Electrical properties of in situ As doped Hg1À x Cdx Te epilayers
grown by molecular beam epitaxy
Y. Selamet,a) C. H. Grein, T. S. Lee, and S. Sivananthan
Microphysics Laboratory and Department of Physics, The University of Illinois at Chicago, Chicago,
Illinois 60607-7059
共Received 2 January 2001; accepted 2 April 2001兲
The electrical properties of extrinsic in situ doped mercury cadmium telluride 共Hg1⫺x Cdx Te兲
epilayers grown by molecular beam epitaxy on (211)B CdTe/Si and CdZnTe substrates are studied.
The doping is performed with an elemental arsenic source. HgCdTe epilayers with a CdTe mole
fraction between 0.23 and 0.36 are grown at substrate temperatures of 175–185 °C. The temperature
dependent Hall effect characteristics of the grown samples are measured by the van der Pauw
technique. A magnetic field of up to 0.8 T is used in these measurements. An analysis of the Hall
coefficient in the temperature range of 40–300 K with a fitting based on a two-band nonparabolic
Kane model, a fully ionized compensating donor concentration, and two independent discrete
acceptor levels is reported. In addition, the fitting results of a three-band modeling of Hall effect
results are compared to published data on p-type Hg1⫺x Cdx Te. Both as-grown and annealed samples
are used in this study. All of the as-grown samples showed n-type characteristics whereas annealed
samples showed p-type characteristics. The minority carrier lifetimes of arsenic doped epilayers
measured by a photoconductive decay method are presented. In this work, an AlGaAs laser of
wavelength 850 nm with a pulse length of 10–90 ns is used. The electron lifetimes obtained from
this study are compared to published minority electron lifetimes in p-type HgCdTe. Theoretical
electron lifetimes of p-type Hg1⫺x Cdx Te material are reported and a comparison to published
electron lifetimes is also given. © 2001 American Vacuum Society. 关DOI: 10.1116/1.1374628兴
I. INTRODUCTION
HgCdTe is one of the most important infrared materials.
Its band gap tailorability, which permits one to tune to atmospheric transmission windows, has attracted special attention. By changing the Cd mole fraction, x, the band gap of
HgCdTe grown by molecular beam epitaxy 共MBE兲 can be
changed from the short wavelength infrared 共SWIR兲 to the
very long wavelength infrared 共VLWIR兲 range from run to
run, or within a single run, with small calibrations. Many
state-of-the-art device structures have already been proposed,
grown, and operated1–5 using this property of MBE crystal
growth. The current technology also requires well-controlled
n- and p-type doping with very well defined junction formation. N-type doping of HgCdTe with indium is well understood and controllable down to concentrations of mid–low
1014 cm⫺3.6 P-type doping of HgCdTe is one of the few
completely unsolved areas. P-type doping with group-I elements gives very good Hall effect results with ⬃100% activation, but high diffusivity of these elements7,8 prevents
them from being used in well defined junctions, as is required for many devices. Group-V elements can be used to
dope HgCdTe p type if the doping species are made to occupy anion sites.9,10 Among those group-V elements, arsenic
is the most widely studied element for employment as a
p-type dopant. High structural and crystalline quality MBE
growth of HgCdTe is restricted to the Te-saturated side of
the existence region.11 Growth away from the Te-saturated
region results in strong twinning, which can be observed in
a兲
Electronic mail: [email protected]
1488
J. Vac. Sci. Technol. B 19„4…, JulÕAug 2001
reflection high-energy electron diffraction 共RHEED兲 patterns. Arsenic doping of HgCdTe under Te-saturated conditions naturally results in arsenic atoms occupying cation sites
and hence behaving as donors rather than the desired acceptor character. When an element As source is used for doping,
it is expected that the molecular flux from the arsenic cell be
in the form of As4 molecules. There is consensus that the As
flux is in the form of tetramers,12,13 however there is ongoing
debate regarding the form in which arsenic incorporates into
the material. Arsenic is also strongly self-compensating, and
requires postgrowth annealing. For As concentrations higher
than 2⫻1018 cm⫺3, the activation ratio rapidly falls off due
to the decreasing solubility of arsenic14 in HgCdTe.
II. EXPERIMENT
Hg1⫺x Cdx Te layers with x values in the midwavelength
infrared 共MWIR兲 共x⬃0.3) and LWIR regions 共x⬃0.2) were
grown with a Riber2300 MBE system. The substrates were
CdZnTe, with 4% Zn to provide good lattice matching, and
CdTe on Si. Both substrates were in (211)B orientation. The
substrates were cleaned in warm tricloroethelene 共TCE兲 for
10 min followed by acetone and methanol rinses. They were
also etched in a Br⫹methanol solution to provide a fresh
surface, followed by methanol and de-ionized water rinses.
The substrates were mounted onto a molybdenum substrate
holder with a graphite solution after a gallium⫹tin contact
共for thermocouple readings兲 was attached on the backside.
After loading into the growth chamber, the substrates were
heated to ⬃300 °C for thermal cleaning and oxide layer removal. All of the substrates studied here gave very clean
1071-1023Õ2001Õ19„4…Õ1488Õ4Õ$18.00
©2001 American Vacuum Society
1488
1489
Selamet et al.: Electrical properties of As doped HgCdT
1489
gas to reduce impurity concentrations, and sealed with a
H2⫹O2 torch. Annealing was performed in a two-zone oven
under isothermal or lower Hg temperature⫹isothermal combination conditions. The carrier lifetimes of the arsenic
doped samples were measured by a photoconductive decay
technique. An AlGaAs laser of 850 nm wavelength with a
pulse duration of 15–90 ns was used for these experiments.
III. RESULTS AND DISCUSSION
FIG. 1. Hall effect results of as-grown and annealed samples. The increase in
the carrier concentration after annealing is due to transfer of arsenic atoms
to anion sites.
streaky RHEED patterns after thermal cleaning and just before the initiation of HgCdTe growth. The substrates, after
thermal cleaning, were cooled down to the growth temperature of 175–185 °C. The growth of HgCdTe layers was done
under Te-saturated conditions and close monitoring of the
RHEED patterns indicated two-dimensional 共2D兲 growth.
All layers studied here were grown at the rate of approximately 2 ␮m/h and a thickness of 4–5 ␮m and doped by
either the conventional method or planar doping. Details of
the growth techniques for planar doping have been reported
previously.15 HgCdTe growth was terminated by growing a
protective CdTe layer of 0.3–0.4 ␮m thickness. The cut-off
wavelengths were measured at room temperature using a
Fourier transform infrared 共FTIR兲 spectrometer to calculate
the x values and the thickness of the HgCdTe layers. X-ray
rocking curves were also measured to check the crystalline
quality. Hall effect measurements on both as-grown and annealed samples were performed by the van der Pauw technique. A temperature range of 40–300 K and magnetic field
strengths of up to 0.8 T were used. Annealing temperatures
were in the 250–450 °C range and were performed in a
closed-tube annealing system. A quartz tube with two partitions 共one for the sample and the other for a Hg drop兲 was
evacuated and purged several times with high purity argon
The carrier concentration versus 1000/T data points obtained from Hall effect measurements are values averaged
over the layer grown. There are many electrically active defects, including vacancies, interstitials and impurity species,
that may give p- or n-type characteristics. Therefore, direct
extrapolation of the data points obtained from the Hall effect
measurements to find the ionization energies of impurity levels usually will not give accurate results. For this reason, a
fitting program using a Kane model with nonparabolic bands
was developed at the Microphysics Laboratory 共MPL兲 at the
University of Illinois. The program assumes two acceptor
levels and a single donor level for fitting. The fitting was
done to the Hall coefficient, given by16
共 p⫺b c n 兲
R H⫽
,
q 共 p⫹bn 兲 2
共1兲
where b and c are parameters in the fitting program representing the donor to acceptor mobility ratio and the ratio
exponent, respectively. CdTe/Si substrates have advantages
over CdZnTe substrates in terms of cost and physical durability. From fittings of Hall effect results we found no clear
distinction between CdZnTe substrates and the CdTe/Si substrate of this study. This could be good news for CdTe/Si
substrates, but it is still a subject of further study. The Hall
effect results of as-grown arsenic doped samples gave n-type
characteristics. This may be a consequence of the Tesaturated growth conditions that force some arsenic atoms to
occupy cation sites where they act as donors, or may be due
to background impurities. After annealing, the samples were
converted to p type, as shown in Fig. 1. The conversion
process may be complex since arsenic does not necessarily
incorporate into HgCdTe as single arsenic atoms as was previously assumed. Theoretical work at MPL has concluded
TABLE I. Fitting of Hall effect results of arsenic doped samples with various Cd mole fractions. After annealing,
a very low V Hg concentration is observed due to isothermal annealing filling those vacancies.
Sample
x
共%兲
N d 共cm⫺3)
⫻1016
donor conc.
N a1 共cm⫺3)
⫻1016
As conc.
N a2 共cm⫺3兲
⫻1016
V Hg conc.
E a1 共meV兲
As ionization
energy
E a2 共meV兲
V Hg ionization
energy
1
2
3
4
5
6
7
8
36.5
35.0
29.9
29.1
28.9
26.9
25.4
23.3
1.5
0.82
0.001
1.02
5.9
6.73
0.033
7.5
3.3
27.0
27.02
4.22
11.0
15.01
4.3
21.0
0.011
0.0026
0.001
0.01
0.000 98
0.48
0.0002
0.001
10.2
5.6
4.2
3.7
7.7
3.5
6.0
3.2
107.6
108.8
101.0
84.0
152.1
32.6
99.9
59.6
JVST B - Microelectronics and Nanometer Structures
1490
Selamet et al.: Electrical properties of As doped HgCdT
FIG. 2. HgCdTe band-edge diagram for x values between 0.2 and 0.4. Levels
due to arsenic impurities are very shallow with ionization energies less than
10 meV. Levels due to V Hg are close to midgap and principally located
60–110 meV from the valance band. The dotted line is the calculated curve
for an arsenic concentration of 1⫻1017 cm⫺3.
that there is evidence that arsenic may be incorporated as
tetramers (As4) or As4 clusters for As4 sources. Calculations
using the general principles of quasithermodynamic theory,
employing the law of mass action to calculate the relations
between the arsenic concentration and the arsenic and Hg
partial pressures, led to this conclusion.17,18 The conversion
to p-type behavior may occur by the breakup of neutral tetramers, resulting in arsenic atoms occupying Te sites.12,18 As
seen from Fig. 1, the increase in the doping level ( 兩 N d
⫺N a 兩 ) would then be due to the transfer of some arsenic
atoms to Te sites, upon which they start to act as acceptors.
The decrease in activation, as measured by secondary ion
mass spectroscopy 共SIMS兲 and the Hall effect, when the arsenic concentrations are in excess of 2⫻1018 cm⫺3, may be
due to the lack of activation of neutral As4 clusters. SIMS
results have already been published and compared to the
doping levels measured by the Hall effect.15 The lower mobility of as-grown layers, showing n-type characteristics,
compared to indium doped n-type layers with the same doping levels is attributed to the presence of additional strong
scattering centers, such as As4 and its clusters. Table I shows
a summary of the fitted Hall effect results. The shallow and
deep levels found by fitting to Hall effect data are plotted in
a HgCdTe band-edge diagram for x values between 0.2 and
0.4 in Fig. 2. The change in the band gap with x value is
calculated by the well accepted formula,19
1490
FIG. 3. Comparison of ionization energies due to arsenic impurities to published results. Lines are the calculated fit using Eq. 共3兲 for specific arsenic
concentrations.
these fittings are in good agreement with other published
results. A fit for the arsenic acceptor level in published
p-type doped HgCdTe data20 is given by
E a ⫽42x⫹1.36⫺1.4⫻10⫺5 N 1/3
a .
共3兲
The expected values of the ionization energy obtained using
this formula are very close to our findings. Figure 3 shows a
comparison between the fit function and our experimental
results for specific arsenic concentrations.
The minority carrier lifetime of the arsenic doped layers
was measured and compared to theoretical values. The dominant Auger mechanism in the p-type layers is the Auger-7
process.21 Recently, some corrections to the old Auger
mechanism calculations of Casselman21 were made.22 Figure
4 shows both theoretical and experimental 80 K minority
E g 共 x,T 兲 ⫽⫺0.302⫹1.93x⫺0.81x 2 ⫹0.832x 3
⫹0.000 535T 共 1⫺2x 兲 ,
共2兲
at 80 K. In our experimental study, shallow level energies
due to arsenic impurities were lower than 10 meV for x
⬍0.36. Deep levels, which were assumed to be V Hg , were
close to the intrinsic level, with most of the points in 60–110
meV range. Ionization energies for arsenic obtained through
J. Vac. Sci. Technol. B, Vol. 19, No. 4, JulÕAug 2001
FIG. 4. Minority carrier lifetimes of arsenic doped HgCdTe layers at 80 T.
Lines are the calculated values for radiative and Auger lifetimes. For Auger
lifetimes, both the old and new theoretical versions are included for comparison.
1491
Selamet et al.: Electrical properties of As doped HgCdT
1491
obtained, from Hall effect results, very shallow levels for
arsenic with ionization energies lower than 10 meV for
MWIR and LWIR materials. Deep levels are very close to
the midgap, with ionization energies principally in the 60–
110 meV range. The concentrations of these midgap levels
must be reduced in order to improve the device characteristics of the HgCdTe layers. Carrier lifetimes of arsenic doped
layers were mostly radiatively limited. Lifetime comparisons
to sample grown with other techniques are also given. The
results obtained are comparable to the best published lifetimes.
T. J. Lyon et al., J. Cryst. Growth 201Õ202, 980 共1999兲.
T. J. Lyon et al., J. Electron. Mater. 27, 550 共1998兲.
3
R. D. Rajavel et al., J. Electron. Mater. 27, 747 共1998兲.
4
P. Ferret, J. P. Zanutta, R. Hanelin, S. Cremer, A. Million, M. Wolny, and
G. Destefanis, J. Electron. Mater. 29, 641 共2000兲.
5
T. J. Lyon et al., J. Vac. Sci. Technol. B 16, 1321 共1998兲.
6
G. Brill 共private communication兲.
7
M. Boukerche, P. S. Wijewarnasuriya, S. Sivananthan, I. K. Sou, Y. J.
Kim, K. K. Mahvadi, and J. P. Faurie, J. Vac. Sci. Technol. A 6, 2830
共1988兲.
8
S. Sivananthan, P. S. Wijewarnasuriya, and J. P. Faurie, Proc. SPIE 2554,
55 共1995兲, and references therein.
9
H. R. Vydyanath, J. A. Ellsworth, and C. M. Devaney, J. Electron. Mater.
16, 13 共1987兲.
10
M. A. Berding and A. Sher, Appl. Phys. Lett. 74, 685 共1999兲.
11
H. R. Vydyanath and C. H. Hiner, J. Appl. Phys. 65, 3080 共1989兲.
12
C. H. Grein, J. W. Garland, S. Sivananthan, P. S. Wijewarnasuriya, F.
Aqariden, and M. Fuchs, J. Electron. Mater. 28, 789 共1999兲.
13
M. A. Berding and A. Sher, J. Electron. Mater. 28, 799 共1999兲.
14
P. S. Wijewarnasuriya and S. Sivananthan, Appl. Phys. Lett. 72, 1694
共1998兲.
15
F. Aqariden, P. S. Wijewarnasuriya, and S. Sivananthan, J. Vac. Sci.
Technol. B 16, 1309 共1998兲.
16
R. A. Smith, Semiconductors 共Cambridge University Press, Cambridge,
1959兲.
17
J. W. Garland, C. H. Grein, B. Yang, P. S. Wijewarnasuriya, F. Aqariden,
and S. Sivananthan, Appl. Phys. Lett. 74, 1975 共1999兲.
18
T. S. Lee, J. W. Garland, C. H. Grein, M. Sumstine, A. Jandeska, Y.
Selamet, and S. Sivananthan, J. Electron. Mater. 29, 869 共2000兲.
19
G. L. Hansen, J. L. Schmit, and T. N. Casselman, J. Appl. Phys. 53, 7099
共1982兲.
20
P. Capper, Properties of Narrow Gap Cadmium-Based Compounds
共INSPEC, London, 1994兲, p. 250.
21
T. N. Casselman, J. Appl. Phys. 52, 848 共1981兲.
22
S. Krishnamurty and T. N. Casselman, J. Electron. Mater. 29, 828 共2000兲.
23
V. C. Lopes, A. J. Syllaios, and M. C. Chen, Semicond. Sci. Technol. 8,
824 共1993兲.
1
2
FIG. 5. Lifetime comparison of MBE grown HgCdTe:As layers at MPL with
those of other growth techniques. The MBE grown sample lifetimes are
comparable to or longer than the lifetimes of samples grown with other
techniques.
carrier lifetimes versus arsenic concentration. For comparison, both the old calculations of the Auger lifetimes and new
corrections to it are plotted. The new Auger lifetime curve
for x⫽0.305 gives Auger lifetimes of an order in magnitude
higher than the old calculations. As can be seen from Fig. 4,
the lifetimes were mostly radiatively limited. For the lifetimes that are higher than theoretical results, photon recycling or persistent photoconductivity may be responsible. We
also compare our MBE grown sample lifetimes to those
grown by other epitaxial techniques23 in Fig. 5. Our results
are comparable to the best published lifetimes.
IV. CONCLUSIONS
We have studied Hall effect and minority carrier lifetimes
of arsenic doped HgCdTe epilayers grown by MBE. Asgrown arsenic doped layers grown under Te-saturated conditions were n type as expected. After annealing under Hg
pressure we achieved conversion from n type to p type. We
JVST B - Microelectronics and Nanometer Structures