Full text of DOI:10.1007/BF02665849

Composition Control of Long Wavelength MBE
Journal of ELECTRONIC MATERIALS, Vol. 30, No. 6, 2001
Special Issue P6a4pe3r
HgCdTe Using In-situ Spectroscopic Ellipsometry
Composition Control of Long Wavelength MBE
HgCdTe Using In-situ Spectroscopic Ellipsometry
DENNIS EDWALL,1,2 JAMIE PHILLIPS,1 DON LEE,1 and JOSE ARIAS1
1.—Rockwell Science Center, 1049 Camino Dos Rios, Thousand Oaks, CA 91360. 2.—e-mail:
ddedwall@rsc.rockwell.com
Improved composition control of Hg1–xCdxTe layers grown by molecular beam
epitaxy using in-situ spectroscopic ellipsometry is described. This has increased
our composition yields from < 40% to approximately 70% for a specification of x
to within 0.0015 of target composition. Knowledge of composition during growth
also enables corrections to effusion cell temperatures so that the in-depth
compositionprofilecanbecontrolled.Furtherimprovementswereobtainedafter
active composition control was implemented whereby the ellipsometer controls
the Te cell temperature to maintain the desired composition.
Key words: HgCdTe, spectroscopic ellipsometry, MBE
INTRODUCTION
EXPERIMENTAL
Accurate composition control is essential for long
wavelength Hg1–xCdxTe (x ~ 0.22, lco > 10 mm at 78 K)
where an accuracy of Dx £ 0.001 is desired in order to
achieve the desired cutoff wavelength for infrared
detection. For epitaxial growth by molecular beam
epitaxy (MBE), this requires knowledge of the CdTe
and Te source temperatures to within ±0.1∞C. But
this knowledge is lacking because the temperature
control thermocouples are not in direct contact with
the source materials, and thermal contact between
the materials and their crucibles constantly changes
due to depletion effects and temperature cycling.
Hence, in general, without in-situ real-time monitor-
ing, the layer composition is unknown. This causes
yield loss in obtaining the desired average composi-
tion, as well as lacking knowledge of the in-depth
composition uniformity and profile. Historically, the
inabilitytoobtainthedesiredlayercompositionisthe
highest yield loss factor in our MBE growth process.1
Hence, real-time monitoring of composition is es-
sential to improve producibility. Spectroscopic
ellipsometry (SE)2–7 has the required sensitivity to
measure x to high accuracy for HgCdTe. This paper
reportsourprogressusingSEinwhichweaddressthe
issues of how well it works, the impact on composition
yield, control of in-depth composition profile, and
implementation of real-time active control in which
the ellipsometer controls the Te cell temperature to
achieve a desired composition.
Layers were grown in a Riber 32P MBE system
installed in our laboratory in 1999. This system and
its operation were discussed in detail by Bajaj et al.8
Substrates were (211)B CdZnTe (3.5 ±1% Zn) with
sizes of 3¥4, 4¥4, or 3¥5 cm2. The substrates were
rotated during growth to improve uniformity, and SE
data were acquired continuously during rotation.
Layercompositionisdeterminedbythetemperatures
of the CdTe and Te effusion cells. The absorbing
layers of interest were typically 10–15 mm thick with
compositions in the range 0.205–0.230 depending on
application. Layer characterization techniques in-
cludedroom-temperatureFouriertransforminfrared
transmission (FTIR) to determine composition x and
layer thickness of the absorber layers. Secondary ion
mass spectroscopy (SIMS) was used to independently
measure in-depth composition profiles.9
SE measurements were performed using a multi-
spectral M-88 ellipsometer system from the J.A.
Woollam Co.3,10–12 Data were acquired continuously
throughout the growth run using 88 wavebands from
1.6 to 4.5 eV. Using an optical constant library built
with data from our HgCdTe layers, the data analysis
using the 88 separate wavelengths as well as data
from the real and imaginary parts of the dielectric
function ensured that a unique solution was found for
the composition to high accuracy.
RESULTS
As an example of how the layer composition can
change unpredictably during a run, Fig. 1 shows the
(Received November 21, 2000; accepted January 16, 2001)
643

644
Edwall, Phillips, Lee, and Arias
Fig.1.Depthcompositionprofilefortheabsorberlayerfromrun3–150.
During the time of this growth, all cell setpoint temperatures were
constant. The transient from time t = 40–60 m was caused by the
transition from a high x buffer layer (off scale).
Fig. 2. Depth composition profile for the absorber layer from run 3–99,
comparing measurements from in-situ SE and post-growth SIMS.
in-depth composition profile for a 3 h absorber layer
duringwhichtheCdTeandTecelltemperatureswere
constant. Off to the left of the graph (t < 40 m), a high-
x buffer layer was grown. While the composition
undershootaroundt=50mcouldhavebeencausedby
actual temperature overshoot of the Te cell or under-
shoot of the CdTe cell (but not observed from the
controlling thermocouples), the long, sustained de-
crease in x throughout most of the growth was caused
by drifts of the fluxes in either or both of the cells. We
have observed that the shapes of profiles obtained for
constant cell temperatures can change radically from
run to run, which is a major reason for the inability to
accurately predict the layer average composition.
To demonstrate that the composition profile of Fig.
1 is real, Fig. 2 shows the profiles of another layer
measured by SE and independently by SIMS. The
very good agreement between the profiles in terms of
features demonstrates the validity of SE.
Figure3showsacomparisonofrun-to-runxcontrol
with and without SE. With SE, the large outliers are
completely eliminated and the distribution is notice-
ably tighter. The standard deviation improved from
0.0081 to 0.0018, a factor of 4.5 improvement. A
discussion of the resulting improvements in cutoff
wavelength control and factors affecting it can be
found in Ref. 13.
Figure 4 shows the relationship between composi-
tions measured by in-situ SE and room temperature
transmission measurements for 50 layers. A rela-
tivelyconstantoffset(xSE–xFTIR)of–0.0025isobserved
(and accounted for during growth), with 90% of the
layers having an accuracy of Dx < ±0.002 from the
desired composition. xSE represents an average using
nine time slices equidistantly spaced from the begin-
ningtoendofthelayer.Foragivenlayer,theaccuracy
of this parameter is ~ ±0.0005, which depends on the
exact time slices selected. Consistency of this offset
parameter is now the key driver controlling the yield
oflayershavingthedesiredcomposition. Thisparam-
eter is plotted in Fig. 5 for 144 of the most recently
grown layers. A similar accuracy of ~0.0005 is esti-
Fig. 3. Comparison of x control for hundreds of x~0.2 layers grown
without SE compared with recent layers grown with SE. With SE, the
large outliers are completely eliminated and the distribution is notice-
ably tighter. The standard deviation improved by a factor of 4.5.
mated for the determination of xFTIR. However, due to
the strong sensitivity of geometrical factors on the SE
measurements and their analysis (e.g., substrate ro-
tation wobble, layer morphology, etc.), we believe x
control is primarily limited by the SE measurements
rather than how accurately x can be determined by
FTIR.
Implementation of Active Control
Figures 1 and 2 above show that the layer composi-
tion cannot be predicted merely by knowing the CdTe
and Te cell temperatures. Run-to-run variations in
in-depth composition complicate achieving the de-
sired average composition as measured by FTIR, and
are potentially detrimental to detector performance
due to reduced quantum efficiency. These factors
have made it desirable to implement active control
whereby the Te cell temperature (which controls x) is
controlled by the output signal of the ellipsometer.
Automated feedback control was achieved through

Composition Control of Long Wavelength MBE
HgCdTe Using In-situ Spectroscopic Ellipsometry
645
Fig.6.Depthcompositionprofilefortheabsorberlayerfromrun3–243,
comparing the linear composition profile with the Te cell temperature
thatwasrequiredtoachieveit. Thedeviationfromalineartemperature
profile approximately represents the degree of active control neces-
sary to achieve the linear composition profile.
Fig. 4. Relationship between compositions measured by in-situ SE
and room temperature transmission measurements for 50 layers. A
relatively constant offset (xSE–xFTIR) of –0.0025 is observed (and
accounted for during growth), with 90% of the layers having an
accuracy of Dx < ±0.002 from the desired composition. These layers
were grown without active control.
and it can even change during a run which then
causes a deviation from linearity in the composition
profile. Using active control, Fig. 6 shows the profile
for one layer with the corresponding Te cell tempera-
ture profile. The cell temperature is determined by a
linearly ramped setpoint plus a correction setpoint
term determined from the ellipsometer output suffi-
cient to maintain a linear composition profile. The
deviation from linearity in the temperature profile
approximately represents the amount of active con-
trol correction that was necessary to keep the compo-
sition profile linear. As Fig. 6 shows, the correction
necessary can be significant.
SUMMARY AND CONCLUSIONS
Accurate control of HgCdTe composition during
MBE growth has been a challenge, but thanks to the
combinedeffortsofmanyintheHgCdTeMBEandSE
communities, the goal is now within reach. In the
work reported here, using SE resulted in a standard
deviation improvement in x control from 0.0081 to
0.0018 for more than 200 layers grown using SE, a
factor of 4.5 improvement. Active control of Te cell
temperature by the ellipsometer was implemented to
furtherimprovecontrolregardlessofdriftsineffusion
cell characteristics. The reproducibility of the offset
parameter, (xSE–xFTIR), is now the key driver control-
ling the composition yield. Work is now underway to
understandthefactorsthatlimitthisreproducibility.
Fig. 5. Offset (xSE–xFTIR) for more than a hundred layers.
separate temperature control of the Te source using
real-time SE data and a control program written
usingLabVIEWsoftwarefromNationalInstruments.13
The user inputs the desired starting and ending
compositionandcorrespondinggrowthtime. Thepro-
gram then determines a target composition value
using a linear fit. A proportional feedback algorithm
makes adjustments during the growth run to the Te
effusion cell temperature in response to the deviation
of composition measured from SE with respect to the
target composition.
Consider the case where a linear composition pro-
file is desired. Without active control, using a linear
Te cell temperature ramp during growth will gener-
ally result in an approximately linear composition
profile. However, the slope of the profile varies from
run to run due to time-dependent cell characteristics,
ACKNOWLEDGEMENTS
This work was supported by Independent Research
and Development funding from the Boeing Corpora-
tion and by AFRL/MLPO Contract No. F33615-99-C-
5432(Wright-PattersonAFB,OH).Theauthorsthank
Blaine Johs and co-workers at the J.A. Woollam Co.
for their development of SE that made this work
possible and for their excellent technical support, as
well as L.A. Almeida at E-OIR Measurements for
many useful technical discussions.

646
Edwall, Phillips, Lee, and Arias
8. J. Bajaj, J.M. Arias, M. Zandian, D.D. Edwall, J.G. Pasko,
L.O. Bubulac, and L.J. Kozlowski, J. Electron. Mater. 25,
1394 (1996).
9. L.O. Bubulac, D.D. Edwall, J.T. Cheung, and C.R.
Viswanathan, J. Vac. Sci. Technol. B 10, 1633 (1992).
10. J.A. Woollam Co., 645 M St., Suite 102, Lincoln, NE 68508.
11. S. Dakshina Murthy, I.B. Bhat, B. Johs, S. Pittal, and P. He,
J. Electron. Mater. 24, 445 (1995).
12. P.G. Snyder, J.A. Woollam, S.A. Alterovitz, and B. Johs, J.
Appl. Phys. 68, 5925 (1990).
13. J. Phillips, D. Edwall, D. Lee, and J. Arias, to be published in
J. Vac. Sci. Technol. B.
REFERENCES
1. D.D. Edwall, M. Zandian, A.C. Chen, and J.M. Arias, J.
Electron. Mater. 26, 493 (1997).
2. B. Johs, C.M. Herzinger, J.H. Dinan, A. Cornfeld, and J.D.
Benson, Thin Sol. Films 313/314, 137 (1998).
3. L.A. Almeida, J.N. Johnson, J.D. Benson, J.H. Dinan, and B.
Johs, J. Electron. Mater. 27, 500 (1998).
4. G.L. Olson, J.A. Roth, P.D. Brewer, R.D. Rajavel, D.M.
Jamba, J.E. Jenson, and B. Johs, J. Electron. Mater. 28, 749
(1999).
5. F. Aqariden, W.M. Duncan, H.D. Shih, L.A. Almeida, and
M.J. Bevan, J. Electron. Mater. 28, 756 (1999).
6. L.A. Almeida and J.H. Dinan, J. Cryst. Growth 201/202, 22
(1999).
7. B. Johs, C. Herzinger, J.H. Dinan, A. Cornfeld, J.D. Bensen,
D. Doctor, G. Olson, I. Ferguson, M. Pelczynski, P. Chow,
C.H. Kuo, and S. Johnson, Thin Sol. Films 313/314, 490
(1998).