ProgressinQuantumElectronics36(2012)342–473 www.elsevier.com/locate/pquantelec Review Progress in focal plane array technologies Antoni Rogalskin InstituteofAppliedPhysics,MilitaryUniversityofTechnology,2KaliskiegoSt.,00-908Warsaw,Poland Availableonline13July2012 Abstract Development of focal plane arrays started in seventies last century and has revolutionized imagingsystemsinthenextdecades.Thispaperpresentsprogressinopticaldetectortechnologyof focalplanearraysduringthepasttwentyyears.Atthebeginningofpaper,emphasisesaregivenon integrated detector assembly and cooling requirements of different types of detectors. Next, the classificationoftwotypesofdetectors(photondetectorsandthermaldetectors)isdoneonthebasis of their principle of operation. This topic is followed by general overview of focal plane array architectures. Themainsubjectofpaperisconcentratedondescribingofmaterialsystemsanddetectorsoperated in different spectral ranges. Special attention is given on recent progress in their detector technologies. Discussion is focused mainly on current and the most rapidly developing focal plane arraysincluding:CdZnTedetectors,AlGaNphotodiodes,visibleCCDandCMOSimagingsystems, HgCdTe heterostructure photodiodes, quantum well AlGaAs/GaAs photoresistors, and thermal detectors. Emphasis is also given on far-infrared and sub-millimetre wave detector arrays. Finally, the outlookfor near-future trendsin optical detectortechnologies ispresented. &2012Elsevier Ltd. Allrights reserved. Keywords: X-ray detectors; Ultraviolet detectors; Visible detectors; Infrared detectors; Monolithic and hybrid focalplanearrays Contents 1. Introduction . .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . . 343 2. Integrateddetectorassembly. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . . 346 nTel./fax:þ48226839109. E-mailaddress:[email protected] 0079-6727/$-seefrontmatter&2012ElsevierLtd.Allrightsreserved. http://dx.doi.org/10.1016/j.pquantelec.2012.07.001 A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 343 2.1. Detector operating temperature .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 347 2.2. Cooler technologies. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 348 2.2.1. Cryocoolers .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . . 349 2.2.2. Peltiercoolers. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . . 356 3. Classification ofdetectors. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 358 3.1. Photon detectors. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 358 3.2. Thermal detectors.. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 367 4. Overview offocal planearray architectures . .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 372 4.1. Monolithic arrays.. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 373 4.1.1. CCDdevices . .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . . 374 4.1.2. CMOS devices .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . . 375 4.2. Hybrid arrays. .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 378 4.2.1. Interconnecttechniques.. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . . 379 4.3. Performance offocal planearrays . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 383 5. X-raydetector arrays.. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 385 5.1. Phosphors, scintilators, andmicrochannel plates . .. .. . .. .. .. .. .. .. . .. .. . 385 5.2. Direct andindirect X-ray detection. .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 387 5.3. Silicon detectorarrays. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 388 5.4. CdZnTe hybrid array . .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 391 6. Ultraviolet detector arrays .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 394 7. Visible detector arrays. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 398 8. Infrareddetectorarrays.. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 404 8.1. InGaAs arrays .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 404 8.2. Schottky-barrier photoemissive arrays. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 407 8.3. Lead saltarrays . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 409 8.4. InSb arrays . . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 411 8.5. HgCdTe arrays.. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 413 8.6. QWIP arrays . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 418 8.7. Type II superlattice arrays.. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 421 8.8. Thermal detectorarrays.. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 426 8.9. Third generationinfrared detectors. .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 431 8.10. Readiness level ofLWIR detectortechnologies .. .. .. . .. .. .. .. .. .. . .. .. . 436 9. Far-IR andsub-mm-wave detectorarrays .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 438 9.1. Schottky barrier arrays .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 444 9.2. Silicon andgermanium extrinsicphotoconductorarrays. .. .. .. .. .. .. . .. .. . 446 9.3. Pairbraking photondetectorarrays.. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 452 9.4. Semiconductor bolometer arrays . . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 455 9.5. Superconducting HEBand TESarrays.. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 458 9.6. Field effecttransistordetector arrays . .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. . 463 10. Conclusions . . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 464 References .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 466 1. Introduction Developments in source and detection of electromagnetic radiation have a very long history. First humans relied on the radiation from the Sun. Cave men used torches (approximately 500,000 years ago). Candles appeared around 1000 BC, followed by gas 344 A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 lighting(1772),andincandescentbulbs(Edison,1897).Radio(1886–1895),X-rays(1895), UV radiation (1901), and radar (1936) were invented in the end of the 19-th and the beginning of the 20-th centuries. The terahertz (THZ) region of electromagnetic spectrum (see Fig. 1) is often described as the final unexplored area of spectrum and still presents a challenge for both electronic and photonic technologies. Looking back over the past several hundreds of years we notice that following the invention and evolution of optical systems (telescopes, microscopes, eyeglasses, cameras, etc.)theopticalimagewasformedonthehumanretina,photographicplate,orfilms.The birth of photodetectors can be dated back to 1873 when Smith discovered photo- conductivity in selenium [1]. Progress was slow until 1905, when Einstein explained the newly observed photoelectric effect in metals, and Planck solved the blackbody emission puzzle by introducing the quanta hypothesis. Applications and new devices soon flourished, pushed by the dawning technology of vacuum tube sensors developed in the 1920s and 1930s culminating in the advent of television. Zworykin and Morton, the celebrated fathers of videonics, on the last page of their legendary book Television (1939) concluded that: ‘‘when rockets will fly to the moon and to other celestial bodies, the first images we will see of them will be those taken by camera tubes, which will open to mankind new horizons.’’ Their foresight became a reality with the Apollo and Explorer missions. Photolithography enabled the fabrication of silicon monolithic imaging focal planes for the visible spectrum beginning in the early 1960s. Some of these early developments were intended fora picturephone, other effortswerefor television cameras, satellite surveillance, and digital imaging. Infrared imaging has been vigorously pursed in parallel with visible imaging because of its utility in military applications. More recently (1997),theCCDcameraaboardtheHubblespacetelescopedeliveredadeep-spacepicture, aresultof10day’sintegration,featuringgalaxiesofthe30thmagnitude—anunimaginable figureevenforastronomersofourgeneration.Probably,thenexteffortwillbeinthebig- band age. Thus, photodetectors continue to open to mankind the most amazing new horizons. Fig.1. Theelectromagneticspectrum. A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 345 Inthis paperopticalradiation isconsidered asa radiation over therange from vacuum ultraviolet to the submillimeter wavelength (25 to 3000mm): 25–200nm Vacuum ultraviolet VUV 200–400nm Ultraviolet UV 400–700nm Visible VIS 700–1000nm Near infrared NIR 1–3mm Short wavelength infrared SWIR 3–5mm Medium wavelength infrared MWIR 5–14mm Long wavelength infrared LWIR 14–30mm Very long wavelength infrared VLWIR 30–100mm Far infrared FIR 100–3000mm Submillimeter SubMM THz radiation is frequently treated as the spectral region within frequency range nE0.1–10THz (lE3mm–30mm) and it is partly overlapping with loosely treated submillimeter (sub-mm) wavelength band nE0.1–3THz (lE3mm–100mm). Both ultraviolet (UV), visible, as well as infrared (IR) detectors technologies have undergone significant maturation, and high-performance detectors are available across mostspectralbandsfrom0.2-to100-mm.Significantimprovementswaspossibleforsensor systems by adding functionality, such as multi- and hyper-spectral response, polarimetric sensitivity, dynamic resolution and sensitivity adaptation, as well as reductions in size, weight, power and cost. The increase in digital processing capabilities, fuelled by the semiconductor industry, is a trend that will continue to have a major effect on sensor system. Development in detector focal plane array (FPA) technology has revolutionized many kinds of imaging [2]. From g rays to the infrared and even radio waves, the rate at which images can be acquired has increased by more than a factor of a million in many cases. Fig. 2 illustrates the trend in array size over the past 40 years. Imaging FPAs have developed in proportion to the ability of silicon integrated circuit (ICs) technology to read and process the array signals, and with ability to display the resulting image. The progress in arrays has been steady and has paralleled the development of dense electronic structures such as dynamic random access memories (DRAMs). FPAs have nominally the same growth rate as DRAM ICs, which have had a doubling-rate period of approximately 18 months; it is a consequence of Moore’s Low, but lag behind in size by about 5–10 years. The graph in insert of Fig. 2 shows the log of the number of pixels per a sensor chip assembly (SCA) as a function of the year first used on astronomy for MWIR SCAs. Charge coupled devices (CCDs) with close to 2gigapixels offer the largest formats. There are many excellent texts at both introductory and advanced levels that deal the fundamentals and applications of optical detectors [3–5]. Our intent is to provide a progress in optical detector FPA technology in the last two decades. In this context, the paperacquirestopicalityofpreviouslypublishedpapersdevotedFPA[2,6].Incomparison withthepreviouspapers,thispaperpresentsadditionalinformationdevoteddevelopment of integrated FPA detector assemblies. 346 A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 1010 109 108 107 106 1011 105 104 ARGUS-IS DRAM (surveillance) 1010 MOS 103 GAIA CCD SNAP(space) 102 (space) Intel 109 (project) DRAM 108 107 106 CCD S O M 105 C e bl si 104 Vi 103 102 101 Fig.2. Imagingarrayformatscomparedwiththecomplexityofsiliconmicroprocessortechnologyanddynamic accessmemory(DRAM)asindicatedbytransistorcountandmemorybitcapacity(adaptedafterRef.[2]with completions). The timeline design rule of MOS/CMOS features is shown at the bottom. CCDs with close to 2gigapixelsofferthelargestformats.NotetherapidriseofCMOSimagerswhicharechallengingCCDsinthe visiblespectrum.Thenumberofpixelsonaninfraredarrayhasbeengrowingexponentially,inaccordancewith Moore’sLawfor30yearswithadoublingtimeofapproximately18months.Ininfrared147megapixelarraysare now available for astronomy applications. Imaging formats of many detector types have gone beyond that requiredforhighdefinitionTV. 2. Integrated detector assembly AFPAiscreatedbyarrangingindividualelementsinalattice-likearray.Typicallyeach pixel has one independent contact and shares the second contact with other pixels in the array. The distribution of the common contacts impacts electrical and readout speed. A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 347 Bump Communications link Optics bonding (wireless or fiber) s) &mitter Dpeitxeecltsor Local processingparallel/serial linkmunications trans Receiver remoteprocessing(serial link) &D sistoprlaagye w. m (co Cooling Fig.3. Schematicrepresentationofanimagingsystemshowingimportantsub-systems(afterRef.[7]). Aseriouslimitationinthedevelopmentofarraysofdetectorsisthatlightiseasilycoupled to neighbouring pixels in an array which lead to the developments of false counts, or crosstalk. There are approaches to mitigate this limitation, but they add additional complexity to the manufacturing. Detectors are only a part of usable sensor systems. Military sensor systems include optics, coolers, pointing and tracking systems, electronics, communication, processing together with information-extraction sub-systems, and displays (see Fig. 3)[7]. So, the process of developing sensor system is significantly more challenging than fabricating a detector array. InIRsystems,two-dimensional(2-D)arraysofdetectorsconnectedwithindiumbumps toa readout integrated circuit(ROIC) chip asahybridstructure areoften called asensor chip assembly (SCA). The FPA industry is not sufficiently large to support the development of a complete set of unique tools. The evolution of the silicon industry can lead to divergence and to gaps inthe FPA tool set. One simpleexample is that the silicon industryhasstandardizedonafieldsizeof22(cid:2)33mm2foritslithographytools.Thedrive to larger pixel counts for FPAs often requires much larger overall FPA sizes which can only be accomplished by abutting multiple fields. Tilling large arrays from smaller chips addresses the practical and economic limits of making larger detector chips. Generally, there are two main types of detectors; cooled and uncooled. Cooled detectors require cooling below ambient temperature. Although uncooled sensors offer significant advantages in terms of cost, lifetime, size, weight and power, cooled sensors offer significantly enhanced range, resolution, and sensitivity as a result of the lower noise operation. In the infrared industry the housing with detector installed is known as and integrated detectorassembly(orIDA).ThehousingonanIDAisbasicallyafancydewar.Thedetectoris locatedonthebaseoftheinnerwallwithawindowinthebaseoftheouterwall.Therearemany design considerations and challenges that go into developing an IDA. 2.1. Detector operating temperature Fig. 4 is a chart depicting infrared operating temperature and wavelength regions spanned by a variety of available infrared detector technologies. Typical operating temperatures range from 4K to just below room temperature, depending on the detector technology. Most modern cooled detectors operate in the temperature range from o10 to 150K, depending on the detector type and performance level. 77K is a very 348 A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 Fig. 4. The operating temperature and wavelength regions spanned by a variety of available IR detector technologies. common temperature because this is relatively easily achievable with liquid nitrogen. Uncooled detectors, despite their title, typically incorporate some degree of temperature control near or slightly below room temperature ((cid:3)250–300K) to minimize noise, optimize resolution, and maintain stable operating temperature. For space mission detectors with the very long wavelength cutoff wavelengths are also required.Duetosmallphotonenergyactivation,theextrinsicdetectormaterialsareused. In comparison with intrinsic photoconductivity,the extrinsic photoconductivity isfar less efficient because of limits in the amount of impurity that can be introduced into semiconductorwithoutalteringthenatureoftheimpuritystates.Implicitinthetreatment oflowbackgrounddetectorsisthatthegenerationoffreecarriersisdominatedbyphoton absorption, not by thermal excitation. As a result, lower temperature is required as the long wavelength cutoff of the detector increases, what can be approximated as [8] 300 K T ¼ ð1Þ max l ½mm(cid:4) c This general trend is illustrated in Fig. 5 for six high performance detector materials suitable for low-background applications: Si, InGaAs, InSb, HgCdTe photodiodes, Si:As (Si:Sb) impuritybandconductor (IBC) detectorsalso calledastheblocked impurityband (BIB) detectors, and Ge:Ga stressed photoconductive detectors. The operating tempera- ture range can be fairly broad due to the wide range of system backgrounds. 2.2. Cooler technologies The method of cooling varies according to the operating temperature and the system’s logistical requirements [9,10]. The two technologies currently available for addressing the A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 349 1000 100 10 1 0.1 Fig. 5. Operating temperatures for low-background material systems with their spectral band of greatest sensitivity. The dashed line indicates the trend toward lower operating temperature for longer wavelength detection. cooling requirements of IR and visible detectors are closed cycle refrigerators and thermoelectric coolers. Closed cycle refrigerators can achieve the cryogenic temperatures required for cooled IR sensors, while thermoelectric coolers are generally the preferred approachtotemperaturecontrolforuncooledvisibleandIRsensors.Themajordifference between the thermoelectric and mechanical cryocoolers is the nature of the working fluid. Athermoelectriccoolerisasolid-statedevicethatuseschargecarriers(electronsorholes) as a working fluid, whereas mechanical cryocoolers use a gas such as helium as the working fluid. The selection of a cooler for a specific application depends on cooling capacity, operating temperature, procurement, cost and maintenance, and servicing requirements. A survey of currently operating cryogenic systems for commercial, military, and space applications are summarized in Fig. 6. 2.2.1. Cryocoolers Cryocoolers can be classified as either recuperative or regenerative. In recuperative systems,gasflowsinasingledirection.Thegasiscompressedatambientfixedtemperature and pressure and allowed to expand through an orifice to the desired cryogenic fixed temperature and pressure. The Joule Thompson and Brayton cycle refrigerators are examples of recuperative systems. In a regenerative system, the gas flow oscillates back and forth between hot and cold regionsdrivenbyapiston,diaphragmorcompressor,withthegasbeingcompressedatthe hot end and expanded on the cold end. Stirling, Gifford-McMahon and Pulse Tube cryocoolers are the most common types of regenerative cryocooler systems. Fig. 7 presents a map of the major cryocooler applications in terms of the temperature and net refrigeration power required. The major commercial applications include cryopumps for semiconductor fabrication facilities, magnetic resonance imaging magnet 350 A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 Thermoelectric Giffort-McMahon Fig.6. Temperaturerangesforcommercialrefrigerators(afterRef.[9]). 106 LNG 105 Accelerators LHiquid Transmission 2 lines 104 1TJ Vacuum Large- size Transformers 2 103 FCL Mid-size Generators 102 SMES MRI Motors Cryopumps FCL 101 Maglev Bearings Cryosurgery MRI Wireless Micro-SMES 100 H2ZBO IR 2ZBO 4 10-1 elecLtTroSnics NElbeNct. IR IR HTS SQUIDs SQUIDs 10-2 Fig.7. Mapofcryocoolerapplicationsinplaneofrefrigerationpowerversustemperature(afterRef.[10]).Note: SMES—Superconducting Magnetic Energy Storage, MRI—Magnetic Resonance Imaging, LTS—Low Tem- perature Superconductivity, SQIDs—Superconducting Quantum Interference Devices, LNG—Liquid Natural Gas, FCL—Freon Coolant Line, IR—Infrared, ZBO—zero boil-off, and HTC—High-Temperature Superconductivity. cooling, and gas separation and liquefaction. The largest low-power application of cryocoolers is covered by IR sensors. Theperformanceofacryocoolerisspecifiedbycoefficientofperformance(COP),which is defined as the ratio of cooling power achieved at a particular temperature to total electricalinputpowertothecryocooler.TheCOPisoftengivenasafractionoftheCarnot efficiency. Fig. 8 compares the relative performance of the different technologies as a fraction of the limiting ideal efficiency. For lower temperatures the efficiency drops significantly.At4Ktypicalefficiencyisabout1%orless.Generally,recuperativesystems haveadvantagesintermsofreducednoiseandvibration,whereastheregenerativesystems A.Rogalski/ProgressinQuantumElectronics36(2012)342–473 351 30 25 Stirling Gifford-McMahon 20 Turbo-Brayton 15 10 5 0 Fig.8. Efficiencyofsmallcryocoolersasafunctionofcoldendtemperature(afterRef.[10]). tend to obtain higher efficiencies and greater reliability at the temperatures of interest for many IR detector applications. Eversincethelate1970s,militarysystemshaveovercometheproblemofLN operation 2 by utilizing Stirling closed-cycle refrigerators to generate the cryogenic temperatures necessary for critical IR detector components. These refrigerators were designed to produce operating temperatures of 77K directly from DC power. Early versions were large and expensive, and suffered from microphonic and electromagnetic interference (EMI) noise problems. The use of cooling engines has increased considerably due to their efficiency, reliability, and cost reduction. Today, smaller and more efficient cryocoolers have been developed and refined. The development of novel adsorber materials and designsaswellasimprovedsealingtechniquesformaintainingthehighpressuresrequired for cryocooler operation have significantly contributed to the improved lifetimes [mean time before failure(MTBF)] nowachievable inmany cryocoolersystems (5000–10,000h). Lifetime of ten years are now possible in space cryocoolers, and five years lifetimes are possible in similar cryocoolers developed for commercial applications [11]. A comparison of cryocooler efficiencies near 80K is shown in Fig. 9. Efficiency as high as about 20% is achieved for large best space cryocoolers, whereas 10–20% is typical for the best commercial cryocoolers. Improvements in heat exchanger, recuperator designs and materials have given COP values in the range to 10% for small cryocoolers. However, despite these advances, the cryocooler remains a major failure point for the integrated IR sensor system. BelowwebrieflydescribedifferenttypeofFPAcoolers[12].Table1presentsadvantages and disadvantages of different cryocoolers for space applications. More information can be found in Refs. [11,13]. 2.2.1.1. Cryogenic dewars. Most 8–14-mm detectors operate at approximately 77K and can be cooled by liquid nitrogen. Cryogenic liquid pour-filled dewars are frequently used for detector cooling in laboratories. They are rather bulky and need to be refilled with