NIR/SWIR FPAs and Applications

Near-infrared and shortwave infrared imaging—0.7 to 3 μm—continues to rise in popularity because of its ability to make use of sky glow to enhance performance during exceptionally dark periods of the night.

Development of SWIR FPAs is continuing in order to meet the needs of both military and commercial applications. Most of these developments make use of InGaAs, although new SWIR HgCdTe (MCT) FPAs are also being developed for applications that require longer cutoff wavelength—above 1.7 μm.

Pixel sizes of InGaAs FPAs are decreasing. Two com-panies reported on their 12.5 μm pitch 1280 × 1024 InGaAs FPAs. As the pixels get smaller surface leak-age current must be controlled in order to keep the dark current low. One company did this with a proprietary dielectric passivation as shown in Fig.1.

Fig. 1

Planar InGaAs/InP pixel structure used to fabricate 1280 × 1024 InGaAs FPA.

The other major source of dark current, junction diffusion current, was controlled by operating at an optimum bias voltage. The result was a total dark current of 0.7 nA/cm² at an operating temperature of 15 °C or 2.0 nA/cm² at an operating temperature of 20 °C.

Another company described its 1280 × 1024 and 640 × 512 InGaAs FPAs with 12.5 μm pixels and efforts to decrease total camera operating power. Rather than eliminating the thermo-electric cooler (TEC) completely, the company retained the TEC but used several temperature setpoints in order to reduce the TEC power. The result for a 640 × 480 camera is that power consumption remains below 2 W across a wide operating range, rising to 3 W at the extreme ends of the temperature range (-40 °C and +80 °C). This allows battery lifetimes of 6 hours or more using just three CR123A cells.

In another presentation, a 640 × 512 pixel, 20 μm pitch, InGaAs array was liquid nitrogen cooled for electroluminescence studies and astronomical applications. Total dark noise was only 5 electrons per pixel.

Two companies presented their results on MCT SWIR FPAs. One company grew high performance low cost MCT SWIR FPAs by MBE on 3” CdTe-buffered silicon substrates. Despite the large lattice mismatch between HgCdTe and Si substrate, the materials and detector performance are better than those reported for III-V mixed crystals. High minority carrier lifetime and low dark current densities were measured. Quantum efficiency exceeding 70% at 2.0 μm, without an-tireflective coating, was measured on single element detectors. FPAs were fabricated with this HgCdTe on Si material and imaging and radiometric characterization performed from 77K through room temperature.

Another company described their MCT SWIR FPAs (0.85 to 1.75 μm) that were grown lattice-matched on in-house fabricated CdZnTe substrates by a Liquid Phase Epitaxy (LPE) dipping process. 640×512 FPAs with 15 μm pixel pitch were processed in planar diode technology with n-on-p polarity. The company has adapted its high operating temperatures (HOT) tech-nology that was developed for MWIR detectors SWIR detectors.

The measured dark current densities are less than one order of magnitude within the values of the empirical Rule 07 for MCT detectors with a cut-off wavelength of 1.75 μm. Thus, at an operating temperature of 240 K the measured dark current density is 2 nA/cm². (A dark current density of 1 nA/cm2 or less is assumed to be required for applications under airglow.)

A comparison of the quantum efficiencies of 640 × 512 SWIR MCT FPAs with 15 μm pitch (at 210 K) and InGaAs FPAs at room temperature are shown in Fig. 2.

Fig. 2

Comparison of quantum efficiencies of InGaAs at room temperature and HgCdTe at 210 K.

For NIR applications, silicon-based imagers are continuing to be improved. In particular, CMOS imagers are used in a wide variety of applications, including large numbers of cell phone cameras which are replacing standalone digital cameras and are also starting to replace video camcorders due to the availability of low cost solid state memory and high quality miniaturized optics.

CMOS sensors have also demonstrated performance unachievable by any other semiconductor imaging array in high end TV, cinematography, slow motion and low light level cameras, thus revolutionizing digital image capture in general.

However, the adoption of CMOS sensors in space and astronomy instruments has been slow. Many new remote sensing systems and telescopes are still being equipped with CCDs, although, for longer exposure time applications, science CMOS image sensors (sC-MOS), have gained wide acceptance in remote sensing, biological imaging, and machine vision.

There may be two reasons why CMOS sensors have not found their way into these applications more frequently: (1) the high cost of space based systems or large telescopes which can lead designers to use field-proven solutions to minimize risk and avoid cost escalation and (2) the fact that state-of-the-art CMOS sensors are highly optimized for images visible to the human eye rather than in the NIR or UV that are critically important for scientific applications.

One of the presentations describes a backside illuminated CMOS image fabricated on high resistivity silicon. The use of a thick photosensitive membrane allows carrier depletion with low background doping concentration while maintaining low dark current and good MTF performance. A 640 × 512 sensor was shown to have high quantum efficiency over a wide spectral range and a fast photodetector response.

Type II Superlattice FPAs

There were a total of eleven papers in the two sessions devoted to Type II Superlattice and Barrier detectors, and several more on this subject in the Poster session. This reflects the continued strong interest in the potential performance advantages that this technology has been predicted to have theoretically—long carrier lifetimes and a high optical absorption coefficient. Experimentally, lifetimes as long as those predicted have not yet been achieved, but improvements have been made recently for material structures that exclude gallium. Lifetimes continue to be shorter than for HgCdTe with comparable bandgaps.

Using InAs/InAsSb superlattices, and avoiding Ga to achieve longer lifetime, LWIR and VLWIR FPAs were reported. Quantum efficiency was reduced however, compared with Ga-containing structures, with LWIR values about 18 % in the 8-10 μm region, but higher in the MWIR band. VLWIR D* was about 10¹⁰ Jones.

LWIR InAs/GaSb superlattices were reported. Homo-junction and hetrojunction designs were compared. Fig. 3 shows the dark current vs. bias and the photoresponse vs. bias for the designs measured. Dark currents in the low 10^-4 A/cm² were achieved for a 10 μm cutoff bandgap.

Fig. 3

Dark current density and photocurrent as a function of bias for homo- and hetro-junction Type II LWIR superlattice structures.

InAs/GaSb Type II superlattice pB_np barrier devices were reported that were designed using a modified k • p model. This paper also confirms the lower quantum efficiency of Ga-free alloys. For cutoff wavelengths in the range of 9.2 to 9.9 μm, the dark current is within about a factor of 10 of HgCdTe rule 07 as shown in Fig. 4.

Fig. 4

Comparions of dark current density values for serveral InAs/GaSb pB_np absorption layer thickness designs with HgCdTe Rule 07.

Bi-spectral, MWIR/MWIR Type II superlattice FPA performance was updated showing the impact of im-proved passivation in reduced high-noise tails in the histograms of NEΔT values—see Fig. 5. Improvements are attributed to better passivation technology. This update covers results from a production program for a threat warning application.

Fig. 5

NEΔT histograms with reduced high-noise tails from an ongoing MWIR/MWIR InAs/GaSb production program for a threat warning application.

MWIR InAs/GaSb superlattice pin photodiodes were evaluated for dark current, quantum efficiency, and the angular photo-response.

In another paper, MWIR InAs/GaSb superlattice detectors were combined with a plasmonic coupler to enhance the quantum efficiency. Fig. 6 illustrates how the plasmonic couplers were integrated into the structure.

Fig. 6

A plasmonic coupler is fabricated on top of an MWIR Type II superlattice detector using a combination of germanium and gold.

Papers were presented discussing surface passivation of Type II superlattices with a nitrogen plasma and with ZnTe. Figure 7 compares the effectiveness of a variety of passivation methods in suppressing dark current in small diodes. Chlorine-doped ZnTe appears to be a promising candidate.

Fig. 7

ZnTe:Cl and ECP provided the lowest dark current for small mesa diodes.

InAs pn photodiodes and nBn detectors were compared on lattice-matched substrates and on mis-matched substrates. Fig. 8 shows the dark current vs. inverse temperature for the case of nBn devices. Lattice matching yields about an order of magnitude dark current reduction.

Fig. 6

Lattice-matched InAs nBn devices show about an order of magnitude lower dark current than those on lattice mismatched substrates.

MWIR Type II superlattice detectors having an ‘N’ structure were described. The InAs-GaSb absorber layers incorporated an asymmetric AlSb barrier that the authors claim increases absorption and decreases dark current.

A model was presented for an ideal InAs nBn detector where the barrier layer is doped n-type and where the dark current is assumed to be only due to Auger-1 and radiative recombination. The model gives guidance for maximizing photocurrent and avoiding depletion regions.

Barrier detectors were compared to HgCdTe photodiodes, especially with respect to high operating temperature performance. Fig. 9 a) and b) show the comparison that was presented for the MWIR and LWIR bands (5 and 10 μm cutoff values). In spite of the many attractive features of the III-V materials, their dark current densities are higher than that of HgCdTe. To increase their competitive position, improvements in lifetime, passivation, and hetrostruc-ture engineering are needed.

Fig. 9

A comparison of NEΔT values for 5 cutoff MWIR (top) and 10 cutoff LWIR (bottom) for barrier detectors and HgCdTe photodiodes.

Advances in Detector Materials

The trend towards ever-larger FPAs for astronomy and other applications has led to the development of larger infrared crystal materials. A paper announced the recent development of 5- and 6-inch InSb crystals. Fig. 10 shows the progression of InSb crystal size.

Fig. 10

Progression of InSb crystal size and the number of wafers that can be cut from them.

SWIR detectors using InGaAs are also endeavoring to increase production capacities, in this case by MOVPE growth on multiple 4-inch InP wafers in a single run. One measure of the success of this approach is the carrier concentration depth profile shown in Fig. 11.

Fig. 11

Carrier concentration depth profile from a 2.5 μm thick InGaAs layer grown on a 4-inch InP substrate.

High Operating Temperature (HOT) FPAs

The goal of increasing the operating temperature of FPAs without sacrificing performance is motivated by the reduction in cooler power, improved cooler efficiency, longer cooler lifetime, smaller imager size, and lighter weight sensor systems that this makes possible. This goal is being pursued using HgCdTe, Type II superlattices, and nBn materials and has relevance especially in the MWIR and LWIR spectral bands.

Efforts to increase the operating temperature of HOT MWIR HgCdTe detectors while also reducing the number of defective pixels were summarized in one paper. Fig. 12 shows the increase in temperature at which the dark current and photocurrent become equal, while Fig. 13 illustrates the corresponding tem-perature dependence in defective pixel count.

Fig. 12

Progress in raising the temperature at which dark current equals photocurrent is illustrated.

Fig. 13

Progress in raising the temperature at which them number of defective pixels increase.

Another paper reported increases in HOT performance of HgCdTe MWIR FPAs. Fig. 14 shows how the NEAT distribution at 155 K with f/4 flux was tightened in the past two years, reducing the high noise tail that reflects the number of defective pixels. Fig. 15 shows imagery at three elevated temperatures — at 217 K the cooling power was less than 1 W.

Fig. 14

Reduction in the high-noise tail in an MWIR HgCdTe NEΔT histogram over the past two years.

Fig. 15

Infrared imagery from an MWIR HgCdTe FPA at three elevated temperatures.

XBn III-V InAsSb barrier detectors are being developed for HOT applications. One of these has a spectral range from 3.4 to 4.2 μm and operates at 150 K. A larger format—1280 × 1024—with 15 μm pixels was described. Fig. 16 shows the integrated detector-cooler package.

Fig. 16

Integrated detector-cooler package for a large-format XBn InAsSb barrier detector.

The absorption coefficient of Type II superlattice materials was the subject of a detailed study. It has been observed that Ga-free materials have longer carrier lifetimes than Ga-containing constituents. The absorption coefficient of a Ga-free MWIR superlattice at 160 K was compared to that of InSb at 80 K as shown in Fig. 17. The superlattice exhibits a softer spectral cutoff compared to direct bandgap alloy materials.

Fig. 17

Optical absorption coefficient comparison between InSb and two Ga-free Type II superlattice designs.

Monolithic PbSe FPAs operating at 230 K with an NEAT of 30 mK, f/1 at 400 Hz were reported. Imagery was shown—see Fig. 18. Pixel operability was over 99.5% for a 240 × 320 array having 60 μm pixels. This PbSe technology is an MWIR photon detector with a potential cost of production that compares favorably to uncooled microbolometers. Typical spectral D* values for PbSe are shown in Fig. 19.

Fig. 18

30 mK infrared imagery from an MWIR PbSe FPA at 230 K and 400 frames per second.

Fig. 19

Spectral D* for PbSe at temperatures between 250 and 78 K. The D* decreases at temperatures below 125 K because the spectral response is increasing—consult a theoretical D* vs wave-length chart..

HgCdTe

The HgCdTe alloy detector—characterized by a high absorption coefficient and a long lifetime—contin-ues to dominate the choice for a broad range of infrared applications. Aside from applications that are ideal for either InSb in the MWIR spectral band, or InGaAs in the 1.7 μm SWIR band, or those that can utilize uncooled FPAs, HgCdTe continues to be the most popular choice. Papers in this section update how HgCdTe is continuing to develop and evolve. Papers on this topic were presented in two sessions on HgCdTe detectors as well as in the HOT session, the Applications sessions, and in the Poster session.

An exciting paper was presented in which spherically-curved FPAs—both cooled HgCdTe and uncooled— were fabricated and tested. Fig. 20 shows the uncooled FPA after it was spherically shaped. Fig. 21 shows the interference pattern from a cooled HgCdTe FPA. Curved FPAs allow for optics designs that are more efficient. Detector properties were little-changed after the process. Improved MTF was illustrated for one particular design comparison.

Fig. 20

An uncooled FPA that has been spherically shaped to make more efficient use of optics designs.

Fig. 21

Interference fringes showing the uni-formity of curvature for a HgCdTe cooled FPA hybrid.

HgCdTe FPA developments for astronomy and spec-troscopy were described, as well for SWIR avalanche photodiodes. Electron-avalanche devices were reported for MOVPE grown devices that have separate absorption and avalanche regions. Fig. 22 shows the excess noise figure for an e-APD. Single-photon detection was reported. Large format arrays of 1032 × 1280 with 15 μm pitch having a source-follower per detector, and a 1048 × 2056 with 17 μm pitch have been made. Also described was an extended-range—2.5 to 10 μm—FPA capable of 1000 frames per second for an imaging Fourier spectrometer.

Fig. 22

Excess noise factors compared for LPE- and MOVPE-grown e-APDs.

New threats are prompting a push to higher resolution and larger format HgCdTe FPAs using MBE growth on GaAs substrates. A progression of FPAs has been proposed as follows:

The pixel size is envisioned to shrink as the format goes up in proportion to maintain the use of the same dewar package. Fig. 23 shows an array of 10 μm pixels in proportion to a human hair.

Fig. 23

An array of 10 μm HgCdTe pixels in comparison to a human hair.

Improvements in the performance and reliability of MWIR and LWIR HgCdTe detectors was described for n-on-p diodes. Among other things, the improved diodes—INP—showed much reduced random telegraph (also called popcorn or burst) noise. Fig. 24 shows the 50 × impact of the improved process on this phenomena. Device stability up to 15,000 hours was documented.

Fig. 24

The number of pixels showing random telegraph noise as a function of temperature for standard- and improved-n-on-p MWIR diodes.

Compact plasmonic metal-insulator-metal (MIM) metamaterial antenna designs for MWIR and LWIR bands were proposed in combination with HgCdTe absorbers.

HgCdTe p-on-n detector technology with extrinsic In- and As-doping has enabled R₀A performance to approach Rule 07 over a broad range of cutoff wave-lengths in the LWIR-to-VLWIR range. Fig. 25 shows the improvements compared to earlier n-on-p devices. A variety of SWIR diode results were also reported in this paper.

Fig. 25

Improved p-on-n technology has enabled an increase in HgCdTe R₀A values approaching Rule 07 over a wide range of LWIR-to-VLWIR cutoff wavelengths.

Reducing the pitch

Smaller pitch HgCdTe FPAs was the topic of one session. The challenges associated with obtaining small pitch— ~5 μm — were outlined, including:

Fig. 26 above shows candidate device structures. The goal is to also include operation at temperatures as high as room temperature. As part of this examination, one other detector material and device technology were compared to HgCdTe diodes varieties:

Fig. 26

Candidate pixel designs for small pitch: a) mesa p+n, b) HDVIP or Loophole, c) planar-implanted junction pIn.

An example of the dark current predicted is illustrated in Fig. 27.

Fig. 27

Modeled dark currents for five candidate detector material and structures between 100 and 300 K.

In a follow-up paper, the results of imaging with 5 pitch FPAs was presented. This was done with an HD-VIP pixel structure. Fig. 28 shows an image from a 720 × 1280 FPA with 5 μm pixels. Quantum efficiency was reported to be 45-55 % and LWIR operability was 99.6 %.

Fig. 28

LWIR image taken with a 720 × 1280 FPA having 5 μm pixels. An f/1 lens was used.

10 μm pitch FPAs in two formats—720 × 1280 and 768 × 1024—were announced for the MWIR band. Selectable charge storage was used to adjust the gain ranges from 0.7- to 4.4 × 10⁶ electrons, together with a dynamic range of 80 dB. Fig. 29 shows NEAT and operability as a function of temperature. Future consid-eration is being given to image dithering to effectively provide 5 μm pixel resolution with 10 μm pixels.

Fig. 29

NEΔT and operability for an MWIR FPA having 10 μm pixels as a function of operating temperature.

The benefits of small pixels were illustrated in a paper that first noted the significant reduction in false alarm rates that is possible because a true alarm will create a point-spread function blur in a group of pixels while a noise spike is likely to be confined to a single pixel. Fig. 30 illustrates this concept. The paper also described the development of a readout with a 2 Mpixel format of 5 μm pixels designed to provide correlation and noise reduction directly in the readout. Fig. 31 shows a photo of the readout.

Fig. 30

An over-sampled true-alarm signal will produce a group of pixels from the signal point-spread function while a noise spike will only be generated in a single pixel.

Fig. 31

A 2 Mpixel ROIC having 5 μm pixels and featuring on-chip correlation and noise reduction.

Another paper on this general topic of small pixels considered the implications of pixel reduction on the modulation transfer function—MTF. Experimental measurements were compared to results from finite el-ement modeling so that a variety of general mitigation factors could be considered. Fig. 32 compares experi-mental data and the model for two pixel sizes and for a variation of diffusion lengths. The impact of surface recombination and crystal potential fields were also considered along with mesa etch depth.

Fig. 32

MTF measurements for 30- and 15 μm pixels having several values of lateral collection length compared to calculations from a finite element model.

QWIP and Q-Dot FPAs

Quantum well and dot structures were discussed in three papers at the conference. In the first of these, an optical ring resonator was considered for boosting QWIP quantum efficiency. A variety of designs were modeled, followed by some experimental verification. Peak quantum efficiency of 30 % was achieved together with about 10 % collection efficiency. Higher values of quantum efficiency were modeled, but the model showed that there was a trend towards lower quantum efficiency with smaller pixels as shown in Fig. 33. A 640 × 512 FPA was demonstrated.

Fig. 33

Quantum efficiency as a function of wavelength for LWIR resonator-QWIPs with five different pixel sizes.

A second paper explored plasmonic structures to enhance quantum dot detectors—one design had narrow spectral response while another had a broader band response.

Colloidal PbS quantum dot devices were studied as a possible low-cost SWIR detector.

Emerging Uncooled

While microbolometers based on VO_x and α-Si have shown remarkable resilience against changes in mar-ket requirements for both performance and cost, there remains considerable interest in finding better ways to sense infrared radiation. Activities include new materials, more efficient absorbers, and innovative concepts. A U.S. university is investigating both nickel oxide and molybdenum oxide as materials having potential for improved process controllability by virtue of reduced sensitivity to the partial pressure of oxygen during material deposition. Molybdenum in particular is remarkable for the fact that the natural logarithm of its resistance varies linearly with temperature, as shown in Fig. 34, implying that its TCR is nearly constant with temperature.

Fig. 34

The natural logarithm of MoO_x resistance prepared at different ppO₂ decreases linearly with increasing of temperature.

In developments in Japan, an asymmetric plasmonic structure has been shown provide wavelength and po-larization selectivity of absorption.

A paper presented by a U.S. university described analysis and experiment that verified the existence of a repulsive electrostatic force sufficient to overcome the Casimir force, which is the source of the “stiction” problem in MEMS devices. That permits fabrication of a unique null-switching infrared detector having inherently digital output. In a poster paper they also analyzed the effect of thermo-mechanical noise on this detector. In this device a bi-material cantilever initially makes contact with an electrode on the substrate. A repulsive electrostatic force is suddenly activated, which moves the cantilever away from the substrate. A clock is then activated, and the applied voltage is ramped downward. The number of clock cycles until the cantilever again contacts the electrode is a measure of the radiation absorbed by the cantilever. The thermodynamic oscillation of the cantilever, primarily at its resonant frequency, is the dominant noise source.

An interesting approach by another U.S. university to achieve very high thermal isolation includes a weblike structure such as that shown in Fig. 35. The balanced mechanical design potentially permits long thermal conduction paths using nano-scale structures. So far, the effort has been limited to analysis, and the structures contemplated are large relative to state-of-the art bolometers. Nevertheless, it represents another arrow in the quiver of the pixel designer.

Fig. 35

Top view of the three designs used for simulation (a) Device 1: simple circular spider web with linear electrodes, (b) Device 2: meandered circular spider web with linear electrodes, and (c) Device 3: meandered circular spider web with meander electrodes.

In a very unusual approach in Germany, devices have been fabricated based on what is believed to be the operating mode of infrared receptors in the Mela-nophila acuminata jewel beetle. A micro chamber between two silicon wafers contains a fluid that expands upon absorption of infrared radiation, and in doing so moves one plate of a capacitor. A narrow channel con-nects the chamber to a reservoir, providing a high-pass filter that eliminates spurious response to gradual ambient temperature changes.

A South Korean institute has demonstrated films of VO_x /ZnO/VO with a TCR of -3.12 % and a resistance of about 80 kΩ/square in a structure of 60 nm VO_x, 10 nm ZnO and 10 nm VO_x. The constituent films were deposited at room temperature, and the composite structure was annealed at 300 °C in oxygen.

A poster paper from Japan presented the results of a design study of 3-dimensional plasmonic infrared ab-sorbers comprising a 2-dimensional array of micropatch antennas on posts that elevated them 150 nm above a bottom plate of the same material, gold. They found that the absorption was sharply peaked at a wavelength approximately proportional to the size of the micropatches, and that it was more sharply peaked if they were square rather than round. While the studied design included a relatively thick 200 nm bottom plate, it can be thinned significantly because of the small skin depth in gold. This technology is intended to enable detectors arrays wherein interspersed subarrays respond to different sub-bands of the IR spectrum. Along similar lines, a poster paper from Turkey showed enhanced absorption by using a planar array of plasmonic structures on the absorber.

A poster paper from a small U.S. business and a university presented results for a bolometer array using gold-doped VO_x, which lowered resistivity while maintaining a respectable TCR. Measured detector performance equates to a system NEΔT of less than 100 mK.

Uncooled

Uncooled IR microbolometer arrays continue to find application in many and varied places. Advancements include reductions in cost as well as improvements in performance, and the factors being addressed include packaging, readouts, pixel design and fabrication, and materials quality. Papers in the Uncooled FPAs and Applications session and the poster session were con-cerned with all these issues.

A new family of 17 μm microbolometer arrays from Germany includes a 640 × 480 version and a previ-ously introduced 320 × 240 version, both incorporating massively parallel 16-bit sigma-delta ADCs located under the active array. The average f/1 NEAT for those arrays is around 70 mK for a 30 Hz frame rate. The arrays are packaged chip scale; i.e., the array die are covered with an IR-transparent window separated from the array by a solder frame, as show in Fig. 36.

Fig. 36

Realization (left) and principle (right) of a chip-scale vacuum package.

The development of pixel-level packaging for very low-cost sensors continues to progress in France. Fig. 37 shows such a device. The pixels of a microbolometer array are individually sealed by depositing an IR-transparent silicon film on top of a second sacrificial layer. Small holes in the silicon film facilitate the release etch, and the holes are afterward plugged by deposition of an anti-reflection coating. An 80 × 80 demonstration array indicates the initial package vacuum supports f/1 NEAT in the range of 70 mK and pixel operability of 99.7 %.

Fig. 37

Pixel-level packaging: left—schematic; center—SEM top view; right—SEM cross section.

Another approach to low-cost sensors is in development in Turkey—see Fig. 38. It uses bulk micromachining for more rigorous compatibility with a standard CMOS process, sacrificing sensitivity in the interest of substantial cost reductions. A standard metal layer serves as the mask for etching away silicon to release the bolometer structure. The sensor under development is a 160 × 120 array of 70 μm pixels operating at 30 Hz. The device power dissipation is 50 mW with analog output, and it is packaged at the wafer level. The f/1 NETD is less than 350 mK in a 1.65-in³, 15-gram camera dissipating 625 mW.

Fig. 38

Illustration of an ideal fabrication process as only one mask process step is required and no deposition step is used since the CMOS devices are used as the temperature sensing active material. The silicon beneath the detector bridge is etched in order to thermally isolate the detector bridge.

A U.S. university is investigating the tradeoffs of re-sistance, TCR and 1/f noise in VO. The presence of 1/f noise in standard bolometric materials remains a limitation of bolometer technology. Increasing material resistivity generally increases TCR, but it also increases 1/f noise. The precise nature of the tradeoff is complex, because dynamic range, power dissipation and variability with ambient temperature also enter into the picture. The understanding is further complicated by the dependence of 1/f noise on the degree of crystallinity and the details of the deposition conditions.

The cost, performance, size and weight of uncooled IR sensors have made possible their implementation in smart munitions. A potentially serious impediment to their widespread use in such applications is the ability of the seemingly fragile pixel structures to withstand the excessive shock environments association with the firing and/or launching of munitions. A U.S. university has demonstrated that a microbolometer sensor, properly configured, can survive a 20,000-g shock and maintain it original performance.

In another development, a Canadian company has grown vanadium oxide films having nonuniformity of less than 2.5% rms across the entire 150 mm wafer, a four-fold improvement over previous results. This improvement was measured using a 4-point probe, and the results were confirmed using resistor bridge structures.