applied physics reviews quantum well photoconductors … · quantum well photoconductors in...

JOURNAL OF APPLIED PHYSICS VOLUME 93, NUMBER 8 15 APRIL 2003

APPLIED PHYSICS REVIEWS

Quantum well photoconductors in infrared detector technologyA. RogalskiInstitute of Applied Physics, Military University of Technology, 2 Kaliskiego St., 00-908 Warsaw, Poland

~Received 20 December 2002; accepted 3 January 2003!

The paper compares the achievements of quantum well infrared photodetector~QWIP! technologywith those of competitive technologies, with the emphasis on the material properties, devicestructure, and their impact on focal plane array~FPA! performance. Special attention is paid to twocompetitive technologies, QWIP and HgCdTe, in the long-wavelength IR~LWIR! andvery-long-wavelength IR~VLWIR ! spectral ranges. Because so far, the dialogue between the QWIPand HgCdTe communities is limited, the paper attempts to settle the main issues of bothtechnologies. Such an approach, however, requires the presentation of fundamental limits to thedifferent types of detectors, which is made at the beginning. To write the paper more clearly forreaders, many details are included in the Appendix. In comparative studies both photon and thermaldetectors are considered. Emphasis is placed on photon detectors. In this group one may distinguishHgCdTe photodiodes, InSb photodiodes, and doped silicon detectors. The potential performance ofdifferent materials as infrared detectors is examined utilizing thea/G ratio, wherea is theabsorption coefficient and G is the thermal generation rate. It is demonstrated that LWIR QWIP’scannot compete with HgCdTe photodiodes as single devices, especially at higher operatingtemperatures~.70 K!. This is due to the fundamental limitations associated with intersubbandtransitions. The advantage of HgCdTe is, however, less distinct at temperatures lower than 50 K dueto problems inherent in the HgCdTe material~p-type doping, Shockley–Read recombination,trap-assisted tunneling, surface and interface instabilities!. Even though QWIP is a photoconductor,several of its properties, such as high impedance, fast response time, long integration time, and lowpower consumption, comply well with the requirements imposed on the fabrication of large FPA’s.Due to a high material quality at low temperatures, QWIP has potential advantages over HgCdTe inthe area of VLWIR FPA applications in terms of array size, uniformity, yield, and cost of thesystems. The performance figures of merit of state-of-the-art QWIP and HgCdTe FPA’s are similarbecause the main limitations come from the readout circuits. Performance is, however, achievedwith very different integration times. The choice of the best technology is therefore driven by thespecific needs of a system. In the case of readout-limited detectors a low photoconductive gainincreases the signal-to-noise ratio and a QWIP FPA can have a better noise equivalent differencetemperature than an HgCdTe FPA with a charge well of similar size. Both HgCdTe photodiodes andQWIP’s offer multicolor capability in the MWIR and LWIR range. Powerful possibilities offered byQWIP technology are associated with VLWIR FPA applications and with multicolor detection. Theintrinsic advantage of QWIP’s in this niche is due to the relative ease of growing multicolorstructures with a very low defect density. ©2003 American Institute of Physics.@DOI: 10.1063/1.1558224#

75

TABLE OF CONTENTS

I. INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . .4356II. FUNDAMENTAL LIMITS TO INFRARED

DETECTOR PERFORMANCE. . . . . . . . . . . . . . . . 4357A. General theory of photon detectors. . . . . . . . . . . 4358B. General theory of thermal detectors. . . . . . . . . . 4359C. Comparison of the fundamental limits of

thermal and photon detectors. . . . . . . . . . . . . . . . 4361III. QWIP’s VERSUS OTHER TYPES OF

INFRARED PHOTODETECTORS. . . . . . . . . . . . 4362A. a/Gth ratio. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4364

4350021-8979/2003/93(8)/4355/37/$20.00

Downloaded 02 Mar 2006 to 148.81.117.137. Redistribution subject to AI

B. Quantum efficiency. . . . . . . . . . . . . . . . . . . . . . .4365C. Dark current andR0A product. . . . . . . . . . . . . . . 4366D. Detectivity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4369E. Background-limited performance. . . . . . . . . . . . . 4370

IV. FOCAL PLANE ARRAY PERFORMANCE. . . . . 4371A. Uniformity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4371B. NEDT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4372C. Charge handling capacity and integration time.. 43D. Cost. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4377E. Reliability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4377

V. MULTIBAND FPA’s . . . . . . . . . . . . . . . . . . . . . . . . .4378A. Dual-band HgCdTe. . . . . . . . . . . . . . . . . . . . . . . .4378

5 © 2003 American Institute of Physics

P license or copyright, see http://jap.aip.org/jap/copyright.jsp

nomt

caic

isdeprre

tpeveca

T

toveas

haiGSbtotrveta

fathve

o-min-d

vol-ve-aveece-er/

sorsatulkhasuchryo-en-and

P’syer

asho-ton

s toes.eredx-

ofsuchloron is

attors

0 K.

4356 J. Appl. Phys., Vol. 93, No. 8, 15 April 2003 A. Rogalski

B. Multiband QWIP’s. . . . . . . . . . . . . . . . . . . . . . . .4380VI. QWIP APPLICATIONS. . . . . . . . . . . . . . . . . . . . . .4383

A. Medicine. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4383B. Astronomy.. . . . . . . . . . . . . . . . . . . . . . . . . . . . .4384C. Defense. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4384D. High-frequency detection. . . . . . . . . . . . . . . . . . . 4385E. Market data.. . . . . . . . . . . . . . . . . . . . . . . . . . . . .4386

VII. CONCLUSIONS. . . . . . . . . . . . . . . . . . . . . . . . . . .4386APPENDIX. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4387

1. Noise equivalent difference temperature. .. . . . . 4387a. Photon detectors. . . . . . . . . . . . . . . . . . . . . . .4388b. Thermal detectors. . . . . . . . . . . . . . . . . . . . . .4388

2. GaAs/AlGaAs QWIP’s. . . . . . . . . . . . . . . . . . . . .4388a. Thermal generation. . . . . . . . . . . . . . . . . . . . .4388b. Quantum efficiency. . . . . . . . . . . . . . . . . . . . 4389

3. HgCdTe photodiodes. . . . . . . . . . . . . . . . . . . . . .43894. InSb photodiodes. . . . . . . . . . . . . . . . . . . . . . . . .43895. Extrinsic photoconductors. . . . . . . . . . . . . . . . . . 4389

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4390

I. INTRODUCTION

Since the initial proposal by Esaki and Tsu1 and the ad-vent of molecular beam epitaxy~MBE!, interest in semicon-ductor superlattices~SL’s! and quantum well~QW! structureshas increased continuously over the years, driven by techlogical challenges and new physical concepts and phenena as well as promising applications. A new class of marials and heterojunctions with unique electronic and optiproperties has been developed. Here we focus on devwhich involve infrared~IR! excitation of carriers in quantumwells. A distinguishing feature of QW infrared detectorsthat they can be implemented in chemically stable wiband-gap materials as a result of the use of intrabandcesses. Among the different types of quantum well infraphotodetectors~QWIP’s!, the technology of GaAs/AlGaAsmultiple-quantum-well ~MQW! detectors is the mosmature.2 Rapid progress has been made recently in theformance of these detectors. Detectivities have improdramatically and they are now high enough for the fabrition of large~e.g., 6403486 and 6403512) focal plane ar-rays~FPA’s! with a long-wavelength IR~LWIR! imaging per-formance comparable to that of state-of-the-art HgCddetectors.3–5

At present, HgCdTe is a variable-gap semiconducmost often used in the production of IR photodetectors. Othe last 40 years it has successfully fought off major chlenges from extrinsic silicon and lead-tin telluride deviceDespite that, however, it has more competitors today tever before. These include Schottky barriers on silicon, Sheterojunctions, AlGaAs multiple quantum wells, GaInstrain layer superlattices, high-temperature superconducand especially two types of thermal detectors: pyroelecdetectors and silicon bolometers. It is interesting, howethat none of these is competitive in terms of fundamenproperties. They may promise to be more easily manutured, but never to provide higher performance or, withexception of thermal detectors, to operate at higher or ecomparable temperatures.


o--

e-les

-o-d

r-d-

e

rr

l-.ne

rs,icr,l

c-en

The main motivation to replace HgCdTe is the technlogical difficulties associated with this material. One of theis the weak Hg–Te bond resulting in bulk, surface, andterface instabilities. Uniformity and yield are still unresolveissues. Slow progress in the development of large phototaic HgCdTe infrared imaging arrays and the rapid achiements of novel semiconductor heterostructure systems hmade it more difficult to predict what types of arrays will breadily available for future system applications. For spaborne surveillance systems, low-background IR seektracker systems as well as reliable and affordable senwith long life are needed which can function effectivelytemperatures higher than 20–30 K currently required by bphoton detectors. The only alternative to HgCdTe thatbeen available so far was extrinsic Si which operates at mlower temperatures where a problematic three-stage ccooler would be required. Improvement in surveillance ssors and interceptor seekers requires highly uniformmulticolor ~or multispectral! large-area IR FPA’s operating inthe LWIR and very-long-wavelength IR~VLWIR ! regions.Among the competing technologies there are the QWIbased on lattice matched GaAs/AlGaAs and strained laInGaAs/AlGaAs material systems.

In this paper we discuss the performance of QWIP’scompared to other types of IR detectors including both pton and thermal detectors. As the QWIP detector is a phodetector, more attention is paid in our comparative studiethis family of detectors, especially to HgCdTe photodiodInSb photodiodes and doped silicon detectors are considtoo. The potential performance of different materials is eamined by utilizing thea/G ratio, wherea is the absorptioncoefficient andG is the thermal generation rate.

The first part of the paper is devoted to different typesdetectors operating as single devices. In the second part,FPA issues as array size, uniformity, operability, multicocapability, and system prices are discussed. The discussi

FIG. 1. Comparison ofD* of various infrared detectors when operatedthe indicated temperature. Chopping frequency is 1000 Hz for all detecexcept the thermopile~10 Hz!, thermocouple~10 Hz!, thermistor bolometer~10 Hz!, Golay cell~10 Hz!, and pyroelectric detector~10 Hz!. Each detec-tor is assumed to view a surrounding hemisphere at a temperature of 30Theoretical curves for the background-limitedD* for ideal photovoltaic andphotoconductive detectors and thermal detectors are also shown.


4357J. Appl. Phys., Vol. 93, No. 8, 15 April 2003 A. Rogalski

TABLE I. Comparison of infrared detectors.

Detector type Advantages Disadvantages

Thermal~thermopile, bolometers, pyroelectric!Light, rugged, reliable, and low costRoom temperature operation

Low detectivity at high frequencySlow response~ms order!

Photon

Intrinsic

IV–VI~PbS, PbSe, PbSnTe!

Easier to prepareMore stable materials

Very high thermalexpansion coefficientLarge permittivity

II–VI ~HgCdTe! Easy band gap tailoringWell developed theory & expt.Multicolor detectors

Nonuniformity over large areaHigh cost in growth and processingSurface instability

III–V ~InGaAs, InAs, InSb, InAsSb! Good material and dopantsAdvanced technologyPossible monolithic integration

Heteroepitaxy with largelattice mismatchLong wavelength cutoff limitedto 7 mm ~at 77 K!

Extrinsic~Si:Ga, Si:As, Ge:Cu, Ge:Hg!

Very-long-wavelength operationRelatively simple technology

High thermal generationExtremely low-temperature operation

Free carriers~PtSi, Pt2Si, IrSi!

Low-cost, high yieldsLarge and close-packed 2D arrays

Low quantum efficiencyLow-temperature operation

Quantumwells

Type I~GaAs/AlGaAs, InGaAs/AlGaAs!!

Matured material growthGood uniformity over large areaMulticolor detectors

High thermal generationComplicated design and growth

Type II~InAs/InGaSb, InAs/InAsSb!

Low Auger recombination rateEasy wavelength control

Complicated design and growthSensitive to the interfaces

Quantum dots InAs/GaAs, InGaAs/InGaP,Ge/Si

Normal incidence of lightLow thermal generation

Complicated design and growth

Itrth

IRtw

ow

arin

edstctn

bec

ritth

otleec.nx-

her-ledorepho-sive,

ofasrs,

ingeretaic,hion.tive

malab-the

mainly concentrated on two competitive technologies, QWand HgCdTe. In such a way, the paper attempts to settlemain issues of both technologies. To present the matemore clearly for readers, many details are included inAppendix.

II. FUNDAMENTAL LIMITS TO INFRARED DETECTORPERFORMANCE

Spectral detectivity curves for a number of availabledetectors are shown in Fig. 1. The wavelengths of theatmospheric windows, 3–5mm @medium-wavelength IR~MWIR!# and 8–14mm ~LWIR!, are the most interesting~atmospheric transmission is the highest in these windand the room-temperature emissivity reaches its maximuml'10 mm!. It should be noted, however, that in recent yeinterest in longer wavelengths has been steadily increasstimulated by space applications.

Depending on the detection mechanisms, the naturthe interaction and material properties, various types oftectors have different characteristics. These characteriresult in advantages and disadvantages when the deteare used in field applications.6–9 Table I shows a comparisoof various IR detectors.

In the class of photon detectors, radiation is absorwithin the material by means of interactions with free eletrons or those bound either to lattice atoms or to impuatoms. The observed electrical output signal results fromchanged electron energy distribution. The response of phdetectors per unit incident radiation power depends setively on wavelength. These detectors exhibit both perfsignal-to-noise performance and a very fast responseachieve this, photon detectors require, however, cryogecooling. Photon detectors with long-wavelength limits eceeding approximately 3mm are usually cooled. This is nec


Pheiale

o

satsg,

ofe-icsors

d-ye

onc-t

Toic-

essary to prevent thermal generation of charge carriers. Tmal transitions compete with optical ones, making noncoodevices very noisy. Cooling is the main obstacle to a mwidespread use of IR systems based on semiconductortodetectors because it makes them bulky, heavy, expenand inconvenient to use.

Depending on the nature of the interaction, the classphoton detectors is further sub-divided into different typesshown in Table I. The most important are intrinsic detectoextrinsic detectors, photoemissive~metal silicide Schottkybarriers! detectors, and quantum well detectors. Dependon how the electric or magnetic fields are developed, thare various modes such as photoconductive, photovolphotoelectromagnetic~PEM!, and photoemissive ones. Eacmaterial system can be used in different modes of operatIn this paper, we focus on photodiodes and photoconducdetectors.

The second class of IR detectors is composed of therdetectors. In a thermal detector the incident radiation issorbed to change the temperature of the material, and

FIG. 2. Model of a photodetector.


do

die

anditlypln

e

ectoha

rou

uirftede

u

thto

er

hste

r-eo

ittoteoe

hemtio

es,-

gand

theingerootans

de-theber

ndnd

re-n-

es

f

alfter-g-


resultant change in a certain physical property is usedgenerate an electrical output. The detector is suspendelegs which are connected to the heat sink. The signal dnot depend upon the photonic nature of the incident ration. Thus, thermal effects are generally wavelength indepdent; the signal depends upon the radiant power~or its rate ofchange! but not upon its spectral content. This would methat the mechanism responsible for the absorption of ration is itself wavelength independent, which is not strictrue in most cases. Three approaches are most widely apin infrared technology: namely, bolometers, pyroelectric, athermoelectric effects. In pyroelectric detectors, a changthe internal electrical polarization is measured, whereasthermistor bolometers a change in the electrical resistancmeasured. In contrast to photon detectors, thermal detetypically operate at room temperature. They are usually cacterized by modest sensitivity and slow response~becauseheating and cooling of the detector is a relatively slow pcess!, but they are cheap and easy to use. They have fowidespread use in low cost applications that do not reqhigh performance and speed. Being unselective, they arequently used in IR spectrometers. Uncooled FPA’s fabricacurrently from thermal detectors have revolutionized thevelopment of thermal imagers.10,11

A. General theory of photon detectors

A photodetector is a slab of homogeneous semicondtor with an actual ‘‘electrical’’ areaAe . It is coupled to abeam of infrared radiation by its optical areaAo ~Fig. 2!.Usually, the optical and electrical areas of the device aresame or close to each other. The use of optical concentramay increase theAo /Ae ratio.

The current responsivity of a photodetector is detmined by the quantum efficiencyh and by the photoelectricgaing. The quantum efficiency value describes how well tdetector is coupled to the radiation to be detected. It is ually defined as the number of electron-hole pairs generaper incident photon. The idea of photoconductive gaing wasput forth by Rose12 as a simplifying concept for the undestanding of photoconductive phenomena and is now widused in the field. The photoelectric gain is the numbercarriers crossing the contacts per generated pair anddefined simply as the ratio of the photoelectron lifetimethe transit time. This value shows how well the generaelectron-hole pairs are used to generate the current respof a photodetector. Bothh and g are assumed here to bconstant over the volume of the device.

The spectral current responsivity is equal to

Ri5lh

hcqg, ~1!

wherel is the wavelength,h is Planck’s constant,c is thelight velocity, andq is the electron charge. Assuming that tcurrent gain for photocurrent and noise current is the sathe current noise due to the generation and recombinaprocesses is12

I n252~G1R!AetD f q2g2, ~2!


toonesa-n-

a-

ieddininisrsr-

-nde

re-d-

c-

ers

-

eu-d

lyfis

dnse

e,n

whereG and R are the generation and recombination ratDf is the frequency band, andt is the thickness of the detector.

The detectivityD* is the main parameter characterizinthe normalized signal-to-noise performance of detectorscan be defined as

D* 5Ri~AoD f !1/2

I n. ~3!

According to Eqs.~1!–~3!,13

D* 5l

hc S Ao

AeD 1/2

h@2~G1R!t#21/2. ~4!

For a given wavelength and operating temperature,highest performance can be obtained by maximizh/@ t(G1R)#1/2 which corresponds to the condition of thhighest ratio of the sheet optical generation to the squareof sheet thermal generation and recombination. This methat high quantum efficiency must be obtained in a thinvice. For a given wavelength and operating temperature,performance may be increased by reducing the total numof generation and recombination acts which is (G1R)3(Aet).

In further considerations we putAo /Ae51. Assuming asingle pass of the radiation and negligible frontside abackside reflection coefficients, the quantum efficiency adetectivity are

h512exp~2at !, ~5!

D* 5l

hc~12e2at!@2~G1R!t#21/2, ~6!

wherea is the absorption coefficient.The highest detectivity is obtained fort51.26/a. For

this condition (12eat)t21/2 reaches its maximum of 0.62a1/2. This thickness is the best compromise between thequirements of high quantum efficiency and low thermal geeration. In this optimum caseh50.716 and detectivity isequal to13

D* 50.45l

hc S a

G1RD 1/2

. ~7!

At equilibrium, the generation and recombination ratare equal to each other, and we have

D* 50.31l

hc S a

GD 1/2

. ~8!

Considerations presented in Ref. 13 indicate

D* 50.31l

hckS a

GD 1/2

, ~9!

where 1<k<2, andk is dependent on the contribution orecombination and backside reflection.

The ratio of the absorption coefficient to the thermgeneration rate,a/G, is the fundamental figure of merit oany material intended for infrared photodetectors. It demines directly the detectivity limits of the devices. This fiure should be used to assess any potential material.


he

etgioethu

ca

ope

.

ro

l

rio

n

uathce

syan

lr

ervec

s

-r-

rs

m

erto-es.y

dia-re-owle.a

d,

inionund

itssrce.

at,in-y anyctriclesthel

willstheua-


An optimized photodetector should consist of tfollowing:14

~i! A lightly doped active~base! region, which acts as anabsorber of IR radiation. Its band gap, doping, and geomshould be precisely selected. The surface of the active remust be insulated from the ambient by a material that dnot contribute to the generation of carriers; in addition,carriers which are optically generated in the absorber shobe kept away from the surfaces where recombinationreduce the quantum efficiency.

~ii ! Electric contacts to the base region which sensetically generated charge carriers and should not contributthe dark current of the device,

~iii ! A backside mirror for double pass of IR radiation

The above conditions may be fulfilled by using hetejunctions like N1 –p–p1 and P1 –n–n1 with heavily dopedcontact regions~symbol ‘‘1’’ denotes strong doping; capitaletter, wider gap!. Homojunction devices~like n–p, n1 –p,p1 –n) suffer from surface problems; excess thermal genetion results in increased dark current and recombinatwhich in turn, reduce the photocurrent.

The total generation rate is the sum of the optical athermal generation rates

G5Gth1Gop . ~10!

To achieve high performance, thermal generation mbe suppressed to the lowest level possible. This is usudone with cryogenic cooling. For practical purposes,ideal situation occurs when thermal generation is redubelow optical generation. The requirements for the thermgeneration rate may be highly reduced in heterodynetems, in which optical excitation by the local oscillator cdominate the generation, even in conditions of high thermgeneration. Optical generation may be due to the signabackground radiation. For infrared detectors, backgrounddiation is usually higher compared to signal radiation. If thmal generation is reduced much below the background lethe performance of the device is determined by the baground radiation~BLIP conditions for background limitedinfrared photodetector!. This condition may be described a

hFBt

t.nth , ~11!

wherenth is the density of thermal carriers at temperatureT,t is the carrier lifetime, andFB is the total background photon flux density~unit cm22 s1) reaching the detector. Rearanging Eq.~11! we obtain the BLIP requirements

Gop

5hFB

t.

nth

t5Gth ; ~12!

i.e., the photon generation rate needs to be greater thanthermal generation rate. Either majority or minority carriecan be considered.

The direct band gap semiconductor photodiode is anority carrier device and in thermal equilibrium

nmin5ni

2

nma j, ~13!


ryons

eldn

-to

-

a-n,

d

stllyed

als-

alora--l,

k-

the

i-

whereni is the intrinsic carrier concentration andnma j is themajority carrier concentration. The ultimate limit on carrilifetime in a direct gap semiconductor is given by band-band recombination, by either radiative or Auger processHumphreys indicated15 that the van Roosbroeck–Shockletheory of radiative recombination underestimates the rative lifetime due to noiseless photon reabsorption. As asult, Auger recombination is the dominant process in narrgap semiconductors like HgCdTe ternary alloy, for examp

The extrinsic semiconductor photoconductor is strictlymajority carrier device.

The background limited detectivity or, the so-calle‘‘photovoltaic’’ BLIP detectivity is given by16,17

DBLIP* 5l

hc S h

2FBD 1/2

. ~14!

DBLIP* for photoconductors isA2 times lower that for photo-diodes. This is attributable to the recombination processphotoconductors which is uncorrelated with the generatprocess contributing to the detector noise. The backgrophoton flux density received by the detector depends onangular field of view~FOV! of the background and on itability to respond to the wavelengths contained in the souPlots ofDBLIP* as a function of wavelength forTBLIP5300 Kand for full 2p FOV are shown in Fig. 1.

B. General theory of thermal detectors

Thermal detectors operate on the simple principle thwhen heated by incoming IR radiation, their temperaturecreases and the temperature changes are measured btemperature-dependent mechanism such as thermoelevoltage, resistance, and pyroelectric voltage. The simprepresentation of the thermal detector is shown in Fig. 3. Tdetector with the emissivitye is represented by a thermacapacitanceCth coupled via thermal conductanceGth to aheat sink at a constant temperatureT. In the absence of aradiation input, the average temperature of the detectoralso beT, although it will exhibit fluctuations around thivalue. When a radiation input is received by the detector,temperature rise is found by solving the heat balance eq

FIG. 3. Thermal detector mounted via legs to heat sink.


ne

e-

to

teile

ia

ct

incaeenth

-ta

fd

heateun

urds

on

tu

mat

nd

alise,

tothe

ave

o

ex-

Init,

-theandat

ture.98

ec-


tion. Assuming the radiant power to be a periodic functiothe change in the temperature of any thermal detector duthe incident radiative flux is16,18

DT5eFo

~Gth2 1v2Cth

2 !1/2. ~15!

Equation~15! illustrates several features of thermal dtectors. Clearly it is advantageous to makeDT as large aspossible. To do this, the thermal capacity of the detec(Cth) and its thermal coupling to its surroundings (Gth) mustbe as small as possible. The interaction of the thermal detor with the incident radiation should be optimized whreducing, as far as possible, all other thermal contacts wits surroundings. This means that a small detector massfine connecting contacts to the heat sink are desirable.

The characteristic thermal response time of a detemay be defined as

t th5Cth

Gth5CthRth , ~16!

whereRth51/Gth is the thermal resistance.The values of the thermal time constant are typically

the millisecond range. This is much longer than the typitime of a photon detector. There is a trade-off between ssitivity DT and frequency response. If one wants a high ssitivity, a low-frequency response is then forced upondetector.

For further discussion we introduce the coefficientK5DV/DT that reflects the efficiency with which the temperature changes translate into the electrical output volof a detector.

The voltage responsivityRv of the detector is the ratio othe output signal voltageDV to the input radiation power anis given by

Rv5KeRth

~11v2t th2 !1/2

. ~17!

In order to determine the detectivity of the detector, tnoise mechanism must be known first. The Johnson noisone of the major noises. Two other fundamental noisesimportant for assessing the ultimate performance of a detor, namely the thermal fluctuation noise and the backgrofluctuation noise.

The thermal fluctuation noise arises from temperatfluctuations in the detector. These fluctuations are causeheat conductance variations between the detector and therounding substrate with which the detector is in thermal ctact.

The spectral noise voltage due to temperature fluctions is16,18

Vth2 5

4kT2D f

11v2t th2

K2Rth . ~18!

The third noise is the background noise resulting froradiative heat exchange between the detector at temper


,to

r

c-

thnd

or

ln--

e

ge

eisrec-d

ebyur--

a-

ure

Td and the surrounding environment at temperatureTb . It isthe ultimate limit of a detector’s performance capability afor a 2p FOV is given by16,18

Vb25

8kesA~Td21Tb

2!

11v2t th2

K2Rth2 , ~19!

wheres is the Stefan–Boltzmann constant.The fundamental limit of the sensitivity of any therm

detector is determined by the temperature fluctuation noi.e., random fluctuations of the detector temperature duethe fluctuations of the radiant power exchange betweendetector and its surroundings. Under this condition we hat low frequencies (v!1/t th)

Dth* 5S e2A

4kTd2Gth

D 1/2

. ~20!

It is assumed here thate is independent of the wavelength, sthat the spectralDl* and blackbodyD* ~T! values are iden-tical.

If the radiant power exchange is the dominant heatchange mechanism, thenG is the first derivative of theStefan–Boltzmann function with respect to temperature.that case, known as the background fluctuation noise limwe have

Db* 5F e

8ks~Td51Tb

5!G 1/2

. ~21!

Note thatDb* is, as expected, independent ofA.Equations~20! and ~21! assume that the detector re

ceives the background radiation from all directions whentemperature of the detector and background are equal,from the forward hemisphere only when the detector iscryogenic temperatures. The highest possibleD* to be ex-pected for a thermal detector operating at room temperaand viewing a background at room temperature is 1

FIG. 4. Theoretical performance limits of LWIR photon and thermal dettors at zero background and background of 1017 photons/cm2 s, as a functionof detector temperature~after Ref. 19!.


-ulsicedlt

b

r

ga.a

io

e

tote

niorsarec

istoIR

retod.o

or

antongra

ngtsceme

a

hi-es

e

qual

tec-

per

ors

vitycu-ym-

ednte-

Te


31010 cm Hz1/2W21. Even if either the detector or background were cooled to absolute zero, the detectivity woimprove only by the square root of two. This is the balimitation of all thermal detectors. Photon detectors limitby background noise have higher detectivities as a resutheir limited spectral responses.

The performance achieved by any real detector willinferior to that predicted by Eq.~21!. The degradation ofperformance will arise from~i! encapsulation of the detecto~reflection and absorption losses at the window!, ~ii ! effectsof excess thermal conductance~influence of electrical con-tacts, conduction through the supports, influence of anyconduction and convection!, or ~iii ! additional noise sources

Typical values of the detectivities of thermal detectors10 Hz range from 108 to 109 cm Hz1/2 W21.

C. Comparison of the fundamental limits of thermaland photon detectors

The temperature dependence of the fundamental limof D* of photon and thermal detectors at different levelsbackground are shown in Fig. 4. As seen the performancintrinsic IR detectors~HgCdTe photodiodes! in the LWIRspectral range is higher than that of other types of phodetectors. HgCdTe photodiodes with a background limiperformance operate at temperatures below'80 K. HgCdTeis characterized by a high optical absorption coefficient aquantum efficiency and a relatively low thermal generatrate compared to QWIP’s and extrinsic silicon detectoBoth QWIP and silicon detectors are extrinsic in their chacter and require more cooling than intrinsic photon dettors having the same long-wavelength limit.

The ultimate detectivity value for thermal detectorsmuch less temperature dependent than for photon detecAt temperatures below 50 K and zero background, LWthermal detectors are characterized byD* values lower thatthose of LWIR photon detectors. However, at temperatuabove 60 K, thermal detectors perform better than phoones. At room temperature, the performance of thermaltectors is much better than that of LWIR photon detectorsis due to the influence of the fundamentally different typesnoise ~generation-recombination noise in photon detectand temperature fluctuation noise in thermal detectors! thatthe dependences of the detectivity on the wavelengthtemperature are different in these two classes of detecPhoton detectors exhibit better performance at lowavelength infrared radiation and lower operating tempetures. Thermal detectors perform better in the very-lowavelength spectral range. The temperature requiremenattain the background-fluctuation-noise-limited performanin general, favor thermal detectors at higher cryogenic teperatures and photon detectors at lower cryogenic temptures.

The above relations, which are shown in Fig. 4 forbackground of 1017 photons/cm2 s, are modified by the influ-ence of the background. It is interesting to note that in tcase the theoretical curves ofD* for photon and thermal detectors show similar fundamental limits at low temperatur


d

of

e

s:

t

tsfof

nd

dn.--

rs.

sne-Itfs

drs.---to,-

ra-

s

.

The relevant figure of merit for FPA’s is the noisequivalent difference temperature~NEDT!, which is the tem-perature change of a scene required to produce a signal eto the rms noise~see Appendix A1!. It is of interest to com-pare the performance of uncooled photon and thermal detors in the MWIR (l55 mm! and LWIR (l510 mm! spec-tral range. We have followed the procedure used in a parecently published by Kinch.20

Figure 5 compares the theoretical NEDT of detectoperating at 290 K for F/1 optics and a 1 mil pixel size.Parameter values typical of micromachined resonant cabolometers given in Appendix A1 are assumed in the callations. A N1 –p – P1 HgCdTe photodiode, first proposed bElliott and Ashley,21 is chosen as a photon detector. The sybol p designates an intrinsic region containing ap-type back-ground dopant concentration of 531014 cm23 with carrierlifetime limited by the Auger 7 process. It is also assumthat the detector node capacity is capable of storing the i

FIG. 5. Theoretical NEDT comparison of uncooled thermal and HgCduncooled photon LWIR~a! and MWIR ~b! detectors~after Ref. 20!.



Downloaded 02 Ma

TABLE II. Essential properties of LWIR HgCdTe photodiodes and QWIP’s atT577 K.

Parameter HgCdTe QWIP~n-type!

IR absorption Normal incidence Eoptical' plane of well requiredNormal incidence: no absorption

Quantum efficiency >70% <10%Spectral sensitivity Wide-band Narrow-band~FWHM '1–2 mm!Optical gain 1 0.4~30–50 wells!Thermal generation lifetime '1 ms '10 psR0A product (lc510 mm! 300 V cm2 104 V cm2

Detectivity (lc510 mm, FOV50! 231012 cm Hz1/2 W21 231010 cm Hz1/2 W21

no

ndAst

det o3

5foalorthnor

h asnshe

nes,

butateithIR0.8

tSi,tun-en-ort-

grated charge provided by the detector dark current.Figure 5 shows that the ultimate performance of u

cooled HgCdTe photon detectors is far superior to thatthermal detectors at wide frame rates and spectral baAlso any other tunable band gap alloy, such as type-II InGaInSb superlattices, could be useful in the developmenuncooled photon detector technology.6,22The ultimate perfor-mance of InAs/InGaSb strain layer superlattice photodiowith optimally doped base regions is comparable with thaHgCdTe photodiodes in the temperature range betweenK and 77 K.23

Comparing both curves for thermal detectors in Fig.we can see that for long integration times excellent permance can be achieved in the LWIR region with NEDT vues below 10 mK for frame rates of 30 Hz. However, fsnapshot systems with an integration time below 2 ms,available NEDT is above 100 mK even in the LWIR regioIn the MWIR band thermal detectors has obvious perfmance limitations at any frame rate.

r 2006 to 148.81.117.137. Redistribution subject to AI

-fs./

of

sf00

,r--

e.-

III. QWIP’s VERSUS OTHER TYPES OF INFRAREDPHOTODETECTORS

Current-cooled IR detector systems use materials sucHgCdTe, InSb, PtSi, and doped Si. As far as IR applicatioare concerned, QWIP is a relatively new technology. Tfirst GaAs/AlGaAs QWIP was demonstrated by Leviet al.24 in 1987. Among these cooled IR detector systemPtSi FPA’s are highly uniform and easy to manufacturethey have very low quantum efficiency and can only operin the MWIR range. The InSb FPA technology is mature wvery high sensitivity and it also can operate in the MWspectral range. Doped Si has a wide spectral range fromto 30 mm and operates only at very low temperatures. PInSb, and doped Si detectors do not have wavelengthability or multicolor capabilities. Both QWIP’s and HgCdToffer high sensitivity together with wavelength flexibility ithe MWIR, LWIR, and VLWIR regions, as well as multicolor capabilities. HgCdTe can also operate in the sh

TABLE III. HgCdTe photodiode architectures used for hybrid FPA’s~after Ref. 25!.

Configuration Junction formation Company

n-on-p VIP Ion implantation formsn-on-p diode inp-typeHgCdTe, grown by Te-solution LPE on CdZnTeand epoxied to silicon ROIC wafer; over the edgecontact

DRS Infrared Technologies~formerly Texas Instruments!

n-p loophole Ion beam milling formsn-type islands inp-typeHg-vacancy-doped layer grown by Te-solutionLPE on CdZnTe, and epoxied onto silicon ROICwafer; cylindrical lateral collection diodes

GEC-Marconi Infrared~GMIRL!

n1-on-p planar Ion implant into acceptor-dopedp-type LPE filmgrown by Te-solution slider

Sofradir ~Societe Francaise deDetecteurs Infrarouge!

n1-n2-p planar homojunctions Boron implant into Hg-vacancyp-type, grown byHg-solution tipper on 3’’ diam. sapphire withMOCVD CdTe buffer; ZnS passivation

Rockwell/Boeing

P-on-n mesa 1. Two-layer LPE on CdZnTe:Base: Te-solution slider, indium-dopedCap: Hg-solution dipper, arsenic-doped2. MOCVD in situ on CdZnTeIodine-doped base, arsenic-doped cap

IR Imaging Systems, Sanders– A Lockheed MartinCompany~LMIRIS!

P–on–n mesa 1. Two-layer LPE on CdZnTe or Si:Base: Hg-solution dipper, indium-dopedCap: Hg-solution dipper, arsenic-doped2. MBE in situ on CdZnTe or SiIndium-doped base, arsenic-doped cap

Raytheon Infrared Center ofExcellence~RIRCoE,formerly SBRC! and HughesResearch Laboratories~HRLs!

P–on–n planar buried heterostructure Arsenic implant into indium-doped N-n or N-n-Nfilm grown by MBE on CdZnTe

Rockwell/Boeing


d

eidr

thjo

f

hesortuna

thge

onne

tacththarteo

tesnd

tumingtondWn-andere-andthin

Aed

ho-hethe

ised.h ato

ndega-

erl

eal

emeticalgap

ur

ec-


wavelength IR~SWIR! region, while QWIP has to go to adirect band-to-band scheme for SWIR operation. Tablecompares the essential properties of HgCdTe and QWIPvices at 77 K.

Different HgCdTe photodiode architectures have befabricated that are compatible with backside and frontsilluminated hybrid FPA technology. The most important achitectures are included in Table III which summarizesapplications of HgCdTe photodiode designs by the maFPA manufactures today.

All QWIP’s are based on ‘‘band gap engineering’’ olayered structures of wide-band-gap~relative to thermal IRenergies! materials. These structures are designed in sucway that the energy separation between two selected statthe structure matches the energy of the infrared photons tdetected. Several QWIP configurations have been repobased on transitions from bound-to-extended states, boto-quasicontinuum states, bound-to-quasibound states,bound-to-miniband states.

Figure 6 shows two detector configurations used infabrication of two-color QWIP FPA’s. The major advantaof the bound-to-continuum QWIP@Fig. 6~a!# is that the pho-toelectron can escape from the quantum well to the ctinuum transport states without being required to tunthrough a barrier. As a result, the voltage bias requiredefficiently collect the photoelectrons can be reduced dramcally, thereby lowering the dark current. Furthermore, sinphotoelectrons do not have to tunnel through them,AlGaAs barriers can be made thicker without reducingphotoelectron collection efficiency. It appears that the dcurrent decreases significantly when the first excited stadropped from the continuum to the well top, bound-tquasibound QWIP, without sacrificing the responsivity.

FIG. 6. Band diagram of demonstrated QWIP structures:~a! bound-to-extended and~b! bound-to-miniband. Three mechanisms creating dark crent are also shown in~a!: ground-state sequential tunneling~1!, intermedi-ate thermally assisted tunneling~2!, and thermionic emission~3!. The grayshading indicates extended states through which current flows.


IIe-

ne-er

ain

beedd-nd

e

-loti-eeekis

-

A miniband transport QWIP contains two bound stawith the higher energy level being resonant with the groustate miniband in the superlattice barrier@see Fig. 6~b!#. Inthis approach, IR radiation is absorbed in the doped quanwells, exciting an electron into the miniband and transportit along the miniband until it is collected or recaptured inanother quantum well. Thus, the operation of this minibaQWIP is analogous to that of a weakly coupled MQbound-to-continuum QWIP. In this device structure, the cotinuum states above the barriers are replaced by a minibof superlattice barriers. The miniband QWIP’s show a lowphotoconductive gain than bound-to-continuum QWIP’s bcause photoexcited electron transport occurs in the minibwhere electrons have to be transported through manyheterobarriers resulting in a lower mobility.

The light-coupling scheme is a key factor in QWIP FPperformance. Different light-coupling mechanisms are usin QWIP’s. A distinct feature ofn-type QWIP’s is that theoptical absorption strength is proportional to an incident pton’s electric-field polarization component normal to tquantum wells. This implies that a photon propagating indirection normal to the quantum wells, whose polarizationentirely in the plane of the quantum wells, is not absorbTherefore, these detectors have to be illuminated throug45° polished facet. For imaging, it is necessary to be ablecouple light uniformly to two-dimensional~2D! arrays ofthese detectors, so a diffraction periodic grating26 or othersimilar structure~such as random reflectors,27 corrugatedreflectors,28 lattice mismatch strain material systems, ap-type materials29! is typically fabricated on one side of thdetector to redirect a normally incident photon into propation angles more favorable for absorption.

Kinch and Yariv30 presented the first investigation evof the fundamental physical limitations of individuamultiple-quantum-well IR detectors as compared to idHgCdTe detectors atlc58.3 mm and 10mm. The dark cur-rent in the QWIP is typically four to five orders of magnitudhigher than in a direct-gap semiconductor with the saband gap and operating at the same temperature. For tacbackground fluxes, this enables the operation of a direct-

-

FIG. 7. Absorption coefficient for various photodetector materials in sptral range of 1–14mm.


fe

hihent-

or

harayct

ffih

tioi

nst

dg

e-bea-

g are-ab-hs

ned

tem-The

d-

-roft at

ient


semiconductor at temperatures about 50° higher while ofing the same performance as the QWIP. This issue canalleviated somewhat by the use of resonant structures, wenable IR absorption for significantly thinner QWIP’s at texpense of band spectral response, but the improvememinimal. The dominant factor favoring HgCdTe in this comparison is the excess carrier lifetime, which forn-typeHgCdTe is above 1ms at 80 K, compared to about 10 ps fthe AlGaAs/GaAs superlattice. Several other papers20,31–34

have been devoted to this issue.

A. aÕGth ratio

The considerations presented in Sec. II A indicate tthe ratio of the absorption coefficient to the thermal genetion rate,a/Gth , is the fundamental figure of merit of anmaterial intended for IR photodetectors because it diredetermines the detectivity limits of the devices.

Figure 7 shows measured intrinsic absorption coecients for various narrow-gap photodetector materials. Tabsorption coefficient and the corresponding penetradepth vary among different semiconductor materials. Itwell known that the absorption curve for direct transitiobetween parabolic bands at a photon energy higher thanenergy gap,Eg , obeys a square-root law

a~hn!5b~hn2Eg!1/2, ~22!

whereb is a constant. As seen in Fig. 7, the absorption e

FIG. 8. Absorption coefficient spectra measured atT5300 K for differentQWIP samples described in Table III.


r-bech

is

t-

ly

-ens

he

e

value in the MWIR spectral region is between 23103 cm21

and 33103 cm21; in the LWIR region it is about 103 cm21.Sincea is a strong function of the wavelength, the wav

length range in which an appreciable photocurrent cangenerated is limited for a given semiconductor. Near the mterial’s band gap, there is tremendous variation causinthree order of magnitude variation in absorption. In thegion of the material’s maximum usable wavelength thesorption efficiency drops dramatically. For wavelengtlonger than the cutoff wavelength, the values ofa are toosmall to give appreciable absorption.

Figure 8 shows the infrared absorption spectra obtaifrom differentn-doped, 50-period GaAs/Alx Ga1-xAs QWIPstructures. The measurements were carried out at roomperature using a 45° multipass waveguide geometry.spectra of bound-to-bound continuum~B-C! QWIP’s~samples A, B, and C! are much broader than those of bounto-bound~B-B! ~sample E! or bound-to-quasibound~B-QB!QWIP’s ~sample F!. Correspondingly, the value of the absorption coefficient of B-C QWIP’s is significantly lowethan that of B-B QWIP’s. This is due to the conservationoscillator strength. The values of the absorption coefficien77 K, peak wavelengthlp , cutoff wavelengthlc , and spec-tral width Dl ~full width at half-ap) are given in Table IV. Itappears that the low-temperature absorption coefficap~77 K!'1.3ap~300 K! and ap ~Dl/l!/ND is a constant

FIG. 9. a/G ratio versus temperature for LWIR (lc510 mm! photon detec-tors.

TABLE IV. Structure parameters for differentn-doped, 50-period AlxGa12xAs QWIP structures~after Ref. 2!.

Sample

Wellwidth~Å!

Barrierwidth ~Å!

Compositionx

Dopingdensity

(1018 cm23)Intersubband

transitionlp

~mm!lc

~mm!Dl

~mm!Dl/l~%!

ap~77 K!(cm21)

ha~77 K!~%!

A 40 500 0.26 1.0 B–C 9.0 10.3 3.0 33 410 13B 40 500 0.25 1.6 B–C 9.7 10.9 2.9 30 670 19C 60 500 0.15 0.5 B–C 13.5 14.5 2.1 16 450 14E 50 500 0.26 0.42 B–B 8.6 9.0 0.75 9 1820 20F 45 500 0.30 0.5 B–QB 7.75 8.15 0.85 11 875 14


n00hatio

orn

isayili

th

sIRththfi-

ons

dTerre-tlyncyong.ge

e

e totralbeor-of

nd,ith

za-ffi-

vityheicalm

-d

geiasThisce.

ated

aley of

on

iod


(ND is the well’s doping!.2 Typical values of the absorptiocoefficient in the LWIR region are between 600 and 8cm21 at 77 K. Comparing Figs. 7 and 8 one may notice tthe absorption coefficients for direct band-to-band absorpare higher than those for intersubband transitions.

The a/G ratio versus temperature is shown in Fig. 9 fdifferent types of tunable materials with the hypothetical eergy gap equal to 0.124 eV (l510 mm!. The proceduresused in the calculations ofa/G for different materials aregiven in Appendixes A2–A5. It is apparent that HgCdTeby far the most efficient detector of IR radiation. One malso notice that QWIP is a better material than extrinsic scon. Figure 9 is complimented by Fig. 10, where thea/Gratio in different materials is presented as a function ofwavelength at 77 K.

B. Quantum efficiency

HgCdTe has a direct-band-gap energy that translatehigh radiative efficiency and may be tailored to absorbradiation at a wavelength of interest. When operating inphotovoltaic mode, the optical gain is close to unity andresponsivity is directly proportional to the quantum efciency @see Eq.~1!# of the photodiode. In this sectionh is

FIG. 10. a/G ratio versus cutoff wavelength for different types of photdetectors operated at 77 K.

FIG. 11. Quantum efficiency versus wavelength for a HgCdTe photodand GaAs/AlGaAs QWIP detector with similar cutoffs.


tn

-

-

e

to

ee

defined as the ratio of the number of photoexcited electrcollected by the readout integrated circuit~ROIC! to thenumber of photons incident at a detector surface. HgChas high optical absorption and a wide absorption band ispective of the polarization of the radiation, which greasimplifies the detector array design. The quantum efficieobtained routinely is around 70% without an antireflecti~AR! coating and is in excess of 90% with an AR coatinMoreover, it is independent of the wavelength over the ranfrom less than 1mm to near the cutoff of the detector. Thwide-band spectral sensitivity with a near-perfecth enablesgreater system collection efficiency and a smaller aperturbe obtained. This makes FPA’s useful for imaging, specradiometry, and long-range target acquisition. It shouldnoticed, however, that the current LWIR staring array perfmance is mostly limited by the charge handling capacitythe ROIC and the background~warm optics!.

Due to intersubband transitions in the conduction bathen-type QWIP detection mechanism requires photons wa non-normal angle of incidence to provide proper polarition for photon absorption. The absorption quantum eciency is relatively small~about 25%! with a 2D grating.Since QWIP is a photoconductive detector, the responsiis proportional to the conversion efficiency—that is, tproduct of the absorption quantum efficiency and the optgain. The optical gain of typical QWIP structures varies fro0.2 ~in the case of 30–50 wells! to more than 1. This indi-cates thath is typically below 10% at the maximum response, rapidly falling off in the direction of both short anlong wavelengths. Figure 11 compares the spectralh of aHgCdTe photodiode to that of a QWIP. A higher bias voltais used to boosth. However, an increase in the reverse bvoltage also causes an increase of the leakage current.limits any potential improvement in the system performanNew grating designs are under study to improveh, such asan enhanced QWIP, antenna gratings, and corruggratings.3,4,32

The strong dependence of the LWIR QWIP signal~and aweaker, but still significant dependence of the MWIR sign!on the polarization of the light is very important in futurreconnaissance systems, because it offers the capabilitdiscriminating between different polarization states.35 The

e

FIG. 12. Current-voltage characteristics at various temperatures for a 12mmcutoff HgCdTe photodiode~after Ref. 38!.


e

rethahaeine

llg

ntt

tiogb

etro

nceov-

eser-rep-datanc-

ur--ple-

e isme

ice,theific

qual-ethe

nel-by-

ter-oring

the

sted

r

s d

n of

odes


first polarization sensitive QWIP thermal imager has beassembled.36

It should be stressed that, so far, most efforts in the aof QWIP’s are focused on tactical applications, wherestructure designs and the doping are optimized to increthe operating temperature and suit the readout chargedling capacity. It is well known that using a smaller numbof quantum wells and bound-to-continuum structures,creased optical gain and improved detector performanclow temperatures are possible. Tidrow presented37 a high-performance QWIP consisting of only three quantum wewith the conversion efficiencies up to 29% at a bias voltaof 20.8 V and a peak wavelength of 8.5mm.

C. Dark current and R0A product

In ideal photodiodes the diffusion current is dominatherefore their leakage current is very low and insensitivethe detector bias. Leakage current is the primary contribuof unwanted noise. Figure 12 shows typical current-voltacharacteristics of a HgCdTe photodiode at temperaturestween 40 and 90 K for a 12mm cutoff detector at 40 K. Theleakage current is less than 1025 A/cm2 at 77 K. The bias-independent leakage current makes it easier to achieve bFPA uniformity, as well as to reduce the detector bias-conrequirements during changes in photocurrent.

FIG. 13. Dependence of theR0A product on the long wavelength cutoff foLWIR p–on–n HgCdTe photodiodes at temperatures<77 K. The solid linesare calculated on the assumption that the performance of photodiodes ito thermal generation governed by the Auger mechanism in the basen-typeregion of photodiodes with thicknesst510 mm and dopingNd5531014

cm23. The experimental values are taken from different papers~after Ref.39!.


n

aesen-

r-at

se

,onee-

tterl

The dependence of the base region diffusion limitedR0Aproduct on the long-wavelength cutoff forp-on-n HgCdTephotodiodes at temperatures<77 K is shown in Fig. 13. Thesolid lines are calculated by assuming that the performaof the photodiodes is determined by thermal generation gerned by the Auger mechanism in then-type base regionwith a thicknesst510 mm and dopingNd5531014 cm23.At 77 K, the highest experimental data are still about 5 timlower than the ultimate theoretical predictions. As the opating temperature of the photodiodes is lowered, the discancy between the theoretical curves and experimentalincreases, which is due to additional currents in the jutions.

Many additional excess mechanisms affect the dark crent of HgCdTe photodiodes.8 They arise from nonfundamental sources located in the base and cap layer, the detion region and the surface. As the operating temperaturlowered, the thermal dark current mechanisms becoweaker and allow other mechanisms to prevail. In practnonfundamental sources dominate the dark current ofpresent HgCdTe photodiodes, with the exception of speccases of near room temperature devices and the highestity 77 K LWIR and 200 K MWIR devices. The main leakagmechanisms of HgCdTe photodiodes are: generation indepletion region, interband tunneling, trap-assisted tuning, and impact ionization. Some of them are causedstructural defects in thep-n junction. These mechanisms receive much attention now, particularly because they demine ultimately the array uniformity and yield and cost fsome applications, especially those with lower operattemperatures.

In Hg12xCdx Te photodiodes withx'0.22, the diffusioncurrent in the zero-bias and low-bias regions is usuallydominant current down to 60 K.39,40 At medium values ofreverse bias, the dark current is mostly due to trap-assi

ue

FIG. 14. Detailed analysis separates the cumulative distribution functioR0A values of LWIRp–on–n HgCdTe photodiodes~fabricated by LPE! intothree regions: good diodes, diodes affected by point defects, and diaffected by metallurgical defects~after Ref. 42!.


reiara

unveat

et-

onths

iliag

alu

nt,a-

ndof

ca-

; theowerch isn

or

in-inre-

hesur-ro-iontheling

byar-thee

ior16ra-

dTta,V;or

ATThe


tunneling. Trap-assisted tunneling dominates the dark curalso at zero bias below 50 K. At high values of reverse bbulk band-to-band tunneling dominates. At low tempetures, such as 40 K, significant spreads in theR0A productdistributions are typically observed due to the onset of tneling currents associated with localized defects. MoreoHgCdTe photodiodes often have an additional surface-relcomponent of the dark current,8 particularly at low tempera-tures.

Chenet al.42 carried out a detailed analysis of the widdistribution of theR0 values of HgCdTe photodiodes operaing at 40 K. Figure 14 shows the cumulative distributifunction R0 obtained in devices with a cutoff wavelengbetween 9.4 and 10.5mm. It is clear that while some deviceexhibit a fair operability withR0 values spanning only twoorders of magnitude, other devices show a poor operabwith R0 values spanning more than five to six orders of mnitude. Lower performance, withR0 values below 73106 Vat 40 K, is usually due to gross metallurgical defects, suchdislocation clusters and loops, pin holes, striations, Te incsions, and heavy terracing. However, diodes withR0 values

FIG. 15. Influence of the dislocation density on the parameters of HgCphotodiodes:~a! R0A product versus EPD, showing the fit of model to dafor a 9.5-mm array~at 78 K!, measured at 120, 77, and 40 K at zero FO~b! 1/f noise current at 1 Hz vs dislocation density measured at 78 K f10.3-mm HgCdTe photodiode array~f/2 FOV! ~after Ref. 41!.


nts,-

-r,

ed

ty-

s-

between 73106 V and 13109 V at 40 K contained no vis-ible defects~Hg interstitials and vacancies!. Dislocations,twins and subgrain boundaries in LWIRn1-p HgCdTe pho-todiodes primarily produce bias dependent dark currewhile Te precipitation and associated dislocation multipliction produces bias dependent noise.43

Dislocations are known to increase the dark current athe 1/f noise current. The reverse bias characteristicsHgCdTe diodes depend strongly on the density of dislotions intercepting the junction. Johnsonet al.41 showed thatin the presence of high dislocations densities theR0A prod-uct decreases as the square of the dislocation densityonset of the square dependence occurs at progressively ldislocation densities as the temperature decreases, whishown in Fig. 15~a!. At 77 K, R0A begins to decrease at aetch-pit density~EPD! of approximately 106 cm22, while at40 K R0A is immediately affected by the presence of onemore dislocations in the diode. The scatter in theR0A data atlarge EPD may be associated with the presence of ancreased number of pairs of ‘‘interacting’’ dislocationssome of those diodes; these pairs are more effective inducing theR0A than individual dislocations.

In general, the 1/f noise appears to be associated with tpresence of potential barriers at the contacts, interior, orface of the semiconductor. Different models have been pposed to explain experimental data including the modulatof the surface generation current by the fluctuations ofsurface potential and the influence of trap-assisted tunneacross a pinched-off depletion region.8 Johnsonet al.41 pre-sented the effect of dislocations on the 1/f noise. Figure 15~b!shows that, at low EPD, the noise current is dominatedthe photocurrent, while at higher EPD the noise current vies linearly with EPD. It appears that dislocations are notdirect source of the 1/f noise, but rather increase this noisonly through their effect on the leakage current. The 1/f noisecurrent varies asI 0.76 ~where I is the total diode current!;similar to the fit of data taken on undamaged diodes.

The average value of theR0A product at 77 K for a10-mm cutoff HgCdTe photodiode at 77 K is around 300Vcm2 and drops to 30V cm2 at 12mm.8,25,44–46At 40 K, theR0A product varies between 105 and 108 V cm2 with 90%above 105 V cm2 at 11.2mm.

In comparison with HgCdTe photodiodes, the behavof dark current of QWIP’s is understood better. Figureshows the current mechanisms in a QWIP. At low tempe

e

a

FIG. 16. The current mechanisms of QWIP: DT is the direct tunneling, Tis the thermally assisted tunneling, and TE is the thermionic emission.dashed lines show the photocurrent mechanism.


o

thncia

rtents

i

defir

rent

edias

atV

asterut

g

thature

reurefor aithm-

Pnic

eers.3For

n-lkan-rityronsd aIP

ur-for

uresilylowivetheigh, actor-rs,ub-etel-

tact

ek-

a


tures (T,40 K, lc510 mm!, the dark current is mostlycaused by defect-related direct tunneling. In the mediumerating range between 40 and 70 K (lc510 mm!, thermallyassisted tunneling dominates. In this case, electrons aremally excited and tunnel through the barriers with assistafrom defects and the triangle part of the barrier at high bAt temperatures higher than 70 K (lc510 mm!, thermallyexcited electrons are thermionically emitted and transpoabove the barriers. It is difficult to block this dark currewithout sacrificing the photoelectrons~transport mechanismof thermionically emitted current and photocurrent are simlar!. Minimizing thermionically emitted current is critical tothe commercial success of the QWIP as it allows highlysirable high-temperature camera operation. Dropping the

FIG. 17. Current-voltage characteristics of a QWIP detector having a presponse of 9.6mm at various temperatures, along with the 300 K bacground window current measured at 30 K with an 180° FOV~after Ref. 48!.

FIG. 18. Current density versus temperature for a HgCdTe photodiodea GaAs/AlGaAs QWIP withlc510 mm ~after Ref. 49!.


p-

er-e

s.

d

-

-st

excited state to the top theoretically causes the dark curto drop by a factor of'6 at 70 K. This compares well withthe fourfold drop experimentally observed for 9-mm cutoffQWIP’s.47

The value of the QWIP dark current could be adjustusing different device structures, doping densities, and bconditions. Figure 17 shows theI–V characteristics for tem-peratures ranging from 35 to 77 K, measured in a devicethe 9.6-mm spectral peak. It shows typical operation at 2applied bias in the slowly varying region of current with bibetween the initial rise in current at low voltage and the larise at high bias. Typical LWIR QWIP dark current is abo1024 A/cm2 at 77 K. Thus, the same current in a 24324mm2 pixel47 will be in the nanoampere range. ComparinFigs. 12 and 17 we can see that a 9.6-mm QWIP must becooled to 60 K to have a leakage current comparable toof a 12-mm HgCdTe photodiode operating at the temperathigher by 25°.

Additional insight into the difference in the temperatudependence of the dark currents is given by Fig. 18. Fig18 shows the current density versus inverse temperatureGaAs/AlGaAs QWIP and a HgCdTe photodiode, both wlc510 mm. The current densities of both detectors at teperatures lower than 40 K are limited by tunneling~tempera-ture independent!. However, the current density of the QWIis lower than that of the HgCdTe photodiode. The thermioemission regime for the QWIP~>40 K! is highly tempera-ture dependent, and ‘‘cuts on’’ very rapidly. At 77 K, thQWIP has a dark current which is approximately two ordof magnitude higher than that of the HgCdTe photodiode

A QWIP typically operates at a bias voltage from 1 toV depending on the structure and the number of periods.the voltage divided by the dark current density, theR0Aproducts are usually larger than 107 V cm2 and 104 V cm2

when operated at 40 K and 77 K, respectively.32 These val-ues indicate a very high impedance.

In low-background and low-temperature operating coditions, a QWIP operates very similarly to extrinsic buphotoconductors. Electrons in the subbands of isolated qutum wells can be visualized as electrons attached to impustates in bulk semiconductors. The photogenerated electdepart from the doped, active QW region and leave behinspace-charge buildup. Thus, in order to operate a QWsteadily, it requires a sufficient dark or background photocrent to replenish the depleted QW’s. This is not an issueQWIP’s operating at high background or high temperatconditions since refilling the space-charge buildup is eaprovided. However, at a low operating temperature and abackground, the high resistivity of the barriers in the actregion impedes electrons from entering the detector fromopposite electrode. As a result, a lower responsivity at hoptical modulation is observed. To overcome this problemnew detector structure, the blocked intersubband dete~BID! has been proposed.50 In this structure, the active region of BID’s consists of QW’s separated by thin barriewhich creates subminibands due to the large overlap of slevel wave functions. Thus, space-charge buildup will gquickly refilled by electrons via sequential resonant tunning along the ground-state minband from the emitter con

ak

nd


into

ionframh

thu

nf othw

veeberrahei

he

ocas

ecr,

is re-uldwe

tec--

A

As/

srent

tedOV


layer. To suppress the current of the device, a thick blockbarrier is placed between the active region and the collec

D. Detectivity

We can distinguish two types of detector noise: radiatnoise and intrinsic noise. Radiation noise includes sigfluctuation noise and background fluctuation noise. For inred detectors, background fluctuation noise is higher copared to the signal fluctuation noise. In photodiodes, the snoise is usually the major type of noise.

In the case of QWIP’s, the major source of noise isdark current, which is a generation-recombination one. Dto high levels of the dark current, the Johnson noise isglected in most cases, especially at high temperatures oeration. At a lower temperature and at smaller pixel sizes,Johnson noise becomes comparable to the dark noise. Oto stable surface properties, very low 1/f noise is observed inQWIP’s.

For FPA’s pattern noise~which results from local varia-tion of the dark current, photoresponse, and cutoff walength! is the major limitation of array performance, espcially at low temperatures. This type of noise mayconsidered as a nonuniformity appearing across the aThis nonuniformity does not vary with time and reflects tintrinsic properties of an FPA. The fixed pattern noiselower in QWIP arrays than in HgCdTe arrays due to a higmaterial quality and better-controlled cutoff wavelength.

Figure 19 compares the detectivities ofp-on-n HgCdTephotodiodes with those of GaAs/AlGaAs QWIP’s. The theretical curves corresponding to HgCdTe photodiodes areculated on the assumption of constant cutoff wavelength10 mm and 11mm. The VLWIR results for HgCdTe~14.8mm at 80 K and 16.2mm at 40 K! and QWIP at 16mm showthe intrinsic superiority of HgCdTe photodiodes. The dettivity of HgCdTe is roughly an order of magnitude highe

FIG. 19. LWIR detector detectivity versus temperature for GaAs/AlGaQWIP’s andp–on–n HgCdTe photodiodes~after Refs. 2 and 51!.


gr.

nal--

ot

ee

e-p-e

ing

--

y.

sr

-l-

of

-

though the advantage decreases as the temperatureduced. The best example of a region where QWIP’s cohave a performance advantage is low temperatures. Ascan see from Fig. 19, at temperatures<45 K QWIP’s offersuperior performance at a 7.7mm peak wavelength relativeto HgCdTe at 10.6mm.

Since 1987, rapid progress has been made in the detivity of long-wavelength QWIP’s, starting with bound-to

s

FIG. 20. Dependence of detectivity on the long wavelength cutoff for GaAlGaAs QWIP’s andn1 –p HgCdTe photodiodes at temperatures<77 K.The dashed lines are calculated forn-doped QWIP’s. The theoretical linefor photodiodes are calculated assuming the Auger-limited diffusion curmechanism~dashed lines! and diffusion and tunneling mechanisms~dash-dotted lines! ~after Refs. 2 and 51!.

FIG. 21. Estimation of the temperature required for background limioperation of different types of photon detectors. In the calculations F530° andTB5300 K are assumed~after Ref. 31!.


ityns

r

Ind

ithfo

-n-ad

ta-e-

atent,21,tedheture

de-ither

than’s

ithe

ise

ure-ad-cynly

a

I


bound QWIP’s characterized by a relatively poor sensitivand culminating in high performance bound-to-quasibouQWIP’s with random reflectors. All the QWIP data are clutered between 1010 and 1011 cm Hz1/2/W at operating tem-peratures of about 77 K. Gunapala and Bandara3 assert thatthe highest possibleD* obtained using random gratings, foexample, is about 231011 cm Hz1/2/W. This appears to beconsistent with the data including the enhanced QWformed by both patterning the multiple quantum wells areducing the number of wells.

The above performance comparison of QWIP’s wHgCdTe in the low temperature range is less favorablephotodiodes in the case ofn1-p structures. At low temperatures, thep-type HgCdTe material suffers from some nonfudamental limitations~contacts, surface, Schockley–Re

FIG. 22. Noise performance of typical HgCdTe photodiodes and QW~after Ref. 38!.


d-

P

r

processes! more than then-type one. Including the influenceof tunneling, the comparison of detectivity is more advangeous for GaAs/AlGaAs QWIP’s in the spectral region blow 14 mm and at temperatures below 50 K~see Fig. 20!.

E. Background-limited performance

The BLIP temperature is defined as the temperaturewhich the dark current equals the background photocurrat a given a FOV and background temperature. In Fig.the calculated temperature required for background limioperation with a 30° FOV is shown as a function of tcutoff wavelength. We can see that the operating temperaof ‘‘bulk’’ intrinsic IR detectors ~HgCdTe and PbSnTe! ishigher than for other types of photon detectors. HgCdTetectors with performance limited by background operate wthermoelectric coolers in the MWIR range. On the othhand, LWIR detectors (8<lc<12 mm! operate at'100 K.HgCdTe photodiodes can operate at temperatures higherextrinsic detectors, silicide Schottky barriers, and QWIPcan. However, the cooling requirements for QWIP’s wcutoff wavelengths below 10mm are less stringent than thosfor extrinsic detectors and Schottky barrier devices.

Figure 22 shows both photon and dark current novoltage for an 8.5-mm cutoff QWIP and a 12-mm cutoffHgCdTe FPA. For the purposes of comparison, the measments are normalized to the same dark current levels byjusting the readout gain. Due to a lower quantum efficienand higher dark current, QWIP detectors become BLIP oat photons fluxes in excess of 231013 photons/cm2 s. Bycomparison, LWIR HgCdTe photodiodes are BLIP down tophoton flux of 131010 photons/cm2 s. This indicates their

P

FIG. 23. Hybrid IR FPA interconnects technique between a detector array and silicon multiplexer:~a! indium bump technique,~b! loophole technique, and~c!scanning electronic microscopy photo of FPA.


pnd

ysIRimunbreninulgen-Tidco

exeracari1E

f

rates

asn-

de

toraytnd

el-IR

Te-

the

-

ent

her ofthengeixelmK,histedis

s-a-

ting.ionberge

R

et isithngde-

ce.

cti


privileged position in the area of strategic and space apcations, where longer-range detection at low backgrouand sensitivity to longer wavelengths is required.

IV. FOCAL PLANE ARRAY PERFORMANCE

The combination of existing high-performance arraand highly developed silicon integrated circuits in hybridarray/silicon planes have proved to be useful for thermalaging systems with the thermal and spatial resolutionmatched by any competing technologies at present. In hyFPA’s, detectors and multiplexers are fabricated on differsubstrates and mated with each other by flip-chip bond~see Fig. 23!. In this case the detector material and the mtiplexer may be optimized independently. Other advantaof hybrid FPA’s are the near 100% fill factor and the icreased signal-processing area on the multiplexer chip.detector array can be illuminated either from the fronts~with the photons passing through the transparent silimultiplexer! or from the backside~with photons passingthrough the transparent detector array substrate!. In general,the latter approach is most advantageous, as the multiplwill typically have areas of metallization, as well anothopaque regions that may reduce the effective optical arethe structure. In HgCdTe hybrid FPA’s, photovoltaic detetors are formed in a thin HgCdTe epitaxial layer on transpent CdTe or ZnCdTe substrates. In HgCdTe flip-chip hybtechnology, the maximum chip size is on the order of318 mm2.52 In order to overcome this problem, the PAC~Producible Alternative to CdTe for Epitaxy! technology isbeing developed with sapphire or silicon as the substrate

FIG. 24. Noise equivalent difference temperature as a function of deteity. The effects of nonuniformity are included foru51023 and 1024 ~afterRef. 2!.


li-s

--idtg-s

heen

er

of-r-d8

or

HgCdTe detectors. When using opaque materials, substmust be thinned to 10–20mm in order to obtain sufficientquantum efficiencies and to reduce the cross talk.

Progress in the area of LWIR HgCdTe photodiodes hbeen relatively slow. In October 1992, in the U.S.A, a cosortium was formed~with DARPA’s support! to developp-on-n double-layer heterostructure HgCdTe photodioFPA’s and two color arrays.53 MBE makes HgCdTe morefavorable for producing high-quality FPA’s.54 Currently,1283128, 2563256, and 3203256 LWIR arrays are com-mercially available, and such large formats as 6403480 onsilicon substrate have been demonstrated.55 Also prototypeMWIR/LWIR two-color assemblies with 1283128 ~Ref. 56!and 2563256 ~Ref. 57! pixel formats have been producedacquire IR imagery. It should be noted, however, that arsize, uniformity, reproducibility, and yield are still difficulissues, especially for low-temperature and low-backgrouoperation.

The relatively new QWIP technology has been devoped very quickly in the last decade. Large-size LWGaAs/AlGaAs FPA’s with up to 6403480 (6403512) pixelshave been demonstrated3–5,58–64 with excellent uniformityand operability.

Among the cooled IR detector systems, only HgCdand QWIP offer wavelength flexibility and multicolor capabilities. Tables V and VI summarize the present status ofperformance of both detector systems.

A. Uniformity

As was discussed by Levine,2 at detectivities approaching values above 1010 cm Hz1/2/W ~see Fig. 24!, FPA perfor-mance is uniformity limited and thus essentially independof detectivity. An improvement in nonuniformity from 0.1%to 0.01% after correction could lower the NEDT~see Appen-dix 1! from 63 to 6.3 mK.

The nonuniformity value is usually calculated using tstandard deviation over the mean, counting the numbeoperable pixels in an array. For a system operating inLWIR band, the scene contrast is about 2%/K of the chain the scene temperature. Thus, to obtain a pixel to pvariation in the apparent temperature less than, e.g., 20the nonuniformity in response must be less than 0.04%. Tis nearly impossible to achieve in the case of an uncorrecresponse of the FPA; therefore, a two-point correctiontypically used.

FPA uniformity influences the complexity of an IR sytem. Uniformity is important for accurate temperature mesurements, background subtraction, and threshold tesNonuniformities require the development of compensatalgorithms to correct the image; but, by consuming a numof analog-to-digital bits they also reduce the dynamic ranof the system.

Tactical IR FPA’s usually require operation in the LWIwindow with a small number of applications in the 3–5mmwindow. The distance between the sensor and the targtypically short, allowing the use of imaging sensors wlarge FPA’s where precise radiometry is not critical. Imagiarrays can usually tolerate some percentage of dead orgraded pixels without the jeopardizing mission performan

v-



Downloaded 02 Ma

TABLE V. Performance specifications for LWIR HgCdTe FPA’s~after SOFRADIR and SBRC data sheets!.

Parameter SOFRADIR SBRC

Format 1283128 3203256 2563256Cut on–cut off~mm! 7.7–10.3 7.7–9.0 8.5–11.0FPA temperature~K! , 85 , 90 77 ~up to 100!Detector pitch~mm! 50 30 30Fill factor ~%! .70Charge handling capacity .1183106 e2 12 or 36 83106 e2 ~min!Frame rate~Hz! to 300 to 400 to 120D* peak rms/Tint/pitch ~average!~cm Hz1/2 W21 !

1.131011

Pixel NETD ~average! 10 mK for 275 Hz 18NEI ~photons/cm2 s! ~max! 1.5231012

Typical FOV f /2 f /2Fixed pattern noise 7% RMSCrosstalk~optical and electrical! ~%! 2Array operability~%! 99 99

gh

ireerA

A’s

i

or

o

hlly

ewitthao

of

or-

ntis

hshese-is

umi-nce

wotheex-

Tactical backgrounds in the IR windows are relatively hiwith about 1016 photons/cm2 s reaching the detector.

The admissible nonuniformity depends on the requments of the operability. A higher requirement on the opability usually leads to a lower uniformity and vice versa.typical uncorrected-response nonuniformity in QWIP FPis 1%–3% with the operability~the fraction of good pixels!higher than 99.9%. For a 1283128 15-mm array fabricatedby Jet Propulsion Laboratory~see Table VI!, the uncorrectedstandard deviation is 2.4% and the corrected nonuniformis 0.05%. For a large 6403486 9-mm FPA the uncorrectednoise nonuniformity is about 6%, and after a two-point crection improves to an impressive 0.04%.60 For an FPA ofthe the same format demonstrated by Lockheed Martin,erability higher than 99.98% was described.58 It is very hardfor HgCdTe to compete with QWIP’s in the area of higuniformity and operability of large-array formats, especiaat low temperatures and in the VLWIR region.

High uniformity and high operability, as shown in thabove examples, demonstrate the maturity of GaAs groand processing technology. In this context, the nonuniformand operability have been an issue for HgCdTe, althoughvalues published recently for Sofradir and SBRC arraysas high as 99%. As we compare the performance of btypes of FPA’s~see Table V and VI!, the array operability ishigher for QWIP’s~above 99.9%!.


--

ty

-

p-

thye

reth

Nonuniformity is a serious problem in the caseVLWIR HgCdTe detectors. The variation ofx across aHg12xCdx Te wafer causes a much larger spectral nonunifmity ~e.g., at 77 K the variation ofDx50.2% givesDlc

50.064 mm at lc55 mm, but Dlc50.51 mm at 14 mm!,which cannot be fully corrected by two or three poicorrections.32 Therefore the required composition controlmuch more stringent for VLWIR than for MWIR.

In the case of QWIP’s, extending the cutoff wavelengtto VLWIR is relatively easy since there is little change in tmaterial properties, growth, and processing. However, arious requirement for maintaining the device performanceto lower the operating temperature. Due to lower quantwell barriers, the dark current of thermionic emission domnates at lower temperatures. In order to achieve performaequivalent to that of a 10-mm cutoff QWIP at 77 K, thetemperature needs to be lowered to 55 K for a 15-mm cutoffand to 35 K for a 19-mm cutoff ~see Table VI and Fig. 19!.

B. NEDT

The noise in HgCdTe photodiodes at 77 K is due to tsources: the shot noise from the photocurrent andJohnson noise from the detector resistance. It can bepressed as35

ing

TABLE VI. Properties of JPL 9-mm and 15-mm GaAs/AlGaAs QWIP FPA’s~after Refs. 5 and 59–61!.

Parameter lc59 mm lc515 mm

Array size 2563256 ~Ref. 61! 3203256 ~Ref. 5! 6403486 ~Ref. 60! 1283128 ~Ref. 59!Pixel pitch ~mm! 38338 30330 25325 50350Pixel size~mm! 28328 28328 18318 38338Optical coupling 2D periodic grating 2D periodic grating 2D periodic grating 2D periodic gratPeak wavelength~mm! 8.5 8.5 8.3 14.2Cutoff wavelength, 50%~mm! 8.9 8.9 8.8 14.9Operability ~%! 99.98 99.98 99.9 .99.9Uncorrected nonuniformity~%! 5.4 5.6 2.4Corrected uniformity, 17–27 °C~%! 0.03 0.04 0.05Quantum efficiency~%! 6.4 6.9 2.3 3D* ~cm Hz1/2 W21) 2.031011 ~70 K! 2.031011 ~70 K! 1.631010 ~55 K!NEDT with f/2 optics~mK! 23 ~70 K! 33 ~70 K! 36 ~70 K! 30 ~45 K!


tioa

ucatee

ck

atio

pu-tols

ed

ter

nis

n-erity.y a

is

the

ll

the

it


I n5AS 2qIph14kTd

R DD f , ~23!

where k is the Boltzmann constant andR is the dynamicresistance of a photodiode. Assuming that the integratime t int is such that the readout node capacity is kept hfull, we have

D f 51

2t int~24!

and then

I n5AS 2qIph14kTd

R D 1

2t int. ~25!

At tactical background levels, the Johnson noise is msmaller than the shot noise from the photocurrent. In the cwhere the number of electrons collected in a frame is limiby the capacity of the ROIC charge well, which is often truthe signal to noise ratio is given by35

S

N5

qNw /2t int

A2qS qNw

2tD 1

2t int

5ANw

2. ~26!

Assuming that the temperature derivative of the baground fluxF can be written to a good approximation as

]F

]T5

hc

lkTB2

Q ~27!

and using Eq.~26!, the NEDT is equal to

NEDT52kTB

2 l

hcA2Nw

. ~28!

In the last two equations,l5(l11l2)/2 is the averagewavelength of the spectral band betweenl1 andl2.

If one assumes a typical storage capacity of 23107 elec-trons,l510 mm, andTB5300 K, Eq.~28! yields NEDT of19.8 mK.

The same estimate can be made for QWIP. In this cthe Johnson noise is negligible compared to the generarecombination noise therefore,

I n5A4qg~ I ph1I d!1

2t int, ~29!

where the dark current may be approximated by

I d5I 0 expS 2Ea

kTD . ~30!

In the above expressions,I d is the dark current,I 0 is a con-stant that depends on the transport properties and the dolevel, andEa is the thermal activation energy, which is usally slightly less than the energy corresponding to the cuwavelength of the spectral response. It should be astressed thatg, I ph , andI 0 are bias dependent.

The signal-to-noise ratio for a storage-capacity-limitQWIP is given by


nlf

hsed,

-

sen-

ing

ffo

S

N5

qNw /2t int

A4qgS qNw

2tD 1

2t int

51

2ANw

g, ~31!

and the NEDT is

NEDT52kTB

2 l

hcA g

Nw. ~32!

Comparing Eqs.~28! and ~32! one may notice that thevalue of NEDT in a charge-limited QWIP detectors is betthan that of HgCdTe photodiodes by a factor of (2g)1/2 sincea reasonable value ofg is 0.4. Assuming the same operatioconditions as for HgCdTe photodiodes, the value of NEDT17.7 mK. Thus, a low photoconductive gain actually icreases theS/N ratio and a QWIP FPA can have a bettNEDT than an HgCdTe FPA with a similar storage capac

The above deduction was confirmed experimentally bresearch group at Fraunhofer IAF.62–64The photoconductionmechanism of a photovoltaic ‘‘low-noise’’ QWIP structureindicated in the inset of Fig. 25~b!, where four zones~1!–~4!of each period are shown. Because of the period layout,detector structure has been called a four-zone QWIP.65 Thefirst two zones~1 and 2! are analogous to the barrier and weof a conventional QWIP@see Fig. 6~a!#. Two additionalzones are present in order to control the relaxation ofphotoexcited carriers: namely, a capture zone~3! and a tun-neling zone~4!. The tunneling zone has two functions;blocks the carriers in the quasicontinuum~carriers can be

FIG. 25. Peak responsivity, gain~a!, and peak detectivity~b! of a low-noiseQWIP versus bias voltage~after Ref. 62!.


thth

20

-

thiviahe

IP

alth

d in

tege,s ised-ewrentherg aP atoutorherheted

65


captured efficiently into a capture zone! and transmits thecarriers from the ground state of the capture zone intoexcitation zone of the subsequent period. In this way,noise associated with the carrier capture is suppressed.

Figure 25 summarizes the performance of a typicalperiod low-noise QWIP with a cutoff wavelength of 9.2mm.The peak responsivity is 11 mV at zero bias~photovoltaicoperation! and about 22 mA in the range between22 and23 V. Between21 and22 V, a gain of about 0.05 is observed. The detectivity has its maximum around20.8 V andabout 70% of this value is obtained at zero bias. Due toasymmetric nature of the transport process, the detectstrongly depends on the sign of the bias voltage. This behior is in strong contrast with a conventional QWIP where tdetectivity vanishes at zero bias.

Based on the photovoltaic low-noise four-zone QWstructure, the Fraunhofer group62–64 has manufactured a2563256 FPA camera operating at 77 K with the 9-mm cut-off wavelength. The camera exhibits record-low NEDT vues of 7.4 mK with 20-ms integration time and 5.2 mK wi

FIG. 26. NEDT improvement due to the skimming architecture~f/1, 25mmpitch, Tb5300 K!.

FIG. 27. NEDT of a 1283128 HgCdTe FPA (f /2 optics, 50% well fill,pitch 50mm! as a function of operating temperatures~after Ref. 67!.


ee

-

etyv-

-

40 ms. It is the best temperature resolution ever obtainethe LWIR regime.

Considerations carried out in the Appendix 1 indicathat for a given frame time and maximum storable charthe way to increase the performance of LW staring systemto reduce the dark current in order to get photon limitsystem performance. Boiset al.66 described their unique approach of using two stacks of identical QWIP’s and a nskimmed architecture accommodating the large dark curof the detector. The QWIP at the top stack produced a higphotocurrent than that of the lower stack. Therefore, usinbridge readout arrangement, the dark current of the QWIthe top stack may be subtracted by the bottom stack withsacrificing much of the photocurrent. With this detectstructure, the FPA should be able to operate at much higtemperatures and with a longer integration time. After tselection of the optimum subtraction rates the expecNEDT is 10 mK at 85 K forlc59.3 mm andf/1 optics. Thisperformance is usually achieved with standard QWIP’s atK ~see Fig. 26!

FIG. 28. NEDT of a 3203256 QWIP FPA (lc58.9 mm, f /2 optics, 50%well fill, pitch 30 mm! as a function of operating temperature~after Ref. 5!.

FIG. 29. NEDT histogram of a 6403486 QWIP FPA (lc58.8 mm! atoperating temperature of 70 K, bias voltage of22 V at 300 K backgroundwith f /2 optics~after Ref. 61!.


iongysm

db

eredocae

inte

as

te

ho

en

ar

0K36/enive

’sceto

ha

fo

poho

neeanen

ting

is

. Inndsh-

anw-IPheing

isfiny

,

cence

e ofcul-es-

-ap-isein-


Two other solutions can be envisaged for the reductof the dark current:~i! to decrease the focal plane operatitemperature, and~ii ! to decrease the cutoff wavelength badjusting the detector material composition. The secondlution allows one to increase the focal plane operation teperature without degrading the performance.

Figure 27 compares the performance of two SofraHgCdTe staring arrays. One of these arrays is sensitivetween 7.7mm and 9mm, the other between 7.7mm and 9.5mm. Higher performance with improved technology has beobtained using, on the one hand, the reduced-dark-curdetector technology and, on the other hand, a new reacircuit architecture maximizing both the charge handlingpacity and the responsivity.68 At a constant temperature, thperformance is higher forlc59 mm due to the fact that thedark current is lower and the integration time can becreased. As a consequence, the focal plane operationperature of the improved arrays withlc59 mm can be in-creased up to 105 K and up to 102 K for arrays withlc

59.5 mm ~NEDT,18 mK in this example!.Figure 28 shows the measured and estimated NEDT

function of the temperature for a 8.9-mm QWIP FPA. Incomparison with a representative HgCdTe FPA~Fig. 27!, thisparameter exhibits a strong temperature dependence. Atperatures,70 K, the signal-to-noise~SNR! ratio of the sys-tem is limited by the multiplexer readout noise and the snoise of the photocurrent. At temperatures.70 K, the tem-poral noise due to the QWIP’s higher dark current becomthe limitation. As mentioned earlier, this higher dark curreis due to thermionic emission and thus causes the chstorage capacitors of the readout circuitry to saturate.

Figure 29 presents a NEDT histogram of a 6403486QWIP FPA (lc58.8 mm! at an operation temperature of 7K, bias voltage of22 V, background temperature of 300with f/2 optics. The mean value of the NEDT histogram ismK. The uncorrected nonuniformity~standard deviationmean! of this unoptimized FPA is only 5.6% including th1% nonuniformity of the readout circuit and the 1.4% nouniformity due to the fact that the cold stop is unable to gthe same field of view to all the pixels in the FPA.

Comparing the values of NEDT for both types of FPA~see also Tables V and VI!, we can see that the performanof LWIR HgCdTe arrays is generally better. However, upnow the best temperature resolution in the LWIR bandbeen obtained with four zone QWIP structures.62–64

C. Charge handling capacity and integration time

One of the simplest and most popular readout circuitsIR FPA’s is the direct injection~DI! input, where the darkcurrent and photocurrent are integrated into a storage caity. In this case, the bias varies across the array by ab6~5–10! mV due to the variations in the transistor thresolds. At 80 K, HgCdTe diodes show very small changesthe leakage current with small changes in reverse biaszero volts~see Fig. 12!. The goal is to fit a capacitor as largas possible into the unit cell, particularly for high tacticapplications, where signal-to-noise ratios can be obtaithrough longer integration times. For high injection efficie


n

o--

ire-

nntut-

-m-

a

m-

t

stge

-

s

r

ac-ut-far

ld

-

cies, the resistance of the field effect transistor~FET! shouldbe small compared to the diode resistance at its operapoint. Generally, it is not difficult to fulfill this inequality forMWIR HgCdTe staring designs where diode resistancelarge (R0A product in the range of more than 106 V cm2),but it may be very important for LWIR designs whereR0 issmall (R0A product about 102 V cm2 !. LWIR HgCdTe pho-todiodes operate with a small reverse bias near zero voltsthis case, a large bias is desirable but its value depestrongly on the material quality of the array. For very higquality LWIR HgCdTe arrays a bias of21 V is possible.

The bias applied to a QWIP array is usually higher ththat applied to a HgCdTe array and is around 2–3 V. Hoever, the readout power consumption is similar for QWand HgCdTe and is negligible compared with that of treadout electronics. The power dissipation in an imag6403480 QWIP FPA is, for example, lower than 150 mW.58

It results from the very high impedance of the QWIP. Thimpedance is in the GV range at 77 K for a pixel size o24324 mm2.32 It makes the readout design very easyterms of achieving high efficiency. The injection efficiencof a 9-mm cutoff 6403486 QWIP array is, for example99.5%.58 However, Singh and co-workers69,70 uncovered apossible problem connected with the offsets~up to'1 V! inthe low-temperature current-voltage characteristics~nonzerocurrent at zero bias!. This observation indicates the presenof a time-constant effect due to the charging of a capacitain the QWIP detector. QWIP’s behave likeRC circuits, witha resistance in series with a capacitance. The presenclarge offsets or large dynamic resistance could cause diffities in matching the impedances of the FPA and ROIC,pecially at low biases and certainly at low temperatures.

The DI input circuit is not generally used for low backgrounds due to injection efficiency issues. The strategicplications often have low backgrounds and require low nomultiplexers interfaced to high-resistance detectors. The

FIG. 30. NEDT versus charge handling capacity~after Ref. 72!.


he

iththr t

hear

e.

toed

rgte

s

rd

ths

thc

heI

ec

lxissste

ce

o

detyof

oon-

ra

cyine-t

s isnt

tiesa-pli-

si-K.

the0all

ardof

per32,en-er

itorsted

t tothe

ias


put circuit commonly used for strategic applications is tcapacitative transimpedance amplification~CTIA! inputcircuit.71 Increased injection efficiency is also possible wbuffered DI. The latter two input circuits also accentuate1/f noise and the operability but they require higher poweoperate.

At high backgrounds, it is often impossible to handle tlarge number of carriers generated over frame times thatcompatible with the standard video frame rates. The useforced to integrate for only a fraction of the full frame timOff-FPA integration of subframes is then used to attainlevel of sensor sensitivity commensurate with the deteclimited FPA detectivity and not the charge-handling-limitsensor detectivity.

Figure 30 shows the theoretical NEDT versus chahandling capacity for HgCdTe FPA’s assuming that the ingration capacitor is filled to half the maximum capacity~topreserve dynamic range! under nominal operating conditionin different spectral bandpasses: 3.4–4.2mm, 4.4–4.8mm.3.4–4.8 mm, and 7.8–10mm. The measured data foTCM2800 at 95 K, other PACE HgCdTe FPA’s at 78 K anrepresentative LWIR FPA’s are also shown. We can seethe measured sensitivities agree with the expected value

The well charge capacity is the maximum amount ofcharge that can be stored in the storage capacitor of eachThe size of the unit cell is limited to the dimensions of tdetector element in the array. In the case of a large LWHgCdTe hybrid array, a mismatch between the thermalpansion coefficients of the detector array and the readoutforce the cell pitch to 20mm or less to minimize lateradisplacement. However, the development of heteroepitagrowth techniques for HgCdTe on Si has opened the pobility of a cost-effective production of significant quantitieof large-area arrays through the utilization of large-diameSi substrates.

Figure 31 shows the charge handling capacity versuspitch. We can see that for a 30330 mm2 pixel size, the stor-age capacities are limited to the range of 1 to 53107 elec-trons depending on the design feature. At a 0.8-mm feature ifthe oxide is thinned to 150 Å to recoup capacity, the read

FIG. 31. Charge handling capacity versus cell pitch~after Ref. 73!.


eo

reis

ar-

e-

at.eell.

Rx-an

ali-

r

ll

ut

yield and device reliability may be reduced due to oxidamage from hot carriers.73 For example, a storage capaciof 53107 electrons requires that the total current densitya detector with a 30330 mm2 pixel size be lower than 27mA/cm2 with a 33-ms integration time.32 If the total currentdensity is in the 1 mA/cm2 range, the integration time has tbe reduced to 1 ms. In LWIR HgCdTe FPA’s the integratitime is usually below 100ms. Since the noise power bandwidth D f 51/2t int , a short integration time causes extnoise during integration.

LWIR QWIP FPA’s using conventional ROIC’s typicallyoperate at 60–65 K. Due to the smaller quantum efficienof a QWIP, filling the charge capacitor is not a problemhigh-background applications. QWIP’s allow a longer intgration time, which yields relatively low values of NEDT. Ahigher temperatures, however, the dark current of QWIP’high and fills the charge capacitor very quickly. The curresubtraction and switched capacitor noise filtering capabiliof ROIC’s permit a low NEDT at higher operating tempertures. In this case, however, the readout circuit is comcated, which limits the array size.

The goal of third-generation imagers is to achieve sentivity improvement corresponding to a NEDT of about 1 mThe analysis of Eq.~A2! ~see Appendix A1! indicates thatthe required charge storage capacity in a 300 K scene inLWIR region with the thermal contrast of 0.04, is above 19

electrons. This high charge-storage density within the smpixel dimensions probably cannot be obtained with standCMOS capacitors.52 Although the reduced oxide thicknessthe submicrometer design rules gives a large capacitanceunit area, the reduced bias voltage, as illustrated in Fig.largely cancels any improvement in the charge storage dsity. Ferroelectric capacitors may provide much highcharge storage densities than the oxide-on-silicon capacnow used, however, such technology is not yet incorporainto standard CMOS foundries. Nortonet al.52 suggested us-ing stacked hybrid structures as an interim solution at leasincorporate the desired charge storage density indetector–readout–capacitor structures.

FIG. 32. Trends for design rule minimum dimensions and maximum bvoltage of silicon foundry requirements~after Ref. 52!.


oin

veroete

in

cothlyot othseidchesue

nthee

Tebes

ena

olsanecn

fth

e-b-r

asveissdAmte

s

hero-

tednots-to

k tor-

andandIPofb-auseth

i-y-

on-er.s aan

e-he

.he

desm-nis

fill

-Te

Te

hesityadyison-

ps.hy-

out


D. Cost

The cost of a FPA depends strongly on the maturitythe technology and varies with the production quantitydifferent companies. So far, large-size LWIR FPA’s habeen developed in R&D laboratories only, without mass pduction experience. According to Sofradir, a continuousfort in the industry has decreased the cost of HgCdTe detors by a factor of 5 to 10.74 The cost of making high-performance cooled components can be broken downthree parts of about equal weight: the chip~detector andROIC!, the Dewar, and the integration and tests.75 In addi-tion, the user must add the cryogenic machine cost. Thisis not negligible compared to the other ones. Even ifdetection circuit is free of charge, the total cost would onbe reduced by about 15%–20%. This explains why the cof PtSi and QWIP detectors is not markedly less than thaphoton detectors of the same complexity, even thoughraw materials~Si or GaAs! are much cheaper than in the caof HgCdTe. Moreover, since a PtSi requires a very woptical aperture to obtain acceptable performance and sinQWIP requires lower operating temperatures than other pton detectors, a possible reduction in the purchase priccounterbalanced by a significant increase in operating co

HgCdTe detectors have been the center of a major indtry during the last three decades. The technology is relativmature in the MWIR region with a size up to 102431024,25

but it is not extended easily to LWIR. To make componewith more pixels, a reduction of the pitch is required or tmastering of the thinning operation needed to withstand thmal cycling~differential thermal expansion between CdZnand silicon!. It seems that, in the future, Si substrates willused, because they offer many well-known advantagecompared to bulk CdZnTe substrates~a much larger sizeavailable at lower cost, a thermal expansion coefficimatching that of Si readout chips, higher purity, and compibility with automated wafer processing/handling methodogy due to their superior mechanical strength and flatne!.Promising results have been achieved in the SWIRMWIR spectral region. During the last four years the defdensity for MWIR layers of HgCdTe grown by MBE osilicon substrates decreased from 2000 cm22 to below 500cm22.52 Currently, MWIR arrays with pixel operability o98% can be produced from this material. For comparison,operability of CdZnTe is typically about 99% or higher. Dfect densities in the LWIR material grown on silicon sustrates continue to limit performance but they have beenduced by an order of magnitude in the past decade.52,55

In comparison with HgCdTe FPA’s, the industry hmuch less experience in QWIP FPA’s; therefore, improments can be expected, especially since this technologyan earlier stage of development. The major challenge iimprove the device performance by means of tuning thevice and grating designs. Because of the maturity of Gagrowth technology and the stability of the material systeno investment is needed for developing QWIP substraMBE growth, and processing technology.32 Four-inch QWIPwafers are commercially available.76 Development of LWIRand multicolor HgCdTe detectors is extremely difficult, e


f

-f-c-

to

ste

stfe

ee ao-ist.s-ly

s

r-

as

tt--sdt

e

e-

-at

toe-s,s,

-

pecially for low-background applications. It means that tcost of the development QWIP technology and QWIP pduction will be less than in the case of HgCdTe.

E. Reliability

In our discussion the reliability issue has been omitdue to the fact that statistical data on this subject areavailable. In several applications, especially military sytems, there is a considerable demand for high reliabilityensure both the success of the mission and minimal risthe user. Two reliability challenges affect both FPA’s: suvival in high-temperature system-storage environmentswithstanding repetitive thermal cycles between ambientcryogenic temperatures. In HgCdTe, as well as in QWFPA’s, indium bumps are used to hybridize both typesdetectors with a silicon multiplexer. However, certain prolems can be expected in the case of QWIP arrays, becindium bumps are known to form many different alloys wiIII–V compounds.

Very large FPA’s may exceed the limits of hybrid relability engineered into contemporary cooled structures. Hbrids currently use mechanical constraints to force the ctraction of the two components to closely match each othThis approach may have limits when the stress reachepoint where the chip fractures. Three approaches offeropportunity to resolve this issue:

~i! Elimination of the thick substrate that limits the dtector active region from deforming at the slower rate of tsilicon readout.

~ii ! Subdivision of the array into a number of regions~iii ! Use of silicon as the substrate for the growth of t

detector material.The process sequence for making HgCdTe photodio

and QWIP’s on silicon substrates would ultimately be siplified because of the maturity level of the existing silicotechnology. However, the major difficulties inherent in thapproach are the following:52

~i! Less area would be available for readout circuitry.~ii ! Microlens arrays would be required to regain the

factor.~iii ! Material quality may not be adequate for low

leakage detectors, particularly in the case of LWIR HgCdphotodiodes.

~iv! Silicon integrated circuits are processed on^100&oriented silicon, but the preferred orientation for HgCdgrowth on silicon is near the211& orientation.

A less demanding approach to the elimination of tthick detector substrate is to use the loophole or high-denvertically-integrated photodetector device structures alrepracticed by GEC Marconi and DRS, respectively. In thapproach, HgCdTe material is glued to the readout and ctacts are made through a thin layer~10–20 mm! after thesubstrate is removed@see Fig. 23~b!#. Another approach is toremove the substrate after hybridization with indium bumSubstrate removal is a standard practice with very largebrid InSb arrays (102431024 pixels!. This approach hasbeen recently adopted for QWIP arrays.4,5,77 After epoxybackfilling of the gaps between the array and the read


m

Ath

en

cecdinenimulin

IRplionctul-inhu’sp

menalllyemsop-

rre-isn-size,

aalistign,

atedtory

ess in

elyds

dep

hefil-

torso-, isnumof

pho-the

.thet toenes;

esof

s orW

y-r. Itb-eirop-be

ted


multiplexer, the substrate is thinned to form a very thin mebrane~'1000 Å!. This allows not only the elimination of thethermal mismatch between the silicon readout and the Gabased detector array but also the complete elimination ofpixel-to-pixel cross talk, and finally, significant enhancemof the optical coupling of IR radiation to QWIP pixels.

V. MULTIBAND FPA’s

Multicolor capabilities are highly desirable for advanIR systems. Systems that gather data in separate IR spebands can determinate both the absolute temperature anunique signatures of the objects in the scene. By providthis new dimension of contrast, multiband detection alsoables advanced color processing algorithms to furtherprove the sensitivity above that of single-color devices. Mtispectral detection permits rapid and efficient understandof the scene in a variety of ways. In particular, two-colorFPA’s can be especially beneficial for threat-warning apcations. By using two IR wave bands, spurious informatisuch as background clutter and sun glint, may be subtrafrom the IR image, leaving only the objects of interest. Mtispectral IR FPA’s can also play many important rolesground and planetary remote sensing, astronomy, etc. Tthe effective signal-to-noise ratio of two-color IR FPAgreatly exceeds that of single-color IR FPA’s for specific aplications.

FIG. 33. Cross section views of unit cells for various back-illuminadual-band HgCdTe detector approaches:~a! bias-selectablen–p–n structurereported by Raytheon~Ref. 78!, ~b! simultaneousn–p–n design reported byRaytheon~Ref. 80!, ~c! simultaneousp–n–n–p reported by BAE Systems~Ref. 83!, ~d! simultaneousn–p–p–p–n design reported by Leti~Ref. 85!,and ~e! simultaneous structure based onp–on–n junctions reported byRockwell ~Ref. 56! ~after Ref. 91!.


-

s-et

traltheg--

-g

-,ed

s,

-

Currently, multispectral systems rely on cumbersoimaging techniques that either disperse the optical sigacross multiple IR FPA’s or use a filter wheel to spectraselect the image focused on a single FPA. These systcontain beam-splitters, lenses, and bandpass filters in thetical path to focus the images onto separate FPA’s cosponding to different IR bands. Also complex alignmentrequired to map the multispectral image pixel for pixel. Cosequently, these approaches are expensive in terms ofcomplexity, and cooling requirements.

At present, the efforts are being made to fabricatesingle FPA with multicolor capability to eliminate the spatialignment and temporal registration problems that exwhenever separate arrays are used, to simplify optical desand reduce size, weight, and power consumption.

Considerable progress has been recently demonstrby research groups at Hughes Research Labora~Raytheon!,78–80 Lockheed Martin ~BAE Systems!,81–84

Rockwell,56 Leti,85 and DRS35 in multispectral HgCdTe de-tectors employing mainly MBE~although LPE and MOCVDare also used! for the growth of a variety of devices. ThQWIP technology also demonstrates considerable progrethe fabrication of multicolor FPA’s.4,5,35,58,86–90Devices forthe sequential and simultaneous detection of two closspaced subbands in the MWIR and LWIR radiation banhave been demonstrated.

A. Dual-band HgCdTe

In backilluminated dual-band detectors, the photodiowith the longer cutoff wavelength is grown epitaxially on toof the photodiode with the short cutoff wavelength. Tshorter-cutoff photodiode acts as a long-wavelength-passter for the longer-cutoff photodiode.

Both sequential mode and simultaneous mode detecare fabricated from multilayer materials. The simplest twcolor HgCdTe detector, and the first to be demonstratedthe bias-selectablen-p-n back-to-back photodiode shown iFig. 33~a!. The sequential-mode detector has a single indibump per unit cell that permits sequential bias selectivitythe spectral bands associated with the operating tandemtodiodes. When the polarity of the bias voltage applied tobump contact is positive, the top~LW! photodiode is reversebiased and the bottom~SW! photodiode is forward biasedThe SW photocurrent is shunted by the low impedance offorward-biased SW photodiode, and the only photocurrenemerge in the external circuit is the LW photocurrent. Whthe bias voltage polarity is reversed, the situation reversonly the SW photocurrent is available. Switching timwithin the detector can be relatively short, on the ordermicroseconds, so the detection of slowly changing targetimagers can be done by switching rapidly between the Mand LW modes.

One bump contact per unit cell, as for single-color hbrid FPA’s, is a big advantage of a bias-selectable detectois compatible with existing silicon readout chips. The prolems with bias selectable devices are the following: thconstruction does not allow independent selection of thetimum bias voltage for each photodiode, and there cansubstantial MW crosstalk in the LW detector.



FIG. 34. Spectral response curves for two-color HgCdTe detectors in various dual-band combinations of MWIR and LWIR spectral bands~after Ref. 38!.

cinft-icren

t i

dea

v-

reeotioinestw

i

oet

he

n-out

ons had

asty.ouse-

ob-nicistedple.theisndis

ndiva-

Many applications require true simultaneity of the detetion in the two spectral bands. This has been achievednumber of ingenious architectures schematically shown, aReine,91 in Figs. 33~b!–33~e!. All these simultaneous dualband detector architectures require an additional electrcontact to the underlying layer in the multijunction structuof both the SW and LW photodiodes. The most importadistinction is the requirement of a second readout circuieach unit cell.

Integrated two-color HgCdTe technology has been undevelopment for nearly a decade with steady progressnow boasts a wide variety of pixel sizes~30–61mm!, arrayformats (64364 up to 3203240) and spectral-band sensitiity ~MWIR/MWIR, MWIR/LWIR, and LWIR/LWIR!. Figure34 shows examples of the spectral response from diffetwo-color devices. Note that there is minimal cross talk btween the bands, since the short-wavelength band absnearly 100% of the shorter wavelengths. One might queswhether the additional processing and complexity of builda two-color detector structure might adversely affect the pformance, yield, or operability of two-color devices. Testructure indicates that the separate photodiodes in acolor detector perform exactly as single-color detectorsterms of the achievableR0A product variation with wave-length at a given temperature. For example, fill factors1283128 MWIR/MWIR FPA’s as high as 80% werachieved using a single mesa structure to accommodate

FIG. 35. NEDT for a two-color camera having 50 mm,f/2.3 lens, as afunction of the operating temperature~after Ref. 38!.


-a

er

al

tn

rnd

nt-rbsn

gr-to-n

f

wo

indium bump contacts required for each unit cell with tsize of 50mm. Band 1~2.5–3.9mm! had the operability of99.9%, with 23 inoperable pixels. Band 2~3.9–4.6mm! hadthe operability of 98.9%, with 193 inoperable pixels. Quatum efficiencies of 70% were observed in each band withthe use of an antireflection coating. TheR0A values for thediodes ranged from 8.253105 to 1.13106 V cm2 at f/2 FOV.The NEDT for both bands is shown in Fig. 35 as a functiof temperature. The camera used for these measurementa 50 mm,f/2.3 lens. Imagery was acquired at temperatureshigh as 180 K with no visible degradation in image quali

Recently, Rockwell has developed a novel simultanetwo-color MWIR/LWIR FPA technology based on a doubllayer planar heterostructure~DLPH! MBE technology@seeFig. 33~e!#. To prevent the diffusion of carriers between twbands, a wide-band-gap 1-mm-thick layer separates these asorbing layers. The diodes are formed by implanting arseas a p-type dopant and activating it with an anneal. Thresults in a unipolar operation for both bands. The implanarea of band 2 is a concentric ring around the band 1 dimBecause the lateral carrier-diffusion length is larger thanpixel pitch in the MWIR material and band 1 junctionshallow, each pixel is isolated by dry-etching a trench arouit to reduce the carrier cross talk. The entire structurecapped with a layer of a material with a slightly wider bagap to reduce surface recombination and simplify passtion. Two-color 1283128 FPA’s with low background lim-

FIG. 36. Detectivity of two-color MWIR/LWIR 1283128 HgCdTe FPA~after Ref. 56!.


he

d-

nt

padiiath

teoficoA

A.5c

The

ths/i-fd

eenledryhy-

ryhenularlord

icem

ue

thdue

arlyheIP

aren beameitye-the

oItom-infor

he%.

lor5 V


ited detectivity performance (1013 cm22 s21) have been ob-tained for the MWIR~3–5mm! regime atT,130 K and forLWIR ~8–10mm! at T'80 K ~see Fig. 36!. The pixels ex-hibit BLIP performance up to 120 K for the two bands. TFPA also exhibits low NEDT values: 9.3 mK for the MWband and 13.3. mK for the LW band, similarly to gooquality single-color FPA’s.

B. Multiband QWIP’s

Sanders was the first to fabricate two-color, 2563256bound-to-miniband QWIP FPA’s in each of four importacombinations: LWIR/LWIR, MWIR/LWIR, near IR~NIR!/LWIR, and MWIR/MWIR—with simultaneousintegration.58,86

Devices capable of simultaneously detecting two serate wavelengths can be fabricated by vertically stackingferent QWIP layers during epitaxial growth. Separate bvoltages can be applied to each QWIP simultaneously viadoped contact layers that separate the MQW detector hestructures. Figure 37 shows schematically the structuretwo-color stacked QWIP with contacts to all three ohmcontact layers. The device epilayers were grown by MBEa 3-in. semi-insulating GaAs substrate. An undoped Galayer, called an isolator, was grown between two AlGastop layers, followed by an Au/Ge ohmic contact of a 0mm-thick doped GaAs layer. Next, two QWIP heterostru

FIG. 37. Structure of a two-color stacked QWIP~after Ref. 86!.

FIG. 38. Typical responsivity spectra at 40 K and a common bias of 1.recorded simultaneously for two QWIP’s in the same pixel~after Ref. 86!.


-f-se

ro-a

-ns

s--

tures were grown, separated by another ohmic contact.stack sensitive to long wavelengths~red QWIP, lc511.2mm! is grown above that sensitive to shorter waveleng~blue QWIP,lc58.6 mm!. Each QWIP is a 20-period GaAsAl xGa12xAs MQW stack in which the thickness of the Sdoped GaAs QW’s~with a typical electron concentration o531017 cm23) and the Al composition of the undopeAl xGa12xAs barriers~'550–600 Å! is adjusted to yield thedesired peak position and spectral width. The gaps betwFPA detectors and the readout multiplexer were backfilwith epoxy. The epoxy backfilling provides the necessamechanical strength to the detector array and a readoutbrid prior to the array’s thinning process.

Most QWIP arrays use a 2D grating, which is vewavelength dependent, and the efficiency gets lower wthe pixel size gets smaller. Lockheed Martin used rectangand rotated rectangular 2D gratings for their two-coLW/LW FPA’s. Although random reflectors have achieverelatively high quantum efficiencies with a large test devstructure, it is not possible to achieve similarly high quantuefficiencies with random reflectors on small FPA pixels dto the reduced width-to-height aspect ratios.5 In addition, it isdifficult to fabricate random reflectors for shorter-wavelengdetectors, when compared to long-wavelength detectors,to the fact that feature sizes of random reflectors are lineproportional to the peak wavelength of the detectors. Tquantum efficiency becomes a more difficult issue for QWmulticolor FPA’s than for single-color ones.

Typical operating temperatures for QWIP detectorsin the range of 40–100 K. The bias across each QWIP caadjusted separately, although it is desirable to apply the sbias to both colors. As is shown in Fig. 38, the responsivof both QWIP’s is around 300–350 mA/W. The spectral rsponse is relatively narrow so that a detector designed forLWIR band only detects light in that band with little or n‘‘spectral cross talk’’ with a detector in the MWIR band.appears that the complex two-color processing has not cpromised the electrical and optical quality of either colorthe two-color device since the peak quantum efficiencyeach of the 20-period QWIP’s was estimated to be'10%.For comparison, a normal single-color QWIP with twice tnumber of periods has a quantum efficiency of around 20A pixel operability for each color is.97% in comparison toa value higher than 99.9% routinely achieved for single-co,

FIG. 39. Simultaneous images from 2563256 MWIR/LWIR QWIP FPA’s.Note appearance of the filter and the soldering iron in the two bands~afterRef. 86!.


Pe1

ehea

/fiW

ionerth

ta

edue

d-th

0-

enab-

esub-tla-e

ceicthe

e

ri-t-

aen-.h-

hell

t-ry

of

or


QWIP’s. The NEDT value was 24 mK for the blue QWIand 35 mK for the red QWIP. The difference was attributto the poor transmission properties of the optics in the 1mm band.

To cover the MWIR range, a strained layer of thInGaAs/AlGaAs material system is used. InGaAs in tMWIR stack produces a high, in-plane, compressive strwhich enhances the responsivity.86,92 The MWIR/LWIRFPA’s fabricated by Sanders consist of an 8.6-mm GaAs/AlGaAs QWIP on top of a 4.7-mm strained InGaAs/GaAsAlGaAs heterostructure. The fabrication process allowedfactors of 85% and 80% to be achieved for the MW and Ldetectors, respectively. The first FPA’s with this configurathad an operability in excess of 97% and NETD valugreater than 35 mK. The excellent imagery in each coloshown in Fig. 39. Note the appearance of the filter andsoldering iron in the two bands.

Recently, Gunapalaet al.77,93have demonstrated the firs8–9 and 14–15mm two-color imaging camera based on6403486 dual-band QWIP FPA, which can be equippwith dual or triple contacts to access the CMOS readout mtiplexer. A single indium bump per pixel is usable only in th

FIG. 40. Conduction band diagram of a LWIR and a VLWIR two-coldetector~after Ref. 93!.


d.2

in

ll

sise

l-

case of the interlace readout scheme~i.e., odd rows for onecolor and even rows for the other!, which uses a existingsingle-color CMOS readout multiplexer. However, the disavantage is that it does not provide the full fill factor for bowavelength bands.

The device structure shown in Fig. 40 consists of a 3period stack~500 Å AlGaAs barrier and a 60 Å GaAs well!of the VLWIR structure and a second 18-period stack~500 ÅAlGaAs barrier and a 40 Å GaAs well! of the LWIR struc-ture separated by a heavily doped 0.5-mm-thick intermediateGaAs contact layer. The VLWIR QWIP structure has bedesigned to have a bound-to-quasibound intersubbandsorption peak at 14.5mm, whereas the LWIR QWIP structurhas been designed to have a bound-to-continuum interband absorption peak at 8.5mm, because the photocurrenand the dark current of the LWIR device structure are retively small compared to those of the VLWIR portion of thdevice structure.

Figure 41 shows a schematic side view of an interladual-band GaAs/AlGaAs FPA. Two different 2D periodgrating structures were designed to independently couple8–9 mm and 14–15mm radiation to detector pixels in theven and odd rows of the FPA. The top 0.7-mm-thick GaAscap layer was used to fabricate the light-coupling 2D peodic gratings for 8–9mm detector pixels, whereas the lighcoupling 2D periodic gratings of the 14–15mm detectorpixels were fabricated using LWIR MQW layers. In suchway, this grating scheme short circuited all the detector ssitive to the 8–9mm radiation in all odd rows of the FPANext, the LWIR detector pixels were fabricated by dry etcing the photosensitive GaAs/AlGaAs MQW layers into t0.5-mm-thick doped GaAs intermediate contact layer. AVLWIR pixels in the even rows of the FPA were shorcircuited. The VLWIR detector pixels were fabricated by detching both MQW stacks into the 0.5-mm-thick heavilydoped GaAs bottom contact layer. After epoxy backfilling

e
FIG. 41. Structure cross section of thinterlace dual-band FPA~after Ref.93!.


Downloaded 02 Ma

TABLE VII. Performance specifications for two-color long wavelength 6403486 QWIP FPA’s~after Ref. 93!.

Parameter LWIR pixels~8–9 mm! VLWIR pixels ~14–15mm!

Array size 6403486Pixel pitch ~mm! 25325 25325Pixel size~mm! 23323 22323Optical coupling 2D periodic grating 2D periodic gratingOperating temperature~K! 40 40Differential resistance~V! at f/2 FOV 2.031012 7.031011

Peak wavelength~mm! 8.4 14.4Peak responsivity~mA/W! at VB522 V 509 382Cutoff wavelength, 50%~mm! 9.1 15.0Operability ~%! 99.7 98.0Uncorrected nonuniformity~%! 5.4Corrected uniformity, 17–27 °C~%! 0.03Quantum efficiency~%! 12.9 8.9Uncorrected nonuniformity ofh ~%! 2 1DBB* ~cm Hz1/2 W21) at f/2 FOV 2.931010 1.131010

NEDT with f/2 optics~mK! 29 74

xAry

irralyixe

s

yerof

edthy,

is

K

-e,

d at

VII.atti-theresIR

t-e

ffi-

Atis

ry/nd,h

the gaps between FPA detectors and the readout multiplethe substrate was thinned and finally the remaining GaAlGaAs material contained only the QWIP pixels and a vethin membrane~'1000 Å!. This allows the whole structureto adapt to the thermal expansion, eliminates thermal mmatch between the silicon readout and the detector aeliminates pixel-to-pixel crosstalk, and finally, significantenhances the optical coupling of IR radiation to QWIP pels. The FPA was back illuminated through a flat thinnsubstrate membrane.

The 6403486 GaAs/AlGaAs gave excellent imagewith 99.7% of the LWIR pixels and 98% of VLWIR pixelsworking, demonstrating the high yield of GaAs technologThe 8–9mm detectors have shown background limited pformance~BLIP! at 70 K ~at a background temperature300 K with a f/2 cold stop!. The 14–15mm detectors showBLIP at 45 K with the other operating conditions unchangAt temperatures below 70 K, the signal-to-noise ratio ofLWIR detector pixels is limited by the array nonuniformitmultiplexer readout noise, and photocurrent noise~for the f/2cold stop!. At temperatures above 70 K, the temporal nodue to the higher dark current of QWIP’s~thermionic emis-sion! becomes the limitation. At temperatures below 40


er,s/

s-y,

-d

.-

.e

e

,

the signal-to-noise ratio of the VLWIR detector pixels is limited by the array nonuniformity, multiplexer readout noisand photocurrent noise.

The performance of these dual-band FPA’s was testea background temperature of 300 K, with af/2 cold stop, andat a 30 Hz frame rate The results are described in TableThe estimated NEDT of the LWIR and VLWIR detectors40 K is 36 and 44 mK, respectively. Due to BLIP, the esmated and experimentally obtained NEDT values ofLWIR detectors do not change significantly at temperatubelow 65 K. The experimentally measured values of LWNEDT equal to 29 mK are lower than the estimated ones@seeFig. 42~a!#. This improvement is attributed to the lighcoupling efficiency of the 2D periodic grating. However, thexperimental VLWIR NEDT value@see Fig. 42~b!# is higherthan the estimated value. It is probably the result of inecient light coupling in the 14–15mm region, readout multi-plexer noise, and the noise of the proximity electronics.40 K, the performance of both types of detector pixelslimited by the photocurrent noise and the readout noise.

At present, a joint Goddard/Jet Propulsion LaboratoArmy Research Laboratory proposal to develop a four bahyperspectral, 6403512 QWIP array is funded by the Eart

-

n

FIG. 42. The uncorrected NEDT histograms of 8–9mm detector pixels~a!and 14–15mm detector pixels~b! of6403486 dual-band FPA. The meaNEDT are 29 mK for LWIR FPA and74 mK for VLWIR FPA ~after Ref.93!.


th

coanprr

caea

edte

trhe

wonhect

tevehA

gy

ltiti

ntrethe

ofedr-herbyerarke

theaf-

per-andnot

an-n.

as18

in-

er-r-ath-

nil-andoortionosture-raof

he

lly,those


Science Technology Office of NASA.90 The four bands ofthe QWIP array are fabricated in a manner similar totwo-band system described above~see Fig. 41!. The four-detector bands are delineated by a deep groove etch totact a specific band, while shorting out the other three-blayers. Experiments are in progress to determine the omum coupling surface treatment and both grating and cogation methods are being considered.

The three-year effort will focus on expanding the spetral imaging capability of the integrated QWIP, utilizingnew Sunpower, Inc. Cryocooler, and developing a linvariable etalon to produce a hyperspectral image from 315.4mm. The four IR bands are 3–5mm, 8.5–10mm, 10–12mm, and 14–15.4mm. These bands are spectrally subdividinto approximately 200 bands. The technical goals are lisbelow:

~i! array size: 5123640, 23mm pixels,~ii ! 4 IR bands:~1! 3–5 mm, ~2! 8.5–10mm, ~3! 10–12

mm, ~4! 14–15.4mm,~iii ! format: each band contains 1283640 pixels,~iv! operating temperature: 43 K~mechanical cryo-

cooler!,~v! f/4, fully reflective optics,~vi! FOV: 25.7°,~vii ! wavelength resolution: 0.02mm at 3 mm to 0.05

mm at 15.4mm,~viii ! resolving power,l/Dl: .100 in all bands,~ix! NEDT,0.020 K,~x! frame rates up to 30/s, variable integration time,~xi! D* .1013 cm Hz1/2/W at 3mm to D* .231011 cm

Hz1/2/W at 14mm.Assuming successful development of the Hyperspec

QWIP Imager, this instrument is intended to fly on tNASA ER-2 ~U2 Plane! to perform high altitude remotesensing of a variety of earth science parameters.

The presented results indicate that QWIP’s have shomuch progress in recent years in terms of their applicatiin the multiband imaging problem. It is the niche in whicthey have an intrinsic advantage due to the relative easgrowing multiband structures by MBE with very low defedensity.

VI. QWIP APPLICATIONS

Since the initial demonstration of QWIP FPA’s in the la1980s, there has been a worldwide sustained effort in deoping the detector into a mainstream IR technology. TQWIP 2000 workshop held in July 2000 in Dana Point, C~see Infrared Physics and Technology,Vol. 42, No. 3–5,2001! covered many team efforts in bringing the technolointo the commercial market.

The preferable range of QWIP FPA applications resufrom their advantages. The metrics of a FPA are its sensity ~}1/NEDT!, the operating temperature~} TBLIP), and thespeed ~}1/t int) determined by the integration timet int .NEDT ~see Appendix A1! can be given by94

NEDT51

C F2g

N S 111

r D 2

1u2G1/2

, ~33!


e

n-d

ti-u-

-

rto

d

al

ns

of

l-e

sv-

where r is the ratio of the photocurrent to the dark curreand u is the nonuniformity. The first term in the squabrackets is the temporal noise and the second term isfixed pattern noise.

In general, higher quantum efficiency improves allthese figures of merit. The minimum NEDT can be achievby having a smallerg. Higher gain increases the photocurent and, therefore, the speed. However, the resulting higg–r noise reduces the sensitivity. Since the gain is giventhe ratio of the electron lifetime to the transit time, highlifetime increases the photocurrent with respect to the dcurrent~i.e., the value ofr! and therefore is able to increasthe operating temperature (TBLIP is its value whenr 51). Ashort transit time, in turn, increases the photocurrent anddark current in equal proportion, and therefore does notfect the operating temperature. If high speed and high oating temperature are essential, high quantum efficiencygain is required. However, high quantum efficiency doesreduce NEDT once the detection is background limited~e.g.,r .1).

QWIP technology is accepted as the commercial stdard for high-performance and large-format LWIR detectioGunapalaet al.5,87 presented large-format QWIP camerholding forth a promise for many applications in the 6–mm wavelength range in science, medicine, defense, anddustry. Possible applications are medical imaging,5,87,95 firefighting,5,87 monitoring volcano activities,5,87 IRastronomy,5,87,96 national and tactical missile defense,5,87,97

monitoring heat emission in certain geological areas, hypspectral imaging,90 and preventive maintenance. More infomation concerning medicine, astronomy, and defense is gered below.

A. Medicine

Fauciet al.95 described the history of infrared imaging iclinical trials. The advent of QWIP technology, the availabity of high-speed and large-memory personal computers,the shift of clinical modality from anatomical imaging tfunctional imaging open a new area of IR imaging fscreening and diagnosing purposes. High thermal resoluis an essential requirement for medical imaging. The mdemanding applications are those that involve the measment of neuronal control of blood perfusion. The cameperformance for these applications requires a sensitivity

FIG. 43. ~a! This clearly shows the tip of the nose is warmer than tsurrounding tissue due to the enhanced metabolic activity~angiogenesis! ofa skin cancer.~b! Shows a face with no skin cancer on the nose. Usuanose and ears are cooler relative to the other parts of the face, becauseextend out of the body~after Ref. 5!.


sithet oteea

desurma

he-orrorfba

rocea

htne

heicee

ns

hleo

rs

tch

A

m-

s aec-ts.

arntage

red-

WIP

a-in-iontheeen

ongs atar-

i-ign

oleop-

atbyd

el-

en-cep-pli-ndent

rgetace-ionscethethe

tivelynt

b

ladum

ks


about 15 mK measured with af/2 50 mm lens at a 20-mintegration time. QWIP is the only large format detector wgreat stability that fulfills this requirement. Moreover, thcost of a similar HgCdTe array is at least 10 times the cosa comparable QWIP detector. The overall camera syscost will, as a result, increase by a factor of 2, which bcomes very important from the point of view of producingcost-effective medical imaging system.

It has been found that cancer cells exude nitric oxiwhich causes changes of the blood flow in the tissuerounding a cancer that can be detected by a sensitive thesensor. Recently, a midformat QWIP has been used ininstrument called BioScan System for dynamic area teletmometry~DAT!.5,87 DAT involves the accumulation of hundreds of consecutive IR images and fast Fourier transfanalysis of the biomodulation of skin temperature and michomogeneity of skin temperature, which measures the pesion of the skin’s capillaries. The BioScan System canalso used in breast tumor detection, brain surgery, skin ccer detection, and leprosy treatment. In general, cancecells have a high metabolic rate and these cancerousrecruit new blood supply; one of the characteristic of a mlignant lesion. Therefore cancerous tissues are sligwarmer than the neighboring healthy tissues. Thus, a setive thermal imager can easily detect malignant skin canc~see Fig. 43!.

B. Astronomy

QWIP’s are attractive for use in astronomy because toffer good sensitivity, are very stable, are large-size devthat can be easily manufactured, and their operating temptures are relatively high when compared to standard extrisilicon photoconductors used for astronomy goals~e.g.,Si:As!. Because of the narrow spectral range, QWIP’s witcutoff wavelength of 10mm operate at 25–30 K and are abto satisfy stringent dark current requirements, such as thimposed on the Next Generation Space Telescope~NGST!.96

Other important properties of QWIP’s are~i! extremely low1/f noise and~ii ! easy mating to existing CMOS multiplexe

FIG. 44. Comparison of a composite near-infrared image obtained2MASS ~a!, with an 8.5-mm LWIR image~b! obtained with a QWIP FPA atprimary focus of the Palomar 5-m hale telescope. The S106 region dispvigorous star formation obscured behind dense molecular gas and coldand extended nebular emission from dust heated by starlight. Therinfrared imaging can be used to assess the prevalence of warm~'300 K!dusty disks surrounding the stars in such regions. Formation of these disan evolutionary step in the development of planetary systems~after Ref. 5!.


fm-

,r-alnr-

m-u-en-uslls-lysi-rs

ys

ra-ic

a

se

~since QWIP’s are thin-layer structures, thermal mismamay cause certain difficulties; 102431024 pixel devices areplanned!.

The first astronomical observations with a QWIP FPwere described by Gunapalaet al.5,87 An 8–9-mm 2563256 QWIP-based astronomical wide-field multicolor caera built at the Jet Propulsion Laboratory~JPL! has beendesignated for operation at thef/3 prime focus of the Palo-mar 5-m telescope. Operation at the prime focus affordvery wide FOV; 2 minutes of arc compared to 10–30 sonds of arc for traditional ground-based MWIR instrumenFigure 44 compares the composite near-infrared image~a!obtained by 2MASS, with an 8.5mm LWIR image ~b! ob-tained with a QWIP FPA at the primary focus of the Palom5-m hale telescope. These images demonstrate the advaof large format, stable~low 1/f noise! LWIR QWIP FPA’s insurveying obscured regions in search for embedded ordened objects such as young forming stars.

The JPL camera is designated to house up to three Qdevices:96 the existing 2563256 pixel 8.5-mm device, a6403512 pixel, 12.5-mm device, and a 10.3-mm device ofunspecified format. This arrangement will allow simultneous three-color imaging and enable rapid selection ofteresting objects based on their MWIR colors. The selectof three bands is motivated by the properties of bothastronomical objects and the Earth’s atmosphere. It has bfound that many objects enshrouded in dust, exhibit a strsilicate absorption. The strength of this absorption peak9.7 mm. The transmission of the atmosphere is, in turn, chacterized by a strong ozone feature at 9.5–10.0mm. Watervapor produces strong absorption at wavelengths,8 mm andCO2 absorbs wavelengths.13mm. To obtain the best senstivity, a compromise solution has been chosen to desQWIP’s with the peak response at 8.5, 10.3, and 12. 5mm.

It is expected that QWIP’s may have an even greater rto play in space-borne missions because of their highererating temperatures. Examples of such missions~NGST andInterstellar Probe! were described by Ressleret al.96

An impressive program of QWIP FPA’s applicationsthe NASA/Goodard Space Flight Center was presentedJhabvala.90 The future of QWIP’s at Goddard is diverse anstill unfolding. The Hyperspectral QWIP Imager was devoped to perform high altitude remote sensing.

C. Defense

The missile defense systems performed either in thedoatmosphere or exoatmosphere. Endoatmospheric intertors and airborne surveillance sensors used for tactical apcations typically observe warm targets with high backgrouirradiance. Such applications require accurate measuremand subtraction of background irradiance to detect the tasignal. In contrast, exoatmospheric interceptors and spbased surveillance sensors used for strategic applicattypically engage cool targets with low-background irradianlevels. The targets are often far away and unresolved atearly stage of detection. In strategic applications, wherescene is a space background and the targets are at relalow temperatures, LWIR and VLWIR are very importa

y

ysst,

al-

is


pl Ic-reorageaoc

o

uteers.argetingd to

isnotcle

ect-

h-ndeld-

era-cle

is

oreld-dy-

en

leth

py,in

is-ast

de-mu-

is

s/cal

IPLWc

eri-fre-


bands. Improvement in surveillance sensors and interceseekers requires highly uniform large-area, multispectraFPA’s covering the LWIR and VLWIR regions. The restritions imposed on pixel uniformity in strategic IR FPA’s amuch more severe, because dead or degraded pixels ccause the target to be missed completely, and absoluteometry, not imagery, is the method by which most tarinformation is obtained. The space-based FPA’s are usudesigned with larger pixels, roughly equal to the blur spotthe optics. Taking into account the advantages of QWIP tenology ~large and uniform arrays!, the Ballistic Missile De-fense Organization~BMDO! supported extensive efforts tdevelop highly sensitive large FPA’s based on QWIP’s.32,33,97

FIG. 45. Blackbody spectral radiant emittance at various temperatures~afterRef. 5!.

FIG. 46. Image of a Delta-II launch vehicle taken with the LW QWRADIANCE during the launch. This clearly indicates the advantage ofQWIP cameras in the discrimination and identification of a cold-launvehicle in the presence of a hot plume during early stages of launch~afterRef. 5!.


torR

ulddi-tllyfh-

The operating temperature and cooldown time constita major consideration in such systems as missile seekReconnaissance systems, however, require medium to lformat FPA’s and use closed-cycle cooling to reach operatemperatures of 77 K. The extra time and power needelower the temperature~e.g., 65 K for LWIR QWIP operation!is not significant if the integrated dewar/cooler assemblydesigned well. In spite of theoretical speculations, it hasbeen shown conclusively that running a modern closed-cycooler at temperatures below 77 K degrades the life expancy of the cooler.

A QWIP RADIANCE camera was used by the researcers at BMDO in a unique experiment to discriminate aidentify clearly a cold-launch vehicle against its hot plumemanating from rocket engines. The temperature of colaunch vehicles is usually about 250 K, whereas the temptures of the hot plume emanating from the launch vehimay reach 950 K. Figure 45 indicates that the 4-mm photonflux ratio of blackbodies at 250 K and 950 K, respectivelyabout 25 000, whereas the same photon flux ratio at 8.5mmis about 115. Therefore it is very clear that one must expllonger wavelengths for better discrimination between cobody and hot plume because the highest instantaneousnamic range of IR cameras is usually 12 bits~i.e., 4096! orless. Figure 46 shows the image of Delta-II launch takwith a QWIP RADIANCE camera.

D. High-frequency detection

Certain applications that were previously not feasibwith HgCdTe detectors, the maximum electrical bandwidof which remains approximately at 3 GHz,7,98 are now pos-sible with QWIP’s. When a CO2-laser local oscillator~LO!with HgCdTe detector is used for molecular spectroscofor example, the molecular line of interest must be withroughly 3 GHz of a CO2 emission line. Because these emsion lines are typically separated by at least 30 GHz, the vmajority of the frequency space between 9 and 11mm isundetectable. A similar problem occurs when a HgCdTetector is used as a front-end mixer in coherent laser comnications, where the maximum digital transmission rateapproximately 1 Gb/s.

In comparison with HgCdTe-based technology, GaAAlGaAs QWIP’s offer several advantages: higher electri

h

FIG. 47. Schematic of the IR heterodyne and microwave mixing expment. The measured signals, in general, consist of all possible mixingquencies of two infrared and one microwave frequency~after Ref. 100!.


ls

to

, a

its

Ilymndth

igt

eclinren-

-

het-df

iors

till

PL,

rcialee-isedis

ofial, a

i-e is,ess

llyioncast

y

theiveies,es-

aptheus-ete

at

ns.m-the

-rll-cro

d


bandwidth, greater robustness and tolerance to high leveLO power, and monolithic compatibility with GaAs HEMTamplifiers.99,100The intrinsic high-speed capability is duethe inherently short carrier lifetimet'5 ps.

In optical mixing experiments described in Ref. 101CO2 laser and a lead-salt tunable diode laser~TDL! wereused as the sources of IR radiation. The CO2 laser frequencywas fixed, while the TDL frequency was tuned by varyingoperating temperature. Liuet al.102 carried out experimentswith an intermediate frequency~IF! of up to 26.5 GHz.

Far beyond 26.5 GHz a direct measurement of thesignal is very difficult because of two impediments; namethe lack of a sufficiently sensitive preamplifier-analyzer cobination and the difficulty in making a low-loss, broadbacircuit that can be used to couple the weak IF signal fromQWIP. To extend the frequency range, Liuet al.103 used amicrowave mixing technique shown schematically in F47. The technique is based on the nonlinearity present inI–V curves of QWIP’s associated with the transport of eltrons in the absence of IR radiation. Because of this nonearity, the microwave signal is rectified and the measuchange of the bias voltage is the response. As is showFig. 47, apart from two IR beams~as in conventional hetero

FIG. 48. Direct infrared heterodyne~squares! and mixed heterodyne frequency with microwave frequency~dots! signal vs heterodyne frequency foa bias voltage of 2 V. The dashed and dotted lines show the expected robehavior. The incident power from the two infrared sources and the miwave source is normalized to about 0.2 and 0.3 mW, respectively. Thevice temperature is 80 K~after Ref. 100!.

FIG. 49. Sensors and systems revenues from 1999 to 2004~after Ref. 104!.


of

F,-

e

.he--din

dyne experiments!, an additional microwave excitation is applied to the QWIP. Letf IR1 and f IR2 denote the two IRfrequencies andf mwave the microwave frequency. The IRheterodyne frequency is thenf het5u f IR12 f IR2u, while thedownconverted signal frequency isu f het2 f mwaveu. In thisheterodyne detection technique very highf het frequencies upto 82.16 GHz have been achieved. The measured opticalerodyne signal~squares! and the up- and downconvertemixing signal ~dots! are shown in Fig. 48 as a function oheterodyne frequency (f het) for a bias voltage of 2 V. Thedashed and dotted lines are the expected roll-off behavdue to the resistance-capacitance~RC! time constant of thedevice and the photocarrier lifetime.

E. Market data

From a market perspective, the QWIP technology is sin its infancy, although it is on the brink of acceptance.104

The QWIP detector process methodology developed at Jculminating in the fabrication of midformat (2563256), por-table LWIR cameras, has been transferred to the commeenvironment through QWIP Technologies LLC under agrment with Caltech. At present, QWIP Technologies LLCinvolved in high-volume III-V-material processes developfor the cellular communications industry. The first productthe QWIP-Chip, a 3203256 format QWIP FPA spectrallycentered at 8.5mm with a 30-mm pitch. Production rampedup to 100 QWIP-Chips per month by the first quarter2001. Also QWIP Technologies has begun low rate initproduction of its second product in the QWIP-Chips line6403512 QWIP FPA.

The high-performance military market is currently domnated by other, more expensive technologies. This stanchowever, changing in favor of larger staring arrays and lexpensive imagers, like those using QWIP FPA’s.

In the U.S. alone, billions of dollars are spent annuaon various types of imaging systems. In 2001, over $2 billwas spent on IR imaging systems, and the market is foreto grow by 31% annually between 1999 and 2004~see Fig.49!. In 1996, IR FPA market was predominantly militar~80% compared to 20% commercial!. It is expected tochange to 66% military and 34% commercial by 2004.

VII. CONCLUSIONS

The intention of this paper has been to compareachievements of QWIP technology with those of competittechnologies, with the emphasis on the material propertdevice structure, and their impact on FPA performance,pecially in LWIR and VLWIR spectral regions.

At present, HgCdTe is the most widely used variable-gsemiconductor that has a privileged position both inLWIR as well as VLWIR spectral ranges. The above discsion leads to a conclusion that LWIR QWIP cannot compwith HgCdTe photodiode as single devices, especiallyhigher operation temperatures~.70 K!, due to the funda-mental limitations associated with intersubband transitioHowever, the advantage of HgCdTe is less distinct at teperatures below 50 K due to the problems inherent in

off-e-


a-bi

lo

oc

a

nd

isona

us-icn

ityIRhfo

itloelrgesar-ne

s

IRoarmowr

lyhoDgeey

arexe-

o-c-on-

etac-

in-ard

lo-elypli-ectorsen-canre.m-

eci-

t-veise

lec-llytemde-

r

atd.for

an-torialelyandnce.


HgCdTe material~p-type doping, Shockley-Read recombintion, trap-assisted tunneling, surface and interface instaties!.

Comparing photovoltaic HgCdTe and QWIP technogies, we arrive at the following conclusions:

~i! Two major issues that impede the performanceQWIP’s should be overcome: optical conversion efficienand dark current.

~ii ! In the case of HgCdTe, the improvement in the arruniformity is necessary.

~iii ! QWIP seems to have more potential for LWIR aVLWIR FPA operation~also with multicolor detection!.

The main drawback of LWIR QWIP FPA technologythat the device performance is limited to long integratitime applications and low operating temperature. The madvantages are linked to the performance uniformity andthe ease of fabrication of large size arrays. The large indtrial infrastructure in III-V material/device growth, procesing, and packaging brought about by the extensive appltion of GaAs-based devices in the telecommunicatioindustry gives QWIP’s a potential advantage in producibiland cost. The only known use of HgCdTe, to date, isdetectors. The main drawback of LWIR HgCdTe FPA tecnology is the unavailability of large size arrays necessaryTV and larger format.

Even though QWIP is a photoconductor, several ofproperties, such as high impedance, fast response time,integration time, and low power consumption, comply wwith the requirements imposed on the fabrication of laFPA’s. Due to a high material quality at low temperaturQWIP has potential advantages over HgCdTe in the areVLWIR FPA applications in terms of the array size, unifomity, yield, and cost of the systems. QWIP FPA’s combithe advantages of PtSi Schottky barrier arrays~high unifor-mity, high yield, radiation hardness, large arrays, lower co!with the advantages of HgCdTe~high quantum efficiencyand long-wavelength response!. It is also anticipated that theinherent QWIP polarization sensitivity can be exploited inimaging systems to discriminate between targets and dec

The performance figures of merit of state-of-the-QWIP and HgCdTe FPA’s are similar because the main litations come from the readout circuits. Performance is, hever, achieved with very different integration times. The veshort integration time of LWIR HgCdTe devices~typicallybelow 300ms! is very useful to freeze a scene with rapidmoving objects. Due to excellent homogeneity and low ptelectrical gain, QWIP devices achieve an even better NEthe integration time, however, must be 10–100 times lonand typically is between 5 and 20 ms. The choice of the btechnology is therefore driven by the specific needs of a stem. Observation of the global market for the last few yeindicates that even though LWIR HgCdTe photodiodeshibit intrinsically higher performance than that of QWIP dtectors, the market tendencies for the future are~i! HgCdTefor small formats~e.g., 1283128), small pitch, high framerates, and low integration times, and~ii ! QWIP for largeformats ~e.g., 6403480 and larger!, low frame rates, andlarge integration time.


li-

-

fy

y

intos-

a-s

-r

sngle,of

t

ys.ti--

y

-T;r

sts-s-

Powerful possibilities of the QWIP technology are assciated with VLWIR FPA applications and multicolor detetion. Three-band and four-band FPA’s have been demstrated.

Until now, efforts in the area of QWIP technology havbeen limited to increasing the operation temperature fortical applications. The discussion presented in this paperdicates, however, that more effort should be directed towdeveloping QWIP for VLWIR strategic applications.

Despite serious competition from alternative technogies and progress slower than expected, HgCdTe is unlikto be seriously challenged as far as high-performance apcations~requiring multispectral capability and fast respons!are concerned. The recent successes of competing detecooled cryogenically are due to technological, not fundamtal issues. The steady progress in epitaxial technologymake HgCdTe devices much more affordable in the futuThe much higher operation temperature of HgCdTe, copared to low-dimensional solid devices, may become a dsive argument in this case.

APPENDIX

1. Noise equivalent difference temperature

Detectivity provides only a limited insight into the suiability of IR detectors for specific IR systems. The objecticriterion of performance of imaging devices is the noequivalent difference temperature~NEDT!; i.e., the differ-ence in temperature of an object required to produce an etric signal equal to the rms noise voltage. While normathought of as a system parameter, detector NEDT and sysNEDT are the same except for system losses. NEDT isfined as

NEDT5Vn~]T/]Q!

~]Vs /]Q!5Vn

DT

DVs, ~A1!

whereVn is the rms noise andDVs is the signal measured fothe temperature differenceDT. It can be shown that105

NEDT5~tChBLIPANw!21, ~A2!

whereC is the thermal contrast which in the MWIR band300 K is 3.5%–4% compared to 1.6% for the LWIR banNw is the number of photogenerated carriers integratedone integration timet int :

Nw5hAdt intQB . ~A3!

The percentage of BLIP,hBLIP , is simply the ratio of photonnoise to composite FPA noise:

hBLIP5S Nphoton2

Nphoton2 1NFPA2

D 1/2

. ~A4!

It results from the above formulas that the charge hdling capacity of the readout, the integration time linkedthe frame time, and dark current of the sensitive matebecome the major issues of IR FPA’s. The NEDT is inversproportional to the square root of the integrated charge,therefore the greater the charge, the higher the performa


’seat

fsed

nly

-te

iat

des

d

le

-

aa

ra

e-

erter-

ofit

cal

wn

tore-

c-


The distinction between integration time and FPAframe time must be noted. At high backgrounds it is oftimpossible to handle the large amount of carriers generover a frame time compatible with standard video rates. OFPA frame integration can be used to attain a level of sensensitivity that is commensurate with the detector-limitD* and not the charge-handling-limitedD* .

a. Photon detectors

The sensitivity of an IR system can be expressed ivariety of ways.106 In this paper we follow the paper recentpublished by Kinch.20

The limiting factor of detector sensitivity is the fluctuation in the relevant carrier concentration since this demines the minimum observable signal. This fluctuationenvisioned by assuming that the detector current is integron a capacitive node, giving a variance

N1/25gF ~ I d1I B!t int

q G1/2

, ~A5!

whereg is the gain,I d is the detector dark current,I B is thebackground flux current, andt int is the integration time.

The noise equivalent flux~also named as the minimumobservable signal! is given by

gDFht intA5N1/2, ~A6!

whereh represents the overall quantum efficiency of thetector, including the internal quantum efficiency. Using latwo equations, the noise equivalent flux can be describe

DF5g~11I d /I F!FB

N1/2. ~A7!

Because

DF5DT

dFB /dT, ~A8!

if we substitute the above, we have for the noise equivadifference temperature

NEDT5g~11I d /I F!FB

~dFB /dT!N1/2, ~A9!

where (dFB /dT)/FB5C is the scene contrast. For unlimited available capacity,N1/2 varies asg, andDF and NEDTare independent of the gaing. However, for a limited maxi-mum well capacityN, both bandwidth~Df! and NEDT willvery asg. These parameters depend on specific systemdetector quantities such as area and bandwidth, where bwidth is defined asD f 51/2t.

b. Thermal detectors

In the case of a thermal detector, the limiting tempeture fluctuation is given by16,18

DTn5S 4kT2RthD f

11v2t th2 D 1/2

. ~A10!


nedf-or

a

r-sed

-tas

nt

ndnd-

-

At low frequenciesDTn5(4kT2RthD f )1/2. It appears thatcorrelated double sampling of this limitation with an intgrated timet int , gives20

DTn5F2kT2@12exp~2t int /RthCth!#

CthG1/2

, ~A11!

where it is assumedD f 51/RthCth . Comparing Eq.~A11!with a signal fluctuation change given by

DTs5S dP

dTDDTF12expS 2t int

RthCthD GRth , ~A12!

one obtains

NEDT5F 2kT2

Cth@12exp~2t int /RthCth!#G1/2 1

~dP/dT!Rth,

~A13!

where (dP/dT) is the differential change in radiated powper scene temperature change in the spectral region of inest. The integration time should fulfill the conditiont int

'2Rth Cth to avoid image smearing. The maximum valueRth is given by the radiative coupling with the upper limequal 43108 K/W for a 1 mil pixel size. For micromachinedresonant cavity bolometers withf/1 optics, there is a finitelimit of detector thickness. Typicallyt53000 Å. For 1 milpixel size a volumetric specific heat is about 2 J/cm3 K.

2. GaAs ÕAlGaAs QWIP’s

a. Thermal generation

The carrier generation rate from the QW’s due to optiphonons may be obtained as107

Gphon5NQW

ns2

t, ~A14!

whereNQW is the number of QW’s,ns2 is the 2D density ofcarriers originating from the excited subband state, andt isthe lifetime of electrons in the upper state. It can be shothat

ns25Lp2S 2pm* kT

h2 D 3/2

expS 2Vb2EF

kT D , ~A15!

t54hepDEcoLp

q2EphonI 1

. ~A16!

In these equations,Lp is the mean well spacing,m* is theelectron effective mass,k is the Boltzmann constant,Ephon isthe phonon energy, andI 1 is a dimensionless integral close2 in the 8–12mm range for a quasibound upper state. Moover, DEcois the cutoff energy defined asDEco5hc/lc

5Vb2E1 (Vb is the barrier energy of the QW, andlc is thecutoff wavelength!, 1/ep51/e`21/e0 (e` and e0 is the di-electric permittivity at infinite and zero frequency, respetively!,

EF2E15ns1h2

4pm*, ~A17!

is the Fermi energy of the well, andns1 is the 2D carrierdensity in the ground state. It should be noted thatGphon


al

f 2d

ze

twffi

o-

nhetin

.

aru

on

e

nshen

mr

taleadature

s

ofs,ble

esra-

ityd,ulkofn

lec-at

-

vel

rr


does not depend onLp and is therefore a more fundamentquantity of a QWIP, in contrast tons2 andt. The followingparameters have been assumed:m* 50.067m0, ep563e0,Ephon536 meV,ns15231011 cm22, and NQW550.

b. Quantum efficiency

An unpolarized double-pass quantum efficiency~at a 45°angle of incidence! is equal to

h512exp~22a l !

2, ~A18!

wherel is the total active superlattice length. The factor oin the exponent of Eq.~A18! accounts for a double pass anthe factor 2 in the denominator accounts for the unpolaribeam.

As Levine has discussed, the behavior ofh results fromthe fact that the net quantum efficiency is composed ofcomponents. One is the optical absorption quantum eciencyha ~e.g., the fraction of the unpolarized incident phtons which are absorbed by the intersubband transition!. Theother factor is the probabilitype that a photoexcited electrowill escape from the quantum well and contribute to tphotocurrent, rather than being recaptured by the originawell. That is, we can express the total~net! quantum effi-ciencyh as

h5hape . ~A19!

For bound-to-continuum transitions,pe varies from 34% to52% at low bias and increases toward unity at high bias

3. HgCdTe photodiodes

The Auger mechanism is more likely to impose fundmental limitations to the LWIR HgCdTe detectoperformance.108 Assuming a nondegenerate statistic, the Ager generation rate is equal to

GA5n

2tA1i

1p

2tA7i

51

2tA1i S n1

p

g D , ~A20!

wheren andp are the electron and hole concentrations,tA1i

andtA7i are the intrinsic Auger 1 and Auger 7 recombinati

times, andg5tA7i /tA1

i is the ratio of Auger 7 an Auger 1intrinsic recombination times.tA1

i is given by

tA1i 5

3.8310218es2~11m!1/2~112m!expF S 112m

11m D Eg

kTG~me* /m!uF1F2u2~kT/Eg!3/2

,

in s, ~A21!

where m5me* /mh* is the ratio of the conduction to thheavy-hole valence-band effective masses,es is the relativestatic dielectric constant, andF1 andF2 are the overlap in-tegrals of the periodic part of the electron wave functioThe value ofuF1F2u is taken as a constant equal to 0.2. Tg term is of high uncertainty. According to Casselman aPetersen109 for Hg12xCdxTe over the range 0.16<3<0.40and 50 K<T<300 K, 3<g<6.

The Auger-dominated detectivity is equal to


d

o-

g

-

-

.

d

D* 5l

21/2hc

h

t1/2S tA1i

n1p/g D 1/2

. ~A22!

The resulting Auger generation achieves its minimufor p5g1/2ni . Sinceg.1, the highest detectivity of Augelimited photodetectors can be achieved with a lightp-typedoping.

The requiredg1/2ni p-type doping is difficult to achievein practice for low-temperature photodetectors~the control ofhole concentration below the 531015 cm23 level is difficult!and thep-type material suffers from some nonfundamenlimitations such as contacts, surface, and Shockley–Rprocesses. These are the reasons why the low-temperdetectors are typically produced from lightly dopedn-typematerials. In contrast,p-type doping is clearly advantageoufor LWIR and near room-temperature detectors.

Rogalski and Ciupa39 have compared the performancen12p and P12n LWIR HgCdTe photodiodes. It appearthat for the lowest doping levels achievable in a controllamanner in the base regions of photodiodes (Na5531015

cm23 for n1-p structure, andNd5531014 cm23 for P1

2n structure! the performance of both types of photodiodis comparable for a given cutoff wavelength and tempeture.

4. InSb photodiodes

InSb photodiodes are generally fabricated by impurdiffusion and ion implantation. Epitaxy is not used; insteathe standard manufacturing technique begins with bn-type single-crystal wafers with a donor concentrationabout Nd51015 cm23. The Shockley–Read recombinatiocenters are an intrinsic characteristic ofn-type InSb. Theelectron and hole lifetimes are equal because inn-type ma-terial there is a single set of recombination centers. The etron and hole carrier lifetimes of high-quality materialtemperatures below 130 K obey the expression110

t54.43108

Nd, in s ~A23!

The thermal generation can be described by

Gth5ni

2

Ndt. ~A24!

5. Extrinsic photoconductors

At the low temperature of operation of impurity photoconductors~when kT!Ed and n!Nd , Na), the thermalequilibrium free-charge carrier in ann-type extrinsic semi-conductor with a partially compensated singly ionized leis equal to111,112

nth5nma j5Nc

2 S Nd2Na

NaDexpS 2

Ed

kTD . ~A25!

HereNc is the density of states in the conduction band,Nd isthe donor concentration,Na is the compensating acceptoconcentration, andEd is the bonding energy of the donorelative to the conduction band.


n--

thl

taffi.

MD.ci

ca

t

R

pl.

s,

. E

s-n

y

nd

r,oc.

ylor,

. F.

ke,

,

z-

De

T.si,

T.

-

. K.vel,.

M.A.

ns,

on,

.-

nd,on,

R.

D.on

. A.E

.ban-

, P.IE

K.


The majority carrier lifetime is determined by the desity of empty ~ionized! donor levels and, for low temperatures, such thatnma j,Na,Nd , is given by

t5~scv thNa!21, ~A26!

wheresc is the capture cross section for electrons intodonor level, andv th5(8kT/pm* )1/2 is the carrier thermavelocity. For shallow-level impurities~B and As! typically,sc'10211 cm2, while for the deep-level impurities~In, Au,Zn! show,sc510213 cm2. By comparison, thesc of intrin-sic photoconductors is about 10217 cm2.

Practical values ofa for optimized extrinsic silicon pho-toconductors are in the range 10–50 cm21. To maximize thequantum efficiency, the thickness of the detector crysshould be not less than about 0.1 cm. Typical quantum eciencies are in range of 10%–50% at the response peak

1L. Esaki and R. Tsu, IBM J. Res. Dev.14, 61 ~1970!.2B. F. Levine, J. Appl. Phys.74, R1 ~1993!.3S. D. Gunapala and K. M. S. V. Bandara, inThin Films, edited by M. H.Francombe and J. L. Vossen~Academic Press, New York, 1995!, Vol. 21,p. 113.

4S. D. Gunapala and S. V. Bandara, inHandbook of Thin Devices,edited byM. H. Francombe~Academic Press, San Diego, 2000!, Vol. 2, p. 63.

5S. D. Gunapala, S. V. Bandara, J. K. Liu, E. M. Luong, S. B. Rafol, J.Mumolo, D. Z. Ting, J. J. Bock, M. E. Ressler, M. W. Werner, P.LeVan, R. Chehayeb, C. A. Kukkonen, M. Ley, P. LeVan, and M. A. FauOpto-Electron. Rev.8, 150 ~2001!.

6M. Razeghi, Opto-Electron. Rev.6, 155 ~1998!.7A. Rogalski,Infrared Detectors~Gordon and Breach, Amsterdam, 2000!.8A. Rogalski, K. Adamiec, and J. Rutkowski,Narrow-Gap SemiconductorPhotodiodes~SPIE Press, Bellingham, 2000!.

9A. Rogalski, Sens. Mater.12, 233 ~2000!.10R. A. Wood and N. A. Foss, Laser Focus World, 101~1993!.11P. W. Kruse,Uncooled Thermal Imaging. Arrays, Systems, and Appli

tions ~SPIE Press, Bellingham, 2001!.12A. Rose, Concepts in Photoconductivity and Allied Problems~Inter-

science, New York, 1963!.13J. Piotrowski and W. Gawron, Infrared Phys. Technol.38, 63 ~1997!.14J. Piotrowski and A. Rogalski, Sens. Actuators A67, 146 ~1998!.15R. G. Humpreys, Infrared Phys.26, 337 ~1986!.16P. W. Kruse, L. D. McGlauchlin, and R. B. McQuistan,Elements of In-

frared Technology~Wiley, New York, 1962!.17P. W. Kruse, inOptical and Infrared Detectorsedited by R. J. Keyes

~Springer-Verlag, Berlin, 1977!, p. 5.18A. Smith, F. E. Jones, and R. P. Chasmar,The Detection and Measuremen

of Infrared Radiation~Clarendon, Oxford, 1968!.19R. Ciupa and A. Rogalski, Opto-Electron. Rev.5, 257 ~1997!.20M. A. Kinch, J. Electron. Mater.29, 809 ~2000!.21T. Ashley and C. T. Elliott, Electron. Lett.21, 451 ~1985!.22F. Fuchs, L. Bu¨rkle, R. Hamid, N. Herres, W. Pletschen, R. E. Sah,

Kiefer, and J. Schmitz, Proc. SPIE4288, 171 ~2001!.23J. Piotrowski and A. Rogalski, J. Appl. Phys.80, 2542~1996!.24B. F,. Levine, K. K. Choi, C. G. Bethea, J. Walker, and R. J. Malik, Ap

Phys. Lett.50, 1092~1987!.25M. B. Reine, inInfrared Detectors and Emitters: Materials and Device

edited by P. Capper and C. T. Elliott~Chapman and Hall, London, 2000!,p. 279.

26J. Y. Andersson, L. Lundqvist, and Z. F. Paska, Appl. Phys. Lett.58, 2264~1991!.

27G. Sarusi, B. F. Levine, S. J. Pearton, K. M. S. V. Bandara, and RLeibenguth, Appl. Phys. Lett.64, 960 ~1994!.

28C. J. Chen, K. K. Choi, M. Z. Tidrow, and D. C. Tsui, Appl. Phys. Lett.68,1446 ~1997!.

29S. S. Li and Y. H. Wang, inQuantum Well Intersubband Transition Phyics and Devices,edited by H. C. Liu, B. F. Levine, and J. Y. Anderso~Kluwer Academic, Dordrecht, 1994!, p. 29.

30M. A. Kinch and A. Yariv, Appl. Phys. Lett.55, 2093~1989!.31A. Rogalski, Infrared Phys. Technol.38, 295 ~1997!.32M. Z. Tidrow, in Proceedings of the 6th Annual AIAA/BMDO Technolog


e

l-

.

,

-

.

.

Readiness Conference and Exhibit, San Diego, CA, 1997.33M. 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, aW. R. Dyer, Opto-Electron. Rev.7, 283 ~1999!.

34A. Rogalski, Infrared Phys. Technol.40, 279 ~1999!.35A. C. Goldberger, S. W. Kennerly, J. W. Little, H. K. Pollehn, T. A. Shafe

C. L. Mears, H. F. Schaake, M. Winn, M. Taylor, and P. N. Uppal, PrSPIE4369, 532 ~2001!.

36D. W. Beekman and J. Van Anda, Infrared Phys. Technol.42, 323 ~2001!.37M. Z. Tidrow, Proc. SPIE2999, 109 ~1996!.38P. R. Norton, Proc. SPIE3379, 102 ~1998!.39A. Rogalski and R. Ciupa, J. Appl. Phys.77, 3505~1995!.40A. Rogalski and R. Ciupa, J. Appl. Phys.80, 2483~1996!.41S. M. Johnson, D. R. Rhiger, J. P. Rosbeck, J. M. Peterson, S. M. Ta

and M. E. Boyd, J. Vac. Sci. Technol. B10, 1499~1992!.42M. C. Chen, R. S. List, D. Chandra, M. J. Bevan, L. Colombo, and H

Schaake, J. Electron. Mater.25, 1375~1996!.43R. S. List, J. H. Tregilgas, A. M. Turner, J. D. Beck, and J. C. Ehm

Proc. SPIE2228, 274 ~1994!.44W. E. Tennant, C. A. Cockrum, J. B. Gilpin, M. A. Kinch, M. B. Reine

and R. P. Ruth, J. Vac. Sci. Technol. B10, 1359~1992!.45M. B. Reine, K. R. Maschhoff, S. P. Tobin, P. W. Norton, J. A. Mroc

kowski, and E. E. Krueger, Semicond. Sci. Technol.8, 788 ~1993!.46J. Bajaj, J. M. Arias, M. Zandian, J. G. Pasko, L. J. Kozlowski, R. E.

Wames, and W. E. Tennant, J. Electron. Mater.24, 1067~1995!.47S. D. Gunapala, J. K. Liu, J. S. Park, M. Sundaram, C. A. Shott,

Hoelter, T. L. Lin, S. T. Massie, P. D. Maker, R. E. Muller, and G. SaruIEEE Trans. Electron Devices44, 51 ~1997!.

48M. Z. Tidrow, J. C. Chiang, S. S. Li, and K. Bacher, Appl. Phys. Lett.70,859 ~1997!.

49A. Singh and M. O. Manasreh, Proc. SPIE2397, 193 ~1995!.50S. V. Bandara, S. D. Gunapala, S. Rafol, D. Ting, J. Liu, J. Mumolo,

Trinh, A. W. K. Liu, and J. M. Fastenau, Infrared Phys. Technol.42, 237~2001!.

51L. J. Kozlowski, K. Vural, J. M. Arias, W. E. Tennant, and R. E. DeWames, Proc. SPIE3182, 2 ~1997!.

52P. Norton, J. Campbell, S. Horn, and D. Reago, Proc. SPIE4130, 226~2000!.

53J. D. Benson, J. H. Dinan, J. Waterman, C. J. Summers, R. D. Benz, BWagner, S. D. Pearson, A. Parikh, J. J. Jensen, O. K. Wu, R. D. RajaG. S. Kamath, K. A. Harris, S. R. Jost, J. M. Arias, L. J. Kozlowski, MZandian, J. Bajaj, K. Vural, R. E. DeWames, C. A. Cockrum, G.Venzor, S. M. Johnson, H. D. Shih, M. J. Bevan, J. A. Dodge, andSimmons, Proc. SPIE2744, 126 ~1996!.

54O. K. Wu, T. J. DeLyon, R. D. Rajavel, and J. E. Jensen, inNarrow-GapII –VI Compounds for Optoelectronic and Electromagnetic Applicatioedited by P. Capper~Chapman and Hall, London, 1997!, p. 97.

55T. J. DeLyon, J. E. Jensen, M. D. Gorwitz, C. A. Cockrum, S. M. Johnsand G. M. Venzor, J. Electron. Mater.28, 705 ~1999!.

56W. E. Tennant, M. Thomas, L. J. Kozlowski, W. V. McLevige, D. DEdwall, M. Zandian, K. Spariosu, G. Hildebrandt, V. Gil, P. Ely, M. Muzilla, A. Stoltz, and J. H. Dinan, J. Electron. Mater.30, 590 ~2001!.

57S. M. Johnson, J. L. Johnson, W. J. Hamilton, D. B. Leonard, T. A. StraA. E. Patten, J. M. Peterson, J. H. Durham, V. K. Randall, T. J. de LyJ. E. Jensen, and M. D. Gorwitz, J. Electron. Mater.29, 680 ~2000!.

58W.A. Beck and T.S. Faska, Proc. SPIE2744, 193–206~1996!.59S. D. Gunapala, J. S. Park, G. Sarusi, T. L. Lin, J. K. Liu, P. D. Maker,

E. Muller, C. A. Shott, and T. Hoelter, IEEE Trans. Electron Devices44,45 ~1997!.

60S. D. Gunapala, S. V. Bandara, J. K. Liu, W. Hong, M. Sundaram, P.Maker, R. E. Muller, C. A. Shott, and R. Carralejo, IEEE Trans. ElectrDevices45, 1890~1998!.

61S. D. Gunapala, S. V. Bandara, J. K. Liu, E. M. Luong, N. Stetson, CShott, J. J. Bock, S. B. Rafol, J. M. Mumolo, and M. J. McKelvey, IEETrans. Electron Devices47, 326 ~2000!.

62H. Schneider, M. Walther, C. Scho¨nbein, R. Rehm, J. Fleissner, WPletschen, J. Braunstein, P. Koidl, G. Weimann, J. Ziegler, and W. Caski, Physica E~Amsterdam! 7, 101–107~2000!.

63H. Schneider, M. Walther, J. Fleissner, R. Rehm, E. Diwo, K. SchwarzKoidl, G. Weimann, J. Ziegler, R. Breiter, and W. Cabanski, Proc. SP4130, 353 ~2000!.

64H. Schneider, P. Koidl, M. Walther, J. Fleissner, R. Rehm, E. Diwo,Schwarz, and G. Weimann, Infrared Phys. Technol.42, 283 ~2001!.


P.

no

w

ra

IE

W.rd,

r,

R.ed

ngg,

. Arys

. AM

. L,

.V

, Aer.

L.

al

M

D.A.

B.

er,

M.n,

S.

W.

A.

s.

,

.

iu,


65H. Schneider, C. Scho¨nbein, M. Walther, K. Schwarz, J. Fleissner, andKoidl, Appl. Phys. Lett.71, 246 ~1997!.

66P. Bois, E. Costard, X. Marcadet, and E. Herniou, Infrared Phys. Tech42, 291 ~2001!.

67A. Manissadjian, P. Costa, P. Tribolet, and G. Destefanis, Proc. SPIE3436,150 ~1998!.

68G. Destefanis, P. Audebert, E. Mottin, and P. Rambaud, J. Cryst. Gro184Õ185, 1288~1998!.

69A. Singh and D. A. Cardimona, Proc. SPIE2999, 46 ~1997!.70D. A. Cardimona, A. Singh, D. Huang, C. Morath, and P. Varangis, Inf

red Phys. Technol.42, 211 ~2001!.71L. J. Kozlowski, S. A. Cabelli, D. E. Cooper, and K. Vural, Proc. SP

1946, 199 ~1993!.72L. J. Kozlowski, Proc. SPIE3179, 200 ~1997!.73L. J. Kozlowski, R. B. Balley, S. E. Cooper, I. S. Gergis, A. C. Chen,

V. McLevige, G. L. Bostrup, K. Vural, W. E. Tennant, and P. E. HowaOpt. Eng.~Bellingham! 33, 54 ~1994!.

74J. P. Chatard, Proc. SPIE3698, 407 ~1999!.75F. Bertrand, J.T. Tissot, and G. Destefanis, inPhysics of Semiconducto

Devices, edited by V. Kumar and S.K. Agarwal~Narosa Publishing HouseNew Delhi, 1998!, Vol. II, p. 713.

76J. M. Fastenau, W. K. Liu, X. M. Fang, D. I. Lubyshev, R. I. Pelzel, T.Yurasits, T. R. Stewart, J. H. Lee, S. S. Li, and M. Z. Tidrow, InfrarPhys. Technol.42, 407 ~2001!.

77S. D. Gunapala, S. V. Bandara, A. Singh, J. K. Liu, B. Rafol, E. M. LuoJ. M. Mumolo, N. Q. Tran, D. Z. Ting, J. D. Vincent, C. A. Shott, J. Lonand P. D. LeVan, IEEE Trans. Electron Devices47, 963 ~2000!.

78R. D. Rajavel, D. M. Jamba, O. K. Wu, J. E. Jensen, J. A. Wilson, EPatten, K. Kosai, P. Goetz, G. R. Chapman, and W. A. Radford, J. CGrowth 175, 653 ~1997!.

79R. D. Rajavel, D. M. Jamba, J. E. Jensen, O. K. Wu, P. D. Brewer, JWilson, J. L. Johnson, E. A. Patten, K. Kasai, J. T. Caulfield, and P.Goetz, J. Cryst. Growth184, 1272~1998!.

80R. D. Rajavel, D. M. Jamba, J. E. Jensen, O. K. Wu, J. A. Wilson, JJohnson, E. A. Patten, K. Kasai, P. M. Goetz, and S. M. JohnsonElectron. Mater.27, 747 ~1998!.

81M. B. Reine, P. W. Norton, R. Starr, M. H. Weiler, M. Kestigian, B. LMusicant, P. Mitra, T. Schimert, F. C. Case, I. B. Bhat, H. Ehsani, andRao, J. Electron. Mater.24, 669 ~1995!.

82P. Mitra, S. L. Barnes, F. C. Case, M. B. Reine, P. O’Dette, R. StarrHairston, K. Kuhler, M. H. Weiler, and B. L. Musicant, J. Electron. Mat26, 482 ~1977!.

83M. B. Reine, A. Hairston, P. O’Dette, S. P. Tobin, F. T. J. Smith, B.Musicant, P. Mitra, and F. C. Case, Proc. SPIE3379, 200 ~1998!.

84H. K. Pollehn and J. Ahearn, Proc. SPIE3698, 420 ~1999!.85J. P. Zanatta, P. Ferret, R. Loyer, G. Petroz, S. Cremer, J. P. Chamon

Bouchut, A. Million, and G. Destefanis, Proc. SPIE4130, 441 ~2000!.86T. Whitaker, Compound. Semicond.5„7…, 48 ~1999!.87S. D. Gunapal, S. V. Bandara, J. K. Liu, E. M. Luong, S. B. Rafol, J.


l.

th

-

,

.t.

..

.J.

.

.

, P.

.

Mumolo, D. Z. Ting, J. J. Bock, M. E. Ressler, M. W. Werner, P.LeVan, R. Chehayeb, C. A. Kukkonen, M. Levy, P. LeVan, and M.Fauci, Infrared Phys. Technol.42, 267 ~2001!.

88M. Sundaram, S. C. Wang, M. F. Taylor, A. Reisinger, G. L. Milne, K.Reiff, R. E. Rose, and R. R. Martin, Infrared Phys. Technol.42, 301~2001!.

89A. Goldberg, T. Fischer, S. Kennerly, W. Beck, V. Ramirez, and K. GarnInfrared Phys. Technol.42, 309 ~2001!.

90M. Jhabvala, Infrared Phys. Technol.42, 363 ~2001!.91M. B. Reine, Proc. SPIE4288, 266 ~2001!.92Ph. Bois, E. Costard, J. Y. Duboz, and J. Nagle, Proc. SPIE3061, 764

~1997!.93S. D. Gunapala, S. V. Bandara, A. Sigh, J. K. Liu, S. B. Rafol, E.

Luong, J. M. Mumolo, N. Q. Tran, J. D. Vincent, C. A. Shott, J. P. LeVaProc. SPIE3698, 687 ~1999!.

94K. K. Choi, Proc. SPIE~to be published!.95M. A. Fauci, R. Breiter, W. Cabanski, W. Fick, R. Koch, J. Ziegler, and

D. Gunapala, Infrared Phys. Technol.42, 337 ~2001!.96M. E. Ressler, J. J. Bock, S. V. Bandara, S. D. Gunapala, and M.

Werner, Infrared Phys. Technol.42, 377 ~2001!.97M. Z. Tidrow and W. R. Dyer, Infrared Phys. Technol.42, 333 ~2001!.98I. Melngailis, W. E. Keicher, C. Freed, S. Marcus, B. E. Edwards,

Sanchez, T. Yee, and D. L. Spears, Proc. IEEE84, 227 ~1996!.99E. R. Brown and K. A. McIntosh, inThin Films,edited by M. H. Fran-

combe and J. L. Vossen~Academic Press, San Diego, 1998!, Vol. 23,p. 173.

100H. C. Liu, in Handbook of Thin Devices,edited by M. H. Francombe~Academic Press, San Diego!, 2000 Vol. 2, p. 101.

101E. R. Brown, K. A. McIntosh, F. W. Smith, and M. J. Manfra, Appl. PhyLett. 62, 1513~1993!.

102H. C. Liu, G. E. Jenkins, E. R. Brown, K. A. McIntosh, K. B. Nicholsand M. J. Manfra, IEEE Electron Device Lett.16, 253 ~1995!.

103H. C. Liu, J. Li, E. R. Brown, K. A. McIntosh, K. B. Nichols, and M. JManfra, Appl. Phys. Lett.67, 1594~1995!.

104C. A. Kukkonen, M. N. Sirangelo, R. Chehayeb, M. Kaufmann, J. K. LS. B. Rafol, and S. D. Gunapala, Infrared Phys. Technol.42, 397 ~2001!.

105L. J. Kozlowski and W. F. Kosonocky, inHandbook of Optics,edited byM. Bass, E. W. Van Stryland, D. R. Williams, and W. L. Wolfe~McGraw-Hill, New York 1995!, Chap. 23.

106J. M. Lloyd, Thermal Imaging Systems~Plenum Press, New York, 1975!.107J. Y. Andersson, J. Appl. Phys.78, 6298~1995!.108P. E. Petersen, inSemiconductors and Semimetals,edited by R. K. Wil-

lardson and A. C. Beer~Academic Press, New York, 1981!, Vol. 18,p. 121.

109T. N. Casselman and P. E. Petersen, Solid State Commun.33, 615~1980!.110R. Schoolar and E. Tenescu, Proc. SPIE686, 2 ~1986!.111P.R. Bratt, inSemiconductors and Semimetals,edited by R.K. Willardson

and A.C. Beer~Academic Press, New York, 1977!, Vol. 12, p. 39.112N. Sclar, Prog. Quantum Electron.9, 149 ~1984!.


applied physics reviews quantum well photoconductors … · quantum well photoconductors in...

Documents