planar network was considered (see Figure 6). Each antenna element had dimensions of 288 by 360 µm. For this structure a maximum gain of 16 dB and a directivity of 18.1 dB were found. This gain can be increased even further by using arrays with more elements. The above mentioned 31 dB should in principle be possible with arrays of somewhat more than 1,000 elements. Yet, a planar feeding network may have losses which are too high to be practical. 4. Dielectric THz Mirrors The jump in refractive index between air and typical indoor interfaces like walls and furniture is not dramatic. Furthermore the roughness of typical indoor materials approaches the dimension of the wavelength in the THz range. Both effects combine and together lead to high THz reflection losses [20]. Hence, it can be expected that indirect non-line-of-sight paths involving reflections have significant diffuse contributions. The reflective properties of indoor interfaces can be enhanced by covering them with dielectric THz mirrors. Such mirrors have first been demonstrated in 2002 [21], [22]. Although they have to be further improved, they can potentially be produced inexpensively and in large areas. Recently we demonstrated the first omnidirectional mirror for the THz range [18]. The structure consists of five 150-µm thick layers of polypropylene with a refractive index of 1.53 and four 63-µm thick layers of high-resistivity silicon with a refractive index of 3.418. As shown on the left side of Figure 7 the layers are clamped together by a metal frame.

TERAHERTZ COMMUNICATIONS: A 2020 VISION

335

Figure 7. Left: Photo of the omni-directional PP/Si mirror. Right: Reflection coefficient of the mirror measured with the fiber-coupled reflection spectrometer for three different angles of incidence and for s-polarization (triangles). Numerical simulations (dotted line) reproduce the experimental data well. Calculations for normal (0°) and gazing incidence (89°) are displayed, too. An omni-directional frequency band of high transmission for all angles of incidence can be determined (shaded areas).

The right side of Figure 7 shows the reflection spectrum of the mirror structure for various angles and s-polarized THz waves. It was obtained with THz-TDS. The spectrum for p-polarized waves looks similar. The experimental data shown as symbols are in good agreement with transfer matrix calculations shown as solid lines. Altogether one finds that the mirror is highly reflecting for all incidence angles and s- as well as p-polarization in the frequency band between 319 and 375 GHz [18]. The use of thin slices of crystalline silicon in this particular mirror reduces its mechanical flexibility. Yet, exactly this property of polymeric mirrors may be required for practical applications. It may, however, be possible to replace these silicon layers by polymer layers the refractive index of which has been considerably enhanced by the addition of highindex microparticles in a compounding process. For example, a fine highresistivity silicon or TiO2 powder could be mixed with polypropylene to obtain a flexible high-index dielectric. In this case structures produced by coextrusion could further improve the quality and uniformity of the mirror. Recently ray-tracing simulations have been performed to demonstrate that THz reflectors can considerably improve the signal level. A typical indoor scenario was assumed. The propagation in a furnished room with people was investigated for the case when walls are covered with dielectric mirrors and for the case without them. The room had dimensions of

336

M. KOCH

6×5×2.5 m. The transmitter was placed in the middle of the room just below the ceiling whereas the position of the receiver was varied in a plane 1 m above the floor. Experimental date from wideband reflection measurements on typical indoor materials [20] and omni-directional dielectric mirrors [18] were used to make the simulations realistic. Figure 8 shows the received power level 1 m above the floor for the case with and without mirrors for TE polarization. The application of dielectric mirrors in office environments enhances the received power levels, on average by 9.5 dB and 6 dB, for TM and TE polarization respectively. Hence, the required gain of the antennas can be somewhat smaller if reflectors are placed on the walls. If antenna arrays were used they could contain fewer elements which, in turn, would make their fabrication easier. Furthermore, simulations show that entire walls need not be covered. Instead there are some “hot spots” where the wall reflectivity should be enhanced by dielectric mirrors.

Figure 8. Received power levels in a room with and without mirrors placed on the walls. (Courtesy of R. Piesiewicz.)

5. Conclusion Future wireless communication systems will have to support much higher data rates. Consequently, they will have to operate at much higher frequencies and therefore will approach the THz range. Although there has been recent progress THz communications is still in its infancy. Fundamental hardware to build a commercial wireless THz communication system does not exist today. We are far from having room temperature THz emitters or THz receivers that cost just several ten € and have the size of a Euro coin. Developing the technology for a short-range THz communication system will be challenging and represents a multidisciplinary and long-term task.

TERAHERTZ COMMUNICATIONS: A 2020 VISION

337

Some basic estimations show that THz communication systems will need a line-of-sight connection between receiver and emitter. To circumvent the blocking of the direct line-of-sight connection these systems will also have to rely on non-line-of-sight paths which involve reflections off the walls. Ray-tracing simulations show that indirect transmission paths between a transmitter and receiver, supported by dielectric mirrors, will provide better signal coverage and will make the THz communication channel much more robust against shadowing. Altogether, it is very likely that we will see indoor communications at a few hundred gigahertz in 10–15 years from now. First systems which rely on a point-to-point connection between stationary emitters and receivers may hit the market even earlier. Yet, the present technology has to be further developed in the years to come and a lot of problems need to be solved. Acknowledgment It is a pleasure to acknowledge my active cooperation with the group of Professor Kürner at the Institut für Nachrichtentechnik of the TU Braunschweig. The channel estimations and ray-tracing simulations have been performed by R. Piesiewicz. Indoor materials and THz mirrors were characterized by N. Krumbholz during a visit in the group of Professor Mittleman at the Rice University in Houston. I further acknowledge discussions with Professor A. Czylwik and Professor J. Schöbel as well as with several of my students including M.N. Islam, F. Rutz, R. Wilk, T. Kleine-Ostmann, S. Wietzke, and I.A. Ibraheem.

References 1. 2. 3. 4.

5. 6. 7.

S. Cherry, Edholm’s law of bandwidth, IEEE Spectr. 41, 50 (2004). see: http://www.ntia.doc.gov/osmhome/allochrt.pdf H. Eisele and G. I. Haddad, Two-terminal millimeter-wave sources, IEEE Trans. Mircow. Theory Techn., 46, 739 (1998). H. Eisele, M. Naftaly, and R. Kamoua, Generation of submillimeter-wave radiation with GaAs TUNNETT diodes and InP Gunn devices in a second or higher harmonic mode, Intern. J. Infrared Millimeter Waves, 26, 1 (2005). D.J.T. Heatley, et al., Optical wireless, the story so far, IEEE Communications Magazine, 36(12), 72–74 (Dec. 1998). D. C. O’Brien et al., High-speed integrated transceivers for optical wireless, IEEE Communications Magazine, 41(3)3, 58–62 (March 2003) V. Jungnickel et al., 155 Mbit/s wireless transmission with imaging infrared receiver, IEE Electronic Letters, 37(5), 314–315 (2001).

338 8. 9. 10.

11. 12.

13.

14. 15.

16.

17.

18.

19.

20. 21. 22.

M. KOCH M. Wolf and D. Kress, Short-Range Wireless infrared transmission: the link budget compared to RF, IEEE Wireless Communications, 10(2), 8–14 (April 2003). J. Faist, F. Capasso, D. L. Sivco, A. L. Hutchinson, and A. Y. Cho, Quantum cascade laser, Science, 264, 553 (1994). R. Köhler, A. Tredicucci, F. Beltram, H. E. Beere, E. H. Linfield, A. G. Davies, D. A. Ritchie, R. Iotti, and F. Rossi, Terahertz semiconductor-heterostructure laser, Nature, 417, 156 (2002). S. Kumar, B. S. Williams, S. Kohen, Q. Hu, and J. L. Reno, Frequency and phase-lock control of a 3 THz quantum cascade laser, Appl. Phys. Lett., 84, 2494 (2004). S. Hoffmann, M. Hofmann, E. Bründermann, M. Havenith, M. Matus, J. V. Moloney, A. S. Moskalenko, M. Kira, S. W. Koch, S. Saito, and K. Sakai, Four-wave mixing and direct terahertz emission with two-color semiconductor lasers, Appl. Phys. Lett., 84, 3587 (2004). T. Kleine-Ostmann, K. Pierz, G. Hein, P. Dawson, and M. Koch, Audio signal transmission over THz communication channel using semiconductor modulator, Electron. Lett., 40, 124 (2004). T. Kleine-Ostmann, K. Pierz, G. Hein, P. Dawson, and M. Koch, Electrically driven room temperature terahertz modulator, Appl. Phys. Lett., 84, 3555 (2004). A. Hirata, M. Harada, T. Nagatsuma., 120-GHz wireless link using photonic techniques for generation, modulation, and emission of millimeter-wave signals, IEEE J. of Lightwave Technol., 21, 2145 (2003). T. Nagatsuma, A. Hirata, Y. Sato, R. Yamaguchi, H. Takahashi, T. Kosugi, M. Tokumitsu, H. Sugahara, T. Furuta, and H. Ito, Sub-terahertz wireless communications technologies, Proceedings of the International workshop on Terahertz Technology 2005 (Osaka, Japan, Nov. 2005). R. Piesiewicz, M. N. Islam, M. Koch, and T. Kürner, Towards short-range terahertz communication systems: basic considerations, in 18th International Conference on Applied Electromagnetics and Communications, Dubrovnik, Croatia, Techn. Digest. p. 153, Oct. 2005. N. Krumbholz, K. Gerlach, F. Rutz, M. Koch, R. Piesiewicz, T. Kürner, and D. Mittleman, Omnidirectional terahertz mirrors: a key element for future THz communication systems, Appl. Phys. Lett., 88, 202905 (2006). M. N. Islam and M. Koch, Terahertz patch antenna arrays for indoor communications, Int. Conference on Next-Generation Wireless Systems 2006, (Dhaka, Bangladesh 2006). R. Piesiewicz, T. Kleine-Ostmann, N. Krumbholz, D. Mittleman, M. Koch, and T. Kürner, THz characterization of building materials, Electron. Lett., 41, 1002 (2005). D. Turchinovich, A. Kammoun, P. Knobloch, T. Dobbertin, and M. Koch, Flexible allplastic mirrors for the THz range, Applied Physics A, 74, 291 (2002). US patent 6.954.309 B2

Theme 5 OVERVIEW

APPLIED TERAHERTZ SCIENCE: THE TECHNOLOGY OF THE FUTURE, AND ALWAYS WILL BE? MARTYN CHAMBERLAIN* Department of Physics, Durham University, Durham DH1 3LE,United Kingdom and Scientific Coordinator of the EU TeraNova Project

Abstract. It is now over a decade since the first papers were published on Terahertz (THz) pulsed imaging. Much was then claimed for this field, and many applications were confidently foreseen in medicine, biology, process engineering, surveillance, and security. During the intervening 10 years, a variety of developments have taken place in both THz science and technology and the subject is maturing rapidly. In this final chapter, the central Themes of this workshop will be briefly reviewed. In addition, a summary will be presented of the key questions that must be addressed in the short term if applicable. THz science is to realise some of the potential which has been claimed for this new window on the electromagnetic spectrum.

Keywords: terahertz, devices, sources, systems, detection, sensing, imaging, interactions

1. Introduction This is the third meeting in the NATO series dealing with Terahertz (THz) science and technology. The last meeting took place some six years ago and during that time, considerable developments have taken place in both technological applications and underpinning science. The present programme was divided, for convenience, into four Themes: Devices, Interaction, Detection and Sensing, and Systems. There were, inevitably, a good number of correlations between these areas; an attempt was made

______

* To whom correspondence should be addressed: J. M. Chamberlain, Department of Physics, Durham University, Durham, UK, DH1 3LE; e-mail: [email protected]

341 R.E. Miles et al. (eds.), Terahertz Frequency Detection and Identification of Materials and Objects, 341–352. © 2007 Springer.

342

M. CHAMBERLAIN

throughout the programme to balance contributions from both the electronics and photonics traditions. Since the Chateau de Bonas Meeting in 2000, a considerable number of incremental steps have occurred in all of these Themes. For example, in Devices, source development from both the optical and electronic areas indicates that it is now convenient to consider a new Terahertz Gap, at approximately 1.0–2.5 THz: here, purely electronic devices are unlikely to be efficient, and difficulties may be experienced in the design and operation of Quantum Cascade Lasers (QCLs). Despite this comment, however, it should be noted that QCLs can now be realised with: an operating temperature of up to 163K for pulsed and 110K for CW; a lowest frequency of 1.38 THz; and a peak pulsed emission power of the order of 200 mW. From the electronics side: an output exceeding 1 THz has been reported for a Resonant Tunnelling Diode; careful heat-management strategies for InP Gunn diodes have resulted in sources operating up to 422 GHz; a 1.2-THz transistor is now envisaged by Intel; and advances in GaN materials processing may soon result in Tunnel Injection devices that can oscillate with useful power up to 700–800 GHz. In Interactions, the background science that is pertinent to new applications was reviewed. The most useful interaction of THz radiation with matter occurs with the motion of groups of relatively large groups of molecules; in consequence, such applications span the range from cosmology through atmospheric sensing to medicine and biology. In the present meeting, advances in biomolecular interactions were reviewed, and this topic is echoed in the next Theme. The ability of THz systems to recognise the binding state of DNA and by extension, to monitor the presence of mutations, opens up significant opportunities for harnessing interactions of this type. However, very severe problems remain which are largely connected with the (necessary) presence of water in biological material and the need to improve instrumental sensitivity. One particularly promising development, which may be a significant enabling technology, is in the field of plasmonics, that is in the use of passive devices which rely on the delivery of THz radiation to very small experimental samples, such as minute quantities of DNA, and which harness the propagation of THz waves at semiconductor surfaces. A continuing interest remains in the interaction of THz radiation with the molecular motion of the constituents of drugs of abuse, and of explosives, in view of the applications in law enforcement. One factor which is now emerging is the need to undertake realistic spectroscopy; that is, to take note of the presence of the scattering effects of mixing materials with which these substances are usually

TERAHERTZ SCIENCE AND TECHNOLOGY – THE FUTURE

343

associated. There has clearly been a considerable diversity in the way that basic spectroscopic information has been presented in the past, and an important outcome of this meeting, to be discussed later in this chapter, is to seek for standardisation of practice. In the third Theme, Detection and Sensing, attention continued to be focused on the major applications of biosensing and surveillance, largely for security purposes. No new fundamental developments were reported in medical sensing and imaging, although system availability has improved considerably and there are now real advances in applications of THz tools in pharmaceutical industry. There is mounting evidence that THz systems are capable of delivering unique information that is of value to security services, but that the stand-off range is likely to be no more than 10 m unless there are some significant breakthroughs in technology. One particularly impressive example of this, although one which is unlikely to be acceptable in civil law enforcement, is the concept of local generation and detection of THz radiation using high-powered near-infrared lasers. The importance of THz techniques in biological sensing is beginning to take a new direction, following the demonstration that THz spectroscopy is sensitive to the dielectric response of small ligands binding to biological molecules, which opens the way for the detection of toxins and pathogens. As a parallel development, DNA-based components can be envisaged that utilise THz-sensitive biomolecules. These, it is envisaged, might lead to such novel applications as: THz smart-dust, which would detect the presence of toxic agents; and ultimately, to THz control of processes at the molecular level. The Systems Theme provided overviews of the development of THz systems for near-field imaging, contraband detection, communications, and spectroscopy using both electronic and optical subsystems. Although some of these topics have been assessed previously from a systems perspective, the prospect of new active or passive components or subsystems, such as QCLs, multipliers or wire waveguides, indicate the need to revisit the topic. Near-field THz systems are now evolving at a considerable pace, and the first genuine microspectroscopic data is beginning to emerge. The extent to which such systems can become valuable to biology researchers does, however, remain a moot point as sample-handling techniques require effort, and a spatial resolution of less than 1 µm is essential. For stand-off surveillance systems, a consensus is now emerging that operational frequencies of order 600–800 GHz may represent the best route for efficient propagation and sensing; given the remarks made earlier, it would seem that all-electronic systems need to be pursued for this application.

344

M. CHAMBERLAIN

Communications systems were once seen as a major driver of THz research, offering considerable bandwidth advantages, albeit at the expense of range. The need for new components for such systems, especially for modulators and amplifiers was evident. Finally, the full exploitation of THz systems, for whatever purpose, does require significant future effort in clarifying health and safety matters and in the establishment of standard calibration protocols. 2. Topics for Future Development Towards the end of the meeting, the advances presented by the speakers were critically discussed and analysed. A list of key questions was then compiled which, it was thought, needed to be addressed before further significant progress could be made. A review of these is now given. 2.1. DEVICES

The topic inevitably is broken down into: sources, detectors, and other components. Historically, source development has demanded the greatest attention, and this pattern continues. The main requirement for sources, of course, is to satisfy the needs of mainstream applications such as imaging, spectroscopy, communications, and sensing. The balance of future effort between CW and pulsed source development was a key issue to debate in this context, and the overall view was that CW or quasi-CW approaches to source development were most likely to be fruitful. Present technological advance, as identified in the previous section, indicates that a new THz gap may be emerging between the highest frequency of operation that can be foreseen for electronic devices (even those deploying advanced materials technology), and the lowest frequency for optical (laser) sources operating at reasonably convenient temperatures. For electronics-based sources, there were deemed to be advantages in the development of modular or quasi-integrated units which could be simply “plugged” into multipliers and which would enable the systems developer to assess the value of novel sources at an early stage. The need to align some aspect of THz source development with silicon-based semiconductor technology was also thought to be worthy of consideration, and it was noted that a 1.2-THz transistor was under consideration for future commercial development. In terms of materials advances, it was evident that GaN-based sources were likely to realise great potential in the next few years: efforts should therefore be directed to overcoming processsing and other problems

TERAHERTZ SCIENCE AND TECHNOLOGY – THE FUTURE

345

so that this higher-energy phonon materials could be used for tunnel diode or other source designs. In general, it was thought that conventional vacuum electronics sources were unlikely to be incorporated into most systems, on account of the need to include heavy or bulky magnets, but that, there remained considerable opportunities at the boundary of nanotechnology and vacuum technology. For example, vacuum devices realised on a silicon chip, beneath a silicon nitride or silicon dioxide enclosure, may be worth investment especially if design options (such as reflex klystrons) that did not require a magnetic field could be pursued. Another example at the boundary of nanotechnology and vacuum electronics was that of carbon nanotubes, and already some indications were emerging from the literature that this offered a new route to the development of oscillator sources at THz frequencies. The outlook for superconducting sources was thought to be very limited indeed, as only limited power extraction was possible from an inductive device. In general, the developments in optical-based sources seemed to be satisfactory: improvements in the efficiency of laser-mixed sources (especially those deploying low-cost telecommunications components), and of the operational temperature for QCLS, were thought to be broadly on target. Despite that remark, there was an urgent need for novel device principles or breakthroughs to be sought, as source development (especially of convenient, cheap, and coherent sources) underpinned almost the whole territory of applicable THz science. Clearly, superlattice devices had made considerable advances in recent years but a large amount of unexplored territory remained to be covered before inversionless gain was seen. Finally, in this Theme, the topics of detector and amplifier development were considered. In general, detector technology appeared to be moving at a reasonable pace but arrays were an area needing further effort. This comment applied to arrays of microbolometers, where operation at room temperature had been suggested, and to arrays of heterodyne detectors which perhaps embodied QCL local oscillator sources. Amplifiers remained an almost unvisited area, and there was a pressing need for low-noise amplifiers and also for power amplifiers. QCLs may possibly offer a way forward for these components but, again, these devices are fraught with the same difficulties and concerns that attend QCL oscillators, namely the frequency and temperature range of operation.

346

M. CHAMBERLAIN

2.2. INTERACTIONS

Reaching a deeper physical understanding of the way in which THz radiation interacts with condensed matter will provide the basis for all future applications of THz sources and systems, together with background information necessary for the development of new devices. Although instrumentation advances have been significant since the last meeting in this series, considerable gaps remain in such areas as: theoretical modelling and simulation of data; interpretation of experimental data arising from studies of “mixed” condensed matter systems, including the extent to which scattering and interference effects can mislead the experimenter; and the limitations of given experimental methods. For example, advanced numerical computation techniques now exist which are beginning to predict the THz optical properties of simple molecules of biological significance; these very promising tools need further development so that they can, for example, offer guidance to the developer of functionalised biological sensor molecules which will respond to THz radiation. The question of how to interpret experimental THz data taken from a macroscopically inhomogeneous medium also raises some fascinating questions in the deployment of effective medium theory. Almost certainly, there is no simple model which can be universally applied to the range of such systems which are of current interest, for example, emulsions, nanocomposites, mixtures of drugs of abuse and cutting agents, or even human tissue. Scattering remains an issue of universal interest, and the development of rapid methods of assessing the importance of this effect in spectra is of clear importance to securing the continued reputation of the technique. Finally, the construction of new types of high-intensity THz sources provided strong impetus for new initiatives in non-linear spectroscopy, which was thought to be of value in studies of inhomogeneous systems. As an additional, but highly pertinent comment on the question of the interpretation of data, the meeting felt that a new protocol should be drawn up in order to present experimental data in a straightforward manner, so that the essential material physical properties at THz frequencies could be unambiguously stated. Such a protocol would of course, require a statement of the conditions under which the measurement was made (e.g. state of hydration, temperature, composition, etc.) and the particular experimental technique that was used (e.g. narrowband Backward Wave Oscillator, QCL, broadband spectroscopy, etc.). This protocol would enable careful intercomparison of data to be undertaken, and standardisation to be achieved for the first time in this part of the spectrum. The protocol was of especial importance for critical data, for example, that on explosives or toxic

TERAHERTZ SCIENCE AND TECHNOLOGY – THE FUTURE

347

substances. The preferred format for this type of data presentation would almost certainly be the real and imaginary parts of the relative permittivity at THz frequencies. The suggestion was also made that a new, secure database should be drawn up along these lines, and that international funding for this initiative should be sought. In terms of specific areas for future experimentation, there was an obvious need (as indicated by the content of the meeting) for spectroscopic studies of biological processes in order to realise novel bio-specific sensors and other devices. This topic is clearly a very large one, and guidance from theoretical modelling experts (as noted above) is required, together with closer interactions with the wider biochemical community. Phase transitions were also highlighted as an area that might profit from deeper study, together with the analysis of new types of explosive and bio-hazard substances that are now being deployed illegally. Evidently, progress in this dangerous but highly important area will require cooperation from law enforcement and security agencies if it is to be successful. The impact of THz science on health issues has perhaps, lessened a little recently. Nevertheless, there have been a number of useful instrumental advances, together with some strong developments in data visualisation and modelling. The need for safety studies at THz frequencies had emerged as a clearly important issue, perhaps as a prelude to the investigation of the therapeutic effects of THz radiation within a framework of clinical integrity. 2.3. DETECTION AND SENSING

Although a very large number of questions could be posed in this particular topic, the following only will be considered, which were selected because of their relevance to developments reported at the meeting. The questions were: will THz technology ever provide a viable route to stand-off detection of explosives; are there inherent advantages in the use of pulsed or CW approaches to sensing and detection; and to what extent can imaging and spectroscopy be combined with advantage? An essential task before attempting to respond to the first of these questions is to consider the user requirements, which differ considerably. For example, within the USA alone, three different requirements have been imposed: ranging from 400 m (for stand-off detection of toxins) to 30 cm for close surveillance of packages. In the UK, only one requirement, namely a distance of 100 m for the recognition of explosives secreted below garments, has been promulgated. Consideration of the current state

348

M. CHAMBERLAIN

of technology suggests that the target of 400 m is unattainable with present technology, using realistic link budget calculations that take into account effects of atmospheric propagation, scattering and diffusion of radiation within a target, and currently achievable signal to noise performance. However, detection at 100-m range could, just, be attainable with present systems at frequencies below 1 THz, although the amount of useful information gathered could be very small and the timescale unacceptably long. A range of 30 m or less is definitely now achievable, and identification of RDX spectra under dry weather conditions is a realistic proposition. The use of pulsed systems for very close inspection of mail packages appears to be the best solution, as this enables information from particular regions at depth within the sample to be analysed more readily, although new deconvolution procedures may be required if this technique is to be optimised. The possibility of revisiting the whole problem of stand-off detection was considered: for example, might there be advantages in deploying sensors remotely (smart dust), which would be activated by the presence of bio-hazardous gases and which could be attached to simple radio systems sending this information back to base? Another route, which may be advantageous in battlefield conditions, but is unlikely to gain favour within the sphere of civil policing, is to generate and to detect THz radiation in the vicinity of the target to be detected. This method requires four-wave mixing, in the ambient air close to the target, and for real success it would need electric fields exceeding 10 million v/m! Alternatively, an interesting solution (at least for systems deployable from small vehicles) may come from a return to the technology of the molecular gas laser, perhaps utilising microwave pumping, or by using a photonic tube fibre laser pumped with a miniaturised carbon dioxide laser operating at 10 µm. The question of the superiority of pulsed or CW techniques for sensing and detection were also considered. The key issue here is the dynamic range of both systems which, with current availability of components, are 50 and 100 dB (120 dB possibly achievable soon) for pulsed and CW systems respectively. Although information gained from both types of systems is, in principle, accessible via mathematical transformations, there are numerous additional questions to be considered such as cost and frequency ranges that are accessible. At this time, CW approaches appear to have lower cost than pulsed systems, but are perhaps not so well developed for use in the lower frequency part of the THz spectrum. In addition, the higher-peak power that is obtainable with pulsed systems does provide a

TERAHERTZ SCIENCE AND TECHNOLOGY – THE FUTURE

349

considerable advantage that has already been recognised, at lower frequencies, in radar systems. In order to gain the greatest benefit from whatever spectroscopic information can be gleaned at the longer stand-off distances noted above, and even to significantly enhance the value of data acquired at shorter ranges, the need to develop sensor fusion approaches was highlighted. In this method, a combination of millimetre wave and THz probes could be deployed to indicate the presence and approximate shape of suspect or anomalous items before spectroscopic analysis was undertaken. Although there are a number of issues surrounding this problem relating to the intellectual traditions of the differing communities of researchers, the key development requirement appears to be in signal processing: here, there has been a conspicuous absence of activity, which is perhaps surprising in view of the wealth of established techniques used in other frequency ranges. In the case of broadband pulsed detection and sensing systems, in particular, the time may now be ripe to develop strategies which reflect the additional information that is carried in the pulse, such as pulse width, delay, or the evidence of emission processes. Finally, considerable information may be present in the polarisation state of the pulses, especially when biological molecules with specific chirality are interrogated; this area is worthy of further study. 2.4. SYSTEMS

THz systems, operating at the borders of electronics and photonics, can clearly profit from the development of component concepts drawn from both fields. This viewpoint informed discussion of this Theme at several levels, together with consideration of the – new opportunities that are provided by more widely available high-powered sources. The following is a brief list of issues of topics that emerged for future consideration: •

• •

• •

What type of components and subsystems are required to improve the way in which THz radiation is guided, controlled, and directed into, and within, operational systems? What are the desirable integration steps that should now be undertaken to facilitate future systems development? What opportunities might arise for the realisation of THz systems operating at the high-frequency end of the THz range, or which utilise high-power sources? What are the standards requirements for systems? What new developments are now necessary to realise useful THz frequency communications systems?

350

M. CHAMBERLAIN

2.4.1. New component and subsystem requirements There was an emerging need to control and compensate for dispersion of THz pulses. One obvious application of such a component would be in a medical imaging system. In such a system, pulses directed into, and out of, the human body might be usefully analysed if the information they conveyed had not been lost through dispersion in a waveguide or other means of delivery. There was a general need to develop methods to deliver and analyse THz radiation in remote or inaccessible locations, with appropriate phase control if required. Non-reciprocal components, such as directional couplers and isolators, were clearly necessary to speed up system design: for both these, and the previously mentioned dispersion-compensator devices, it would be helpful to translate operational principles from other frequency regimes. The guidance and direction of THz radiation from within a device, or between devices, was now becoming important; already considerable efforts were being expended to ensure efficient outcoupling of THz radiation from QCLs using plasmonic guidance, and this principle might usefully be extended to other components. Plasmonics guidance was also likely to be of importance in the development of new near-field probes, for biological and medical research, and future efforts could profitably be focussed in this area. Finally, systems development at THz frequencies would benefit from the availability of s-design software that could handle, for example, the transport of THz pulses through a large number of components and subsystems. 2.4.2. Desirable integration steps Integration could usefully be considered at two levels: subsystems and systems. For the former, these included: integration of lasers and photomixers on the same chip; development of silicon-based subsystems, using CMOS sources to drive a multiplier; the modularisation of subsystems used in parametric oscillators; and the development of better interconnect. For systems integration, future efforts might concentrate on multi-modal approaches which would lead to systems with wider operational capability. For example, systems might be constructed which would operate in both pulsed and CW modes, or over different frequency ranges, or which might incorporate data fusion from a number of channels.

TERAHERTZ SCIENCE AND TECHNOLOGY – THE FUTURE

351

2.4.3. High-frequency and high-power systems Although there was an emerging consensus that stand-off detection systems would operate at frequencies below 1 THz, in order to minimise propagation and scattering losses, a consideration of the abundance of fingerprint spectroscopic lines, together with smaller available linewidths, suggested that there may be a need for the development of compact systems for molecular sensing, perhaps in forensic studies, in the 10–20 THz range. The growing availability of high-powered sources needed to be recognised by systems designers as they offered opportunities to explore non-linear effects. For example, if control of high-power pulse sequences at THz frequencies could be achieved, might the way then be opened for spin-echo and related techniques to be applied to a range of inhomogeneous samples, for example, to biological materials? Some efforts in this area therefore seem to be warranted. 2.4.4. Standards requirements The need for standards in systems operation had already been discussed in another part of this forum. In particular, there was a strong requirement to establish power and wavelength standards at THz frequencies, where none exist at present, and this was a task for national standards laboratories. In parallel with this effort, experimental samples should be examined in a variety of laboratories, using a range of systems, and the results circulated amongst the community to ensure uniformity. Health and safety standards were also essential, as it was no longer appropriate to extrapolate the infrared and microwave standards to a region where unexpected tissue responses may occur. 2.4.5. Communications systems Communications systems had been considered in some detail at this meeting. The major systems requirements were considered to be the development of: a modulator, that would be capable of modulating nearinfrared radiation at THz frequencies; a device that would be able to rapidly modulate the amplitude, and also the phase, of a THz signal; low-noise amplifiers, operating at ambient temperature at THz frequency; arrays of detectors; and a variety of smart antennas and reflective components. Systems of this type would offer high bandwidth transmission for the last mile, or (perhaps more realistically) in a secure manner within a building.

352

M. CHAMBERLAIN

3. Conclusion The field of THz technology continues to develop apace. New opportunities are arising from a variety of application areas, notably biological sensing and imaging, security, and surveillance. The field as a whole continues to translate device concepts from other frequency ranges, but also exploits the unique capabilities of electromagnetic radiation in this region. A number of impediments to progress have been overcome, but a considerable range of problems remain at the component, subsystem, and system levels. This meeting has reviewed progress across a broad front, with the exception of astronomy and earth-sensing, and has identified some areas where future efforts might be rewarded.

LIST OF SPEAKERS Dr. Roger Appleby QinetiQ, Bldg A Rm 302, Malvern Technology Centre, St. Andrews Road, Malvern, Worcestershire WR14 3PS, UK

Dr. Ian Gregory TeraView Ltd., Platinum Building, St. John’s Innovation Park, Cambridge, CB4 0WS, UK Prof. Peter Haring-Bolivar Institute of High Frequency and Quantum Electronics, University of Siegen, Hölderlinstr. 3, D-57068 Siegen, GERMANY

Prof. Martyn Chamberlain Department of Physics, Science Laboratories, University of Durham, Durham DH1 3LE, UK

Prof. Kodo Kawase Engineering Department, Nagoya University, Chikusa, Nagoya, 464-8603, JAPAN

Dr. Howard Cummins HMGCC, Hanslope Park, Hanslope, Milton Keynes, MK19 7BH, UK

Prof. Martin Koch Institut für Hochfrequenztechnik, Technische Universität Braunschweig, Schleinitzstr. 22, 38106 Braunschweig, GERMANY

Prof. Giles Davies School of Electronic Engineering, University of Leeds, Leeds LS2 9JT, UK Dr. Heribert Eisele School of Electronic Engineering, University of Leeds, Leeds LS2 9JT, UK

Prof. Arunas Krotkus Semiconductor Physics Institute, A. Gostauto 11, 2600, Vilnius, LITHUANIA

Prof. Jerome Faist University of Neuchatel, Rue A.L. Breguet 1, 2000 Neuchatel, NE, SWITZERLAND

Dr. Andrea Markelz University at Buffalo, 239 Fronczak Hall, Buffalo, NY 14260, USA

Dr. Richard Green NEST CNR-INFM & Scuola Normale Superiore, Piazza dei Cavalieri 7, 56126 Pisa, ITALY

Prof. Robert E. Miles School of Electronic Engineering, University of Leeds, Leeds LS2 9JT, UK

353

354

LIST OF SPEAKERS

Dr. Daniel Mittleman Rice University, ECE Dept., MS-366, 6100 Main St. Houston, TX 77005, USA Dr. Paul Planken University of Technology Delft, Faculty of Applied Sciences, Dept. of Imaging Science and Technology, Lorentzweg 1, 2628 CJ Delft, THE NETHERLANDS Prof. Hartmut Roskos Physikalisches Institut der Johann Wolfgang Goethe-Universitaet, Frankfurt am Main, Robert-Mayer-Str. 2-4, D-60054 Frankfurt/Main, GERMANY Prof. Carlo Sirtori Matériaux et Phénomènes Quantiques, Université Denis Diderot - Paris 7, 75005, Paris, FRANCE Prof. Jan Stake Chalmers University of Technology, Department of Microtechnology and Nanoscience, Kemivägen 9, SE-412 96 Göteborg, SWEDEN

Dr. Bill Truscott School of Electrical and Electronic Engineering, University of Manchester, Manchester, M60 1QD, UK Prof. Peter Uhd-Jepsen Nanophotonics, COM Research Center, Technical University of Denmark, DK-2800 Kgs. Lyngby, DENMARK Prof. Daniel van der Wiede Department of Electrical & Computer Engineering, University of Wisconsin, 1415 Engineering Dr, Madison WI 53711, USA Dr. Dwight Woolard U.S. Army Research Laboratory, Army Research Office, RTP, NC 27709, USA Prof. X-Cheng Zhang Physics Department, Rensselaer Polytechnic Institute, 1108th Street, Troy, NY 12180, USA

LIST OF PARTICIPANTS Dr. K. N. Alekseev Department of Physical Sciences, FI-90014 University of Oulu, FINLAND

Dr. Andrei Malcoci Universitatea Politehnica din Timisoara, Facultatea de Electronica si Telecomunicatii, Bd. Vasile Parvan, nr. 2 300223, Timisoara, ROMANIA

Naomi Alexander Alfa Imaging, S.A., 28006 Madrid, SPAIN

Dr. Vitaly Malevich Institute of Physics, National Academy of Sciences of Belarus, 68 Nezalezhnasti Ave., 220072 Minsk, BELARUS

Prof. Dr. René Beigang Department of Physics, Kaiserslautern Technical University, Erwin-Schroedinger-Str., 67663 Kaiserslautern, GERMANY

Prof. Patrick Mounaix Centre de Physique Moléculaire Optique et Hertzienne, Université Bordeaux I - CNRS, 351 cours de la Liberation, 33405 Talence CEDEX, FRANCE

Carlos Callejero Alfa Imaging, S.A., 28006 Madrid, SPAIN Dr. Farid Dowla Advanced Signal Processing Group, Lawrence Livermore National Laboratory, 7000 East Avenue, Livermore, CA 94550, USA

Dr. Daryoosh Saeedkia University of Waterloo, 200 University Ave. West, Waterloo, Ontario, CANADA N2L 3G1

Dr. Franco Fiore NATO C3 Agency, Command and Control Systems Division, Surveillance & Reconnaissance Resource Centre, Oude Waalsdorperweg 61 2597AK The Hague, THE NETHERLANDS

Prof. Alexander Shkurinov Department of Physics and International Laser Centre, M V. Lomonosov Moscow State University, Leninskie gory, Moscow 119899, RUSSIA

Dr. Vladimir Gayvoronsky Institute of Physics NASU, pr. Nauki 46, Kiev 03028, UKRAINE

Dr. Josip Vukusic Nanofabrication Laboratory, Department of Microtechnology and Nanoscience, Chalmers University of Technology, Kemivägen 9, SE-412 96 Göteborg, SWEDEN

Prof. Boris Knyazev Budker Institute of Nuclear Physics, Russian Academy of Science, 630090 Novosibirsk, RUSSIA 355

Speakers, Participants and Partners NATO ARW Spiez, Switzerland July 2006

357

INDEX absorption coefficients 110 absorption due to barrier materials 215 absorption lines 189 absorption parameters 111 Active (Radar) 214 active imaging 222 adenine 99 alcohol 154 amorphous condensed-phase systems 148 amplifier facet 51 analytical spectroscopy and gas sensing 167 angular divergence 50 antenna gain 331 Anthrax simulant 210 atmospheric absorption 139, 188, 215 atmospheric and free-space damping 329 atmospheric attenuation 206 atomic rearrangements 92 avalanche breakdown 71

biotin-avidin binding 124 bioweapons 139 Bloch effect 73 Bloch frequency 31 Bloch gain 31 Bloch-gain lasers 34 borosilicate glasses 110 bulk electro-optic (EO) rectification 92 bulk water 124 BWO 70, 245 cadmium mercury telluride 14 capacitance-voltage characteristic 18 carbon nanotubes 345 CF2 twisting vibration 94 CF2 wagging vibration 95 chalcogenide glasses 109 Challenges in the Security Domain 223 chemical identification 242 chemical-biological mail threat 209 chemically-specific absorption spectra 241 clothing 225 coherence length 178 coherent tunable THz waves 243 concealed ceramic knife blade 176 continuous-wave (CW) 137 contrast model 230 cost sensitivity 169 crystalline organic substances 242 cw-THz homodyne detection 173

backward-wave oscillators 18 ballistic movement 12 ballistic photocurrent surge effect 12 bandwidth in wireless shortrange communication 325 barrier layers 20 billiard channels 332 binding state of DNA 342 bioactive proteins 299 biochip sensing 315 biomedical imaging 254 biosensing 254 359

360

cw-THz photomixer systems 170 cytidine 98 data filtering techniques 187 dc-to-RF conversion efficiencies 70 Debye velocity of sound 108 Debye Waller factors 132 decomposition kinetics 291 Delrin 109 density functional perturbation theory 147 DFB devices based on etched gratings 44 dielectric fibers 56 diffuse reflectance spectroscopy 257 diffuse reflection 190 diffuse scatterer 196 dioxane-water mixture 159 directional couplers 350 discrimination 209 discrimination parameters 179 distributed feedback (DFB) gratings 43 DNA 293 domain formation 34 dynamic cut-off frequency 22 dynamic range 348 elastance 22 electrical-field-induced optical rectification 4 electro thermal HBV device model 25 electron-ion interactions 151 end mirrors 43 environmental radiometric temperature 232 ethanol-water mixtures 160 excitonic absorption lines 34 explosives 139, 227

INDEX

Far Infrared Polarizing Michelson Interferometer 227 fibre optic coupling 175 figure-of merit for varactors 19 free-running oscillators 81 free-space transport of the terahertz beam 56 frequency gaps 193 frequency multiplier 18 frequency multiplier chains 70 Fresnel transmission term 130 full field screening 34 fundamental device limitations 71 future wireless communication 336 GaAs LO phonon resonance 97 gas lasers 168 germanium 15 glass transition temperature 113 group velocity 60 group velocity dispersion 61 guanine 101 Hankel function 59 harmonic balance 82 harmonic generation 19 HBV device model 25 HBV multiplier circuits 24 HBV quintupler 24 HBV tripler 24 hen egg white lysozyme (HEWL) 126 hertzian dipole antenna 127 heterodyne receivers 225 Heterostructure Barrier Varactor 18 high density polyethylene (HDPE) 109 high electron mobility transistors (HEMTs) 331

INDEX

high power millimeter wave tripler 25 higher harmonic multiplication factors 18 higher-power sources 138 Home Made explosives (HMX) 218 homodyne detection 170 horizontally polarized THz pulses 57 hydrated protein dynamics 125 hydrogen-bonded networks 101 i.f.bandwidth 195 illuminating chamber 233 Imaging at 100GHz 233 imaging electronic components 208 IMPATT 70 Improvised Explosive Device (IED) 217 InAs 8 indicative and background images 199 indoor imaging 232 infrared-active lattice vibrations 7 injection seeded TPG (is-TPG) 242 integration 349 intermolecular bonds 240 intermolecular hydrogen bonds 152 intermolecular vibrational modes 252 intervalley scattering 12 intramolecular modes 148 ionic dangling bonds 112 isolators 350 K-K relationship 269 Kleinman-Bylander 151

361

Kramers-Kronig 293 Kramers-Kronig (K-K) transform 269 Kroemer’s rule 33 Kubelka-Munk 258 laboratory demonstrator 174 lasing spectrum 43 lateral photocurrent component 6 librational modes 140 LiTaO3 pyroelectric detector 246 LO-phonon emission 38 low-frequency vibrational spectra 92 low-loss polymers 56 lyophilized powders 126 maltose 97 Master Oscillator Power Amplifier (MOPA) 50 medical imaging and cancer characterisation 178 metal tubes 56 metal waveguide 57 metallic losses 142 millimeter wave image 206 modal dispersion 142 modular approach 183 molecular transitions 242 multi-barrier structures 22 multiple reflection artifacts 124 multiple targets 140 negative differential resistance 70 new protocol 346 noise level 94 non ionizing 206 nonlinear crystal 242 nonlinear spectroscopy 345 nonlinear transmission line (NLTL) 140

362

non-periodic molecular clusters 150 non-reciprocal components 350 object movement 198 omnidirectional mirror 335 optical delay retro-reflector 174 optical flats 109 optical rectification 4 optical semiconductor switches 168 optimum slit width 47 outdoor imaging 229 paper 120 parametric amplification 19 Passive (Radiometer) 214 passive imaging 238 penetration in common building materials 212 Perspex 109 pharmaceutical research 251 phase calibration 198 phase with no long-range order 154 phase-matched difference frequency mixing 95 phonon processes 108 Photo-Dember effect 5 piezoelectric fibre modulator system 175 plane-wave density functional method 151 plasma oscillation 7 plasmonics 342 plastic explosive 177 plastic ribbons 56 polaritons 242 polarity reversal 58 polarized hot-electron photoluminescence 6 policy of no consumables 168

INDEX

polymerisation front model 118 polymerisation in SU8 118 polymorphs 169, 254 power budget 188, 193 Principal Component Analysis 189 protein-ligand binding 123 protein-ligand interaction 125 protein-protein binding 123 PTFE 109 public multimedia station 326 pulse reshaping 56 pump-probe 33 purine 99 Q-switched Nd:YAG laser 243 quanta of terahertz energy 188 quasi-integrated units 344 quasi-optical propagation 138 QWITT 73 radially polarized mode 58 radiative transitions of electrons between sub-bands 42 radiometers 219 radiometric properties 226 rapid parallel assays 124 rate of tuning 47 Rayleigh’s equation 234 RDX plastic explosive 180 real world situations 188 receiver thermal sensitivity 234 reflex klystrons 345 relaxation times 79 resonant tunneling diodes 72 RNA 293 scattering 214, 342 Schottky Diode 18 second order gratings 48 secondary reflection 110 seeder wavelength 244

INDEX

self-referenced reflection THzTDS 154 semiconductor diode lasers 169 short-range THz communication 327 silica glass 111 single interband laser diode 328 single mode operation 43 small-signal conductivity 33 smart antennas 333 solid-state reactions of drug substances 276 Sommerfeld wire wave 59 space group symmetry 147 spatial resolution 191 spectral lines 192 spectral purity 43 spectral resolution 172 spectroscopic signatures 140 spectroscopy of explosives 255 spontaneous polarization 78 standing wave effects 140 standoff detection 142 stopband 46 SU8 109 sucrose crystal 153 superlattice electronic devices (SLED) 73 surface field effect 5 surface field generation 92 surface plasmon gratings 46 surface plasmon polariton (SPP) 61 surface selection rule 271 symmetric varactors 19 system thermal sensitivity 235 systems of high symmetry 149 tablet imaging 168 target signature 226 Technical Security Counter Measures (TSCM) 207

363

TEDs 70 temporal THz waveform 96 terahertz dielectric spectroscopy 124 Terahertz Gap 342 threat detection at range 207 threat detection close in 207 three barrier HBV material 22 thymine 101 THz antennas 332 THz communications 138 THz differential time-domain spectroscopy (THz-DTDS) 304 THz endoscope 65 THz modulator 330 THz phonon 242 THz polarization 4 THz pulses 94 THz pulses on a metal wire 57 THz quantum cascade 41 THz refractive indices 110 THz smart-dust 343 THz spectral database for explosive sensing 264 THz spectral fingerprints 263 THz spectroscopy for pharmaceutical applications 275 THz spectroscopy of proteins 297 THz spectroscopy of small biomolecules 293 THz transients 32 THz wave parametric generators (TPG) 242 THz wave parametric oscillator (TPO) 242 THz-beam amplification 33 Ti:sapphire laser 94 tomographic techniques 187 TPX 109 transient current 92

364

transit-time diodes 76 transverse wave fields 242 traveling-wave tubes (TWTs) 139 two-stage identification process 189 ultra-broadbandwidth THz radiation 93 ultrafast switching of photoconductive (PC) emitters 92 undercoupled DFB 46 unusual blue-shift 163 unusual dispersive behavior 65 user-selectable beat frequency 139

INDEX

vacuum-technologies 345 varactor 19 vertical emission spectra 49 vibrational dynamics 92 waveguide losses 50 Willenberg formula 36 wireless IR links 327 ZnTe EO detector 92 zone-centre phonon modes 151