Full text of DOI:10.1007/BF02665839

Tennant, Thomas, Kozlowski, McLevige, Edwall, Zandian,
5Jo9ur0nal of ELECTRONIC MATERIALS, Vol. 30, No. 6, 2001
Special Issue Paper
Spariosu, Hildebrandt, Gil, Ely, Muzilla, Stoltz, and Dinan
A Novel Simultaneous Unipolar Multispectral Integrated
Technology Approach for HgCdTe IR Detectors and
Focal Plane Arrays
W.E. TENNANT,1,5 M. THOMAS,1 L.J. KOZLOWSKI,1 W.V. MCLEVIGE,1
D.D. EDWALL,1 M. ZANDIAN,1 K. SPARIOSU,1 G. HILDEBRANDT,1 V. GIL,1
P. ELY,2 M. MUZILLA,2 A. STOLTZ,3 and J.H. DINAN4
1.—Rockwell Science Center, 1049 Camino Dos Rios, Thousand Oaks, CA 91360. 2.—Boeing
Electronic Systems and Missile Defense, Anaheim, CA 92803. 3.—E-OIR Measurements, Inc.,
Spotsylvania, VA 22553. 4.—Night Vision & Electronic Sensors Directorate (NVESD), Fort Belvoir,
VA 22060-5806. 5.— wetennant@rsc.rockwell.com
In the last few years Rockwell has developed a novel simultaneous unipolar
multispectral integrated HgCdTe detector and focal plane array technology that
is a natural and relatively straightforward derivative of our baseline double
layer planar heterostructure (DLPH) molecular beam epitaxial (MBE) technol-
ogy. Recently this technology was awarded a U.S. patent. This simultaneous
unipolar multispectral integrated technology (SUMIT) shares the high perfor-
mance characteristics of its DLPH antecedent. Two color focal plane arrays with
low-1013 cm–2s–1 background limited detectivity performance (BLIP D*) have
been obtained for mid-wave infrared (MWIR, 3–5 mm) devices at T > 130 K and
for long-wave infrared (LWIR, 8–10 mm) devices at T ~ 80 K.
Key words: HgCdTe, MBE, MWIR, LWIR
INTRODUCTION
IRFPAs for specific applications. An integrated tech-
Various applications are beginning to demand si-
multaneous detection in multiple spectral bands.
Simultaneity (with full pixel coverage) assists vari-
ous algorithms quickly to distinguish threats from
temporal background clutter. Having a single (uni-)
polarity allows the input cell of each spectral band to
be optimized for its particular function without re-
gardtothepresenceofothercolordetectorswithinthe
pixel. Multispectral detection permits rapid and effi-
cient understanding of the scene in a variety of ways.
In particular, two-color infrared focal plane arrays
can be especially beneficial for threat-warning appli-
cations. By using two infrared wavebands, spurious
information, such as background clutter and sun-
glint, may be subtracted from an infrared image,
leavingonlytheobjectsofinterest. Thus, theeffective
signal-to-noise ratio of two-color infrared focal plane
arrays (IRFPAs) greatly exceeds that of single-color
nology greatly assists the producibility and high pixel
densityneededfortheconstantlyincreasingdemands
ofpixelcountandeconomy.Awidevarietyofhigh-end
commercial and military applications would likely
exist for a technology with this flexibility.
We have developed a two-color sensor technology
that is an evolutionary extension of current single-
colortechnology.Muchofthelayerstructure,detector
processing, and readout design are identical to their
single-color counterparts, but each has been modified
slightly to provide simultaneous two-color operation.
Because much of the single-color technology is pre-
served, RockwellandBoeing’ssimultaneousunipolar
multispectral integrated technology (SUMIT)1 archi-
tecture is able to overcome many of the difficulties
and limitations of other approaches.2,3
FABRICATION
Our two-color architecture is an extension of
Rockwell’s single-color double layer planar hetero-
(Received November 21, 2000; accepted January 13, 2001)
590

A Novel Simultaneous Unipolar Multispectral Integrated Technology
Approach for HgCdTe IR Detectors and Focal Plane Arrays
591
structure (DLPH) architecture.4-8 The important ele-
mentsoftheDLPHstructurethataremaintainedare
an In-doped (n-) Hg1–xCdxTe absorber layer with a
wider bandgap cap layer grown by molecular beam
epitaxy (MBE), arsenic implanted (p+) diodes, a pas-
sivation layer, metal contacts to the diode, and in-
dium bumps for mating to the readout integrated
circuit(ROIC)multiplexer(MUX).Thesubstrateused
for MBE growth is typically CdZnTe, but use of
Si substrates may be possible. The SUMIT two-color
architecture extends this design in several aspects.
To accommodate two colors, extra layers are grown in
the heterostructure (Fig. 1).
The Band 1 absorber (shorter wavelength) is grown
first, with the Band 2 absorber (longer wavelength)
grownontop.Awide-bandgapbarrierlayerseparates
these absorbing layers to prevent diffusion of carriers
between them. The entire structure is capped with a
slightly wider bandgap layer to reduce surface recom-
bination and simplify passivation. The SUMIT fabri-
cation process differs from the DLPH process prima-
rily in two steps. As in DLPH, SUMIT diodes are
formed by implanting arsenic as a p-type dopant and
activating with an anneal (giving unipolar operation
for both bands). However, the Band 1 absorber must
firstbeexposedbeforeimplantation, aprocessaccom-
plished by wet chemical etching for the devices re-
ported here. In addition, because the lateral carrier
diffusion lengths are larger than the pixel pitch in the
mid-wave infrared (MWIR) material and the Band 1
junctions are small, the pixels must be isolated by
dry-etching a trench around each pixel [a process
developed by the Night Vision & Electronic Sensors
Directorate(NVESD)]. Thisetchstepisnotnecessary
for two-color operation, but is necessary to improve
the modulation transfer function (MTF) for these
MWIR-MWIR devices. The remainder of the process
is identical to the DLPH process, with a passivation
layer deposited, metallic contacts made to the diodes,
and indium bumps deposited for ROIC hybridization.
Amicrographshowingacrosssectionofatypicalpixel
is shown in Fig. 2.
Since the pixel architecture consists of two inde-
pendent p-on-n diodes, the MUX is a very simple
extension of a standard single-color device. Each
MUXcellconsistsoftwoseparatesingle-colorreadout
cells. In fact, initial demonstrations of the SUMIT
devices used a single-color MUX mated to an array of
detectors with pixels rotated 45∞ and increased by ÷2
in linear dimensions to match the cell inputs to the
detector indium bumps. A readout designed specifi-
callyforthe128¥ 128two-colorarrays,theTCM1060,
is actually a 256 ¥ 128 single-color MUX, albeit with
a non-square cell pitch. The TCM1060 is fabricated in
0.5 mm complementary metal-oxide semiconductor
(CMOS) technology using Conexant’s (previously
Rockwell’s) CMOS fabrication line. The MUX archi-
tecture is a simple direct injection input cell. The
integrating capacitor for each color is the same size,
but the capacitors could easily be fabricated asym-
metric to account for different flux levels in the differ-
ent bands. In fact, each color could, in principle, have
an entirely different input cell architecture and sepa-
rate timing and biases. The Band 1 and Band 2
detectors may be integrated simultaneously—an im-
portant feature for threat-warning systems. With no
fundamental changes, the simple MUX architecture
also allows extension of the SUMIT approach to more
than two colors.
RESULTS
Our major focus has been on MWIR-MWIR
two-color arrays. Before hybridizing and testing the
focal plane arrays (FPAs), we measured optical and
electrical performance of test devices on the two-color
layers. The measured optical characteristics include
spectral response and quantum efficiency, while the
electrical characteristics measured are R0A, dark
a
b
Fig. 1. (a) Single-color (DLPH) and (b) two color (SUMIT) structures.
Light enters through the substrate absorber. The Band 2 implanted
area is a concentric ring around the Band 1 dimple. V-groove pixel
isolation trenches deep enough to cut into the (slightly graded) Band
1 layer surround the pixel to reduce carrier cross-talk. The first and
secondbandabsorbersareseparatedbyawide-bandgapbarrierlayer
~1 mm thick. The second band is capped with a slightly wider-bandgap
layer for passivation, as in the DLPH structure.
Fig. 2. SUMITscanningelectronmicrograph(SEM)crosssection. The
‘V’-shaped features are the dry-etched pixel-isolation trenches. The
Band1diodeisimplantedatthebottomofthewet-etcheddimpleinthe
center. The Band 2 diode is implanted at the top of the mesa
surrounding the Band 1 etch. The metallic contacts are visible, as are
the indium bumps for hybridization.

Tennant, Thomas, Kozlowski, McLevige, Edwall, Zandian,
Spariosu, Hildebrandt, Gil, Ely, Muzilla, Stoltz, and Dinan
592
a
b
Fig. 3. D* histograms at (a) 120 K and nominal operating background and (b) peak D* showing background limited infrared photoconductor (BLIP)
operation of 40 mm pixel pitch array up to 120 K.
current, and minority carrier diffusion length. These
values complement the characteristics measured on
two-color FPAs.
Current-voltagecharacteristicsofvariousdiodesat
78KprovidedarkcurrentandR0Avalues.Eveninthe
“dark” of our standard measurement dewars, the
residual light leakage gave a background flux on the
order of 2 ¥ 1010 photons-cm–2s–1 which dominated the
device characteristics. Measurement equipment lim-
ited determination of R0A values to a lower bound
estimate of ~1 ¥ 108 Wcm2. Although some differences
between the SUMIT architecture and our baseline
DLPH single-color architecture are significant, these
SUMIT devices appear to be as good as any single
Fig. 4. Captured still image from two-color FPA. The image on the right
color devices. Comparison of the effective optical area
is from Band 2 and shows the twin engines of a small private aircraft
on a landing approach. The left image is from Band 1 and shows lower
MTF (a blurry image) due to an insufficient depth of the pixel isolation
etch. The frame rate was 110 Hz with f/2.3 optics and 2 ms integration
with the implant area gave a diffusion length in the
shorter-wavelength layer of 43 mm, and in the longer
wavelengthlayer~50mm.Theselongdiffusionlengths
time.
require a pixel isolation etch deep enough to reduce
spatial cross talk in Band 1. Without sufficient pixel
isolation, photo-generated carriers in one pixel could
easilydiffuseintoanadjacentpixelandbedetectedby
its diode only 40 mm away.
etching,buttoothinalayerwillreduceQE.Moreover,
since Band 2 is partially removed to provide access to
Band 1, the fill factor of Band 2 is slightly decreased.
The estimated fill factor of the Band 2 detector is
~87%. Together these effects will typically reduce the
effective QE (that is QE x fill factor) of the Band 2
detector. The Band 1 diodes show a typical single-
color spectral response (with a cutoff wavelength
The Band 1 quantum efficiency (QE) is 72%, near
the theoretical value of 76% (there is no antireflection
coating on the substrate). Slight over-etching of the
large opening in the Band 2 layer needed to fabricate
the large area diode used for the first color QE mea-
surement and imperfections (scattered light) in the
lensusedforthemeasurementmayhavecausedafew
percent underestimate of the test diode QE. The FPA
measurement at 3.4 mm, on the other hand, gave 79%,
essentially the theoretical maximum allowed by sur-
face reflection. This FPA value may have been a few
percent high due to narcissus from the 12% reflecting
filter located about 1.5 cm from the FPA during the
flood measurement. The QE measurement for Band 2
was not completed on the test diodes. However, Band
2 may have a smaller QE than Band 1. Two effects
could account for this difference. A thin, extended-
wavelength Band 2 absorber layer simplifies trench
l
co1 = 3.9 mm at 78 K). In the process evaluation chips
(PECs) Band 2 diodes exhibit response in the spectral
region of Band 1, due to a combination of long lateral
collection length and lifetime dominated by radiative
processes. The Band 1 layer absorbs the shorter-
wavelength light, but a fraction of the generated
carriers recombine and radiate into the Band 2 ab-
sorber (with cutoff wavelength lco2 = 5.9 mm at 78 K).
The large amount of lateral collection from the bulk of
the semiconductor in the vicinity of the diode can
make this spectral cross talk quite pronounced in
isolated diodes. The density of diodes in an array,
however, suppresses this effect, since carriers must

A Novel Simultaneous Unipolar Multispectral Integrated Technology
Approach for HgCdTe IR Detectors and Focal Plane Arrays
593
a
b
Fig. 5. QE map for another FPA with ~10 mm pixel isolation etch. (a) Band 1, (b) Band 2. Gray scale enhances appearance of non-uniformity, but
actual non uniformity (standard deviation/mean) is 8% in Band 1 and 5% in Band 2. Operability (response >50% of mean) is about 99% in each
band.
diffuse some significant distance to recombine and in
an array will much more likely be collected by a
nearby junction. Moreover, this spectral cross talk
will decrease at higher temperatures because the
holes diffuse more rapidly to the diode. The FPA
analysis of our 40 mm devices reveals that the effect is
less than 6% of the in-band response at 78 K and less
than 3% at 120 K. Note that this cross talk is asym-
metric, being almost non-existent in the longer-to-
shorter band direction.
The very high R0A values mean that the pixels have
low dark current and exhibit background limited
performance (BLIP) up to 120 K in both bands. Figure 3
shows the peak D* for the two bands at the nominal
operating background for the detectors at 120 K and
the temperature dependence of D* under two differ-
ent illumination conditions. This figure shows the
good operability and sensitivity of the array. The FPA
also exhibits low NEDT values: 9.3 mK for Band 1 and
13.3 mK for Band 2, similar to good quality single-
color FPAs. Finally, the broadband QE for Band 1 is
79% with only a 4% variation (sigma/mean) over the
pixels, and 75% for Band 2, with a 2% variation over
the pixels. The larger variation in Band 1 appears to
have been limited by the effect of cross talk on pixels
abutting a defective pixel. This was a result of the
trench etch being shallower than intended, as men-
tioned above. This causes a degraded modulation
transfer function and appears as a fuzzy image, as
shown in Fig. 4. Shorted pixels can act as current
sinks and thereby reduce the response of adjacent
pixels. This effect does not occur for the Band 2 pixels
because the isolation etch is deep enough to suffi-
ciently isolate Band 2 pixels. A deeper pixel-isolation
etch in a more recent array has given higher perfor-
mance in several respects. The Band 1 pixels have
a higher operability (~99% in each band—Fig. 5)
while both bands show reasonable non-uniformity
(8% SD/mean in Band 1 and 5% SD/mean in Band 2).
Thespatialcrosstalkappearstobereducedcompared
with the imaged FPA of Fig. 4. This latter conclusion
comes from comparing similar QE maps for the Band
1 diodes of the Fig. 4 array to that shown for the new
FPAs made from these detectors were 128 ¥ 128
devices with 40 mm pixel pitch. Their performance
confirms the test diode evaluation described above.
Fig. 6. Detectivity of a two-color MWIR-LWIR SUMIT FPA with 84
micron effective pixel pitch made using a 128 ¥ 128 single-color MUX
with a 60 mm pixel pitch.

Tennant, Thomas, Kozlowski, McLevige, Edwall, Zandian,
Spariosu, Hildebrandt, Gil, Ely, Muzilla, Stoltz, and Dinan
594
array in Fig. 5. The earlier array maps showed evi-
dence of spatial cross talk via a depressed response of
thenearestneighbordetectorstoashorteddiode. The
MTF of this Fig. 5 FPA should also be superior to that
of the imaged array.
CONCLUSION
We have demonstrated that the SUMIT architec-
ture is well-suited to threat warning applications. We
have mitigated the spatial cross talk seen in initial
FPAs by using a deeper pixel-isolation etch. This
architecturehasproducedtwo-colordevicesthatequal
or exceed the best we have seen in single-color devices
from any laboratory. Although the results presented
here are for MWIR-MWIR detectors, there is no limi-
tation in either the detector design or the ROIC
design to these wavelengths. The SUMIT architec-
ture should work equally well for long wavelength,
strategic applications. The challenge lies in extend-
ing the architecture to large format (small-pixel)
FPAs and simultaneous detection in more than two
spectral bands.
The spectral cross talk from the Band 1 to Band 2 is
5.6% at 78 K, decreasing to 2.4% at 120 K. This shows
that the large effect observed in isolated diodes or
diode groups does not apply to arrays, where the
lateral collection is limited by the isolation etch. We
note that at 120 K we appeared to be limited in both
bands by the photocurrent from the 120 K thermal
backgroundoftheclampingstructure.Thecoldshield
(at ~80 K) had much lower in-band emissivity. As a
result, the diodes on the two corners of the array
nearest the clamp showed higher dark current that
increased with FPA and clamp temperature. If this is
confirmed, thecentraldiodesnotascloselyexposedto
the radiation from the clamp appear to be operating
abovetheradiativeperformancelimit,asHumphreys
predicted.9
Recent devices have all been MWIR-MWIR. How-
ever, prior to our fabricating the “true” two-color
MUX for the FPAs discussed above, we used a single
color high capacity MUX to demonstrate two color
FPA performance with MWIR-LWIR (long-wave in-
frared). This device (made without the pixel isolation
etches) had 60 mm pitch for single color, giving an
84 mm two-color pixel pitch (obtained by rotating the
detector pixel by 45∞ and increasing the linear dimen-
sion by 21/2) and giving the FPA a 128 ¥ 128/2 format.
We had some difficulty with hybridization on this
earlydevice,withtheresultthatonlyabout70%ofthe
pixels were connected. As in the case of its later
descendants, this device gave inferred R0A values
(from averages of the connected pixels) comparable to
state-of-the-art single-color detectors at 4.7 mm and
8.7 mm, with FPA detectivity approaching BLIP to
over 100 K. Figure 6 shows the temperature depen-
dence of the detectivity for the operable elements of
this array.
ACKNOWLEDGEMENTS
The MUXes used in this work were developed with
efforts sponsored by the U.S. Navy (NRL contract
numbersN00014-94-C-2179andN00014-97-C-2068).
We gratefully acknowledge the support of Boeing for
Independent Research and Development funds.
REFERENCES
1. W.E. Tennant and W.V. McLevige, U.S. patent 6,034,407
(2000).
2. E.F. Schulte, U.S. patent 5,113,076 (1992).
3. K. Kosai and G.R. Chapman, U.S. patent 5,457,331 (1995).
4. J.M. Arias, Properties of Narrow Gap Cadmium-based Com-
pounds, ed. Peter Capper, EMIS Datareview Series No. 10
(London: Inspec 1994), pp. 30–35.
5. J.M. Arias, J.G. Pasko, M. Zandian, J. Bajaj, L.J. Kozlowski,
R.E. DeWames, and W.E. Tennant, Proc. SPIE Symp. on
ProducibilityofII-VIMaterialsandDevices,Vol.2228(1994).
6. J. Bajaj, J.M. Arias, M. Zandian, J.G. Pasko, L.J. Kozlowski,
R.E. DeWames, and W.E. Tennant, J. Electron. Mater. 24,
1067 (1995).
7. J.M. Arias, J.G. Pasko, M. Zandian, S.H. Shin, G.M. Will-
iams, L.O. Bubulac, R.E. DeWames, and W.E. Tennant,
Appl. Phys. Lett. 62, 976 (1993).
8. W.E. Tennant and C. Cabelli, Mater. Res. Soc. Symp. Proc.
494, 221 (1998), and references contained therein.
9. R.G. Humphreys, Infrared Phys. 26, 337 (1986).