LARGE FORMAT DUAL-BAND QUANTUM WELL ...etd.lib.metu.edu.tr/upload/3/12610936/index.pdfApart from...

91

Transcript of LARGE FORMAT DUAL-BAND QUANTUM WELL ...etd.lib.metu.edu.tr/upload/3/12610936/index.pdfApart from...

  • 1

  • LARGE FORMAT DUAL-BAND QUANTUM WELL INFRAREDPHOTODETECTOR FOCAL PLANE ARRAYS

    A THESIS SUBMITTED TOTHE GRADUATE SCHOOL OF NATURAL AND APPLIED SCIENCES

    OFMIDDLE EAST TECHNICAL UNIVERSITY

    BY

    YETKİN ARSLAN

    IN PARTIAL FULFILLMENT OF THE REQUIREMENTSFOR

    THE DEGREE OF MASTER OF SCIENCEIN

    ELECTRICAL AND ELECTRONICS ENGINEERING

    SEPTEMBER 2009

  • Approval of the thesis:

    LARGE FORMAT DUAL-BAND QUANTUM WELL INFRARED

    PHOTODETECTOR FOCAL PLANE ARRAYS

    submitted by YETKİN ARSLAN in partial fulfillment of the requirements for thedegree of Master of Science in Electrical and Electronics Engineering Department,Middle East Technical University by,

    Prof. Dr. Canan ÖzgenDean, Graduate School of Natural and Applied Sciences

    Prof. Dr. İsmet ErkmenHead of Department, Electrical and Electronics Engineering

    Prof. Dr. Cengiz BeşikçiSupervisor, Electrical and Electronics Engineering Dept., METU

    Examining Committee Members:

    Prof. Dr. Tayfun AkınElectrical and Electronics Engineering Dept., METU

    Prof. Dr. Cengiz BeşikçiElectrical and Electronics Engineering Dept., METU

    Prof. Dr. Nevzat G. GençerElectrical and Electronics Engineering Dept., METU

    Asst. Prof. Dr. Haluk KülahElectrical and Electronics Engineering Dept., METU

    Prof. Dr. Raşit TuranPhysics Dept., METU

    Date: 11.09.2009

  • I hereby declare that all information in this document has been obtained andpresented in accordance with academic rules and ethical conduct. I also declarethat, as required by these rules and conduct, I have fully cited and referenced allmaterial and results that are not original to this work.

    Name, Last Name: YETKİN ARSLAN

    Signature :

    iii

  • ABSTRACT

    LARGE FORMAT DUAL-BAND QUANTUM WELL INFRAREDPHOTODETECTOR FOCAL PLANE ARRAYS

    Arslan, YetkinM.S., Department of Electrical and Electronics Engineering

    Supervisor : Prof. Dr. Cengiz Beşikçi

    September 2009, 76 pages

    Quantum Well Infrared Photodetectors (QWIPs) are strong competitors to other de-

    tector technologies for future third generation thermal imagers. QWIPs have inherent

    advantages of mature III-V material system and well settled fabrication technology,

    as well as narrow band photo-response which is an important property facilitating the

    development of dual-band imagers with low crosstalk. This thesis focuses on the de-

    velopment of long/mid wavelength dual band QWIP focal plane arrays (FPAs) based

    on the AlGaAs/GaAs material system.

    Apart from traditional single band QWIPs, the dual-band operation is achieved by

    proper design of a bias tunable quantum well structure which has two responsiv-

    ity peaks at 4.8 and 8.4 µm for midwave infrared (MWIR) and longwave infrared

    (LWIR) atmospheric windows, respectively. The fabricated large format (640x512)

    FPA has MWIR and LWIR cut-off wavelengths of 5.1 and 8.9 µm, and it provides

    noise equivalent temperature differences (NETDs) of ∼ 20 and 32 mK (f/1.5 at 65 K)in these bands, respectively.

    The employed bias tuning approach for the dual-band operation requires the same

    iv

  • fabrication steps established for single band QWIP FPAs, which is an important ad-

    vantage of the selected method resulting in high-yield, high-uniformity and low-cost.

    Results are encouraging for fabrication of low cost, large format, and high perfor-

    mance dual band FPAs, making QWIP a stronger candidate in the competition for

    third generation thermal imagers.

    Keywords: Quantum-well infrared photodetectors, MWIR / LWIR, Dual-Band ther-

    mal imaging,

    v

  • ÖZ

    GENİŞ FORMATLI ÇİFT BANTLI KUANTUM KUYULU KIZILÖTESİ ODAKDÜZLEM DİZİNLERİ

    Arslan, YetkinYüksek Lisans, Elektrik Elektronik Mühendislig̃i Bölümü

    Tez Yöneticisi : Prof. Dr. Cengiz Beşikçi

    Eylül 2009, 76 sayfa

    Kuantum Kuyulu kızılötesi fotodedektörler (KKKF) gelecek üçüncü nesil termal gö-

    rüntüleme sistemlerinde yer almak için dig̃er dedektör teknolojileri arasında kuvvetli

    adaylardandır. AlGaAs/GaAs KKKFler, olgunlaşmış bir malzeme sistemiyle üretilme,

    gelişmiş ve oturmuş üretim teknolojisine sahip olmak gibi avantajlarının yanında çift

    bantlı çalışan dedektörler için önemli bir nokta olan dar bantlı tepki özellig̃ine de

    sahiptir. Bu tez çalışması AlGaAs/GaAs malzeme sistemi üzerinde, uzun ve orta

    dalgaboyu kızılötesi bantlarında çalışan çift bantlı KKKF odak düzlem dizinlerinin

    (ODD) geliştirilmesine odaklanmıştır.

    Tek bantlı KKKFlerden farklı olarak, çift bantlı çalışma, orta dalgaboyu kızılötesi

    (ODK) atmosferik penceresinde 4.8 µm, uzun dalgaboyu kızılötesi (UDK) atmos-

    ferik penceresinde ise 8.4 µm tepe duyarlılık dalgaboyu göstermek üzere özel tasar-

    lanan kuantum kuyularının eg̃imleme gerilimlerinin anahtarlanması prensibiyle elde

    edilmiştir. Üretilen geniş formatlı (640x512) ODD, ODK ve UDK pencerelerinde

    sırasıyla 5.1 µm ve 8.9 µm üst kesim dalgaboyuna sahip olup, bu bantlarda sırasıyla

    20 mK ve 32 mK (f/1.5, 65K) gürültü eşdeg̃er sıcaklık farkı göstermiştir.

    vi

  • Çift bantlı çalışma için uygulanan eg̃ilim gerilimi anahtarlanması yaklaşımı stan-

    dart tek bantlı KKKF ODD ile aynı üretim sürecine tabi olup, bu sayede yüksek

    üretim verimi, yüksek düzgünlük ve düşük maliyet avantajlarına sahiptir. Elde edilen

    sonuçlar düşük maliyetli, geniş formatlı ve yüksek performanslı çift bantlı KKKF

    ODD üretimi açısından ümit verici olup, KKKFleri dig̃er teknolojilere kıyasla üçüncü

    nesil görüntüleme sistemlerinde bir adım öne çıkarmaktadır.

    Anahtar Kelimeler: Kuantum kuyulu kızılötesi fotodedektörler, MWIR / LWIR, Çift

    bantlı termal görüntüleme

    vii

  • To my family

    viii

  • ACKNOWLEDGMENTS

    First, I would like to thank my advisor Prof. Dr. Cengiz Beşikci for his enlightening

    guidance and providing me the possibility to work in such a sophisticated facility.

    I would like to thank Prof.Dr. Raşit Turan, Prof. Dr. Nevzat Gençer and Asst. Prof.

    Dr. Haluk Külah for being on my thesis committee.

    I would like to thank Prof. Dr. Tayfun Akın for being on my thesis committee and

    sharing his laboratories with us.

    I would like to thank Dr. Oray Orkun Cellek , Dr. Selçuk Özer and Mr. Ümid

    Tümkaya and Mr. Burak Aşıcı for sharing with me their knowledge, their invaluable

    comments and opinions as well as their friendship.

    I would like to express my gratitude to Mr. Süleyman Umut Eker and Mr. Melih

    Kaldırım for all the talks, all the jokes and all the sleepless growth nights. Your

    invaluable friendly support kept me going, thank you for every moment that we have

    spent throughout this research.

    I would like to thank to Mr. Emre Onuk, Mr. Alp Tolungüç, Mr. and Mrs. Ag̃aog̃lu

    for their friendships and help in all the laboratory.

    I would like to thank to Mr. Özgür Şen for his unbelievable efforts to keep the labo-

    ratory working.

    Last but the best, I want to express my deep love and gratitude to my family and

    Irmak, for being always with me.

    ix

  • TABLE OF CONTENTS

    ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv

    ÖZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

    DEDICATON . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii

    ACKNOWLEDGMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

    TABLE OF CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x

    LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii

    LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii

    CHAPTERS

    1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

    1.1 Infrared Radiation and Thermal Imaging . . . . . . . . . . . 2

    1.2 Infrared Detectors . . . . . . . . . . . . . . . . . . . . . . . 7

    1.3 Figure of Merits of Infrared Detectors and Imaging Systems . 10

    1.4 Scope and Objective of This Work . . . . . . . . . . . . . . 14

    2 Dual-Band Detection with QWIPs and Design Approach . . . . . . . 16

    2.1 Advantages of Dual/Multi-Band Detection . . . . . . . . . . 16

    2.2 State-of-The-Art Dual-Band Detectors . . . . . . . . . . . . 18

    2.3 Advantages of QWIPs for Dual/Multi-Band Detection . . . . 22

    2.4 QWIP Basics . . . . . . . . . . . . . . . . . . . . . . . . . . 23

    2.4.1 Theory of Operation . . . . . . . . . . . . . . . . 23

    2.4.2 Material Systems for QWIPs . . . . . . . . . . . . 30

    2.5 Design Approach . . . . . . . . . . . . . . . . . . . . . . . 34

    2.5.1 Selection of Dual-Band Operation Type . . . . . . 34

    2.5.2 Operation of Voltage Tunable Detector Structure . 40

    x

  • 2.5.3 Design Targets . . . . . . . . . . . . . . . . . . . 42

    2.5.4 Design . . . . . . . . . . . . . . . . . . . . . . . . 44

    3 Implementation and Characterization of GaAs Based MWIR/LWIRDual-Band QWIP FPAs . . . . . . . . . . . . . . . . . . . . . . . . . 47

    3.1 Final Design . . . . . . . . . . . . . . . . . . . . . . . . . . 47

    3.2 Growth of the Device Epilayer Structure . . . . . . . . . . . 48

    3.2.1 Molecular Beam Epitaxy Basics . . . . . . . . . . 48

    3.2.2 Optimization of Growth Parameters and DopingCalibration . . . . . . . . . . . . . . . . . . . . . 50

    3.2.3 X-ray Diffraction Characterization . . . . . . . . . 51

    3.3 FPA Fabrication . . . . . . . . . . . . . . . . . . . . . . . . 53

    3.4 Pixel and FPA Level Characterization and Discussions . . . . 58

    3.4.1 Pixel Level Characterization . . . . . . . . . . . . 58

    3.4.2 FPA Characterization . . . . . . . . . . . . . . . . 63

    4 CONCLUSION AND FURTHER WORK . . . . . . . . . . . . . . . 71

    REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

    xi

  • LIST OF TABLES

    TABLES

    Table 2.1 Mature III-V material systems used for QWIP fabrication. . . . . . . 31

    Table 2.2 Specifications of the ISC9803 ROIC. . . . . . . . . . . . . . . . . . 43

    Table 3.1 Comparison of the dual-band focal plane arrays. . . . . . . . . . . . 70

    xii

  • LIST OF FIGURES

    FIGURES

    Figure 1.1 Spectral photon exitance of blackbody at various temperatures . . 4

    Figure 1.2 Atmospheric transmission . . . . . . . . . . . . . . . . . . . . . . 5

    Figure 1.3 Band structure of type II strained superlattices . . . . . . . . . . . 10

    Figure 2.1 Pixel architecture for TLHJ type dual-band detectors . . . . . . . . 19

    Figure 2.2 Pixel architecture with multiple electrical contact. . . . . . . . . . 20

    Figure 2.3 Multiple contact pixel architecture by CEA/LETI. . . . . . . . . . 20

    Figure 2.4 Layer diagram of the four-band QWIP device structure. . . . . . . 21

    Figure 2.5 An image recorded with the four-band detector. . . . . . . . . . . . 22

    Figure 2.6 Typical QWIP diagram. . . . . . . . . . . . . . . . . . . . . . . . 24

    Figure 2.7 Typical band diagram of the QWIP. . . . . . . . . . . . . . . . . . 24

    Figure 2.8 Al mole fraction dependence of energy states AlGaAs/GaAs . . . 25

    Figure 2.9 Typical spectral responses of different QWIP types. . . . . . . . . 26

    Figure 2.10 Illustration of the effect of diffraction grating. . . . . . . . . . . . . 27

    Figure 2.11 Corrugated pixel structure . . . . . . . . . . . . . . . . . . . . . . 27

    Figure 2.12 Illustration of the capture and emission mechanisms in QWIPs . . 28

    Figure 2.13 Temperature dependence of the detectivity of the QWIPs. . . . . . 30

    Figure 2.14 Dependence of the NETD on detector bias . . . . . . . . . . . . . 32

    Figure 2.15 Thermal image recorded with the 640x512 InP/InGaAs FPA. . . . . 33

    Figure 2.16 Responsivity spectrum of AlInAs/InGaAs/InP QWIPs. . . . . . . . 34

    Figure 2.17 Multi-contact pixel schemes for co-registered dual-band operation . 35

    Figure 2.18 Illustration of space sharing pixel placement . . . . . . . . . . . . 36

    Figure 2.19 Dual-band detector structure employing one contact per pixel. . . . 37

    xiii

  • Figure 2.20 Band diagram of a voltage-tunable two-color detector . . . . . . . 37

    Figure 2.21 Responsivity spectra of a MWIR/LWIR detector . . . . . . . . . . 38

    Figure 2.22 Voltage tunable MWIR/MWIR spectral response . . . . . . . . . . 39

    Figure 2.23 Characteristic I-V curve of standard AlGaAs/GaAs QWIP. . . . . . 41

    Figure 2.24 Operation points of the series connected QWIP stacks. . . . . . . . 41

    Figure 2.25 Simplified unit cell schematic of the ISC9803. . . . . . . . . . . . 44

    Figure 2.26 Generic layer structure for voltage-tunable QWIP structure. . . . . 45

    Figure 3.1 Final dual-band QWIP epilayer structure. . . . . . . . . . . . . . . 48

    Figure 3.2 Riber Epineat III-V MBE Reactor utilized in this work. . . . . . . . 49

    Figure 3.3 Rocking curve of the detector epilayer structure. . . . . . . . . . . 52

    Figure 3.4 Illustration of fabrication steps of detector hybrid . . . . . . . . . . 54

    Figure 3.5 FC-150 flip-chip aligner/bonder. . . . . . . . . . . . . . . . . . . . 56

    Figure 3.6 Transmission of the wide-band AR coating. . . . . . . . . . . . . . 57

    Figure 3.7 Pictures of the FPA after polishing and AR coating. . . . . . . . . . 57

    Figure 3.8 Dark current and photo current characteristic of a single pixel. . . . 59

    Figure 3.9 Responsivity and detectivity measurement setup diagram. . . . . . 60

    Figure 3.10 Absolute spectral responsivity of test pixels. . . . . . . . . . . . . 61

    Figure 3.11 Bias sharing between dual-band detector stacks. . . . . . . . . . . 63

    Figure 3.12 NETD histogram of the dual-band QWIP in LWIR mode. . . . . . 64

    Figure 3.13 Outdoor images taken with the MWIR/LWIR dual-band sensor. . . 66

    Figure 3.14 Thermal images of a man holding a 8-12 µm band-pass filter. . . . 67

    Figure 3.15 Outdoor images in two bands and image fusion. . . . . . . . . . . 68

    xiv

  • CHAPTER 1

    Introduction

    Infrared (IR) radiation was unknown to mankind until Sir Frederick William Her-

    schel’s experiments with thermometers and a simple glass prism, 200 years ago. What

    he called ”calorific rays” at that time kept being an important field of interest to re-

    searchers all around the world.

    Today’s infrared detectors started their evolution in the form of thermal detectors in

    1820’s by August Ludwig Friedrich Wilhelm Seebeck with the main purpose being

    simple temperature measurement. Advancements like construction of thermopiles by

    Leopoldo Nobili followed, and Samuel Pierpont Langley’s bolometer employing a

    Wheatstone bridge was able to detect the radiation emitted by a cow from quarter

    miles away in late 1890’s [1].

    Even though photoconductivity of Selenium was well known since 1897 after W.Smith,

    first synthesized photon detectors have their origin in the work of T.W. Case where

    he made a systematic search for ”light-active substances” in 1917 [2, 3]. Some of to-

    day’s major infrared materials’ appearances are InSb in mid 1950’s, HgCdTe in early

    1960’s and AlGaAs/GaAs quantum well infrared detectors in late 1980’s.

    Its development well driven by military applications in years, infrared detection tech-

    nology has evolved from first generation single element (or 1-D arrays) scanning

    configurations to second generation 2-D detector arrays in late 1970’s owing to the

    establishment of reproducible bulk growth techniques. Recently, in mid 1990’s, with

    improved fabrication techniques, focal plane arrays (FPAs) appeared with improved

    capabilities like very large number of pixels, decreased pixel pitch and multi-band

    spectral operation. Although its definition is not well established, these talented de-

    1

  • tectors are accepted to constitute the third generation of infrared systems. [4]

    Latest efforts in the development of third generation infrared systems are toward

    higher resolution, higher sensitivity and multi-band operation at lower cost. At the

    low cost end of performance-cost dilemma, quantum well infrared photodetectors

    (QWIPs) based on the fairly mature GaAs material system display big potential.

    QWIPs are also inherently advantageous as a multi-band detector alternative as they

    have narrow band spectral response by their nature. Mainly influenced by these two

    facts, in this thesis work, longwavelength / midwavelength dual-band QWIPs employ-

    ing bias tuning approach are designed, implemented, and their FPA level performance

    is assessed after fabrication of very large format (640x512) FPAs.

    This chapter focuses on the fundamentals of infrared radiation and infrared detectors

    including the present status of detector technology. The second chapter presents the

    design criteria and approach adopted in this work, as well as the basics of QWIPs.

    The third chapter is devoted to demonstration of the dual-band FPA, characterization,

    results and discussion. Finally, fourth chapter includes the conclusions that can be

    drawn from this work and further work.

    1.1 Infrared Radiation and Thermal Imaging

    Infrared radiation is an electromagnetic radiation, and its location in the electromag-

    netic spectrum starts from high side of the visible part at 700-800 nm and ends at the

    low side of terahertz radiation which is at 100 µm. Spanning roughly two orders of

    magnitude in wavelength, subdivisions and their nomenclatures are varying for dif-

    ferent application fields. Commonly used names in infrared sensor related studies are

    as follows [5]

    • Near infrared (NIR): 0.7 to 1.0 µm.• Short-wave infrared (SWIR): 1.0 to 3.0 µm.• Mid-wave infrared (MWIR): 3 to 5 µm.• Long-wave Infrared (LWIR): 8 to 12, or 7 to 14 µm.• Very-long wave infrared (VLWIR): 12 to 30 µm.• Far infrared (FIR) : 30 to 100 µm.

    2

  • The above division is mainly based on the atmospheric transmission related issues

    which will be discussed later.

    As an electromagnetic wave, infrared radiation obeys all Maxwell’s equations. From

    a practical point of view, infrared radiation is subject to some extra rules dictated by

    its interaction with matter, its sources and environment. Unlike the human vision,

    NIR and partly SWIR systems, which are concerned with the reflected light from

    objects, passive infrared (thermal) imaging senses the radiation emitted by the object

    itself. This emitted radiation by each object with a temperature above absolute zero

    is described by Planck’s law of radiation. Planck’s law relates the spectral exitance

    of a blackbody to its temperature where a blackbody is defined as an idealized object

    which absorbs all the radiation falling on it without any reflection or transmission.

    Planck’s law is given as

    Mp(λ,T ) =2πhc2

    λ51

    ehc/λkT − 1 (1.1)

    where Mp(λ,T ) is the spectral radiant exitance in Watt/cm2µm, c is speed of light, h is

    the Planck constant, k is the Boltzmann constant, and T is the object temperature. It is

    also possible to express spectral exitance in terms of number of photons/cm2 emitted

    in unit time per wavelength interval. Spectral photon exitance is plotted in Fig. 1.1

    for various blackbody temperatures.

    There are several important observations that can be extracted by investigating the

    spectral photon exitance distribution. The first observation is that most of the energy

    radiated from near room temperature objects is in the IR region with a peak around

    10 µm. Dependence of the peak wavelength of the blackbody radiation to object

    temperature is explained by Wien’s displacement law as follows

    λmax =2897.7µmK

    T. (1.2)

    Another observation is that the change in the exitance with varying wavelength is

    higher in the MWIR (3-5 µm) region than that in LWIR (8-12 µm) region, resulting

    in a higher imaging contrast in the MWIR band. Even though there is less power in

    this window, higher contrast may be useful in some cases.

    3

  • Figure 1.1: Spectral photon exitance of blackbody at various temperatures

    In reality, the blackbody assumption of Planck’s law is seldom valid. Instead, real-

    life objects exhibit reflection and transmission properties resulting in deviations from

    the blackbody assumption. In order to characterize this behavior, emissivity (ε) of

    a material is defined to be the ratio of the total energy emitted by a material to the

    total energy radiated from a perfect blackbody at the same temperature. Emissivity of

    a material shows wavelength and view angle dependence. However, for engineering

    purposes it is common to consider the emissivity of a surface to be wavelength and

    angle independent at least for a certain wavelength range. This is known as greybody

    assumption.

    After the infrared radiation is emitted by matter, it is also important to examine its

    propagation through the transmission medium which is generally air. Infrared ra-

    diation is subject to different events that could affect its propagation. These events

    include the absorption by atmospheric gases, scattering by aerosols and reflections.

    Absorption of the atmospheric gases is the most significant mechanism, and it is

    greatly varying with atmospheric conditions. Absorbing gases include H2O, CO2,

    CH4, N2O, O3, NH3, etc. For low attitude horizontal paths, H2O and CO2 are the

    dominant absorbing gases. Sample atmospheric transmission is given in Figure 1.2.

    4

  • For passive thermal imaging systems there are two important spectral regions, MWIR

    and LWIR where the atmosphere is transparent. However, H2O absorption isolating

    these two windows is strongly dependent on the water vapor content of the air which

    is very different from place to place and time to time.

    Figure 1.2: Atmospheric transmission and absorbing molecules, measured at sea leveland through 1800 m horizontal path [6].

    Scattering is the next important factor in atmospheric propagation. It is due to in-

    teraction of photons with atmospheric molecules and aerial particles like fog, dust

    and droplets. Rate of scattering of photons by particles smaller than wavelength is

    inversely proportional to wavelength. However scattering rate becomes wavelength

    independent as particle size becomes much larger than radiation wavelength. Con-

    sequently, infrared imaging becomes more advantageous over visible imaging when

    only the scattering mechanisms are considered.

    Taking these mechanisms into account, a more sophisticated model can be constructed.

    In this model the amount of radiation falling on a detector due to a distant emitting

    target is expressed as

    I = F(r) ·∫

    λ

    [τa(λ, r){εtM(λ, Tt) + ρlM(λ,Tae)} + [1 − τa(λ, r)]M(λ, Ta)] dλ (1.3)

    In the above equation,

    • F(r) is a geometric factor depending on target’s projected area and distance,• τa is the atmospheric transmission factor as a function of wavelength and target

    distance,

    5

  • • εt is the emissivity of the target,

    • M(λ, Tt) is the spectral exitance,

    • ρl is the reflectivity of the target,

    • M(λ, Tae) is the spectral exitance of background,

    • 1 − τa is the emissivity of the atmosphere (emissivity+transmissivity=1 whenreflection is ignored),

    • M(λ, Ta) is the temperature dependent exitance of the atmosphere,

    • Ta is the temperature of air,

    • Tt is the temperature of target,

    • Tae is the temperature of background.

    This model also accounts for the reflections of background from target and the emis-

    sion of the atmosphere itself. Such artifacts should be considered when designing an

    imaging system.

    Thermal imaging systems basically consist of an input optics stage, detector stage,

    and electronics. Optic stage is responsible from construction of the image on the

    detector, and it is specially designed to work in a selected spectral window. Passive

    imaging systems usually cover 3-5 µm, 8-12 µm or directly 3-12 µm windows. With

    the help of cold shield, optical stage determines the optical power falling on the unit

    area of the detector or field of view of the system. The detector converts incoming

    radiation power into electrical signal. Detectors operate at near room temperature or

    at cryogenic temperatures. In both cases, detectors are vacuum sealed for maximum

    isolation from environmental conditions. In the hybrid technology, the detector is

    coupled to a readout integrated circuit (ROIC). There also exist monolithic technolo-

    gies where the readout circuitry and the detectors are fabricated on the same chip.

    ROIC is responsible from driving the detector array, reading the output from each

    detector and multiplexing output to the first stage of electronics. Typically, external

    electronics exist in two stages in an imaging system. The front stage is usually located

    near the detector-ROIC hybrid. This stage drives the ROIC and preamplifies output

    to the back stage. Back stage electronics constructs the image and streams the data to

    the user in desired protocol and form.

    6

  • Operation principles for the various types of detectors, figures of merit and present

    detector technologies are investigated in further detail in the next section.

    1.2 Infrared Detectors

    It is possible to build an infrared detector by exploiting many physical phenomena.

    Some of these are thermoelectric effect (thermocouples), change in the electrical con-

    ductivity with temperature (bolometers), gas expansion (Golay cell), pyroelectricity,

    intrinsic absorption, impurity absorption, and intersubband transitions. In some of

    these effects, incoming photons are absorbed into heat, and the resulting temperature

    change is detected by monitoring some physical properties of the detector material.

    Detectors based on this principle are called thermal detectors. On the other side, in

    some of the listed phenomena, photons are interacting with electrons causing some

    change in the electronic state of the material, and an electrical signal is output as the

    indicator of the infrared radiation. These types of detectors are called photon detec-

    tors.

    Thermal detectors are wide band detectors in the sense that detector response is given

    to all wavelengths. Operating region of thermal detectors is dictated by optics. Ther-

    mal detectors do not require cryogenic cooling, but thermoelectric coolers are used

    for temperature stabilization. Currently microbolometers are the detectors of choice

    for applications fitting into specifications of thermal detectors. Microbolometers’s

    theory of operation is based on monitoring of the thermally induced change in the

    electrical resistance of the active material. Advantages of microbolometers over pho-

    ton detectors covering similar spectral range can be listed as follows

    • Small and lightweight imager structure

    • Low power consumption

    • Immediate video output after power on

    • Very long mean time between failure (MTBF)

    • Low cost.

    Main disadvantages of thermal detectors can be given as follows

    7

  • • Low sensitivity

    • High speed imaging is not possible due to large response time

    • Dual/multi-band operation can not be implemented at detector level

    • Difficulties in decreasing the pixel size for very high resolution.

    Performance limits for microbolometers arise from structural properties of the pixels

    and active material’s physical properties. Roughly, sensitivity of a microbolometer is

    limited by thermal conductance, and the response time is dictated by the heat capacity

    over thermal conductance ratio for a single pixel [7].

    In photon detectors, not like thermal detector where photons are absorbed into heat,

    absorption of photons occurs through excitation of electrons from one energy state to

    another one. Cryogenic cooling requirement of photon detectors arises from the fact

    that, thermal excitations must be suppressed for photo-excitations to stay dominant.

    Excited electrons are collected with the help of an electric field to create detector

    signal. Photon detectors are further divided into two subgroups according to the gen-

    eration of this electric field. In the first group, called photovoltaic detectors, built in

    electric field, as in a p-n junction, is employed to collect excited charge carriers. In

    the second group, an external bias potential is applied to collect excited carriers, and

    this type of photon detectors are called photoconductive detectors.

    Excitations of charge carriers in a photon detector can be between bands, subbands or

    minibands. Today, most commonly used technologies for fabrication of infrared pho-

    ton detectors utilize low bandgap materials like HgCdTe and InSb or QWIP structures

    of the AlGaAs/GaAs material system. Type II strained superlattice of InAs/InGaSb

    material system is also showing great advancement. These technologies are briefly

    discussed below.

    InSb is a binary compound which has band to band transition energy corresponding

    to a wavelength in the MWIR region. InSb is a mature technology, and InSb pho-

    todiodes have been available since late 1950’s [8]. Main drawback of the InSb is

    impossibility of wavelength tuning since it is a binary compound. InSb is preferred

    especially in astronomy and megapixel size FPA applications. High quality InSb

    FPAs are demonstrated and commercialized by several manufacturers [9, 10].

    8

  • HgxCd1−xTe is a very important infrared material and has been studied extensively

    over many years. Since it is a ternary compound, its bandgap can be tailored to cover

    the 1-20 µm spectral range. Its high quantum efficiency together with the possibility

    of tuning responsivity spectrum make HgCdTe currently the first choice of technol-

    ogy where performance is the primary requirement. Main problems with the HgCdTe

    technology are production cost, unavailability of high quality lattice matched sub-

    strates, nonuniformity (especially in LWIR band) and difficulties in epitaxial growth.

    In order to overcome these closely correlated difficulties, there is significant amount

    of research on high quality growth of HgCdTe on alternative substrates like Si, Ge

    and GaAs. Currently HgCdTe arrays with megapixel (1280x720) size and dual-band

    operation (MWIR/LWIR) ability have been demonstrated [11].

    Type II strained superlattice technology (SLS) progressed very quickly in the last cou-

    ple of years and nowadays offers the high quantum efficiency of low bandgap detec-

    tors combined with mature fabrication methods of III-V compound semiconductors.

    Type II SLSs consist of stacked InAs/(In)GaSb layers. In this structure, bottom of the

    conduction band of InAs goes below of the top of the valence band of (In)GaSb as in

    Figure 1.3. In this broken gap structure, minibands are formed due to overlapping of

    wave functions. Spacing of these bands are controlled with thickness of layers over a

    wide range. The fact that SLS is the only material that can theoretically outperform

    HgCdTe attracted researchers, and breakthrough advancements have been made in

    this material over the past few years. Main obstacles in front of the performance of

    the SLSs were surface passivation for LWIR detectors, low quantum efficiency, and

    high leakage. With the recent advancements, passivation quality is greatly increased

    with SiO2 coating, and good quality, high quantum efficiency layers are demonstrated

    [12]. Currently MWIR/MWIR dual color mid-format SLS FPAs are in production

    for European military aircrafts by AIM in Germany. SLSs are also considered as a

    future solution in VLWIR imaging area where HgCdTe has very serious uniformity

    problems.

    Quantum well infrared photodetectors constitute an important share of the current

    detector market. QWIP’s theory of operation bases on the intersubband excitations

    of electrons by photons. These subbands are created by 1-D confinement in the con-

    duction band of alternating AlGaAs/GaAs layers. Relying on mature GaAs material

    9

  • system, QWIPs are relatively low cost photon detectors with high yield, good uni-

    formity and stability. Main drawback of the QWIPs is the low quantum efficiency

    of the structure. Integration time should be increased to keep the sensitivity at ac-

    ceptable level, and therefore, very high speed imaging is not feasible. With proper

    selection of material system, spectral response of QWIPs covers the entire infrared

    region, and photoresponse is narrow band type unlike low bandgap photon detectors.

    This property of narrow band type response is an advantage for multi band imaging

    decreasing spectral cross talk and facilitating spectral tuning. QWIP technology has

    reached megapixel size with dual-band ability [13]. QWIPs are investigated in further

    detail in the next chapter.

    Figure 1.3: Band structure of type II strained superlattices [14].

    1.3 Figure of Merits of Infrared Detectors and Imaging Systems

    In order to assess the performance of an IR detector or imaging system one needs

    quantitative measures. In this section, fundamental figures of merit for an IR detector

    and imaging system are presented. Main performance parameters for an IR detector

    can be listed as follows

    • Responsivity

    10

  • • Noise level

    • Detectivity.There are also parameters that are defined at system level, and commonly used ones

    can be given as follows

    • Noise equivalent temperature difference (NETD)

    • Minimum resolvable temperature difference (MRTD)

    • Modulation transfer function (MTF).

    Responsivity:

    Responsivity is the magnitude of the response of the detector to unit radiation power.

    It provides information on gain, linearity, dynamic range, and saturation level of the

    detector. Responsivity is a measure of the transfer function between the input signal

    photon power or flux and the detector signal output

    < = outputsignalinput f lux

    =iphotoφ

    . (1.4)

    Output signal is expressed in volts or amperes, and input signal is expressed in ei-

    ther watts or photons/sec. Responsivity can be defined as a function of wavelength in

    which case it is called spectral responsivity. Responsivity can be also defined inde-

    pendent of wavelength and detector’s spectral response, assuming that response is flat

    all over the blackbody spectrum at a specific temperature. This interpretation is called

    blackbody responsivity, and it is used in conjunction with a blackbody temperature. It

    is possible to switch from blackbody responsivity to spectral responsivity with a mul-

    tiplicative factor called peak factor, which is the ratio of the integrated normalized

    blackbody spectrum to detector’s integrated response spectrum.

    For photon detectors, responsivity is commonly defined in units of ampere/watts and

    is expressed as follows

    < = qηg λhc. (1.5)

    In this equation, η is the quantum efficiency meaning how many electrons are excited

    per incoming photon. Symbol ’g’ is the photoconductive gain meaning how many

    electrons are collected at the external circuit for a single excited electron in the de-

    tector. For photovoltaic detectors g is close to unity. Photoconductive detectors have

    11

  • variable gain values, and g is expressed as the ratio of the average drift distance to

    total device length.

    Noise:

    Noise level in the system determines the system performance. Most pronounced noise

    sources for IR detectors are generation-recombination (g-r) noise, shot noise, photon

    noise, 1/f noise, and thermal noise.

    Even though it is not well understood yet, 1/f noise is thought to arise from material

    instabilities. For GaAs QWIPs, it is shown experimentally that 1/f noise seldom limits

    detector performance and thermal noise, which is inherent to all resistive devices, has

    a negligible contribution [15].

    Shot noise is statistical noise and arises from random arrival rates of the electrons in

    the device. It is expressed as follows

    i2n = 2qIdevice∆ f (1.6)

    where Idevice is the device current, ∆f is the bandwidth, and q is the unit charge.

    Dominant noise mechanisms in QWIPs are photo- and dark current originated g-r

    noise. Randomness of the generation and the recombination events creates the noise

    called g-r noise. The g-r noise is expressed as

    i2n = 4qgnoise(Iphoto + Idark)∆ f . (1.7)

    In this equation gnoise is the noise gain, and under normal operation it is taken to

    be equal to photoconductive gain [16]. Iphoto + Idark are the two components of the

    current flowing through the device. Idark is the current which is unavoidably flowing

    through the detector even without any illumination. Dark current is dominated by

    thermal activation of electrons and is minimized by lowering the detector temperature.

    The condition where Iphoto � Idark is called background limited performance (BLIP)condition, and it defines the operation temperature of the detector.

    12

  • Detectivity:

    Although responsivity, by itself, gives how much signal is generated by the detector

    to incoming irradiance, it gives no indication of minimum amount of radiant flux that

    can be detected. Detection of small amount of irradiance is inhibited by the noise

    level of the detector. A convenient descriptor for minimum detectable signal is the

    noise equivalent power (NEP) which is the required radiant flux to give signal output

    equal to detector noise [17]. NEP is expresssed as

    NEP =inoise< . (1.8)

    However, NEP is a situation specific descriptor and does not allow direct comparison

    of different detectors. For this purpose, the noise term in the NEP definition is nor-

    malized to detector area and measurement bandwidth, and the inverse of the result is

    defined as specific detectivity (D∗) as

    D∗ =

    √A∆ f

    inoise< [cm√Hz/Watt] (1.9)

    In order to study the theoretical detectivity limit for photoconductive detector, con-

    sider the incoming radiant power given as follows

    φ = φpAhcλ

    (1.10)

    where φp is the photon flux similar to that expressed in Eqn. 1.3, and A is the detector

    area. Also consider the signal generated by the detector which is expressed as follows

    iphoto = qηgphotoAφp (1.11)

    Employing Eqn. 1.10, 1.11, 1.7 and Eqn. 2.13 and making the assumptions of

    Iphoto � Idark and gphoto ∼ gnoise yield the ultimate performance of the photocon-ductive detectors as

    D∗BLIP =λ

    2hc

    √η

    φp. (1.12)

    Noise Equivalent Temperature Difference:

    NETD is the target-to-background temperature difference that produces a signal equal

    to rms background noise. Since the change in the photon exitance with temperature

    will be different at different temperatures, NETD values are also specified with a

    13

  • background temperature. Smaller NETD values indicate better thermal sensitivity.

    NETD is a measure of the thermal sensitivity of the whole imaging system and ig-

    nores spatial resolution [17]. Its analytical expression is given as follows

    NET D =(4 f 2 + 1)

    √∆ f√

    A∫λ

    D∗(λ)dMp(λ,T )dT dλ[Kelvin]. (1.13)

    In this equation f is the f-number of the optics which describes the field-of-view of the

    optical aperture. This equation is valid under detector limited performance situations.

    Under BLIP condition, NETD will depend on the f-number linearly because D∗(λ)

    will also exhibit linear f-number dependence. Taking nonuniformity of the array as

    an additional noise component, NETD for a system may also be uniformity limited

    [18].

    Minimum Resolvable Temperature Difference:

    MRTD is subjective measure of both spatial resolution and thermal sensitivity. MRTD

    measurement encompasses all the components of the system including the user itself.

    Its measurement includes an ensemble of observers looking to a standardized 4-bar

    pattern with an aspect ratio of 7:1. Observers determine the minimum temperature

    difference required to resolve 4-bar pattern from the background. Analytically MRTD

    is directly proportional to NETD and inversely proportional to MTF. MRTD is also

    affected by parameters like equivalent integration time of human eye [17, 19].

    Modulation Transfer Function:

    MTF characterizes spatial resolution and image quality by means of spatial frequency

    response of an imaging system. It has complex measurement routines, and it is a

    function of detector pitch, pixel to pixel crosstalk, fill factor, quality of optics, etc.

    MRTD measurement is preferred over MTF alone because MRTD does not regard for

    a noise level.

    1.4 Scope and Objective of This Work

    The third generation concept brings the requirements of improved reconnaissance per-

    formance, dual-band capability, and reduced cost to the IR photon sensors. Main mo-

    14

  • tivation in this work is to develop and demonstrate a dual-band large format quantum

    well infrared photodetector focal plane array in order to satisfy these requirements.

    The dual-band capability of the detector is obtained by using the voltage tuning of

    the spectral response principle. Even though there are several methods to adjust the

    spectral response of the detector by changing the bias voltage, the mechanism used

    in this work is not well investigated in the literature. Therefore this work has an im-

    portant contribution to the literature in the sense that it demonstrates the application

    of its principles, especially at the final product level. An important advantage of the

    voltage-tuning principle employed in this work is that it is fully compatible with the

    present single band detector technology, therefore time and effort to put this technol-

    ogy on the market is drastically reduced.

    In the first phase of the work, detector structure is designed and the proper opera-

    tion of the design is justified. Complete fabrication and the characterization of the

    designed detector structure were performed in the second phase of the work. Next

    chapter discusses the dual-band imaging with QWIPs and describes the design phase

    carried out in this work.

    15

  • CHAPTER 2

    Dual-Band Detection with QWIPs and Design Approach

    In this chapter of the thesis, first the requirement for dual-band imaging is justified

    by listing the advantages and exemplifying the important applications. Following this

    justification, present status of the dual-band detector technology is presented with

    examples from the literature. Thirdly, necessary background is given to understand

    the basics of operation and fabrication of QWIPs. Finally design process is described

    in details including auxiliary studies and investigations.

    2.1 Advantages of Dual/Multi-Band Detection

    Dual/multi-band imaging (sometimes called hyperspectral imaging) provides plenty

    of benefits and reveals new potential applications. Improved reconnaissance perfor-

    mance is the main factor calling for simultaneous imaging in different bands or differ-

    ent spectral regions in the same band. Advantages provided by hyperspectral imaging

    and some of potential applications are discussed in this section.

    MWIR band is well known for its high maritime performance and its immunity to the

    degradations caused by water vapor. Also MWIR band is preferred in narrow field-of-

    view applications due to its lower diffraction limit. However, when a scene contains

    hot sources or reflections from hot sources, imaging quality in the MWIR band is

    greatly degraded by saturation of the readout circuit capacitors. On the other side,

    LWIR band is advantageous when the background radiation is too low for MWIR

    imaging or the detection range is degraded by atmospheric scattering. LWIR band is

    also insensitive to stray light effects and saturation by hot sources due to its high dy-

    16

  • namic range. In a dual/multi-band imaging system, these inherent advantages of each

    band are combined in a single system. Therefore optimum imaging quality is ensured

    under many scenarios and circumstances where performance of a single band detec-

    tor is not sufficient [20]. Improved image quality and reconnaissance performance

    are especially desirable for early warning systems. Utilization of dual/multi-band de-

    tectors will decrease false alarm rates, and reaction time will be shortened compared

    to a single band system.

    Imaging in more than one band allows new interpretations to emissivity and black-

    body characteristics. Considering the emissivity specific features of a scene, target

    recognition and disclosure of camouflaged targets can be described by algorithms

    thus automated. On the other side, long-range accurate temperature measurement

    is possible by discarding the emissivity changes. Accurate long-range temperature

    measurements are useful for a number of important applications. As an example, in

    battlefield and emergence situations, such as those requiring triage, vital physiolog-

    ical parameters, such as skin temperature, must be measured when direct contact is

    not possible [1, 21].

    Another advantage of the hyperspectral imaging is that it allows the detection of spe-

    cific emissions by spectral comparison. Specific emissions in a certain spectral region

    can be caused by combustion, explosions, lasers or other emitters. These sources can

    be differentiated from blackbody radiation by checking their counter part in another

    band. Such a differential operation yields very high prediction accuracy.

    MWIR and LWIR signature of rocket vehicles and their plumes are of great interest

    for detection, recognition and interception. Rocket plumes present a large MWIR

    signal that can be detected and tracked at hundreds of kilometers away. However, for

    interception it is necessary to image the hardbody of the missile which radiates mostly

    in the LWIR band. LWIR signature of the plume is stronger than that of hardbody

    but it is weaker than plume’s MWIR signature. Therefore, it is advantageous for an

    interceptor to have imaging ability in both MWIR and LWIR regions [22].

    With various additional potential applications not mentioned here, dual-band imaging

    is an active research area and significant amount of research is in progress for low cost

    and more talented dual-band detector arrays.

    17

  • 2.2 State-of-The-Art Dual-Band Detectors

    Currently different research groups are working on various infrared materials to im-

    plement dual-band detector arrays. Even though each detector technology has its

    own advantages and shortcomings, general trend is toward megapixel size arrays with

    highly customized ROICs.

    In the year 2006, Raytheon Vision Systems (RVS) has developed and demonstrated

    the first 1280 x 720 pixel dual-band MWIR/LWIR FPAs under the U.S. Army’s Dual-

    Band FPA Manufacturing program. These dual-band detector arrays are based upon a

    mercury-cadmium-telluride (HgCdTe) n-p-n triple-layer heterojunction (TLHJ) back-

    to-back diode detector structure. The ”bottom” n-type absorbing layer has a shorter

    cutoff (MWIR) than the ”top” n-type absorbing layer (LWIR). Detector arrays are

    fabricated by etching mesa structures in the TLHJ and extending beyond the bottom

    p-n (MWIR) junction. Only a single indium bump per pixel is then required for ROIC

    interconnections. The cross section of the stack diode structure is shown in Figure 2.1.

    This stacked structure permits either of the two p-n junctions within each detector

    element to be reverse-biased by applying the appropriate voltage to the mesa top

    contact. By alternating between two pre-determined bias voltage values, the photo-

    response of the diode stack can be switched between MWIR and LWIR bands. The

    bias switching is either performed on an alternating frame basis or many times within

    a single frame period. The 1280 x 720 detector arrays have LWIR cut-off wavelengths

    longer than 10.5 µm at 78 K. These FPAs have demonstrated high-sensitivity at 78

    K with MWIR NETD values less than 20 mK and LWIR NETD values less than 30

    mK with f/3.5 aperture. Pixel operability greater than 99.9% has been achieved in the

    MWIR band and greater than 98% in the LWIR band.

    Jet Propulsion Laboratory (JPL) demonstrated a megapixel (1024x1024) dual-band

    QWIP FPA based on the standard AlGaAs/InGaAs/GaAs material system [13]. Full

    width at half maximum (FWHM) of the MWIR part covers 4.4-5.1 µm and that of the

    LWIR detector covers 7.8-8.8 µm. The FPA was fabricated with co-registered pixels

    with two indium bumps per pixel approach. MWIR part consists of coupled-quantum

    wells to broaden the spectral response. At 90 K FPA temperature, MWIR stack is

    under BLIP condition with f/2.5 optics up to -1 V bias and exhibits a peak detectivity

    18

  • of 4x1011 cmHz1/2/Watt. LWIR stack is BLIP at 72 K temperature with -1 V bias

    and f/2.5 aperture. Peak detectivity for LWIR stack is 1x1011 cmHz1/2/Watt at 70 K

    FPA temperature. They report NETD values of 27 and 40 mK at 70 K FPA tempera-

    ture with f/2 optics for MWIR and LWIR band, respectively. Main problem with the

    approach is the complexity of the fabrication procedure (thirteen layer photolithog-

    raphy) which results in low operability. Operability of the reported FPA is around

    90 % according to 3σ definition for both bands, and low operability is explained by

    discontinuities of the metal lines.

    Figure 2.1: Pixel architecture for TLHJ type dual-band detectors. Band 1 is designedto respond to MWIR radiation and Band 2 is designed to respond to LWIR radiation[11].

    These two demonstrations by RVS and JPL represent the latest published results in

    the dual-band FPA field. There are other mid- and large-format FPAs previously

    reported by several groups. Lockheed Martin demonstrated 256 x 256 pixel dual-

    band FPAs with 40 µm pitch size featuring three indium bumps per pixel [23]. The

    reported MWIR/LWIR FPA shows detection in two separate bands peaked at 5.1 µm

    and 8.5 µm. The pixel operability of the FPA is larger than 97%, NETD is better than

    35 mK (T = 65 K, f/2 optics, 295 K background, and 100 Hz frame rate), and the

    uncorrected uniformity (sigma/mean) is smaller than 4% for both colors. The pixel

    reading scheme of the FPA is illustrated in Figure 2.2. They have also demonstrated

    FPAs in MWIR/MWIR and LWIR/LWIR configurations with the same approach.

    19

  • Figure 2.2: Pixel architecture with multiple electrical contact. The responses fromred QWIP and blue QWIP are integrated into different capacitors [23].

    CEA/LETI achieved the dual-band imaging capability with multiple contact pixel

    architecture using HgCdTe detector technology [24]. The epilayer structure of the

    reported detector consists of a p-type MWIR HgCdTe layer grown on top of a LWIR

    one. These two layers are isolated by a layer with larger bandgap (SWIR). The fab-

    rication starts with the etching of the holes in order to reach the buried LWIR region.

    Two photodiodes are then formed by n-type ion implantation (Figure 2.3). The re-

    maining p-type layers are connected together as the detector common node (Figure

    2.3). The reported FPA has cut-off wavelengths of 4.8 and 9.7 µm in the MWIR

    and LWIR bands, respectively. The fabricated FPA is hybridized to a custom ROIC

    designed for this structure, and characterized at 77 K detector temperature with f/2

    optics. NETD values of 12 mK and 25 mK are reported with operabilities of 99.1 %

    and 95.1 % in MWIR and LWIR bands, respectively.

    Figure 2.3: Multiple contact pixel architecture by CEA/LETI. Two photodiodes areformed in different p-type regions by ion implantation (right), p-type regions are con-nected together as the detector common node (left) [24].

    20

  • A different approach to achieve multi-band capability is demonstrated by Gunapala

    et.al [25]. In this approach, the pixels responsive in different bands are spatially

    separated on the same array. A four-band device structure is achieved by the growth

    of multi-stack QWIP structures separated by heavily doped n+ contact layers on a

    GaAs substrate. Device parameters of each QWIP stack are designed to respond in

    different wavelength bands. The schematic of the device structure is shown in Figure

    2.4. Four separate detector bands are defined by a deep trench etch process, and the

    unwanted spectral bands are eliminated by a detector short circuiting process. The

    unwanted top detectors are electrically shorted by gold coated etched gratings. The

    unwanted bottom QWIP stacks are electrically shorted at the end of each detector

    pixel row. The FPA is hybridized to a standard CMOS multiplexer. The spectral

    response of the four stacks are reported to be between 4-5.5, 8.5-10, 10-12, and 13.5-

    15 µm. The NETDs are 21.4, 45.2, 13.5, and 44.6 mK (at 40 K) in these bands,

    respectively. An example image recorded with the four-band detector is shown in

    Figure 2.5.

    Figure 2.4: Layer diagram of the four-band QWIP device structure. Each pixel rep-resent a 640 x 128 pixel area of the four-band focal plane array [25].

    21

  • Figure 2.5: An image recorded with the four-band detector [25].

    2.3 Advantages of QWIPs for Dual/Multi-Band Detection

    Different methods are present in the literature in order to implement dual-band ca-

    pability in different detector technologies. Advantages and disadvantages of these

    methods and technologies should be mentioned before investigating the advantages

    of QWIPs over them.

    QWIPs are mostly challenged by low bandgap infrared materials. Cut-off wavelength

    is the only adjustable parameter for the spectral response of a low bandgap material,

    and material is sensitive to all the wavelengths below this limit. It is too difficult to

    isolate the spectral regions in a dual-band detector with a low bandgap semiconductor

    because the material with higher cut-off wavelength tends to cover all the response

    from the material with lower cut-off. In the triple layer heterojunction type dual-band

    detection, spectral regions are separated by exploiting the high absorption quantum

    efficiency of the material. The layer with higher cut-off wavelength is placed first

    on the path of illumination, and behaves as a filter for the layer with lower cut-off

    by absorbing all the photons above its bandgap energy. However, spectral isolation

    with this approach is limited, and it is very difficult to create spectral gaps if not

    impossible. At this point, QWIPs are clearly very advantageous since the cut-on

    and cut-off wavelengths can be adjusted independently for each stack, and precise

    22

  • tailoring of the spectral response is very easy.

    HgCdTe, which is one of the most popular detector materials, still suffers from ma-

    terial instabilities and lack of high quality large area substrates after decades of its

    discovery. These facts increase the cost of the HgCdTe based detectors by decreasing

    yield and production volume. Lattice matched 7x7 cm2 CdZnTe substrates are pro-

    nounced as state of the art substrate sizes [11]. On the other side, GaAs based QWIP

    technology is very mature and well proven. Fabrication methods are well investigated

    in the literature, and high quality large substrates are widely available at affordable

    prices. Resulting high volume production capability with low cost and high yield

    makes QWIPs very favorable as an infrared detector solution.

    Thermal imaging performance of the QWIPs is shown to be more than satisfactory

    at moderate frame rates, and dual-band imaging capability can be easily integrated

    into the present technology without compromising any of the advantages of the sin-

    gle band QWIPs. Therefore, QWIPs are very promising as a solution to dual-band

    imaging requirements of the new generation systems with low cost at the sensor level.

    2.4 QWIP Basics

    Design of QWIP structures requires the knowledge on the key performance parame-

    ters which will be presented in this section.

    2.4.1 Theory of Operation

    Theory of operation of QWIPs is briefly introduced in section 1.2. QWIP structures

    consists of alternating layers of a large bandgap barrier material and low bandgap well

    material. This well-barrier stack is placed between two highly doped contact layers

    as shown in Figure 2.6. Discontinuities in the conduction band and valence band of

    the structure produce a well shaped confinement in one dimension. This confinement

    leads to quantization of the energy levels and discrete energy states are formed. The

    band diagram of a typical QWIP structure is shown in Figure 2.7.

    Incoming radiation is detected by the interaction of the quantized electrons with the

    23

  • photons. Photoexcited electrons form the photocurrent flowing through the device.

    Under dark conditions, excitations are mostly due to thermionic emission which con-

    stitutes the device dark current. Schrödinger equation with proper boundary condi-

    tions should be solved to find the energy levels. Energy levels are determined by the

    width and depth of the well and the effective mass of the electron in the well material.

    Width of the well is controlled with the thickness of the well material, and depth of

    the well is dependent on the band discontinuity which is controlled with the composi-

    tions of well and barrier materials. Transitions between ground state and first excited

    state are of interest for regular QWIP operation. Change of energy levels and corre-

    sponding peak absorption wavelength with well width are depicted in Figure 2.8 for

    different compositions of AlxGa1−xAs/GaAs material system.

    Figure 2.6: Typical QWIP structure [26].

    Figure 2.7: Typical band diagram of the QWIP [26].

    24

  • Figure 2.8: Dependence of energy states [left] and peak absorption wavelength [right]on well width for different Al mole fraction [27].

    QWIPs are classified according to the position of the first excited energy state. QWIPs

    in which the first excited energy state is below the edge of the barrier are called bound-

    to-bound QWIPs. First excited state can be located at the edge of the barrier, then it

    is called a bound-to-quasibound QWIP. In these two cases, excited electron needs

    to tunnel through the tip of the barrier where high electric field is required to force

    excited electrons to tunnel. QWIPs in which the first excited state is above the edge of

    the barrier are called bound-to-continuum QWIPs, and small electric field is required

    only to collect the electrons excited to continuum. As the excited state moves into

    continuum, detector responsivity becomes less dependent on bias voltage, and the

    spectral response becomes wider. The typical spectral responses of different QWIP

    types are shown in Figure 2.9. Transition rate from ground states to excited states

    can be found via Fermi’s Golden Rule. Using the transition rate, theoretical quantum

    efficiency is estimated as follows [15, 18]

    η =q2hS in2θ

    Cosθn2D f δ(E2 − E1 − ~ω). (2.1)

    In this equation, n2D is two dimensional electron density in the well, θ is the angle

    of incidence, and f is the oscillator strength. Dependence of quantum efficiency on

    the angle of incidence predicts no response to normal incidence. The quantum effi-

    ciency is maximized if the radiation is perpendicular to growth direction which is not

    practically possible.

    25

  • Figure 2.9: Typical spectral responses of different QWIP types.

    In practice, QWIPs show non-zero response to normal incident radiation, and this is

    explained by auxiliary mechanisms and imperfections. Light coupling schemes are

    utilized to increase the quantum efficiency of QWIPs which is in the range of 5-10%.

    Implementing grating structures on top of the pixel is a common method to increase

    light coupling of QWIPs as shown in Figure 2.10. Incoming radiation is diffracted by

    the grating structure creating a component parallel to the quantum wells. Diffraction

    gratings are applicable at the FPA level and commonly employed. Another method to

    increase light coupling efficiency in QWIPs is the corrugated pixel structure. In this

    method, FPA pixels are structured so that incoming photons are reflected from walls

    and propagate parallel to the active region. Illustration of the structure is depicted in

    Figure 2.11.

    As reflected by Eqn.2.1, quantum efficiency of the device is increased with the well

    doping density. However, dark current in the device also increases with increasing

    doping density. Another mechanism determining the dark current level in the device

    is the transition type. Dark current density is increased as the excited level is moved

    into continuum. Therefore, the position of the excited state determines both the dark-

    and photo-current at a specific bias. As shown in Figure 2.9, αp∆λ/λpeak

    NDis almost

    constant where αp is the peak absorption coefficient, and ∆λ is the full width at half

    maximum of the responsivity spectrum.

    26

  • Figure 2.10: Illustration of the effect of diffraction grating on the incoming radiation[26].

    Figure 2.11: Corrugated pixel structure [28].

    Capture and emission processes in QWIPs are illustrated in Figure 2.12. Capture

    probability (pc) in a QWIP is defined as the probability of an electron to be captured

    by a well while it is flowing through the device.

    If the device current is dominated by photocurrent (Ip), the current captured by each

    well (pcIp) must be equal to the current emitted (Iem). The emitted current is expressed

    as

    Iem = qφηsingle well = pcIp (2.2)

    where φ is the number of photons falling on the detector in one second, and ηsinglewell

    is the quantum efficiency of a single well.

    27

  • Figure 2.12: Illustration of the capture and emission mechanisms in QWIPs [26].

    The total photocurrent expression generated by unit area of the detector is given as

    Ip = qφηgphoto (2.3)

    where η is the quantum efficiency of the overall structure, and gphoto is the photocon-

    ductive gain. Therefore, the gain of a QWIP is expressed as

    g =1pc

    ηsingle well

    η. (2.4)

    It is known that quantum efficiency of a single well is very low (∼ 1%), thus it isconvenient to assume η � ηsingle wellN, where N is the total number of wells [18].

    Therefore, the gain expression reduces to

    gphoto =1

    N pc. (2.5)

    Photoconductive gain is also given by [18]

    gphoto =τLτt

    =τLυdri f t

    τtυdri f t=

    Ldri f tLdevice

    (2.6)

    where τL is the lifetime of the photoexcited carriers, τt is the transit time of the carriers

    through the QWIP structure, υdri f t is the drift velocity of the carriers, Ldri f t is the drift

    distance of the excited carriers, and Ldevice is the length of the quantum well stack.

    More intuitively, this expression states that the gain of the device is equal to the ratio

    of the total number of excited electrons to the number of electrons collected at the

    contacts.

    28

  • The Johnson noise and 1/f noise are not dominant in QWIPs under typical operating

    conditions. The G-R noise is generally the dominant noise source in QWIPs, and

    this noise arises from fluctuations in the device current due to both dark- and photo-

    current generation mechanisms. Total G-R noise in a QWIP is therefore expressed

    as

    in =√

    4qgnoise(qnthermalυdri f tAd + qφηgphoto)∆ f (2.7)

    where nthermal is the thermally excited electron density, and the products qφηgphoto and

    qnthermalυdri f tAd represent the photocurrent and the dark current flowing through the

    device, respectively.

    Under dark current limited conditions caused by low illumination level or high detec-

    tor temperature, expression of detector noise is reduced to

    in =√

    4q2gnoisenthermalυdri f tAd∆ f . (2.8)

    Detectivity of the QWIP is therefore expressed by using Eqns. 1.5, 1.9 and 2.8 as

    D∗dark =η

    2hc/λ

    √τL

    nthermalLdevice. (2.9)

    The temperature dependence of the detectivity of the QWIP is given in Figure 2.13.

    The detectivity increases with decreasing detector temperature until the BLIP tem-

    perature is reached. This increase in the detectivity is due to the suppression of

    the thermionic emission with decreasing temperature. After the BLIP temperature

    is reached, detectivity is background limited and does not considerably depend on

    detector temperature. Under this (BLIP) condition the detectivity is given as

    D∗BLIP =λ

    2hc

    √η

    φp. (2.10)

    The above equation shows that the BLIP detectivity is independent of the device gain.

    However, it depends considerably on the quantum efficiency of the detector.

    29

  • Figure 2.13: Temperature dependence of the detectivity of the QWIPs [26].

    2.4.2 Material Systems for QWIPs

    The major advantage of the QWIP technology is its low cost due to the relatively

    mature III-V semiconductor technologies utilized in the fabrication of focal plane

    arrays. While the III-V semiconductor famility offers a wide range of heterostructure

    systems suitable for QWIP fabrication, the low cost nature of the technology can

    only be preserved if the device is based on mature III-V semiconductor systems such

    as AlGaAs/GaAs or InP/InGaAs(P). Table 2.1 lists the alternatives offered by these

    material systems for different applications.

    Material systems on InP provides certain advantages for single and dual-band QWIPs.

    While the InP/InGaAs and InP/InGaAsP systems provide higher responsivity than

    their GaAs based counterparts in the LWIR band, AlInAs/InGaAs system provides

    a lattice matched solution in the MWIR band [29]. Higher responsivity of the InP

    based QWIPs is a result of the larger gain offered by this material system. This larger

    gain arises from the higher drift distance of the photoexcited electrons which was

    shown by detailed ensemble Monte Carlo simulations to be due to longer lifetime of

    the photoexcited carriers [30].

    30

  • Table 2.1: Mature III-V material systems used for QWIP fabrication. Eg1 and Eg2are the bandgaps of the materials and ∆Ec is conduction band discontinuity of theresulting heterostructure.

    Material System Eg1 (eV) Eg2 (eV) ∆Ec Comments

    AlxGa1−xAs/GaAs Variable 1.43 Variable Used for LWIR QWIPs∆Ec is insufficient forMWIR QWIPswith acceptable x

    AlxGa1−xAs/InyGa1−yAs Variable Variable Variable Used for MWIR QWIPsInP/In0.53Ga0.47As 1.35 0.75 0.25 Used for LWIR QWIPs

    High

  • NETDs of the FPA are as low as 14 and 40 mK with integration times as short as 1.8

    ms and 250 µs (f/1.5, 65 K). Without field of view correction, the uncorrected DC

    signal and NETD nonuniformities of the FPA are 5.5% and 15%, respectively [34].

    Figure 2.14 shows the NETD of the FPA calculated using the measurements on the

    test detectors which are identical to the FPA pixels. The integration times shown at

    each bias voltage correspond to half-filled ROIC capacitors (1.1x107 electron capac-

    ity). A thermal image recorded with an integration time of 500 µs is shown in Figure

    2.15 displaying the successful operation of the FPA with very short integration times.

    Figure 2.14: Dependence of the NETD on detector bias [34].

    32

  • Figure 2.15: Thermal image recorded with the 640x512 InP/InGaAs FPA at 65 Kdetector temperature with an integration time of 500 µs and f/2 optics [34].

    In the MWIR band, we have shown that cut-off wavelength of AlInAs/InGaAs QWIPs

    on InP substrates can be adjusted very flexibly from 4.15 µm to 5.1 µm by changing

    only the well width without deviating from the lattice matched composition [35].

    Figure 2.16 shows the normalized responsivity spectrum of various InAlAs/InGaAs

    MWIR QWIPs with well widths ranging from 22 to 30 Å. We have demonstrated

    the performance of the AlInAs/InGaAs material system in the MWIR band at the

    FPA level. A large format (640x512) FPA was fabricated using an epilayer structure

    consisting of thirty 23 Å thick, Si doped (ND = 4 x 1018 cm−3) In0.53Ga0.47As quan-

    tum wells sandwiched between 300 Å thick Al0.52In0.48As barriers. The FPA yielded

    excellent sensitivity with an NETD as low as 22 mK (τint = 20 ms) and a BLIP tem-

    perature as high as 115 K with f/1.5 aperture and 300 K background [29].

    33

  • Figure 2.16: Responsivity spectrum of AlInAs/InGaAs/InP QWIPs with differentwell widths [29].

    2.5 Design Approach

    Our design process of a MWIR/LWIR dual-band QWIP FPA is started with the selec-

    tion of the virtually most feasible approach for the fabrication of large format dual-

    band detector arrays. The detector epilayer structure is designed based on the theoret-

    ical expectations and experimental observations. Once the measurements on the test

    detectors reflected acceptable performance, large format FPAs were fabricated with

    the optimized epilayer structure.

    2.5.1 Selection of Dual-Band Operation Type

    In order to implement dual-band sensing capability on a single detector array, there

    are mainly three schemes implemented in the literature. The first approach is to em-

    ploy a multi-contact pixel geometry [13, 23, 24, 36]. In this method, two independent

    active regions are grown one on the top of other, and these active regions are driven

    by two or three electrical contacts. The main advantage of this method is that it

    34

  • provides temporal and spatial coherence at the same time. However, fabrication of

    such a structure deviates from the standard single contact process, and it is greatly

    complicated. Furthermore, placing at least two In bumps on each pixel practically es-

    tablishes a lower limit for the pitch size making it impossible to obtain high resolution

    with an acceptable FPA size. Pixel structure with two and three electrical contacts are

    illustrated in Figure 2.17.

    Figure 2.17: Multi-contact pixel schemes for co-registered dual-band operation. (a)Fully isolated pixels with three electrical contacts. (b) Mid-contact layer is shortedto ground contact layer as the detector common. Passivation layer exists betweendetector stacks [13].

    The second approach is to employ space shared pixel placement [37]. In this ap-

    proach, two independent active regions are grown. One stack of each pixel is deacti-

    vated either by removing (etching) the stack or short circuiting both ends of the stack.

    Pixels are placed in the form of a checkerboard or line by line. This approach is not

    preferred because spatial coherence is crucial in most of the applications. Addition-

    ally, operating conditions (detector bias, integration time and well capacity) can not

    be optimal for both bands at the same time, unless a special ROIC is used. Further-

    more, fill factor of the FPA is greatly degraded. A sample pixel placement for space

    sharing dual-band operation approach is illustrated in Figure 2.18. Demonstration of

    implementation of this approach has been limited in the literature.

    35

  • Figure 2.18: Space sharing pixel placement. Pixels are placed in lines or as in acheckerboard [38].

    The third approach is to obtain dual-band operation by tuning of the spectral response

    of the detector with the applied detector bias [11, 39, 40, 41]. In this method, detector

    structure is specially designed so that the peak responsivity wavelength shifts from

    one band to the other with the changing detector bias. The detector structure used

    in this approach is illustrated in Figure 2.19. The main advantage of bias tuning

    scheme is the utilization of the same fabrication process with single band detectors.

    Consequently, production is simple, yield is high, cost is low, and detector pitch can

    be decreased freely to keep the resolution diffraction limited. The main issue with

    this approach is the lack of temporal coherence. Pixels can be operated in one band

    at a time. In order to obtain the scene in two bands simultaneously, the detector

    bias is continuously switched in consecutive frames during imaging. Therefore, there

    exists one frame delay between the frames of the same band. This reading scheme

    is problematic only at rapidly changing scenes. The time delay between the frames

    of each band can be reduced by decreasing the integration time of the other band.

    However, it should be noted that the detector’s sensitivity is sacrificed in this case.

    Noting the need for low cost dual-band sensors in affordable third generation thermal

    imagers, this approach is selected in this work because of its fabrication simplicity,

    smaller array size, high resolution and standard ROIC advantage.

    36

  • Figure 2.19: Dual-band operation employing one contact per pixel. Two active re-gions exist in each pixel and they are activated with the applied bias.

    Several methods are demonstrated in the literature to obtain dual-band voltage tun-

    able spectral response. Choi et.al. demonstrated a 256x256 voltage tunable AlGaAs/

    InGaAs/GaAs MWIR/LWIR QWIP FPA with NETD performances of 27 mK (τint =

    33 ms) in the MWIR band and 90 mK (τint = 2 ms) in the LWIR band for f/2.44 optics

    and 50 K FPA temperature [40]. Spectral tuning mechanism was based on photocur-

    rent asymmetry in a double-supperlattice quantum well structure. In this structure,

    two quantum well stacks with thin barriers are separated by a graded barrier in the

    middle as shown in Figure 2.20.

    Figure 2.20: Band diagram of a voltage-tunable two-color detector. The numeralswithout units are either energies in milli-electron volts or thicknesses in angstroms.The Al and In molar ratios are denoted by x and y, respectively [40].

    37

  • There is no voltage drop on the stacks because of the sequential tunneling through

    ground minibands, and the resistance of the structure is determined by the middle

    contact. When a bias is applied to the structure, the electron injection into the barrier

    is controlled by the photoexcitation from one of the stacks only. Therefore spectral

    response of the overall structure is switched from one stack to the other by switching

    the bias polarity as shown in Figure 2.21. However, the main drawback of the struc-

    ture is the large thermally assisted tunneling which results in high dark current and

    low operating temperature. Currently, this fundamental problem exists as a limitation

    to reach a feasible operating temperature with this approach.

    Figure 2.21: Responsivity spectra under various bias voltages [40].

    Zhang et.al. also published promising results on voltage tuning of spectral response

    [42]. They reported the single pixel characterization of a combined photovoltaic (PV)

    and photoconductive (PC) QWIP structure. PV stack (MWIR) consists of AlGaAs/

    AlAs/GaAs double barrier quantum wells, and PC stack (LWIR) consists of standard

    bound-to-continuum AlGaAs/GaAs quantum wells. At zero external bias, responsiv-

    ity of the PV stack is close to saturation, and PC stack does not contribute to response,

    since there is no electric field to collect excited carriers. At positive bias, PC stack

    starts to show its response, but response of the PV stack is decreased because of the

    reduced drift distance of excited electrons [42]. Authors have no published work on

    the FPA level characterization of this detector structure.

    38

  • We reported a MWIR/MWIR dual-color QWIP FPA in which voltage-tunable spectral

    response has been achieved through series connection of two eight-well stacks of

    AlGaAs/InGaAs epilayers grown by molecular beam epitaxy on GaAs substrate [39].

    The peak responsivity wavelength of the detectors is shifted from 4.1 µm (color 1)

    to 4.7 µm (color 2) as the bias is increased from -1 V to -3.5 V as shown in Figure

    2.22. The operability of the FPA is ∼99.5%, and NETDs of ∼ 60 and 30 mK (f/1.5)are achieved in color modes 1 and 2, respectively. The voltage-tunable shift in the

    responsivity spectrum is attributed to two different QWIPs acting in series with each

    other. The distribution of the applied bias voltage between the two QWIPs (based on

    relative resistances) and their relative responsivities determine the spectral response

    of the structure [43].

    Figure 2.22: Voltage tunable MWIR/MWIR spectral response reported by Kaldirimet.al [39].

    In this thesis work, we adopted a similar approach which exploits the saturation of

    the responsivity and dark current limited operation in QWIPs based on the AlGaAs/

    InGaAs/GaAs material system. This approach will be detailed in the following sec-

    tion.

    39

  • 2.5.2 Operation of Voltage Tunable Detector Structure

    If a multi quantum well stack sensitive in the MWIR atmospheric window is con-

    nected in series with a LWIR sensitive stack, the spectral response of the resultant

    structure depends on the distribution of the applied bias voltage between these stacks

    and the relative responsivities of the stacks. On the other hand, the division of the bias

    voltage between the two stacks depends on the individual stack resistances. Under

    dark current limited conditions, the stack resistances do not considerably depend on

    the photocurrents flowing through them. Similarly, under background limited condi-

    tions, stack currents do not depend considerably on the detector bias after the respon-

    sivity is saturated. Therefore, if the structure is designed properly to switch significant

    portion of the total bias voltage from one stack to the other under two different bias

    voltages, the desired spectral response can be achieved.

    In order to explain the concept in further details, consider the characteristic I-V be-

    havior of the standard AlGaAs/GaAs QWIPs under illumination, as illustrated in Fig-

    ure 2.23. In region I, current increases strongly with increasing bias. Under BLIP

    conditions, this increase in the total current is mostly due to the increase in the pho-

    tocurrent which is indeed a result of increase in the responsivity. When biased in this

    region, dynamic resistance of the detector is relatively small. In region II, dynamic

    resistance increases due to saturation of the responsivity. When the responsivity of

    the stack is saturated and the dark current is negligible with respect to the photocur-

    rent (BLIP condition), dependence of the total detector current on the applied bias

    voltage is decreased. However, due to the high responsivity, detector current is very

    sensitive to the incoming radiation power. In Region III, dark current dominates over

    photocurrent and BLIP condition is no longer valid. Detector behavior is vanished

    unless extreme illumination is encountered.

    In our approach, dual-band capability is achieved by sequential growth of specially

    designed one color detectors separated by a conducting layer. It is experimentally

    shown that this behavior of stacked devices corresponds to individual devices simply

    connected in series [43].

    40

  • Figure 2.23: Characteristic I-V curve of standard AlGaAs/GaAs QWIP.

    In our stack design, we have adjusted the dc and dynamical resistances for both of

    the stacks such that, under low bias LWIR stack operates in region I and MWIR stack

    operates in region II. The total detector current is mostly controlled by the stack which

    has the larger dynamical resistance. At low bias, MWIR stack controls the detector

    current, and a response in the MWIR band is observed from the overall structure.

    When the bias is increased, MWIR stack is forced to enter the dark current dominated

    regime, and the photoresponse of the LWIR part is almost saturated. In other words,

    dynamical resistance of the MWIR stack is decreased when it moves from region II

    to region III, and the dynamical resistance of the LWIR stack is increased by moving

    from region I to region II. Therefore, control of the detector current is not at MWIR

    stack, and overall detector response is LWIR dominated. The corresponding operating

    points under low and high biases are illustrated in Figure 2.24.

    Figure 2.24: Operation points of the series connected QWIP stacks in MWIR andLWIR modes.

    41

  • 2.5.3 Design Targets

    In this section design targets which we have determined prior to design are described

    with justifications. Design targets include the spectral characteristics of the detector

    structure, spectral switching requirement, thermal imaging performance and feasible

    operation conditions.

    The spectral response of each stack should be properly designed to cover the neces-

    sary portions of the corresponding atmospheric windows. It is beneficial to extend

    the cut-off wavelength of the MWIR part toward 5.0 µm as much as possible due

    to significant increase of photon flux from room temperature objects. Cut-on wave-

    length should be selected taking into account scattered background from sun, stray

    light sources and hot sources. Absorption/emission band of CO2 should be also taken

    into account. In our design we targeted a peak response wavelength above 4.5 µm

    (upper end of the CO2 absorption region) and a cut-off wavelength around 5 µm. In

    the design of the spectral response of the LWIR part, there is an important trade-off

    between long range performance and operation temperature. Long range performance

    in the LWIR band improves significantly by increasing the cut-off wavelength above

    9.0 µm, especially under water vapor rich conditions. However, background limited

    performance is reached at lower detector temperatures as the cut-off wavelength is

    extended. This arises from the fact that the thermionic emission is stronger from

    shallower wells. In our design, we targeted a cut-off wavelength around 9 µm. Ac-

    ceptable limit for the operation temperature in our case is 65 K. Close cycle stirling

    coolers can reach this temperature with reasonable power consumption and lifetime.

    There exists an important restriction that the switching of the spectral response should

    be achieved within the applicable bias limit of a standard ROIC. In this work, com-

    mercially available ISC9803 ROIC from Indigo/FLIR Systems is used. Specifications

    of this ROIC are given in the Table 2.2.

    42

  • Table 2.2: Specifications of the ISC9803 ROIC.

    ISC9803Array Size 640x512Array Pitch 25 µmInput Circuit Direct InjectionInput Polarity p-on-nElectron Well Capacity 11.2 Me−

    Input Noise at Lowest Gain 550 e−

    Gain 2 bits (x1, x1.3, x2, x4)Integration Type SnapshotIntegration Modes Integrate-While-Read

    Integrate-Then-ReadNumber of Outputs Selectable 1,2 or 4Maximum Power Consumption 180 mW

    The ISC9803 uses a direct injection input circuit as shown in Figure 2.25. Detector is

    connected between the common node (Vdetcom) and the unit cell input by hybridiza-

    tion. Detector current flows through the input gate transistor and charges up the inte-

    gration capacitor. The voltage on the integration capacitor is sampled and multiplexed

    to the column amplifier. The column amplifiers provide sample/hold, amplification,

    and skimming functions. The signal from the column amplifiers are multiplexed to

    the output buffers through output multiplexer [44].

    The maximum allowed bias that can be applied to the detector by changing the Vdetcom

    voltage is 3.5 V. Therefore, the dominant response should switch from one stack to the

    other with adequate responsivity within the bias range of 0-3.5 V. Furthermore, the

    spectral modes should be separated with minimum crosstalk. Crosstalk mechanisms

    are investigated in further detail in the next chapter when discussing the experimental

    results.

    43

  • Figure 2.25: Simplified unit cell schematic of the ISC9803 [44].

    2.5.4 Design

    Overview of the voltage-tunable dual-band QWIP structure and design variables are

    shown in Figure 2.26. The AlGaAs/InGaAs/GaAs system is the standard material

    system used in MWIR QWIPs. Spectral response is determined by the width of the

    well and band discontinuity. In the design of the MWIR stack, reported results in the

    literature are investigated. In order to achieve the required band discontinuity, the Al

    mole fraction in the barriers and the In mole fraction in the wells are selected as x

    = 0.35 and y = 0.2, respectively. Our target sp