Handbook of Infrared Detection Technologies
Handbook of Infrared Detection Technologies
This Page Intentionally Left...
391 downloads
3092 Views
35MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Handbook of Infrared Detection Technologies
Handbook of Infrared Detection Technologies
This Page Intentionally Left Blank
Handbook of Infrared Detection Technologies
Edited by Mohamed Henini and Manijeh Razeghi
ELSEVIER
UK USA JAPAN
Elsevier Science Ltd, The Boulevard, Langford Lane, Kidlington, Oxford OX 5 1GB, UK Elsevier Science Inc, 360 Park Avenue South, New York. NY 10010-1710, USA Elsevier Science Japan, Tsunashima Building Annex, 3-20-12 Yushima, Bunkyo-ku, Tokyo 113. Japan Copyright ~ 2002 Elsevier Science Ltd. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means: electronic. electrostatic, magnetic tape, mechanical, photocopying, recording or otherwise, without prior permission in writing from the publishers.
British Library Cataloguing in Publication Data Handbook of infrared technologies 1 .Infrared technology I.Henini, Mohamed II.Razeghi, M(Manijeh) technologies 621.3'62
III.Infra-red
ISBN 1 85617 388 7
Library of Congress Cataloging-in-Publication Data Handbook of infrared technologies / edited by Mohamed Henini and Mahijeh Razeghi. p. cm Includes bibliographical references and index. ISBN 1-85617-388-7 (hardcover) 1. Infrared technology-Handbooks, manuals, etc. I. Henini, Mohamed. II. Razeghi, M. TA 1 5 7 0 . H 3 5 8 2002 621.36'2-dc21
2002040767
No responsibility is assumed by the Publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Published by Elsevier Advanced Technology, PO Box 15() Kidlington Oxford OX5 1AS. UK Tel.: +44(0) 1865 8 4 3 0 0 0 Fax: +44(0) 1865 843971 Typeset by Variorum Publishing Limited, Lancaster and Rugby Printed and bound in Great Britain by Biddles Limited, Guildford and King's Lynn
Contents
List of contributors
xiii
Chapter 1
Introduction
Chapter 2
Comparison of photon and thermal detector performance 2.1 Introduction 5 2.2 Fundamental limits to infrared detector performance 6 2.2.1 Photon detectors 9 2.2.2 Thermal detectors 18 2.2.3 Comparison of the fundamental limits of photon and thermal detectors 21 2.3 Focal plane array performance 23 2.4 FPAs of photon detectors 26 2.4.1 InSb photodiodes 30 2.4.2 HgCdTe photodiodes 33 2.4.3 Photoemissive PtSi Schottky-barrier detectors 36 2.4.4 Extrinsic photoconductors 36 2.4.5 GaAs/A1GaAs OWIPs 38 2.4.6 OWIP versus HgCdTe in the LWIR spectral region 39 2.5 Dual-band FPAs 52 2.5.1 Dual-band HgCdTe 53 2.5.2 Dual-band QWIPs 54 2.6 FPAs ofthermal detectors 58 2.6.1 Micromachined silicon bolometers 58 2.6.2 Pyroelectric arrays 61 2.6.3 Thermoelectric arrays 64 2.6.4 Status and trends of uncooled arrays 65 2.7 Conclusions 67 Appendix 70 References 75
vi
Handbookof Infrared Detection Technologies
Chapter 3
GaAs/AIGaAs based quantum well infrared photodetector focal plane arrays
3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9
Chapter 4
Introduction Detectivity D* comparison Effect of nonuniformity 640 • 512 pixel long-wavelength portable OWIP camera 640• 486 long-wavelength dual-band imaging camera 640 • 512 pixel broad-band OWlP imaging camera 640 x 512 spatially separated four-band QWIP focal plane array OWIPs for low background and low temperature operation Summary Acknowledgements References
83 89 91 92 99 102
110 113 115 117 117
GalnAs(P) based QWIPs on GaAs, InP and Si substrates for focal plane arrays 121 4.1 Introduction
4.2
4.3
4.4
4.5
4.6
4.1.1 Overview of infrared detector 4.1.2 Quantum well infrared photodetector 4.1.3 State-of-the-art Fundamentals of OWlP 4.2.1 Intersubband absorption 4.2.2 OWIP parameters 4.2.3 Comparison of n-type and p-type QWIPs 4.2.4 Growth. fabrication and device characterization of a single QWIP device Fabrication of infrared FPA 4.3.1 Infrared FPA fabrication steps 4.3.2 Indium solder bump fabrication steps 4.3.3 ROIC for infrared FPA p-type OWIPs 4.4.1 p-type MWIR QWIPs 4.4.2 p-type LWlR QWIPs n-type OWIPs 4.5.1 n-type LWIR QWIPs 4.5.2 n-type VLWIR OWIPs 4.5.3 Multi-color OWIPs Low cost QWIP FPA integrated with Si substrate 4.6.1 Overview of OWIPs on Si 4.6.2 Growth of GaInAs/InP QWIP-on-Si 4.6.3 Detector performance of GaInAs/InP OWIP-on-Si How to fabricate a monolithic integrated FPA 4.6.4 with Si substrate
121 123 125 126 126 130 133 134 135 135 139 140
141 142 144 144 144 146 147 147 147 148 148 151
Contents
4.7 4.8
New approaches of OWIP Conclusions References
vii
151 153 154
Chapter 5
InAs/(Galn)Sb superlattices: a promising material system for infrared detection 5. ] Introduction 159 5.2 Materials properties 159 5.2.1 BandstructureoflnAs/(GaIn)Sbsuperlattices 159 5.2.2 X-ray characterization 163 5.2.3 Interfaces 165 5.2.4 Sample homogeneity 16 7 5.2.5 Residual doping 16 7 5.3 Superlattice photodiodes 169 5.3.1 Diode structure 169 5.3.2 Diode processing 172 5.3.3 Photo response 172 5.3.4 I-V measurements 173 5.3.5 C-V measurements 181 5.3.6 Noise measurements 183 5.4 Summary and outlook 18 7 References 18 7
Chapter 6
GaSb/InAs supperlattices for infrared FPAs 6.1 Type-II heterostructures 6.1.1 Historical review 6.1.2 Definition oftype-II band alignment 6.1.3 Features oftype-II band alignment and their applications 6.2 Type-II infrared detectors 6.2.1 Principle of operation 6.2.2 Band structure of type-II superlattices 6.2.3 Optical absorption in type-II superlattices 6.2.4 Modeling and simulation of type-II superlattices 6.3 Experimental results from type-II photoconductors 6.3.1 Uncooled type-II photoconductors in the X=8-12 l~m range 6.3.2 Cooled type-II photoconductors for )v > 2013m 6.4 Experimental results from type-II photodiodes 6.4.1 Uncooled type-II photodiodes in the )v=8-12 ~m range 6.4.2 Cooled type-II photodiodes in the )v > 14 l.tm range 6.5 Future work References
191 191 191 192 193 193 193 194 19 7 201 201 208 211 211 215 227 231
viii
Handbook of Infrared Detection Technologies
Chapter 7
Chapter 8
MCT properties, growth methods and characterization 7.1 Preface 7.2 Introduction 7.2.1 Brief history 7.3 MCT Characteristics and material properties 7.3.1 Composition and crystal structure 7.3.2 Bandgap 7.3.3 Intrinsic carrier concentration 7.3.4 Doping and impurities 7.3.5 Carrier mobility 7.3.6 Carrier lifetime 7.3.7 Defects 7.4 MCT crystal gowth methods 7.4.1 Phase diagrams 7.4.2 Bulk growth 7.4.3 Epitaxial growth 7.5 Material characterization methods 7.5.1 Material composition 7.5.2 Measurements of carrier concentration and mobility 7.6 Summary References HgCdTe 2D arrays- technology and performance limits 8.1 Introduction 8.1.1 Historical perspective 8.2 Applications and sensor design 8.3 Comparison of HgCdTe with other 2D array materials 8.4 Multiplexers for HgCdTe 2D arrays 8.4.1 Photocurrent injection techniques 8.4.2 Scanning architectures 8.4.3 Future trends 8.5 Theoretical foundations for HgCdTe array technology 8.5.1 Thermal diffusion currents in HgCdTe 8.5.2 Leakage currents 8.5.3 Photocurrent and quantum efficiency 8.6 Technology ofHgCdTe photovoltaic devices 8.6.1 Materials growth technology 8.6.2 Junction forming techniques in homojunction arrays 8.6.3 Device structures 8.7 Measurements and figures of merit for 2D arrays 8.7.1 NETD - theoretical calculation 8.7.2 NETD- experimental measurement 8.7.3 Relationship of NETD with other figures of merit
233 233 234 235 235 236 238 239 240 241 245 245 246 247 248 256 256 260 260 260
269 270 271 274 276 277 278 279 279 280 281 282 282 284 287 288 292 293 294 294
Contents
8.8 8.9
8.10 8.11
8.12
HgCdTe HgCdTe 8.9.1 8.9.2
2D arrays for the 3-5 lum (MW) band 2D arrays for the 8-12 ~tm (LW) band Array design issues Introduction to performance limitations in LW arrays 8.9.3 Cause of defective elements in HgCdTe 2D arrays HgCdTe 2D arrays for the 1-3 ~m (SW) band Towards GEN III detectors 8.11.1 Two-colour array technology 8.11.2 Higher operating temperature (HOT) device structures 8.11.3 Retina level processing Conclusion and future trends Acknowledgements References
ix
296 296 296 297 298 299 301 301 303 304 304 305 305
Chapter 9. Status of HgCdTe MBE technology 9.1 Introduction 9.2 HgCdTe MBE equipment and process sensors 9.2.1 Vacuum equipment and sources 9.2.2 HgCdTe MBE process sensors 9.3 HgCdTe MBE growth process 9.3.1 Substrate preparation 9.3.2 Growth conditions 9.3.3 Defects 9.3.4 Doping 9.4 Device applications 9.4.1 Multispectral HgCdTe infrared detectors 9.4.2 Near-infrared avalanche photodiodes High-performance MWIR detectors 9.4.3 9.4.4 Large-format arrays on silicon substrates Acknowledgements References
309 310 310 312 318 318 320 320 322 324 324 331 33 7 341 348 348
Chapter 10 Silicon infrared focal plane arrays 10.1 Introduction 10.2 CooledFPAs 10.2.1 Schottky-barrier FPAs 10.2.2 Heterojunction internal photoemission FPAs 10.3 UncooledFPAs 10.3.1 Silicon On Insulator (SOI) diode FPAs 10.3.2 Si-based resistance bolometer FPAs 10.3.3 Thermopile FPAs 10.4 Summary References
353 354 354 3 71 374 374 379 383 386 387
x
Handbookof Infrared Detection Technologies
Chapter 11 Infrared silicon/germanium detectors 11.1 Introduction 11.2. Near Infrared detectors 11.2.1 General operation principle 11.2.2 Detector growth and fabrication 11.2.3 Results and discussion 11.3 Mid-and long-wavelength SiGe IR detectors 11.3.1 Introduction 11.3.2 Principle of operation of HIP detectors 11.3.3 Growth and material characterization 11.3.4 Experimental results and discussion 11.3.5 Calculation of optical properties of SiGe HIP detectors 11.3.6 R6sum6 and outlook for SiGe MWIR detectors Acknowledgements References Chapter 12 PolySiGe uncooled microbolometers for thermal IR detection 12.1 Introduction 12.1.1 Uncooled resistive microbolometers 12.1.2 Microbolometer terminology 12.1.3 Microbolometer process options 12.2 Structural, thermal and electrical properties ofpolySiGe 12.2.1 Deposition of polySiGe 12.2.2 Structural properties 12.2.3 Thermal properties 12.2.4 Electrical properties 12.2.5 High-temperature vs. low-temperature polySiGe 12.3 PolySiGe bolometer pixels 12.3.1 Process development 12.3.2 Absorber comparison and trade-offs 12.3.3 Pixel optimization 12.3.4 Vapor HF processing 12.3.5 Stiffness enhancement techniques 12.4 Readout and system development 12.4.1 Introduction 12.4.2 Readout of polySiGe bolometer arrays 12.5 Zero-level vacuum packaging 12.5.1 Introduction 12.5.2 Indent-Reflow Sealing using metal solder 12.5.3 Zero-levelpackagingusingBCB 12.5.4 Hermeticitytestingusingmicrobolometers 12.6 Conclusions and outlook Acknowledgements References
393 398 398 403 408 411 411 412 413 421 438 444 445 445
449 449 451 452 453 453 454 455 455 457 457 457 459 460 462 463 467 467 468 472 472 473 475 475 476 477 477
Contents
xi
Chapter 13 Fundamentals of spin filtering in ferromagnetic metals with application to spin sensors
13.1 13.2
13.3 13.4
13.5
Index
Introduction Theoretical IMFP variation 13.2.1 The simplest m o d e l - mathematical bases of the calculation 13.2.2 Amore complete treatment 13.2.3 An intuitive derivation 13.2.4 Comparison with the Sch6nhense and Siegmann model Experimental study of A c~ Spin precession and spin filters 13.4.1 Density-operator formalism 13.4.2 Electron transmission through ferromagnetic bilayers 13.4.3 The bilayer with collinear magnetizations 13.4.4 The bilayer with perpendicular magnetizations Discussion and conclusion Acknowledgements References
481 483 483 488 492 495 495 498 499 501 502 503 504 506 506 509
This Page Intentionally Left Blank
List of Contributors
Ian M. BAKER, BAE Systems Infrared Ltd., P.O. Box 217, Southampton, Hampshire SO 15 0EG, UK S. V. BANDARA, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91 ] 09, USA L. BURKLE, Fraunhofer Institut f/ir Angewandte Festk6rperphysik, Tullastrasse 72, D-79108 Freiburg, Germany Henri-Jean DROUHIN, Laboratoire de Physique de la Mati~re Condens~e (UMR 7643-CNRS), Ecole Polytechnique, 9 ] 128 Palaiseau cedex, France F. FUCHS, Fraunhofer Institut fiir Angewandte Festk6rperphysik, Tullastrasse 72, D-79108 Freiburg, Germany S.D. GUNAPALA, Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109, USA
M. HENINI, Department of Physics and Astronomy, University of Nottingham, Nottingham, UK Chris van HOOF, IMEC Kapeldreef 75, B-30()1 Haverlee, Belgium and ESATINSYS Department, University of Leuven, Belgium J. E. JENSEN, HRL Laboratories, 3011 Malibu Canyon Road, Malibu, CA 90265. USA
J. JIANG, Center for Quantum Devices, Electrical and Computer Engineering Department, Northwestern University, Evanston, Illinois 60208, USA
xiv Handbook of Infrared Detection Technolo,qies Masafumi KIMATA, Senior Technology Department, Advanced Technology, R&D Center, Mitsubishi Electric Corporation, 8-1-1, Tsukaguchi-Honmachi, Amagasaki, Hyogo 661-8661. Japan Randolph E. LONGSHORE, Raytheon Missile Systems, P.O. Box 1137, MS 8 4 0 / 7 Tuscon, AZ 85734, USA Terry de LYON, HRL Laboratories. 3 O l l Malibu Canyon Road, Malibu, CA 90265, USA H. MOHSENI, Center for Quantum Devices. Electrical and Computer Engineering Department, Northwestern University. Evanston. Illinois 60208. USA Piet de MOOR, IMEC, Kapeldreef 7 5. B- 3()() 1 Heverlee, Belgium Hartmut PRESTING, DaimlerChrysler Research (REM/C), Dep. FT2/H, WilhelmRunge Strasse 11, D-89081 ULM, Germany R. D. RAJAVEL, HRL Laboratories, 3011 Malibu Canyon Road. Malibu, CA 90265, USA
Manijeh RAZEGHI, Center for Quantum Devices. Electrical and Computer Engineering Department, Northwestern University. Evanston. Illinois 6()2()8, USA Antoni ROGALSKI, Institute of Applied Physics, Military University of Technology, 2 Kaliskiego Str.. ()()-908 Warsaw. Poland J. A. ROTH, HRL Laboratories, 3011 Malibu Canyon Road, Malibu, CA 90265. USA
Chapter 1
Introduction M. R a z e g h i a n d M. H e n i n i
Nature has provided numerous examples of efficient detection systems. Almost all types of life, from bacteria, to plants, to h u m a n beings, have evolved some type of optoelectronic detection system for perceiving the world around them. These systems have had millions of years to develop, and demonstrate a seamless integration of optoelectronics with biological systems. The jewel beetle (Melanophila acuminata) thrives on the remnants of forest fires. Its larva feed on the dead wood, which gives evolutionary incentive for the beetle to find dead wood before other species. Towards this end, the beetles have developed an infrared detection system which allows them to sense a 10 hectare forest fire from up to 12 km away. As shown in Figure 1.1, a pit organ, called a sensilla, is located on either side of the beetle's thorax, which allows both intensity and directional information to be obtained. Absorption of infrared (2.4-4 lxm wavelength) light triggers a mechanical expansion which triggers nerve impulses. Obviously, this system must be small and easy to use. Further, as
Figure 1.1 The jewel beetle and its infrared sensor.
2
Handbookof Infrared Detection Technologies
a beetle does not have a large built-in power supply or cryogen, the system must be power efficient and be uncooled. Our eyes are also excellent examples. Nature has provided a multi-spectral detection system based on microscopic variation in detector design. These differentiated detector cells add another dimension to the versatility of the eye. With a broadband detector, there is no way to differentiate between the intensity of a source and its emissivity at different wavelengths. This is akin to trying to pick out a matching wardrobe with a black and white camera. Multispectral systems allow separate waveband analysis of objects, which allows faster and more accurate identification to be made. On an evolutionary perspective, this ability allows more efficient target identification, allowing faster response to a potentially hazardous situation. The goal of science is to enhance our senses and better understand the universe around us. Infrared detectors broaden our vision into the realm of heat, allowing remote sensing of an object's temperature. This has had a dramatic impact on how we perceive our environment, and has led to many types of thermal imaging, including night vision, infrared astronomy, medical diagnostics, and failure analysis. These newfound abilities have spurred the development of m a n y new systems, as shown in Figure 1.2. Infrared detectors have seen a remarkable surge in interest over the past several decades. This is thanks in part to the successful development of highperformance devices which have become the core of all the infrared systems listed above. The natural progression of these systems is a multispectral, uncooled, infrared camera, which can, by itself, address most of these applications. As in nature, a good system should be flexible, power efficient, lightweight, and easy to use. While we cannot expect to match the sophistication of natural systems, we can be inspired by them.
Figure 1.2 Examples of mainstream thermal imaging systems.
Introdl~ction
3
One inspiration involves the exploitation of quantum size effects for higher efficiency and added functionality. Most infrared photon detectors have a limited photocarrier lifetime and peak detection wavelength that is fixed by the bandgap of the material. Without changing the chemical composition of the material, patterning on an atomic scale can allow an increase in carrier lifetime and tuning of the peak detection wavelength. This type of effect has already been demonstrated in the form of the type-II InAs/GaSb semiconductor detector. Used in another way, similar to the eye, microscopic alterations can be made to the lateral size of individual detectors to demonstrate multispectral sensitivity in a single focal plane array. The purpose of this book is to present current methods and future directions in infrared detection. By bringing together experts in physics, material science, fabrication technology, and application, we will develop a well-rounded view of how far we have progressed towards the goal of an integrated, versatile, infrared detection system.
This Page Intentionally Left Blank
Chapter 2
Comparison of photon and thermal detector performance A. Rogalski
2.1 Introduction At present, HgCdTe is the most widely used variable gap semiconductor for infrared (IR) photodetectors. Over the last forty years it has successfully fought off major challenges from extrinsic silicon and lead-tin telluride devices, but despite that it has more competitors today than ever before. These include Schottky barriers on silicon, SiGe heterojunctions, A1GaAs multiple quantum wells, GaInSb strain layer superlattices, high temperature superconductors and especially two types of thermal detectors: pyroelectric detectors and silicon bolometers. It is interesting, however, that none of these competitors can compete in terms of fundamental properties. They may promise to be more manufacturable, but never to provide higher performance or, with the exception of thermal detectors, to operate at higher or even comparable temperatures. The main motivations to replace HgCdTe, are technological problems of this material. One of them is a weak Hg-Te bond, which results in bulk, surface and interface instabilities. Uniformity and yield are still issues. The slow progress in the development of large photovoltaic HgCdTe infrared imaging arrays and the rapid achievements of novel semiconductor heterostructure systems have made it more difficult to predict what types of arrays will be readily available for future systems applications. For spaceborne surveillance systems, low background IR seeker/tracker systems, reliable and affordable sensors with long life are needed which can function effectively at temperatures higher than the 2 0 - 3 0 K currently required by bulk photon detectors. The only alternative to HgCdTe that had been available so far was extrinsic Si, which operates at much lower temperatures where a problematic three-stage cryocooler would be required. Improvement in surveillance sensors and interceptor seekers requires large area size, highly uniform and multicolour (or multispectral) IR focal plane arrays (FPAs) involving long wavelength IR (LWIR) and very long wavelength IR
6 Handbookof Infrared Detection Technologies
(VLWIR) regions. Among the competing technologies are the quantum well infrared p h o t o c o n d u c t o r s (QWIPs) based on lattice matched GaAs/A1GaAs and strained layer InGaAs/A1GaAs material systems. In comparison with photon detectors, thermal detectors have been considerably less exploited in commercial and military systems. The reason for this disparity is that thermal detectors were popularly believed to be rather slow and insensitive in comparison with photon detectors. As a result, the world-wide effort to develop thermal detectors has been extremely small relative to that of photon detectors. In the last ten years, however, it has been shown that extremely good imagery can be obtained from large thermal detector arrays operating uncooled at TV frame rates. The speed of thermal detectors is quite adequate for non-scanned imagers with two-dimensional detectors. At present, uncooled, monolithic FPAs fabricated from thermal detectors, revolutionise the development of low cost thermal imagers. In this paper, we discuss the performance of photon detectors as compared to thermal detectors. In comparative studies, more attention is paid to a wide family of photon detectors, especially to HgCdTe photodiodes and QWIPs. The potential performance of different materials used for photon detectors is examined utilizing the o~/G ratio, where cz is the absorption coefficient and G is the thermal generation. Different types of detectors operated as single element devices, are considered. Also such FPA issues as array size, uniformity, operability, multicolour capability and cost of systems, are discussed.
2.2 Fundamental limits to infrared detector performance Spectral detectivity curves for a number of available IR detectors are shown in Figure 2.1. Interest has centered mainly on the wavelengths of the two atmospheric windows 3-5 ~m [middle w a v e l e n g t h IR (MWIR)] and 8 - 1 4 pm (LWIR region) (atmospheric transmission is the highest in these bands and the emissivity maximum of the objects at T,~3()()K is at the wavelength ;.~10 micron), though in recent years there has been increasing interest in longer wavelengths stimulated by space applications. Depending on the detection mechanism, nature of interaction and material properties, the various types of detectors have their own characteristics. These characteristics result in advantages and disadvantages when the detectors are used in field applications 1-4. Table 2.1 shows a comparison of various IR detectors. Progress in IR detector technology is connected with s e m i c o n d u c t o r IR detectors, which are included in the class of photon detectors. In this class of detectors the radiation is absorbed within the material by interaction with electrons either bound to lattice atoms or to impurity atoms or with free electrons. The observed electrical output signal results from the changed electronic energy distribution. The photon detectors show a selective wavelength dependence of response per unit incident radiation power. They exhibit both perfect signal-to-noise performance and a very fast response. But to
Comparison of photon and thermal detectors performance
7
1012
101,
10'~ o
"D 109
108
1
1.5
2
3
4
5 6 7 8 9 10 Wavelength (lum)
15
20
30
40
Figure 2.1 Comparison of the D* of various infrared detectors when operated at tile indicated temperature. Chopping frequency is 1000 Hz.for all detectors except the thermopile (10 Hz), thermocouple (10 Hz), thermistor bolometer (10 Hz), Golay cell (10 Hz) and p!lroelectric detector (10 Hz). Each detector is assumed to view a hemispherical surround at a temperature of 300 K. Theoretical curves for the backgroundlimited D'for ideal photovoltaic and photoconductive detectors and thermal detectors are also shown.
achieve this, the photon detectors require cryogenic cooling. Photon detectors having long-wavelength limits above about 3 pm are generally cooled. This is necessary to prevent the thermal generation of charge carriers. The thermal transitions compete with the optical ones, making non-cooled devices very noisy. Cooling requirements are the main obstacle to the more widespread use of IR systems based on semiconductor photodetectors, making them bulky, heavy, expensive and inconvenient to use. Depending on the nature of the interaction, the class of photon detectors is further sub-divided into different types as shown in Table 2.1. The most important are: intrinsic detectors, extrinsic detectors, photoemissive detectors (PtSi Schottky barriers), and quantum well detectors. Depending on how the electric or magnetic fields are developed, there are various modes such as photoconductive, photovoltaic, photoelectromagnetic (PEM), and photoemissive ones. Each material system can be used for different modes of operation. In this paper we focus on photodiodes. Photodiodes with their very low power dissipation, easy multiplexing on focal plane silicon chip and less stringent noise requirements for the readout devices and circuits, can be assembled in twodimensional (2D) arrays containing a very large number of elements, limited only by existing technologies. Current cooled IR detector systems use material such as HgCdTe, InSb, PtSi, and doped Si. OWIP is a relatively new technology for IR applications. Among these cooled IR detector systems, PtSi FPAs are highly uniform and
3 Table 2.1
2
Comparison o f infrared detectors
0 ?7
Iletector Type
Advantages
Disadvantages
Thermal (thermopile, bolometers, pyroelectric)
Light, rugged, reliable, a n d low cost Room temperature operation
Low detectivity at high frequency Slow response (ms order)
I'hoton
Easier to prepare More stable materials
Very high thermal expansion coefficient Large permittivity
r.,
Easy bandgap tailoring Well developed theory and cxp. Multicolour detectors
Non-uniformity over large area High cost in growth a n d processing Surface instability
*
(;ood material and dopant.s Advanced technology I'ossiblc monolithic inlcgriition
Heteroepitaxy with large I;itticr mis~iiatcli I.ong wavelength cut-offlimited to 7 p m ( a t 77 K )
Very long wavelength operation Kelativcly simple technology
lligh thermal generation 1:xtremely low temperature operation
I,ow-cost, high yields 1.arge a n d close packed 21) arrays
I.ow q u a n t u m el'liciency [,ow temperature operation
'I'ypc I ((;aAs/Al(;aAs. In(;aAs/Al(;aAsJ)
Matured material growth (;ood uniformity over large area Multicolour detectors
High thermal generation Complicated design and growth
'l'y pe I I (InAs/ln(;aSh. InAs/lnAsShJ
1,ow Auger recombination rate Iksy wavelength control
Complicated design and growth Sensitive to the interfaces
Intrinsic
2
IV-VI (I'bs. I'bSc. I'bSn'l'e)
'5 2 2, u -,
-. 9 ? .
Y
5
0 9. s,
Ill-V (In(;:iAs. InAs. IliSb. InAsShJ
1:rec carriers (I'tSi. l't2Si, IrSi) ()uantum wells
- --
~- --
~
-
-~
~
~
~
-~-.-ppp-p
Comparison of photon and thermal detectors performance 9
manufacturable, but have very low quantum efficiency and can only operate in the MWIR range. The InSb FPA technology is mature with very high sensitivity, but it also can operate in the MWIR spectral range. Doped Si has a wide spectral range from 0.8 to 30 pm and it can only operate at very low temperatures. PtSi, InSb, and doped Si detectors do not have wavelength tunability or multicolour capabilities. Both QWIPs and HgCdTe offer high sensitivity with wavelength flexibility in the MWIR, LWIR and VLWIR regions, as well as multicolour capabilities. HgCdTe can also operate in the short wavelength IR (SWIR) region, while QWIP has to go to a direct band-to-band scheme for SWIR operation. The second class of IR detectors is composed of thermal detectors. In a thermal detector the incident radiation is absorbed to change the temperature of the material, and the resultant change in some physical property is used to generate an electrical output. The detector is suspended on lags, which are connected to the heat sink. The signal does not depend upon the photonic nature of the incident radiation. Thus, thermal effects are generally wavelength independent; the signal depends upon the radiant power (or its rate of change) but not upon its spectral content. This assumes that the mechanism responsible for the absorption of the radiation is itself wavelength independent, which is not strictly true in most instances. Attention is directed toward three approaches which have found the greatest utility in infrared technology; namely, bolometers, pyroelectric and thermoelectric effects. In pyroelectric detectors a change in the internal electrical polarization is measured, whereas in the case of thermistor bolometers a change in the electrical resistance is measured. In contrast to photon detectors, the thermal detectors are typically operated at room temperature. They are usually characterized by modest sensitivity and slow response (because heating and cooling of a detector element is a relatively slow process), but they are cheap and easy to use. They have found widespread use in low cost applications, which do not require high performance and speed. Being unselective, they are frequently used in IR spectrometers. Uncooled FPAs fabricated currently from thermal detectors revolutionize the development of thermal imagers s'6. 2.2.1 Photon detectors
The photodetector is a slab of homogeneous semiconductor with the actual 'electrical' area, A e, that is coupled to a beam of infrared radiation by its optical area, Ao (Figure 2.2). Usually, the optical and electrical areas of the device are the same or close. The use of optical concentrators can increase the Ao/Ae ratio. The current responsivity of the photodetector is determined by the quantum efficiency, r/, and by the photoelectric gain, g. The q u a n t u m efficiency value describes how well the detector is coupled to the radiation to be detected. It is usually defined as the number of electron-hole pairs generated per incident photon. The idea of photoconductive gain, g, was put forth by Rose 7 as a simplifying concept for the understanding of photoconductive phenomena and is now widely used in the field. The photoelectric gain is the number of carriers passing contacts per one generated pair. This value shows how well the
10 Handbook of Infrared Detection Technologies
Figure 2.2 Model of a photodetector. generated electron-hole pairs are used to generate the current response of a photodetector. Both values are assumed here as constant over the volume of the device. The spectral c u r r e n t responsivity is equal to (1)
R i - -~ccq g
where 2 is the wavelength, h is Planck's constant, c is the light velocity, and q is the electron charge. Assuming that the current gains for p h o t o c u r r e n t and noise c u r r e n t are the same, the current noise due to generation and recombination processes is 7 12 - 2(G + R)AetAfq2g 2
(2)
where G and R are the generation and recombination rates, Af is the frequency band and t is the thickness of the detector. Detectivity D* is the main p a r a m e t e r characterizing normalized signal-tonoise performance of detectors and can be defined as D* --
R , ( A o A f ) 1/2 I,
(3)
According to eqs ( 1 )-(3)s D* -
-
2(A~ 1/2 hc Aee ~[2(G+ R)t] -1/2
(4)
Comparison of photon and thermal detectorsperformance 11 For a given wavelength and operating temperature, the highest performance can be obtained by maximizing q/[t(G+R)] 1/2 which corresponds to the condition of the highest ratio of the sheet optical generation to the square root of sheet thermal generation-recombination. This means that high q u a n t u m efficiency must be obtained with a thin device. For a given wavelength and operating temperature, the performance can be optimized by reducing the total number of generation and recombination acts, which is (G + R)(Aet). In further considerations we put that Ao/Ae = 1. Assuming a single pass of the radiation and negligible frontside and backside reflection coefficients, the quantum efficiency and detectivity are r/= 1 - exp(-c~t)
(5)
D* - hcc "~ (1 - e-at) [2(G + R)t] - 1/2
(6)
where a is the absorption coefficient. The highest detectivity can be obtained for t = l . 2 6 / a for which (1-eat)t -1/2 achieves a maximum value of 0.62a 1/2. This thickness is the best compromise between the requirements of high quantum efficiency and low thermal generation. In this optimum case 17= O. 716 and detectivity is equal 8
D* - 0.4S
2
) 1/2 ot C+R
(7)
To achieve a high performance, the thermal generation must be suppressed to possibly the lowest level. This is usually done by cryogenic cooling of the detector. For practical purposes, the ideal situation occurs when the thermal generation is reduced below the optical generation. At equilibrium, the generation and recombination rates are equal, and we have
d. (or) 1/2
D* - 0.31~c ~
(8)
Considerations carried out in ref. 8 indicate that, for a double pass of radiation, the detectivity of an optimized photodetector, leads to the following expression
~ (~)1/2 D* -- 0"31~cck
(9)
where 1 ~
12
Handbook of Infrared Detection Technologies
10~
~10 4
300 K 77K "'-..,
--.
10-' l t
E
410~
" ' C ....
PbSe
r
1101"~
.g_
102 &
~1.
InAs~!Te~ L i I
101~
1
I
2
I',
3
I:
4
:
5
i ~// t Hgc,,Cdo,-Te 1
6
!
7
8
J
HgI Cd [ e
9
10 11 12 13 ,14,03t
Wavelength (lxm)
Figure 2.3 Absorption coef~cientfor various photodetector materials in spectral range 1-14 lzm.
Figure 2.3 shows the measured intrinsic absorption coefficients for various narrow gap photodetector materials. The absorption coefficient and corresponding penetration depth vary among the semiconductor materials. It is well known that the absorption curve for direct transitions between parabolic bands at photon energy greater than energy gap, Eq, obeys a square-root law ~(hlJ) - ~ ( h l ) -
Eg) 1/2
(ao)
where/~ is a constant. As can be readily seen in Figure 2.3, in the MWIR spectral region, the absorption edge value changes between 2 x l ( ) 3 c m -1 and 3 x 103 cm- 1. in the LWIR region it is about 10 ~cm- 1. Since cz is a strong function of the wavelength, for a given semiconductor the wavelength range in which an appreciable photocurrent can be generated, is limited. Near the material's band gap, there is tremendous variation causing a third order of magnitude variation in absorption. In the region of the material's maximum usable wavelength, the absorption efficiency drops dramatically. For wavelengths longer than the cut-off wavelength, the values of a are too small to give appreciable absorption. Figure 2.4 shows the infrared absorption spectra for different n-doped, 50 period GaAs/Al• OWIP structures measured at room temperature using a 45 ~ multipass waveguide geometry. The spectra of the bound-to-bound continuum (B-C) QWIP (samples A, B, and C) are much broader than the boundto-bound (B-B) (sample E) or bound-to-quasibound (B-QB) OWIP (sample F). Correspondingly, the value of the absorption coefficient for the B-C OWIP is significantly lower than that of the B-B OWIP, due to conservation of oscillator strength. The values of the absorption coefficient at 77 K, peak wavelength ).p, cut-off wavelength 2c (long wavelength for which c~drops to half C~p),and spectral
Comparison of photon and thermal detectors performance
700
1 /~1
/
E600
I
/B ~
I
r
g
1600
T = 300 K 1200E
Z" 500 c._o E- 400 .~
'
13
g 800
300
g
o 200
400
100
6
8
10 12 Wavelen9th (gin)
14
16
Figure 2.4 Absorption coefficient spectra measured at T= JO0 K for different QWIP samples described in Table 2.2 (after ref. 9).
width A2 (full width at half etp) are given in Table 2.2. It appears that the lowtemperature absorption coefficient ~ p ( 7 7 K)~l.3C~p(300 K) and % ( A 2 / ) . ) / N D is a constant (ND is the well's doping). 9 A typical value of the absorption coefficient in 7 7 K in the LWIR region is between 600 and 800 cm -]. Comparing the last two figures (Figures 2.3 and 2.4) we can notice that the absorption coefficients for direct band-to-band absorption is higher than that for intersub-band transitions. The ot/Gth ratio versus temperature for different types of tunable materials with hypothetical energy gap equal 0.25 eV (2 - 5 ~tm) and 0 . 1 2 4 eV (2 - 10 ~m) is shown in Figures 2.5 and 2.6. Procedures used in calculations of Gth for different materials are given in Appendix A1-A4. It is apparent that the HgCdTe is by far the most efficient detector of IR radiation" it is characterized by high absorption coefficient and relatively low thermal generation rate. We can also notice that OWIP is better material than extrinsic silicon. The above two figures are completed by Figure 2.7, where the ol/Gth ratio of dependence on wavelength is presented for different materials at 77 K. An optimized photodetector should consist of ] (): 9 9
9
A lightly doped active (base) region, which acts as an absorber of IR radiation. Its band gap E,, doping and geometry should be selected. Electric contacts to the base region, which sense optically generated charge carriers and which should not contribute to the dark current of the device. Surfaces of the absorber regions which must be insulated from the ambient by a material, which also does not contribute to the generation of carriers; in addition the carriers, which are optically generated in the absorber,
Table 2.2 Structure parameters for different n-doped. 5 0 period AI,Gal .As QWlP structures (after ref. 9) Samplc
A
I3 C' I< I:
Wcll width
I3arricr width
C'ornposition
(A)
(A)
Y
Iloping density (lOiXcrn '1
Intersub-hand transition
i,, (pm)
40 40 (10 50 45
500 500 500 500 500
0.Lh 0.25 0.15 0.2h 0. 30
1.O 1.6 0.5 0.42 0.5
13-CT 13-C' 13-cT 13-13 13-013
9.0 9.7 5 8.6
- -1.1
I.,
(1.1rn) 10. 3 10.9 14.5 9.0 8.1 5
Ai (~un)
A ('Kt)
u,,( 7 7 K ) (crn ' )
3.0 2.9
33 30 1 (1 9 11
410 hi0 450 1x20 Xi5
2.1 0.75 0.85
Comparison of photon and thermal detectors performance
100
~
10 "-
~
Xc= 5 lam ~ .
d~ 10-8E
r
C0
_
15
\\
.
HgCdTe
.
QWlP
.
Si
\ ".. ".
0_12
".
\\
10-~6..
10_~o_ 1 0 .24
".....
........
~ -.......... ' " . . . .... ..
-- _.
......
...............
100 150 Temperature (K)
50
200
Figure 2.5 o~/Gthratio versus temperature for MWIR (,;.r 5/lm) photon detectors.
Xc =
0-11_
-
.
-
.
-
.
-
.
.
101am HgCdTe QWlP
~- 10 '~E o 10 '~-
"'..
~ 9
""'"".........
~ ~ ~ ~
..~ ...~ .~....~
1 0 -23
--.-........ '"".................... ......................
1 0 -27
50
100 '
15 0
Temperature (K)
200 '
Figure 2.6 c~/Gth ratio versus temperature for L WIR ( ~.~.=10 Ilm ) photon detectors.
should be kept away from surfaces where recombination can reduce the quantum efficiency. A backside mirror for double pass of IR radiation.
16
Handbook of Infrared Detection Technologies 10 ~ T-77K 10 3
10 -3
oo
E 0 CO
HgCdTe
QWIP
\
10 -~
"". \ \ '"\i \ \
10 -~5
10
9 ',nS0 .9.
-2~
~ \ \ 9. ~ '-. "~
.~. 9-....
"-~---..... ~..
..............................-.:
1 0 -27
0
5
10 15 Wavelength (~um)
20
25
Figure 2.7 ot/Gth ratio versus cut-off u,aveh'ngth for different t!lpes of photon detectors operated at 77 K.
The above conditions can be fulfilled using heterojunctions like N+-p-p + and P + - n - n + with heavily doped contact regions (symbol '+' denotes strong doping, capital l e t t e r - wider gap). Homojunction devices (like n-p, n+-p, p+-n) suffer from surface problems: excess thermal generation results in increased dark current and recombination, which reduces the photocurrent. To achieve a high performance, the thermal generation must be suppressed to possibly the lowest level. This is usually done by cryogenic cooling of the detector. For practical purposes, the ideal situation occurs when the thermal generation is reduced below the optical generation. The requirements for the thermal generation rate can be highly reduced in heterodyne systems, in which optical excitation by the local oscillator can dominate the generation, even for high thermal generation. The total generation rate is a sum of the optical and thermal generation G - Gth + Got,
( 1 O)
The optical generation may be due to the signal or background radiation. For infrared detectors, usually background radiation is higher than the signal radiation. If the thermal generation is reduced much below the background level, the performance of the device is determined by the background radiation (BLIP conditions for Background Limited Infrared Photodetector). This condition can be described as qOPB T t
> nrl,
(11)
Comparison of photon and thermal detectors performance
17
where nth is the density of thermal carriers at temperature T, and r is the carrier lifetime, and ~R is the total background photon flux density (unit cm -2 s -1) reaching the detector. Rearranging we have, for the BLIP requirements, 17r B
G~
t
=
>
tlth
r
-
G.,
(12)
i.e., the photon generation rate per unit volume needs to be greater t h a n the thermal generation rate per unit volume. The carriers can be either majority or minority in nature. The direct bandgap semiconductor photodiode is a minority carrier device and in thermal equilibrium t/rain
--
tlmaj
(13)
where ni is the intrinsic carrier concentration, and n,,,,,j is the majority carrier concentration. The ultimate limit on carrier lifetime in a direct gap semiconductor is given by band-to-band recombination, by either radiative or Auger processes. Humphreys indicated 11 that the van Roosbroeck and Shockley theory of radiative recombination underestimates the radiative lifetime due to noiseless photon re-absorption. As a result, Auger recombination is the dominant process in narrow gap semiconductors like, e.g., HgCdTe ternary alloy. The extrinsic semiconductor photoconductor is strictly a majority carrier device. The background limited detectivity, or so-called 'photovoltaic' BLIP detectivity, is given by 12'13
DRuP = hc
(14)
D'BLIP for photoconductors is v ~ times lower than for photodiodes. This is attributable to the recombination process in photoconductors, which is uncorrelated with the generation process, which contributes to the detector noise. The background photon flux density received by the detector depends on its angular field of view (FOV) of the background and on its ability to respond to the wavelengths contained in this source. Plots of D*~LIp as a function of wavelength for TR/.ip= 300 K and for full 2x FOV are shown in Figure 2.1 BLIP temperature is defined when the device is operating at a temperature at which the dark current is equal to the background photocurrent, given an FOV, and a background temperature. In Figure 2.8. plots of the calculated temperature required for background limited (BLIP) operation in 30 ~ FOV are shown as a function of cut-off wavelength. We can see that the operating temperature of 'bulk' intrinsic IR detectors (HgCdTe and PbSnTe) is higher than for other types of photon detectors. HgCdTe detectors with background limited performance operate with thermoelectric coolers in the MWIR range. However, the LWIR detectors (8~<)., ~< 12 lam) operate at ~1()()K. HgCdTe photodiodes
18
Handbook of Infrared Detection Technologies
"~\
-
FOV = 30 ~ Scene temperature = 300 K
",~..\ ~ p-on-n HgCdTe Auger limited photodiodes ',,~..\.~ N~= 5x10 TM cm~, t = 10 ~m ',',,-.k--
200
I
\
",,~..~ ~ n*-on-p HgCdTe Auger limited photodiodes ',,~,.",~j/... N. = 5x10 '5 cm 3, t= 10 lum
\
\\
,,z' v
I-
._1 m
"*.X,. ~HgCdTe Auger limited photoconductors x .. X 14 -3 " ' , ~ ~ ....... N , = 3 x l 0 cm , t = 1 0 p m "', "".":b<.~,.. n'-on-p PbSnTe Auger limited photodiodes ', "'~~' N~ = 10" cm3, t = 10 pm ",,, "--f. ......
100 Schott
Q = 10 ~~ [__L
00
i
i
]
L__
5
L
.
~
L
10 Cutoff wavelength (lum)
. . . . . . . .
15
1
......
20
Figure 2.8 Estimation of the temperature required for background limited operation of different types of photon detectors. In the calculations FOV= 3() ~ and TB- 300 K are assumed (after ref. 14).
exhibit higher operating temperatures compared to extrinsic detectors, silicide Schottky barriers and QWIPs. However, the cooling requirements for QWIPs with cut-off wavelengths below 10 l.tm are less stringent in comparison with extrinsic detectors and Schottky barrier devices. 2.2.2 Thermal detectors Thermal detectors operate on a simple principle, that when heated by incoming IR radiation their temperature increases and the temperature changes are measured by any temperature-dependent mechanism, such as thermoelectric voltage, resistance, pyroelectric voltage. The simplest representation of the thermal detector is shown in Figure 2.9. The detector is represented by a thermal capacitance Cth coupled via a thermal conductance Gth to a heat sink at a constant temperature T. In the absence of a radiation input, the average temperature of the detector will also be T, although it will exhibit a fluctuation about this value. When a radiation input is received by the detector, the rise in temperature is found by solving the heat balance equation. Assuming the radiant power to be a periodic function, the change in temperature of any thermal detector due to incident radiative flux is 12.1 s AT--
(15)
Comparison of photon and thermal detectors performance
19
Figure 2.9 Thermal detector mounted via lags to heat sink.
Equation (15) illustrates several features of the thermal detector. Clearly it is advantageous to make AT as large as possible. To do this, the thermal capacity of the detector (Cth) and its thermal coupling to its surroundings (Gth) must be as small as possible. The interaction of the thermal detector with the incident radiation should be optimized while reducing as far as possible all other thermal contacts with its surroundings. This means that a small detector mass and fine connecting wires to the heat sink are desirable. A characteristic thermal response time for the detector can therefore be defined as Cth 75th -- ~Gth -- CthRth
(1 6)
where R t h - - 1 / G t h is the thermal resistance. The value of the thermal time constant is in the millisecond range. This is m u c h longer than the typical time of a photon detector. There is a trade-off between sensitivity, AT, and frequency response. If one wants a high sensitivity, then a low frequency response is forced upon the detector. For further discussion we introduce the coefficient K=AV/AT, which reflects how good the temperature change translates into the electrical output voltage of the detector. The voltage responsivity Rv of the detector is the ratio of the output signal voltage A V to the input radiation power and is given by
20
Handbookof InfraredDetectionTechnologies KsRth
R,, =
(17)
In order to determine the detectivity of the detector, it is necessary to define a noise mechanism. One major noise is Johnson noise. Two other f u n d a m e n t a l noise sources are i m p o r t a n t for assessing the ultimate performance of a detector: thermal fluctuation noise and b a c k g r o u n d fluctuation noise. Thermal fluctuation noise arises from t e m p e r a t u r e fluctuations in the detector. These fluctuations are caused by heat c o n d u c t a n c e variations between the detector and the s u r r o u n d i n g substrate with which the detector element is in thermal contact. The spectral noise voltage due to t e m p e r a t u r e fluctuations is ~2.1 s
4kT 2 A f K2
V~, = 1 4- ,02 r~,
(18)
Rt,,
A third noise source is b a c k g r o u n d noise which results from radiative heat exchange between the detector at t e m p e r a t u r e Td and the s u r r o u n d i n g e n v i r o n m e n t at t e m p e r a t u r e Tl,, that is being observed. It is the ultimate limit of a detector's performance capability and is given for a 2 rr FOV by 12.1 s
V~,
-
8ksc~A(T2 + T~,) K2 , 1 + ofir~,
R~-,,
(19)
where c~ is the Stefan-Boltzmann constant. The f u n d a m e n t a l limit to the sensitivity of any thermal detector is set by t e m p e r a t u r e fluctuation noise, i.e., r a n d o m fluctuations in the t e m p e r a t u r e of the detector element due to fluctuations in the radiant power exchange between the detector and its surroundings. Under this condition at low frequencies (to < < 1/rth) w e have
Dt,, -
4kTdGt,,,j
(20)
It is assumed here that E is independent of wavelength, so that the spectral D*; and blackbody D*(T) values are identical. If radiant power exchange is the d o m i n a n t heat exchange mechanism, then G is the first derivative with respect to temperature of the Stefan-Boltzmann function. In that case, k n o w n as the background fluctuation noise limit, we have
, Db--
I.
8 8ka(T dg -4- Tb~)
] 1,/2
(21)
Comparison of photon and thermal detectors performance
21
Note that D*b is independent of A, as is to be expected. Equations (20) and (21) assume that background radiation falls upon the detector from all directions when the detector and background temperature are equal, and from the forward hemisphere only when the detector is at cryogenic temperatures. The highest possible D* to be expected for a thermal detector operated at room temperature and viewing a backgrounds at room temperature is 1 . 9 8 • l~ cmHz 1/2 W -1. Even if the detector or background, not both, were cooled to absolute zero, the detectivity would improve only by the square-root of two. This is the basic limitation of all thermal detectors. The background-noiselimited photon detectors have higher detectivities as a result of their limited spectral responses. The performance achieved by any real detector will be inferior to that predicted by equation (21 ). The degradation of performance will arise from: 9 encapsulation of detector (reflection and absorption losses at the window), 9 the effects of excess thermal conductance (influence of electrical contacts, conduction through the supports, influence of any gas-conduction and convection), 9 the additionalnoise sources. Typical values of detectivities of thermal detectors at 10 Hz change in the range between ]08 and 109 cmHz ~/2 W -~. 2.2.3 Comparison of the fundamental limits of photon and thermal detectors
The temperature dependence of the fundamental limits of D* of photon and thermal detectors for different levels of background are shown in Figures 2.10 and 2.11. From Figure 2.10 the results show that in the LWIR spectral range, the performance of intrinsic IR detectors (HgCdTe photodiodes) is higher than for other types of photon detectors. HgCdTe photodiodes with background limited performance operate at temperatures below about 80 K. HgCdTe is characterized by a high optical absorption coefficient and quantum efficiency and a relatively low thermal generation rate compared to extrinsic detectors and OWIPs. The extrinsic photon detectors require more cooling than intrinsic photon detectors having the same long wavelength limit. The theoretical detectivity value for the thermal detectors is much less temperature dependent than for the photon detectors. At temperatures below 50 K and zero background, LWIR thermal detectors are characterized by D* values lower than those of LWIR photon detectors. However, at temperatures above 60 K, the limits favour the thermal detectors. At room temperature, the performance of thermal detectors is much better than LWIR photon detectors. The above relations are modified by the influence of background, shown in Figure 2.10 for a background of 1017 photons/cm2s. It is interesting to notice that the theoretical curves of D* for photon and thermal detectors show similar fundamental limits at low temperatures.
22
Handbook of Infrared Detection Technologies
1014~-\\ i
,_~
QB-"
#
.
-
~,\~
-OLO-
\
101'Ph/cm~s
-'----,,
/nermal
~
HgCdTe
~
-1- 1 E
-
10 8 ~-
"~xtrinsic
_
-
14 pm
;Lc =
10 6
,
10
1
30
,
~ 1
.
.
_ .
.
50 70 90 Temperature (K)
110
130
Figure 2.10 Theoretical performance limits of LWIR photon and thermal detectors at zero background and background of 1017 photons/cm2s, as afimction of detector temperature (after ref. 16).
, o1,ii,',,'-..'-., , "7
10,~_ .
",,",, .
.
.
QB- 10'~Ph/cm2s Q~ - 0 ,"-- ......
Thermal
..............
otodiode
~N 10 ~~ "1" E
~
10 ~ ~.~trinsic
10 ~
lo
Xc = 28 p m
3o
~;o
~
zo
~;o
Temperature (K)
1~o
l ao
Figure 2.11 Theoretical performance limits of VL WIR photon and thermal detectors at zero a background and background of 10 lz photons/cm2s, as afunction of detector temperature (after ref. 16).
Comparison of photon and thermal detectors performance 2 3
Figure 2.12 Thermal imager system configuration.
Similar experiments have been carried out for very long wavelength infrared (VLWIR) detectors operated in the 1 4 - 5 0 p m spectral range. Results of calculations are presented in Figure 2.11. Detectors operating within this range are cryogenic Si and Ge extrinsic photoconductors and cryogenic thermal detectors, usually bolometers. Nevertheless, theoretical prediction for intrinsic detectors (HgCdTe photodiodes) is also included. Figure 2.11 shows that the theoretical performance limit of a VLWIR thermal detector at zero and high backgrounds in a wide range of temperatures, equals or exceeds that of photon detectors. The comparison of both types of detector indicates that theoretical performance limits for thermal detectors are more favourable as the wavelength of operation moves from the LWIR to the VLWIR. It is due to the influence of a fundamentally different type of noise (generation-recombination noise in photon detectors and temperature fluctuation noise in thermal detectors), these two classes of detectors have different dependencies of detectivities on wavelength and temperature. The photon detectors are favoured at long wavelength infrared and lower operating temperatures. The thermal detectors are favoured at very long wavelength spectral range. The temperature requirements to attain background fluctuation noise performance, in general favour thermal detectors at the higher cryogenic temperatures and photon detectors at the lower cryogenic temperatures.
2.3 Focal plane array performance For FPAs the relevant figure of merit is the noise equivalent temperature difference (NEDT), the temperature change of a scene required to produce a signal equal to the rms noise. The configuration of the basic thermal-imager system is shown in Figure 2.12. The spectral photon incidence for a full hemispheric surround is
24 Handbookof Infrared Detection Technologies
O-
r(;t)O(T, 2)d2
(22)
).
if a zero-emissivity bandpass filter having in-band transmission, r(2), cut-on wavelength kl, and cut-off wavelength k2 is used (zero emissivity is practically obtained by cooling the spectral filter to a temperature where its self-radiation is negligible). The photon flux density incident on the detector focal plane arrays is
OB__=
1
] + 4(f/#) 2
0__
(23)
where f / # is the ratio of the focal length to the diameter of the limiting aperture or lens. Under these conditions the background-induced photocurrent in any photon detector of area Ad is:
Iph
1 + 4(f/#) 2 ;. r(2)0(2)0(r, 2)d2
(24)
where rl()~) is a spectral response per photon (quantum efficiency) and z(k) is an optics (filter) transmission spectrum. The thermal contrast is one of the important parameters for IR imaging devices. It is the ratio of the derivative of spectral photon incidence to the spectral photon incidence
c
=
~176176 o
(2s)
Figure 2.13 is a plot of C for several MWIR sub-bands and the 8 - 1 2 pm LWIR spectral band. We can notice, that the contrast in the MWIR bands at 300 K is 3.5-4 percent compared to 1.6 percent for the LWIR band. The noise equivalent difference temperature of a detector represents the temperature change, for incident radiation, that gives an output signal equal to the rms noise level. While normally thought of as a system parameter, detector NEDT and system NEDT are the same except for system losses. NEDT is defined as NEDT -
Vn(3T/3O) AT (3Vs/3O) = V,, A Vs
(26)
where Vn is the rms noise and AV~ is the signal measured for the temperature difference AT. It can be shown that 17 N E D T - (rC~n1,,pV~w)-'
(27)
Comparison of photon and thermal detectors performance
25
0 . 0 6 0 .. _
...... 9
.9. . . . . . . . ..
0.050
0.040
cO
-. ,,,",.,
.....
3.5-4.1
- .... 3.5-5.0 ........
---~
.........
i
... 9
pm
pm
4 . 5 - 5 . 0 lam 8.0-12.0 pm
.. . . . . . . . . . . . . . . ".iiiiiii
E 0.030 r !--
0.020
0.010 250
~
l 270
l
l L l I 290 310 S c e n e t e m p e r a t u r e (K)
1 330
I
350
Figure 2.13 Spectral photon contrast in the MWIR and L WIR (after ref. 17).
where Nw is the number of photogenerated carriers integrated for one integration time, tint:
Nw -
rlAcltintQB
(2
The percentage of BLIP, FPA noise
rlst,e
rlBLIp,is simply the
~Nvhoto,,+NFi,A2) 9
2
r
o
w
_
_
8)
ratio of photon noise to composite
(29)
It results from the above formulas that the charge handling capacity of the readout, the integration time linked to the flame time, and the dark current of the sensitive material, become the major issues ofIR FPAs. The NEDT is inversely proportional to the square-root of the integrated charge and therefore the greater the charge, the higher the performance. The distinction between the integration time, and the FPA flame time must be noted. At high backgrounds it is often impossible to handle the large amount of carriers generated over flame time compatible with standard video rates. Off-FPA flame integration can be used to attain a level of sensor sensitivity that is commensurate with the detector-limited D* and not the charge-handlinglimited D*. It is of interest to compare the performance of uncooled photon and thermal detectors in the MWIR (2= 5 lam) and LWIR (k= 10 pm) spectral range. The paper recently published by Kinch as follows us in this comparison. Figure 2.14 compares theoretical NEDT of detectors operated at 290 K for f/1 optics and a
26 Handbook of Infrared Detection Technologies I mil pixel size. Parameters typical for micromachined resonant cavity bolometers given in Appendix A5 are assumed in calculations. As a photon detector, N+-1r-P + HgCdTe photodiode is chosen, first proposed by Elliott and Ashley. 19 7r designates an intrinsic region containing a p-type background dopant equal to 5x 1014 cm -3 with carrier lifetime limited by the Auger 7 process. It is also assumed that the detector node capacity can store the integrated charge due to detector dark current. Figure 2.14 shows that the ultimate performance of the uncooled HgCdTe photon detectors is far superior to the thermal detectors at wide flame rates and spectral bands. Also, any other tunable bandgap alloy, such as type-II InAs/ GaInSb superlattices, could be worthwhile with regard to the development of an uncooled photon detector technology. 1'2~ The ultimate performance of HgCdTe photodiodes with optimally doped base region are comparable with that of InAs/ InGaSb strain layer superlattices in the temperature range between 300 K and 77K. 21 Comparing both curves of Figure 2.14 for thermal detectors, we see that for long integration times in the LWIR region, excellent performance is achieved, with NEDT values below 10 mK for frame rates of 30 Hz. However, for snapshot systems with integration time below 2 ms, the available NEDT is above 100 mK even at the LWIR region. For the MWIR band the thermal detector has obvious performance limitations at any frame rate.
2.4 FPAs of photon detectors There are a number of architectures used in the development of IR FPAs. In general, they may be classified as hybrid and monolithic. The central design questions involve performance advantages versus ultimate producibility. Each application may favour a different approach depending on the technical requirements, projected costs, and schedule. In the monolithic approach, some of the multiplexing is done in the detector material itself rather than in an external readout circuit. The basic element of a monolithic array is a metal-insulator-semiconductor (MIS) structure as shown in Figure 2.15(c). An MIS capacitor detects and integrates the IR-generated photocurrent. Although efforts have been made to develop monolithic FPAs using narrow-gap semiconductors, silicon based FPA technology with Schottkybarrier detectors is the only technology, which has matured to a level of practical use. Microbolometer FPAs are also based on silicon technology (see Figure 2.15(d)). Hybrid FPAs detectors and multiplexers are fabricated on different substrates and mated with each other by the flip-chip bonding or loophole interconnection (Figure 2.16). In this case we can optimize the detector material and multiplexer independently. Other advantages of the hybrid FPAs are; near 100% fill factor and increased signal-processing area on the multiplexer chip. In the flip-chip bonding, the detector array is typically connected by contacts via indium bumps to the silicon multiplex pads. The detector array can be illuminated from either
l O 2 ~(a)
102p-
LWlR 300 K
-3
lo4
'
l
L
"
Integration time (s)
lo1' t
I
MWlR 300 K
lo'/
" " i lI
t
10'
8
1
,
'
'
L
'
I
,
' ~ " ' ' ' ~
n
~
"
"
'
~
l
10' Io2 Integration time (s)
lo3
'
I
I C " '
1
1:igrirr 2 . 14 Tllrorctic~crlNE:'L)Ti.o~tlpirrisc,,~ o$lir~cool~d tlrernlul crtlrl HgCd'l'r ~ir~cool~~ilpl~oton 1,WIK ( u )citld MWlK ( 1 1 ) il(~1i~c~tor.s (riftiv' rqf. 18).
28 Handbookof Infrared Detection Technologies
Figure 2.15 Monolithic IR FPAs: (a) all-silicon: (b) heteroepitaxy-on-silicon: (c) non-silicon (e.g.. HgCdTe CCD): (d) microbolometer.
the frontside or backside (with photons passing through the transparent detector array substrate). In general, the latter approach is most advantageous. In HgCdTe hybrid FPAs, photovoltaic detectors are formed on thin HgCdTe epitaxial layer on transparent CdTe or CdZnTe substrates. For HgCdTe flip-chip hybrid technology, the maximum chip size is of the order of 10 mm square. In order to overcome this problem, the technology is being developed with saphire or silicon as the substrate of HgCdTe detectors. When using opaque materials, substrates must be thinned to 1 0 - 2 0 ~tm in order to obtain sufficient quantum efficiencies and reduce the crosstalk. There is a large research activity directed towards 2D staring arrays detectors consisting of more than 106 elements. IR FPAs have nominally the same growth rate as dynamic random access memory (RAM) integrated circuits (ICs) (it is a consequence of Moore's Law, which predicts the ability to double the transistor integration on each IC about every 18 months) but lag behind in size by about
Comparison of photon and thermal detectors performance 29
Figure 2.16 Hybrid IR FPA with independently optimized signal detection and readout: (a) indium bump technique; (b) loophole technique.
5 - - 1 0 years. ROIC's are somewhat analogous to dynamic R A M m o n l y readouts require a m i n i m u m of three transistors per pixel compared to one per memory cell. Consequently, whereas various 6 4 x 6 4 FPAs were available in the early 1980s, several vendors are now producing monolithic FPAs in the TVcompatible 1 0 4 0 x 1040 formats. Figure 2.17 illustrates the trend of array size over the past 25 years and some projections of w h a t will evolve in the coming decade. Rockwell has developed the world's largest HgCdTe SWIR FPA for astronomy and low background applications. The format of the device is a hybrid 2 0 4 8 x 2 0 4 8 with a unit cell size of 1 8 ~ t m x l 8 ~ t m . Table 2.3 contains a description of representative IR FPAs that are commercially available as standard products and/or catalogue items from the major manufacturers. Twenty years ago, high quality single element detectors were often priced over $ 2 0 0 0 , but now, some current IR FPA production costs are less than $0.1 per detector and even greater reductions are e x p e c t e d in the near future. As the commercial market for uncooled imagers expands, the cost of commercial systems will inevitably decrease. At present, the cost of 3 2 0 x 2 4 0 , 50• 50 ~tm2, bolometer arrays for thermal imagers is $15 0 0 0 - 2 0 000. Two types of silicon addressing circuits have been developed: CCDs (CMOS) switches. The and complementary m e t a l - o x i d e - s e m i c o n d u c t o r photogenerated carriers are first integrated in the well, formed by a photogate, and subsequently transferred to slow (vertical) and fast (horizontal) CCD shift registers.
30
Handbook of Infrared Detection Technologies
101~
4 megapixel infrared
109- focal plane arrays .-Q" c- 10" O t._ 10~ Cu oo 106 .-=,_ .Q 0
~
before 2000
_.~
-
I1~ e s i c c D
~
-
DRAM 1 ~ " / PtSi.......[] .....~ K .production/~ ....[] ......... ~ 2x512 ~A480x640
10 5
0 x
il. 10'
_ ......~ r ~
103 4 102
K ,
.~..-'OHiOdTe
~ J
S
QwlP
.........
_
'''Cr''''" ~
L
72 74 z6 78 80 82
,
~
,
88 90 92 94 96 98 2000 &
Year
~
o4 06 68 2010
Figure 2.17 Increase in array format size over the past 2 ~ years and projections for the coming decade. PtSi, InSb, and HgCdTe have been following the pace of d!lnamic RAM. offset by about a decade. O WIP detectors have been recently reported in sizes as large as 640• 480 pixels (after ref. 22).
An attractive alternative to the CCD readout is coordinative addressing with CMOS switches. The advantages of CMOS are that existing foundries, which fabricate application specific integrated circuits, can be readily used by adapting their design rules. Design rules of 0.18 pm are in production with pre-production runs of the 0.12 lum design rules. At present, CMOS with m i n i m u m feature ~<0.5 l.tm is also enabling monolithic visible CMOS imagers.
2.4.7 InSb photodiodes InSb material is more mature t h a n HgCdTe and good quality bulk substrates of larger t h a n 7 cm diameter are commercially available. 23 InSb photodiodes have been available since the late fifties and they are generally fabricated by impurity diffusion (usually Cd) and ion implantation (usually Be + or B+). Epitaxy is not used; instead, the standard m a n u f a c t u r i n g technique begins with bulk n-type single crystal wafers with donor concentration about 101 ~ cm -3. Wimmers et al. have presented the status of InSb photodiode technology for a wide variety of linear and F P A s . 24'2 s A typical InSb photodiode RA product at 77 K is 2 x 106 ~ c m 2 at zero bias and 5 • facm 2 at slight reverse biases of approximately 1 0 0 m V . This characteristic is beneficial when the detector is used in the capacitive discharge mode. As element size decreases below 10 -4 cm -2, some slight degradation in resistance due to surface leakage occurs. InSb photodiodes can also be operated in the temperature range above 77 K. Of course, the RA products degrade in this region. At 120K, R A products of 104 Eacm2 are still achieved with slight reverse bias, making BLIP operation
Table 2.3 Representative IR FPAs offered by some major manufactures Manufacturerjweb site
Pixel size (~m)
Detector material InSb InSb %:As BIB HgCdTe HgCdTe VO,(bolometer) I'yro (BST) HgCd'l'e HgCdTe HgCd'l'e HgCdTe HgCd'l'e VO,(bolorneter)
BAE Systems
HgCdTe HgCdl'e HgCd'l'e HgCdrl'e
Spectral range (~m)
Oper. temp. (K)
( c ~ H z ~ ~ ~ ~ w ) NETD (mK)
W
Table 2.3 (continued) ManufacturerIWeb site
N
Pixel size ( ~ m )
Detector material Amorphous Si(bolometer)
1 IgCd'l'e HgC'd'l'c l'yro ( I'S'I') l'yro ( I'S'l7)
Sensors llnlimited/ www.sensorsinc.com H -- hybrid. M -monolithic
Spectral range ( ~ m )
Oper. temp. (K)
Comparison of photon and thermal detectorsperformance 3 3
possible. The quantum efficiency in InSb photodiodes optimized for this temperature range remains unaffected up to 160 K. InSb photovoltaic detectors are widely used for tactical applications, groundbased infrared astronomy and for applications aboard the Space Infrared Telescope Facility. An array size of 1 0 2 4 • is possible (see Table 2.3) because the InSb detector material is thinned to less than 10 ~tm (after surface passivation and hybridization to a readout chip) which allows it to accommodate the InSb/silicon thermal mismatch. 26 Linear array formats of 64, 128 and 256 elements are also produced with frontside-illuminated detectors for both highbackground and astronomy applications. Element sizes depend on device format and range from 20 x 2 0 - 2 0 0 • 200 Bm. The cryogenically cooled InSb and HgCdTe arrays have comparable array size and pixel yield at MWIR spectral band. However, wavelength tunability and high quantum efficiency have made HgCdTe the preferred material. 2.4.2 HgCdTe photodiodes
During the past four decades, HgCdTe has become the most important semiconductor for the middle and long wavelength (k=3-30~tm) IR photodetectors. The short wavelength region has been dominated by III-V compounds (InGaAs, InAsSb, InGaSb). Spectral cut-off can be tailored by adjusting the HgCdTe alloy composition. Epitaxial techniques are the preferred technique to obtain device-quality HgCdTe material for IR devices. Epitaxial growth of the HgCdTe detector array on a Si substrate, rather than CdZnTe, has emerged as a particularly promising approach to scale up wafer dimensions and achieve cost-effective production. 27 MBE offers unique capabilities in material and device engineering including the lowest growth temperature, superlattice growth and potential for the most sophisticated composition and doping profiles. The growth temperature is less than 200~ for MBE but around 350~ for MOCVD, making it more difficult to control the p-type doping in the MOCVD due to the formation of Hg vacancies. Different HgCdTe photodiode architectures have been fabricated that are compatible with backside and frontside illuminated hybrid FPA technology. 3'28 The fabrication of HgCdTe photodiodes was usually based on the most common n+-p and P+-n double layer heterojunction (DLHJ) structure. In these photodiodes the base p-type layers (or n-type layers) are sandwiched between CdZnTe substrate and high-doped (in n+-p structure) or wider-gap (in P+-n structure) regions. Due to backside illumination (through CdZnTe substrate) and internal electric fields (which are 'blocking' for minority carriers), the influence of surface recombinations on the photodiodes performance is eliminated. Both optical and thermal generations are suppressed in the n+-region due to the Burstein-Moss effect and in the P+-region due to wide gap. The influence of surface recombination is also prevented by the use of suitable passivation. Passivation of HgCdTe has been achieved by several techniques and a comprehensive review was given by Nemirovsky and Bahir. 29 Recently,
36 Handbook of Infrared Detection Technologies Photovoltaic HgCdTe FPAs are based on both p-type and n-type materials. Linear ( 2 4 0 , 2 8 8 , 4 8 0 , and 960 elements), 2D scanning arrays with time delay and integration (TDI), and 2D staring formats from 64 • up to 2048 • have been made. 18 Pixel sizes ranging from 18 ~m square to over 1 mm have been demonstrated. The best results have been obtained using hybrid architecture. 2.4.3 Photoemissive PtSi Schottky-barrier detectors
The most popular Schottky-barrier detector is the PtSi detector which can be used for the detection in the 3-512m spectral range. 3~.~4 Radiation is transmitted through the p-type silicon and is absorbed in the metal PtSi (not in the semiconductor), producing hot holes which are then emitted over the potential barrier into the silicon, leaving the silicide charged negatively. Negative charge of silicide is transferred to a CCD by the direct charge injection method. The effective q u a n t u m efficiency in the 3-_S ~m atmospheric window is very low, of the order of 1%, but useful sensitivity is obtained by means of near full flame integration in area arrays. The q u a n t u m efficiency has been improved by thinning PtSi film and implementation of an 'optical cavity'. Due to very low q u a n t u m efficiency, the operating temperature of Schottky barrier photoemissive detectors is lower than another types of IR photon detectors (see Figure 2.8). Schottky photoemission is independent of such factors as semiconductor doping, minority carrier lifetime, and alloy composition, and, as a result of this, has spatial uniformity characteristics that are far superior to those of other detector technologies. Uniformity is only limited by the geometric definition of the detectors. Most of the reported Schottky-barrier FPAs have the interline transfer CCD architecture. Using a 1.2 txm charge sweep device technology, a large fill factor of 71% was achieved with a 2 6 • 2 pixel in the 5 1 2 • monolithic structure. 34 The NETD was estimated as 0.()33 K with f/1.2 optics at 300 K. The 1040• element CSD FPA has the smallest pixel size (17• 17 l.tm2) among two-dimensional IR FPAs. Current PtSi Schottky barrier FPAs are mainly manufactured in 150 mm wafer process lines with around 1 l~m lithography technologies; the most advanced Si technology offers 200 mm wafers process with 0.25 ~tm design rules. However, the performance of monolithic PtSi Schottky-barrier FPAs has reached a plateau, and slow progress is expected from now on. 2.4.4 Extrinsic photoconductors
Extrinsic photoresistors are used in a wide range of the IR spectrum extending from a few~m to approximately 300pro. They are the principal detectors operating in the range ;k > 20 l~m. 3s Detectors based on silicon and germanium have found the widest application as compared with extrinsic photodetectors on other materials. Si has several advantages over Ge: for example, three orders of
Comparison of photon and thermal detectorsperformance 3 3
possible. The quantum efficiency in InSb photodiodes optimized for this temperature range remains unaffected up to 160 K. InSb photovoltaic detectors are widely used for tactical applications, groundbased infrared astronomy and for applications aboard the Space Infrared Telescope Facility. An array size of 1 0 2 4 • is possible (see Table 2.3) because the InSb detector material is thinned to less than 10 ~tm (after surface passivation and hybridization to a readout chip) which allows it to accommodate the InSb/silicon thermal mismatch. 26 Linear array formats of 64, 128 and 256 elements are also produced with frontside-illuminated detectors for both highbackground and astronomy applications. Element sizes depend on device format and range from 2 0 x 2 0-2 O0 x 2 O0 ~m. The cryogenically cooled InSb and HgCdTe arrays have comparable array size and pixel yield at MWIR spectral band. However, w a v e l e n g t h tunability and high quantum efficiency have made HgCdTe the preferred material. 2.4.2 HgCdTe photodiodes
During the past four decades, HgCdTe has become the most important semiconductor for the middle and long wavelength (~.=3-3()~m) IR photodetectors. The short wavelength region has been dominated by III-V compounds (InGaAs, InAsSb, InGaSb). Spectral cut-off can be tailored by adjusting the HgCdTe alloy composition. Epitaxial techniques are the preferred technique to obtain device-quality HgCdTe material for IR devices. Epitaxial growth of the HgCdTe detector array on a Si substrate, rather than CdZnTe, has emerged as a particularly promising approach to scale up wafer dimensions and achieve cost-effective production. 27 MBE offers unique capabilities in material and device engineering including the lowest growth temperature, superlattice growth and potential for the most sophisticated composition and doping profiles. The growth temperature is less than 200~ for MBE but around 350~ for MOCVD, making it more difficult to control the p-type doping in the MOCVD due to the formation of Hg vacancies. Different HgCdTe photodiode architectures have been fabricated that are compatible with backside and frontside illuminated hybrid FPA technology. 3'2s The fabrication of HgCdTe photodiodes was usually based on the most common n+-p and P+-n double layer heterojunction (DLHJ) structure. In these photodiodes the base p-type layers (or n-type layers) are sandwiched between CdZnTe substrate and high-doped (in n+-p structure) or wider-gap (in P+-n structure) regions. Due to backside illumination (through CdZnTe substrate) and internal electric fields (which are 'blocking' for minority carriers), the influence of surface recombinations on the photodiodes performance is eliminated. Both optical and thermal generations are suppressed in the n+-region due to the Burstein-Moss effect and in the P+-region due to wide gap. The influence of surface recombination is also prevented by the use of suitable passivation. Passivation of HgCdTe has been achieved by several techniques and a comprehensive review was given by Nemirovsky and Bahir. 29 Recently,
34
Handbook of Infrared Detection Technologies
however, most laboratories have been using CdTe or CdZnTe (deposited by MBE, MOCVD, sputtering and e-beam evaporation) for photodiode passivation. 3o For example, the important element of the DLHJ structure is an In-doped (Na.~.lO is cm -3) HgCdTe absorber layer with a wider bandgap cap layer grown by LPE (MBE or MOCVD), arsenic diffusion (or implantation: N . ~ I O is cm-3), a passivation layer, metal contacts to the diode, and indium bumps for mating to the readout integrated circuit multiplexer. It appears that, for the lowest doping levels achievable in a controllable manner in the base regions of photodiodes (N,,= 5 x l 015 cm- 3 for n+-p structure, and Nd=5• 14cm -3 for P+-n structure), the performance of both types of photodiodes is comparable for a given cut-off wavelength and temperature. 31 1081
\
~\
~ k
~
p-on-n HgCdTe N,=5x 10 TM cm -3 t= 10 ~tm
106
A(35K)
--Q
04
9
9
o
_
Q '~o
(40K) (40K)o
10" " ~ 0
2
rr 1 0 -
~o
N(80K/
o"
(40K)e
"~,
o
50K
~176
+ "
9 0
Zx +
~,
~ x
100
60K
(6OK)
,,.
o
9
o
A
7K
x x
+
o (70K) u .~.
(60oK)
1 0 2L-
8
I
12
]
1
16 kc (pm)
~
1
20
I
24
Figure 2.18 Dependence of the RoA product on the long wavelength cut-off for L WIR p-on-n HgCdTe photodiodes at temperatures <<.77 K. The solid lines are cah'ulated assuming that the performance of photodiodes are due to thermal generation governed by the Auger mechanism in the base n-type region of photodiodes with t= 10 #m and Na = 5 • 1014 cm- ~. The experimental values are taken from different papers.
Comparison of photon and thermal detectors performance
37
magnitude higher impurity solubilities are attainable, hence thinner detectors with better spatial resolution can be fabricated from silicon. Si has lower dielectric constant than Ge, and the related device technology of Si has now been more thoroughly developed, including contacting methods, surface passivation and mature MOS and CCD technologies. Moreover, Si detectors are characterized by superior hardness in nuclear radiation environments. Figure 2.20 illustrates the spectral response for several extrinsic detectors. Although the potential of large extrinsic silicon focal plane arrays for terrestrial applications has been examined, interest has declined in favour of HgCdTe and InSb, with their more convenient operating temperatures. Strong interest in doped silicon continues for space applications, particularly in low background flux and for wavelengths from 13 to 20 pm, where compositional control is difficult for HgCdTe. The shallower impurity energies in germanium allow detectors with spectral response up to beyond 100 ~m wavelength and major interest still exists in extrinsic germanium for wavelengths beyond about 20 Hm. To maximize the quantum efficiency and detectivity of extrinsic photoconductors, the doping level should be as high as possible. This idea is realized in blocked impurity band (BIB) devices. The longer spectral response of the BIB Si:As device compared with the bulk Si:As device (see Figure 2.20) is due to the higher doping level in the former that reduces the binding energy of an electron. BIB devices made from either doped silicon or doped germanium are sensitive in the infrared wavelength range of 2 - 2 2 0 ~m. BIB devices in large staring array formats are now becoming commercially available. Hybrid FPAs with Si:As BIB detectors operating in the 4 - 1 0 K temperature range have been optimized for
10 0
(D
r.f) cO c~ oo
CD "-' 1 (I.)
01
-
>
. ~ . , .
c~ CD
:/
n"
10 .2 0
, Si:ln 5
,
,
Si:Ga '.~
.i
10
15
20
25
Wavelength (pm)
30
Figure 2.20 Examples of extrinsic silicon detector spectral response. Shown are Si:In, Si:Ga, and Si:As bulk detectors and a Si:As BIB (after ref. 36).
38 Handbook of Infrared Detection Technologies low, moderate, and high IR backgrounds. 37 The 2 5 6 • format with 30~tm pixels and 2 4 0 • format with 50 ~tm pixels are available for low- and high background applications, respectively. Antimony-doped silicon (Si:Sb) arrays and 128 • 128-pixel Si:Sb hybrid FPAs having response to wavelengths > 40 ~m have been also demonstrated, primarily for use at low and moderate backgrounds. Germanium BIB devices have been developed on an experimental basis, but they have not been reported in large 2D array formats yet. 2.4.5 GaAs/AIGaAs QWlPs
QWIP technology is based on the well-developed A3BS material system, which has a large industrial base with a number of military and commercial applications. OWIP cannot compete with the HgCdTe photodiode as a single device, especially at higher temperature operation ( > 70 K) due to fundamental limitations associated with intersub-band transitions.3S However, the advantage of HgCdTe is less distinct in the temperature range below 50 K due to problems involved in a HgCdTe material (p-type doping, Shockley-Read recombination, trap-assisted tunnelling, surface and interface instabilities). Even though OWIP is a photoconductor, several of its properties such as high impedance, fast response time, long integration time, and low power consumption, comply well with the requirements of large FPAs fabrication. Due to the high material quality at low temperature, OWIP has potential advantages over HgCdTe for VLWIR FPA applications in terms of the array size, uniformity, yield and cost of the systems. A key factor in QWIP FPA performance is the light-coupling scheme. A distinct feature of QWlPs is that the optical absorption strength is proportional to an incident photon's electric-field polarisation component normal to the q u a n t u m wells. For imaging, it is necessary to be able to couple light uniformly to 2D arrays of these detectors, so a diffraction grating or other similar structure is typically fabricated on one side of the detectors to redirect a normally incident photon into propagation angles more favourable for absorption. The pixels of 2D arrays are thinned to about 5 ~tm in thickness. The thinning traps diffracted light inside the illuminated pixels, increasing responsivity and eliminating crosstalk.
Table 2.4
Essential properties of LWIR HgCdTe p h o t o d i o d e s and QWIPs at T=77 K
Parameter
HgCdTe
QWlP (n-type)
IR absorption
Normal incidence
Eoptic, d _Lplane of well required
Quantum efficiency Spectral sensitivity Optical gain Thermal generation lifetime RoA product (2c=10 ~tm) Detectivity (2c= 10 lam,FOV=O)
>~70% Wide-band 1 ~ 1 ~s 300 f2cm2 2 • 1012 cmHz12 W-1
Narrow-band (FWHM~ 1+2 lain) 0.4 ( 30-50 wells) 10 ps 104 ~ c m 2 2 x 1()1o cmHz1,,_,W- 1
Normal incidence: no absorption <~10%
Comparisonof photon and thermaldetectorsperformance 39
The thinning also allows the detector array to stretch and accommodate the thermal expansion mismatch with the Si ROIC. At present QWIP FPAs are into a mainstream IR technology. The OWIP 2000 workshop on 2 7-29 July 2000 in Dana Point, CA, USA (see Infrared Physics and Technology, Vol. 42, No. 3-5, 2001) covered many team efforts in bringing the technology into the commercial market. 2.4.6 QWIP versus HgCdTe in the LWIR spectral region
Table 2.4 compares the essential properties of HgCdTe and QWIP devices at 77 K. Quantum efficiency HgCdTe has large optical absorption and a wide absorption band irrespective of the light polarization, greatly simplifying the detector array design. Quantum efficiency is routinely produced around 70% without anti-reflection (AR) coating and in excess of 90% with AIR coating, and is spectrally constant from less than 1 l~m out to near the cut-off of the detector. The wide-band spectral sensitivity with near-perfect rl enables greater system collection efficiency (smaller aperture) making the FPA useful for imaging, spectral radiometry, and long-range target acquisition. It should be noticed, however, that the current LWIR staring array performance is mostly limited by the charge handling capacity on the IROIC and the background (warm optics). Due to intersub-band transitions in the conduction band, the n-type QWIP detection mechanism requires photons with non-normal angle of incidence to provide proper polarization for photon absorption. The absorption quantum efficiency is relatively small, about 20% using 2D grating. Since OWIP is a photoconductive detector, the responsivity is proportional to the conversion efficiency, which is the product of the absorption quantum efficiency times the optical gain. The optical gain of OWIP structures is typically 0.4. From this it follows that the r/ is typically below 10% at the maximum response, rapidly rolling of both the short and long wavelength sides of the peak. Figure 2.21 compares the spectral r/of the HgCdTe photodiode with a QWIP. Dark current and RoA product Figure 2.22 shows typical current-voltage characteristics of HgCdTe photodiode at temperatures between 40 and 90K for a 12 l~m cut-off detector at 40K. Leakage current is less than 10 -5 A/cm 2 at 77K. The biases-independent leakage current, aids in achieving FPA uniformity as well as reducing detector bias-control requirements during changes in photocurrent. Usually for Hgl_xCdxTe photodiodes with x~().22, in the zero-bias and lowbias region, diffusion current is the dominant current down to 60K. 3 For medium reverse bias, trap-assisted tunnelling produces the dark current, and also dominates the dark current at zero bias below 50 K. For a high reverse bias, bulk band-to-band tunnelling dominates. At low temperatures, such as 40 K, large spreads in RoA product distributions are typically observed due to onset of tunnelling currents associated with localized defects. Moreover, the HgCdTe
40
Handbook of Infrared Detection Technologies
100 800
c
HgCdTe Xo = 9.2 gm
~ ~1 ~
:3 r
m=3
0
60-
,i
//
X~ = 8.5 g m
/ [FWHM - 1.4 .p...m
/
l
6 ~ c
~r 4 E ~
//
20-
r
-
0 o
2
\1
11
40-
0
8
]QWlP ]
. - - .
9 E
10
,,
P "~ "
4
i- . . . . . . . .
6
"i'-
-
""
"
8 10 Wavelength (#m)
I
12
0
14
Figure 2.21 Ouantum efficiency versus wavelength for a HgCdTe photodiode and GaAs/A1GaAs OWIP detector with similar cut-off.
photodiodes often have additional dark current, particularly at low temperature, which is related to the surface. 3 The average value of RoA product at 77 K for a 10 ~tm cut-off for HgCdTe photodiodes at 77K is around 300 ~ c m 2 and drops to 30 ~ c m 2 at 12 l~m. At 40 K, the RoA product varies between 1()~ and 10 s ~ c m 2 with 90% above 10 ~ ~2cm2 at 11.2 ll,m. 28'3~ In comparison with HgCdTe photodiodes, the behaviour of the dark current of QWIPs is better understood. At low temperatures (T < 40 K for 2,,= 10 ~m), the dark current is mostly caused by defect related direct tunnelling. In the medium operating range between 40 and 70 K (for 2,,= 10 pro), the thermally assisted tunnelling dominates. In this case, electrons are thermally excited and tunnel through the barriers with assistance from the defects and the triangle part of the barrier at high bias. At high temperature (> 70K for kc=lOlam), thermally excited electrons are thermionically emitted and transport above the barriers. It is difficult to block this dark current without sacrificing the photoelectrons (transport mechanisms of thermionically emitted current and photocurrent are similar). Minimizing thermionically emitted current is critical to the commercial success of the QWIP, as it allows the highly desirable high-temperature camera operation. Dropping the first excited state to the top, theoretically causes the dark current to drop by a factor of ~ 6 at a temperature of 70 K. This compares well with the four-fold drop experimentally observed for 9 ~m cut-off QWIPs. 4~ The value of the QWIP dark current could be adjusted using different device structures, doping densities, and bias conditions. Figure 2.23 shows the I-V characteristics for a range of temperature between 3 5 and 77 K measured on a device with 9.6 ~m spectral peak. Typical operation at 2 V applied bias in the
Comparison of photon and thermal detectors performance
10
-~
Operation bias/,,',."
/,',.":
/,,'//
-25mY
r
41
E 10-~
o
90 K 77 K
~~r 1 0-5
i/:
/," /:'
-,,1i ,' v; :
"" "~
I
I
II I, I"
~_1 ~
::3
0-7
..........
'''
o
10 .9 -0.5
......
"\.i
40 K -0.4
i
60 K
"----.--v.-: . . . . . . . . .
-0.3 -0.2 Bias voltage (V)
-0.1
"',,
0.0
0.1
Figure 2.22 Current-voltage characteristics at various temperatures for a 12 I~m cut-off HgCdTe photodiode (after ref. 22 ).
slowly-varying region of current with bias between the initial rise in current at low voltage and the later rise at high bias. Typical LWIR OWIP dark current at 7 7 K is about 10 -4 A/cm 2, which is in the n a n o a m p e r e range for 2 4 x 2 4 ~tm2 pixel. 42 Comparing Figures 2.22 and 2.23 we can see that a 9.6 pm OWIP must be cooled to 60 K to achieve leakage current comparable to a 12 pm HgCdTe photodiode operating at 2 5 degrees warmer. QWIP operates at a bias voltage from 1 to 3 V depending on the structure and periods of the devices. Using the voltage divided by the dark current density, the RoA products are usually larger t h a n 107 fZcm 2 and 104 f2cm 2 w h e n operated at 40 K and 77 K, respectively. 42 These values indicate very high impedance.
Detectivity We can distinguish two types of detector noise" radiation noise and intrinsic noise. Radiation noise includes signal fluctuation noise and background fluctuation noise. For infrared detectors, background fluctuation noise is higher compared to the signal fluctuation noise. Usually for photodiodes, shot noise is the major noise. In the case of QWIPs, the major source of noise is the dark current. Due to high dark current, Johnson noise is neglected in most cases, especially at high temperature operation. But at lower temperature and w h e n the array pixel size is smaller, Johnson noise becomes comparable to dark noise. Owing to stable surface properties, there is very little 1 ff noise observed in OWIPs. At FPA level, the pattern noise (which results from local variation of the dark current, photoresponse, and cut-off wavelength) is the major limitation to the
42
Handbook of Infrared Detection Technologies
1O~I
300 K FOV=2x
~_
Eo _
I
................
....
104
77K
t ~
/
/
,~..-'""
'""-.
.-""' 66 K
oB 10-~t "~
10 -1~
f
-5
I
-4
I
-3
1
-2
1
1
I
-'1 0 1 2 Bias voltage (V)
I
3
1
4
5
Figure 2.23 Current-voltage characteristics of a QWIP detector having a peak response of 9.6 I~m at various temperatures, along with the 300 K background window current measured at 30 K with a 180 ~ FOV (after ref. 41).
array performance, especially at low temperature. This type of noise is a nonuniformity appearing across the array, which does not vary with time and reflects the intrinsic properties of a FPA. The fixed pattern noise is smaller for QWIP arrays than for HgCdTe arrays due to their material quality and better controlled cut-off wavelength. Figure 2.24 compares the detectivities of p-on-n HgCdTe photodiodes with GaAs/A1GaAs OWIPs. The theoretical curves for HgCdTe photodiodes are calculated assuming constant cut-off wavelengths of 10 pm and 11 ~tm. The VLWIR results for HgCdTe (14.8 lam at 80 K and 16.2 pm at 40 K) and the QWIP at 16 pm show the intrinsic superiority of the HgCdTe photodiodes. HgCdTe has roughly an order of magnitude higher detectivity, although the advantage decreases as the temperature is reduced. The best example of where the QWIP could have a performance advantage, is at low temperature. As we see from Figure 2.24, the QWIP at 7.7 Bm peak wavelength offers superior performance relative to a ,~ 10. 6 l~m HgCdTe at temperature ~<45 K. Since 1987, 44 rapid progress has been made in the detectivity of long wavelength QWIPs, starting with bound-to-bound QWIPs, which had relatively poor sensitivity, and culminating in high performance bound-to-quasibound OWIPs with random reflectors. All the OWIP data is clustered between 101~ and 1011 cmHzl/a/W at about 77 K operating temperature. The above performance comparison of QWIPs with HgCdTe in the low temperature range is less profitable for photodiodes in the case of n+-p structures due to the same non-fundamental limitations (contacts, surface, SchockleyRead processes). Including the influence of tunnelling, the comparison of
Comparison of photon and thermal detectors performance
43
detectivity is more advantageous for GaAs/A1GaAs OWIPs in the spectral region below 14 ~m and at temperature below 50 K. ]4 Noise equivalent difference temperature The noise in HgCdTe photodiodes at 77 K is due to two sources; the shot noise from the photocurrent and the Johnson noise from the detector resistance. It can be expressed as
(30) where k is Boltzmann's constant and R is the dynamic resistance of the photodiode. Assuming that the integration time, "tint, is such that readout node capacity kept half full, we have 1
Af
(31)
- - - ,
2tint
and then
10.6 ~m LPE HgCdTe (1990)
1015
~
"
10
'". .. ~'_"-~"
01.
1
.>
1012
O
"-.
-,.
E
v
9 ~""".,
y--v.
TM
T-
a "r'
...... ...... 9 1991 AIGaAs QWlP (qCp=2.3%)
- 4 1 - - Est: 10 ~m MBE HgCdTe --O--- Est: 11 ~m MBE HgCdTe _A 15 p.m QWlP (Gunapala) ~ Est: 15.8 ~m MBE HgCdTe I...................10.7 ~m QWlP (Levine) h. . . . . 19 ~m QWlP (Levine) i: - HgCdTe Theory @ 11 ~m i . . . . . . . . . . HgCdTe Theory @ 10 pm various S-o-A 9-10 pm QWlP's
0 "O
(D a_
I
"
101~ 10 ~
30
", ""
",.x
'-v ~::~ .......
[] ",,
_4'0
(~]
"-
~'50
6()
,~
80
100 1 89
150
Temperature (K) Figure 2.24 LWIR detector detectivity versus temperature for GaAs/AIGaAs QWIPs and p-on-n HgCdre photodiodes (after refs. 9 and 43).
44 Handbook of Infrared Detection Technologies
~/(4_~Ta) In --
1
2qlph + ~
2tint"
(32)
At tactical background levels, the Johnson is much smaller than the shot noise from the photocurrent. In the case where the number of electrons collected in a flame is limited by the capacity of the ROIC charge well, which is often true, the signal-to-noise ratio is given by S
qNw/2r
N
(qNw) l
=~
(33)
2 q \ 2r ,] 2--r Assuming that the temperature derivative of the background flux can be written to a good approximation as O__QO=h__f_cQ aT
(34)
A:T2
and using equation (2 6), the NEDT under these conditions is NEDT -
2kT2X
(35)
hc2x/X~w
In the last two equations ~=(21+}.2)/2 is the average spectral band between 21 and 22. For a typical storage capacity of 2 x 107 electrons, 2 - 1 0 from equation (35) we have an NEDT of 19.8 mK. The same estimations can be made for OWIP. In this case negligible compared to the generation-recombination noise,
/
1
wavelength of the pm, and Tu= 300 K, the Johnson noise is and then (36)
where dark current can be approximated by Ia-Ioexp
( Eo) -~
.
(37)
In the above expressions, g is the photoconductive gain, Id is the dark current, Io is a constant that depends on the transport properties and the doping level, and Ea is the thermal activation energy, which is usually slightly less than the energy corresponding to the cut-off wavelength of the spectral response, g, Ivh, and Io should also be bias-dependent parameters.
Comparison of photon and thermal detectorsperformance 45 The signal-to-noise ratio for a storage capacity-limited QWIP is given by
s
qNw/2__ ~4
(~r~)
1
Nw
(38)
1 , ~ ~~'
and the NEDT is NEDT-
2kT~ i~ ~ h---~ ,
(39)
Comparing equations (35) and (39) we notice that the NEDT value for chargelimited OWIP detectors is better than for HgCdTe photodiodes by a factor of (2g) 1/2 since a reasonable value of g is 0.4. Assuming the same operation conditions as for HgCdTe photodiodes, the NEDT is 17.7 mK. So, a low photoconductive gain actually increases the S/N ratio and a QWIP FPA can have a better NEDT than an HgCdTe FPA with similar storage capacity. This deduction was experimentally confirm by Schneider et al. 4 s Using a photovoltaic 'low-noise' OWIP structure in which g is only 0.05, the group achieved a NEDT of 7.4 mK with 20 ms integration time and 5.2 mK with 40 ms. For a system operating in the LWIR band, the scene contrast is about 2%/K of change-in-scene temperature. Therefore, to obtain a pixel-to-pixel variation in apparent temperature of less than, e.g. 20 mK, the nonuniformity in response must be less than 0.04%. This is nearly impossible to obtain in the uncorrected response of the FPA, so a two-point correction is typically used. The nonuniformity can be different depending on the specification of operability; e.g., a higher requirement on the operability usually leads to a lower uniformity and vice versa. Typical uncorrected response nonuniformity in QWIP FPAs is 1-3% with an operability (the fraction of good pixels) greater than 99.9%. For the 1 2 8 x 1 2 8 151am array fabricated by the Jet Propulsion Laboratory (see Table 2.5), the uncorrected standard deviation is 2.4% and the corrected nonuniformity 0.05%. For recently described large 6 4 0 x 4 8 6 9 pm FPA, the uncorrected noise nonuniformity is about 6%, and after two-point correction improves to an impressive 0.04%. For the same format FPA, demonstrated by Lockheed Martin, an operability of greater than 99.98% was described. 5~ It is very hard for HgCdTe to compete with OWIP for high uniformity and operability with large array format, especially at low temperature and VLWIR. The variation of x across the Hgl_xCdxTe wafer causes a much larger spectral nonuniformity (e.g. at 77 K, a variation of Ax=0.2% gives a A2,,=0.064 Bm at 2c=5 ~tm, but A2~=0.51 ~m at 14 ~m), which cannot be fully corrected by the two or three-point corrections. Therefore, the required composition control is much more stringent for VLWIR than for MWIR. High uniformity and high operability, as shown in the above examples, demonstrate the maturity of GaAs growth and processing technology. In this
46 Handbook of Infrared Detection Technologies context, the n o n u n i f o r m i t y and operability h a v e been an issue for HgCdTe, a l t h o u g h recently published values for Sofradir and SBRC a r r a y s are as high as 99%. W h e n we c o m p a r e the p e r f o r m a n c e of both types of FPAs (see Tables 2.5 and 2.6), the a r r a y operability is h i g h e r for OWIPs, above 99.9%. In the case of OWIPs, e x t e n d i n g cut-off w a v e l e n g t h to VLWIR is relatively easier since there is little c h a n g e in material properties, g r o w t h and processing. However, a serious r e q u i r e m e n t for m a i n t a i n i n g the device p e r f o r m a n c e is to lower the o p e r a t i n g t e m p e r a t u r e . Due to lower q u a n t u m well barriers, the dark c u r r e n t of t h e r m i o n i c emission d o m i n a t e s at a lower t e m p e r a t u r e . In order to achieve e q u i v a l e n t p e r f o r m a n c e of a 1 0 ~ m cut-off OWIP at 7 7 K , the t e m p e r a t u r e needs to be cooled d o w n to 55 K for a 15 ~tm cut-off and 35 K for a 19 ~m cut-off (see Table 2.5 and Figure 2.24). Figure 2.25 c o m p a r e s the p e r f o r m a n c e of two Sofradir's HgCdTe staring arrays, sensitive b e t w e e n 7.7 ~m and 9 ~m, and 7.7 ~m and 9.5 ~m. Higher p e r f o r m a n c e with improved t e c h n o l o g y has been obtained using, on the one h a n d , a reduced dark c u r r e n t detector t e c h n o l o g y and, on the o t h e r hand, n e w r e a d o u t circuit a r c h i t e c t u r e w h i c h maximizes both c h a r g e h a n d l i n g capacity and responsivity. 51 For a c o n s t a n t t e m p e r a t u r e , the p e r f o r m a n c e is h i g h e r for 2c-9 ~m due to the fact t h a t dark c u r r e n t is lower and i n t e g r a t i o n time can be increased. In c o n s e q u e n c e , the focal plane o p e r a t i n g t e m p e r a t u r e of improved arrays with 2c=9 ~m can be increased up to 105 K and to 102 K for a r r a y with 2~=9.5 ~tm (for NEDT < 18 mK, for the given example). The next figure (Figure 2.26) shows the m e a s u r e d and estimated NEDT as a function t e m p e r a t u r e for 8.9-~m OWlP FPA. In c o m p a r i s o n with a r e p r e s e n t a t i v e HgCdTe FPA (Figure 2.2:3), this p a r a m e t e r exhibits strong
Table 2.5
Properties of JPL 9 ~tm and 15 ~tm GaAs/AIGaAs QWIP FPAs (after refs. 4 6 - 4 9 ) .
Parameter
2,,=9 Jam
Array size
256x256 (ref. 46) 38x38 28x28 2D periodic grating 8.5 8.9
320x256 (ref. 47) 3()x 30 28x28 2D periodic grating 8.5 8.9
640x486 (ref. 48) 25x25 18x 18 2D periodic grating 8.3 8.8
128x 128 (ref. 49) 50x 50 38x 38 2-D periodic grating 14.2 14.9
99.98 5.4
99.98
99.9 5.6
>99.9 2.4
0.04
0.05
2.3
3 1.6x 10 l~ (55K) 30 (45K)
Pixel pitch (~m) Pixel size (lxm) Optical coupling Peak wavelength (lxm) Cutoffwavelength, 50% (lain) Operability (%) Uncorrected nonuniformity (%) Corrected uniformity. 17-27~ Quantum efficiency (%) D*(cmHz1/2W- 1) NEDTwith f/2 optics (mK)
2,,= 15 ~m
0.03 6.4 2.0x l 0 ll (7OK) 23 {70 K)
6.9
2 . 0 x 1011
33 (70 K)
(7OK) 36 (70 K)
Comparison of photon and thermal detectors performance 47
temperature dependence. At temperatures < 70K, the signal-to-noise (SNR) ratio of the system is limited by multiplexer readout noise, and shot noise of the photocurrent. At temperatures > 70 K, temporal noise due to the QWIP's higher dark current becomes the limitation. As mentioned earlier, this dark higher current is due to thermionic emission and thus causes the charge storage capacitors of the readout circuitry to saturate. Comparing the values of NEDT parameters for both types of FPA (see also Tables 2.5 and 2.6), we can see that the performance of LWIR HgCdTe arrays is better. The well charge capacity is the m a x i m u m a m o u n t of charge that can be stored on the storage capacitor of each cell. The size of the unit cell is limited to the dimensions of the detector element in the array (of large LWIR HgCdTe hybrid array, a mismatch in the coefficient of thermal expansion between detector array and the readout can force the cell pitch to 20 pm or less to minimize lateral displacement). However, the development of heteroepitaxial growth techniques for HgCdTe on Si has opened up the possibility of cost-effectively producing significant quantities of large-area arrays t h r o u g h utilization of large-diameter Si substrates. For a 3 0 x 3 0 l~m2 pixel size, the storage capacities are limited to 1 to 5 • 107 electrons. For example, for a 5 • 107 electron storage capacity, the total current density of a detector with a 30• 30 lam a pixel size has to be smaller t h a n 2 7 pA/ cm 2 with a 33 ms integration time. 42 If the total current density is in the 1 m A / c m 2 range, the integration time has to be reduced to 1 ms. For the LWIR HgCdTe FPAs the integration time is usually below 1 O0 ps. Since the noise power bandwidth Af= 1/2t~nt, a small integration time causes extra noise in integration. Usually, LWIR OWIP FPAs using conventional ROIC typically operated at 6 0 65 K. Due to a smaller q u a n t u m efficiency of OWIP, filling the charge capacitor is not a problem at high background application. OWIP allows a longer integration Table 2.6
Performance specifications for LWIR HgCdTe FPAs (after SOFRADIR and SBRC
d a t a sheets) Parameter
SOFRADIR
Format Cut on--cut off (gm) FPA temperature (K) Detector pitch (gin) Fill factor (%) Charge handling capacity Frame rate (Hz) D* peak RMS/T~,,t/pitch (average) (cmHzX/aW-1) Pixel NETD (average) NEI (photons/cmas)(max) Typical FOV Fixed pattern noise Crosstalk (optical and electrical) (%) Array operability (%)
128 x 128 7.7-10.3 < 85 50 > 70 > 118 106 eto 300 1.1 • 1 ()11
SBRC 3 2 0 x 2 56 7.7-9.0 < 90 30
256x256 8.5-11.0 77 (upto 100) 30
12 or 36 to 4 0 0
8 x 106 e- (min) to 120
l O m K for 275 Hz
18
f/2 7% RMS 2 99
f/2
1.52x1012
99
48
Handbook of Infrared Detection Technologies
50
j/
9.5 pm / 9 Std MC techno. /
40-_
/
/
E,E
Improved
/
//
-
I-
pm
MCT techno. 9.5
,,If /
/
. / o 9 pm
m 20Z
10-
~ !
i
75
I
I
80
85
,
~'1
I
I
*
90 95 100 FPA temperature (K)
105
110
115
Figure 2.25 NEDT of 1 2 8 x 128 HgCdTe FPA (f/2 optics. ~0% well Jill. pitch ~0 l~m) as a fimction of operating temperatures (after ref. f~1 ).
time, which gives a relatively lower NEDT. However, at higher temperatures, the dark current of OWIP is high and fills the charge capacitor very quickly. The current subtraction and switched capacitor noise filtering capabilities of ROICs permit low NEDT at higher operating temperatures. In this case however, the readout circuit is complicated which limits the size of array. A goal of third-generation imagers is to achieve sensitivity improvement corresponding to NEDT of about 1 mK. From equation (2 7) we see that in a 300 K scene in the LWIR region with thermal contrast of 0.04, the required charge storage capacity is above 109 electrons. This high charge-storage density within the small pixel dimensions is probably not possible with standard CMOS capacitors. Norton et al. 52 have suggested using of stacked hybrid structures as at least an interim solution to incorporate the desired charge storage density in detector-readout-capacitor structures. Cost
The cost of a FPA depends strongly on the maturity of the technology and varies with production quantity in different companies. So far, large size LWIR FPAs are developed in R&D laboratories without mass production experience. According to Sofradir, HgCdTe experience, by continuous effort in the domain of industrialization decreased the cost of HgCdTe detectors by a factor of 5-10. 53 The cost of making high performance cooled components can be broken down into three parts of about equal weight: the chip (detector and ROIC): the dewar;
Comparison of photon and tllermal detectors performance
49
50
40~" 30-
a
o 20UJ Z
=
10
Estimated Experiment
_
0
50
I
55
1
60 65 70 Temperature (K)
75
80
Figure 2.26 NEDT of 3 2 0 • QWIP FPA ().c=8.9 I~m. f/2 optics. 50% well fill, pitch 30 I~m) as a function of operating temperature (after ref. 4 7).
integration and tests, s4 In addition, the user must add the cryogenic machine cost that is not negligible compared to those of the component. Even if the detection circuit is free of charge, the total cost would only be reduced by about ] 5-20%. This explains why the cost of PtSi and QWIP detectors is not markedly less than that of q u a n t u m detectors of the same complexity, even though the raw materials (Si or GaAs) are m u c h less than for HgCdTe. Moreover, since PtSi requires a very wide optical aperture to obtain acceptable performance, and since OWIP requires lower operating temperatures than other photon detectors, a possible reduction in the purchase price is counterbalanced by a significant increase in operating costs. HgCdTe detectors have been the centre of a major industry for the last three decades. The technology is relatively mature at MWIR but it does not fold over to LWIR. To make components with more pixels requires reducing the pitch or mastering the thinning operation needed to withstand the thermal cycling (differential thermal expansion between CdZnTe and silicon). In the future, a more advantageous approach would seem to be the use of Si substrates, which offer m a n y well-known advantages relative to bulk CdZnTe substrates (much larger available size at lower cost, a thermal expansion m a t c h to Si readout chips, higher purity, and compatibility with automated wafer processing/ handling methodology due to their superior mechanical strength and flatness). Promising results have been achieved in the SWIR and MWIR spectral region. During last four years the defect density for MWIR layers of HgCdTe grown by MBE on silicon substrates has decreased from 2 0 0 0 cm -2 to below 500 cm-2. s2 Currently MWIR arrays with pixel operability of 98% can be produced from this material. For comparison, CdZnTe material typically has operability of 99% or
50 Handbookof Infrared Detection Technologies better. Defect densities for LWIR material grown on silicon substrates continue to limit performance, but they have been reduced by an order of magnitude in the past decade. ~2.~~ In comparison with HgCdTe FPAs, the industrial experience in OWIP FPAs is lower and improvements can be expected because this technology is at a lower stage of development. The major challenge is changing the device and grating designs to improve the device performance. Because of the maturity of the GaAs growth technology and stability of the material system, no investment is needed for developing OWIP substrates, MBE growth, and processing technology. 42 Development of LWIR and multicolour HgCdTe detectors are extremely difficult, especially for low background applications. It means a lower cost in OWIP technology development and production compared with HgCdTe. Reliability In our discussion the reliability issue has been omitted due to the fact that statistical data on this subject is not available. In several applications, especially military systems high reliability is required to ensure both the success of the mission and minimal risk to the user. Two reliability challenges affect both FPAs; survival in high temperature system storage environments and withstanding repetitive thermal cycles between ambient and cryogenic temperatures. In HgCdTe as well as in OWIP FPAs the indium bumps are used to hybridize both types of detectors with a silicon multiplexer. However, certain problems can be expected in the case of QWIP arrays, since the indium bumps have many known alloys with III-V compounds. Very large FPAs may exceed the limits of hybrid reliability engineered into current cooled structures. Hybrids currently use mechanical constraints to force the contraction of the two components to closely match each other. This approach may have limits, when the stress reaches a point where the chip fractures. Three approaches offer an opportunity to resolve this issue:
9
to eliminate the thick substrate which limits the detector active region from deforming at the slower rate of the silicon readout, 9 to subdivide the array into a plurality of regions, 9 to use silicon as the substrate for growth of the detector material.
A technology for making HgCdTe photodiodes and QWIPs on silicon a substrate would be the ultimate, simply because of the vast existing silicon technology. However, major issues with this approach are: ~2 9 9 9
less area would be available for readout circuitry, microlens arrays would be required to regain the fill factor, material quality may not be adequate for low-leakage detectors, particularly for LWIR HgCdTe photodiodes, 9 silicon integrated circuits are processed on (100) oriented silicon, but e.g. the preferred orientation for HgCdTe growth on silicon is near the (211/ orientation.
Comparison of photon and thermal detectors performance
51
Less demanding approaches to the elimination of the thick detector substrate is the loophole or high density vertically integrated photodetector device structures already practised by GEC Marconi and DRS, respectively. In this approach, HgCdTe material is glued to the readout and contacts are made through the thin layer (1()-20 ~tm) after the substrate is removed (see Figure 2.16(b)). Another approach is to remove the substrate after hybridization with indium bumps. Substrate removal is standard practice with very large hybrid InSb arrays ( 1 0 2 4 • pixels). This approach has been recently adopted for OWIP arrays. 46'47's6 After epoxy backfilling of the gaps between the array and the readout multiplexer, the substrate is thinned to a very thin membrane ( ~ 1 0 0 0 A). This not only eliminates the thermal mismatch problem between the silicon readout and the GaAs based detector array, but also completely eliminates pixel-to-pixel crosstalk, and finally, significantly enhances an optical coupling of IR radiation into OWIP pixels.
Summary
LWIR QWIP cannot compete with the HgCdTe photodiode as a single device, especially at temperatures above 70 K due to fundamental limitations associated with intersub-band transitions. However, the advantage of HgCdTe is less distinct at temperatures below 50 K. Comparing photovo|taic HgCdTe and OWIP technologies, we arrive at the following conclusions"
9 two major issues that impede the performance of OWIPs should be overcome: optical conversion efficiency and dark current, 9 for HgCdTe, improvement of the array uniformity is necessary, 9 QWIP has more potential to realize VLWIR FPA operation (also with multicolour detection). The main drawbacks of LWIR QWIP FPA technology are the performance limitation for low integration time applications and the low operating temperature. Their main advantages are linked to performance uniformity and to availability of large size arrays. Next, the main drawback of LWIR HgCdTe FPA technology is the unavailability of large size arrays necessary for TV and larger formats. Several properties of OWIP such as high impedance, fast response time, long integration time, and low power consumption, comply well with requirements for fabrication of large FPAs. Due to the high material quality at low temperature, OWIP has potential advantages over HgCdTe for VLWIR FPA applications in terms of the array size, uniformity, yield and cost of the system. Three-band and four-band FPAs will be demonstrated in the near future. State of the art QWIP and HgCdTe FPAs provide similar performance figures of merit, because they are predominantly limited by the readout circuits. The performance is, however, achieved with very different integration times. The very short integration time of LWIR HgCdTe devices of typically below 300 l~s is very useful in order to freeze a scene with rapidly moving objects. OWIP
52
Handbookof Infrared Detection Technolo#ies
devices achieve, due to excellent homogeneity, an even better NEDT, however, the integration time must be 1()-1 ()() times longer for that, and typically 5-20 ms. Choice of the best technology is therefore driven by the specific needs of a system. Observation of the global market over the past several years has indicated that even HgCdTe photodiodes intrinsically exhibit higher performance than the OWIP detectors, the market tendencies for the future are: 9 HgCdTe for small formats (e.g. 128 x128), small pitch, high flame rates and low integration times, 9 OWIP for large formats (e.g. 64()• and larger), low flame rates and large integration time. Despite serious competition from alternative technologies and slower progress than expected, HgCdTe is unlikely to be seriously challenged for highperformance applications and applications requiring multispectral capability and fast response.
2.5 Dual-band FPAs Multicolour capabilities are highly desirable for advanced IR systems. Systems that gather data in separate IR spectral bands can discriminate both absolute temperature and unique signatures of objects in the scene. By providing this new dimension of contrast, multiband detection also enables advanced colour processing algorithms to further improve sensitivity above that of single-colour devices. Multispectral detection permits rapid and efficient understanding of the scene in a variety of ways. In particular, two-colour IR FPAs can be especially beneficial for threat-warning applications. By using two IR wavebands, spurious information, such as background clutter and sunglint, may be subtracted from an IR image, leaving only the objects of interest. Multispectral IR FPAs can also play many important roles in Earth and planetary remote sensing, astronomy, etc. Thus, the effective signal-to-noise ratio of two-colour IR FPAs greatly exceeds that of single-colour IR FPAs for specific applications. Currently, multispectral systems rely on cumbersome imaging techniques that either disperse the optical signal across multiple IR FPAs or use a filter wheel to discriminate spectrally the image focused on single FPA. These systems contain beam-splitters, lenses, and bandpass filters in the optical path in order to focus the images onto separate FPAs, responding to different IR bands. Also, complex alignment is required to map the multispectral image pixel for pixel. Consequently, these approaches are expensive in terms of size, complexity, and cooling requirements. At present, considerable efforts are directed to fabricate a single FPA with multicolour capability to eliminate the spatial alignment and temporal registration problems that exist whenever separate arrays are used, to simplify optical design, and reduce size. weight, and power consumption. Considerable progress has been recently demonstrated by research groups at Hughes Research
Comparison of photon and thermal detectors performance
53
Laboratory (Raytheon), 57'58 Lockheed Martin (BAE Systems), s~ DRS Infrared Technology 60 AIM,61 Rockwell ~2 and Leti ~ 3 in multispectral HgCdTe detectors employing mainly MBE (although LPE and MOCVD are also used) for the growth of a variety of devices. Also QWIP's technology demonstrates considerable progress in fabrication of multicolour FPAs. 4 7. s~.-36.64-66 It is, perhaps, the niche in which OWIPs have an intrinsic advantage due to relative ease of growing multi-band structures by MBE with very low defect density. Devices for the sequential and simultaneous detection of two closely spaced sub-bands in the MWIR and LWIR radiation have been demonstrated. 2.5.1 Dual-band HgCdTe In the back-illuminated dual-band detectors, the photodiode with longer cutof wavelength is grown epitaxially on top of the photodiode with the short cut-off wavelength. The shorter cut-off photodiode acts as a long-wavelength pass filter for the longer cut-off photodiode. Both sequential mode and simultaneous mode detectors are fabricated from multi-layer materials. The simplest two-colour HgCdTe detector, and the first to be demonstrated, is the bias-selectable n - p - n back-to-back photodiode shown in Figure 2.2 7(a). The sequential-mode detector has a single indium bump per unit cell that permits sequential bias-selectivity of the spectral bands associated with operating tandem photodiodes. When the polarity of the bias voltage applied to the bump contact is positive, the top (LW) photodiode is reverse biased and the bottom (SW) photodiode is forward biased. The SW photocurrent is shunted by the low impedance of the forward-biased SW photodiode, and the only photocurrent to emerge in the external circuit is the LW photocurrent. When the bias voltage polarity is reversed, the situation reverses: only SW photocurrent is available. Switching times within the detector can be relatively short, on the order of microseconds, so detection of slowly changing targets or imagers can be done by switching rapidly between the MW and LW modes. One bump contact per unit cell. as for single-colour hybrid FPAs, is the big advantage of the bias-selectable detector. It is compatible with existing silicon readout chips. The problems with the bias selectable device are that its construction does not allow independent selection of the optimum bias voltage for each photodiode, and there can be substantial MW crosstalk in the LW detector. Many applications require true simultaneity of detection in the two spectral bands. This has been achieved in a number of ingenious architectures described by Reine. 67 All these simultaneous dual-band detector architectures require an additional electrical contact to an underlying layer in the multijunction structure of both the SW and LW photodiode. The most important distinction is the requirement of a second readout circuit in each unit cell. Integrated two-colour HgCdTe technology has been developed for nearly a decade with a steady progression having a wide variety of pixel size ( 30-6113m), array formats ( 6 4 • up to 32()• and spectral-band sensitivity (MWIR/ MWIR, MWIR/LWIR and LWIR/LWIR). Figure 2.2 8 shows examples of spectral response from different two-colour devices. Note that there is a minimal crosstalk
54 Handbook of Infrared Detection Technologies
Figure 2.27 Cross-section views of unit cells for various l~ack-illuminated dual-band HgCdTe detector approaches: (a) bias-selectable n-p-n structure reported b!t Ra!ltheon: ~7 (b) simultaneous n-p-n design reported b!t Raytheon: ~ (c) simultaneous l ~ n - n - p reported b!t BAE S!lstems: ~9 (d) simultaneous n - l ~ p - p n design reported by Leti: 6 ~ and (e) simultaneous structure based on p-on-n junctions reported b!l Rockwell 62 (after ref. 6 7).
between the bands, since the short wavelength band absorbs nearly 1 ()()% of the shorter wavelengths. Test structure indicates that the separate photodiodes in a two-colour detector perform exactly as single-colour detectors in terms of achievable R o A product variation with wavelength at a given temperature. 2.5.2 Dual-band QWIPs
Sanders was first to fabricate two-colour. 2 56 • 2 56 bound-to-miniband OWIP FPAs in each of four important combinations: LWIR/LWIR. MWlR/LWlR, near IR (NIR)/LWIR and MWIR/MWIR-with simultaneous integration, s(~'~s
Comparison of photon and thermal detectors performance
55
A device capable of simultaneously detecting two separate wavelengths can be fabricated by vertical stacking of the different OWlP layers during epitaxial growth. Separate bias voltages can be applied to each OWIP simultaneously via the doped contact layers that separate the MOW detector heterostructures. Figure 2.29 shows schematically the structure of a two-colour stacked OWIP with contacts to all three ohmic-contact layers. The device epilayers were grown by MBE on a three-inch semi-insulating GaAs substrate. An undoped GaAs layer, called an isolator, was grown between two A1GaAs stop layers, followed by Au/ Ge ohmic contact of a 0.5 ~m thick doped GaAs layer. Next, the two OWIP heterostructures were grown, separated by another ohmic contact. The long wavelength sensitive stack (red OWIP, ;.,= 11.2 pm) is grown above the shorter wavelength sensitive stack (blue QWIP. ,;.,=8.6 lam). Each OWIP is a 20-period GaAs/AlxGal_xAs MQW stack, in which the thickness of the Si-doped GaAs OWs (with typical electron concentration 5 • 1 ()17 cm-~) and the A1 composition of the undoped AlxGal_• barriers (~ 5 5()-6()() A) is adjusted to yield the desired peak position and spectral width. The gaps between FPA detectors and the readout multiplexer were backfilled with epoxy. The epoxy backfilling provides 1.0 0.8 c 0
Q.
4.5 #m utoff
cutoff
14.9 lam\ ~-om0.8-cutoff. l / 0
0.6
Q.
~ 9 0.4
0.2 0.02
0.6-
> 9 0.4-
~ !
m
n"
cutoff
. . . .
3
J
n, 0.2 4
. . . .
Wavelength (pm)
0.0 ' ~ ' ................ 2 3 4 5 6 7 8 910 11121314
5
Wavelength (gm)
1.0-
o t) c 0
0.8
9
~11.4
pm
0.6
=> 9 0.4 rr 0.2
j
0.0 . . . . . . . . . . . . . . . . .
~
...................
2 3 4 5 6 7 8 910 11121314
Wavelength (gm)
Figure 2.28 Spectral response curves for two-colour HgCdTe detectors in various dual-band combinations of MWIR and L WIR spectral bands (after ref. 22 ).
56 Handbook of Infrared Detection Technologies the necessary mechanical strength to the detector array and readout hybrid, prior to the array's thinning process. Most OWIP arrays use 2D grating, which is very wavelength dependent so that the efficiency gets lower when the pixel size gets smaller. Lockheed Martin has used rectangular and rotated rectangular 2D gratings for their two-colour LW/ LW FPAs. Although random reflectors have achieved relatively high quantum efficiencies with a large test device structure, it is not possible to achieve the similar high quantum efficiencies with random reflectors on small FPA pixels due to the reduced width-to-height aspect relations. 47 In addition, it is difficult to fabricate random reflectors for shorter wavelength detectors relative to long wavelength detectors due to the fact that feature sizes of random reflectors are linearly proportional to the peak wavelength of the detectors. The quantum efficiency becomes a more difficult issue for OWIP multicolour FPA than for single colour. The typical operating temperature for QWIP detectors is in the region of 4 0 80K. The bias across each QWIP can be adjusted separately, although it is desirable to apply the same bias to both colours. As shown in Figure 2.30, the responsivity of both OWIPs is around 3()()-35() mA/W. It appears that the complex two-colour processing has not compromised the electrical and optical quality of either colour in the two-colour device since the peak quantum efficiency for each of the 2()-period OWIPs was estimated to be ~1()% in comparison with a normal single-colour OWIP with twice the number of periods which has a quantum efficiency of around 2 ( ) % . A pixel operability for each colour is > 9 7% in comparison with the value of > 99.9% routinely achieved for single-colour OWIPs. The NEDT value was 24 mK for the blue OWIP and 35 mK for the red OWIP. The difference was assigned to the poor transmission properties of the optics in the 11.2 Bm band. An accurate methodology is needed to design
Figure 2.29 Structure o.f two-colour stacked (_)iVIP ( ~(ft('r r(:l: 6 ~ ).
Comparison of photon and thermal detectors performance
57
the detector structure properly to meet different requirements. In the production process, the fabrication of gratings is still quite involved, and its efficiency is rather uncertain in small pixels and in pixels with thick material layers. To cover the MWIR range a strained layer InGaAs/A1GaAs material system is used. InGaAs in the MWIR stack produces a high in-plane compressive strain which enhances the responsivity. 6~.6~ The MWIR/LWIR FPAs fabricated by Sanders consist of an 8.6 ~tm GaAs/A1GaAs OWIP on top of the 4.7 l.tm strained InGaAs/GaAs/A1GaAs heterostructure. Recently, Gunapala et al. 69"7() have demonstrated the first 8 - 9 and 1 4 - 1 5 l.tm two-colour imaging camera based on a 6 4 0 x 486 dual-band OWIP FPA, which can be processed with dual or triple contacts to access the CMOS readout multiplexer. A single indium bump per pixel is usable only in the case of an interlace readout scheme (i.e., odd rows for one colour and even rows for the other colour) which uses an existing single colour CMOS readout multiplexer. However, the disadvantage is that it does not provide a full fill factor for both wavelength bands. The 6 4 0 • GaAs/A1GaAs gave excellent images with 99.7% of the LWIR pixels and 98% of VLWIR pixels working, demonstrating the high yield of GaAs technology. The estimated NEDT of LWIR and VLWIR detectors at 40 K are 36 and 44 mK, respectively. Due to BLIP, the estimated and experimentally obtained NEDT values of the LWIR detectors do not change significantly at temperatures below 65 K. The experimentally measured values of LWIR NEDT equal to 29 mK are lower than the estimated ones. This improvement is attributed to the 2D periodic grating light coupling efficiency. However, the experimental VLWIR NEDT value is higher t h a n the estimated value. It is probably a result of the inefficient light coupling in the 1 4 - 1 5 Bm region, readout multiplexer noise, and
Figure 2.30 Typical responsivity spectra at 40 K and a comnlon bias of 1.5 V. recorded simultaneously for two QWIPs in the same pixel (after ref. 6:5).
58 Handbookof Infrared Detection Technologies the noise of the proximity electronics. At 40 K the performance of both bands' detector pixels are limited by photocurrent noise and readout noise.
2.6 FPAs of thermal detectors IR semiconductor imagers use cryogenic or thermoelectric coolers, complex IR optics, and expensive sensor materials. Typical costs of cryogenically cooled imagers of around $ 5 0 0 0 0 restrict their installation to critical military applications which allow conduction of operations in complete darkness. Very encouraging results have been obtained with micromachined silicon bolometer arrays s'6'72'73 and pyroelectric detector arrays. 6'74-76 Several countries have demonstrated imagers with NEDT below 1 ()() mK, and the cost of simple systems is sometimes below $10 000. It is expected that high-performance imager system costs will be reduced to less than $1()()(), 7~ and above IR cameras will become widely available in the near future. Although developed for military applications, low-cost IR imagers are used in nonmilitary applications such as: drivers aid, aircraft aid, industrial process monitoring, community services, firefighting, portable mine detection, night vision, border surveillance, law enforcement, search and rescue, etc. 2.6.7 Micromachined silicon bolometers
The most popular thermistor material used in fabrication of the micromachined silicon bolometers is vanadium dioxide, V02. From the point of view of IR imaging application, probably the most important property of V02 is its high negative temperature coefficient of resistance (TCR) at ambient temperature, which exceeds 4% per degree for a single element bolometer and about 2% for FPA. There are two reasons for not using VOx (.x"> 2) with substantially higher temperature coefficient: first, reproducibility of properties suffers in the higher x-value films, second, heating becomes a problem with high resisitivity films. The final microbolometer pixel structure is shown in Figure 2.31. The microbolometer consists of a 0.5 l~m thick bridge of Si 3N4 suspended about 2 l~m above the underlying silicon substrate. The use of a vacuum gap of approximately 2.5 ~tm, together with a quarter wave resonant cavity between the bolometer and the underlying substrate, can produce a reflector for wavelengths near 10 ~tm. The bridge is supported by two narrow legs of Si 3N4. The Si3N4 legs provide the thermal isolation between the microbolometer and the heat-sink readout substrate and support conductive films for electrical connection. A bipolar input amplifier is normally required, and this can be obtained with biCMOS processing technology. Encapsulated in the centre of the Si3N4 bridge is a thin layer (500 A) of polycrystalline VOx. Honeywell has licensed this technology to several companies for the development and production of uncooled FPAs for commercial and military systems. At present, the compact 3 2 0 • microbolometer cameras are produced by Raytheon, Boeing, and Lockheed-Martin in the United States. The
Comparison of photon and thermal detectors performance 59
U.S. government allowed these manufactures to sell their devices to foreign countries, but not to divulge manufacturing technologies. In recent years, several countries, including the United Kingdom, Japan. Korea, and France have picked up the ball, determined to develop their own uncooled imaging systems. As a result, although the U.S. has a significant lead, some of the most exciting and promising developments for low-cost uncooled IR systems may come from non-U.S, companies, e.g., microbolometer FPAs with series p-n junction elaborated by Mitsubishi Electric. 77 This approach is unique, based on an allsilicon version of the microbolometer. The 2 4 0 • arrays of 501urn microbolometers are fabricated on industrystandard wafer (4 inch diameter)complete with monolithic readout circuits integrated into underlying silicon. Radford et al. 7~ have reported a 2 4 ( ) x 3 2 0 pixel array with 50 lam square vanadium oxide pixels and thermal time constant of about 40 ms, for which the average NETD (f/1 optics) was 8.6 mK. However, there is a strong system need to reduce the pixel size to achieve several potential benefits. The detection range of m a n y uncooled IR imaging systems is limited by pixel resolution, rather than sensitivity. Therefore, there is a strong system need to reduce the size of the pixels. Because the cost of the optics made of Ge, the standard material, depends approximately upon the square of the diameter, so reducing the pixel size reduces the cost of the optics. These reductions in optics size would have a major benefit in reducing the overall size, weight and cost of manportable IR systems. In addition the reduction in pixel size allows a significantly larger number of FPAs to be fabricated on each wafer. However, the NEDT is inversely proportional to the pixel area, thus, if the pixel size is reduced from 50 • 50 ~tm to 2 5 • 2 5 btm, and everything else remained the same, the NEDT would increase by a factor of four. Improvements in the readout electronics are needed to compensate for this. For future arrays, the f/1 NEDT performance of 25 l.tm pitch microbolometer FPAs is projected to be below 20 mK (see Figure 2.32).79'~() The development of highly sensitive 2 5 ~m microbolometer pixels, however, presents significant challenges in both fabrication process improvements and in pixel design. Microbolometer pixels
Figure 2.31 Bridge structure of Hone!twell microbolometer (after ref. 72).
60 Handbook of Infrared Detection Technologies fabricated with conventional single-level micromachining processes suffer severe performance degradation as the unit cell is reduced below 40 ~m. This problem can be mitigated to some degree if the microbolometer process capability (design rules) is improved dramatically. Table 2.7 summarizes the design and performance parameters for Raytheon's VOx microbolometers. A similar performance has been described by Altman and colleagues at BAE Systems (Lockheed Martin); 82.83 they reported a 6 4 ( ) x 4 8 0 FPA with 2 8 x 2 8 l,tm 2 pixels with NETD (f/1 optics) of below 55 mK. At present, several research programmes are focused towards enhancement of performance level in excess of 109 cmHz 1/2 W -1. It is anticipated that new materials will form the basis of the next generation of semiconductor film bolometers. The most promising material appears to be amorphous silicon (a-Si). The temperature coefficient of resistance values for a-Si, range from - 0 . 0 2 5~ -~ for doped, low resistivity films at room temperature to - 0 . 0 8 ~ - 1 for high resistivity materials. 6 So, although high TC1R values are attainable, they are accompanied by a high level of 1/f noise, s4 Properties of the films depend upon the method of preparation and the type of dopant. Amorphous hydrogenated silicon (a-Si:H) has a metastable state caused by defects arising from prolonged illumination (Staebler and Wronski effect). This is an undesirable feature that requires a specific annealing cycle during preparation (the methodology for reliability enhancement is described in ref. 8 5). If not removed, it adversely affects longterm reliability. Nevertheless, progress in the development of a-Si:H uncooled FPAs has been reported. 320 x 240 arrays with 4 5 lam pitch have been developed with an average NETD of 70 mK (for f/1 aperture and 50Hz imagery
Figure 2.,32 VOw.FPA development (after ref. 79).
Comparison of photon and thermal detectors performance 61 Table 2.7 Performance characteristics of Raytheon's VO~ microbolometers (after refs. 80 and 81)
Performance parameter
Capability (f/1 and 300 K scene)
Array configuration Pixel size (~m 2) Spectral response (~m) Signal responsivity (V/W)
320 x 240 5{}x50 8-14 > 2.5 x 1()' V/W or 50 mK/K......... < 20 < 150 p-p 1.() rms > 40
320 x 240 25x25 8--14 > 2.5xlO'V/W or 20 mK/K...... < 50 < 15() p-p 1.() rms > 1{}{}
or 25 mK/K......
>98 2{}{} 25
>98 15{} 25
>98 390 25
NEDT ~, f/1 {mK) Offset nonuniformity (mV) Output noise (mV) Intrascene dynamic range (~ f/1 (K) Pixel operability (%) Power dissipation (mW) Nominal operating temperature (~
640 x 480 25x25 8--14 > 2.SxlOTV/W
< 15() p-p 0.6 rms > 1{}{}
f r e q u e n c y ) . 85 Using a n i m p r o v e d t e c h n o l o g i c a l stack, fully c o m p a t i b l e w i t h i n d u s t r i a l process, good q u a l i t y 3 2 0 • FPA w i t h 35 ~tm pitch a n d f/1 optics h a s b e e n p r e s e n t e d w i t h NEDT close to 3 5 inK. 86 Because the typical resisitivity of a m o r p h o u s silicon films is several orders of m a g n i t u d e h i g h e r t h a n t h a t of VO• a-Si finds a p p l i c a t i o n in u n c o o l e d a r r a y s in w h i c h bias is c o n t i n u o u s r a t h e r t h a n pulsed w i t h o u t excessive J o u l e a n h e a t i n g of the FPA. ~'
2.6.2 Pyroelectric arrays The i m a g i n g s y s t e m s based on p y r o e l e c t r i c a r r a y s , u s u a l l y n e e d to be o p e r a t e d w i t h optical m o d u l a t o r s , w h i c h c h o p or defocus the i n c o m i n g r a d i a t i o n . This m a y be a n i m p o r t a n t l i m i t a t i o n for m a n y a p p l i c a t i o n s in w h i c h c h o p p e r l e s s o p e r a t i o n is h i g h l y desirable (e.g., guided m u n i t i o n s ) . Most of t h e p y r o e l e c t r i c m a t e r i a l s c o n s i d e r e d for t h e r m a l d e t e c t o r a r r a y s are the l e a d - b a s e d ' p e r o v s k i t e ' oxides s u c h as lead t i t a n a t e [PbTiO3-PT]. These m a t e r i a l s h a v e s t r u c t u r a l similarities w i t h the m i n e r a l p e r o v s k i t e (CaTiO3 }. The basic f o r m u l a is ABO3: w h e r e A is lead, O is o x y g e n a n d B m a y be one, or a m i x t u r e , of c a t i o n s e.g. lead z i r c o n a t e t i t a n a t e [Pb(ZrTi)O~-PZT], b a r i u m s t r o n t i u m t i t a n a t e [BaSrTiO3-BST], lead s c a n d i u m t a n t a l a t e [Pb{Sccj.~Ta{}.~)O3PST] a n d lead m a g n e s i u m n i o b a t e [Pb(Mgl/~Nb2/~)O~-PMN]. Often d o p a n t s are a d d e d to t h e s e basic f o r m u l a t i o n s to e n h a n c e or t u n e the m a t e r i a l properties. A b o v e Curie t e m p e r a t u r e , To, t h e s e m a t e r i a l s form a s y m m e t r i c n o n - p o l a r , cubic s t r u c t u r e . On cooling t h e y u n d e r g o a s t r u c t u r a l p h a s e t r a n s i t i o n to form a polar, ferroelectric phase. The a b o v e m a t e r i a l s c a n be f u r t h e r sub-divided into t w o g r o u p s . The ' c o n v e n t i o n a l ' p y r o e l e c t r i c m a t e r i a l s , s u c h as PT a n d PZT, o p e r a t e at r o o m t e m p e r a t u r e well below their Curie t e m p e r a t u r e w i t h o u t the need for an applied field. R e q u i r e m e n t s for d e t e c t o r t e m p e r a t u r e stabilization is m i n i m a l or c a n be
62 Handbookof Infrared Detection Technologies eliminated since there is little variation in detector performance over quite a large temperature range. It is, however, possible to operate ferroelectrics at or above To, with an applied bias field, in the mode of a 'dielectric bolometer'. This second group of materials (including BST, PST and PMN) have Tc slightly below the detector operating temperature, resulting in minimal pyroelectricity. In this case a constant dc field is applied during operation to induce a polarization and hence regain pyroelectric properties. Barium strontium titanate (BST) ceramic is a relatively well behaved material with a very high permittivity. Texas Instruments (Raython) has improved the performance of pyroelectric FPAs using a bias voltage applied to maintain and optimize the pyroelectric effect near the phase transition. ~7 Figure 2.33 shows details of the completed pyroelectric detector device structure. The 3 2 ( ) x 2 4 0 hybrid arrays with pixels 48.5 IJm are characterized by NEDT as low as 40 mK (array average) and the production average is between 70 mK and 8() mK. A demonstrated sustained production rate in excess of 5()() units per month is a small fraction of factory capacity. For the United Kingdom array programme lead scandium tantalate (PST) material has been chosen, s~ A hybrid pyroelectric/ferroelectric bolometer detector was the first to enter production, and is the most widely used type of thermal detector (in the U.S., the Cadillac Division of General Motors has pioneered this application, selling thermal imagers to the customer for just under $2()()()). 6 Although many applications for this hybrid array technology have been identified, and imagers employing these arrays are in mass production, no hybrid technological advances are foreseen. The reason is that the thermal conductance of the bump bonds is so high that the array NETD (f/1 optics) is limited to about 50 mK. The best NEDT achieved with a hybrid array is about 38 mK, which is consistent with thermal conductance of approximately 4 IJW/K. Pyroelectric array technology, therefore, is moving toward monolithic silicon microstructure technology. The monolithic process should have fewer steps and a shorter cycle time. Most ferroelectrics tend to lose their interesting properties as the thickness is reduced. However, some ferroelectric materials seem to maintain their properties better than others. This seems particularly true for PT and related materials, whereas for BST, the material does not hold its properties well in thin-film form. Thin-film ferroelectric (TFFE) detectors have the performance potential of microbolometers with minimum NEDT of about 1 mK. ~ Figure 2.34 shows NEDT calculated as a function of pixel pitch assuming a constant 30 Hz signal and a thermal time constant of about 15 ms. Reducing the pixel pitch from 50 IJm to 2 5 IJm requires a concomitant increase in thermal isolation by about a factor of four. This seriously impacts the possibility of high performance arrays with small pixels. The TFFE device approach appears remarkably similar to the VOx microbolometer structure developed by Honeywell. However, there are several key futures that distinguish it from that technology. 'ss-~ Since the device is a capacitor rather than a resistor as in a bolometer, the electrodes are located above and below the face of the pixel, are transparent and do not obscure the
Comparison of photon and thermal detectors performance
63
Figure 2. ~ ~ BST dieh'ctric l~olometer pixel ( ~(fter ref. 8 7).
active optical area. Usually, the electrical resistance of the leads can be quite large without degrading the signal-to-noise ratio, since the detector capacitance is approximately 3 pF. This enables the use of thin, poorly conducting electrode materials to minimize thermal conductance. A key feature of the design is that the ferroelectric film is self-supporting" there is no underlying m e m b r a n e necessary to provide mechanical support. In such a way, with the use of transparent oxide electrodes, the ferroelectric material can dominate thermal conductance. It is well known that absorption of IR radiation is accomplished by means of a resonant optical cavity. In a monolithic bridge structure, the cavity is located within the ferroelectric itself or in the space between ferroelectric and the ROIC. This can be realized in two ways. s9 9
9
The bottom electrode must be highly reflective, the top electrode must be semi-transparent, and the ferrolectric must be approximately 1 btm thick for optimal tuning of the cavity for 1 ()-12 btm radiation. Both electrodes must be s e m i - t r a n s p a r e n t , a reflective mirror must be present on the ROIC under each pixel, and the pixel must be located approximately 2 l.tm above the ROIC.
A key factor to performance of the ceramic thin films is the high temperature processing required in achieving the correct ferroelectric crystal phase. The TFFEs of interest are refractory, and require annealing at elevated temperatures to crystallize and develop good pyroelectric properties. Thermal treatments at temperatures that exceed about 45()~ may lead to an adverse interaction between the silicon and a l u m i n i u m interconnections. Various techniques for the deposition of thin ferroelectric films have been investigated including spin-on metal-organic decomposition, radio frequency m a g n e t r o n sputtering, dual ion
64
Handbook of Infrared Detection Technologies
102 ,. ,m9 v/'ek 101 v
E
Is UJ Z
10~
,
I I
101 100
10 ~
102 103 Thermal conductance (nW/K)
10 ~
Figure 2.34 The relationship between thermal isolation and performance of uncooh'd IR detectors in the thermal fluctuation noise limit (~ffter r~:L ,~9/.
beam sputtering, sol-gel processing, and laser ablation. Also a number of surface rapid thermal annealing techniques have been investigated to obtain optimum material response whilst leaving the underlying silicon substrate undamaged. ~{I Monolithic arrays produced to date have demonstrated poor sensitivity. Raytheon's group has demonstrated 32()x 24() array with an NEDT of about 400 mK with operability in excess of 9 5 %. Research group from DERA (UK) has developed 'integrated' and 'composite' detector technology. ~'9~1 In the first technology, the detector material was deposited as a thin film onto free standing micro-bridge structure defined on the surface of the silicon ROIC. The 'composite' technology combines elements of hybrid and integrated technologies (see Figure 2.35). Microbridge pixels are fabricated in a similar fashion to the integrated technology and are next formed onto a high density interconnected silicon wafer. The interconnect wafer uses materials that can withstand the intermediate high temperature processing stage during fabrication of thin ferroelectric films and contains a narrow conducting channel via for every pixel, permitting electrical connection to the underside. Finally, the detector wafer is solder-bump bonded to the ROIC as per the established hybrid array process. It is predicted that, using a PST film, an NEDT of 25 mK (50 Hz image rate and f/1 optics)can be achieved. 2.6.3 Thermoelectric arrays
Thermopile detectors, 91 while only of limited use for imaging applications, have a combination of characteristics that make them well suited for some low power applications. They are highly linear, require no optical chopper, and have D* values comparable to resistive bolometers and pyroelectric detectors. They
Comparison of photon and thermal detectors performance 65
operate over a broad temperature range with little or no temperature stabilization. They have no electrical bias, leading to negligible 1/f noise and no voltage pedestal in their output signal. However, much less effort has been made in their development. The reason is that their responsivity and noise are orders of magnitude less and thus their applications in thermal imaging systems require very low-noise electronics to realize their potential performance. Thermoelectric detectors found almost no use as matrix arrays in TV flame rate imagers. Instead, they are employed as linear arrays that are mechanically scanned to form an image of stationary or nearly stationary objects. The wide operating-temperature range, lack of temperature stabilization, and radiometric accuracy make thermopiles well situated for same space-based scientific imaging applications. Two-dimensional thermopile arrays have been reported only by two groups. 92'93 In both cases, a desire for low cost and manufacturability led to the use of polysilicon thermoelectric materials, which have relatively low thermoelectric figures of merit. Fote and co-workers ~4'9s have improved the performance of thermopile linear arrays by combining Bi-Te and Bi-Sb-Te thermoelectric materials. Compared with most other thermoelelectic arrays, their D* values are highest, as is shown in Figure 2.3 6. This technology has been developed to improve the performance of 2D arrays using a three-level structure with two sacrificial layers. In such way it is possible to improve fill factor and incorporate a large n u m b e r of thermocouples per pixel. Figure 2.37 shows the thermopile detector structure. The structure allows almost 100% fill factor and the model suggests that optimized detectors will have D* values over 109 c m H z l / 2 / W . Further efforts are continued to fabricate high performance 128 • 128 FPAs. 96 2.6.4 Status a n d trends of u n c o o l e d arrays
The information gathered here follows Kruse. ~ Table 2.8 illustrates the status and trends of uncooled arrays for military and commercial applications. As was
Figzlre 2.35 Schematic cross-section of the "composite" detector arra!l design (after reJ. 90).
66
Handbook of Infrared Detection Technologies
previously mentioned, microbolometer arrays containing 64()• pixels are under development. The pixel size is being reduced to 2"3 • 5 ~tm (in order to reduce the optics cost): 35 • 3 "3pm is an intermediate objective. The NEDT goal is 10 mK for high performance applications, instead low cost performance applications (e.g. security sensors) use 16()• 120 pixels, 5()• 50 pm, and NEDT of 100 mK. The commercial systems (microbolometer imagers and radiometers and ferroelectric imagers) derive from military systems that are too costly for widespread use. Imaging radiometers employ linear thermoelectric arrays operating in the snapshot mode: they are less costly than the TV-rate imaging radiometers employing microbolometer arrays. ~7~'~ In the large volume production, the cost of commercial systems will inevitably decrease (see Table 2.8). It seems likely that microbolometer FPAs will become dominant in uncooled detector technology in the near future. They do not require a radiation chopper and have very low thermal conductance which results in a high responsivity and a low NEDT. Moreover, changes in the shape of the lags of the monolithic construction allow trade-off between speed of response and NEDT. It is predicted that the performance of monolithic ferroelectric bolometers can be considerably improved, unlike hybrid pyroelectric detectors where only little development effort is under way. The hybrid pyroelelctric FPAs are not commercially available.
10 ~~
......."
9 Bi-Te/Bi-Sb-Te (this work) 9 Constantan/chromel 9 Silicon
~ .109
-
. .--"" 9 s
0 . 0 "
-
N
"I-
v
E o
V
"O 108
I
7 1 0
0
.
.
.
.
.
.
.
.
.
.
10 ~
.
.
.
.
.
I
10 ~
i
L
l ,
.
1
0 3
L L:
R e s p o n s e time (ms) Figure 2.36 Representative data from literature showing reported I)* values as a fiinction of response timeJor thin film thermopile linear arrays. The dashed line represents the Fote and ]ones results. Its slope indicates D* proportional to the square-root of response time. which is typical for thermopiles or bolometers with different geometries and the same material system (after ref. 9 ~ ).
Comparison of photon and thermal detectors performance
67
Figure 2.37 Schematic diagram of thermopile detector structure. The top diagram shows two pixels viewed from the top, with part of the left pixel cut awa!l to show the underl!tinq structure. The lower diagram shows a cross-section side view of two pixels (after ref. 9 6 ).
2.7 Conclusions The intention of this paper has been to provide investigations of the performance of both photon and thermal detectors, with emphasis on the material properties, device structure, and their impact on FPA performance, especially in the LWIR and VLWIR spectral regions. At present, HgCdTe is a widely used variable gap semiconductor and has a privileged position both in LWIR as well as VLWIR spectral ranges. Figure 2.38 shows a plot of the thermal detectivity (3()()K, ()~ FOV) versus operating temperature for the most prominent detector technologies. The thermal detectivity is used here to compare the various technologies for equivalent NETD irrespective of wavelength. The thermal D* figure of merit for photon detectors was obtained by equating the NETD of an ideal thermal detector for a given D* to the NETD of an ideal photon detector with a given D~ . The various regions P show the appropriate applications including 'low cost' uncooled therma 1 detectors, 'high performance uncooled' for night vision e n h a n c e m e n t and earth reconnaissance, 'tactical' for most imaging uses, and 'strategic' for various military-type instruments. Strategic sensors generally detect point targets, so the D* must be as high as possible within the constraint that the cooler must not cause overriding size, weight, reliability and cost issues. High performance near infrared has similar performance requirements, but can only provide a m i n i m u m of cooling because cost and weight minimization is critical. The extrinsic silicon
68
Handbookof Infrared Detection Technologies
Table 2.8
S t a t u s a n d t r e n d s o f c o m m e r c i a l u n c o o l e d arrays (after ref. 6)
Status Feature
Cost ($)
Commercial marketing of military thermal imagers Overspecified for commercial applications 320 x 2 4 0 pixel. 5 0 x 50 ~m bolometer arrays for thermal imagers 3 2 0 x 2 4 0 pixel, 50 x 50 ~tm bolometer arrays for imaging radiometers 120 x 1 pixel. 50 x 5013m thermoelectric arrays for imaging radiometers 3 2 0 x 2 4 0 pixel, 50 x 50 ~m hybrid ferroelectric bolometer a r r a y imagers for drivers' vision e n h a n c e m e n t
1 5 ()()()-2() ()()() 2() ()()()-5() ()()() 1 5 ()()() 2()()()-4()()()
Trends Development of new. low-cost thermal imagers and imaging radiometers Designed for specific commercial applications and meeting commercial requirements 160 • 120 pixel. 50 x 50 lam bolometer arrays for thermal imagers 160 • 120 pixel. 50 • 50 Bm bolometer arrays for imaging radiometers 160 • 120 pixel. 50 • 50 ~m bolometer arrays for driver's vision e n h a n c e m e n t systems 160 x 120 pixel. 50 x 50 Bm bolometer arrays for driver's vision e n h a n c e m e n t systems in extremely large volumes
< 1()()() < 5()()() < 2()()()
1 ()()()
detectors offer very high sensitivity, but at the very low operating temperature which is prohibitive in the most applications. The cryogenically cooled InSb and HgCdTe arrays have comparable array size and pixel yield at MWIR spectral band. However, wavelength tunability and high quantum efficiency have made HgCdTe the preferred material. Thus, the associated cooling and system power requirements can be optimally distributed. The monolithic PtSi Schottky barrier FPAs lead all other technologies with respect to array size (more than 1 ()~' pixels). However, the thermal mismatch barrier in hybrid FPAs has been overcome by developers fabricating InSb and HgCdTe arrays. Detector maturity is a function of accumulated experience and development effort, the complexity of the device required, and the inherent difficulty presented by the material technology. At present, HgCdTe photodiodes and BIB extrinsic silicon detectors are not fully mature. PtSi technology is mature and has reached a plateau. Other two detector technologies such as InSb and silicon bolometers are still evolving significantly as applications for larger array configurations and smaller pixel sizes continue to push the boundaries of technology. Despite serious competition from alternative technologies and slower progress than expected, HgCdTe is unlikely to be seriously challenged for highperformance applications and applications requiring multispectral capability and fast response. The recent successes of competing cryogenically cooled detectors are due to technological, not fundamental issues. The steady progress in epitaxial technology would make HgCdTe devices much more affordable in the near future. The much higher operation temperature of HgCdTe, compared to
Comparison ~f photon and thermal detectors performance
69
Figure 2. $8 Thermal D* versus operating temperature for d(~erent FI'A technologies (~(fter r~:f. 4 ~ ).
Schottky barrier devices and low-dimensional solid devices, may become a decisive argument in this case. The fundamental performance limits of HgCdTe photodiodes have not been reached yet. Continued development of the in situ vapour phase epitaxy methods (MBE and MOCVD) will allow bandgap engineering heterojunction devices of increasing quality and complexity. Also, continued development of epitaxial growth on alternative substrates such as silicon will reduce the cost of 2D arrays. Development of dual-band arrays will continue and three-band detectors will soon be demonstrated. To provide high resolution spectroscopic imaging larger HgCdTe FPAs will be used in Fourier-transform (FT) interferometers. Photodiodes will replace photoresistors for detection out to 15 btm since they are characterized by a more linear response. State of the art QWlP and HgCdTe FPAs provide similar performance figure of merit, since they are predominantly limited by the readout circuits. A low photoconductive gain actually causes a better NEDT of OWIP FPA than that of HgCdTe FPA with a similar storage capacity. The performance is, however, achieved with very different integration times. The integration time of OWIP devices is 1 0 - 1 0 0 times longer than that of HgCdTe arrays, and is typically 5-20 ms. Powerful possibilities of QWIP technology are connected with large LWIR and VLWIR FPA applications and with multicolour detection. Three-band and four-band FPAs will soon be demonstrated in near the future.
70
Handbook of Infrared Detection Technologies
Thermal detector arrays will increase in size and improve in thermal sensitivity to a level satisfying high performance applications at ambient temperature. It is supposed that the silicon microbolometer arrays and the monolithic ferroelectric arrays will capture the low-cost markets. Current uncooled bolometer FPAs have achieved NEDT of less than 1()mK (with 5()~m pixels and f/1 optics), opening the door to the use of less expensive slower optical systems. It is supposed that sales of IR thermal imaging equipment to the automobile market will begin rapidly to change the relative ratio between military/ government and commercial IR markets. Today only about 1 ()% of the market is commercial. After a decade the commercial market can grow to over 7()% in volume and 40% in value, largely connected with volume production of uncooled imagers for automobile driving.-'-' For large volume production for automobile drivers, the cost of uncooled imaging systems will decrease to below $1000. Of course, these systems will cover other segments of the transportation industry" trucks, trains, ships, barges, buses, and airplanes. For the same applications requiring uncooled detectors, the slow response speed is unacceptable. Recently, a number of concepts (e.g., a non-equilibrium device, 99 multi-junction HgCdTe photodiodes, lc1(I optical immersion) and new materials (InAsSb, InAs/GaSb-based type II superlattices) 1 have been proposed to improve the performance of photon detectors operating at near room temperature. The measurements show the possibility to achieve detectivity of ~ l x l ( ) 9 c m H z l / 2 / W at the 8 - 9 ~m range and potentially, the devices can be assembled in large FPAs. A new IR detector concept is micro electromechanical structures (MEMSs). This technology is a marriage of photolithography and mechanics. FPAs based on MEMS technology and a visible optical readout system may offer lower-cost LWIR imaging systems. ~(~1 Finally, considerable development of signal processing function into FPAs can be anticipated.
Appendix A1 HgCdTe photodiodes The Auger m e c h a n i s m is more likely to impose fundamental limitations to the LWIR HgCdTe detector performance. ~(J2 Assuming a non-degenerate statistic, the Auger generation rate is equal to
GA
.
. . . . 2r~1 + 2rl~7
2ri.tl
n+
(A1)
where n and p are the electron and hole concentrations, r ,41 i and r~t are the intrinsic Auger 1 and Auger 7 recombination times and y = r ,-17 i /r{t 1 is the ratio of Auger 7 and Auger I intrinsic recombination times, r!~1 is given by
Comparison olphoton and thermal detectors performance
.rAli
=
G (1 + bt)
(1 + 2/,)exp
(m,, , /,,,)IF1F212(kT/E,,) ~/2
1 + IJ el kT]
in seconds
71
(A2)
where # - m e ~rot, is the ratio of the conduction to the heavy-hole valence-band effective mass, ~, is the relative static dielectric constant, and F, and F2 are the overlap integrals of the periodic part of the electron wave functions. The value of IF1F21 is taken as a constant equal to ().2. The y term is of high uncertainty. According to Casselman and Petersen ~()2 for Hg~_xCdxTe over the range 0 . 1 6 ~ x ~ ( ) . 4 0 a n d 5()K~
2 r/ 217-2hctl/2
n
+ P/YJ
(A3)
The resulting Auger generation achieves its m i n i m u m for p - yl/211i. Since y > 1, the highest detectivity of Auger limited photodetectors can be achieved with a light p-type doping. The required y l / 2 n i p-type doping is difficult to achieve in practice for lowtemperature photodetectors (the control of hole concentration below the 5 x l O l s c m -3 level is difficult) and the p-type material suffers from some n o n f u n d a m e n t a l limitations such as contacts, surface, and Shockley-Read processes. These are the reasons why the low-temperature detectors are typically produced from lightly doped n-type materials. In contrast, p-type doping is clearly advantageous for LWIR and near-room-temperature detectors.
A2 InSb p h o t o d i o d e s
The standard m a n u f a c t u r i n g technique of InSb photodiodes begins with bulk ntype single crystal wafers with donor concentration about N,t-1 ()is cm-3. The Shockley-Read recombination centres are an intrinsic characteristic of n-type InSb. The electron and hole lifetimes are equal because in n-type material there is a single set of recombination centres. The electron and hole carrier lifetimes of high quality material at temperatures below 1 3()K obey the expression 1() r-
4.4 x 10 8
N,t
in seconds
(A4)
The thermal generation can be described by
Gt,, = n]
N,T
(A5)
72
Handbook of Infrared Detection Technologies
A3 GaAs/AIGaAs QWlPs The carrier generation rate from the QWs due to optical p h o n o n s may be obtained as 104 Gpl, o. -
Nowns-Z2
(A 6)
T
where Now is the n u m b e r of OWs, n+2 is the two-dimensional (2D) density of carriers originating from the excited sub-band state, and r is the lifetime of electrons in the upper state. It can be shown, that
ns2 - Lp2 (27rm * kT)
exp -
v,,-kT ~+,)
r -- 4hgpAEcoLp
(A7) (A8)
q2EphonI1
In the equations, Lp is the mean well spacing, m* is the electron effective mass, k is the Boltzman constant, Epl,o,, is the p h o n o n energy. I~ is a dimensionless integral close to 2 in the 8 - 1 2 ~m range for a quasibound upper state. Moreover, AEco is the cut-off energy defined as AEa, = hc/;.c = Vt, - El, (Vl, is the barrier energy of the OW, and ;.c is the cut-off wavelength ). 1/Ep = 1 / ~ - 1/~o ( ~ : and Eo is the dielectric permittivity at infinite and zero frequency, respectively), and
Ef. - E1 = n~lh2
(A9)
47rm*
is the Fermi energy of the well. and n~ is the 2D carrier density in the ground state. It should be noted that Gpl, does not depend on Lp. and is therefore a more f u n d a m e n t a l quantity of a OWIP. in contrast to n~e and r. The following parameters have been assumed: tn*=().()67mo. Ej,=63~o, Evl,o,,=36 meV, n~=2• 11 cm-2, andN0~v = 50.
A4 Extrinsic photoconductors At the low t e m p e r a t u r e of operation of impurity p h o t o c o n d u c t o r s (when kT << Ea and n << Nd, N~), the thermal equilibrium free-charge carrier in an n-type extrinsic semiconductor with a partially compensated singly ionized level is equal to 3 s.1 (~s Pith -- n,,,aj -- --~
Na
exp
-
Here N,. is the density of states in the conduction band. Na is the donor concentration, N,, is the compensating acceptor concentration, and E,~ is the bonding energy of the donor relative to the conduction band.
Conlparison of photon and thernlal detectors perfornlance
73
The majority carrier lifetime is determined by the density of empty (ionized) donor levels, and for low temperatures, such that n,,,,o < Na < Nd, is given by r9 -
(oTvthN,~)
(A11)
-~,
where ere is the capture cross-section for electrons into the donor level, and Vth = (8kT/zrm*)l/2 is the carrier thermal velocity. The shallow-level impurities (B and As) show typically crc~lO -11 cm 2, while the deep-level impurities (In, Au, Zn) show er,.=lO -13 cm 2. By comparison, the or,. of intrinsic photoconductors is about 10-17 c m 2" Practical values of ot for optimized extrinsic silicon photoconductors are in the range 1 0 - 5 0 c m -~. To maximize q u a n t u m efficiency, the thickness of the detector crystal should be not less than about 0.1cm. Typical q u a n t u m efficiencies are in range of 10 to 50% at the response peak. A5 Noise equivalent difference temperature Photon detectors Detectivity provides only a limited insight into the suitability of the IR detectors for specific IR systems. The objective criterion of the performance of imaging devices is the noise equivalent difference temperature (NEDT), i.e. the difference in temperature of the object required to produce an electrical signal equal to the rms noise voltage. The sensitivity of an IR system can be expressed in a variety of ways. 1~ In this paper, we follow the paper recently published by Kinch. 18 The limiting factor of detector sensitivity is the fluctuation in the relevant carrier concentration since this determines the minimum observable signal. This fluctuation is envisioned by assuming that the detector current is integrated on a capacitive node, giving a variance N1/2 __ g
.(Id -+- IB)Tint
,
(A12)
q
where g is the gain, Ia is the detector dark current, IB is the background flux current, and "tint is the integration time. The noise equivalent flux (also named as the minimum observable signal) is given by g A Q r l z i n t A -~ N 1/2
(A13)
where ~/represents the overall q u a n t u m efficiency of the detector, including the internal q u a n t u m efficiency. Using last two equations, the noise equivalent flux can be described as AO -- g(1 + Id/I~,)OB -~
N1/2
(A14)
74 Handbookof Infrared Detection Technologies Because AO=
-"
AT dQt3/dT
(A15)
so substituting above, we have for the noise equivalent difference temperature
~(1 + Ia/Ii3)Qu
NEDT -
(A16)
(dQI3/dT) N 1/2
where (dQR/dT)/QR=C, is the scene contrast. For unlimited available capacity N 1/2 varies as g, and AQ and NEDT are independent of the gain, g. However, for a limited m a x i m u m well capacity N, both bandwidth (Aft and NEDT will vary as g. These parameters depend on specific system and detector quantities such as area and bandwidth, where bandwidth is defined as Af= 1/2 r.
Thermal detectors In the case of thermal detector, the limiting temperature fluctuation is given by12.15 AT,,
\ 1 + w---2r~7J
(A1 7)
At low frequencies AT,,=(4kT2RthAf) 1/2. It appears that correlated double sampling of this limitation with an integrated time Tint, gives 18 2kT2[1 - exp(--rint/RthCth)]) 1/2 Cth
A T,, -
(AlS)
where it was assumed Af=l/RthCth. Comparing equation (A7) with a signal fluctuation change given by
ATs--
(dP)
--~ AT 1 - e x p (
(A 19 )
qn_t '~ R,,,
RthCthJ
we can obtain
2kT2 NEDT -
Cth[1 -- exp(--rint/RthCth)]
) 1/2 (dP/dT)Rth
(A20)
where (dP/dT) is the differential change in radiated power per scene temperature change in the spectral region of interest. The integration time should fulfil the condition ri,,t ,~ 2RthCth t o avoid image smearing. The m a x i m u m value of Rth is given by radiative coupling with the upper limit equal to 4 x 1 ()~ K/W for a 1 mil pixel size. For micromachined resonant cavity bolometers with f/1 optics, there is
Comparisonof photon and tlwrmaldetectorsperformance 75 a finite limit of detector thickness, typically t=3000 A. For 1 mil pixel size a volumetric specific heat is about 2 J/cm ~K.
References 1. M. Razeghi, Current status and future trends of infrared detectors, OptoElectron. Rev. 6, 1 5 5 - 1 9 4 (1998 ). 2. A. Rogalski, Infrared Detectors, Gordon and Breach Science Publishers, Amsterdam (2000). 3. A. Rogalski, K. Adamiec, and J. Rutkowski, Narrow-Gap Semiconductor Photodiodes, SPIE Press, Bellingham (2()0()). 4. A. Rogalski, Infrared detectors at the beginning of the next millennium, Sensors and Materials 12, 2 3 3 - 2 8 8 (2 ()00). 5. R.A. Wood and N.A. Foss, Micromachined bolometer arrays achieve lowcost imaging, Laser Focus World, 101-1 ()6 (June, 1993). 6. P.W. Kruse, Uncooled Thermal Imaging. Arrays. S!tstems, and Applications, SPIE Press, Bellingham (2001). 7. A. Rose, Concepts in Photoconductivit!l and Allied Problems, Interscience, New York (1963). 8. ]. Piotrowski and W. Gawron, Ultimate performance of infrared photodetectors and figure of merit of detector material, Infrared Phys. Technol. 38, 6 3 - 6 8 (1997). 9. B.F. Levine, Ouantum-well infrared photodetectors, ]. Appl. Phys. 74, 1111t81(1993). 10. ]. Piotrowski and A. Rogalski, New generation of infrared photodetectors, Sensors and Actuators A 6 7 , 1 4 6 - 1 5 2 (1998). 11. R.G. Humpreys, Radiative lifetime in semiconductors for infrared detection, InfraredPhys. 2 6 , 3 3 7 - 3 4 2 (1986). 12. P.W. Kruse, L.D. McGlauchlin and R.B. McOuistan, Elements of Infrared Technology, Wiley, New York ( 1962 ). 13. P.W. Kruse, The photon detection process, in Optical and Infrared Detectors, pp. 5-69, edited by R. J. Keyes, Springer-Verlag, Berlin (1977). 14. A. Rogalski, Comparison of the performance of quantum well and conventional bulk infrared photodetectors, Infrared Phys. Technol. 3 8, 2 9 5 - 3 1 0 (1997). 15. A. Smith, F.E. ]ones and R.P. Chasmar, The Detection and Measurement of Infrared Radiation, Clarendon, Oxford (1968). 16. R. Ciupa and A. Rogalski, Performance limitations of photon and thermal infrared detectors, Opto-Electron. Rev. 5, 2 5 7 - 2 6 6 ( 1997). 17. L. ]. Kozlowski and W. F. Kosonocky, Infrared detector arrays, in Handbook of Optics, Chap. 23, edited by M. Bass, E. W. Van Stryland, D. 1t. Williams and W. L. Wolfe, McGraw-Hill, Inc. New York ( 1995 ). 18. M.A. Kinch, Fundamental physics of infrared detector materials, ]. Electron. Mater. 2 9 , 8 0 9 - 8 1 7 (2000).
76 Handbookof Infrared Detection Technologies
19. T. Ashley and C. T. Elliott, Non-equilibrium devices for infrared detection, Electron. Lett. 2 1 , 4 5 1 - 4 5 2 (1985). 20. F. Fuchs, L. B~rkle, R. Hamid, N. Herres, W. Pletschen, R.E. Sah, R. Kiefer, and J. Schmitz, Optoelectronic properties of photodiodes for mid- and far-infrared based on the InAs/GaSb/A1Sb materials family, Proc. SPIE 4 2 8 8 , 1 7 1 - 1 8 2 (2001). 21. J. Piotrowski and A. Rogalski, Comment on 'Temperature limits on infrared detectivities ofInAs/InxGa~ _xSb superlattices and bulk HgxCdl_xTe', [J. Appl. Phys. 74, 4774 ( 1993)] ]. Appl. Phys. gO. 2 5 4 2 - 2 544 (1996). 22. P. R. Norton, Status of infrared detectors, Proc. SPIE 3 3 7 9 , 1 0 2 - 1 1 4 (1998). 23. W.F.M. Micklethwaite and A.J. Johnson, InSb: Materials and devices, in Infrared Detectors and Emitters: Materials and Devices, pp. 1 7 7 - 2 0 4 , edited by P. Capper and C.T. Elliott, Kluwer Academic Publishers, Boston (2000). 24. J. T. Wimmers and D. S. Smith, Characteristics of InSb photovoltaic detectors at 77 K and below, Proc. SPIE 3 64, 12 31- 31 ( 1983). 25. J. T. Wimmers, R. M. Davis, C. A. Niblack and D. S. Smith, Indium antimonide detector technology at Cincinnati Electronics Corporation, Proc. SPIE 9 3 0 , 1 2 5 - 1 3 8 (1988). 26. A. M. Fowler, I. Gatley, P. McIntyre, F. J. Vrba and A. Hoffman, ALADDIN, the 1 0 2 4 x 1024 InSb array: design, description, and results, Proc. SPIE 2 8 1 6 , ]50-160(1996). 2 7. J.B. Varesi, R.E. Bornfreund, A.C. Childs, W.A. Radford, K.D. Maranowski, J.M. Peterson, S.M. Johnson, L.M. Giegerich, T.J. de Lyon, and J.E. Jensen, Fabrication of high-performance large-format MWIR focal plane arrays from MBE-grown HgCdTe on 4" silicon substrates, ]. Electron. Mater. 30, 5 6 6 - 5 7 3
(2001). 28. M.B. Reine, Photovoltaic detectors in MCT, in Infrared Detectors and Emitters: Materials and Devices, pp. 2 79-312, edited by P. Capper and C.T. Elliott, Kluwer Academic Publishers, Boston (2000). 29. Y. Nemirovsky and G. Bahir, Passivation of mercury cadmium telluride surfaces, J. Vac. Sci. Technol. A 7 , 4 5 0 - 4 5 9 (1989). 30. W. E. Tennant, C. A. Cockrum, ]. B. Gilpin, M. A. Kinch, M. B. Reine and R. P. Ruth, Key issues in HgCdTe-based focal plane arrays: an industry perspective, ]. Vac. Sci. Technol. B l O , 1 3 5 9 - 3 6 9 ( 1 9 9 2 ) . 31. A. Rogalski and R. Ciupa, Long wavelength HgCdTe photodiodes: n+-on-p versus p-on-n structures, ]. App1. Phys. 7 7 . 3 5 0 5 - 3 5 1 2 (1995). 32. T.J. de Lyon, R.D. Rajavel, J.A. Vigil, J.E. Jensen, O.K. Wu, C.A. Cockrum, S.M. Johnson, G.M. Venzor, S.L. Bailey, I. Kasai, W.L. Ahlgren, and M.S. Smith, Molecular-beam epitaxial growth of HgCdTe infrared focal-plane arrays on silicon substrates for midwave infrared applications, ]. Electron. Mater. 27, 5 5 0 555(1998). 33. M. Kimata and N. Tsubouchi, Schottky barrier photoemissive detectors, in Infrared Photon Detectors, pp. 299-349, edited by A. Roga]ski, SPIE Optical Engineering Press, Bellingham (1995).
Comparisonof photon and thermaldetectorsperformance 77
34. M. Kimata, Metal silicide Schottky infrared detector arrays, in Infrared Detectors and Emitters: Materials and Devices, pp. 77-98, edited by P. Capper and C.T. Elliott, Kluwer Academic Publishers, Boston (2000). 35. N. Sclar, Properties of doped silicon and germanium infrared detectors, Prog. Ouant. Electr. 9 , 1 4 9 - 2 5 7 (1984). 36. P. R. Norton, Infrared image sensors, Opt. Eng. 30, 1 6 4 9 - 1 6 6 3 (1991). 3 7. S. Solomon, A. Tribble, N. Lum, J. Venzon, G. Domingo, A. Hofman, and M. Smith, High-background, long-wave Si:As IBC 32()x240 IR focal plane array, Proc. SPIE 2 8 1 6 , 1 6 1 - 1 6 8 ( 1 9 9 6 ) . 38. A. Rogalski, Assessment of HgCdTe photodiodes and quantum well infrared photoconductors for long wavelength focal plane arrays, Infrared Phys. rechnol. 4 0 , 2 7 9 - 2 9 4 (1999). 39. J. Bajaj, J.M. Arias, M. Zandian, J.G. Pasko, L.J. Kozlowski, R.E. De Wames, and W.E Tennant, Molecular beam epitaxial HgCdTe material characteristics and device performance: Reproducibility status, J. Electron. Mater. 24, 1 0 6 7 1076(1995). 40. S. D. Gunapala, J. K. Liu, J. S. Park, M. Sundaram, C. A. Shott, T. Hoelter, T. L. Lin, S. T. Massie, P. D. Maker, R. E. Muller, and G. Sarusi, 9 l-tm cutoff 256 x 256 GaAs/AlxGal_xAs quantum well infrared photodetector hand-held camera, IEEE Trans. Electron Devices 44, 5 1 - 57 ( 1997 ). 41. M.Z. Tidrow, J.C. Chiang, S.S Li, and K. Bacher, A high strain two-stack two-color quantum well infrared photodetector, Appl. Phys. Lett. 70, 8 5 9 - 8 6 1 (1997). 42. M.Z. Tidrow, W.A. Beck, W.W. Clark, H.K. Pollehn, J.W. Little, N.K. Dhar, P.R. Leavitt, S.W. Kennerly, D.W. Beekman, A.C. Goldberg, and W.R. Dyer, Device physics and focal plane applications of QWIP and MCT, Opto-Electron. Rev. 7 , 2 8 3 - 2 9 6 (1999). 43. L.J. Kozlowski, K. Vural, J.M. Arias, W.E. Tennant, and R.E. DeWames, Performance of HgCdTe, InGaAs and quantum well GaAs/A1GaAs staring infrared focal plane arrays, Proc. SPIE 3 1 8 2 , 2-13 ( 199 7). 44. B. F. Levine, K. K. Choi, C. G. Bethea, J. Walker, and R. J. Malik, New 10 gm infrared detector using intersubband absorption in resonant tunneling GaA1As superlattices, Appl. Phys. Lett. 50, 1 0 9 2 - 1 0 9 4 (1987). 45. H. Schneider, M. Walther, J. Fleissner, R. Rehm, E. Diwo, K. Schwarz, P. Koidl, G. Weimann, J. Ziegler, R. Breiter, and W. Cabanski, Low-noise OWIPs for FPA sensors with high thermal resolution, Proc. SPIE 4 1 3 0 , 3 5 3 - 3 6 2 (2()()()). 46. S.D. Gunapala, S.V. Bandara, J.K. Liu, E.M. Luong, N. Stetson, C.A. Shott, J.J. Bock, S.B. Rafol, J.M. Mumolo, and M.J. McKelvey, Long-wavelength 256 x 2 5 6 GaAs/A1GaAs quantum well infrared photodetector (QWIP) palm-size camera, IEEE Trans. ElectronDevices 4 7 , 3 2 6 - 3 3 2 (2()()0). 47. S.D. Gunapala, S.V. Bandara, J.K. Liu, E.M. Luong, S.B. Rafol, J.M. Mumolo, D.Z. Ting, J.J. Bock, M.E. Ressler, M.W. Werner, P.D. LeVan, R. Chehayeb, C.A. Kukkonen, M. Ley. P. LeVan, and M.A. Fauci, Recent developments and applications of quantum well infrared photodetector focal plane arrays, Opto-Electron. Rev. 8, 1 5 0 - 1 6 3 (2()() 1 ).
78 Handbookof Infrared Detection Technologies 48. S.D. Gunapala, S.V. Bandara, J.K. Liu, W. Hong, M. Sundaram, P.D. Maker, R.E. Muller, C.A. Shott, and R. Carralejo, Long-wavelength 6 4 0 • GaAs/A1GaAs quantum well infrared photodetector snap-shot camera, IEEE Trans. Electron Devices 45, 1 8 9 0 1 8 9 5 (1998). 49. S.D. Gunapala, J.S. Park, G. Sarusi, T.L. Lin, J.K. Liu, P.D. Maker, R.E. Muller, C.A. Shott, and T. Hoelter, 15-Bm 128 x 128 GaAs/AlxGa 1- xAs quantum well infrared photodetector focal plane array camera, IEEE Trans. Electron Devices 44, 4 5 - 5 0 (1997). 50. W. A. Beck and T. S. Faska. Current status of quantum well focal plane arrays, Proc. SPIE 2 7 4 4 , 1 9 3 - 2 0 6 (1996). 51. A. Manissadjian, P. Costa, P. Tribolet, G. Destefanis, HgCdTe performance for high operating temperatures, Proc. SPIE 3 4 3 6 , 150-161 (1998). 52. P. Norton, J. Campbell, S. Horn, and D. Reago, Third-generation infrared imagers, Proc. SPIE 4 1 1 0 , 2 2 6 - 2 3 6 (2()()()). 53. J.P. Chatard, LW MCT IRFPA cost optimization, Proc. SPIE 3 6 9 8 , 4 0 7 419(1999). 54. F. Bertrand, J.T. Tissot, and G. Destefanis, Second generation cooled infrared detectors. State of the art and prospects, in Physics of Semiconductor Devices, Vol. II, pp. 7 1 3 - 7 2 0 , edited by V. Kumar and S.K. Agarwal, Narosa Publishing House, New Delhi (1998). 55. T.J. DeLyon, J.E. Jensen, M.D. Gorwitz. C.A. Cockrum, S.M. Johnson, and G.M. Venzor, MBE growth of HgCdTe on silicon substrates for large-area infrared focal plane arrays: A review of recent progress. ]. Electron. Mater. 28, 7 0 5 - 7 1 1 (1999). 56. S.D. Gunapala and S.V. Bandara, Quantum well infrared photodetectors, in Handbook of Thin Devices, edited by M.H. Francombe, Vol. 2, pp. 63-99, Academic Press, San Diego (2()()()). 57. R.D. Rajavel, D.M. Jamba, J.E. Jensen, O.K. Wu, P.D. Brewer, J.A. Wilson, J.L. Johnson, E.A. Patten, K. Kasai, J.T. Caulfield, and P.M. Goetz, Molecular beam epitaxial growth and performance of integrated multispectral HgCdTe photodiodes for the detection of mid-wave infrared radiation, J. Crystal Growth 184, 1 2 7 2 - 1 2 7 8 ( 1 9 9 8 ) . 58. R.D. Rajavel, D.M. Jamba. J.E. Jensen. O.K. Wu, J.A. Wilson, J.L. Johnson, E.A. Patten, K. Kasai, P.M. Goetz and S.M. Johnson, Molecular beam epitaxial growth and performance of HgCdTe-based simultaneous-mode two-color detectors, ]. Electron. Mater. 2 7, 74 7 - 751 ( 1 9 9 8 ). 59. M. B. Reine, A. Hairston, P. O'Dette, S. P. Tobin, F. T. J. Smith, B. L. Musicant, P. Mitra, F. C. Case, Simultaneous MW/LW dual-band MOCVD HgCdTe 64 x 64 FPAs, Proc. SPIE 3 3 79.2()()-212 (1998). 60. M.A. Kinch, HDVIP | FPA technology at DRS, Proc. SPIE 4 3 6 9 , :366-578 (2001). 61. W. Cabanski, R. Breiter, R. Koch, K.H. Mauk, W. Rode, J. Ziegler, H. Schneider, M. Walther, and R. Oelmaier, 3rd gen. focal plane arrays IR detectors modules at AIM, Proc. SPIE 4 5 6 9 , 5 4 7 - 5 5 8 (2()() l ). 62. W.E. Tennant, M. Thomas, L.J. Kozlowski, W.V. McLevige, D.D. Edwall, M. Zandian, K. Spariosu, G. Hildebrandt, V. Gil, P. Ely, M. Muzilla, A. Stoltz, and J.H.
Comparison of photon and thermal detectorsperformance 79
Dinan, A novel simultaneous unipolar multispectral integrated technology approach for HgCdTe IR detectors and focal plane arrays, ]. Electron. Mater. 30, 590-594(200]). 63. J.P. Zanatta, P. Ferret, R. Loyer. G. Petroz, S. Cremer, J.P. Chamonal, P. Bouchut, A. Million, and G. Destefanis, Single and two colour infrared focal plane arrays made by MBE in HgCdTe, Proc. SPIE 4 1 3 0 , 4 4 1 - 4 5 1 (2000). 64. H.K. Pollehn and J. Ahearn, Multi-domain smart sensors, Proc. SPIE 3 6 9 8 , 4 2 0 - 4 2 6 (1999). 65. T. Whitaker, Sanders' OWIPs detect two colors at once, Compound Semiconductors 5(7), 4 8 - 5 1 (1999 ). 66. A.C. Goldberg, S.W. Kennerly, J.W. Little, H.K. Pollehn, T.A. Shafer, C.L. Mears, H.F. Schaake, M. Winn, M. Taylor, and P.N. Uppal, Comparison of HgCdTe and OWIP dual-band focal plane arrays, Proc. SPIE 4 3 6 9 , 5 3 2 - 5 4 6 (2OOl). 67. M.B. Reine, HgCdTe photodiodes for I1R detection: A review, Proc. SPIE 4288, 266-277 (2001). 68. Ph. Bois, E. Costard, J.Y. Duboz, and J. Nagle, Technology of multiquantum well infrared detectors, Proc. SPIE 3061,764-771 ( ] 997). 69. S.D. Gunapala, S.V. Bandara, A. Sigh, J.K. Liu, S.B. Rafol, E.M. Luong, J.M. Mumolo, N.Q. Tran, J.D. Vincent, C.A. Shott, J. Long, P.D. LeVan, 8-9 and 14-] 5 l~m two-color 640x486 quantum well infrared photodetector (QWIP) focal plane array camera, Proc. SPIE 3698, 687-697 (1999). 70. S.D. Gunapala, S.V. Bandara, A. Singh, J.K. Liu, B. Rafol, E.M. Luong, J.M. Mumolo, N.O. Tran, D.Z. Ting, J.D. Vincent, C.A. Shott, J. Long, and P.D. LeVan, 640x486 long-wavelength two-color GaAs/A1GaAs quantum well infrared photodetector (OWIP) focal plane array camera, IEEE Trans. Electron Devices 47, 963-971 (2000). 71. P.W. Kruse, Principles of uncooled infrared focal plane arrays, in Semiconductors and Semimetals, Vol. 4 7, pp. 17-42, edited by P. W. Kruse and D. D. Skatrud, Academic Press, San Diego ( 199 7 ). 72.1R. A. Wood, Monolithic silicon microbolometer arrays, in Semiconductors and Semimetals, Vol. 47, pp. 4 5 - 1 2 1 , edited by P. W. Kruse and D. D. Skatrud, Academic Press, San Diego ( 199 7). 73. R.A. Wood, Uncooled microbolometer infrared sensor arrays, in Infrared Detectors and Emitters: Materials atld Devices, pp. 1 4 9 - 1 7 4 , edited by P. Capper and C.T. Elliott, Kluwer Academic Publishers, Boston (2000). 74. C. M. Hanson, Hybrid pyroelectric-Ferroelectric bolometer arrays, in Semiconductors and Semimetals, Vol. 4 7, pp. 12 3-174, edited by P. W. Kruse and D. D. Skatrud, Academic Press, San Diego (1997). 75. D.L. Polla and J.R. Choi, Monolithic pyroelectric bolometer arrays, in Semiconductors and Semilnetals, Vol. 4 7, pp. 1 75-2() 1, edited by P. W. Kruse and D. D. Skatrud, Academic Press, San Diego(1997). 76. R.W. Whatmore and R. Watton, Pyroelectric materials and devices, in Infrared Detectors and Emitters: Materials alut Devices, pp. 9 9 - 1 4 7 , edited by P. Capper and C.T. Elliott, Kluwer Academic Publishers, Boston (2000).
80 Handbookof hlfrared Detection Technologies 77. T. Ishikawa, M. Ueno, K. Endo, Y. Nakaki, H. Hata0 T. Sone, and M. Kimata, Low-cost 3 2 0 • uncooled IRFPA using conventional silicon IC process, Opto-Electron. Rev. 7 , 2 9 7 - 3 ( ) 3 (1999). 78. W. Radford, D. Murphy, A. Finch, K. Hay. A. Kennedy, M. Ray, A. Sayed. ]. Wyles, R. Wyles, ]. Varesi, E. Moody, and F. Cheung. Sensitivity improvements in uncooled microbolometer FPAs. Proc. SPIE 3 6 9 8 . 1 1 9 - 1 3 ( ) (1999). 79. J. Anderson, D. Bradley, D.C. Chen. R. Chin, K. ]urgelewicz, W. Radford, A. Kennedy, D. Murphy, M. Ray, R. Wyles, ]. Brown, and G. Newsome, Low cost microsensors program, Proc. SPIE 4 3 6 9 , 5 5 9 - 5 6 5 {2()() 1 ). 80. D. Murphy, M. Ray, R. Wyles, J. Asbrock, N. Lum, A. Kennedy, J. Wyles, C. Hewitt, G. Graham, W. Radford, ]. Anderson, D. Bradley, R. Chin, and T. Kostrzewa, High sensitivity {25 Bm pitch) microbolometer FPAs and application development, Proc. SPIE 4 3 6 9 , 2 2 2 - 2 3 4 (2()() 1 }. 81. W. Radford, D. Murphy, A. Finch, K. Hay. A. Kennedy, M. Ray, A. Sayed, ]. Wyles, R. Wyles, ]. Varesi, E. Moody, and F. Cheung, Sensitivity improvements in uncooled microbolometer FPAs, Proc. SPIE 3 6 9 8 , 1 1 9 - 1 3 0 (1999). 82. M. Altman, B. Backer, M. Kohin. R. Blackwell, N. Butler, and ]. Cullen, Lockheed Martin 6 4 0 • 48() uncooled microbolometer camera, Proc. SPIE 3 6 9 8 , 137-143(1999).
83. M.N. Gurnee, M. Kohin, R. Blackwell. N. Butler, ]. Whitwan, B. Backer, A. Leary, and T. Nielson, Developments in uncooled IR technology at BAE Systems, Proc. SPIE 4 3 6 9 , 2 8 7 - 2 9 6 (2()()1). 84. B.I. Craig, R.J. Watson, and M.H. Unewisse. Anisotropic excess noise within a-Si:H, Solid-State Electrollics. 39, 8()7-812 {1996). 85. ].L. Tissot, ].L. Martin, E. Mottin, M. Vilain, ]. ]. Yon, and ].P. Chatard, 320• microbolometer uncooled IRFPA development, Proc. SPIE 4 1 3 0 , 4 7 3 - 4 7 9 (2000). 86. E. Mottin, ].L. Martin, ].L. Ouvrier-Buffet. M. Vilain, A. Bain, ].]. Yon, ].L. Tissot, and ].P. Chatard, Enhanced amorphous silicon technology for 32()• microbolometer arrays with a pitch of 35 btm. Proc. SPIE 4 3 6 9 , 25()-256 (20 01 ). 87. C. M. Hanson, Uncooled thermal imaging at Texas Instruments, Proc. SPIE 2 0 2 0 , 3 3 0 - 3 3 9 (1993). 88. R.K. McEwen and P.A. Manning, European uncooled thermal imaging sensors, Proc, SPIE 3 6 9 8 , 3 2 2 - 3 3 7 {1999). 89. ].F. Belcher, C.M. Hanson, H.R. Beratan, K.R. Udayakumar, and K.L. Soch, Uncooled monolithic ferroelectric IRFPA technology, Proc. SPIE 3 4 3 6 , 6 1 1 622(1998). 90. M.A. Todd, P.A. Manning, O.D. Donohue, A.G. Brown, and R. Watton, Thin film ferroelectric materials for microbolometer arrays, Proc. SPIE 4 1 3 0 , 1 2 8 - 1 3 9 (2000). 91. N. Teranishi, Thermoelelctric uncooled infrared focal plane arrays, in Semiconductors and Semimetals, Vol. 47, pp. 2() 3-218, edited by P. W. Kruse and D. D. Skatrud, Academic Press. San Diego i 1997 ). 92. T. Kanno, M. Saga, S. Matsumoto, M. Uchida. N. Tsukamoto, A. Tanaka, S. Itoh, A. Nakazato, T. Endoh, S. Tohyama, Y. Yamamoto, S. Murashima, N.
Comparison of photon and thermaldetectorsperformance 81
Fujimoto, and N. Teranishi, Uncooled infrared focal plane array having 128 x 128 thermopile detector elements, Proc. SPIE 2 2 6 9 . 4 5 0 - 2 5 6 (1994). 93. A.D. Oliver, W.G. Baer, and K.D. Wise. A bulk-micromachined 1024element uncooled infrared imager, Proc. 8th hit. Conf. Solid-State Sensors and Actuators (Transducers '95), and Eurosensors IX, pp. 6 3 6 - 6 3 9 1995. 94. M.C. Fote, E.W. Jones, and T. Caillat. Uncooled thermopile infrared detector linear arrays with detectivity greater than 1 ()~)cmHz 1/2/W, IEEE Trans. Electron Devices 4~, 1 8 9 6 - 1 9 0 2 (1998). 95. M.C. Fote and E.W. Jones, High performance micromachined thermopile linear arrays, Proc. SPIE 3 3 7 9 , 1 9 2 - 1 9 7 (1998). 96. M.C. Fote and S. Gaalema, Progress towards high-performance thermopile imaging arrays, Proc. SPIE 4 3 6 9 , 35()-354 (2()() 1 ). 97. P.W. Kruse, Application of uncooled monolithic thermoelectric linear arrays to imaging radiometers, in Setniconductors and Semimetals, Vol. 47, pp. 2 9 7 - 3 1 8 , edited by P. W. Kruse and D. D. Skatrud, Academic Press, San Diego (1997). 98. T. McManus and S. Mickelson, Imaging radiometers employing linear thermolelectric arrays, Proc. SPIE 3 6 9 8 , 352-36() (1999). 99. C.T. Elliott, Photoconductive and non-equilibrium devices in HgCdTe and related alloys, in Infrared Detectors alld Etnitters: Materials and Devices, pp. 2 7 9 312, edited by P. Capper and C.T. Elliott. Kluwer Academic Publishers, Boston
(2000).
l O0. J. Piotrowski, M. Grudziefl. Z. Nowak. Z. Orman, ]. Pawluczyk, M. Romanis, and W. Gawron, Uncooled photovoltaic Hg~ _• LWIR detectors, Proc. SPIE. 4 1 1 0 , 1 7 5 - 1 8 4 (2()()()). 101. T. Perazzo, M. Mao, O. Kwon, A. Majumdar, J.B. Varesi, and P. Norton, Infrared vision using uncooled micro-optomechanical camera, Appl. Ph!ls. Lett. 74, 3 5 6 7 - 3 5 6 9 (2001). 102. T. N. Casselman and P. E. Petersen, A comparison of the dominant Auger transitions in p-type (Hg,Cd)Te, Solid State Comnlutl. 33, 61:3-619 (198()). 103. R. Schoolar and E. Tenescu, Analysis ofInSb photodiode low temperature characteristics, Proc. SPIE 686, 2-11 (1986). 104. J.Y. Andersson, Dark current mechanisms and conditions of background radiation limitation of n-doped A1GaAs/GaAs quantum-well infrared detectors, J. Appl. Phys. 72, 6 2 9 8 - 6 3 0 4 ( 1 9 9 5 ) . 105. P. R. Bratt, Impurity germanium and silicon infrared detectors, in Semiconductors and Semimetals, Vol. 12. pp. 39-142. edited by R. K. Willardson and A. C. Beer, Academic Press, New York ( 19 77). 106. J.M. Lloyd, Thermal Imagillg Systems. Plenum Press. New York ( 1975 ).
This Page Intentionally Left Blank
Chapter 3
GaAs/AIGaAs based quantum well infrared photodetector focal plane arrays S. D. Gunapala and S. V. Bandara
3.1 Introduction The idea of using a q u a n t u m well to detect light can be understood by the basic principles of q u a n t u m mechanics. 1 An electron in a square q u a n t u m well is the classic particle-in-a-box of basic q u a n t u m mechanics. The electron cannot have just any energy inside the well: rather, it is constrained to reside in certain discrete energy levels, i.e., its energy is quantized. These allowed energy levels, which depend on the electron mass and the size and shape of the q u a n t u m well, can be calculated by solving the time-independent SchrSdinger wave equation. For a square q u a n t u m well, the energy levels depend, in a straightforward way, on the well dimensions (width and potential depth). At low temperatures, an electron will reside in the lowest energy level (the ground state) of the well. W h e n a photon strikes the well, it can impart its energy to the ground state electron and excite it to the next allowed energy level (the first excited state), a process called intersubband absorption. In order for this to happen, the photon energy must equal the energy separation between the ground state and the first excited state. A photon with a different energy is not absorbed because there is no allowed energy state available for the photon to excite the ground state electron. The simplest q u a n t u m well infrared photodetector (QWIP) design uses a simple square q u a n t u m well designed to hold just two states: a ground state deep inside the well, and the first excited state near the well top. A voltage bias can now be applied to sweep out the photoelectron: under constant illumination a steadystate photocurrent (a measure of the incident photon flux) thus flows t h r o u g h the detector. The spectral response of the QWIP is therefore quite narrow, its sharpness determined by the sharpness of the two energy states involved. Making a QWIP to detect light of a different wavelength is then simply
84 Handbook of Infrared Detection Technologies accomplished by changing the width and potential depth of the well in such a way that the two energy states are separated by the corresponding photon energy. Such a square quantum well can be created in the lattice-matched GaAs/ Al• material system by sandwiching a layer of GaAs between two layers of AlxGal_xAs. The bandgap of Al•215 being larger than that of GaAs, results in a square quantum well for electrons in the conduction band' the GaAs layer is the well layer; the AlxGal_xAs layers on either side are the potential barriers. The depth of the potential well can be precisely controlled by controlling the A1 mole fraction x in the AlxGal_• barrier layers. State-of-theart crystal growth techniques like molecular beam epitaxy (MBE) permit the epitaxial growth of such layers, on a typically six-inch diameter GaAs substrate. with ultra-high purity and with control of layer thickness down to a fraction of a molecular layer. This allows the width of the GaAs well layer to be precisely controlled, and the interfaces between the well and barrier layers to be made truly sharp to produce a textbook square quantum well. A controlled number of ground state electrons are provided by doping the GaAs well with Si (an n-type dopant) during the MBE growth. It is worth noting that the above design also creates a square quantum well (with the same width but a smaller potential depth) for holes in the valence band. Ground state holes can be provided by doping the GaAs wells with Be (a p-type dopant) instead of Si, resulting in a QWIP in which holes, rather than electrons, are the photocarriers. The much heavier hole has poor transport once it has been ejected out of the well by a photon, resulting in p-OWIPs being inferior in performance to n-QWIPs. We shall henceforth confine the discussion to electrons. It is important to note that the quantum well exists only in one spatial dimension: the growth direction normal to the layers. While the electron energies are quantized in this direction, the electrons are essentially free to move about in the planes of the GaAs well layers. This has important consequences for the coupling of incident light, as will be explained later. The OWIP is thus a vertical device in that there is vertical transport of the electrons, perpendicular to the layers. The quantum wells in this material system can be designed to detect light at wavelengths longer than 3 ~m (the wells cannot be made deep enough to detect shorter wavelength light). Several quantum wells (sandwiched between barriers) are usually grown stacked on top of each other to increase photon absorption. The upper limit on this number for a typical detector structure in this material system is around 50, about the number of wells and barriers that a photoelectron can traverse in an electric field without being captured by a well, downstream from the well from which the electron originated. Fewer quantum wells implies less photon absorption' more implies fewer photoelectrons collected. Such a stack of quantum wells is called a multi-quantum well (MQW) structure. The entire MOW stack is sandwiched between heavily-doped top (called the emitter) and bottom (called the collector) GaAs layers which provide electrical contacts to the device. The location of the first excited state relative to the top of the well can have a dramatic effect on device performance. By reducing the quantum well width it is
GaAs/A1GaAs based quantum ~vellinfrared photodetector focal plane arra!ts 8 5
possible to p u s h this state up above the well top and into the continuum resulting in a strong b o u n d - t o - c o n t i n u u m intersubband absorption. The major advantage of the bound-to-continuum OWIP is that the photoelectron can escape from the q u a n t u m well to the continuum transport states without being required to tunnel through the barrier. As a result, the voltage bias required to efficiently collect the photoelectrons can be reduced dramatically, thereby lowering the dark current (see below). Furthermore, since the photoelectrons do not have to tunnel through them, the AlxGa~_xAs barriers can be made thicker without reducing the photoelectron collection efficiency. Increasing the barrier thickness up to 500 A reduces the ground state sequential tunneling component of the dark current by an order of magnitude. Using all these improvements, Levine et al. 2 at AT&T Bell Laboratories successfully demonstrated the first bound-tocontinuum OWIP operating at l()13m with a dramatic improvement in performance. In addition to the photocurrent, all detectors including OWIPs produce a dark current which must be minimized to achieve high performance. The dark current is the current flowing through the detector when it is in the dark (i.e. with no photons impinging on it) and is ideally zero. In most applications the total current flowing through the detector is measured and there is no way to distinguish the dark current from the photocurrent. Though this dark current can be approximately subtracted in the image-processing electronics, a high dark current implies that the detector blinds itself even when it sees no photons: when it does see photons the image-processing electronics are swamped by the dark electrons with very little capacity left to process the photoelectrons. In QWIPs, the dark current originates from three different mechanisms. The dark current arising from the first process is due to q u a n t u m mechanical tunneling of ground state electrons from well to well through the AlxGa~_xAs barriers (sequential tunneling). This process is independent of temperature. Sequential tunneling dominates the dark current at very low temperatures ( < 30 K). The second mechanism is thermally-assisted tunneling which involves thermal excitation of a ground state electron followed by its tunneling through the tip of the barrier into the continuum energy levels. This process governs the dark current at medium temperatures. The third mechanism is classical thermionic emission (the emission of electrons over a finite potential barrier due to their finite temperature) and dominates the dark current at higher temperatures (> 45 K). Reducing the dark current due to this mechanism is critical to the commercial success of the OWIP since it enables the, highly desirable, higher temperature camera operation. The thermal generation rate of electrons is determined partly by their lifetime, which in turn is determined by the thickness of the AlxGal_xAs barriers. The dark current can be reduced by increasing the barrier thickness to the point beyond which any further increase would reduce the photocurrent (by a reduction in the efficiency of collection of photoelectrons). The thermal generation rate also depends on the number of electrons (i.e., doping density) in the well. Increasing this number increases the photocurrent" unfortunately, it also increases the dark current. The authors of this chapter found that the dark
86
Handbook of Infrared Detection Technologies
current could be reduced by m a n y orders of magnitude by lowering the well doping density, without significantly reducing the performance of the detectors. 3 For a OWIP designed to detect light at a desired wavelength, there is thus an optimum barrier thickness and well doping density which will minimize the dark current without sacrificing too much in terms of lost photocurrent. Another trick to reduce the dark current due to thermionic emission, and optimize the performance of QWIP is to use b o u l l d - t o - q u a s i b o u n d intersubband absorption (occurring when the first excited state is in resonance with the top of the well) as shown in Figure 3.1.4 This transition maximizes the intersubband absorption while maintaining excellent electron transport. The major advantage of this design lies in the fact that it increases the energy barrier to thermionic emission compared to the case of the bound-to-continuum OWIP. The energy barrier to thermionic emission is the potential height from the ground state to the well top. In the case of the b o u n d - t o - c o n t i n u u m OWIP, the energy barrier to thermionic emission is several meV less than the barrier to photon absorption (since the first excited state is several meV above the well top). In the case of the bound-to-quasibound OWIP, the two barriers are equal. If the two QWIPs are designed to detect the same photon energy, the bound-to-quasibound design thus has a several meV higher energy barrier to thermionic emission than the b o u n d - t o - c o n t i n u u m OWIP, an increase which will drop its dark current significantly (i.e., Ia :x e -:~I:/kT ~ e-1 for T= 7() K). One could push the first excited state deeper into the well to increase the energy barrier to therrnionic emission
Figure 3.1 Schematic diagram of the conduction band in a bol~nd-to-quasibound (.)V~'IP in an ex'ternall!t applied electric field. Absorption of lR photons can photoe.rcite electrons.l?om the ground state of the quantum well into the continuum, causing a photocurrent. Three dark current mechanisms are also shown: ground state tunneling ( 1 ): thermall!t assisted tunneling (2): and thermionic emission ( ~ ). The inset shows a cross-section transmission electron mierograph of a QWIP samph'.
GaAs/A1GaAs based quantum well infi'ared photodetector focal plane arra!ts
87
even further, but this would drop the photocurrent to unacceptably low levels. The bound-to-quasibound QWIP preserves the photocurrent while reducing the dark current. In order to be absorbed by the electrons in the q u a n t u m wells, the incoming light should have an electric field component in the q u a n t u m well direction, i.e., in the growth direction, normal to the layers. Only in this situation is the electric field of the light coupled to the quantized electron m o m e n t u m , enabling a photon to excite an electron and get absorbed in the process. Light being a transverse wave (whose electric field is perpendicular to the direction of travel), this selection rule means that light striking the layers normally (the most direct way to illuminate an imaging array of detectors) is not absorbed. If the light is sent through the thin (,-~500 l.tm) edge of the detector it can be absorbed since it now has a component of its electric field in the correct direction. The edge is sometimes given a 45 ~ wedge and the incident light focused normally on the polished edge facet, s Both schemes result in incident light being piped laterally t h r o u g h the detector. This clearly limits the configuration of detectors to linear arrays and single elements. For imaging, it is necessary to be able to couple light uniformly to two-dimensional arrays of detectors. An elegant way to couple normally incident light to an imaging array of OWIPs is to use a grating to bend some of the light inside the detectors. This is accomplished by putting a special reflector on the detector top and illuminating the detector from the back. If the reflector is a smooth mirror, it is useless: the normally incident light passes through the detector, strikes the mirror, and is reflected straight back out of the back side with none of the light being absorbed by the confined electrons. To be useful, the mirror has to be rough (on the scale of the wavelength of the light in the detector's GaAs material). This roughness may be either periodic r or r a n d o m 7 (Figure 3.2). A rough mirror scatters or sprays the incident light back in a cone (i.e., the roughness ensures that the angle of reflection no longer equals the angle of incidence). The details of the cone depend on the details of the roughness. This cone now strikes the bottom side. Those rays that are within a critical angle of the normal (17 ~ for the GaAs-air interface) refract or escape back into the air. The rest suffer total internal reflection with the back surface acting as a smooth mirror. The internally reflected rays are once again reflected off the top rough mirror. W h a t
Figure 3.2 (a) Two-dimensional light coupling grating on ()WIP pixel. (b) OWIP pixel with a random reflector.
88 Handbookof Infrared Detection Technologies happens next depends on whether the roughness of the top mirror is periodic or random. If it is periodic, the top mirror will scatter or bend these rays so that they are all normal to the q u a n t u m well layers again. These rays pass t h r o u g h the detector and out of the back side. A randomly roughened mirror, on the other hand, will randomly reflect or scatter all the rays internally reflected on to it from the bottom side each time, thereby allowing the incident light to bounce back and forth between the detector top and back surfaces several times. Only light within a 17 ~ (from normal) cone escapes out of the back side. Clever design can reduce the amount of light in the escape cone but cannot eliminate it altogether. For instance, if the random reflector is designed with two levels of rough surfaces having the same areas but located a quarter wavelength (;.c;~A~/4) apart, the normally reflected light intensities from the top and bottom surfaces of the reflector are equal and 180 ~ out of phase. This maximizes the destructive interference at normal reflection and lowers light leakage through the escape cone. On each pass through the detector, following the first reflection off the top rough mirror, some of the light is absorbed by the confined electrons since most of the light rays now have an electric field component in the correct direction (normal to the q u a n t u m well layers). A periodically roughened mirror thus yields two useful passes; a randomly roughened mirror offers several. One measure of the resulting efficiency of the detector in absorbing photons is its responsivity (the photocurrent for a given incident photon flux). Under identical conditions, a top reflector with one-dimensional periodic roughness will give a QWIP about the same responsivity that it would have if illuminated through a 45 ~ edge facet; both the reflector with two-dimensional periodic roughness as well as the random reflector, will approximately double this responsivity. Squeezing the m a x i m u m light trapping ability from the reflector requires increasing the detector aspect ratio, i.e.. the ratio of detector diameter to height, a design feature accomplished by thinning the ,-~5()() ~m thick GaAs substrate on top of which the detector material is grown, to sub-micron level. The resulting optical cavity is calculated and measured to improve the responsivity of a OWIP with a random reflector significantly over the 4 5 ~ case. s Thinning enhances the responsivities of the OWIPs with 1D and 2D periodically rough reflectors to be respectively about 2 and 4 times better than the 4"5~ case. The random reflector is fabricated on the OWIPs using standard photolithography and selective dry etching. The advantage of the photolithographic process over a completely random process is the ability to accurately control feature size and preserve the pixel-to-pixel uniformity necessary for very sensitive imaging focal plane arrays. Due to this high performance and the excellent uniformity of GaAs-based OWIPs, several groups have demonstrated two-dimensional long wavelength IR (LWIR) imaging arrays ~-2s operating at temperatures up to 77K. Infrared imaging systems that work in the 3 - 1 6 ~ m band have many applications. including night vision, surveillance, navigation, flight control, early warning systems, 26 Earth and planetary remote sensing, 2" industrial product quality control, etc. Furthermore, this technology has introduced novel thermographic techniques into medicine, which is clearly evident by the recent approval of a
GaAs/A1GaAs based quantum well infrared photodetector focal plane arrays
89
OWIP-based breast cancer detection instrument by the US Food and Drug Administration, for clinical use.
3.2 Detectivity D* comparison The blackbody detectivity D~ is basically the signal-to-noise ratio of a radiation detector normalized to unit area and operating bandwidth of the detector is given by, Ax/-A-~-f
(1)
i. with RB = ~I21R(2)W(2)d2
(2)
~/'IW (2) d2 where the responsivity R, can be written in terms of q u a n t u m efficiency and photoconductive gain g as (3)
a = (e/hv)o,,g
The photoconductive gain of QWIps can be written as (4)
g=L/l
where L is the hot electron mean free path and 1 is the thickness of the MOW region, and the temporal noise current in of a single element radiation detector is given by,
(5) where e is the charge of an electron, fl=2 for a photovoltaic detector (generation only) and fl=4 for a photoconductor (generation and recombination), the phoconductive gain #=1 for a photovoltaic detector and g(QWIP) is typically 0 . 2 - 0 . 5 (depends on the device structure). Ix~is the detector dark current and Ip, the detector photocurrent, is given by Ip = e~7,~g*A
(6)
where ~ is the photon flux. (See references 18 and 2 5 for details) and (dPR/dT) is the change in the incident integrated blackbody power in the spectral range of detector with temperature. The integrated blackbody power PR, in the spectral range from k I to k2, can be written as PB - Asin 2
cos4>
'J).l
W(2)d2
(7)
90
Handbook of Infrared Detection Technolo~lies
where 0, 4~, and W (2) are the optical field of view, angle of incidence, and blackbody spectral density respectively, and are defined by equations ( 8 ) and (9)
P(2) - W(2)sin2 (O/2 )AFcos(/)
(8)
where A is the detector area, 4~ is the angle of incidence, 0 is the optical field of view angle (i.e., sin 2 (0/2)=(4f2+1) --1 where f is the f number of the optical system; in this case 0 is defined by the radius p of the blackbody opening at a distance D from the detector, so that tan(0/2)=p/D), F represents all coupling factors and F = TI(1 - r)C where Tt is the transmission of filters and windows, r--28% is the reflectivity of the GaAs detector surface, C is the optical beam chopper factor (C=0.5 in an ideal optical beam chopper), and W(2) is the blackbody spectral density given by the following equation (i.e., the power radiated per unit wavelength interval at wavelength ~ by a unit area of a blackbody at temperature TR).
W().) - (27rc2h/)~ 5) (e hc/2kT~ - 1) -1
(9)
Let us consider a background limited condition. At this condition
ID < Ie
( 1 O)
By combining equations (1). (3), (4). (5), (6). and (9) the detectivity D* can be written as,
D* - G 1 v/rla ~,
(11)
Thus, D*IDEAL
/
D*OWlp
/ 2 r/a,IDEAL V/ r/~;a-~IP
(12)
The lowest absorption q u a n t u m efficiency (1/,,) of QWIP is typically 15% (including 30% reflection loss). The ~/,, of an ideal detector is 70% (assume 30% reflection loss). Thus equation (13) reduces to ~
OIDEAL D*o_wlP
=2.67
(13)
Thus, this analysis clearly shows the photoconductive gain is irrelevant at background limited performance (BLIP) condition, and therefore, the detectivities scales solely as a function of absorption q u a n t u m efficiencies of the detectors.
GaAs/AIGaAs based quantum well infrared photodetector focal plane arra!ls 91
3.3 Effect of nonuniformity The general figure of merit that describes the performance of a large imaging array is the noise equivalent temperature difference (NEAT). NEAT is the m i n i m u m temperature difference across the target that would produce a signalto-noise ratio of unity and it is given by 1r 17 NEAT -
Ax/-A-Af
(14)
D*R(dPn/dT)
Before discussing the array results, it is also important to understand the limitations on the FPA imaging performance due to pixel nonuniformities. ~s The total noise I,, of a focal plane array is given by,
i,e, - z,'; +
+ I,,) 2
(15)
Where u is the nonuniformity of the focal plane array, is given by, (7
(9"
#
Ip + ID
(16)
Where # is the mean total signal and a is the standard deviation of the histogram of total signal versus n u m b e r of pixels. Now the focal plane array detectivity or NEAT can be obtained by following equations,
DI~I,A =
(17)
In
NEATFpA =
v/AB D;,pa(dPR/dr)
(18)
Where I,, is the total noise of the focal plane array and it is given by equation (14). The figures of merit such as D*, NEAT, NEP, NEI, etc. are different representations of the basic signal-to-noise ratio of radiation detectors normalized in different ways. The signal-to-noise ratio of a focal plane array can be written as,
SNR -
IZ -
I,
Ie
V/i + ,,a (I,, + I1,)2
(19)
Under background limited condition (use equation (12)) this reduces to, SNR ~-
1 U
(20)
92 Handbookof Infrared Detection Technoloqies
This analysis clearly shows the importance of the array uniformity in the focal plane array total signal-to-noise ratio. This point has been discussed in detail by Shepherd 19 for the case of PtSi infrared FPAs 2(~ which have low response, but very high uniformity. The general figure of merit to describe the performance of a large imaging array is the noise equivalent temperature difference NEAT, including the spatial noise which has been derived by Shepherd, 2~Jand given by NEAT -
N,, dNb/dTb)
(21)
where Tb is the background temperature, and N,, is the total number of noise electrons per pixel, given by -
N
+
+
N, f
(22)
The photoresponse independent temporal noise electrons is Nt, the shot noise electrons from the background radiation is N~,, and residual nonuniformity after correction by the electronics is u. The temperature derivative of the background flux can be written to a good approximation as dNb = hcNb drb k)or 2
(2 3)
b
where )~=(),1+22)/2 is the average wavelength of the spectral band between 2a and ,<2. When temporal noise dominates. NEAT reduces to equation (14). In the case where residual nonuniformity dominates, equations (21) and (23) reduce to NEAT -
u~T 2 b 1.44
(24)
The units of the constant is cm K, 7. is in cm and Tt, is in K. Thus, in this spatial noise-limited operation NEATvcu and higher uniformity means higher imaging performance. Levine is has shown as an example, taking Tt,=300 K, 2=10 pm, and u-0.1% leads to NEAT=63 mK, while an order of magnitude uniformity improvement (i.e., u=O.Ol%) gives NEAT-6.3 mK. By using the full expression equation (11) Levine is has calculated NEAT as a function of D* as shown in Figure 3.3. It is important to note that when D*>~IO 1(~ cmv/Hz/W, the performance is uniformity limited and thus essentially independent of the detectivity, i.e., D* is not the relevant figure of merit. 2
3.4 640x512 pixel long-wavelength portable QWIP camera In order to detect LWIR radiation, we have designed the following MQW structure. Each period of the multi-quantum well (MQW) structure consists of a
GaAs/A1GaAs based quantum well infrared photodetector focal plane arrays
,oolL \
::X) ~" \ I t IL
,
;k
= lOpm
93
,
U-- 10 .3
u = 10 .4 ,,,
10 9
10~o
~
1011
1012
DETECTIVITY (cmd'--Hz/W)
Figure 3.3 Noise equivalent temperature difference NEAT as a function of detectivity D*. The effects of nonuniformity are included for u - 1 O- ~ and 10 4. Note that for D* > 101~ cm v/Hz/W detectivity is not the relevant figure of merit for FPAs. (Taken from ref. 18).
45 A well of GaAs (doped n - 5 • 1017 c m - ~) and a 500 A barrier of Alo. 3Gao.7As. Stacking m a n y identical q u a n t u m wells (typically 50) together, increases photon absorption. Ground state electrons are provided in the detector by doping the GaAs well layers with Si. This photosensitive MOW structure is sandwiched between 0 . 5 p m GaAs top and bottom contact layers doped n - 5 x l ( ) ]7 cm -3, grown on a semi-insulating GaAs substrate by molecular beam epitaxy (MBE). Then a 0.7 pm thick GaAs cap layer on top of a 300 A AI().3Ga().rAs stop-etch layer was grown in s i t u on top of the device structure to fabricate the light coupling optical cavity. The MBE grown material was tested for absorption efficiency using a Fourier Transform Infrared (FTIR) spectrometer. Figure 3.4 shows the measured absorption q u a n t u m efficiency of this material at room temperature. The epitaxially grow material was processed into 200 pm diameter mesa test structures ( a r e a = 3 . 1 4 • 1() -4 cm 2) using wet chemical etching, and Au/Ge ohmic contacts were evaporated onto the top and bottom contact layers. The detectors were back illuminated t h r o u g h a 45 ~ polished facet s and a responsivity spectrum is shown in Figure 3.5. The responsivity of the detector peaks at 8.5 pm and the peak responsivity (Rp) of the detector is 83 mA/W at bias VB -- --1.1 V. The spectral width and the cutoff wavelength are A 5 . / 5 . - 1()% and ,;.,.- 8 . 9 p m respectively. The measured absolute peak responsivity of the detector is small, up to about V R - - 0 . 5 V. Beyond that it increases nearly linearly with bias reaching Rp - 4 2 0 mA/W at V13 - - 5 V. This type of behavior of responsivity versus bias is typical for a bound-to-quasibound QWIP. The peak q u a n t u m efficiency was 1.4% at bias VR - - 1 . 1 V for a 45 ~ double pass. The lower q u a n t u m efficiency is due to the lower photoconductive gain at lower
94
Handbook of Infrared Detection Technologies 12 10 ILl O
8
e.
.2 L
6
O
"
4
6
7
8
9
10
11
W a v e l e n g t h (~Lm)
Figure ].4 Absorption quantmn e .~cienc!l o.lthe Qi~'II" material at room temperature.
0.1
B i a s = -1.1 V
A
<: ~.,
, m
(n 'o r u) (1)
0.08 0.06 0.04 0.02
-
0 6
7
8
Wavelength
9
10
(micron)
Figure 3.5 Responsivity spectrmn of a bound-to-quasibozind I.i~IR QiI'II ~ test structure at temperature T = 7 7 K. The spectral response peak is at Y,. ~ 1~m and the lon[l waveh, n#th t'lit-ofli.~ at 8.9/3 m.
operating bias, which is required to suppresses the dark current. Due to lower readout multiplexer well depth {i.e.. 1 1 x 1 ()~' electrons) a lower dark current is m a n d a t o r y to achieve a higher operating temperature and longer integration times. In BLIP condition the noise equivalent temperature difference NETD improves with the increasing integration time. In this case, the lowest operating temperature of 6 5 K was determined by the cooling capacity of the small Sterling cooler used in an Indigo Phoenix @ camera. The peak detectivity is defined as D~,- Rpx/-A-B/i,,, where Rp is the peak responsivity, A is the area of the detector and A = 3.14 x 1 ()-4 cm 2. The measured peak detectivity at bias V B = - I . 1 V and temperature T = 6 5 K is l x l ( ) ~ cmv/Hz/W. Figure 3.6 shows the bias dependence of peak detectivity as a function of temperature. These detectors show BLIP at bias V~ = - 1 . 1 V and temperature T = 72 K for 3()() K background with f~2 optics. Although r a n d o m reflectors have achieved relatively high q u a n t u m efficiencies with large test device structures, it is not possible to achieve the similar high q u a n t u m efficiencies with random reflectors on small focal plane array pixels due to the reduced width-to-height aspect ratios. In addition, it is
GaAs/A1GaAs based qttantum well infrared photodetector focal phme arra!ls
95
1.E+12 ~" 1.E+11 3 N ~:
1.E§
1 .E+09
-
60
300 K f/2
Background Optics Bias
-1.1 V
70
80
90
Temperature (K) Figure 3.6 Detectivit!t as a function of temperat ures at bias of 1.2 V.
difficult to fabricate r a n d o m reflectors for shorter wavelength detectors relative to very long-wavelength detectors (i.e., 1 5 [tm) due to the fact that feature sizes of r a n d o m reflectors are linearly proportional to the peak wavelength of the detectors. For example, the m i n i m u m feature size of the r a n d o m reflectors of 15 ~m cutoff and 9 btm cutoff FPAs were 1.2 5 and ().6 l~m respectively and it is difficult to fabricate sub-micron features by contact photolithography. As a result, the r a n d o m reflectors of the 9 ~m cutoff FPA were less sharp and had fewer scattering centers compared to the random reflectors of the 1 5 ~m cutoff QWIP FPA. Alternatively, r more IR light can be coupled to the OWlP detector structure by incorporating a two-dimensional periodic grating surface on top of the detectors which also removes the light coupling limitations and makes twodimensional OWIP imaging arrays feasible. This two-dimensional grating structure was fabricated on the detectors by using standard photolithography and CCI2F2 selective dry etching. After the 2D grating array was defined by the lithography and dry etching, the photoconductive OWIPs of the 64()x 512 FPAs were fabricated by wet chemical etching through the photosensitive GaAs/AlxGa~_xAs m u l t i - q u a n t u m well layers into the 0.5 ~tm thick doped GaAs bottom contact layer. The pitch of the FPA is 2 5 l.tm and the actual pixel size is 2 3 • 2 3 btme. The 2D gratings on top of the detectors were then covered with Au/Ge and Au for Ohmic contact and reflection. Figure 3.7 shows twelve processed OWIP FPAs on a three-inch GaAs wafer. Indium bumps were then evaporated on top of the detectors for Si readout circuit (ROC) hybridization. A single ()WIP FPA was chosen and hybridized (via indium bump-bonding process) to a 64()• 512 CMOS multiplexer (ISC 980 3) and biased at V/~ - - 1 . 1 V. At temperatures below 72 K, the signal-to-noise ratio of the system is limited by array non-uniformity, multiplexer readout noise, and photo current (photon flux) noise. At temperatures above 72 K, temporal noise due to the OWIP's higher dark current becomes the limitation. As mentioned earlier this higher dark current is due to thermionic emission and thus causes the charge storage capacitors of the readout circuitry to saturate. Since the QWIP is a high impedance device, it should yield a very high charge injection coupling
96 Handbook of Infrared Detection Technologies
Figure 3.7 Tweh,e 6 4 0 x ~ 12 Oi~'IP focal plane arra!!s on a three-inch GaAs wr
efficiency into the integration capacitor of the multiplexer. In fact, Bethea et al. ~ have demonstrated charge injection efficiencies approaching 9()%. Charge injection efficiency can be obtained from. '~
gmRDet r]inj -- I --1-gmRDet
1+
,1
j~C'DetRD,,t
where g,,, is the transconductance of the MOSFET and it is given by g,,, - elD,.t/kT. The differential resistance Rl2,.t of the pixels at - 1.1 V bias is 4.3 • 1 () 1(~ ohms at T - 65K and detector capacitance Ci~,.t is 3.()• l()-14 F. The detector dark current IDet = 1 . 5 pA under the same operating conditions. According to equation (25) the charge injection efficiency rli,,i - 9()% at a frame rate of 30 Hz. The FPA was back-illuminated through the fiat thinned substrate membrane (thickness ..~1300 A). This initial array gave excellent images with 99.92% of the pixels working (number of dead pixels ~2 5()), demonstrating the high yield of GaAs technology. The operability was defined as the percentage of pixels having noise equivalent differential temperature less than 1 ()()mK at 300 K background and in this case operability happens to be equal to the pixel yield.
GaAs/A1GaAs based quantum well infrared photodetector focal plane arraz3s 97
We have used the following equation to calculate the noise equivalent temperature difference NETD of the FPA. NETD -
v/AB D*B(dPB/dT)sin2(0/2)
(26)
where D~3 is the blackbody detectivity, dP~/dT is the derivative of the integrated blackbody power with respect to temperature, and 0 is the field of view angle (i.e., sin2(0/2)=(4f2+ 1)-1, where f is the f number of the optical system). Figure 3.8 shows the NETD of the FPA estimated (using equation (26)) from test structure data as a function of temperature for bias voltages Vl3 = - 1.1 V. The background temperature TI3 = 300 K, the area of the pixel A=(2313m) 2, the f number of the optical system is 2, and the flame rate is 30 Hz. Figure 3.9 shows the measured NETD of the FPA at an operating temperature of T=65 K, 16 ms integration time, bias V R = - I . 1 V for 300 K background with f/2 optics and the mean value is 24 mK. This agrees reasonably with our estimated value of 10 mK based on test structure data. The data taken from a test setup has shown mean NETD < 10 mK (the higher NETD of the portable imaging camera is due to the system noise, which includes the electronic noise and the Stirling cooler noise) at an operating temperature of T=65K and bias V ~ = - I . 1 V. for a 3OOK background. The uncorrected photocurrent non-uniformity (which includes a 1% non-uniformity of the ROC and a 1.4% non-uniformity due to the cold-stop in front of the FPA not yielding the same field of view to all the pixels) of the 327 680 pixels of the 6 4 0 x 5 1 2 FPA is about 14% {=sigma/mean). The non-uniformity after twopoint (17 ~ and 27 ~ Celsius) correction improves to an impressive 0.3%. It is worth noting that these data were taken from the first 6 4 0 x 512 OWIP FPA which we produced. Thus, we believe that there is plenty of room for further improvements of these FPAs. The net peak q u a n t u m efficiency of the FPA was 1.4% (lower focal plane array q u a n t u m efficiency is attributed to lower photoconductive gain at lower operating bias) and this corresponds to an 1 .E+03 Background Optics Bias
~" 1 . E + 0 2 E
300K
f/2
-1.1 V
~
/
~'~
I
'
tint
121
uJ Z 1.E+01 -
1 .E+00
I
I
60
70
1
80
Temperature (K)
90
Figure 3.8 Noise equivalent temperature difference NETI) estimated fi'om test structure data as a filnction of temperature for bias voltage V B = - 1.1 V. The background temperature TB= 300 K and the area of the pixel A=(23 i~m) 2.
98
Handbook of Infrared Detection Tecllnologies
60,000 50,000 40,000 E
30,000 20,000 10,000
0
I 10
1 20
1 30
k 40
I 50
I 60
I 70
I 80
I 90
NEDT (mK) Figure J. 9 NETD histogram of the 32 7 6 8 0 pixels of the 640 • ~ 12 arra!/ showing a high uni/brmit!l of the FPA. The uncorrected non-uniformit!l (-standard deviation ~mean) ~/the FPA is onl!l 14% including 1% nonuniformit!l of ROC and 1.4"/,, non-un(formit!l due to the cold-stop not being abh' to give the same.held o.f view to all the pixels in the FPA.
average of three passes of IR radiation (equivalent to a single 45: pass) through the photosensitive multi-quantum well region. It is worth noting that, under BLIP conditions, the performance of the detectors is independent of the photoconductive gain. and it depends only on the absorption q u a n t u m efficiency. A 640• OWIP FPA hybrid was mounted onto a 330 mW integral Sterling closed-cycle cooler assembly and installed into an Indigo Phoenix @ camerabody, to demonstrate a hand-held LWIR camera (shown in Figure 3.10). The Phoenix | infrared camera system has been developed by Indigo Systems Corporation to meet the needs of the research, industrial and ruggedized OEM communities. The system is comprized of a camera head and a selection of two video processing back ends. The camera head was made of Indigo's standard 6 4 0 • 512 format readout ISC 980 3. mated to long-wavelength OWIP detector materials. Two video processing units are the Real Time Imaging Electronics (RTIE) that provide conventional NTSC video as well as corrected parallel digital video out at video rates, and the Digital Acquisition System (DAS) that provides high-speed (40 MHz) raw digital data acquisition and output with limited real time video for system setup and focusing. The other element of the camera is a 100 mm focal length germanium lens, with a 5.5 degree field of view. It is designed to be transparent in the 7-1413m wavelength range, to be compatible with the QWIP's 8.:5 Bm operation. The digital acquisition resolution of the camera is 14-bits, which determines the instantaneous dynamic range of the camera (i.e., 16 384). However, the dynamic range of OWIP is 85 decibels. Its nominal power consumption is less than 4 "5watts.
GaAs/A1GaAs based quantmn well in l?ared photodetector focal phme arra!ls
99
Figure :~. 10 Picture of the 640 x 512 hand-held long waveh'ngth O I~'IP camera ( QWIP Phoenix~ ).
Video images were taken at a frame rate of 30 Hz at temperatures as high as T - 70 K, using a ROC capacitor having a charge capacity of 11 • 1()~ electrons (the m a x i m u m number of photoelectrons and dark electrons that can be counted in the time taken to read each detector pixel). Figure 3.11 shows one flame of a video image taken with a 9 ~m cutoff 64()x "512 pixel QWIP Phoenix | camera. The focal plane array performance data reported in this paper was taken with a laboratory measurement setup and OWIP Phoenix ~ camera. Estimates based on the single pixel data show that these FPAs should be able to provide 7 mK NETD with 30 ms integration time, which can be achieved at l. ] V bias. Focal planes were integrated with the close cycle stirling coolers, then these FPA/dewar assemblies were integrated with Phoenix ~ camera electronics. Thus we expect the OWIP Phoenix | camera to achieve a performance of NETD < 2 0 m K with shorter integration time ( < 16 ms with high bias), and a performance of NETD < l OmK with longer integration time at 30 Hz (30 ms integration with low bias).
3.5 640• 486 long-wavelength dual-band imaging camera The LWIR and very long-wavelength infrared (VLWIR) dualband QWIP device structure described in this section processed into interlace simultaneously readable dualband FPAs (i.e., odd rows for one color and the even rows for the other color). 26 The device structure consists of a 30 period stack, of VLWIR QWIP structure and a second 18 period stack of LWIR OWIP structure separated by a heavily doped O. 5 ~m thick intermediate GaAs contact layer. The first stack (VLWIR) consists of 30 periods of a 500 A AlxGal_xAs barrier and a 60 A GaAs well. Since the dark current of this device structure is dominated by the longer wavelength portion of the device structure, the VLWIR OWIP structure has been designed to have a bound-to-quasibound intersubband absorption 2s peak at 14.5 Bm. The second stack (LWIR) consists of 18 periods of a 500 A AlxGa~_xAs
100 Handbook of Infrared Detection Technologies
Figure 3.11 Oneframe of video image taken with the 9 I~m cutqfl640 x ~ 12 pixel O.WIP Phoenix ~ camera.
barrier and a n a r r o w 4 0 A GaAs well. This LWIR OWIP structure has been intersubband absorption peak at designed to have a b o u n d - t o - c o n t i n u u m 8.5 ~m, since photo current and dark current of the LWIR device structure is relatively small compared to the VLWIR portion of the device structure. This whole dualband OWIP structure is then sandwiched between (). 5 ~m GaAs top and bottom contact layers doped with tl= 5 • 1 ()1 - c m - 3, and has been grown on a semi-insulating GaAs substrate by MBE. Then a 3()()A AI~.3Ga~).7As stop-etch layer and a 1.0~tm thick GaAs cap layer were grown in situ on top of the device structure. GaAs wells of the LWIR and VLWIR stacks were doped with n - 6 • and 2 . 5 • cm -~ respectively. All contact layers were doped to n=5 • 1017 cm -3. The GaAs well doping density of the LWIR stack was intentionally increased by a factor of two to compensate for the reduced number of q u a n t u m wells in the LWIR stack. 2~ It is worth noting that the total current (dark c u r r e n t + p h o t o current) of each stack can be independently controlled by carefully designing the position of the upper state, well doping densities, and the number of periods in each MOW stack. All of these features were utilized to obtain approximately equal total currents from each MQW stack. Simultaneously measured responsivity spectrums of these vertically integrated dualband OWIPs are shown in Figure 3.12. Based on single element test detector data, the LWIR detectors show BLIP at bias VR=-2 V and temperature T - 72 K for a 300 K background w i t h f / 2 cold stop. The VLWIR detectors show BLIP under the same operating conditions at 45 K operating temperature. 28 Two different 2D periodic grating structures were designed to independently couple the 8 - 9 and 14-1 5 ~m radiation into detector pixels in even and odd rows of the FPAs. The FPA fabrication process is described elsewhere. 26 These dualband FPAs were tested at a background temperature of 300 K. with f / 2 cold stop, and at 30 Hz frame rate. As expected (due to BLIP), the estimated and experimentally obtained NEAT values of the LWIR d e t e c t o r s do not change
CaAs/AIGaAs based quantum well infrared photodetector focal plane arra!ts
101
0.6 VBIAS = 2V
0.4 v >I> cO Z
o
13_ 09
uJ cr
0.2
5
6
7
I
8
9
10
11
12
13
14
15
16
17
WAVELENGTH (#m)
Figure 3.12 Simultaneousl!l measured responsivit!! sl~ectrmn of verticall!! integrated LWIR and VLWIR dualband OWIP detector (Taken from reference 26 ).
significantly at temperatures below 65 K. The estimated NEAT of LWIR and VLWIR detectors at 40K are 36 and 4 4 m K respectively (See Figures 3.13 and 3.14). These estimated NEAT values based on the test detector data agree reasonably well with the experimentally obtained values. The experimental LWIR NEAT value is lower than the estimated NEAT value of 36mK. This improvement is attributed to the 2D periodic grating light coupling efficiency. On the other hand the experimental VLWIR NEAT value is higher than the estimated NEAT value of 44 mK. The authors believe this degradation is due to the inefficient light coupling at the 14-15 l~m region, readout multiplexer noise,
Figure 3.13 The uncorrected noise equivalent temperature d(ffl'rence (NEAT) histogram of 8-9 l~m detector pixels of the 640 • 486 dualband FPA. The mean NEAT is 29 mK.
102
Handbook of Infrared Detection Technologies
Figure 3.14 The uncorrected noise equivalent temperature tti!ti'rem'e (NEAT) histogram of l 4 - 1 5 ltm detector pixels of the 640 • 4 8 6 dtmlband FPA. The mean .\'EAT is 74 inK.
and noise of the proximity electronics. At 4() K the performance of both LWIR and VLWIR detector pixels of this dualband FPA are limited by photo current noise and readout noise. Video images were taken at a frame rate of 3() Hz. at temperatures as high as T= 74 K, using a ROC capacitor having a charge capacity of 9 • 1 ()~ electrons (the m a x i m u m n u m b e r of photoelectrons and dark electrons that can be counted in the time taken to read each detector pixell. Figure 3.1 5 shows simultaneously acquired 8 - 9 and 14-1 5 micron images using this two-color camera. 2s
3.6 640•
pixel broad-band QWIP imaging camera
A broadband QWIP device structure is designed by repeating a unit of several q u a n t u m wells with slightly different parameters such as well width and barrier height. 27 The positions of ground and excited states of the q u a n t u m well are determined by the q u a n t u m well width (L,,,) and the barrier height, i.e. the AI mole fraction (x) of the barrier 2- Since each single set of parameters for a boundto-quasibound q u a n t u m well 4 corresponds to a spectral band pass of about 1.5 ~m, three different sets of values are sufficient to cover a 1 ()-16 ~m spectral region. As shown in Figure 3.16. the MOW structure consists of m a n y periods of these three-quantum-well units separated by thick barriers. The device structure reported here involved 3 3 repeated layers of GaAs threequantum-well units separated by L t ~ 5 7 5 A thick AlxGa~ xAs barriers (See Figure 3.9). The well thickness of the q u a n t u m wells (see Figure 3.16) of threequantum-well units are designed to respond at peak wavelengths around 1 3, 14, thick and 151um respectively. These wells are separated by Lz, ~ 7 5 A AlxGal_xAs barriers. The AI mole fraction (xl of barriers t h r o u g h o u t the
GaAs/A1GaAs based quantum xvell itll)ared photodetector Jbcal plane arra!ls
1() 3
Figure :~. 15 Both pictures showy (flame-simHltaneousl!l acquired) t~vo-color images ~,i~11the 640 • 486 t~vocolor OWIP camera, hnage on the h'ft is from 14-1 ~ 12m infrared and the image on the right isJ?om 8-9 lxm infrared. Pix'el pitch of the FPA is 2 ~ 12m. The 14-1 ~ micron image is less sharp due to the difJ?action limited spot size being larger than the pix'el pitch of the FPA ( Taken l?om r~:ference 28).
Figure 3.16 Broad-band MOW structz~re is designed b!t repeating a zlnit of several quantum ~vells ~vilh slightl!l different parameters slwh as ~rell ~vidth and barrier height. The excited state energ!l levels broadened due to overlap of the ~vavefunctions associated ~vith excited states ofqlumtum ~vells separated b!l thin barriers.
structure was chosen such that the /.p=l 3 l~m quantum well operates under bound-to-quasibound conditions. ~J The excited state energy level broadening has been further enhanced due to the overlap of the wavefunctions associated with excited states of quantum wells separated by thin barriers. Energy band calculations based on a two-band model show excited state energy levels spreading about 2 8 meV.
104
Handbook of Infrared Detection Technologies
The sample was grown on a semi-insulating three-inch GaAs substrate by molecular beam epitaxy. It consists of the device structure described above sandwiched between top and bottom contact layers. Transport carriers (electrons) were provided by doping all GaAs wells and contact layers with Si. In order to measure the dark c u r r e n t - t e m p e r a t u r e curve, spectral responsivity (see Figures 3 . 1 7 - 3 . 1 9 ) and noise, 200 ~tm diameter mesas were fabricated using wet chemical etching and Au/Ge ohmic contacts were evaporated onto the top and bottom contact layers. The responsivity spectra of these detectors were measured using a 1()00 K black-body source and a grating monochromator. The detectors were back illuminated through a 45 ~ polished facet to obtain normalized responsivity spectra at different bias voltages. Then the absolute spectral responsivities were obtained by measuring total photocurrent from a calibrated black-body source. In Figure 3.18, responsivity curve at V j ~ = - 2 . 5 V bias voltage shows broadening of the spectral response up to A;. ~ 5.5 Bm, i.e. the full width at half
1 .E-06 "~ 1.E-07
E r (9 L
,-
1.E-08
1.E.09 ~-
O
~" 1.E-10~ r~
1 .E-11 1 .E-12
30
35
40
45
50
55
60
65
70
T e m p e r a t u r e (K)
Figure 3 . 1 7 Dark current verses temperature curve of 1 O- 1 5.4 lira broadband O_WIP at bias \:l~= - 2 . 5 V. Data were taken with a 2 0 0 ~tm diameter test device and normalized to 2 S • 2 S u m 2 pix'el.
0.5 0.4
~o. E
0.3 0.2 0.1
8
10
12
14
16
Wavelength (micron)
Figure 3.18 Responsivity spectrum of a broadband OIVIP test device at temperature T= 5 5 K. The spectral response peak is at 13.5 l~m and the long wavelength cutolJis at 1 ~.4 /2m.
GaAs/AIGaAs based quantum well infrared photodetector focal plane arrays
105
0.6
A
0.4
n
Iz
0.2
0~.
-v
0
1
2
3
4
Bias Voltage (V) Figure 3.19 Peak responsivit!t as a function of bias voltage at temperature T-- :~ ~ K.
m a x i m u m from 1 0 . 5 - 1 6 gm. This b r o a d e n i n g A2/2p ,-~42% is about a 4 0 0 % increase compared to a typical b o u n d - t o - q u a s i b o u n d QWIP. 4 Unlike n a r r o w - b a n d OWIPs, these detectors show spectral peak shifts from ,~ - 11.5 jJm to ,~. - 15.1 gm as negative bias voltage increased from VR - - 1 V to VB - - 5 V and similar behavior (~. - 11.5 IJm to ,~. - 14.7 IJm for VR - + 1 V to VB - + 5 V) was observed u n d e r positive bias voltages as well. This suggests that there is no substantial carrier depletion due to the applied electric field within the t h r e e - q u a n t u m - u n i t s because the direction of peak shift r e m a i n s the same u n d e r both positive and negative biases. The responsivity of the detector peaks at 13.5 l~m and the peak responsivity (Rp)of the detector is 2 5 0 m A / W at bias VR - 2 . 5 V. The bias dependent peak responsivity of the detector is s h o w n in Figure 3.19. The m e a s u r e d absolute peak responsivity of the detector is small, up to about VB--0.5 V. Beyond t h a t it increases nearly linearly with bias r e a c h i n g Re - 580 m A / W at V B = - 3 . 5 V. This type of behavior ofresponsivity versus bias is typical for a b o u n d - t o - q u a s i b o u n d OWIP. The peak q u a n t u m efficiency was 11% at bias V~--2.5 V for a 45 ~ double pass. The lower q u a n t u m efficiency is due to the lower well doping density (2 x 10 ~7 c m - 3) as it is necessary to suppress the dark c u r r e n t at the highest possible operating t e m p e r a t u r e . A peak q u a n t u m efficiency as high as 25% has already been achieved with regular well doping density (i.e., 1• cm-3). Due to lower readout multiplexer well depth (i.e., 11 • electrons) a lower dark c u r r e n t is m a n d a t o r y to achieve a higher operating t e m p e r a t u r e . In this case, the highest operating t e m p e r a t u r e of 45 K was determined by the well depth of the readout multiplexer. The dark c u r r e n t noise i,, of the device was m e a s u r e d using a s p e c t r u m analyzer at f - 55 K as a function of bias voltage. The noise gain g,, can n o w be obtained using the g - r noise calculated based on s t a n d a r d noise expression" i,, = x/4eI~g, AB w h e r e In is the dark c u r r e n t and AB is the b a n d w i d t h . Using experimental m e a s u r e m e n t s of noise and responsivity, one can n o w calculate specific detectivity D* form D * - R~/-A~/i,,, w h e r e A is area of the detector. Calculated D* value for the present device (,~. - 15.4 l~m) at, f - 55 K, and VR = 2.5 V is 3 x 10 l~ c m v ~ z / W . Even with broader response, this D* is c o m p a r a b l e to
106 Handbook of Infrared Detection Technoloqies previously reported D* of OWIPs with narrow spectral response. Figures 3.20 and 3.21 show the detectivity D* and the noise equivalent temperature difference (NEAT) as a function of the operating temperature of the device. It is well known that QWIPs do not absorb radiation incident normal to the surface unless the IR radiation has an electric field component normal to the layers of superlattice (growth direction) s. As we have discussed before s m a n y more passes of IR light inside the detector structure can be obtained by incorporating a randomly roughened reflecting surface on top of the detectors which also removes the light coupling limitations and makes two-dimensional OWIP imaging arrays feasible. This random structure was fabricated on the detectors by using standard photolithography and CCI,F, selective dry etching. After the r a n d o m reflector array was defined by the lithography and dry etching, the photoconductive OWIPs of the 64()• 512 FPAs were fabricated by dry etching through the photosensitive GaAs/AlxGa~_xAs m u l t i - q u a n t u m well layers into the 0.5 l,tm thick doped GaAs bottom contact layer. The pitch of the FPA is 2 5 l.tm and the actual pixel size is 2 3 x 2 3 lain2. The random reflectors on
1.E+12
A
1.E+11 N
-IE
~ 1.E+10
1 .E+09 30
40
50
60
70
T e m p e r a t u r e (K)
Figure 3.20 Detectivity as a fimction o.f temperatures at bias volta#e V~=-2. :~V.
1.E+04 1 .E+03 E a
uJ z
1.E+02 1.E+01 1.E+00 I 30
, 40
, 50
, 60
70
Temperature (K)
Figure 3.21 Noise equivalent temperature difference as a fi~nction o[temperatures at bias volta#e VB=-2.
~ V.
CaAs/A1GaAs based quantum well infrared photodetector focal plane arralls
1()7
top of the detectors were then covered with Au/Ge and Au for Ohmic contact and reflection. Figure 3.7 shows twelve processed OWIP FPAs on a three-inch GaAs wafer. Indium bumps were then evaporated on top of the detectors for Si readout circuit (ROC) hybridization. A single OWIP FPA was chosen and hybridized (via indium bump-bonding process) to a 64()x512 CMOS multiplexer (Indigo Systems 9803) and biased at VR = - 2 . 5 V (see Figure 3.22). At temperatures below 48 K, the signal-to-noise ratio of the system is limited by array nonuniformity, multiplexer readout noise, and photo current (photon flux) noise (see Figure 3.23). At temperatures above 48 K, temporal noise due to the OWIP's higher dark current becomes the limitation. As mentioned earlier this higher dark current is due to thermionic emission and thus causes the charge storage capacitors of the readout circuitry to saturate. Since the OWIP is a high impedance device, it should yield a very high charge injection coupling efficiency into the integration capacitor of the multiplexer. In fact Bethea et al. ~ have demonstrated charge injection efficiencies approaching 90%.
Figure 3.22 .4 size comparison of the 640 • ~ 12 long-wavelength QWIP FPA to a quarter.
30K
1.0E-07 -
35K
Q. E
"'Jr- 40K
,=I: 1.0E-09 -
Y.
c
o 'o L_
] OE-]] -
45K 50K
.
- e - - 55K
1.0E-]g 1.5
2
2.5
3
3.5
4
photo
Bias Voltage (V) Figure 3.23 Dark-current and the photo-current of 10-15.412m broadband QWIP as a fi~nction of bias voltage.
lO8 Handbookof InfraredDetectionTechnoloqies Charge injection efficiency can be obtained from equation (2 5 ), where g,,, is the transconductance of the MOSFET and is given by g,,, - eli)et/kT. The differential resistance RDet of the pixels at - 2 V bias is 5 . 4 • l~ ohms at T - 4 5 K and detector capacitance CDet is 1.4 • 1 O-14 F. The detector dark current IDet - 8 pA under the same operating conditions. According to equation (1) the charge injection efficiency rli,,i-99.5% at a frame rate of 30 Hz. The FPA was backilluminated t h r o u g h the flat thinned substrate membrane (thickness ~ 1 3 0 0 A). This thinned GaAs FPA membrane has completely eliminated the thermal mismatch between the silicon CMOS readout multiplexer and the GaAs based QWIP FPA. Basically, the thinned GaAs based QWIP FPA membrane adapts to the thermal expansion and contraction coefficients of the silicon readout multiplexer. Thus, thinning has played an extremely important role in the fabrication of large area FPA hybrids. In addition, this thinning has completely eliminated the pixel-to-pixel optical cross-talk of the FPA. This initial array gave very good images with 99.9% of the pixels working, demonstrating the high yield of GaAs technology. The operability was defined as the percentage of pixels having noise equivalent differential temperature less than lOOmK at 3OOK background with f/2 optics and in this case operability happens to be equal to the pixel yield. We have used equation (26) to calculate the NEAT of the FPA. where Dt*3 is the blackbody detectivity, dPR/dT is the derivative of the integrated blackbody power with respect to temperature, and 0 is the field of view angle (i.e., sine(0/ 2)-(4f2+1) -1, where f is the f number of the optical system). The background temperature T B - 30OK, the area of the pixel A=(231.tm) 2, the f number of the optical system is 2, and the frame rate is 30 Hz. Figure 3.24 shows the experimentally measured NEAT histogram of the FPA at an operating temperature of T -- 35 K, bias VR - - 2 . 5 V at 300 K background with f/2 optics and the mean value is 55 mK. This agrees reasonably well with our estimated value of 25 mK based on test structure data. The read noise of the multiplexer is 500 electrons. The factor of two short-fall of NEAT is mostly attributed to decrease in bias voltage across the detectors during charge accumulation (common in m a n y direct injection type readout multiplexers) and read noise of the readout multiplexer. The experimentally measured peak q u a n t u m efficiency of the FPA was 9.5% which agrees well with the 11% q u a n t u m efficiency estimated from the single element detector data. A 6 4 0 • 512 OWIP FPA hybrid was mounted onto an 84-pin lead-less chip carrier and installed into a laboratory dewar which is cooled by liquid neon, to demonstrate a LWlR imaging camera (FPA was cooled to 35 K). The other element of the camera is a 100 mm focal length AR coated germanium lens, which gives a 9.2~ 6.9 ~ field of view. It is designed to be transparent in the 8 12 ~tm wavelength range (which is not fully compatible with the 1 0 - 1 5 ~tm broadband OWIP array). An image processing station was used to obtain clock signals for the readout multiplexer and to perform digital data acquisition and non-uniformity corrections. The digital data acquisition resolution of the camera is 14-bits, which determines the instantaneous dynamic range of the camera (i.e., 16 384), however, the dynamic range of QWIP is 85 decibels.
GaAs/A1GaAs based quantum well infrared photodetector focal plane arrays
109
14000 12000 C
o
10000
O
8000
=
6000
r
4000
O
2000
0 -0.2
-0.1
0
0.1
0.2
0.3
NEAT (K) Figure 3.24 Noise equivalent temperature difference (NEAT) histogram of the 327 680 pix'els of the 6 4 0 x 5 1 2 array showing a high un(formit!t of the t:PA. The uncorrected non-uniformit!l (=standard deviation/mean) of this unoptimized FPA is onl!l 6.3% includin~l 1% non-uniformity of ROC and 1.4% nonuniformit# due to the cold-stop not being abh' to ~3ivethe sameJield of vicar to all the pixels in the FPA.
The measured mean NEAT of the OWIP camera is :35 mK at an operating temperature of f - 35 K and bias Vt~- - 2 . 5 V at 3 0 0 K background with f/2 optics (see Figure 3.24). This is in good agreement with expected focal plane array sensitivity due to the practical limitations on charge handling capacity of the multiplexer, read noise, bias voltage and operating temperature. The uncorrected NEAT non-uniformity (which includes a 1% non-uniformity of the ROC and a 1.4% non-uniformity due to the cold-stop in front of the FPA not yielding the same field of view to all the pixels) of the 327 680 pixels of the 6 4 0 x 5 1 2 FPA is about 6.3% ( - s i g m a / m e a n ) . The non-uniformity after twopoint (17 ~ and 27 ~ Celsius) correction improves to an impressive ().1%. As mentioned earlier, this high yield is due to the excellent GaAs growth uniformity and the mature GaAs processing technology. Video images were taken at a frame rate of 15 Hz at temperatures as high as T - 35 K using a ROC capacitor having a charge capacity of 11 x 1 ()~ electrons (the m a x i m u m number of photoelectrons and dark electrons that can be counted in the integration time of each detector pixel). Figure 3.2 5 shows a flame of video image taken with this 1 0 - 1 5 . 4 ~m 64()• 512 broadband OWIP imager. These high resolution images comparable to standard TV, demonstrate the high operability (i.e., 99.9%) and the stability (i.e., lower residual uniformity and l f f noise) of the 6 4 0 • long-wavelength OWIP staring array camera. It should be noted that these initial unoptimized FPA results are far from optimum. The light coupling gratings were not optimized (as described earlier) for maximum
1 ]0
Handbook of Infrared Detection Technolo~lies
Figure 3.25 This picture sho~vs one J?ame of video clip taken ~vith the 640• 512 pixel I 0 - I 5.41~m broadband QWIP focal plane arra!l. This ima~le sho~vs the liquid h, vel of a soda can and some.fingerprints on the C~II.
light coupling efficiency, no anti-reflection coatings were used on the backside of the FPA,
3.7 640• array
pixel spatially separated four-band QWIP focal plane
One unique feature of this spatially separated four-band focal plane array is that the four infrared bands are independently readable on a single imaging array. This feature leads to a reduction in instrument size, weight, mechanical complexity, optical complexity and power requirements, since no moving parts are needed. Furthermore, a single optical train can be employed, and the focal plane can operate at a single temperature. This four-band device structure was achieved by the growth of multi-stack QWIP structures separated by heavily doped contact layers, on a GaAs substrate. Device parameters of each QWIP stack were designed to respond in different wavelength bands. Figure 3.26 shows the schematic device structure of a fourcolor OWIP imager. A typical OWIP stack consists of a MOW structure of GaAs quantum wells separated by AlxGal_• barriers. The actual device structure consists of a 15 period stack of 3-5 pm QWIP structure, a 2 5 period stack of 8.5]Opm QWIP structure, a 25 period stack of ]()-12 ~tm OWIP structure and a 30 period stack of 13-15.:3 jam OWIP structure. Each photosensitive MOW stack was separated by a heavily doped (thickness ().2-().813m) intermediate GaAs contact layer (see Figure 3.26). Since the dark current of this device structure is dominated by the longest wavelength portion of the device structure, the VLWIR QWIP structure has been designed to have a bound-to-quasibound intersubband
GaAs/A1GaAs based quantmn well infrared photodetector focal plane arrags
111
Figure .3.26 Lager structure diagram ql four-band ()~'II ~ device and the 2I) periodic grating structure. Each pixel represents a 640 x 128 pixel area of the four-balld focal plane arra!t.
absorption peak at 14.0 ~m. Other OWIP device structures have been designed to have a bound-to-continuum intersubband absorption process, because the photo current and dark current of these devices are relatively small compared to the VLWIR device. This whole four-band OWIP device structure was then sandwiched between 0.513m GaAs top and bottom contact layers doped with n - 5 • 1017 cm- 3 and was grown on a semi-insulating GaAs substrate by MBE. The individual pixels were defined by photolithographic processing techniques (masking, etching, chemical vapor deposition, metal deposition, etc.). Four separate detector bands were defined by a deep trench etch process and the unwanted spectral bands were eliminated by a detector short-circuiting process. The unwanted top detectors were electrically shorted by gold coated reflective 2D etch gratings as shown in Figure 3.26. In addition to shorting, these gratings serve as light couplers for active QWIP stack in each detector pixel. The design and optimization of these 2D gratings to maximize QWIP light coupling were extensively discussed in ref. 5. The unwanted bottom detectors were electrically shorted at the end of each detector pixel row. Typically, in single-band QWIP FPAs. quarter wavelength deep (h -- 2p/4nGaAs) grating grooves are used for efficient light coupling. However, in the four-band FPA, the thickness of the quarter wavelength deep grating grooves are not deep enough to short-circuit the top three MQW QWIP stacks (e.g., three top OWIP stacks on 14-15.5 gm OWIP in Figure 3.26). Thus, three-quarter wavelength groove depth 2D gratings (h - 3;.p/4n~;,~.,) were used to short the top unwanted detectors over the 1 ()-12 and 14-15.5 microns bands. This technique optimized the light coupling to each OWIP stack at corresponding bands while keeping the pixel (or mesa) height at the same level, which is essential for the indium bump-bonding process used for detector array and readout multiplexer hybridization. Figure 3.27 shows the normalized spectral responsivities of all four spectral bands of this four-band focal plane array.
112
Handbook of Infrared Detection Technologies
Figure 3.2 7 Normalized spectral respo~lse of the lbltr-hand OWIt~.lbcal plane arrajt.
Few OWIP FPAs were chosen and hybridized to a 64()x 512 CMOS multiplexer (ISC 9803) and biased at V~ - - 1 . 1 V. At temperatures below 83 K, the signalto-noise ratio of the 4 - 5 l.tm spectral band is limited by array non-uniformity, multiplexer readout noise, and photo current (photon flux) noise. At temperatures above 45 K, temporal noise due to the 14-15.5/.tm QWIP's higher dark current becomes the limitation. The 8-1() and 1 0 - 1 2 ~m spectral bands have shown BLIP performance at temperatures between 45 and 83 K. The FPAs were back-illuminated through the fiat thinned substrate m e m b r a n e (thickness ~ 1 3 0 0 A). This initial array gave excellent images with 99.9% of the pixels working (number of dead pixels ~ 2 5()), demonstrating the high yield of GaAs technology. The operability was defined as the percentage of pixels having noise equivalent differential temperature less than 1 ()()mK at 30()K background and in this case operability happens to be equal to the pixel yield. A 640• pixel four-band OWIP FPA hybrid was mounted onto an 84-pin leadless chip carrier and installed into a laboratory dewar which is cooled by liquid helium to demonstrate a four-band simultaneous imaging camera. The FPA was cooled to 45 K and the temperature was stabilized by regulating the pressure of gaseous helium. The other element of the camera is a 1()() m m focal length AR coated g e r m a n i u m lens, which gives a 9 . 2 ~ ~ field of view. It is designed to be transparent in the 8 - 1 2 l.tm wavelength range. The SEIR | image processing station was used to obtain clock signals for readout multiplexer and to perform digital data acquisition and non-uniformity corrections. The digital data acquisition resolution of the camera is 14-bits, which determines the instantaneous dynamic range of the camera (i.e., 16 384), however, the dynamic range of OWIP is 85 decibels. Video images were taken at a frame rate of 30 Hz at temperatures as high as T - 45 K, using a ROC capacitor having a charge capacity of 11 • 1 ()r electrons {the m a x i m u m n u m b e r of photoelectrons
GaAs/A1GaAs based quantum well itffrared photodetector focal plane arra!ls
113
and dark electrons that can be counted in the time taken to read each detector pixel). Figure 3.28 shows one flame of a video image taken with a four-band 640 x 512 pixel OWIP camera.
3.8 QWIPs for low background and low temperature operation Although QWIP performs exceptionally well under high background levels, the detector exhibits anomalous behavior when it operates under stringent low irradiance and low temperature conditions {i.e. at extremely low photocurrent). Singh et al. have modeled and explained some of the irregular behaviors by treating OWIP as a resistance in series with a capacitance. 32'33 Arrington et al. 34 have shown experimentally the non-fiat frequency response curves of OWIP at low background and low operating temperature conditions. Under such conditions, responsivity decreases as the detector operating frequency increases. The roll-off frequency is completely determined by the device design and operational conditions. ~4 This behavior is empirically similar to dielectric relaxation effects observed in bulk extrinsic silicon and germanium photoconductors under similar operational conditions. ~s. ~, In principle, QWIP operates very similar to extrinsic bulk photoconductors. Electrons in the subbands of the isolated quantum wells can be visualized as electrons attached to impurity states in bulk photoconductors. As photogenerated electrons depart the active doped quantum well regions, they leave behind a space-charge buildup. Thus, in order to operate QWIP steadily, requires a sufficient dark or background photo current to replenish the depleted quantum wells. This is not an issue for QWIP detectors operating at high background conditions or high temperatures, because high photocurrent or high dark current can easily provide carriers to re-fill the space-charge buildup. However,
Figure 3.28 One frame of video image taken ~vith the 4-1 ~. ~ 12m cutoff four-band 640• ~ 12 pixel O.WIP camera. The image is barely visible in the 14-1 ~. ~ 12m spectral band due to the poor optical transmission of the anti-reflection layer coated germanium lens.
114
Handbookof Infrared Detection Technoloqies
at low operating temperatures for low-background irradiance levels, high resistivity of thick barriers in the active region impedes electrons from entering the detector from the opposite electrode. This could lead to a delay in refilling the space-charge buildup, and result in a lower responsitivity at high optical modulation frequencies. In order to overcome this problem we have designed a new detector structure, the Blocked Intersubband Detector (BID) with separate active quantum well region and blocking barrier. 37 As shown the Figure 3.29, the active region of BID consists of quantum wells separated by thin barriers, which creates sub-minibands due to large overlap of sub-level wave functions. Thus, space-charge buildup will get quickly refilled by electrons via sequential resonant tunneling along the ground state miniband from the emitter contact layer. A thick, impurity free blocking barrier is placed between the active region and collector to suppress the dark current of the device. We have fabricated a BID device consisting of 50 period MOW photosensitive regions and a 1 l~m thick A1.2Ga l_.2As blocking barrier layer. Each quantum well period contains a 60 A GaAs well and 1 ()()A A1.2Gal_.2As barrier. The whole device structure is sandwiched between two GaAs contact layers (Figure 3.29). The sample was grown on a semi-insulating GaAs substrate by molecular beam epitaxy. Transport carriers (electrons)were provided by doping all GaAs wells and contact layers with Si to a density of n=4 • 1017 cm-3. Mesas 20()13m in diameter were fabricated using wet chemical etching in order to measure dark current-voltage curves, spectral responsivity and noise. The responsivity spectra of these detectors were measured using a 10()0 K black-body source and a grating monochromator. To obtain normalized responsivity spectra at different bias voltages, detectors were back illuminated through a 45 ~ polished facet. Then the absolute spectral responsivities were calculated by measuring total photocurrent due to a calibrated black-body source. Figure 3.30 shows responsivity curves at different bias voltages from VI~ = -(). :3 to l. :3 V. Unlike in typical OWIPs, the spectral response of BID shows extra broadening because the photoexcitation occurs between sub-minibands instead of localized sub-levels. Figure 3.30 also illustrates the absorption quantum efficiency spectrum of the BID, measured using a 45 ~ wave guide geometry. Although the BID shows comparable absorption quantum efficiency, it shows much smaller responsivity than thick barrier OWIP devices. Therefore, the photoconductive gain of the BID (g = 0.06 at V = - 1 V bias voltage) is a few
Figure 3.29 Energy band diagram of GaAs/Al(;aAs Blocked Intersubband Detector (BID).
GaAs/A1GaAs based quantum well infrared photodetector focal plane arra~ts 115
Figure 3.30 Responsivity and absorption quantum efficienc!l spectra of the BID measured using a 4 ~~ wave guide geometry. Shown responsivity curves are for different bias voltages from VB=--O.5 to 1.5 V with 0.25 Vsteps.
times smaller (g ~ 0.2 - 0.5 at V = - 3 V bias voltage). This can be attributed to the lower electric field across the active region of the BID. Due to higher resistivity of the blocking layer, compared to the active superlattice, much of the applied bias voltage drops across the blocking layer. Thus, the majority of the photoexcited electrons relax within the superlattice before reaching the blocking layer. In contrast, the typical QWIP has uniform resistivity along the active layer, which allows a substantial uniform electric field across the device. In order to improve the carrier collection efficiency, we have designed a BID with a thinner blocking barrier. Although, this device operates at a lower bias voltage, the electric field distribution across the active layers is much higher than in the previous devices. Figure 3.31 shows the comparison of optical gains for two different devices obtained by measuring absorption quantum efficiency and responsivity. Figure 3.32 shows dark currents measured at V = - 1 V bias voltage and at different operating temperatures. The estimated detectivity D* at different operating temperatures, for a low background, with photon flux 4)= 4 • 108 photons/sec/pixel is also illustrated in Figure 3.32. This flux simulates 120 K space background withf/2-optics aperture.
3.9 Summary In summary, we have discussed the importance of FPA uniformity in NEAT, the general figure of merit that describes the performance of large imaging arrays. It is important to note that when D*>~IO l(~ cmv/Hz/W, the performance is uniformity limited and thus essentially independent of the detectivity, i.e., D* is not the relevant figure of merit. 2 Furthermore, we have demonstrated the 6 4 0 • pixel LWIR portable OWIP camera based on bound-to-quasibound device structure, the first 640• 512 pixel four-band FPA, 640• 486 pixel dual-band imaging camera, and
116
Handbook of Infrared Detection Technologies
0.5
.c_ 0 ,,,==
Blocking Layer Thickness 9600 A
0,4-
0.3
~. 0.2
O
0.1
0
i
I
i
!
4
8
12
16
20
Effective Electric Field (kVIpm) Figure 3.31 Optical gain obtained for two d(fferent B/D devices b!l measuring absorption quantum eSficienc!t and responsivity, hnprovement in optical gain is due to efficient carrier collection in the newest design.
E e-
~
L L
1 .E-07
.E+13
1.E-09-
.E+12 -
1.E-11
.E+11
1.E-13
.E+10 -
-
o .a
1 1.E-15 30
I
I
I
40
50
60
Temperature
(K)
.E+09 70
30
,
,
,
40
50
60
Temperature
70
(K)
Figure 3.32 BID dark current and peak detectivit!! D* against operating temperature. The dark current was measured with 200 I~m diameter mesa and normalized to a 30 • 30 i~m 2 pix'el. D* is calculated based on responsivit# and dark currents for a background with photon flux 4)=4 • 10 '~photons/s/pixel which simulates 120 K space background and f/2-optics.
640• pixel broad-band imaging camera. Furthermore, we have demonstrated that OWIP is able to operate at low temperatures, with a higher sensitivity under high background irradiance levels. The anomalous behavior of QWIP w h e n operating under stringent low irradiance levels at low temperatures was addressed. A new blocked intersubband detector (BID) structure, with active superlattice and a thick blocking barrier, has been proposed to overcome this problem. Experimental m e a s u r e m e n t s show similar absorption q u a n t u m efficiency, with a lower photoconductive gain. This can be attributed to a lower voltage drop across the active superlattice of the BID. The real advantage of the BID detector is that its infrared sensing photoemitter is a GaAs/A1GaAs based
GaAs/A1GaAsbasedquantum well infraredphotodetectorfocal plane arrays l 17 superlattice, thus its cutoff wavelength can be easily tuned by band gap engineering (i.e., tailorable cutoff wavelengths). Thus, if one seeks shorter wavelength operation such as 1() or 15 microns cutoff, the MOW based BID detectors will operate at much higher cryogenic temperatures such as 30K which can be achieved passively at space low background conditions.
Acknowledgements The research described in this paper was performed by the Jet Propulsion Laboratory, California Institute of Technology, and was jointly sponsored by the Breakthrough Sensors and Instrument Component Technology Thrust of NASA Cross Enterprise Technology Development Program, Air Force Research Laboratory, and the Ballistic Missile Defense Organization.
References 1. C. Weisbuch, Semicon. Semimet. 24, 1 - 1 3 3 (1987). 2. B. F. Levine, J. Appl. Phys. 74, R1 (1993 ). 3. S. D. Gunapala, B. F. Levine, L. Pfeiffer, and K. West, Dependence of the Performance of GaAs/A1GaAs Q u a n t u m Well Infrared Photodetectors on Doping and Bias, J. Appl. Phys. 69, 6517 ( 1991 ). 4. S. D. Gunapala, J. K. Liu, J. S. Park, T. L. Lin, and M. Sundaram, Infrared Radiation Detecting Device, US Patent No. 6,211.529. 5. S. D. Gunapala and S. V. Bandara, Q u a n t u m Well Infrared Photodetector (OWIP) Focal Plane Arrays, Setniconductors and Semimetals 62, 1 9 7 - 2 8 2 , Academic Press (1999). 6. J. Y. Andersson, L. Lundqvist, and Z. F. Paska, J. Appl. Phys. 71, 3600 (1991). 7. G. Sarusi, B. F. Levine, S. J. Pearton, S. V. Bandara. and R. E. Leibenguth, ]. Appl. Phys. 76, 4 9 8 9 (1994). 8. C. G. Bethea, B. F. Levine, V. O. Shen, R. R. Abbott, and S. J. Hseih, IEEE Trans. Electron Devices 3 8, 1118 ( 1991 ). 9. C. G. Bethea and B. F. Levine, B. F. In Proceedings of SPIE International Symposium on Optical Applied Science and Engineering, San Diego, CA ( 1992 ). 10. C. G. Bethea, B. F. Levine, M. T. Asom, R. E. Leibenguth, J. W. Stayt, K. G. Glogovsky, R. A. Morgan, J. Blackwell, and W. Parish, IEEE Trans. Electron Devices 40, 1957 ( 1993 ). 11. W. A. Beck, T. S. Faska, J. W. Little, J. Albritton, and M. Sensiper, Proceedings of the Second International Symposium on 2-20 l~m Wavelength Infrared Detectors and Arrays: Physics and Applications, Miami Beach, Florida (1994). 12. S. D. Gunapala, J. S. Park, G. Sarusi, True-Lon Lin, J. K. Liu, P. D. Maker, R. Muller, C. A. Shott, and T. Hoelter, 15 Bm 128 x 128 GaAs/AIGaAs Q u a n t u m Well Infrared Photodetector Focal Plane Array Camera, IEEE Trans. Electron Devices 44, 4 5 - 5 0 (1997).
118 Handbookof Infrared Detection Technologies 13. S. D. Gunapala, J. K. Liu, J. S. Park. M. Sundaram, C. A. Shott, T. Hoelter, True-Lon Lin, S. T. Massie, P. D. Maker, R. E. Muller, and G. Sarusi, 9 ~m Cutoff 2 5 6 x 2 5 6 GaAs/AlxGal_xAs Q u a n t u m Well Infrared Photodetector Hand-Held Camera, IEEE Trans. Electron Devices 44, 5 1 - 5 7 (1997). 14. S. D. Gunapala, S. V. Bandara, J. K. Liu, W. Hong, M. Sundaram, P. D. Maker, R. E. Muller, C. A. Shott, and R. Carralejo, Long-Wavelength 6 4 0 x 486 GaAs/AlxGal_xAs Q u a n t u m Well Infrared Photodetector Snap-shot Camera, IEEE Trans. Electron Devices 45, 1890 ( 1998 ). 15. J. Y. Andersson, J. Alverbro, J. Borglind, P. Helander, H. Martijn, and M. Ostlund, 3 2 0 x 2 4 0 Pixels Q u a n t u m Well Infrared Photodetector (QWIP) Array for Thermal Imaging: Fabrication and Evaluation. SPIT 3 0 6 1 , 7 4 0 - 7 4 8 (199 7). 16 K. K. Choi, A. C. Goldberg, N. C. Das, M. D. Jhabvala, R. B. Bailey, and K. Vural, S P I E 3 2 8 7 , 1 1 8 - 1 2 7 (1998). 17. R. Breiter, W. Cabanski, 1R. Koch. W. Rode. and J. Ziegler, SPIT 33 79, 423 (1998). 18. R. H. Kingston, Detection of Optical and Infrared Radiation Springer, Berlin (1978). 19. A. Zussman, B. F. Levine, J. M. Kuo, and J. de Jong, 1. Appl. Phys. 70, 5101
(1991). 20. F. D. Shepherd, In Infrared Detector and Arrays, SPIT 9 3 0 (SPIT Orlando, FL), p. 2 (1988). 21. J. M. Mooney, F. D. Shepherd. W. S. Twins. J. E. Murgia, and J. Silverman, Opt. Eng. ZS, 1151 (1989). 22. Grave and A. Yariv, Intersubband Transitions in Ouantum Wells, edited by E. Rosencher, B. Vinter, and B. Levine, Cargese. France (Plenum, New York), p. 15 (1992). 23. B. F. Levine, C. G. Bethea, K. G. Glogovsky, J. W. Stayt, and R. E. Leibenguth, Semicond. Sci. Technol. 6, C114 ( 1991 ). 24. M. T. Asom, C. G. Bethea, M. W. Focht, T. R. Fullowan, W. A. Gault, K. G. Glogovsky, G. Guth, R. E. Leibenguth, B. F. Levine, G. Lievscu, L. C. Luther, J. W. Stayt, Jr., V. Swaminathan, Y. M. Wong, and A. Zussman. Proceedings of the IRIS Specialty Group on Infrared Detectors I. 13 ( 1991 ). 25. V. Swaminathan, J. W. Stayt, Jr., J. L. Zilko, K. D. C. Trapp, L. E. Smith, S. Nakahara, L. C. Luther, G. Livescu, B. F. Levine, R. E. Leibenguth, K. G. Glogovsky, W. A. Gault, M. W. Focht, C. Buiocchi, and M. T. Asom, Proceedings of the IRIS Specialty Group on Infrared Detectors, Moffet Field, CA ( 1992). 26. D. Duston, BMDO's IS&T faces new hi-tech priorities, BMD Monitor, 1 8 0 183, May 19 (1995). 2 7. M. T. Chahine, Sensor requirements for Earth and Planetary Observations, Proceedings of Innovative Long Wavelength Infrared Detector Workshop, Pasadena, California, pp. 3-31, April 24-26 (1990). 28. S. D. Gunapala, S. V. Bandara, A. Singh, J. K. Liu, S. B. Rafol, E. M. Luong, J. M. Mumolo, N. 0. Tran, J. D. Vincent, C. A. Shott, J. Long, and P. D. LeVan, 6 4 0 x 4 8 6 Long-wavelength Two-color GaAs/A1GaAs Quantum Well Infrared Photodetector (OWIP) Focal Plane Array Camera IEEE Trans. Electron Devices
4 7 , 9 6 3 - 9 7 1 (2ooo).
GaAs/A1GaAs basedquantum well infraredphotodetectorfocal plane arrays 119 29. S. D. Gunapala, S. V. Bandara, ]. K. Liu, S. B. Rafol, C. A. Shott, R. ]ones, S. Laband, ]. Woolaway II, ]. M. Fastenau, and A. K. Liu, 9 IJm Cutoff 6 4 0 x 512 Pixel GaAs/Al• Quantum Well Infrared Photodetector Hand-held Camera, to be published in SPIE Proceedings 4 7 2 1 (2002). 30. S. V. Bandara, S. D. Gunapala, ]. K. Liu, E. M. Luong, ]. M. Mumolo, W. Hong, D. K. Sengupta, and M. ]. McKelvey, 10-16 IJm Broadband Quantum Well Infrared Photodetector, Appl. Phys. Lett. 72, 2427 (1998). 31. S. V. Bandara, S. D. Gunapala, ]. K. Liu, S. B. Rafol, ]. M. Mumolo, and D. Z. Ting, Array of OWIPs With Spatial Separation of Multiple Colors, NASA Tech Briefs 26 No. 5, 8a (2002). 32. A. Singh and D. A. Cardimona, SPIE, 2 9 9 9 , 46 (1997). 33. A. Singh and D. A. Cardimona, Opt. Eng. 38, 1424 (1999). 34. D. C. Arrington, ]. E. Hubbs, M. E. Gramer, Gary A. Dole, SPIE 4 0 2 8 , 288 (2000). 35. M. Ershov, S. Satou, and Y. Ikebe, 1. Appl. Phzds. 86, 6442 (1999) 36. N. M. Haegel, C. R. Brennan, and A. M. White, 1. Appl. Phys. 80, 1510 (1996). 37. S. Gunapala, S. Bandara, ]. Bock, M. Ressler, ]. Liu, ]. Mumolo, Sir Rafol, D. Ting and M. Werner, Large Format Long-wavelength GaAs/A1GaAs Multiquantum Well Infrared Detector Arrays for Astronomy SPIE 4 2 8 8 , 2 7 8 - 2 8 5 (2001).
This Page Intentionally Left Blank
Chapter 4
GalnAs(P) based QWIPs on GaAs, lnP, and Si substrates for focal plane arrays J. Jiang and M. Razeghi
4.1 Introduction 4.7.1 Overview of infrared detector
Every object emits infrared light. The hotter an object is, the more radiation it will emit. Detection and processing of infrared radiation can provide a wealth of information about an object that is not available in other regions of the spectrum. Infrared detectors are very important in both civilian and military applications and have been extensively studied over the past century. In general, infrared detectors can be categorized as either thermal detectors or photon detectors (as shown in Figure 4.1). Photon absorption in thermal detectors leads to an increase in temperature, resulting in a measurable change in certain material properties. The operation of the photon detectors, on the other hand, is based on the measurement of an electrical photocurrent generated by photon absorption in a semiconductor. A comparison between thermal and photon detectors is given in Table 4.1. The atmosphere has a few important transmission windows in the infrared region (as shown in Figure 4.2). These atmospheric windows are important for almost all infrared detector applications, allowing detection of objects at long distances. The short wavelength infrared (SWIR) window extends from the visible region up to 2.5 pm, the middle wavelength infrared (MWIR) window lies between 3 and 5 l~m, and the long wavelength infrared (LWIR) window ranges from 8 to 14 pm. The location of the LWIR window is fortunate since room temperature objects have a peak wavelength of almost 101~m. Detection of wavelengths larger than 141um, so called very long wavelength infrared (VLWIR), is mainly used for outer space application.
122 Handbook of Infrared Detection Technologies
I FerroelectrJc-PyroelectricI Resistive Thermal (Bo]ometers) ]
Infrared]
IV-V! ~PbSe,PbSnTe) ]
D
I
t ''-v'
I
t ![[-\: (In~. InAsSb,InTlSb.Ia,SbBi) ]
-~QuantumWells
Type II(InSl~'InA~Sb,IaAs;C_ralnSbl[
Figure 4.1 Infrared detector categorization based on their method of detection.
Near infrared Middle infrared " 44 3, i
i
! ......
-i
Far infrared ....
i
.
g~ m
i" I m
Q,
t ,,
IL t,tt t,__,,,,. ,.,tt',. .,,tt~
Wavelength (~m)
~
~
,,t .~
~,
Absorption Molecule
Figure 4.2 The transmission spectrum of the atmosphere over a horizontal 6000ft. path length at sea level. The regions of high transmission, called atmospheric windows, are evident. Quantum well infrared photodetector.
Almost all the real-world applications of the infrared detector are based on thermal imaging using the infrared focal plane array (FPA) camera. An FPA is an optical sensor placed at the focal plane of an optical system such as a camera, spectrometer or telescope. The infrared FPA is composed of an infrared detector array, which can be designed and manufactured to be sensitive to wavelength range from SWIR to VLWIR based on both thermal and photon detectors. Infrared FPAs are combined with a read out integrated circuit (ROIC) or multiplexer which allows the electronic access of every pixel in the array.
GalnAs(P) based OWIPs on GaAs. InP. and Si substrates for focal plane arrays Table 4.1
123
Pros and c o n s o f thermal and p h o t o n infrared detectors
Detectors
Advantages
Disadvantages
Thermal
9 Relatively low cost imaging possible 9 Sensitive over a wide infrared r a n g e 9 Light, rugged, reliable, and convenient to use 9 A d e q u a t e response time for some imaging applications 9 Easy to tailor the b a n d g a p of the alloys to cover the entire infrared region 9 High detectivity 9 Well developed theoretical and experimental results 9 Good material properties 9 Fast response
9 Slow response {ms o r d e r ) d u e to reliance on h e a t i n g of atomic s t r u c t u r e 9 Relatively low detectivity (D*~ 1 ()s cmHz 1,2/W )
Photon
9 Difficulty in device processing 9 High cost in g r o w t h and device processing
4.1.2 Quantum well infrared photodetector What is a Q WIP?
The q u a n t u m well infrared photodetector (OWIP) is a semiconductor infrared photon detector relying on intersubband absorption within either the conduction band (n-type) or the valence band (p-type). The idea of utilizing a q u a n t u m well for infrared detection was first presented by Esaki and Sakaka in 19 7 7 and can be explained by using the basic principles of quantum mechanics. 1 The quantum well is equivalent to the well-known particle in a box problem in quantum mechanics, which can be solved by the time independent SchrSdinger equation. The solutions to this problem are the eigenvalues that describe energy levels inside the q u a n t u m well in which the particle is allowed to exist. OWIPs are built from quantum wells of wide-bandgap materials that in bulk do not absorb in the MWIR or LWIR. However, electron (hole) excitation may occur between the ground state and excited states in a conduction (valence) band quantum well (see Figure 4.3), making intersubband absorption possible in the MWIR or LWIR regions. The quantum well structure is designed so that these photoexcited carriers can escape from the quantum well and can be collected as photocurrent. These detectors afford greater flexibility because the peak and cutoff wavelength can be continuously tailored within a range by varying layer thickness (quantum well width) and barrier composition (barrier height). With appropriate choice of the well and barrier material, the detection wavelength of a QWIP can be tailored to any wavelength from MWIR to VLWIR. Why QWIP?
So far, the 'king' of infrared detectors has been the HgCdTe intrinsic photodetector (MCT). HgCdTe is the most studied semiconductor material for IR detectors and it is the standard against which all of other IR photon detectors are matched against. 2'3'4 It has very high quantum efficiency and detectivity: at
124
Handbook of Infrared Detection Technologies
Figure 4.3 Schematic of intersubband absorption, which takes place entirely within the valence band (for ptype doping, from H~ to H2) or conduction band (for n-type doping, from E1 to E2) of a quantum well. respectively. 7 7 K, a reported q u a n t u m efficiency exceeding 70%, and a detectivity exceeding 1012 cmHzl/2W-1. Focal plane arrays (FPAs) as large as 6 4 0 x 4 8 0 and dualcolor 1 2 8 x 1 2 8 FPAs have been demonstrated, s'6 However, the future of this material is uncertain because of the difficulties associated with its growth, processing, and device stability (see Table 4.2). Yields as low as 1% for 128 • 128 arrays produce costs per array as high as $10 0 0 0 - - 6 0 000. 7 On the other hand, QWIPs use the III-V material system, which is very mature due to the previous development of GaAs- and InP-based lasers, LEDs, and microwave circuitry. The growth and fabrication of QWIPs can be borrowed directly from these technologies. The advantages of QWIPs are also shown in Table 4.2, including their use in high frequency applications, s Although most QWIPs must be operated at cryogenic temperature, advances in closed cycle cryo-coolers have reduced both the size and cost of such systems. QWIP FPAs show better or comparable array performance to HgCdTe at a much lower cost. 9 The first n-type OWIP device based on intersubband absorption was demonstrated by Levine et al. on GaAs/AIGaAs material system in 1987.1~ Its peak spectral response occurs at l O.8~tm. The first p-type QWIP was demonstrated by Levine et al. in 1991.11 Also in 1991, only four years after the first demonstration of the QWIP, Bethea et al. obtained the first infrared image using a ten-element linear scanning OWIP array. 12 W h y Al-ffee?
GaAs/A1GaAs historically has been the most studied III-V material system because of its simplicity. Today, most of the reported FPAs were based on GaAs/ A1GaAs OWIP structures. Al-free III-V materials, on the other hand, have some inherent advantages, like reduced surface recombination velocity, less stringent
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arrays Table 4.2
125
C o m p a r i s o n b e t w e e n i n f r a r e d HgCdTe a n d QWIP d e t e c t o r .
Detector type
Advantages
Disadvantages
HgCdTe
9Excellent quantum efficiency 9 Very high detectivity
9 Poor array operability and uniformity 9 Material defects result in low RoA for many VLWIRpixels. Ro non-uniform 9 Unstable alloy and difficult material growth control 9 Radiation-hard arrays are difficult due to narrow bandgap and defects in material 9 Low yield and high costs for large-area arrays 9 Reproducibility is poor due to large sensitivity of bandgap to composition
9 Bandgap can be adjusted to vary detection wavelength
QwIP
Mature III-V growth technology 9Lower quantum efficiency than intrinsic HgCdTe 9 Wide-bandgap material is better for radiation-hard applications. 9Excellent array uniformity 9 High RoA allows long integration time 9 Narrow photoresponse spectrum 9Multi-color arrays demonstrated 9Very fast response for ultra-high frequency applications
9 Requires lower sensor temperature than intrinsic detectors for ). < 12 Hm 9 Require light coupling for n-type QWIPs
processing r e q u i r e m e n t s , a n d a b s e n c e of DX centers. To take a d v a n t a g e of these properties, people started to g r o w OWIPs w i t h o t h e r m a t e r i a l systems, like I n G a P / G a A s , G a l n A s P / I n P , G a I n A s P / G a A s , et a/. 13'14'1s Table 4.3 gives a c o m p a r i s o n b e t w e e n Al-free a n d Al-based s e m i c o n d u c t o r materials. W i t h the a d v a n t a g e s of Al-free m a t e r i a l s given in Table 4.3, high-reliability a n d low-cost infrared OWIP FPAs based on these m a t e r i a l s y s t e m are h i g h l y expected. In addition, the Al-free OWIP g r o w n on silicon s u b s t r a t e also s h o w e d better d e t e c t o r p e r f o r m a n c e t h a n t h a t of Al-based OWIP on silicon (see Section 4.6), w h i c h t r a n s l a t e s to better m o n o l i t h i c i n t e g r a t i o n infrared FPA based on Al-free QWIPs. As a result, the Al-free GalnAs(P) m a t e r i a l system, as a n i m p o r t a n t a l t e r n a t i v e to A1GaAs m a t e r i a l s y s t e m for OWIP s t r u c t u r e , has been receiving m o r e a n d m o r e a t t e n t i o n in r e c e n t years. In the rest of this c h a p t e r , we will focus on the GaInAs(P) OWIP s t r u c t u r e g r o w n on v a r i o u s substrates, s u c h as GaAs, InP, a n d silicon.
4.1.3 S t a t e - o f - t h e - a r t
The detectivity (D*) is the p r i m a r y figure of merit used in e v a l u a t i n g the single e l e m e n t p h o t o d e t e c t o r p e r f o r m a n c e . Today. the D* for 4 pm a n d 9 pm
126
Handbookof Infrared Detection Technologies
Table 4.3
C o m p a r i s o n b e t w e e n Ai-free and Al-based s e m i c o n d u c t o r m a t e r i a l s 16't7
Al-free material
Al-based materials
9 No oxidation, which results in higher device reliability 9 Simple device processing
9 Passivation needed for device fabrication
9 Low material growth temperature 9 Reduced surface recombination 9 DX-defect centers associated with AI are avoided 9 Potential for growth of monolithic integration with Si based device.
9 Higher material growth temperature which results in interdiffusion problem 9 Poor device reliability
OWIPs measure about 1.5 • 1012 cm-Hzl/2/W and 2x 1()~ cm-Hzl/2/W at 77 K, respectively. 1s.19 In addition, thanks to the narrow absorption peak exhibited by QWlPs, multicolor operation is possible by simply stacking different types of q u a n t u m wells. Tidrow et al. reported a four-color OWIP detector based on stacked InGaAs/ A1GaAs and GaAs/A1GaAs m u l t i - q u a n t u m wells (MQWs). 2~ OWIP FPAs using a different material system have been demonstrated by m a n y groups. 9'21'22 These FPAs have shown excellent imagery in the LWlR atmospheric window. Large format LWlR OWIP FPAs up to 6 4 0 x 4 8 0 pixels have been commercialized by the Jet Propulsion Laboratory (JPL). 23.24 JPL also demonstrated a 15 pm VLWlR OWIP FPA. 2s For the portable market, a handheld LWIR infrared camera based on a 2 5 6 x 2 5 6 GaAs/AIGaAs FPA was realized. 26 For multi-color FPAs, two-color LWIR/VLWlRFPAs were demonstrated. This FPA gave excellent images with 99.7% of the LWIR pixels and 98% ofVLWIR pixels working. 27 As an example, Figure 4.4 shows a thermal image taken by a 2 5 6 • infrared OWlP FPA camera. The handprint on the book can be seen very clearly.
4.2 Fundamentals of QWIP 4.2.1 Intersubband absorption Traditional interband optical absorption involves photoexciting carriers across the band gap E~, i.e., promoting an electron from the valence-band ground state to the conduction-band excited state (see Figure 4.5). An intrinsic semiconductor photon detector utilizing the interband absorption is only sensitive to the photons whose energy is larger than the semiconductor's energy gap. For the detection of MWIR and LWIR regions of spectrum, a narrow E~ material must be used. It is well known that these low band gap materials are more difficult to grow and process than large band gap semiconductors such as
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arrays
127
Figure 4.4 Thermal image taken by a 256 • ~60WIPfocal plane array camera. The handprint on the book can be seen clearly. (Courtesy of Dr. Murzy ]habvala ).
conduction band absorbed photon
~
-
valence band
Figure 4.5 intrinsic semiconductor detectors, an excited electron must cross the entire bandgap to reach the conduction band.
GaAs. To take advantage of the superior growth technology that exists with the wide bandgap semiconductor material, intersubband absorption inside the quantum wells is employed for detection of MWIP, or LWIR. For intersubband absorption, photon absorption takes place from ground state to excited states in quantum wells, as shown in Figure 4.3. The energy separation between the two bound states, E1 and E2 or H1 and H2, is much smaller than the bandgap of the well or barrier material,which translate to a much longer intersubband absorption wavelength. Thus, by controlling the location of the confined energy levels, the absorbed infrared wavelength can be
128 Handbookof Infrared Detection Technologies controlled. For a one-dimensional periodic q u a n t u m well structure, the position of confined energy levels can be computed by the Kronig-Penney model. 2s'29' 3o The relationship between the energy states and the well and barrier characteristics is given by:
1(
cosxL - coskwcoshabb + -~ rl - 1 )
sinkwsinha,,b
(1)
where X is a solution parameter, L is the period of superlattice thickness (well plus barrier width), k = v / 2m,,,E/~-. m;,,is the effective mass of the well material, E is 9 , the energy, /i is the reduced Planck's constant, w is the well thickness, ab = v/2m*b(Vo- E)ffi2, m~ is the effective mass of the barrier material, V(, is the conduction band offset, b is the barrier thickness, and 71 - m,*,,ab/m~,k. The values of allowed energy levels E can be obtained by numerical solution of equation ( 1 ). From (1), we can see that the absorbed infrared wavelength can be tailored by the bandgap and thickness of well and barrier material. For n-type QWIPs, which utilize an intersubband transition between F-valley derived states, the oscillator strength of a m u l t i - q u a n t u m well structure is given by:
f12 ~ I(FllT"-p[F2}[ 2
2
where F1 and F2 are the electron envelope functions, -K is the polarization vector for the incident infrared light, and - f the m o m e n t u m operator. If we assume the q u a n t u m well growth direction is along the z-axis, then the envelope functions should depend on z only, with F1-Fl(Z) and F2-F2(z). It is seen readily that for normal incidence of infrared radiation, for which 7=(e,.. e.,/, 0), results in zero absorption. One remarkable note to this q u a n t u m selection rule is that it only applies to n-type doped q u a n t u m wells. For p-type q u a n t u m wells, there is strong mixing among the heavy holes and lights holes at non-zero wave vector in the valence band which makes absorption of normal incidence infrared light possible. 31.32 One requirement for detection using q u a n t u m wells is that the well material must be doped. Because the energy of an absorbed photon is less than the bandgap, it cannot produce an excited photocarrier by itself. With the addition of carriers in the well, excitation only needs to happen from the first bound state to the next bound state. For n-type doping, excitation occurs entirely within the conduction band" for p-type doping, excitation occurs entirely within the valence band. Photocurrent is produced w h e n an excited photocarrier is able to escape the well whereupon an externally applied bias can sweep it out to be collected by one of the contacts. For n-type QWIPs, depending on the position of the first excited state in the q u a n t u m well 33 and barrier layer structure (see Figure 4.6), intersubband transitions in OWlPs can be classified to four types" bound-to-bound state (BB), 1~ bound-to-quasibound state (BQB), 34 b o u n d - t o - c o n t i n u u m state (B-C), 3~ and bound-to-miniband (B-M). 36 These structures all exhibit different degrees of
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arrays
129
Figure 4.6 Schematic of energ!t band diagram for B-B. B-C. BOB. and B-M OWIPs.
electron confinement within the q u a n t u m well. In general, less confined upper states exhibit a larger photoresponse bandwidth and weaker peak detectivity. The B-B OWIPs have intersubband absorption occurring between two bound states contained within the q u a n t u m well. However, the carriers excited to the higher bound state still need a way to escape from the well and get swept out. The two possible ways are tunneling through the barrier and thermionic emission. This sort of OWIP has a narrow absorption spectra and large dark current. The large dark current is related to the thin barriers used to facilitate the tunneling of photocarriers out of the q u a n t u m well. In a B-C QWIP, the intersubband absorption is based on the bound-toc o n t i n u u m states transition. The peak absorption is relatively small because the wave functions of the upper excited states spread over the barrier region. The absorption spectrum is expected to be wider due to the broad final states band above the barrier. In a BQB QWIP, the wave function of the upper excited states is aligned with the top of barrier. Since the barrier height also limits the cut-off wavelength, there is a sharp cut-off for this type of detector. In a B-M QWIP, the transition is from the localized bound ground state in enlarged wells to a resonant-coupled miniband of superlattice barriers. Since the well width in a B-M QWIP is m u c h wider than that of B-C and BOB QWIPs, there is a large overlap of wave functions between the initial and final states, and the interaction is strong. A blue shift of the peak detection wavelength is obtained by
130
Handbook of Infrared Detection Technologies
aligning the excited state to the top of the miniband, while a maximum bandwidth is achieved when the excited state is lined up in the middle of the miniband. Thus, B-M QWIPs offer more flexibility and fine tuning of the detection wavelength and spectral bandwidth than other types of OWIPs. For p-type QWlPs, the absorption and responsivity are much lower than those of n-type OWIP due to the large hole effective mass and low hole mobility. Reduced absorption stems from the absorption coefficient, which is inversely proportional to the effective mass of holes. On the other hand, a key benefit, as mentioned before, is the ability to absorb normal incidence radiation. To improve the intersubband absorption of the p-type QWlP, tensile strained-layer (TSL) or compressive strained-layer (CSL) quantum wells have been utilized. 37-4~J The strong mixing between light and heavy hole states substantially increases the normal incidence absorption in the strained-layer p-type OWIPs. In TSL p-type QWIPs, the intersubband absorption occurs between the ground light-hole (LH1) state to heavy-hole continuum states. Since the ground states are occupied by the light holes with smaller effective mass, the linear optical absorption is greatly enhanced over the unstrained case. In CSL p-type QWIPs, the intersubband absorption occurs between the ground (HH1) heavy-hole state to excited heavy-hole states (e.g. HH 3). Apart from the reduction of hole effective mass by compressive strain in the quantum well, the conducting holes behave like light holes with higher hole mobility due to the resonance between heavyhole and light-hole states. The energy band diagram for TSL and CSL p-type OWIP is shown in Figure 4.7. 4.2.2 Q W i P p a r a m e t e r s
Single OWIP detectors are usually rated by four figures of merit" absorption, dark current, responsivity, and detectivity. The photocurrent I t, generated in a QWIP can be written as Ip - npqv
(3)
Figure 4.7 Schematic energy band diagram for p-t!tpe tensile strained-la!ler (TSL) and compressively strained-layer ( CSL ) QWIP.
GalnAs(P) based O_WIPs on (;aAs. InP. and Si substrates for focal plane arra!ts
1] 1
where np is the volume density of photogenerated carriers, q the electron charge, and v is the transport velocity through the OWIP. For example in n-type OWIPs, np is given by Hp
z
Pcos0c~ Pe l'I, hid
where P is the incident optical power on the detector, p,, is the probability of escape from the q u a n t u m well, 0 is the angle between the incident light and the direction perpendicular to the q u a n t u m wells, and rL is the recapture lifetime. From this information we can calculate the detector's responsivity, which is defined as the ratio of the electrical current generated by the detector to the total optical power absorbed. Expressed mathematically, the peak responsivity Rp m the m a x i m u m value of responsivity as a function of photon energy - - is Rp =
Ip
Pcos0
=
q O.p,,g hv
(4)
where 77,, is double-pass absorption q u a n t u m efficiency and g is the photoconductive gain. Given the length of the m u l t i - q u a n t u m well region as 1. r/,, and g are defined as -
1 - exp(-2cd) 2
(s)
and VrL = _ra = _L
g-
1
(6)
1
rT
where r r = l / v is the transit time and L is the hot electron mean free path. The overall q u a n t u m efficiency 17is calculated as 77 - rl,,pe -
Rp h 1) qg
(7)
This gives an indication of how many carriers are being generated, which may contribute to the photocurrent, versus how many photons are absorbed. Another very important type of current in a OWIP. the dark current, can be calculated from I,(v)
- ,*(V)qv(V)A
(8)
where n*(V) is the effective number of electrons excited from the well into the c o n t i n u u m state, v ( V ) is the average transport velocity, and A is the detector area. n * ( V ) i s expressed as: 41
132
Handbook of Infrared Detection Technologies
~ f(E)T(E, V)dE n* ( V) - ~ 2m*Lp J E~
(9)
where Lp is the superlattice period. T(E. V) is the tunneling current transmission factor for a single barrier, andf(E)is the Fermi factor, given by
f(E) --
1 + [ e x p ( E - E1 - E,:)/kT]
where E is relative to the bottom of the q u a n t u m well, Et: is the Fermi level, k is Boltzmann's constant, and T is the temperature. The average transport velocity v(V) is given as
p.F -
V/1 + (#F/v,) 2 where # is the mobility, F is the average electric field, and vs is the saturated drift velocity. Related to the dark current is the noise current, which ultimately becomes the limiting factor for the sensitivity of a detector: the signal (responsivity) of a detector can be externally amplified, but the noise current would be amplified as well, leaving the signal-to-noise ratio unchanged. Several kinds of noise have been identified in photodetectors, though there are just two that can dominate the noise current in a QWIP. 42 The first is called generation-recombination noise, or GR noise. It is composed of r a n d o m thermal excitation and decay of carriers in the q u a n t u m wells, leading to fluctuations in the n u m b e r of carriers. The second is Johnson noise, which is a strictly thermal noise, and causes fluctuations in the velocity of carriers. Since noise is a r a n d o m process, the addition of noises obeys the root-mean-square (rms) system, that is, the square of the total noise equals the sum of the squares of the component noises. The total noise current i,, in a OWIP, then, is
In 29
_
i2R 27 llohnson .2
The GR noise current and the Johnson noise current are given by
icR -- v/4qlagAf and
~
( 1 O)
T
where Af is the m e a s u r e m e n t bandwidth and R is the series resistance of the QWIP.
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arra!ts 13 3
Lastly, the specific detectivity gives the signal-to-noise ratio of the detector. It is an indication of how small a signal a detector can pick up. The specific detectivity D* is given by D* = Rp x/-AAf 9 In
(12)
It is a standardized figure of merit for both photovoltaic and photoconductive detectors and is independent of detector area or peak absorption wavelength. The general figure of merit of a 0WIP FPA is the noise equivalent temperature difference (NEAT). NEAT is the m i n i m u m temperature difference across the camera target that would create a signal-to-noise ratio of unity. It's given by 4 ~ NEAT =
v~Af
(13)
D*(dPB/dT)
where D* is the detector detectivity and P I~ is the background photon power. When D* is larger than 101() cm-Hzl/2/W, the NEAT is limited by the uniformity of FPA. 44 4.2.3 Comparison of n-type and p-type Q WIPs
Today, almost all the OWIP FPAs are made of n-type doped OWIP detectors (see Section 4.1.3). The main advantages for n-type 0WIPs are higher responsivity and detectivity due to the higher electron mobility, higher optical absorption and photoconductive gain related with the smaller electron effective mass compared with p-type OWIPs. Today, the values of detectivity for n-type OWIPs are generally higher than that of p-type QWIPs at the same peak wavelength. The obvious drawback in n-type OWIPs is the zero normal incidence absorption due to the q u a n t u m mechanical selection rules. Different light coupling schemes (such as gratings) are thus needed for the n-type QWIPs to obtain better detector performance. This usually complicates the fabrication process of 0WIP FPAs and also limits the pixel size. The normal incidence absorption feature of p-type QWIPs allows the fabrication of grating-less FPAs. Another advantage of p-type 0WIPs is the Fermi Table 4.4
n-type
p-type
Comparison b e t w e e n n-type and p-type QWIP detector
Advantages
Disadvantages
9Higher responsivity and detectivitv due to the higher electron mobility and low electron effective mass 9Detectionrange from SWIRto VLWIR 9Normal incidence absorption 9Grating-less FPA 9Lowerdark current
9Light-couplingrequired 9ComplicatedFPA processing and limited pixel size 9Lowerresponsivity and detectivity 9Difficultyfor VLWIRdetection
134 Handbook of Infrared Detection Technologies level pinning effect at higher doping density, in which the Fermi level above the ground bound state in the OW is nearly independent of the doping density in the OW. Thus, it is possible to dope the OW at a higher doping density to increase the linear absorption coefficient and the responsivity without increasing the dark current in the detector. 4.2.4 Growth, fabrication and device characterization of a single QWIP device
The most widely used growth technologies for OWIP devices are molecular beam epitaxy (MBE) and metalorganic chemical vapor deposition (MOCVD). MBE is an ultra-high vacuum thin film technology developed in the early 1970s. 4s Both gas source (GSMBE) and solid source (SSMBE) can be used for material growth. The advantages of this growth technology are an ill sitll controllable growth process, extremely precise control over layer thickness and doping profile, and high uniformity ( > 3 inch). Its main disadvantage is the high expense. MOCVD has established itself as a unique and important epitaxial crystal growth technique which yields high quality low dimensional structures for fundamental semiconductor physics research and devices. 4~' This growth technique is attractive in its ability to grow uniform layers, sharp interfaces, and for its commercial applications. The details of MBE and MOCVD technology have been described in many publications. 4 s-4 r In general, a OWIP structure is grown on a semi-insulating substrate (e.g. InP, GaAs) and consists of 2 0 - 5 0 periods of multiple q u a n t u m well (MOW) sandwiched between a top and bottom contact layer. The q u a n t u m well is doped n or p-type, as desired. Once a OWIP structure has been grown either by MBE or MOCVD, it must undergo a fabrication process to create a working detector. The basic fabrication procedures are: 1. Cleaning of sample in the order of trichloroethylene (TCE), acetone, methanol, and DI water. 2. Positive photolithography to define the detector pixel and light coupling structure (if needed). 3. Etching of the exposed semiconductor, either by dry etching {e.g. electron cyclotron resonance-enhanced reactive ion etch (ECR-RIE)) or wet chemical etching. 4. Negative photolithography for the metal contact definition. 5. Metal contact evaporation. 6. Lift-off process and alloying. Z Device packaging (die-bonding and wire-bonding). A finished GalnAs/InP OWIP mesa detector is shown in Figure 4.8. As discussed in Section 4.2.1 for n-type QWIPs, since there is no absorption for normal incidence, a light coupling scheme needs to be applied on the OWIP device for detection of the desirable normal incidence illumination. For a single n-type OWIP detector, as one step of device fabrication, a 45 ~ facet is usually polished on the substrate for light coupling. Different coupling schemes,
GalnAs(P) based QWIPs on (;aAs. InP. and Si substrates for focal plane arra!ls
13 5
Figure 4.8 A schematic diagram of the GalnAs/InP QWIP on InP substrate. This is how the device appears after fabrication, with the mesas etched and the metal contacts deposited.
for mull-element detectors, like diffraction gratings, 4~ random scattering reflectors, 49 microlenses, ~(~ and corrugated structures ~1 have also been developed. Several device characteristics, like the absorption spectra, relative spectral response, blackbody responsivity, noise current, and dark current, are usually measured to evaluate the performance of OWlP devices. The packaged OWIP device is usually tested inside a liquid nitrogen cryostat. The absorption coefficient given by an absorption spectra m e a s u r e m e n t can be used to calculate the room temperature (T=300K) and low temperature ( T = 7 7 K ) absorption q u a n t u m efficiency using r/~.~2 The relative spectral response gives information about a detector's peak absorption wavelength, cut-off wavelength, and spectral linewidth. Both the absorption spectra and relative spectral response can be performed with a Fourier Transform Infrared (FTIR) spectrometer. The absolute magnitude of the responsivity can be accurately determined by measuring the photocurrent, Ip, with a calibrated blackbody source that is at a specific temperature. Noise current m e a s u r e m e n t can be performed on a fast fourier transform (FFT) spectrum analyzer. The dark current of the OWlP can be measured with a semiconductor parameter analyzer and a dark shield placed around the OWIP inside the cryostat. Once the noise current and dark current are known, the overall q u a n t u m efficiency, photoconductive gain, and specific detectivity of a detector can be calculated by the formulas given in Section 4.2.2.
4.3 Fabrication of infrared FPA 4.3.1 Infrared FPA fabrication steps
As we mentioned in Section 4.1.1, most of the infrared detector applications are based on FPA cameras. So, how to make an operational infrared FPA also becomes an important issue for OWIP researchers. In this section, we will give a
136
Handbook of Infrared Detection Technologies
Figzire 4.9 Schematic qf ir(lrared FPA fabrication process.
GalnAs(P) based O WIPs on (;aAs. InP. and Si substrates for focal plane arrays
137
general description of the infrared FPA fabrication process. A schematic of infrared FPA fabrication is shown in Figure 4.9. Many technologies and much equipment are involved in this complicated fabrication process. The details of this process are described below step by step. 1. Growth of an infrared detector structure on the appropriate substrate. The growth is performed by MOCVD or MBE technology. This step basically determines the final FPA~sperformance. Different characterization will be performed to ensure that the detector's properties meet the requirements. 2. Fabrication of detector pixels. This includes the UV-photolithography for the sample patterning and the etching (like ECR-RIE dry etching or wet chemical etching) for the pattern transfer. For large format FPAs, the pitch size (array spacing) is usually between 30 and 50 l~m. The detector pixel size is even smaller. 3. Detector pixel passivation. Si02 or Si3N4 are common choices for the passivation material. The deposition of Si02 and Si 3N4 can be performed by plasma enhanced chemical vapor deposition (PECVD). The passivation layer will neutralize the surface states and insulate the covered areas. 4. Patterning and etching of the passivation layer. The purpose of this step is to make openings for the metal connection to each detector pixel. The opening size depends on the pixel size. 5. Fabrication of indium solder bumps. This is the most complicated step during the FPA fabrication and will be described in more detail in the next section. Indium bumps will be used as the interconnection between detector and ROIC pixels. 6. Flip-chip bonding of the detector substrate with a Si-based ROIC. The schematic of the alignment procedure is shown in Figure 4.10. An optical head will be brought between the upper and lower chips. The surface image of both upper chip and lower chip will be displayed on a monitor with the help of two video
Figure 4.10 Schematic offlip-chip bonding procedure.
138
Handbook of Infrared Detection Technoloqies
Figure 4.11 An indimn joint ~(fter reflow process.
cameras. The alignment can be done either manually or automatically. After fine alignment and parallelism, the optical head will be removed and the upper and lower chip will be pressed together with preset pressure and temperature. The indium joint needs to be refiowed at the preset temperature within an inert atmosphere. A refiowed indium joint is shown in Figure 4.11. 7. Underfill dispensing between FPA and ROIC. Underfill is usually based on an epoxy system. 53 The epoxy underfill provides the necessary mechanical strength to the detector array and readout hybrid, prior to the thinning process. It also reduces the effect of global thermal expansion mismatch between the detector array and readout chip and protects the FPA hybrid, from moisture, ionic contaminants, and hostile operating environments such as shock and vibration. 8. Detector substrate thinning process. This is usually performed by abrasive polishing or wet chemical etching. In some cases, the substrate is completely removed. The purpose of this step is to further reduce or completely eliminate (for substrate removal) the thermal expansion mismatch between detector array and readout chip. This basically to allow the OWlP FPAs to go through an unlimited number of temperature cycles without any indium bump breakage or delamination. Furthermore, the substrate thinning process also eliminates the pixel-to-pixel optical cross-talk and significantly enhances the optical coupling of infrared radiation into the OWIP pixels. 54 9. Packaging of the FPA hybrid. The hybrid will be loaded onto a lead-flee ceramic chip carrier (LCCC) by the die-bonding process and input/output metal pads on the readout chip will be connected to the LCCC pins by a wire-bonding process. 10. The final step of OWIP FPA fabrication is the test and evaluation of the FPA hybrid's performance. The test system is usually composed of a liquid-nitrogen Dewar, a cold clamp assembly, a camera head, digital electronics, and computerbased testing software.
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arra!ls
139
4.3.2 Indium solder bump fabrication steps
Indium has become the most important mating material for infrared FPAs due to the fact that it stays ductile at liquid helium temperature, is easy to work with, and forms a good bond at low temperature. Highly-uniform and tall indium bumps are required for the infrared FPA fabrication. It has been shown that taller indium bumps can increase the FPA hybrid's reliability and thus its operating life. ~~ An indium solder bump is chiefly fabricated by two technologies: evaporation and electroplating. The comparison between the two technologies is shown in Table 4.5. In this section, the low-cost electroplated indium bump technology will be described. Figure 4.12 shows the schematic representation of the fabrication process for an electroplated indium solder bump. ~6-~s The details of an indium solder bump fabrication process are given below step by step. 1. Passivation of detector pixels. This is same as step three in the FPA fabrication described above. 2. Deposition of a Ti shortcut layer. This layer serves as a current path for the later electroplating process. 3. Thick photoresist ( > 10 ~m) lithography. The opening on the thick photoresist provides the mold for the indium electroplating process. Very strict control of photoresist spinning and UV-exposure is required to assure a good mold profile, s9 4. Electroplating of Under-Bump-Metallurgy (UBM) layers. A high reliability solder bump interconnect is comprised of two features: the UBM, also known as the Ball Limiting Metallurgy (BLM), and the solder ball itself. UBM is usually composed of several different metal layers and it provides the following features: good adhesion to the detector pixel metal pad, an effective solder diffusion barrier, and a solder wettable metal of appropriate thickness. 6~ For indium solder bumps, Ti/Au can be used as UBM layers. 5. Electroplating of indium solder bumps. This is performed in an appropriate indium plating bath (e.g. indium sulfamate) under a controlled current density. The plated indium bumps usually have a mushroom-shaped profile before reflow (see Figure 4.13 (a)). Table 4.5
Comparison between evaporated and electroplated solder bumping technology Advantages
Disadvantages
Evaporated solder bumping technology
9 Highly uniform bump 9 Excellent metallurgical control of bump
Electroplated solder bumping technology
9 Lower equipment costs 9 Simple fabrication process
9 High cost due to expensive evaporation equipment 9 Limited to material with high vapor pressure 9 Complicated fabrication process 9 More pollution to environment. Why? 9 Low t h r o u g h p u t 9 Non-uniform bump height 9 Induced subsurface wafer stress 9 Poor alloy control
140 Handbookof Infrared Detection Technologies
Figure 4.12 Schematic of the indium solder bump fabrication process.
6. Removal of the thick photoresist. This can be done with Acetone or photoresist stripper. 7. Reflow of the indium solder bumps. The purposes of reflow are to increase bump height by reshaping the indium into a sphere (see Figure 4.13(b)) and to facilitate the flip-chip bonding alignment. Reflowed indium bumps have also been shown to have higher reliability. 61 Reflow is usually performed in a rapid thermal process oven (RTP) under N2 atmosphere. Some novel reflow methods utilizing ECR-RIE and low-pressureMOCVD were also demonstrated. 62'123 These methods can effectively remove the indium oxide on the surface of indium bumps to ensure the high quality reflow. 8. Removal of Ti shortcut layer. This can be done by wet chemical etching with dilute HF (2.5 %). 4.3.3 ROIC for infrared FPA
As we described in Section 4.3.1. infrared FPAs are mated via indium bump technology to silicon based readout integrated circuits (ROIC) which allow electronic access to every single detector pixel in the array. The ROIC serves as an electrical interface between the infrared FPA and off-focal plane signal
GalnAs( P) based OWIPs on GaAs. InP. and Si substrates for focal plane arrays
141
Figure 4.13 (a) Mushroom-shaped indium sohh'r bump before reJlo~v. (b) Indium solder ball after reflow process.
processing electronics. A ROIC is usually a hybrid between a multiplexer and an array of low power circuits which, at a minimum, provide a precise bias voltage to the detectors and extract the photocurrent from them with the least possible noise. During the FPA operation, the integration capacitor in the ROIC unit cell stores the charges provided by an individual detector photocurrent within a fixed integration time. The resulting signal is multiplexed to a serial stream of sampleand-hold signals within the fixed readout time. Additional focal plane circuitry is usually required to improve the uniformity, signal-to-noise, and versatility of the FPA. In most ROICs, the signal is pre-amplified at the front-end stage to avoid further corruption by other noise sources. A number of readout preamplifiers, such as self-integrator (SI), source-follower detector (SFD), direct injection (DI), source-follower direct injection (SFDI), and capacitor feedback trans-impedance amplifier (CTIA), have been utilized for infrared FPA applications. (~3 Figure 4.14 shows a ROIC unit cell with the most popular DI type preamplifier. 64 The detector bias voltage is controlled by the Vbias_adj gate and the signal is multiplexed out from the sample and hold capacitor during the readout period. The anti-bloom gate keeps the input circuit from saturating. Today, most ROICs are built by advanced sub-micron cmOS technology which has the advantages of higher operation frequency and capability for random access. 65 Many types of large format ROICs are currently commercially available. 66'67'126 The general requirements of ROICs for infrared FPAs are: stable detector bias; small input resistance: low nonlinearity: low power consumption; and large dynamic range. (~ For a specific type of infrared detector, the ROIC has to be tailored to the detector characteristics.
4.4 p-type QWIPs Hole intersubband absorption is attractive because the strong mixing of heavy and light holes in the valence band (at k~:()) allows the desirable normal incidence illumination geometry to be used. To fully utilize the advantage of
142 Handbook of Infrared Detection Technologies
Figure 4.14 Simplified schenmtic of a ROICmlit cell based on a direct injection input circuit.
p-type intersubband transition, p-type OWIPs based on different material systems have been studied by m a n y groups. In this section, we will give some examples of p-type OWlP for detection of different wavelength range. All of the OWIP structures reported here are based on Al-free GalnAs(P) material systems. 4.4.1 p-type MWIR QWIPs p - t y p e GaAs/GalnP
J. Hoff et al. reported MWIR p-type OWIP based on a lattice-matched (no strain) GaAs/GaInP MQW structure. 69 Three p-type OWIP structures were grown on semi-insulating GaAs substrate by a low-pressure MOCVD reactor. Each structure contained 50 GaAs q u a n t u m wells separated by nominally 280 A wide Gao.51 Ino.49 P barriers. These superlattices were sandwiched between thick GaAs layers used for top (0.5 l~m) and bottom (1.() l~m) contacts. All GaAs layers were doped with Zn to a net acceptor concentration of 1 x 1 ()lg cm-3. Different GaAs well widths (22A, 32A, and 55A) were used for the three different structures. 4 0 0 x 4 0 0 l~m mesa detectors were fabricated using photolithography and wet chemical etching, and 1 5 0 x l S ( ) l ~ m square Au/AuZn electrodes were evaporated and alloyed. The devices were front-illuminated under normal incidence. The spectral response shifts towards longer wavelength as the well width is reduced, which is expected as the hole ground state is pushed up towards the top of the barrier. Dark current was measured as a function of bias for the 22 A well at a different temperature (see Figure 2 in reference 69). At low bias, tunneling is negligible and the dark current originates from thermionic emission above the top of the barrier. The activation energy of holes in the q u a n t u m well can be calculated directly from the dark current data. Since dark current is dominated by thermionic emission from the q u a n t u m well at temperatures greater than 4 5 K, there is an Arrhenius dependence of dark current with temperature.
GalnAs( P ) based OWIPs on GaAs, InP, and Si substrates for focal plane arrays
l n ( I D / T ) oc - E A / k T
14 3
(14)
Where EA is the activation energy out of the quantum well. The EA revealed from this method for the three samples are 2 3 5 , 3 0 0 . and 330 meV for the 22, 32, and 55 A wide well QWIP respectively. These activation energies correspond to wavelength of 5.28, 4.13, and 3.75 l~m, respectively. Those values closely match to observed cut-off wavelength in photoresponse curves (see Figure 1 in reference 69). The detectivity of 8 x l O ~ cmHzl/2W-] at 2.5 l~m and T=77K was obtained for the 22 A well QWIP. This low detectivity is related to its small gain and absorption quantum efficiency (,-, 1.4 x 1 ()- 3) at 2.5 Bm. Each of these devices was found to be background limited at liquid nitrogen temperatures ( 77 K) or higher. Typical p-type GaAs/A1GaAs OWIPs peak within 3 0 - 5 0 meV above the cut-off wavelength. 7~ The magnitude of the photoresponse decreases with decreasing wavelength thereafter. However. the peak response in these GaAs/ GaInP QWIPs is typically 2 0 0 m e V or more above the cut-off wavelength. Moreover, the 22 A wide well sample appears to have two peaks to its spectral response. These unique spectral shapes were proven, through the device modeling by a Kane Hamiltonian, to be caused by the influence of the spin splitoff band. Therefore, since there is significant coupling between the heavy hole and the split-off and between the heavy hole and the light hole, the observed spectral shapes result from the dual influence of split-off extended state absorption with light-hole extended state absorption. The lower energy peak seen in the 22 A wide sample corresponds to the appearance of the second lighthole state in the continuum of the standard q u a n t u m well. The 32 A and 55 A wide well samples are wide enough for the second light-hole state to be buried in the standard quantum well so it cannot contribute to photoresponse. p-type GaAs/GalnAsP QWIP
To increase the cut-off wavelength, a lattice-matched GaAs/ Gao.71Ino.29Aso.39Po.61 OWIP was grown by LP-MOCVD with 50 periods of 30 A wide GaAs quantum wells separated by 280 A wide GaInAsP (Eg=l.8 eV at T= 300 K) barriers. 72 The effect of the quaternary alloy was to reduce the valence band offset relative to lattice matched GaAs and thus increase the cut-off wavelength. A photovoltaic effect is observed from the photoresponse curve (see Figure 1 in reference 72). This arises from an asymmetric quantum well potential profile. Such asymmetry could be related to either a structural difference in the two q u a n t u m well interfaces or to dopant migration during the material growth. The photoresponse has a peak around 4 lum with broad maxima. Compared with the GaAs/GaInP OWIP reported in the last section, the peak wavelength increased slightly. This OWIP remains background limited up to a detector temperature of 100 K for biases between - 7 . 5 V and +2.5 V. The activation energies derived from Arrhenius plots of clark current, 170, 204. and 226 meV for bias voltages +9, - 9 , and 0.1 V respectively, agree with the cut-off wavelengths observed in the photoresponse curve. 72
144 Handbookof Infrared Detection Technologies p-type GalnAs/InP QWIP Sengupta et al. reported a lattice-matched GaInAs/InP p-type QWIP. 7~ The
QWIP structure was grown by GSMBE on InP substrate and consisted of 30 periods 10 A uniformly Be-doped (p= 3 x 1 () 1~ cm- 3) Ga(~.47In(}.s ~As wells and 500 A thick, uniformly Be-doped (1 x 1 ()l 7 cm- ~) InP barriers. 200 ~m diameter mesa was fabricated by wet chemical etching. The 4 5 ~ incidence responsivity at 80 K was measured as a function of wavelength at different bias voltage (see Figure 2 in reference 73). The peak photoresponse shifts slightly with bias and an increase in magnitude of response was observed with positive bias voltage. They attributed the increase in response with magnitude of the bias voltage to the increased photoexcited carrier escape and the increased carrier drift velocity. 4.4.2 p-type LWIR Q WIPs p-type GalnAsP/GalnAsP QWlP
To achieve LWIR detection, J. Hoff et al. proposed a fully quaternary ptype O W I P . 74'7~ The result, according to theory, should be a QWIP with a cut-off wavelength of approximately 1() ~m. p-type lattice-matched Gao.87Ino.13Aso.74Po.26/Ga(~.~,2In(~.3sAs(j.22P(j.7~ QWIP with 30 A wells and 2 8 0 A barriers was grown by LP-MOCVD on GaAs substrate. One remarkable feature of this q u a t e r n a r y - q u a t e r n a r y QWIP is that the photoresponse is extremely broad, from 2.5 lam to 1 () btm. 74 This unique spectral shape occurs because of the dual influence of the light-hole extended state and the split-off extended state. It also indicates that high quality quaternaryquaternary superlattices should be inherently capable of multi-color photoresponse.
4.5 n-type QWIPs Compared to p-type QWIPs, it is easier to realize longer infrared wavelength detection with n-type OWIPs due to the smaller conduction band discontinuity. Also, due to the smaller electron effective mass and higher mobility, n-type OWIPs have shown excellent detector performance. In this section, we give some n-type QWIP examples that cover the wavelength range from MWIR to VLWIR. In addition, a multi-color QWIP example is also given at the end. 4.5.1 n-type t WIR Q WIPs
LWIR is the most important atmospheric window since room temperature objects have a 10 l.tm peak wavelength. LWIR QWlPs based on m a n y different material systems have been demonstrated. In this section, detector performance of three n-type GaInAs/InP OWlP samples will be described.
GalnAs(P) based OWIPs on GaAs, InP. and Si substrates for focal plane arra!ts
145
n - t y p e GalnAs/InP Q W I P
GalnAs/InP OWIP was firstly fabricated by Gunapala et al. in 1991. 76 Compared to the most popular GaAs/AlxGal_xAs OWIPs, the Al-free GaInAs/InP QWIPs have the following advantages" 1. more straightforward device fabrication process since no passivation is required; 2. higher electron mobility due to the lower effective electron mass: 3. binary InP barrier layers have an inherently lower defect density which results in lower dark current. Samples (A, B, and C) were grown on semi-insulating InP subsrate by a LPMOCVD. 77 Each QWIP structure consists of 20 periods of 60 A GaInAs wells, separated by 500 A InP barriers. Only the middle 50 A of each well was doped with Si and doping densities are 1.7• 5• and 1.7x1018 cm -3 for sample A, B, and C, respectively. 4 0 0 x 4 0 0 Bm OWIP mesa detectors were fabricated by photolithography and ECR-RIE dry-etching. Since each mesa has 4 5 ~ sidewall after dry-etching (with BCI ~/C12/Ar etch chemistry), the OWIPs were measured with normal incidence with front-illumination geometry. The dark current for the three OWIPs was measured as a function of bias voltage (see Figure 3 in reference 77).The large increase of dark current with the doping density in the well is due to the decrease in activation energy. The peak detection wavelength was measured at 9.013m for all three samples (see Figure ] in reference 77). One noteworthy aspect to point out is the exponential dependence of the responsivity with applied bias (see Figure 2 in reference 77). '
I
'
I
100 ~~ >
j~
10
oo
/JJ
jl
ill /
Z
~mm /
. m
oo c
'
1
m~ m m~
/
0.1 0.01
.f
!
nh
1 E-3
, 0
I
2
,
Bias (V)
I
4
,
-
6
Figure 4.15 Peak responsivit!l of a corrugated (;alnAs/InP C)iVII~with corrugation period 10 I~m. T= 77 K.
146 Handbookof Infrared Detection Technologies This behavior is very different from what is usually observed in GaAs/AIGaAs QWIPs, whose responsivities increase linearly with bias then saturate at higher bias. A likely cause is avalanche gain, which is due to impact ionization of electrons excited out of the quantum wells. 7s Sample B yields a maximum detectivity of 3 . 5 x l O l ~ at an operating bias 0.75V. Further, by changing the doping density by a factor of 10, the detectivity changes by over a factor of 20 (see Figure 6 in reference 77). With a more effective light coupling scheme, the detector performance of GaInAs/InP OWIPs can be improved dramatically. Figure 4.15 shows a peak responsivity curve for a GaInAs/InP OWIP which is covered by corrugation on its detector mesa. 79 This QWlP has the identical structure parameters as sample B, and the corrugation period is 1()~m. The corrugation is created with dryetching by a ECR-RIE and has approximately 60 ~ sidewall. Compared with a peak responsivity of 33.2 A/W at a bias voltage of + 5 V for sample B, 77 a record high responsivity of 218 A/W was obtained at bias of + 5 V (mesa top positive) for this corrugated GaInAs/InP QWIP. 4.5.2 n-type VLWIR Q WIPs
n-type InGaAsP/InP Due to the smaller conduction band offset between GaInAsP and InP, peak photoresponse wavelengths ranging from 8-2()IJm and beyond are possible by changing the composition of the quaternary, g() Two lattice-matched GaInAsP/ InP QWIPs (A and B: E,=O.82eV) and one GaInAs/InP (C, the same structure parameter as sample A from Section 4.4.1) was grown by LP-MOCVD. gl The GaInAsP/InP OWIP structure consists of 20 periods of 65 A GaInAsP wells, separated by 500 A InP barriers. The bandgap of well material for sample A and B are 0.95 eV (2=1.3 ~tm) and 0.80 eV (,;.=1.55 lam). The n doping level in the well is 1.7 x 10 ~7 c m - 3. The relative response curves were measured by a FTIR at 80 K (see Figure 4 in reference 81 ). For a bias of - 1 V (mesa top negative), the 50% cut-off wavelengths for the three samples are 9.3, 10.7, and 14.2 ~m for A, B, and C, respectively. The FWHM (as Ak/)~) of the three samples are 36%, 10%, and 5.5% for A, B, and C, respectively. The switch from a bound-to-bound to a bound-to-continuum transition between sample B and A is evident. The measured peak detectivity for the three samples were 1.1 x l 0 s cmHzl/2W -1 (T=3OK), 2x l 0 s cmHzl/2W-~ (T=80 K), and 8x 1 0 9 cmHzl/2W -1 (T=80 K) for sample A, B, and C, respectively. n-type InGaAs/GaAs QWIP The longest detection wavelength for OWIP was reported by Perera et al. on a InGaAs/GaAs QWIP with cut-off wavelength at 35 ~tm. 82 The OWIP structure consisted of 20 periods of 9 8 A In().()sTGa().gj3As wells (Si doped to n=4xlOl~ -2) and undoped 4 1 8 A GaAs barriers. This QWIP structure showed a bound-to-quasibound intersubband transition. 240 • 240 ~m detector mesas were fabricated by wet chemical etching. The peak photoresponse was
GalnAs(P) based QWIPs on GaAs. InP, and Si substrates for focal plane arrags
14 7
measured about 35~tm at T=4.2K and bias O.15V. The measured peak at responsivity and detectivity were 0.3 A / W and l.O• T=4.2 K. 4.5.3 Multi-color QWIPs
Compared to QWIPs based on GaAs substrates, a major advantage of InP-based OWIPs is that both 3-5 ~m and 8-20/3m spectral bands can be realized with lattice-matched n-type devices (see Figure 9 in reference 83 ), which is not true for GaAs-based QWIPs. A two-color n-type QWIP was grown on semi-insulating InP by GSMBE.s4 The MWIR MOW consisted of 20 periods of 3 5 0 A Ino.s2Alo.48As barriers and 35A Int).s3Gat~.47As (n-doped to 5 x 1 0 1 7 c m -3) quantum wells. The LWIR MOW consisted of 20 period 500 A InP barriers and 55A Ino.s3Gao.47As wells (n-doped to 5xl()lTcm-~). 4 ( ) 0 x 4 0 0 p m detector mesas were fabricated by ECR-RIE dry etching. Since no middle metal contact was made, this QWIP operated as a voltage tunable two-color detector. At T= 7 7 K, for bias less than 7 V, only the 8.5 Jam peak was observed and for bias larger than 7 V, the 4 lum peak was observed.
4.6 Low cost QWIP FPA integrated with Si substrate 4.6.7 Overview of QWIPs on Si
Currently, most QWIPs are grown on GaAs (GaAs/A1GaAs heterostructure) and InP (GaInAs/InP h e t e r o s t r u c t u r e ) s u b s t r a t e s . Unfortunately, the readout integrated circuits (ROICs) for FPAs are silicon-based. As seen in Section 4.3, complicated techniques, like flip-chip bonding, are necessary to hybridize the FPAs with the silicon ROIC. GaAs/A1GaAs OWIPs directly grown on Si have been reported, ss-87 OWIPs directly grown on Si can make use of large-area substrates, with higher thermal conductivity and mechanical strength. Most importantly, this arrangement makes monolithic integration with the ROIC possible. The major obstacle to this goal is the large lattice mismatch between silicon and III-V materials (4.1% between GaAs and Si and 8.1% between InP and Si), which creates very high densities of threading dislocation in the epitaxial layer. These dislocations dramatically deteriorate the III-V material's optical and electrical qualities. To reduce the threading dislocation densities, many growth techniques, such as ex situ or in situ thermal annealing, low temperature nucleation, lateral epitaxial overgrowth, and insertion of a very thick, graded, or superlattice buffer, have been developed to improve the quality of GaAs and InP grown on Si. ss-91 Thermal annealing effectively reduces threading dislocation densities both in InP and GaAs epitaxial layers grown on Si. Annihilation and coalesence of dislocations are caused by dislocation movement under the thermal stress induced by annealing. 9~ Due to the larger immobile In atom in InP, stress induced
148 Handbook of Infrared Detection Technologies by thermal annealing can reduce threading dislocation density in InP-on-Si with less interdiffusion problems than in GaAs. 4.6.2 Growth of GalnAsllnP QWlP-on-Si
In this section, growth process of a OWIP-on-Si will be described in detail. Two 20 period GaInAs/InP OWIP structures were grown on Si substrate and on InP substrate by LP-MOCVD, respectively. 1~ The Si substrates used are GaAs-on-Si manufactured by Kopin Inc. with a GaAs coating thickness of 1 ~m. The {1 ()0) Si substrate has a 4 ~ miscut toward the [()11] direction. This misorientation is necessary to prevent the formation of anti-phase domains in the growth of III-V materials o n Si. 46 The growth of this QWIP-on-Si starts with substrate deoxydation and deposition of two different thin buffer layers. The substrate is heated up to 550~ under AsH3 flow, followed by growth of a thin GaAs buffer layer at 510C. After that, the substrate is cooled to 4()()~ and a 5()0 A thick InP nucleation layer is grown at growth rate of 3() A/min. Next, a thick buffer is grown, followed by a thermal annealing cycle. The temperature is raised to 500~ and a 6 ~m InP thick buffer layer is grown at growth rate of 180 A/rain. After every 1 ~m of InP. the growth is interrupted and an in situ thermal cyclic annealing (TCA) is performed. From room temperature, every annealing cycle consists of a temperature increase to 55()~ for 3 rain, followed by a decrease to room temperature. This procedure is repeated 3 times for every TCA. After TCA, the temperature is raised to 500~ and the next 1 ~m InP layer is grown. PHi/He is flown during the whole TCA process. Threading dislocation densities and rms surface roughness, for a 2 0 x 2 0 l.tm field estimated by atomic force microscopy (AFM) image lc~4 for thick InP buffer layer on Si, are 2 • 10 5 c m - 2 and 42 A respectively. After the growth of the entire 6 ~m InP thick buffer layer, the temperature is decreased to 480~ and a GaInAs/InP is grown. The details of the GaInAs/InP OWIP structure was described elsewhere. 77 X-ray diffraction spectra for QWIPon-Si and OWIP-on-InP are shown in Figure 4. ] 6. Satellite peaks generated by the periodicity of the MOW structure were observed. Satellite peaks were observed to the ] 7 th and 8 ~h order for ()WIP-on-InP and QWIP-on-Si samples respectively. More satellite peaks correspond to better interface quality and layer thickness uniformity. 4.6.3 Detector performance of GalnAsllnP QWIP-on-Si
Figure 4.17 shows the dark current Id measured as a function of bias voltage for a 20 period GaInAs/InP QWIP-on-Si sample. 4()()• detector mesas are fabricated with the process described in Section 4.2.4. For comparison, the dark current of a GaInAs/InP OWIP-on-InP with an identical structure is drawn on the same figure. At a bias voltage of - 1 .()V, the Ia of the QWIP-on-Si sample is 2.039 l.tA at T = 7 7 K . With the temperature change from 77K to 120K, the I,lOf the OWIP-on-Si increases only one order of magnitude. This behavior is very
GalnAs(P) based OWIPs on (;aAs. InP. and Si substrates for focal plane arrays
149
Figure 4.16 The measured (002) X-ray d(ffraction spectra.for ()WIP-on-Si and OWIP-on-InP samples.
different compared to the QWlP-on-InP. which changes by three orders of magnitude from 7 7 K to 120 K. This temperature-insensitivity can be attributed to the threading dislocations present in the QWlP-on-Si. The threading dislocations contribute a large temperature-insensitive leakage current under bias. Figure 4.18 shows a comparison of I,l at a bias of - 1 V for both the QWlPon-Si and the QWIP-on-InP as a function of temperature. For the purpose of comparison, a leakage current of 2.0 ~A was subtracted from the Ia for the QWIP-on-Si sample. Figure 4.18 shows almost the same I,t versus T relationship within the whole temperature range for the two samples. ~ s The GaInAs/InP QWIP-on-Si detector has a peak response at 7.80 ~tm, while the peak response of the GaInAs/InP OWIP-on-InP is at 8.15 ~tm. The small blueshift (0.35 ~tm) is due to the presence of some residual strain in the epitaxial layers which leads to a small change of q u a n t u m well parameters. This blueshift was also observed in a GaAs/A1GaAs OWIP on Si. ~~'~ 7 The 77 K peak responsivity for the QWIP-on-Si and QWIP-on-InP as a function of bias is shown in Figure 4.19. The two detectors have identical peak responsivities up to a bias voltage of (). 75 V. Beyond that. the responsivity of the QWIP-on-Si starts to trail off. At a bias voltage of 4.0 V, the peak reponsivity of 29.96 A/W for QWIP-on-InP is about 18 times higher than the 1.64 A/W responsivity for the QWIP-on-Si. The large increase of responsivity at higher bias voltage is attributed to an avalanche mechanism in the QWIP detectors. 78 Due to the presence of threading dislocations, the efficiency of the impact ionization
150
Handbook of Infrared Detection Technologies
~20~-~a mple C QWIP-On-Si ,
103
.
,
.
,
9
,
~
.
i1
,
......... l]
77___~
103
~- 10-~
101 120 77
10_9 1011
05
=
,
-4
-2
0 Bias (V)
2
-In 4
10 -1
lo10-7 10 -9 1011
Figure 4.17 Dark current measured as afimction of bias voltage for QWIP-on-Si and QWIP-on-InP sample.
Figure 4.181 d at bias - 1 V versus Tfor QWIP-on-InP sample (solid squares) and QWIP-on-Si sample (open circles, subtract 2.0 i~A for Id).
process in QWIP-on-Si seems lower than that of the OWIP-on-InP. Similar tendencies are also observed for M-based GaAs/AIGaAs OWIPs-on-Si, although the InP-based OWIPs-on-Si have three times higher responsivitiy. Lastly, the GaInAs/InP QWIP-on-Si detector showed a high detectivity of 2.3 • 109 cmHzl/2/W at a field of 0.89 kV/cm at T=77 K. This detectivity is one
GalnAs(P) based QWIPs on GaAs. InP. and Si substrates for focal plane arrays i
151
9
sample QWIP-on-lnP_
> (/) E 8. (/)
/~
n~ -
sample QWIP-on-Si e ~ e j
~ q l ~
OjO /
0.1
0.01
1E_3
u
0
,
,
2
,
I
4
Bias (V) Figure 4.19 Tilepeak responsivity comparison between OWIP-on-InP (squares) and OWIP-on-Si( circles).
order of magnitude lower than that of the QWIP-on-InP, which is 3 . 3 • 10 l~ cmHzl/2/W. 4.6.4 How to fabricate a monolithic integrated FPA with Si substrate
With the successful growth of a GaInAs/InP OWIP on Si substrate, a monolithically integrated infrared focal plane array becomes feasible. In Figure 4.20, we show a proposed fabrication process based on n-type GaInAs/InP OWIPs. 1. Fabrication of ROIC on Si wafer. 2. Passivation of ROIC surface. 3. Thinning and polishing the backside of Si ROIC wafer. 4. Growth of n-type GaInAs/InP OWIP structure on the backside of Si substrate. 5. Fabrication of detector array by standard lithography and etching process. 6. Passivation of surface of detector array. 7. Lithography and etch through the Si substrate. 8. Deposition of connection metal between every detector pixel and its corresponding ROIC.
4.7 New approaches of QWIP N e w characterization techniques of QWIPs
Despite the rapid development of the quantum well (OW) infrared technology, the intrinsic properties of the QWIPs have not been directly measured under the operating conditions of the detector. However, these details would be extremely useful for detector optimization. New techniques need to be explored for better QWIP characterization. K.K. Choi et al. utilized the surface corrugation to probe
152
Handbook of Infrared Detection Technologies
Figure 4.20 Schematic of the proposed monolithic FPA fi~brication process.
the absorption coefficient and the photoconductive gain of a OWIP under different operating conditions. 9s.~, With a series of varied corrugation period QWIPs, by measuring the detector responsivity as function of period of corrugation, absorption coefficient ~ and gain g can be deduced by curve fitting accurately. Novel QWlP devices
For most applications of n-type QWIPs, effective light-coupling schemes are very important. Based on a corrugated structure. N.C. Das et al. incorporated dielectric or metal coverage on the top to improve performance of the corrugated quantum well infrared photodetectors (C-QWlP) in two wavelength regimes. 97 With non-absorbing Si3N4 coverage, the responsivity of the C-OWIP was improved 3.3 times in the 8 Ftm range and 1.8 times in the 14 ~m range. K.K. Choi et al. improved the C-OWIP detector performance by adding a center vertical trench in each of the corrugations. 9s Ting et al. also proposed a submonolayer QWIP structure for improved n-type normal absorption. ~8 With the insertion of a submonolayer of large lattice mismatched material in the quantum well, the wave function within the quantum well will be modified and therefore make the normal incidence possible. Development of two-color or multi-color OWIP detectors is highly desirable for future high performance IR systems. Without the stacking of different MOW, a postgrowth bandgap engineering technique was recently utilized to tune the wavelength of OWIPs and to achieve multi-color detection. L. Fu et al. used single
GalnAs(P) based QWlPs on GaAs. InP. and Si substrates for focal plane arrays
15 3
high-energy proton implantation and rapid thermal annealing to tune the spectral response ofOWIPs. A redshift as large as 1.8 l~m was observed. 99 With integration of OWIPs with other semiconductor devices, m a n y novel applications are made possible. Liu et al. reported a OWIP-LED for pixelless largearea imaging applications. 1~176OWIP (MWIR or LWIR) and LED (near-IR emission) structures are grown on top of each other. Under forward bias, the decrease of serial resistance of OWIP with the detection of MWIR or LWIR radiation increases the bias voltage on the LED, thus increasing its near-IR (NIP,) emission. This device is actually a MWIR and FWIR to NIR converter. The resulting NIR emission from theLED can be imaged easily using the well developed Si-CCD. Down to nanoscale
It is well know that dipole selection rules forbid absorption of photons polarized in the plane of the wells. Different light coupling schemes have been developed. To solve this problem fundamentally, people start to create the lateral electron confinement with different nanoscale structures, like q u a n t u m dot, q u a n t u m wire, and q u a n t u m grid, in addition to the usual vertical confinement by the material layers. S. Kim et al. demonstrated an InGaAs/InGaP q u a n t u m dot detector for MWIR detection. 1~176 The InGaAs q u a n t u m dots were formed by intrinsic strain due to a lattice mismatch. Normal incidence photoconductivity was observed at a peak wavelength of 5.5 ~m. X.O. Liu et al. demonstrated a 9 l~m A1GaAs/GaAs q u a n t u m wire infrared detector which was based on a V-grooved substrate. 1~ L. P. Rokhinson et al. demonstrated a q u a n t u m grid infrared detector, l~ The grid pattern, with a very narrow linewidth in the OWIP active region, was fabricated by electron-beam lithography and reactive ion beam etching. The coupling efficiency increase at very narrow linewidth, gave evidence for lateral q u a n t u m confinement.
4.8 Conclusions In this chapter, we have given a brief review of current Al-free GalnAs(P) QWIPs grown on InP, GaAs, and Si substrates. Device physics, fabrication, and m e a s u r e m e n t have been described in detail. Fabrication processes for infrared FPAs and indium solder bumps are also described in detail. After more t h a n 10 years of research and development, infrared OWIP detectors have moved from the laboratory to the thermal imaging market. Today, OWIP research is still very active and requires more development to reach the QWIP's full potential. Higher operating temperature and multi-spectral OWlP cameras are the most straightforward goals. With new processing technology, a m u c h smaller and larger format infrared FPA based on Al-free OWIP structures is likely to be available on the market soon. Even more promising are the same FPAs monolithically integrated on Si, which, as discussed (or should have been), can lead to a simpler process and a significant reduction in camera costs.
154 Handbookof InfraredDetectionTechnologies
References 1. L. Esaki and H. Sakaki, IBM Tech. Disc. Bull. 20, 2 4 5 6 ( 1977). 2.W. Mason and ]. R. Waterman, Appl.Phys.Lett. 74, 1633 (1999). 3. K. Adamiec, A. Rogalski and ]. Rutkowski, J.Tech.Phys. 38, 431 (1997). 4. A. Rogalski, Infra.Pbys.Technol. 3 8 , 2 9 5 (1997). 5. T. Tung, L. V. DeArmond, R. F. Herald. P. E. Herning, M. H. Kalisher, D. A. Olson, R. F. Risser, A. P. Stevens, and S. ]. Tighe, Proc. SPIE 173 5, 109 ( 1992). 6. T. ]. Lyon, ]. A. Vigil, ]. E. ]ensen, O. K. Wu, ]. L. Johnson, E. A. Fatten, K. Kosai, G. Venzer, V. Lee, and S. M. Johnson, J. Vac. Sci. Tecbnol. B 16, 1321 (1998). 7. T. Naegele, Electronics 7, 76 ( 1989). 8 H. C. Liu, C. Y. Song, A. Shen, R. Dudek, M. Gao, E. Dupont, Z. R. Wasilewski, M. Buchanan, P .H. Wilson, B. ]. Robinson, D. A. Thompson, G. ]. Brown, F. Szmulowicz, and J. Ehret, Proc. SPIE 4 2 8 8 , 130 (2001 ). 9. S. D. Gunapala, S. V. Bandara, J. K. Liu. E. M. Luong, S. B. Rafol, J. M. Mumolo, D. Z. Ting, J. J. Bock, M. E. Ressler, M. W. Werner, P. D. LeVan, R. Chehayeb, C. A. Kukkonen, M. Levy, P. LeVan, and M. A. Fauci, Infra. Phys. and Technol. 4 2 , 2 6 7 (2001). 10. B. F. Levine, K. K. Choi, C. G. Bethea, J. Walker, and R. J. Malik, Appl. Phys. Lett. 50, 1092 (1987). 11. B. F. Levine, S. D. Gunapala, J. M. Kuo, S. S. Pei, and S. Hui, Appl. Phys, Lett. 59,1864(1991). 12. C. G. Bethea, B. F. Levine, V. O. Shen, R. R. Abbott, S. J. Hseih, IEEE Trans. Electron. Dev. 38, 1118 (1991). 13. J. Hoff, C. Jelen, S. Slivken, E. Mihel, O. Duchemin, E. Bigan, M. Razeghi, G. Brown, and S. M. Hegde, Appl. Phys. Lett 6~, 1130 (1994). 14. J. Hoff, S. Kim, M. Erdtmann, R. Williams, J. Piotrowski, E. Bigan, and M. Razeghi, Appl.Phys.Lett. 67, 22 (1995 ). 15. S. D. Gunapala, B. F. Levine, D. Ritter, R. Harem, and M. Panish, Appl. Phys. Lett. 60, 637 ( 1992). 16. M. Razeghi, H. Yi, J. Diaz, S. Kim, and M. Erdtmann, Proc. SPIE 3 0 0 1 , 243 (1997). 17. C. Jelen, S. Slivken, ]. Hoff, M. Razeghi, and G. J. Brown, Appl. Phys. Lett. 70,360(1997). ] 8. G. Hasnain, B. F. Levine, D. L. Sivco, and A. Y. Cho, Appl. Phys. Lett. 56, 770(1990). 19. S. D. Gunapala, S. V. Bandara, J. K. Liu, E. M. Luong, N. Stetson, C. A. Shott, J. J. Bock, S. B. Rafol, J. M. Mumolo, and M. J. McKelvey, IEEE Tran. Electron. Dev. 47, 326 (2000 ). 20. M. Z. Tidrow, X. Jiang, S. S. Li, and K. Bacher, Appl. Phys. Lett. 74, 1335 (1999). 21. S. D. Gunapala, S. V. Bandara, J. K. Liu, W. Hong, M. Sundaram, P. D. Maker, R. E. Muller, C. A. Shott, and R. Carralejo, IEEE Trans. Electron. Dev. 45, 1890(1998).
GalnAs(P) basedQWIPs on (;aAs. InP. and Si substrates forfocal plane arra!ls 15 5 22. H. Schneider, M. Walther, C. Sch6nbein. R. Rehm, J. Fleissner, W. Pleteschen, J. Braunstein, P. Koidl, G. Weimann, J. Ziegler, and W. Cabanski, Physica E 7, 101 (2000). 23. C. A. Kukkonen, M. N. Sirangelo, R. Chehayeb, M. Kaufmann, J. K. Liu, S. B. Rafol, and S. D. Gunapala, Infra. Ph!ls. Technol. 42, 39 7 (2001 ). 24. C. A. Kukkonen, S. D. Gunapala, S. V. Bandara, J. K. Liu, and J. Lorens, Proc. SPIE 3 6 9 8 , 706 (1999). 25. S. D. Gunapala, J. S. Park, G. Sarusi, T. L. Lin, J. K. Liu, P. D. Maker, R. E. Muller, C. A. Shott, and T. Hoelter, IEEE Trans. Electron. Dev. 44, 45 (1997). 26. S. D. Gunapala, J. K. Liu, J. S. Park, M. Sundaram, C. A. Shott, T. Hoelter, T. L. Lin, S. T. Massie, P. D. Maker, R. E. Muller. and G. Sarusi, IEEE TrAns. Electron. Dev. 44, 51 (1997). 2 7. S. D. Gunapala, S. V. Bandara, A. Singh, J. K. Liu, S. B. Rafol, E. M. Luong, J. M. Mumolo, N. O. Tran, J. D. Vincent, C. A. Shott, J. Long, and P. D. LeVan, Proc. SPIE 3 6 9 8 , 687 (1999). 28. P. Yuh and K. L. Wang, Phys. Rev. B 3 8, 38 ( 1988 i. 29. G. Bastard, Phys. Rev. B 24, 5693 (1981 ). 30. G. Bastard and J. A. Brum, IEEEJ. Ouantum Electron. OE-22, 1625 (1986). 31. J. R. Hoff, M. Razeghi, and G. J. Brown, Phys. Rev. B 54, 10773 (1996 ). 32. Y. C. Chang and R. B. James, Phys. Rev. B 3 9 , 1 2 6 72 (1989). 33. S. D. Gunapala and S. V. Bandara. Microelectronics J. 30, 1057 (1999). 34. S. Gunapala and K. M. S. V. Bandara, Physics of Thin Fihns 21, Aademic Press. New York, 113 ( 1995). 35. B. F. Levine, C. G. Bethea, G. Hasnain, V. O. Shen, E. Pelve, R. R. Abbott, and S. J. Hsieh, Appl. Phys. Lett. 56, 851 (1990). 36. L.S. Y u a n d S . S. Li, Appl. Phys. Lett. 59, 1332 (1991). 37 J. Chu and S. S. Li, IEEE ]. Ouant. Elect. 33(7), 1104 ( 199 7). 38. Y. H. Wang, S. S. Li, J. andChu, P. Ho, Appl. Phys. Lett. 64, 727 (1994). 39. J. Chu, S. S. Li, and P. Ho, Appl. Phzds. Lett. 69, 1258 (1996). 40. Y. H. Wang, J. Chu, S. S. Li, and P. Ho, J. Appl. Phys. 76, 6009 (1994). 41. B. F. Levine, C. G. Bethea, and G. Hasnain, Appl. Phys. Lett. 56, 851 (1990). 42. W. A. Beck, Appl. Phys. Lett. 63, 3589 ( 1993). 43 A. Zussman, B. F. Levine, J. M. Kuo, and J. de Jong, J. Appl. Phys. 70, 5101
(199]).
44. B. F. Levine, J. Appl. Phys. 74, R 1( 1993 ). 45. A. Y. Cho, The Technology and Physics of Molecular Beam Epitaxy. New York: Plenum Press ( 1985). 46. M. Razeghi, The MOCVD Challenge Vol 1. Bristol: Adam Hilger (1989). 4 7. M. Razeghi, The MOCVD Challenge Vol 2. Bristol: Adam Hilger (1995). 48. G. Hasnain, B. F. Levine, C. G. Bethea, R. A. Logan, J. Walker, and R. J. Malik, Appl. Phys. Lett. 54, 2515 (1989). 49. G. Sarusi, B. F. Levine, S. ]. Pearton, K. M. S. Bandara, and R. E. Leibenguth, Appl. Phys. Lett. 6 4 , 9 6 0 ( 1994 ). 50. N. T. Gordon, SemicondSci. Technol. 6, C106 (1991 ).
15 6 Handbookof InfraredDetection Technologies 51. C. J. Chen, K. K. Choi, M. Z. Tidrow, and D. C. Tsui, Appl. Phys. Lett. 68, 1446(1996). 52. G. Hasnain, B. F. Levine, C. G. Bethea, R. R. Abbott, and S. J. Hsieh, I. Appl. Phys. 67, 4361 (1990). 53. J. H. Lau, Low Cost Flip-Chip Technologies for DCA, WLCSP, and PBGA Assemblies, New York: McGraw-Hill (2000). 54 .S.D. Gunapala, S. V. Bandara, A. Singh, J. K. Liu, S. B. Rafol, E. M. Luong, J. M. Mumolo, N. O. Tran, J. D. Vincent, C. A. Shott, J. Long, and P. D. LeVan, IEEE Transactions on Electron Devices, 47, 963 (2000). 55. M. ]. Varnau, S. Yeh, Flip Chip/BGA Workshop, Binghamton, New York (October 1996 ). 56. T. Flynn, C. W. Argento, and ]. O'Brien, Proc. SPIE 3 9 0 6 , 8 (1999). 57. A. B. Frazier and M. G. Allen, ]. Microelectromech. Syst. :2, 87 ( 1993). 58. E. K. Yung and I. Turlik, IEEE Trans. Components, Hybrids, and Manufacturing Technology 14, 549 ( 1991 ). 59. W. W. Flack, S. White, and B. Todd, Proc. SPIE 3 6 7 8 , 49 (1999). 60. J. H. Lau (Ed.) Flip Chip Technologies. New York: McGraw-Hill (1996). 61. Y.-H. Kim, J.-H. Choi, K.-S. Choi, H. C. Lee, and C.-K. Kim, Proc. SPIE ] 0 6 1 , 60 (1997). 62. J. Jiang and M. Razeghi (unpublished). 63. H. H. Martijn, U. Halldin, P. Helander, ]. Alverbro, and ]. Y. Anderson, Proc. SPIE 3698, 789 (1999). 64. N. Y. Aziz, R. F. Cannata, G. T. Kincaid, R. J. Hansen, ]. L. Heath, W. J. Parrish, S. M. Petronio, and ]. T. Woo]away, Proc. SPIE 3698, 766 (1999). 65. F. Bertrand, ]. L. Tissot, and G. Destefanis, Proc. SPIE 3316, 713 (1998). 66. R. F. Cannata, R. ]. Hansen, A. N. Costello, and W. ]. Parrish, Proc. SPIE 3698,756(1999). 67. N. Y. Aziz, G. T. Kincaid, W. J. Parrish, J. T. Woolaway, J. L. Heath, Proc. SPIE 3 3 6 0 , 80 (1998). 68. H. Kulah and T. Akin, Proc. SPIE 3 6 9 8 , 778 (1999). 69. J. Hoff, X. He, M. Erdtmann, E. Bigan, and M. Razeghi, ]. Appl. Phys. 78, 2126(1995). 70. F. Szmulowiz and G. Brown, Appl. Phys. Lett. 66, 1659 (1995). 71.F. SzmulowizandG. Brown, Phys. Rev. B 52, 13203 (1995). 72. ]. Hoff, S. Kim, M. Erdtmann, R. Williams, I. Piotrowski, E. Bigan, and M. Razeghi, Appl. Phys. Lett. 67, 22 ( 1995). 73. D. K. Sengupta, S. L. Jackson, D. Ahmari, H. C. Kuo, ]. I. Malin, S. Thomas, M. Feng, G. E. Stillman, Y. C. Chang, L. Li, and H. C. Liu, Appl. Phys. Lett. 69, 3209(1996). 74. J. Hoff, C. Jelen, S. Slivken, G. ]. Brown, and M. Razeghi, Proc. SPIE 2 6 8 5 , 62(1996). 75. I. R. Hoff, M. Razeghi, and G. Brown, Phys. Rev. B 54, 10773 (1996). 76. S .D. Gunapala, B. F. Levine, D. Ritter, R. Harem, and M. B. Panish, Appl. Phys. Lett. 58, 2 0 2 4 (1991 ). 77. M. Erdtmann, A. Matlis, C. Jelen, M. Razeghi, and G. Brown, Proc. SPIE 3948,220(2000).
GalnAs(P) basedOWIPs on GaAs. InP. and Si substratesfor focal plane arrays 15 7 78. B. F. Levine, K. K. Choi, C. G. Bethea, J. Walker, and R. J. Malik, App1. Phys. Lett. 5 1 , 9 3 4 (1987). 79. J. Jiang and M. Razeghi, (unpublished). 80. S. D. Gunapala, B. F. Levine, D. Ritter, R. A. Hamm, and M. B. Panish, App1. Phys. Lett. 6 0 636 (1992). 81. M. Erdtmann, J. Jiang, A. Matlis, A. Tahraoui, C. Jelen, M. Razeghi, and G. Brown, Proc. SPIE 3 9 4 8 , 227 (2000). 82. A. G. U. Perera, S. G. Matsik, H. C. Liu, M. Gao, M. Buchanan, W. J. Schaff, and W. Yeo, Infra. Phys. and Technol. 42, 157 (2001 ). 83. M. Razeghi, M. Erdtmann, C. Jelen, F. Guastavinos, G. J. Brown, and Y. S. Park, Infra. Phys. and Technol. V42, 135 (2001 ). 84. C. Jelen, Gaxln~_.,.AsyP~_,j-Based n-type Long Wavelength Quantum Well Infrared Photodetectors: Growth, Characterization, and Fabrication. NU Doctoral Thesis, Evanston (1998). 85. D. K. Sengupta, W. Fang, J. I. Malin, J. Li, T. Horton, A. P. Curtis, K. C. Hsieh, S. L. Chuang, H. Chen, M. Feng, G. E. Stillman, L. Li, H. C. Liu, K. M. S. V. Bandara, S. D. Gunapala, and W. I. Wang, AppI. Phys. Lett. 71, 78 (1997). 86. E. R. Brown, F. W. Smith, G. W. Turner. K. A. Mclntosh, and M. J. Manfra, Proc. SPIE 173 5 , 228 ( 1992 ). 87. D. K. Sengupta, M. B. Weisman, M. Feng, S. L. Chuang, Y. C. Chang, L. Cooper, I. Adesida, I. Bloom, K. C. Hsieh, W. Fang, J. I. Malin, A. P. Curtis, T. Horton, G. E. Stillman, S. D. Gunapala, S. V. Bandara, F. Pool, J. K. Liu, M. Mckelvey, E. Luong, W. Hong, J. Mumolo, H. C. Liu, and W. I. Wang, J. Electron. Mater. 2 7 , 8 5 8 (1998). 88. T. E. Crumbaker, H. Y. Lee, M. J. Hafich, and G. Y. Robinson, ]. Vac. Sic. Technol. BS, 261 ( 1990). 89. Y. Ababou, P. Desjardins, A. Chennouf, R. Leonelli, D. Hetherington, A. Yelon, G. L'EspOrance, and R. A. Masut, J. App1. Phys. 8 0 . 4 9 9 7 (1996). 90. M. Yamaguchi, A. Yamamoto, M. Tachikawa, Y. Itoh, and M. Sugo, App1. Phys. Lett. 53, 2293 (1988). 91. S. Naritsuka, T. Nishinaga, M. Tachikawa, and H. Mori, Jpn. J. Appl. Phys. 34,1432(1995). 92. M. Razeghi, J. Jiang, C. Jelen, G. J. Brown, submitted for Proceeding of SPIE Photonic West 2 0 0 2 Symposium for Photodetectors Materials and Devices VII Conference, San Jose, CA (Jan 2002). 93. M. Erdtmann, GalnAs/InP quantum well infrared photodetectors on Si substratefor low-costfocalplane arrays. NU Doctoral Thesis, Evanston (2001). 94. J. Jiang, C. Jelen, M. Razeghi, and G. J. Brown, IEEE Photonic Tech. Lett. 14, 372(2002). 95. K. K. Choi, C. J. Chen, and D. C. Tsui, ]. Appl. Phys. 88, 1612 (2000). 96. C. J. Chen, K. K. Choi, L. Rokhinson, W. H. Chang, and D. C. Tsui, Appl. Phys. Lett. 75, 3210 ( 1999). 9 7. N. C. Das and K. K. Choi, IEEE Trans. Electron. Dev. 47, 653 (2000). 98. D. Z.-Y. Ting, S. V. Bandara, S. D. Gunapala, J. K. Liu, S. B. Rafol, and J. M. Mumolo, InfraPhys. and Technol. V 4 2 , 2 0 5 (2001 ).
158 Handbookof Infrared Detection Technologies 99. L. Fu, H. H. Tan, C. Jagadish, Na. Li, N. Li, X. Liu, W. Lu, and S. C. Chen, Appl. Phys. Lett. 78, 10 (200]). 100. H. C. Liu, Proc. SPIE 3 9 7 5 , 35 (20()()) 101. S. Kim, H. Mohseni, M. Erdtmann, E. Michel, C. Jelen, and M. Razeghi, Appl. Phys. Lett. 7 3 , 9 6 3 (1998) 102. S. Kim, M. Erdtmann, and M. Razeghi, J. Korean Phys. Soc. 35, 303 (1999). 103. S. Kim, M. Erdtmann, and M. Razeghi, Proc. SPIE 3 6 2 9 , 371 (1999) 104. X. O. Liu, N. Li, Z. F. Li, W. Lu, S. C. Shen, Y. Fu, M. Willander, H. H. Tan, C. Jagadish, and J. Zou, lpn. I. Appl. Ph!ls. 39, "3124 (20()()). 105. L. K. Rokhinson, C. J. Chen, D. C. Tsui, G. A. Vawter, K. K. Choi, Appl. Phys. Lett. 7 4 , 7 5 9 ( 1 9 9 9 ) . 106. M. J. Varnau and S. Yeh, Flip Chip/BGA Workshop, Binghamton, New York (October 1996). 107. T. Flynn, C. W. Argento, and J. O'Brien, Proc. SPIE 3 9 0 6 , 8 (1999). 108. A. B. Frazier and M. G. Allen, J. Microelectromech. Syst. 2, 87 ( 1993 ). 109. E. K. Yung and I. Turlik, IEEE Trans. Components. Hybrids, and Manufacturing Technology 14, 549 (1991 ). 110. W. W. Flack, S. White, B. Todd, Proc. SPIE 3 6 7 8 . 4 9 (1999). 111. J. H. Lau (Ed.), Flip Chip Technologies. New York: McGraw-Hill (1996). 112. Y.-H. Kim, J.-H. Choi, K.-S. Choi, H. C. Lee, and C.-K. Kim, Proc. SPIE 3061,60(1997). 113. J. Jiang and M. Razeghi (unpublished). 114. H. H. Martijn, U. Halldin, P. Helander, J. Alverbro, J. Y. Anderson, Proc. SPIE 3 6 9 8 , 789 (1999). 115. N. Y. Aziz, R. F. Cannata, G. T. Kincaid, R. J. Hansen, J. L. Heath, W. J. Parrish, S. M. Petronio, and J. T. Woolaway, Proc. SPIE 3 6 9 8 , 766 (1999). 116. F. Bertrand, J. L. Tissot, and G. Destefanis. Proc. SPIE 3 3 1 6 , 7 1 3 (1998). 117. R. F. Cannata, R. J. Hansen, A. N. Costello, and W. J. Parrish, Proc. SPIE 3698,756(1999). 118. N. Y. Aziz, G. T. Kincaid. W. J. Parrish, J. T. Woolaway, and J. L. Heath, Proc. SPIE 3 3 6 0 , 8 0 ( 1 9 9 8 ) . 119. H. Kulah and T. Akin, Proc. SPIE 3 6 9 8 , 778 ( 1999 ).
Chapter 5
InAs/(Gain)Sb superlattices: a
promising material system for infrared detection L. BOrkle and F. Fuchs
5.1 Introduction InAs and GaSb constitute a nearly lattice matched material system offering great flexibility in the design of optoelectronic devices. Because of the negative band overlap and similarities to the HgTe/CdTe material system, InAs/GaSb superlattices have been considered by Sakaki and Esaki as early as 1978.1 In 198 7 the material system was proposed for infrared detection devices by Smith and Mailhiot. 2 Intense work focussing on mid infrared laser devices started in the early 90s, 3 followed by first reports on detection devices with promising electrooptical properties comparable to the established Mercury-Cadmium-Telluride (MCT) material system. 4's However, for the practical use of InAs/(GaIn)Sb superlattices in infrared imaging devices, stable and reproducible growth conditions and device processing technology have to be established. Most demanding is the development of a stable and robust passivation. Regarding the latter, up to now only one report has been available."
5.2 Materials properties 5.2.1 Bandstructure of InAs/(Galn)Sb
superlattices
Because of their small difference in lattice constant, InAs and (GaIn)Sb form an ideal material system for the growth of semiconductor heterostructures. E.g., Gal_xInxSb with an indium concentration of ] 5% grows compressively strained on a GaSb substrate with a lattice mismatch of A a / a - 0.94%, while InAs is
160 Handbook of Infrared Detection Technologies under tensile strain with a lattice mismatch of A a / a = - 0 . 6 2 % . In an InAs/ (GaIn)Sb superlattice the compressively strained Gal_xInxSb layers can compensate for the tensile strain in the InAs layers. The average lattice mismatch of an InAs/(GaIn)Sb superlattice is given by ( Aa~
\aj
_ 2_
~2dsL
1
(1)
nMLaO
where ao is the lattice constant of the GaSb substrate, dsL is the superlattice period, and nML designates the number of atomic monolayers in a superlattice period. For a certain layer thickness ratio and indium mole fraction x, it is possible to achieve strain compensation on a GaSb substrate. InAs is a low band gap semiconductor with a direct energy gap of (). 35 eV, while GaSb has a direct gap of ().74eV at room temperature. 7 The relative alignment of conduction and valence bands is also of utmost importance. InAs and (GaIn)Sb establish a broken gap type II system, where the conduction band edge of InAs is lower in energy than the (GaIn)Sb valence band edge. The relative alignment of the band edges is shown in Figure 5. l(a). The band gap energy of the ternary Gal_xInxSb alloy depends on the In molar fraction x, and is well described by 8
E~(x) -
813.3 - 991.3x + 413x 2
(2)
[meV]
while the valence band offset of Gal_xlnxSb relative to the binary GaSb compound is given by
800 813.3
f
+67.4 - - - -
7
712.1
673.9
400 ...................
r(9 UJ
o
t
231.4
-T
J
T
GaSb InAs
40.3
NN~ LH
-28.8
349.5
410 -400
+29.2
Gao 8~lno1rSb
(a) unstrained
InSb
LH +31 7 ~.~-.-.-.-..---, -17.7 HH
InAs
Gao 851no~Sb
(b) strained
Figure 5.1 Band alignment between the constituents of an InAs/(Galn)Sb superlattice in: (a) unstrained condition: and (b) coherently strained to the GaSb substrate. The conduction band edge of unstrained InAs was chosen as zero energy.
InAs/(Galn)Sb superlattices: a promising material s!lstemfor infrared detection 161
AEv(x) = 234.7x-
78.7x 2
[meV]
(3)
In the case of the binary compound GaSb the band overlap with InAs is about 140 meV. 9 Because of the type II band alignment, the effective band gap of InAs/ (GaIn)Sb superlattices is smaller than the band gaps of its constituents. The band gap of an InAs/(GaIn)Sb superlattice can be continuously adjusted in a range between 0 and about O. 3 eV. -~ In particular, in the case of thick individual layers the superlattice can become semi-metallic with a vanishing band gap energy. Since electrons are mainly located in the InAs layers of the superlattice, whereas holes are confined to the (GaIn)Sb layers, optical transitions occur spatially indirectly and, thus, the optical matrix element for such transitions is relatively small. In addition, the wave function overlap strongly decreases with increasing superlattice period, which is inconvenient for detection applications in the far infrared. In order to compensate for the spatial separation of electrons and holes, the superlattices are commonly grown with ternary (GaIn)Sb layers instead of binary GaSb. By introducing indium into the binary GaSb material, these layers become compressively strained. The important influence of strain on the band alignment in the superlattice is shown in Figure 5.1 (b). Biaxial strain can be decomposed into a hydrostatic and a uniaxial component. The compressive hydrostatic strain component in the (Gain)Sb layers raises the band edge energies, while the uniaxial component leads to a splitting of heavy and light hole states, such that the heavy hole state becomes the valence band ground state. On the other hand, the tensile strain in the InAs layers leads to a similar effect for InAs but with opposite signs, thus leading to an overall reduction of the superlattice band gap. The band structure of an InAs/(GaIn)Sb superlattice is shown schematically in Figure 5.2. The band gap of the superlattice is determined by the energy difference between the electron miniband E1 and the first heavy hole state HH 1 at the Brillouin zone centre and can be varied continuously in a range between 0 and about 0.3 eV by adjusting the individual layer thicknesses and the indium molar fraction in the ternary (GaIn)Sb layers. 2 The dispersion of the heavy hole states exhibits a strong anisotropy, while the other subbands are nearly isotropic. The electronic structure of InAs/(GaInSb) superlattices can be calculated with satisfactory accuracy by using a three-band envelope-function approximation (EFA) formalism. 1~ Table 5.1 shows the results of such a calculation for a superlattice comprising 13 monolayers InAs and 8 monolayers Ga~.~sln~.~sSb. The superlattice is coherently strained to the lattice constant of the GaSb substrate and shows a residual net strain of Aa/a = - 0 . 8 9 • 1()-~. The zero of the energy scale is set to the value of the unstrained InAs conduction band minimum. Values in parentheses show the deviation to the results of a full EFA 8 • 8k.p EFA calculation. ~ For the electron miniband E~ and the first heavyhole band HH1, which determine the superlattice band gap, the values of the three-band EFA calculation are only some millielectronvolts larger than those of the 8 • 8 band calculation. Hence, the deviation for the fundamental band gap is
162
Handbook of Infrared Detection Technolo#ies
Figure 5.2 Band structure of a superlattice consisting of 1 ~ monola!lers InAs and 8 monolayers Gao.~ slno. 1sSb.
Table 5.1 R e s u l t s o f a t h r e e - b a n d EFA c a l c u l a t i o n for a s u p e r l a t t i c e c o m p r i s i n g 15 m o n o layers I n A s ($.995 n m layer t h i c k n e s s ) a n d 8 m o n o l a y e r s Gao.sslno.lsSb layers ( 2 . 4 6 2 n m layer t h i c k n e s s ) . The s u p e r l a t t i c e is c o h e r e n t l y s t r a i n e d to a GaSb s u b s t r a t e a n d s h o w s a r e s i d u a l n e t s t r a i n o f A a / a = - O . 8 9 x l O --3. T h e zero o f Energy is c h o s e n at t h e c o n d u c t i o n b a n d m i n i m u m o f u n s t r a i n e d InAs. Values in p a r e n t h e s e s s h o w t h e d e v i a t i o n to t h e r e s u l t s o f a full 8 xSk.p EFA c a l c u l a t i o n ll Band
EL HH1 LH1 HH2
Zone center q = () [meV] 242.3 9"3.6 -67.2 -205.3
{+3t (+5) (+42) (+18)
Zone edge q: = 7l'/ds-l. [meV] 386.1 9-5.6 -174.6 -203.7
1+7t {+51 I+36) i+17)
Miniband width AE [meV] 143.8 ().(14 1()7.4 1.6
1+41 1-().()1 ) (-16) t-l)
Effective mass m-, 0.026 -96.5 -().()32 2.4
(-().(1(12 t (+16.71 [ +().()()2) (+().61)
only 2 meV, which corresponds to a relative error of about 1%. However, for higher subbands the deviation from the results of a full 8 • 8 band calculation drastically increases. For superlattices with a band gap energy in the mid and far infrared, the individual layer thicknesses are typically in a range between :3 and 10 monolayers, leading to a strong coupling between electron wave functions of adjacent potential wells. An example for the distribution of the electron and hole wave functions is shown in Figure "3.3 for a superlattice consisting of 13 monolayers InAs and 8 monolayers Ga~.ssIn~.l sSb. The probability density of the electrons is concentrated in the InAs layers. However, due to the strong coupling in the electronic system there is a significant delocalization of the electron wave function over the superlattice. Typical values of the electron miniband width easily exceed 1 ()()meV. On the other hand, the heavy holes are strongly localized in the (GaIn)Sb layers. Thus. as a consequence of the type II
InAs/(Galn)Sb superlattices: a promising material S!lstem for infrared detection 9
I
'
9
InAs
0.6
I
9
(Galn)Sb
I
9
I
InAs
9
!
16 }
" i
(Galn)S
r ,%.,,,,. ..___.,
0.4
o,I
c-
O
13 r ::3
/
0.2
\
0.0
i/ i
9
| |
!
2
f~-~"-
L
6
I
8
,
I
10
112
Position z [nm] Figure 5. ] Probability densit!t distribution o.f the first electron miniband El and the first heavy hole state HH1 in a 13 ML InAs/8 ML Gao.~cslno. 1~Sb superlattice.
band alignment of the material system, electrons and holes are spatially separated. This is particularly disadvantageous for optical absorption, where a significant overlap of electron and hole wave functions is needed. The delocalization of electrons and, consequently, the optical absorption in an InAs/ (GaIn)Sb superlattice is promoted by a reduction in the electronic confinement. This can be achieved either by growing thinner (Galn)Sb barriers or by introducing more indium into the (GaIn)Sb layers, causing a reduction in the energy of the barrier material. ~(~ On the other hand, the effective masses in an InAs/(GaIn)Sb superlattice are not directly dependent on the band gap energy, as it is the case in a bulk semiconductor, j2 Compared to the electron effective mass of HgCdTe, the electron effective mass (m*,~O.()2 in an InAs/(GaIn)Sb superlattice is larger, leading to a reduction in tunnelling contributions in the leakage currents of an IR photodiode. Also related to the larger effective mass in the superlattice is a higher combined density of states, which in turn compensates for the smaller optical matrix element of the superlattice. Thus, despite the type II character of the optical transitions in an InAs/(Galn)Sb superlattice, the optical absorption is comparable to that of HgCdTe. In addition, larger minority carrier lifetimes are expected because of smaller Auger recombination rates compared to MCT. 13
5.2.2 X-ray characterization High-Resolution X-ray diffraction is an important tool for the structural characterization of the MBE grown superlattices. In Figure 5.4 the X-ray diffraction pattern of an InAs/(GaIn )Sb superlattice, close to the (004) diffraction order of the GaSb substrate, is shown. The superlattice peaks can be observed up to the seventh order, indicating an excellent materials quality. From the angular
1 6 4
Handbook of Infrared Detection Technologies
t-
'
,
'
w
,
,
,
lOSl
lo o
10 [_
.~
1(~
,
\
2
lO4
8
,
,106
SL-O GaSh
9
~o~[
)1
SL-%AGasbI1 w
9
10 2
i
I I
lo'
I
j/ I
5
30.3 30.4
+5 .
I . 24
.
/
26
.
.I 28
,
10 3
10~ 10~ . 22
;
.
I
30
Incidence Angle
,
I
32
34
,
,
36
j
+6 -.., .~ 38
o~ [~
Figure 5.4 High-resolution X-raft diffraction pattern of an InAs/((;ahl)Sb superlattice close to the (004) reflection of the GaSh substrate. The superlattice has a period of 4.56 nm and a residual lattice mismatch of + 0 . 8 8 x 1 0 -~
spacing Am of adjacent peaks the superlattice period dsI. can directly be calculated using 2
dsL = 2 Ao)cos|
(4)
where 2 is the wavelength of the x-ray beam and O designates the average diffraction angle between the peaks considered for evaluation. The net strain in the superlattice