Infrared focal plane array technology has evolved dramatically over the last 50 years. The author has been privileged to participate in this remarkable evolution, working totally within the confines of one of the most significant and remaining US players in the focal plane game, namely Texas Instruments, later to become DRS Technologies. This presentation describes a journey from the Common Module through second and third generation infrared systems in the USA up to the exciting developments of the present day ultra-small pixel technology. It represents an attempt to detail both the technology development of the time together with some of its associated drama as viewed from the author’s particular perspective. Thoughts on the lessons learned from this journey and their possible impact on future technology development will be discussed.
Infrared detector pixel pitch has been decreasing, driven by interest in higher resolution, larger displays, and decreased cost. Previous generations of focal plane arrays (FPAs) were on 50, 40, 30, and 20μm pitch. 12μm pitch FPAs are now available. DRS Network and Imaging Systems has developed ultra-small 5μm pitch infrared detectors for the long-wave infrared (LWIR) and medium-wave infrared (MWIR) bands as part of the DARPA AWARE Lambda Scale effort. The smaller pitch was achieved using DRS’ high-density vertically integrated photodiode (HDVIP®) architecture. This technology is a major advance in the state of the art for infrared imaging sensors. The pixel density of 4 million pixels/cm2 enables the production of lower cost FPAs from HDTV resolution up to many millions of pixels. Dark current, collection efficiency, cross-talk, and operability are similar to larger pitch HDVIP FPAs.
Large area HgCdTe focal plane arrays (FPAs) are now available with a 5um pitch, and excellent performance. This analysis examines the benefits associated with ultra-small pixels in enabling not only a reduction in system size, weight, and power, but also an improvement in system thermal performance. A comparison is made between today’s III-V and HgCdTe materials technologies, regarding both performance in today’s FPAs, and the ultimate achievement of background- and diffraction-limited photon detection at room temperature for all spectral bands.
In imaging systems, whether visible or infrared, the pixel dimension plays a crucial role in determining critical
system attributes such as size, weight, and Power (SWaP). Smaller pixels enhance the value proposition of the
imager through reduced cost Focal Plane Arrays (FPAs) and/or added system functionality for a given spatial
footprint. For systems that operate at temperatures in which FPA cold shield efficiency is relevant an additional
benefit to performance is achieved with the faster optics mandated by use of small pixels. Ultimate pixel dimensions
are limited by diffraction effects from the aperture and are in turn wavelength dependent. Limits to the reduction in
pixel dimensions will be explored and related to the historical trends in system design with accompanying
performance attributes. Key challenges in realizing ultimate pixel dimensions in focal plane array design will be
discussed. Progress toward these limits at DRS will be reviewed for LWIR HgCdTe Focal Plane arrays fabricated
with 5 micron pixel dimensions. Possible system implications tied to the success of these shrinking pixel FPAs will
The behavior of the gain-voltage characteristic of the mid-wavelength infrared cutoff HgCdTe linear mode avalanche photodiode (e-APD) is discussed both experimentally and theoretically as a function of the width of the multiplication region. Data are shown that demonstrate a strong dependence of the gain at a given bias voltage on the width of the n− gain region. Geometrical and fundamental theoretical models are examined to explain this behavior. The geometrical model takes into account the gain-dependent optical fill factor of the cylindrical APD. The theoretical model is based on the ballistic ionization model being developed for the HgCdTe APD. It is concluded that the fundamental theoretical explanation is the dominant effect. A model is developed that combines both the geometrical and fundamental effects. The model also takes into account the effect of the varying multiplication width in the low bias region of the gain-voltage curve. It is concluded that the lower than expected gain seen in the first 2×8 HgCdTe linear mode photon counting APD arrays, and higher excess noise factor, was very likely due to the larger than typical multiplication region length in the photon counting APD pixel design. The implications of these effects on device photon counting performance are discussed.
Reducing an array’s pixel pitch reduces the size and weight of the focal plane array (FPA) and its associated dewar,
cooler and optics. Higher operating temperatures reduce cool-down time and cooler power, enabling reduced cooler size
and weight. High operating temperature small pitch (≤15 um) infrared detectors are therefore highly desirable. We have
characterized a large number of MWIR and LWIR FPAs as a function of temperature and cutoff wavelength to
determine the impact of these parameters on the FPA’s dark current, 1/f noise and defects. The 77K cutoff wavelength
range for the MWIR arrays was 5.0-5.6 um, and 8.5-11 um for the LWIR arrays. DRS’ HDVIP® FPAs are based on a
front-side illuminated, via interconnected, cylindrical geometry, N+/N/P architecture. An FPA’s 1/f noise is manifested
as a tail in the FPA’s rmsnoise distribution. We have found that the model-independent nonparametric skew
[(mean–median)/standard deviation] of the rmsnoise distribution is a highly effective tool for quantifying the magnitude
of an FPA’s 1/f noise tail. In this paper we show that a standard FPA’s 1/f noise varies as ni (the intrinsic carrier
concentration), in agreement with models that treat dislocations as donor pipes located within the P-volume of the unit
cell. Nonstandard FPAs have been observed with systemic 1/f noise which varies as ni2.
The utilization of the non-equilibrium photodiode concept for high operating temperature (HOT) FPAs is discussed, both generically, and with regard to the specific example of MWIR HgCdTe. The issues of dark current, surface passivation, and 1/f noise are considered for three different architectures, namely N+/N-/P+, N+/P-/P+, and nBn. These architectures are examined with regard to possible FPA performance limitations, and potential difficulty in reduction to practice. Performance data obtained at DRS for the N+/N-/P+ and N+/P-/P+ HgCdTe architectures will be presented.
The High Operating Temperature Auger suppressed infrared detector concept is being pursued using the high density
vertically integrated photodiode (HDVIP®) architecture and an n+-p device structure. Dark current densities as low as 2.5
mA/cm2 normalized to a 5 μm cutoff at 250K have been demonstrated on these diodes. These dark currents imply
minority carrier lifetimes in excess of 300μsec. 1/f noise in these devices arises from the tunneling of charge into the
passivation interface, giving rise to a modulation in the surface positive charge and hence to the width of the depletion
region in the p-side of the device and a modulation in the total dark current. The measured 1/f noise is in agreement with
the predictions of this model, with very low noise being observed when the lifetimes are high.
Work on HgCdTe began at Texas Instruments in the early 1960s, and continued through 1997 when TI's defense business was sold first to Raytheon, and subsequently in 1998 to DRS Technologies. This presentation traces the history of HgCdTe's evolution throughout this timeframe to the present day, as viewed through the eyes of the author and several of his TI contemporaries who have survived the experience. The materials technology will be traced from the early days of bulk growth by the solid state recrystalization technique, through the traveling heater method of growth, to liquid phase epitaxy from large Te-rich melts, to vapor phase growth by molecular beam epitaxy and metal organic chemical vapor deposition. The evolution of detector device architectures at TI over the years will be discussed, from the early, successful days of photoconductors and the Common Module System, through the somewhat problematic and relatively unsuccessful foray into charge coupled and charge injection devices for 2nd generation FPAs for the Javelin program, to the outstandingly successful development of the vertically integrated photodiode (VIP) and high density VIP FPA architectures for mono-color and multi-color 3rd generation systems. The versatile, and unique nature of this infrared semiconductor materials system will be highlighted by reference to current work at DRS Technologies into electron avalanche photodiodes (EAPDs), for use in active/passive IR systems, and high operating temperature (HOT) detectors, which threaten to eventually offer BLIP photon detection at uncooled operating temperatures, over the whole IR spectrum from 1 to 12um.
The operation of the mid-wave infrared (MWIR) HgCdTe cylindrical electron injection
avalanche photodiode (e-APD) is described. The measured gain and excess noise factor are
related to the to the collection region fill factor. A 2D diffusion model calculates the time
dependent response and steady state pixel point spread function for cylindrical diodes, and
predicts bandwidths near 1 GHz for small geometries. A 2 μm diameter spot scan system
was developed for point spread function and crosstalk measurements at 80 K. An electron
diffusion length of 13.4 μm was extracted from spot scan data. Bandwidth data are shown
that indicate bandwidths in excess of 300 MHz for small unit cells geometries. Dark current
data, at high gain levels, indicate an effective gain normalized dark density count as low as
1000 counts per μs per cm2 at an APD gain of 444. A junction doping profile was
determined from capacitance-voltage data. Spectral response data shows a gain independent
DRS LPE-grown SWIR, MWIR and LWIR HgCdTe material are fabricated in the High-Density Vertically
Integrated Photodiode (HDVIP) architecture. Instruments manufactured for certain strategic applications have
severe constraints on excess low frequency noise due to the effect the noise has on the image quality with
subsequent consequences on the period of calibration. This paper will present data and analysis of excess low
frequency noise in LWIR (&lgr;c ~ 10.5 &mgr;m @ 60 K) HDVIP HgCdTe detectors.
The vehicle for noise measurements is a multiplexed 320 x 6 array of 40 &mgr;m x 50 &mgr;m, 10.5 &mgr;m cutoff, HgCdTe
detectors. Noise has been measured on a column of 320 detectors, at 60 K, as a function of frequency at zero and 50
mV reverse bias. Integration time for the measurement was 1.76 ms. Output voltage for the detectors was sampled
every 10th or every 100th frame. 32,768 frames of time series data were collected for a total record length of 98
minutes. Since the total time for collecting the 32,768 time data series points is 98 minutes, the minimum frequency
is 170 &mgr;Hz. Time series and Fourier transform data on individual detectors at 0 mV and 50 mV reverse bias in the
dark have been studied. Examination of the detector current time series and Fourier transform curves thereof, reveal
a variety of interesting characteristics: (i) time series displaying switching between four states characteristic of
random telegraph signal (RTS) noise, the noise current power spectrum having Lorentzian type characteristics; (ii)
time series data exhibiting slight wave-like characteristics with the noise current power spectrum being 1/f-like at
low frequencies; (iii) pronounced wave-like characteristics in the time series with the noise current power spectrum
being 1/f2-like at low frequencies; and (iv) time series having a mean value independent of time with the noise
current power spectrum being white. The predominance of detectors examined had minimal excess low frequency
noise down to ~ 10 mHz. In addition some isolated diodes had characteristics that lay between the four main types
DRS uses LPE-grown SWIR, MWIR and LWIR HgCdTe material to fabricate High-Density Vertically Integrated
Photodiode (HDVIP) architecture detectors. 2.5 μm, 5.3 μm and 10.5 μm cutoff detectors have been fabricated into
linear arrays as technology demonstrations targeting remote sensing programs. This paper presents 320 x 6 array
configuration technology demonstrations' performance of HDVIP HgCdTe detectors and single detector noise data. The
single detector data are acquired from within the 320 x 6 array. Within the arrays, the detector size is 40 μm x 50 μm.
The MWIR detector array has a mean quantum efficiency of 89.2% with a standard deviation to mean ratio, σ/μ = 1.51%. The integration time for the focal plane array (FPA) measurements is 1.76 ms with a frame rate of 557.7 Hz.
Operability values exceeding 99.5% have been obtained. The LWIR arrays measured at 60 K had high operability with
only ~ 3% of the detectors having out of family response. Using the best detector select (BDS) feature in the read out
integrated circuit (ROIC), a feature that picks out the best detector in every row of six detectors, a 320 x 1 array with
100% operability is obtained. For the 320 x 1 array constituted using the BDS feature, a 100% operable LWIR array
with average NEI value of 1.94 x 1011 ph/cm 2/s at a flux of 7.0 x 1014 ph/cm2/s has been demonstrated.
Noise was measured at 60 K and 50 mV reverse bias on a column of 320 diodes from a 320 x 6 LWIR array.
Integration time for the measurement was 1.76 ms. Output voltage for the detectors was sampled every 100th frame.
32,768 frames of time series data were collected for a total record length of 98 minutes. The frame average for a
number of detectors was subtracted from each detector to correct for temperature drift and any common-mode noise.
The corrected time series data was Fourier transformed to obtain the noise spectral density as a function of frequency.
Since the total time for collecting the 32,768 time data series points is 98.0 minutes, the minimum frequency is 170 μHz.
A least squares fit of the form (A/f + B) is made to the noise spectral density data to extract coefficients A and B that
relate to the 1/f and white noise of the detector respectively. In addition noise measurements were also acquired on
columns of SWIR detectors. Measurements were made under illuminated conditions at 4 mV and 50 mV reverse bias
and under dark conditions at 50 mV reverse bias. The total collection time for the SWIR detectors was 47.7 minutes.
The detectors are white noise limited down to ~10 mHz under dark conditions and down to ~ 100 mHz under
Electron injection avalanche photodiodes in SWIR to LWIR HgCdTe show gain and excess noise properties indicative of a single ionizing carrier gain process. The result is an electron avalanche photodiode (EAPD) with "ideal" APD characteristics including near noiseless gain. This paper reports results obtained on long-wave, mid-wave, and short wave cutoff infrared HgCdTe EAPDs that utilize a cylindrical "p-around-n", front side illuminated, n+/n-/p geometry that favors electron injection into the gain region. These devices are characterized by a uniform, exponential, gain voltage characteristic that is consistent with a hole-to-electron ionization coefficient ratio, k, of zero. Gains of greater than 1000 have been measured in MWIR EAPDS without any sign of avalanche breakdown. Excess noise measurements on MWIR and SWIR EAPDs show a gain independent excess noise factor at high gains that has a limiting value less than 2. At 77 K, 4.3 μm cutoff devices show excess noise factors of close to unity out to gains of 1000. The excess noise factor at room temperature on SWIR EAPDs, while still consistent with the k = 0 operation, approaches a gain independent limiting value of just under 2. The k = 0 operation is explained by the band structure of the HgCdTe. Monte Carlo modeling based on the band structure and scattering models for HgCdTe predict the measured gain and excess noise behavior. A noise equivalent input of 7.5 photons at a 10 ns pulsed signal gain of 964, measured on an MWIR APD at 77 K, provides an indication of the capability of the HgCdTe EAPD.
An overview on DRS' approaches towards realization of HgCdTe photonic infrared detectors based on DRS's proven HDVIP technology is given. The first approach involves the use of a silicon microlens array attached to the detector array, and the second reduction of dark currents in each detector itself. Recent progress is presented.
An overview of the DRS HDVIP architecture for realization of large-area infrared focal plane arrays (IRFPAs) is given. Improvements needed to meet more stringent application requirements are discussed and modeled. Both theoretical and experimental data are presented.
Cu-doped HDVIP detectors with different cut-off wavelengths are routinely manufactured. The DRS HDVIP detector technology is a front-side-illuminated detector technology. There is no substrate to absorb the visible photons as in backside-illuminated detectors and these detectors should be well suited to respond to visible light. However, HDVIP detectors are passivated using CdTe that absorbs the visible light photons. CdTe strongly absorbs photons of wavelength shorter than about 800 nm. Detectors with varying thickness of CdTe passivation layers were fabricated to investigate the visible response of the 2.5-μm-cutoff detectors. A model was developed to predict the quantum efficiency of the detectors in the near infrared and visible wavelength regions as a function of CdTe thickness. Individual photodiodes (λc = 2.5 μm) in test bars were examined. Measurements of the quantum efficiency as a function of wavelength region will be presented and compared to the model predictions.
Detector characteristics of Au- and Cu-doped High Density Vertically Integrated Photodiode (HDVIP) detectors are presented in this paper. Individual photodiodes in test bars were examined by measuring I-V curves under dark and illuminated conditions at high bias values. Noise as a function of frequency has been measured on Au- and Cu-doped MWIR [λc(78 K) = 5 μ] HDVIP HgCdTe diodes at several temperatures under dark and illuminated conditions. No excess currents are observed above the photocurrents for reverse bias values out to 500 mV. Both Au- and Cu-doped detectors measured at 85 K, exhibit gain values between 40 and 50 at 8 V reverse bias. Gain values fell in this same range even when the flux incident on each type of detector was varied. The excess noise factor for the Cu-doped detectors ranged from 1.35 to 1.69 depending on the incident flux. Variation is probably due to measurement error. The noise at 8 V reverse bias is white for the Cu-doped detectors. The Au-doped detectors exhibited 1/f noise at 8 V reverse bias. At higher frequencies where the noise spectrum was quasi-white, the excess noise factor for the Au-doped detector was in the 1.0 to 1.5 range.
Detector characteristics of Cu- and Au-doped High Density Vertically Integrated Photodiode (HDVIP) detectors as well as Cu-doped HDVIP Focal Plane Arrays (FPAs) are presented in this paper. Individual photodiodes in test bars were examined by measuring I-V curves and the associated resistance-area (RA) product as a function of temperature. The Au-doped MWIR [λc(78 K) = 5 μm] HDVIP detectors RoA performance was within a factor of two or three of theoretical. Noise as a function of frequency has been measured on Au-doped MWIR HgCdTe HDVIP diodes at several temperatures under dark and illuminated conditions. Low-frequency noise performance of the Au-doped MWIR diode in the various environments is characterized by the ratio α of the noise current spectral density at 1 Hz to the value of the diode current. For photocurrent at 140 K, αPHOTO = 1.8 x 10-5. The value of αPHOTO is the same at both zero bias and 100 mV reverse bias. At 160 K, αPHOTO is slightly lower but still in the low 10-5 range. Excess low-frequency noise measured at 140 K and 100 mV reverse bias in the dark has αDARK = 1.4 x 10-5. At 160 K and 100 mV reverse bias, αDARK is in the mid 10-5 range. At 140 K,the dark current at 8.2 V reverse bias was equal to the photocurrent at 100 mV reverse bias and close to the photocurrent at zero bias. αDARK = 1.85 x 10-3 at -8.2 V. This ratio is two orders of magnitude greater than αPHOTO. At 8.2 V reverse bias, the current was amplified by avalanche processes. Similar results were obtained on the Au-doped diode at 160 K. Diffusion current dominates dark current at 100 mV reverse bias at T = 185 K and T = 220 K. The ratio, αDARK approximately αPHOTO in the low to mid 10-5 range, i.e. dark diffusion current generates excess low frequency noise in the same manner as photocurrent. In addition, 256 x 256 Cu-doped detector arrays were fabricated. Initial measurements had seven out of ten FPAs having operabilities greater than 99.45% with the best 256 x 256 array having only two inoperable pixels.
This paper reports results obtained on mid-wave IR x equals 0.3 Hg1-xCdxTe avalanche photodiodes (APDs) that utilize a cylindrical 'p-around-n' front side illuminated n+/n-/p geometry. This 'p-around-n' geometry favors electron avalanche gain. These devices are characterized by a uniform, exponential, gain voltage characteristic that is consistent with a hole to electron ionization ratio, k equals (alpha) h/(alpha) e, of zero. At 6 bias and 77 K, gains are typically near 50, and gains of over 100 have been measured at higher biases. Response times have been modeled and measured on these devices. The modeling indicates that the geometry and dimensions of the diode control the diffusion limited device bandwidth. Rise times of less than 0.35 nsec should be possible according to this analysis. TO dat 10 percent to 90 percent rise times as low as 1 nsec have been measured. The gain is approximately noiseless up to gains of over fifty which is consistent with insignificant hole ionization. The noiseless gain behavior reported here is inconsistent with the original theory of McIntyre that predicts an excess noise factor of 2 for the k equals 0 case. The explanation for these results will require application of the modified 'history dependent' theory for excess noise later proposed by McIntyre.
Third generation IR focal plane arrays will be required to operate at significantly higher temperatures than utilized today. The ultimate aim is operation at room temperature, for nay desired cutoff wavelength in the complete IR spectral bandwidth of 1 to 14 micrometers , with performance characteristics equivalent to those achieved today at 77K. Thermal detectors offer a limited capability of meeting these requirements, particularly for any system not operating at LWIR with a slow frame rate. However, the HOT detector concept, first proposed by Elliott and Ashley, offers the promise of uncooled photon detector across the complete range of the IR spectrum at high speeds. This paper discusses the materials and device properties that are important to successfully reduce this concept to practice, together with the sate of the art in HOT detectors today.
The HgCdTe high-density vertically integrated photodiode (HDVIPTM) concept developed at DRS Infrared Technologies is described. This technology is currently in production in both large-area scanning and staring focal plane array (FPA) formats. Detector models are presented and compared to performance data from scanning and staring FPAs. Performance data from 256 X 256 and 640 X 480 LWIR and MWIR staring FPAs, in keeping with these models, is presented with responsivity and D* operabilities in excess of 99.9%. Third generation system requirements mandate megapixel FPA operation at high temperatures, with multi-color capability, and high frame rates. To this end operation of 640 X 480 MWIR HgCdTe FPAs has been demonstrated at temperatures in excess of 150 K, and the push to these higher operating temperatures, with its effect on system cost, is discussed. The technology has also been extended into the realm of simultaneous two-color detection with large area formats, and this effort is described.
The fundamental parameters of IR photon detection are discussed relevant to the meaningful comparison of a wide range of proposed IR detecting materials systems. The thermal generation rate of the IR material is seen to be the key parameter that enables this comparison. The simple materials physics of (1) intrinsic direct bandgap semiconductors, (2) extrinsic semiconductors, (3) quantum well devices, including types I, II, and III superlattices, (4) Si Schottky barriers, are examined with regard to the potential performance of these materials as IR detectors, utilizing the thermal generation rate as a differentiator. The possibility of room temperature photon detection over the whole IR spectral range is discussed, and comparison made with uncooled thermal detection.