Infrared Detectors for the Year 2000

!!~:: :~:+ ~:~ ::~:::':::$:::: :" ~: :'::$: :$: ::::::::: ::: :::::::::::::::::::::: ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::~i$?.~::$ ~:;'~~:::::~:::;~$:$ ~:::::.<::.~:~ ::'¢:':::'::::::::'~ ~k x ~::::::~:~:::::::::::::::::::::::::::::::::::::::::::::?~:~:::~: 3::::: ~;:~: :: $::: :::::::::::::::::::::::::::::: :?$~: ::::::::::::::::::::::::::::: :::] ~::~i~i~::~::~!~ii~i~::~!::~::~!~::~i~::i~::~::~~~i~i~::~i~::::~::~::~::~i~::::~::~::::~::~¢:~::!::~::~ii::~i~::!~::~ii::!i::!::::~::~::i::i~::~:::i~i~;~::~:~::~i~::~::~::~::~::~::~::~i::~::~::~::~::~

I.M. Baker, M.P. Hastings, L.G. Hipwood, C.L. Jones, and P. Knowles GEC-Marconi Infra-Red Limited

GEC-Marconi Infra-Red Limited (GMIRL) is Europe's largest supplier of infrared detectors. Its leading position in this field is illustrated in this article by describing the key detector technologies that will supply major projects into the year 2000, and beyond.

he last two decades have seen rapid development in the T technology of thermal imaging, and the introduction into service of many new infrared detector types for military, industrial, and space applications. The semiconductor cadmium mercury telluride (CMT) has been employed in the great majority of high-performance detectors, and continues to be the material of choice for advanced infrared detectors from nearly all major suppliers. For the highest performance, CMT detectors require cooling to operational temperatures around 80K (-193°C). This intro- duces the need for miniaturized Figure la. CMT bulk crystal and slices. cryogenic cooling techniques and a rugged vacuum-insulated encapsulation. silicon read-out circuit, allowing a Expk~sk~l det(mtOf Conventional thermal imaging sys- much greater array size than conven- // / Gor~e oh~ tems employ a variety of photocon- tional photoconductors, but with an ray ductor arrays as detectors. Among increase in complexity of detector ard these, the TED or SPRITE detector, manufacture. The successful manu- ~es manufactured exclusively by GMIRL facture of these advanced devices m and its licensees, has achieved con- depends critically on bringing to- ~e siderable success in the UK Common gether the technologies of crystal Module programme and in many growth, array processing, and multi- other countries. Its range of applica- plexer design. m=c tions has been broadened by improv- The aim of GMIRL's research, ~ket ing performance over time and which is strongly supported by links extending the range of array formats with UK Government research agen- and encapsulations in which it is cies, is in the direction of higher- star supplied. performance, lower-cost detector ar- lion The drive towards higher perfor- rays and its scope embraces "flip- nyo mance in thermal imaging has stimu- chip" hybrids, heterostructures, and late lated the development of Focal Plane multiple quantum wells. Photodiode Arrays. These detectors combine arrays of photodiodes of an Figure lb. Bridgman infrared sensitive material with a furnace for crystal growth.

0961-1290/96/$15.00 ©1996 Elsevicr Science Ltd. PageS0li~i~iii l!ii!i~ii~:/:~iii.i"]VolgNo2 !i ;~i!',::,j!ii~,,~&~,!!~ ~,-ii~i~ ...... :~ .**:~::::~..,. :~.:. :. :. :. ,.,. :.:. :.:.~. :~..~: ...... ~...... ~...:~ ...... ~...... ~. ~,~ ~A :~ ~~.~- ~!~-#~ •~.~-~..~ ...... • -.-.-..-.-- --.,-- -.---.--.-----.--.-- ...-.-.-:.-...-.~ ~-.-....~ .~ ..... :::::::::::::::::::::::::::::::::::::::: ======:::::.:::~::~<~::~4~:~::::~::~::~;;~4~:.:::~ j

Some basic principles In the absence of light from the sun or artificial sources, the predominant source of electromagnetic radiation in the terrestrial scene is the uni- versally-present thermal radiation which is emitted naturally from all objects, and is described in an idea- lized form as black body radiation. Real objects depart from the ideal and their thermal signatures are characterized by a mixture of temperature, emissivity and reflectivity. Except at very short range, the absorption properties of the atmosphere restrict the useful range of wavelengths for the observation of thermal emissions to two bands -- the medium (3-5 I~m) and long wavelength (8-12 I~m) infrared bands. Figure 2. CMT photoconductor array. The 48 elements are .50 ~m square. These are distinct from the short wavelength (c. 1 I~m) or near-infrared band, which borders the visible task of camouflage very difficult erates excess mobile charge carriers range and covers most applications of compared with the visible. This is (electron-hole pairs) which, in the fibre-optic communications and do- because of a mixture of effects. In the most sensitive types, persist for per- mestic infrared remote controls. outdoors, temperature variations oc- iods up to several microseconds CMT infrared detectors exhibit cur throughout the 24-hour cycle of before recombining. The transient strong absorption of radiation up to solar heating, and objects of interest change in conductivity can be de- a characteristic cut-off wavelength may be either warmer or cooler than tected and amplified into a signal which corresponds to the semicon- their surroundings. Climatic varia- proportional to the infrared intensity. ductor band-gap between the va- tions induce a wide range of back- The construction of simple photo- lence and conduction bands. In ground temperatures, and effective conductor arrays generally involves CMT, this cut-off wavelength can be sky temperatures (which influence the mounting of a CMT die on a engineered through the alloy compo- reflections) vary widely with weather sapphire carrier, delineation of ele- sition of the semiconductor crystal, conditions. Finally, human skin is ments, and the application of ohmic to match the precise requirement of effectively a black body in the long contacts that extend to remote wire- the application, and can be extended wavelength infrared and, as a natu- bonding pads on the carrier (Fig. 2). out to the longer wavelengths of rally warm surface, is highly visible. For imaging applications, this type of interest by virtue of its narrow The crystal growth of CMT is detector needs the image to bc band-gap properties. problematic, and no single growth mechanically scanned, in swathes, to The long wavelength band contains method has emerged which domi- build up a picture in TV format. a large fraction of the total thermal nates the industry/11. In GMIRL, the The detecting elements of focal radiation from objects in the ambient older methods of bulk crystal growth plane photodiode arrays (FPPAs) arc temperature range and is unsur- by the vertical Bridgman technique p-n junction photodiodes. This type passed for the detection and imaging and the accelerated crucible rotation of detector, although technologically of small temperature or emissivity technique (ACRT) are still used to more complex, has the low power differences in the ambient scene. The provide a lowcost, volume source for dissipation and high impedance ne- medium waveband contains only a the production of photoconductor cessary for the realization of large few per cent of this radiation, but arrays. The in-house growth capabil- arrays of many thousands of ele- also contains the peak emissions ity includes zone refining and distilla- ments, interfaced by a direct injec- from artificial sources at a few tion of the elemental starting tion scheme to a silicon multiplexer hundred degrees C. The medium ingredients. These methods produce circuit. Time division multiplexing of waveband is therefore an excellent a cylindrical ingot which is sawn into the diode outputs means that the medium for the detection of hot slices and mechanically polished to silicon chip can be serviced by bodies against a cooler background. device thickness before division into relatively few electrical connections. In addition, the small fraction of individual detector dies (Fig. 1). In GMIRL, CMT material of the ambient scene radiation in this band All of GMIRL's CMT detectors are required crystal perfection for these can be converted into sensitive ima- photoconductor or photodiodc devices is supplied by the technique gery by the most advanced types of types. The simpler photoconductor of liquid phase epitaxy (LPE). Diode large detector arrays. arrays, manufactured in a wide vari- formation techniques are described The intimate combination of tem- ety of-formats and sizcs, share a in later sections, but in general, for perature and emissivity which deter- common operating principle. The detector types currently in develop- mines infrared signatures makes the absorption of infrared radiation gen- ment, diode formation proceeds after

Page 51 [~r|~/ VOl 9 No 2 achieved by tuning the detector bias to equalize the image scan velocity and the drift velocity of the photo- generated minority carriers (holes). As a result, TDI is performed intern- ally and the integrated signal is read- out between a pair of contacts at one end of the element. In effect, the whole area of the long element is useful in gathering signal while preserving the spatial resolution characteristic of a small element. This results in an eight-channel system of equivalent performance to one using many tens of conventional elements, bias supplies, and ampli- tiers.

Figure 3a. Common module detector. the growth process, into a die of p- type material. A third technique of crystal growth, practised as part of GMIRL's research programme, is metal-organ- ic vapour phase epitaxy (MOVPE). In this technique, successive epitaxial layers can be grown with control over thickness, alloy composition and chemical doping. This provides a means of producing device structures directly as part of the growth process and is a possible route to Figure 3b. Thermal imager for naval applications. future detectors of superior performance. of detector was a multi-element Several developments of the TED/ Common module photoconductive array for Class I SPRITE have improved its perfor- and, for Class II, the 8-element TED/ mance and extended its range of detectors.., and beyond SPRITE detector[3]. Class I systems application. Mid-waveband and dual- Developments in infrared imaging are generally man-portable and the waveband versions have been devel- have been driven largely by military display is formed by an array of light- oped, and customized types incor- requirements and this continues to emitting diodes in conjunction with porating special array formats, be the trend. The range of military the scanning mechanism. Class II matched radiation shields, or special applications is wide - surveillance, systems have generally higher perfor- window coatings are now frequently reconnaissance, navigation, target ac- mance and are capable of being designed into specific project appli- quisition and tracking, weapon indirectly viewed via a TV monitor. cations. sights, personal sights, fire control In both cases the encapsulation is Reliable miniature Joule-Thomson and missile guidance - and is cooled by a miniature Joule-Thomson liquefiers of the type originally devel- matched by the variety of platforms. cooler. oped by Parkinson at the Royal Radar There has been an attempt in the The TED/SPRITE detector is a Establishment in the 1950s are avail- UK and separately in the US and unique example of advanced photo- able commercially, notably from the France to standardize military sub- conductor design (Fig. 4). Each ele- Hymatic Engineering Company in the systems using modular design princi- ment of the parallel-scanned array of UK. These operate by the cooling ples. The UK Thermal Imaging eight long elements achieves a signal- effect from a high pressure gas Common Module (UK TICM) evolved to-noise ratio and spatial resolution escaping to atmosphere through a from this approach in the early equivalent to a row of serially- small orifice, and can be employed 1980s, and the common module scanned simple square elements, but for the liquefaction of the gas (Fig. 5). detector was developed as part of without the need for external time The principal features of the min- the programme TM (Fig. 3). The choice delay and integration (TDI). This is iature open-cycle cooler for the Pa0e521 IVo0No2 grated Detector Cooler Assembly (IDCA), illustrated in Fig. 6 Integrated detector cooler assembly. By combin- ing the dewar stem and engine cold finger into a single component, the detector array can be mounted directly on the cold tip of the cooler and the detector vacuum encapsulation can be constructed as an extension of the cooler itself. This configuration brings cost savings and major reductions in size, weight, and power consumption. GMIRL has collaborated with several engine manufacturers to produce a range of IDCA types, including TED/SPRITEs to the same performance specifica- tions as conventional encapsulation types. The success of the TED/SPRITE Figure 4a. TED/SPRITE array (700 ~m long elements). detector can be judged from the number supplied into service, which is now in the region of 5000 systems.

Non-imaging types Non-imaging detector applications include instrumentation, fuses, missile guidance including homing heads, beam riders, and the guidance of intelligent munitions. Although the detector arrays for these types may be fairly simple in format, or even single elements, the encapsulation and cooling requirements are varied and frequently very demanding. One example for instrumentation is the thermoelectrically-cooled TED/ SPRITE detector which has been in quantity production for many years. Figure 4b. This application area contains re- Thermal quirements for the fastest possible image cool-down times to cryogenic oper- common module detector are: been produced for use with several ating temperatures, for example to 1) a helically-wound, finned, high- types of Stirling engine coolers. bring detectors into operation fob pressure tube surrounding a central These engines initially enjoyed a lowing initiation of cooling at missile mandrel which tightly fits the inter- reputation for short service life, launch. One example of a fast-cool nal precision bore of the detector requiring frequent maintenance at detector is that for the guidance dewar, forming a counter-current intervals of hundreds of hours, and system of the medium-range TRIGAT heat exchanger; and for suffering rapid degradation as a missile (Fig. 7). The detector cooler 2) the precision orifice, which consequence of mechanical wear or assembly incorporates a one-shot gas terminates the tube just behind the contamination by lubricants. The bottle charged to high pressure with detector substrate, forming a valve problem of service life has shown argon gas. for the expansion of high-pressure itself to be soluble through good Argon is the preferred refrigerant air. engineering design, and the trend in in fast-cool applications. Its Joule- Self-regulation is achieved by a infrared imaging is now towards Thomson coefficient is about double needle valve mechanism operated increasing numbers of engine-cooled that of air, which results in more by thermal expansion of a sealed systems which are favoured for their rapid extraction of heat, and it volume of gas in a bellows. By this independence of compressed pure liquefies at a temperature of 90K (- means the gas consumption is mini- gas supplies and the regime of pure 183°C) which is similar to the beha- mized with respect to heat load after gas hygiene. viour of air, although slightly inferior the initial cool-down period. The most compact of engine- from the point of view of detector Vacuum encapsulations have also cooled configurations is the Inte- operation. By minimizing the thermal mass of the cooled detector assem- Pa0e531 IVogNo2 ] ~~~::~.~:. ~:~. ~::~.'~.~-~:---~,'-:--.~;:, ~o- ~ ~ ~ :~'~:~'., ~ ...?~,.:~..>-~---~...... ~ ~ ~ ~..~:.,.o o.~. ~- ~ ~...:...~i:~:~!:!~*~!:~::~..~:..~:~;~i~:~~!:.~g~.i~~:~::.~:~N~:~~g ~;~i:iN"~~:~:i: :::N~N,::~I

high performance, often in low infrared background flux, and with a diversity of wavelengths. This pre- sents design constraints that are often more demanding than tej'res- trial applications. However, there is considerable freedom to optimize the temperature of operation. CMT photoconductors preceded photodiodes into space, with the first photodiodes from GMIRL being launched in the ISAMS instrument in 1991. CMT device characteristics are such that photoconductors are pre- Figure 5. Joule-Thomson coolers. ferred for wavelengths above about 8 ~tm, and photodiodes below.

Figure Z Detector cooler assembly for missile guidance.

Scan Radiometer, flown on the US The space business is an expanding Synchronous Meteorological Satellite area for GMIRL, and commitment to in the 1970s. This was followed by satisfying the need for high standards the ESA Meteosat programme, and its of cleanliness for space projects is extension to Transitional Meteosat, illustrated by the dedicated Space Figure 6. Integrated detector cooler providing thermal imaging for weath- Assembly and Test Facility which assembly. er forecasting. The latest study phase was commissioned in 1994 (Fig. 8 of the more advanced Meteosat GMIRL space projects assembly facil- Second Generation (MSG), scheduled ity). bly, it is possible by this means to for launch in the year 2000, antici- Current projects include the Mi- achieve cool-down times of a fraction pates the development of new in- chelson Interferometer for Passive of a second. struments in which several of the Atmospheric Sounding (MIPAS), spectral channels will be served by scheduled for launch in 1998, and Detectors for space CMT detectors. Other programmes the High Resolution Infrared Spectro- for which GMIRL detectors have meter (HRIS). GMIRL has been involved in the been successfully developed are the development and supply of infrared Along-the-Track Scanning Radio- detectors for space applications for meter (ATSR), flown on the Earth Focal plane arrays more than twenty years. Early pro- Resources Satellites (ERS-1 and ERS- First-generation infrared detectors, grammes used tri-glycene-sulphate 2) and used to map sea surface based on TED/SPRITEs and photo- (TGS) pyroelectric detectors, for ex- temperature accurately, the Im- conductive CMT, are firmly estab- ample in an experimental horizon- proved Stratosphere And Mesosphere lished as the main technology for crossing indicator and in radiometers Sounder (ISAMS) monitoring the thermal imaging. It is likely that this flown in "Nimbus E" and "Nimbus F", earth's atmosphere for gas contami- situation will continue for some time which were early American weather nants, and the Pressure Modulated into the future, in view of the satellites. Infrared Radiometer (PMIRR) for the maturity of the production processes The first CMT detector fabricated NASA Mars Observer Programme. and the relatively low cost of the for a space application was used in Space application typically calls for detectors. There are, however, an the thermal band of the infrared Spin Pa0es,l [Vo0No2 ::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::I ~ ~ : ~- " ~?&~i~%~{:~ ~;~i~ ~;~,~:~;~.~ ~i$ ~ ~ ~ ~ iiiii:~ :ii]I

waveband, the preferred process at GMIRL is the proprietary "loophole process" which inherits its reliability from well-established photoconductor fabrication techniques (Fig. 10). Currently, LPE wafers for FPPAs are grown on 20 mm x 30 mm substrates and equipment with a capa- city for larger 30 mm x 40 mm substrates is under development. Cadmium zinc telluride substrates containing 4% of zinc are used to achieve an almost perfect match with the crystal lattice constant of CMT. These are polished in very clean conditions to minimize nucleation defects at the start of growth. A surface flatness of _+ l pm is needed to achieve the desired layer thickness uniformity. After growth, the cadmium zinc teiluride substrates are removed from the layers, which are then carefully polished to device thickness and divided into individual detector dies. The thermal expansion mismatch problem, encountered in all hybrid structures, has been solved in the loophole process by using these thin dies of CMT, which are typically 9 pm thick. These are bonded rigidly to the silicon multiplexer, so that Figure 8. GMIRL space projects assembly facility. strain is taken up elastically by the CMT. The aim is to make the devices mechanically very robust, and arrays increasing number of applications that cannot be met by photoconductive arrays. These require higher infrared sensitivity. A hybrid focal plane approach H, using an array of CMT photodiodes interconnected with a custom-designed multiplexer, allows a very large number of detector elements to be located on the focal plane, and this provides the higher sensitivity (Fig. 9). CMOS multiplexers are the best choice to perform the integra- Figure 9. 1024 element-long linear CMT/CMOS hybrid detector array (actual size of device tion and signal processing for linear c.18 mm overall). or two-dimensional arrays. In recent years, CMOS processes with 1 pm or smaller design rules have permitted the design of lowcost, high-density or less, CMOS is the only practical up to 15 mm in length have been multiplexers. solution. In addition, there are ad- shown to be unaffected by multiple Although CCDs are a well-known vantages resulting from fewer drive cycling to cryogenic temperatures. approach for multiplexed arrays, in voltages, fewer clocks, and much Once the CMT is bonded to the CCD technology it is difficult to lower operating voltages. multiplexer, detectors are individu- envisage ways of implementing some Focal-plane photo-diode arrays ally processed in small batches. A of the advanced functions designed must be as reliable as photoconduc- single photolithographic stage is into GMIRL's multiplexers, because tor arrays. In particular, survival of used to define a photoresist film with of the poor packing density of CCD temperature cycling, mechanical a matrix of holes of 8 pm diameter. processes. For long linear arrays with shock, and high temperature storage Using ion beam milling, the exposed pixel pitches of the order of 30 pm are very important. For the long CMT is eroded to the underlying

Page551 ...... Vo19No2 direction. In addition, the multiplexer can be configured to provide dead element deselection (DED) so that only the good diodes in each channel are read out. This approach, which overcomes the inherent CMT defect problem, is a very powerful tool for optimizing the NETD in each detector channel. Several new projects, including the Infrared Search and Track detector for the European Fighter Aircraft, have adopted this route. The alternative to a scanned array is a two-dimensional staring array. The largest two-dimensional FPPAs currently available in GMIRL are 128 x 128 element arrays. Larger arrays of 384 x 288 elements on a pitch of 20 ~tm are under development for the EUCLID programme, a European collaborative initiative in defence technology. These are medium waveband arrays, for which the two- Figure 10. Loophole array construction. The pitch is typically 30 Ilm. dimensional format has some advan- aluminium contact pads on the multiplexer. The erosion rate of the aluminium is much slower than the CMT, ensuring a wide latitude in stop time. The device is then coated with a metal. On removal of the photoresist, metal remains in the holes forming individual electrical contacts between the CMT and the multi- CMT h~Lyer -- )lexer pads. The loophole junction (p-type;, is formed around the hole during the ion beam milling process. An advan- tage of forming junctions by ion milling is that the whole process takes place at room temperature, ~,._= f which preserves the original material Indimn'buml:~ quality. Under the correct growth conditions, the dislocation density of the Figure 11. Flip-chip array construction.The pitch is typically 30 Ilm. grown layer is no greater than in the substrate, and ranges from 3 to 7 x 10 4 cm -2. The substrate defect level is range of a few hundredths of a tages. Although the technical chal- not easy to reduce and devices there- degree. lenge in making such large detectors fore exhibit defect levels in the mid The favoured design for many of is great, when combined with micro- 10 4 cm -z range resulting in about 1% these systems is a long linear FPPA scanning of the image field they offer of photodiodes being defective. providing instantaneous coverage of a route to full PAL TV imagery in a For the highest performance ima- one row or column of the image, sensor head without any conven- ging applications, full TV resolution is which is scanned in one dimension tional moving parts. This develop- required, combined with very high to achieve the desired image format. ment will reduce an infrared imager resolution of temperature. The latter This configuration leads to a compact towards the size and complexity of is limited by the signal-to-noise level imager and optics. Single, multi- the TV camcorder, and is widely seen of the detector and is described by plexed long linear arrays up to 1024 as the way to more affordable infra- the Noise Equivalent Temperature elements have been developed by red systems in the medium wave- Difference (NETD). Whereas imagers GMIRL for this type of application. length band. At the same time, the using photoconductor arrays may For higher sensitivity, several parallel thermal resolution of such a detector resolve temperature differences rows of diodes can be combined with with more than 100 000 elements down to 0.1 or 0.2 degrees, second the appropriate multiplexer design may approach 0.01 degrees. generation requirements are in the to form TDI channels in the scan

Pa0oS01 l Vo, gNo2 ~ I

are deposited on the array and cleaving through a diode and etching multiplexer surfaces. Proces- the cleaved surface to generate topo- sing can be performed on a graphical variations between n and p- wafer scale up to this stage type regions. A cross-section through before dicing into individual a diode produced by ion beam detector components. Finally, milling is shown in Fig. 12. the detector array and multi- The mismatch in thermal expan- plexer are aligned in a dedi- sion between the CMT array and the cated flip-chip bonder and the silicon multiplexer stresses the struc- Figure 12. Cleaved CMT planar diode revealing two surfaces are pressed to- ture when hybrid detectors are the p-n junction. The maximum junction diameter gether to cold-weld the indium cooled to their operating tempera- is c. 25 I~m. bumps (Fig. 11). The fused tures. The detector is required to bumps form the mechanical and survive many thousands of thermal The next generation electrical connections between the cycles. Elastic deformation of the The loophole process produces de- two components. The flip-chip hy- CMT does not appear to affect the tectors of proven reliability and brid is illuminated through the back- device performance, but if disloca- performance and shares several ele- side of the detector array, on which tions are introduced then perfor- ments of the older photoconductor is deposited an anti-reflection coat- mance may suffer. The strain technology. Characteristically, the ing. produced in such an array can be process involves several polishing In common with the loophole enough to shatter an array substrate stages for the CMT and manual process, diode formation can be such as cadmium zinc telluride upon handling of individual dies, and de- achieved by photomasking and ion cooling to 80K. The ideal case would tectors are processed individually. milling an array of exposed CMT be to use a CMT layer grown epitaxi- The introduction of heterojunction windows. This results in local type ally on a silicon substrate and this has photodiodes offers the possibility of conversion to produce an array of n- been achieved for growth using the higher radiometric performance. type regions in the p-type LPE or MOVPE technique. Another option is GMIRL's research programme con- MOVPE layer. The p-n junctions can the removal of the array substrate tains elements of work aimed at: often be viewed by Nomarski micro- after fabrication of the hybrid device. • improving the radiometric perfor- scopy (a microscopic technique for Both possibilities have been shown to mance through structural and ma- enhancing topographical detail) after work successfully at GMIRL terial properties, and • cost reducing the manufacturing Figure 13. MOVPE process by eliminating some of the crystal grower. labour-intensive stages and intro- ducing a greater degree of wafer- scale processing.

Further cost reductions may be gained from the use of MOVPE- grown CMT for medium wavelength applications, and from the introduction of new types of detector using gallium arsenide-based materials to replace CMT.

Lowcost processing Futurc FPPAs will use technologies that are capable of producing very large linear and two-dimensional arrays, in both medium and long wave- bands, at low cost and with the best possible performance. In particular, fabrication using silicon-like wafer- scale processing offers production at the lowest cost. The loophole process is unique to GMIRL. An "alternative approach to FPPA manufacture which has been the subject of considerable interest elsewhere is the flip-chip hybrid [5]. In this structure, the CMT remains on its thick growth substrate for diode array formation, and indium bumps

Page 57 I ~!+~*""~:~;:t,,%ii~l:is~i~...... Vol 9 No 2 Heterostructures Loophole and flip-chip arrays have to be cooled to around 80K to ensure that the photo-signal is not swamped by dark currents. Any reduction in 3~rn --'--"-4 the cooling requirements would help ::::::::::::::::::::::::::::::::::::: to reduce the power consumption, weight, size and cost of an infrared system and it was with this aim in mind that research work on heterostructure diodes began in 1990. The practical realization of these devices ,~,::~,~::;i~:,:i~:~/.:! has been possible only by using MOVPE growth. The GMIRL system J shown in Fig. 13 was designed and built in-house and is fully computer controlled. Heterostructures consist of two or more layers of CMT with different band-gap and doping in each layer[6]. Position There are two basic structures of interest. The first is wide gap, heavily- Figure la. Electron (n) and hole (p) profiles through a heterostructure diode at zero bias doped p-type on narrow gap, lightly- (subscript O) and reverse bias. doped n-type (I2-n) and the second narrow gap, lightly-doped p-type sandwiched between wide gap, heavily-doped p and n (p-n-n). A £-n diode should have a lower dark current than a p-n diode because thermal generation in the p-region is virtually eliminated by making it wide gap. A £-n-n diode offers the possibility of a further reduction in dark current by suppression of Auger generation in the n-region. The operation of the p- n-n device can be understood with the help of Fig. 14 which shows how the electron and hole concentrations vary through the device. In reverse bias electrons are extracted from the n-region at the n-n junction and cannot be re-supplied readily at the p-x junction. As a result, the electron concentration fails and so does the dark current. The device behaves as though it is at a lower temperature and is known as a "non-equilibrium" Figure 15. Current-voltage characteristic of a non-equilibrium heterostructure diode. device. Non-equilibrium operation has been demonstrated using £-n-n devices with response extending out to used for laser range finding and Quantum well infrared 10 ~tm at room temperature. The, tracking. current-voltage characteristics of Heterostructures may also lead to photodetectors these devices have a region in re- totally new products. For example, Quantum well infrared photodetec- verse bias where the leakage current when operated in forward bias, p-x-n tors (QWIPs) belong to a new class of falls as voltage increases, which pro- devices emit infrared and so it should device made possible by the advent vides a graphic illustration of the be possible to make gas sensors of molecular beam epitaxy for mate- extraction phenomenon (Fig. 15). consisting of two diodes, one an rials growth. This technology allows The £-n-n device has proved to be emitter and the other a detector. By very thin layers of a few atoms noisier than expected at low fre- suitable choice of CMT composition, (quantum wells) to be grown whilst quencies and so the first applications the sensor could be made sensitive to preserving abrupt interfaces between are likely to be in uncooled versions particular gases, for example sulphur the layers, and opens up the possibi- of heterodyne detectors of the type dioxide or ozone. lity of new devices based on this

Page 58 ...... ii I ...... VOI 9 NO 2 list manufacturing has led to a wide product range and a broad spectrum of detector technology. This, combined with GMIRL’s strong research and development programme, is enabling the Company to meet the new challenge with innovative and high quality products.

Acknowledgements GEC-Marconi h&a-Red Ltd. would like to thank the European Space Agency, the UK DRA, the Hymatic Engineering Co., GEC-Marconi Sen- sors Ltd., for their support and co- operation in detector development programmes and in the preparation of this article.

References [ 1] Pmpetiies oJ Nawow Gap Cad mium- based Compoz/ndq P. Capper, Figure 16.160 x 120 element quantum well detector array (actual size of device c. (ed.) EMIS Datareviews Series, No. 10, 7 mm x 9 mm overall). 1994, Ch. Al. [2] A.P. Davis, UK thermal imaging common modules class II - an update on detector and related component en- hancements, SPIE,1157 (1989) 176. technique of band-gap engineering. well devices must be cooled to [3] C.T. Elliott, D. Day and D.J. Wilson, In particular, long waveband detector reduce the dark currents. In fact the /nfrareaP&@ 22 (1982) 31. structures can be designed in materi- cooling requirements are rather [4] I.M. Baker, G.J. Crimes, J.E. Parsons als such as gallium arsenide, provid- more severe than for CMT, necessi- and ES. O’Keefe, SPIE,2269 (1994) 636. ing the infrared detector industry tating the operation of a cooling [5] DA. Scribner, M.R. Kruer and J.M. with a new materiall’l. engine at around 6OK. Killiany, Infrared Focal Plane Array Gallium arsenide’s great attraction The device operates as a photo- Technology. P~oc /XVi$, 79( 1) (1991) 66. is in its wide range of existing conductor, but at sufficiently low [b]/?-o~eti&r of Narrow Gap Cad applications in fibre optics, high- temperatures the dark currents be- mzum-based Compozznd~ P. Capper, speed electronics, and microwave come insignificant and it is possible (ed.) EMIS Datareviews Series, No. 10, sources which have already provided to interface the quantum well array 1994, Ch. A8.4 a cheap, reliable source of device with a multiplexer in the same way as [ 71 Semr’co~a’&Ltor Quantum we/l’s grade material. GEC is well posi- a CMT hybrid. In GMIRL, hybridized and SuperZatiicesfor Long Wavelen@ tioned to exploit this technology arrays of 160 x 120 elements on a InzaredDefec~om M.O. Manesreh (ed.), with infrared expertise at GMIRL, 50 urn pitch have been demon- Artech House, 1993. combined with a gallium arsenide strated (Fig. 16), and work is progres- [8] S.R. Andrews and B.A. Miller, Experi- foundry at GMMT, Caswell, and with sing towards half-PAL TV format mental and theoretical studies of the theoretical input from the Hirst arrays (384 x 288 elements) of performance of quantum-well infrared Research Centre, where early devel- 20 urn pitch. photodetectors, J: Apple P!yx, 70(2) opment took placel’]. (1991) 993. The simplest form of QWIP con- sists of a stack of doped quantum Summary wells separated by thicker undoped The photoconductive CMT detectors Footnote barrier layers of a larger band-gap that have successfully supplied the This article was originally published material. Typical structures would needs of the Common Module Pro- in the GEC Review and we acknowl- consist of a stack containing several gramme and mainstream thermal edge GEC-Marconi Ltd. for permis- tens of quantum well/barrier layer imaging applications for many years sion to publish and would like to pairs. This structure can be grown as have now reached practical limits of thank the Editor, for his assistance in a thin film on a gallium arsenide performance and complexity, and the preparation of this feature. substrate and is amenable to fairly determine the performance limits of standard wafer processing techni- their host systems. A new generation Peter Knowles ques for the formation of device of higher performance imagers is Tel: f44J 90) 1703 316732. arrays. In common with other high within view, using very different Fax: (441 (0) 1703 316732. performance detectors, quantum detectors. GMIRL’s history of specia-

Page 60 Vol 9 No 2