INSTITUTE OF PHYSICS PUBLISHING SEMICONDUCTOR SCIENCE AND TECHNOLOGY

Semicond. Sci. Technol. 17 (2002) 1282–1287 PII: S0268-1242(02)36427-7

Defect-induced optical transitions in CdTe

and Cd0.96Zn0.04Te
R Grill, J Franc, P Hlı́dek, I Turkevych, E Belas and P Höschl
Institute of Physics, Charles University, Ke Karlovu 5, Prague 2, CZ-121 16, Czech Republic
E-mail: grill@karlov.mff.cuni.cz

Received 3 May 2002, in final form 2 October 2002

Published 13 November 2002
Online at stacks.iop.org/SST/17/1282

Abstract
Within a model of the compensated shallow acceptor classical impurity
band, we study the optical absorption of CdTe and Cd1−x Znx Te, at energies
of 20–70 meV below the energy gap connected with acceptor–conduction
band transitions. The shape of the absorption edge is fitted by Monte Carlo
numerical simulations and we estimate the total density of charged
impurities, which is hardly detectable by other experimental techniques.
The activation energy ≈59 meV and density ∼1014 cm−3 of the dominant
shallow acceptor level are reported.

1. Introduction by [6] using phonon-assisted direct excitonic absorption. The

The wide bandgap binary semiconductor CdTe is used as a
source material for the preparation of detectors of gamma and
X- radiation, as a substrate for fabricating detectors of infrared
radiation on the basis of (CdHg)Te and also as a material for
electro-optical modulation of signals.
In order to optimize the function of the above-mentioned
elements, it is necessary to improve both structural and
material parameters of the single crystalline CdTe. According
to the state-of-the-art technology for growing CdTe single
crystals, these demands lead to minimization of the influence
of point defects, which can affect, for example, the mobility
and lifetime of carriers. While at high temperatures native
point defects play a dominant role in undoped crystals, room-
temperature and low-temperature optical and galvanomagnetic
properties are strongly influenced by residual impurities.
Photoluminescence (PL) is a favoured method used to study
defect structure. A systematic effort has been devoted
to find the binding energy of shallow acceptors by low-
temperature PL, PL excitation spectroscopy and magneto-
optical analysis [1–3]. Doping experiments performed on
high-purity materials have shown that all studied acceptors
are related to substitutional impurities [1]. Although there
are a large amount of papers devoted to PL on CdTe
and related compounds, optical absorption studies are rare.
Infrared absorption (400–2000 cm−1 ) in Cu-doped CdTe has
been analysed by [4], while combined PL and absorption
measurements of deep bands related to Te vacancies and their
complexes have been performed by [5]. Irregularities in the
structure of the absorption edge in CdTe have been explained
shape of the absorption spectra as well as the interpretation,
which agrees qualitatively with our paper, has been reported
for CdTe by [7] and for ZnTe by [8]. In this paper, we evaluate
quantitatively the absorption below the absorption edge of
CdTe and Cd1−x Znx Te (x = 0.04). We use a model of the
totally charged classical acceptor impurity band, where the
Coulomb interaction is screened by mobile charges located in
a partially filled level.
2. Experimental details
CdTe and Cd1−x Znx Te (with Zn content ∼4%) single crystals
have been fabricated in our laboratory by the vertical gradient
freeze method [9]. We used 6N purity starting elements in
growth runs. The Cd pressure was controlled during the
growth by adding the corresponding amount of Cd in the
ampoule. Crystals were very good quality single crystals
filling about 90% of the crucible volume without macroscopic
defects (grain boundaries, twins), p-type with the room-
temperature concentration ∼1013 cm−3 , corresponding to the
Fermi level position ∼0.4 eV above the valence band. No
intentional doping was applied. For our PL and optical
absorption studies, samples with thicknesses of 1.5–3 mm
were prepared by cutting, polishing and etching in a Br–
methanol solution. The PL measurements were performed
using a He–Ne laser (λ = 632.8 nm) with power of 8 mW
and an average power flux of 4 W cm−2 . Transmittance
spectra (with a resolution of 0.12 meV) were measured by
a Fourier transform spectrometer IFS66/s (Bruker) in a He
flow cryostat VSK 4-300 (Leybold). The discussed weak

0268-1242/02/121282+06$30.00 © 2002 IOP Publishing Ltd Printed in the UK 1282

 Defect-induced optical transitions in CdTe and Cd0.96Zn0.04Te

4 10 0
(A,X)

(D,X)
3 (A,X)-LO 10 -1

Luminescence (a.u.)
Absorption (cm-1)
(A,X)-2LO; (e,A) X
X-LO

2 DAP 10 -2
X-2LO

1 10 -3

0 10 -4
1.54 1.56 1.58 1.60
Energy(eV)

Figure 1. Absorption and PL spectra of CdTe at 4.8 K. The theoretical value of α deviates from the experimental value at 1.565 eV due to
the complicated course of the absorption edge, which is not included in the model.

absorption structure has been observed on samples of the best mechanism in item (3) can be expected to describe the observed
quality, and a detailed quantitative analysis of two specimen effects.
is presented in this paper. In the case of impure samples, the The absorption constant due to the impurity–band
structure is less pronounced or is hidden in the broadening of transition is calculated in a standard way according to
the fundamental band edge. We analyse the low-temperature [10, 11] in the form
data T < 20 K, where the defect-induced transitions are very
2 5/2 
16e2 P 2 g h̄ h̄ω − Eg + EI
visible. The broadening of the fundamental absorption at α1 (h̄ω) = ∗  2 4 ,
higher temperature results in the attenuation of the effect. 3cε0 nh̄ ω 2m aB
2 2
h̄
+ h̄ω − E + E
2m∗ a 2 g I
B

3. Theory
The structure of the absorption edge below the principal band-
edge energy Eg is generally determined by:
(1) the exciton absorption forming a free exciton with binding
energy in CdTe of 10.6 meV, and the donor and acceptor
bound excitons near 13 and 17 meV, respectively;
(2) the broadened band edge due to the lattice disorder, which
results in the exponential tail in the density of states;
(3) the absorption connected with the electron transfer
into/from the localized states in the gap initiated by lattice
defects;
(4) the phonon-assisted processes at low temperature
characterized by the emission of optical phonon at the
energy of 21.2 meV in CdTe, which can accompany all
the above transitions. The effect results in the repeated
characteristic structure in the spectra.
The purpose of this paper is to explain the absorption 20–
70 meV below Eg at low temperature (4–20 K), where
remarkable details have been found (see figures 1 and 2). We
can eliminate the excitonic effects, which are not observed at
such energies. The exponential broadening of the absorption
edge forms the background absorption in this region. The
(1)
where P = 8.5 × 10−8 eV cm is the momentum matrix
element [12], n is the refractive index, and e, c, ε0 and h̄
are elemental constants. We show in the discussion that
the absorption is connected with the acceptor–conduction
band (CB) transitions, where acceptors are charged due to
compensating donors. In this case, the acceptor degeneracy
g = 4. EI (≡EA ) and aB are the acceptor binding energy
and Bohr radius in the effective mass approximation and
m∗ = 0.096m0 is the CB effective mass.
At low temperatures, when the free particle density can
be neglected, the potential at the localized states is influenced
by the Coulomb interaction with charge on other point defects.
Consequently, EA fluctuates around its ideal single-acceptor
value and the impurity band (IB) with the density of states
G(E) is formed. The energy deviation of the respective
acceptor is obtained in the form
e Zs
δEA (r) = − s , (2)
4πε0 εr |r − rs |
where Zs is the charge of the respective site. The total
absorption constant α is obtained by integrating
 ∞
α= α1 (h̄ω + X − EA )G(X) dX, (3)
Eg −h̄ω

characteristic wavy shape is not, however, explained by this where G(X) influences the start of the absorption, which is
mechanism. Based on given remarks, we suggest that the smoothed compared to the sharp increase in the case of the

1283
R Grill et al
4 10 1
(A,X)-2LO; (e,A)
3
Absorption (cm-1)
DAP
2 10 -2
1.54 1.56
Energy (eV)
Figure 2. Absorption and PL spectra of (CdZn)Te at 4.8 K. The theoretical value of α is not resolved from experiment in this scale. The
enhanced experiment–theory deviation 10α is plotted at the bottom of the figure.
isolated acceptor level. The fitting of α to the experimental
data allows us to estimate the shallow defect parameters.
The magnitude of α determines the defect density and the
start of the absorption gives the position of the respective
activation level. The shape of the absorption edge is used for
the examination of the relevant IB. The least-squares method
is used directly at the fit of α without it being necessary to
manipulate with experimental data so that no information is
lost at the numerical treatment.
The impurity levels in CdTe are of very complicated
structure forming both shallow and deep donor and acceptor
states. The exact distribution of the levels is not known in
our samples. Because of this, we use a minimal model here
assuming a set of i shallow acceptors with density NAi , and
one compensating deep donor level placed below the middle
of the gap. The energies of deep donor levels, which could
fit in our model, have been observed in several elements: Ti,
EDD = Ec − (0.91 − 1.05) eV; Ge, EDD is in the mid-gap;
Pb, EDD = Ec − 1.3 eV [13].
The deep donor density NDD is set above the total acceptor
density NDD > i NAi , which results in the pinning of the
Fermi energy near the deep donor energy. In these conditions,
all acceptors and some of the donors are ionized to fulfil
the neutrality condition p + NDD +
= i NAi , where p is the
hole density. All impurities are distributed randomly without
correlation to other point defects for simplicity. As is apparent
from the low intensity of PL, which corresponds to the donor–
acceptor pairs (DAP) and the donor to bound exciton (D, X)
band, the density of shallow donors ND is low and is neglected
in our model. In figure 3 we show the distribution of the
impurity levels and the relevant optical transitions expected in
the model.
Because only some of the deep donors are charged, the
charge on the donors (+) can be distributed in many different
configurations, each configuration being characterized by the
total energy given by the Coulomb interaction between all
charged donors and charged acceptors. The equilibrium
1284
(A,X)
(A,X)-LO
10 0
Luminescence (a.u.)
(D,X)
X 10 -1
X-LO
10 -3
10 -4
1.58 1.60 1.62
Conduction band
LO
ND
or
N DD
N A1
N A2
Valence band
Figure 3. The energy level scheme. The arrows show the relevant
optical transitions. Dashed lines outline two possibilities to explain
the absorption if several structures in α are observed; see figure 2.
charge distribution at finite temperature is then obtained by
Monte Carlo simulations following the procedures given in
[14–16] and averaging over many random configurations. The
simulations produce the density of states G(E) of the acceptor
impurity band.
The final states in the CB are also disturbed by potential
fluctuations. Their description is much more complicated,
which means that it is necessary to solve the problem of
localization in disordered systems. The solution of this task is
out of the scope of this paper. We estimate such an effect here
using the well-known relation between binding, kinetic and
potential energy, Eb , T  and V , respectively, which in the
hydrogen-like states reads Eb = T  = −V /2. Relying on
this, we assume that the potential fluctuations of the CB edge
manifest by one half in the transition energy and decrease
by two times the effective width of G(E) in comparison to
equation (2). In contrast, the opposite action to G can be
observed due to the unknown charged defects, both donors
and acceptors, which increase the potential fluctuations and
result in the increase of the width of G. At first, we neglect
such effects in our calculations, assuming a flat CB, and we
only include the acceptor levels that are directly observed in
the absorption. The corrections are subsequently discussed to
become the more realistic results.

4. Results and discussion
The first task to be solved is the decision about the type of point
defects that are responsible for the sub-gap transitions. As seen
in figures 1 and 2, the absorption can be well approximated by
the square root of energy in agreement with the numerator in
equation (1). The denominator must have only weak influence.
Such a demand is easily fulfilled in the case of acceptors for

which h̄2 2m∗ aB2 ∼ 0.3 eV for a hydrogen-like acceptor
is significantly greater than the energy interval in which we
are interested. The opposite situation is met in the case of
transitions from the valence band to donors, for which the
factor above is significantly suppressed and α would have
to pass maximum within the monitored interval. Another
argument can be based on experimental findings [1, 17], which
report acceptor levels close to our observations. No donor level
was reported in this region.
We present the results of the fit of α on two p-type
samples at 4.2 K, and we compare these with the luminescence
measurements. The measurements and fit of α on the CdTe
sample are plotted in figure 1 (Eg = 1.605 eV). The typical
characteristic of the absorption edge is evident at 1.55 eV
and the fit produces EA1 = 59.0 meV, which agrees well
with ENa = 58.7 meV [1], ELi = 58.0 meV [1], EAs =
58 ± 2 meV [17], and ESb = 61 ± 2 meV [17]. The
interpretation is supported in the luminescence by the position
of the electron–acceptor recombination maximum (e,A) and
the donor–acceptor peak (DAP), which is shifted to lower
energy by the donor binding energy. The acceptor density of
this level NA1 = 1.4 × 1014 cm−3 was calculated according
to equations (1) and (3), where the Bohr radius aB1 = 10.5 Å
was used [17]. Another possible acceptor level corresponding
to the absorption edge near 1.57 eV is not well pronounced
and cannot be fit within our model. Its position is about
20 meV above EA1 , which supports an alternative
interpretation of this part of the spectrum as a result of the
absorption from level EA1 accompanied by an emission of
optical phonons, discussed later in this paper.
The very good agreement of the theoretical fit with
the experimental absorption, shown together with the PL in
figure 2, was found for the Cd0.96Zn0.04Te sample, for which
Eg = 1.625 eV was established. We see the characteristic
advancement of α near 1.57 eV, which fits very well within the
presented model. The position of the level EA1 = 58.6 meV
is again in good relation to the DAP. The acceptor density
NA1 = 2.8 × 1014 cm−3 . A similar motif is repeated once
more at higher energy 1.59 eV. We have tried to describe this
part by a new acceptor level with EA2 = 38.2 meV. Such a
level has not been reported in the literature1 and its Bohr radius
aB2 is not known.
√ We connected aB2 with aB1 by the relation
aB2 = aB1 EA1 /EA2 , which yields aB2 = 13.0 Å and the
resulting acceptor density NA2 = 3.2 × 1014 cm−3 .
Some characteristic PL energies and their standard
assignments [19] are summarized in table 1. Here, X
denotes free exciton recombination (more precisely, the PL
minimum between upper and lower branches of the excitonic
polariton [20]); (A,X) represents the PL maximum due to the
recombination of excitons bound to neutral acceptors; X-LO is
1 An acceptor level E
A = 35 meV in CdTe(P) [18] was not confirmed;
EP = 68.2 meV [1].
Defect-induced optical transitions in CdTe and Cd0.96Zn0.04Te
Table 1. Characteristic energies (eV) and assignment of transitions
observed in the PL.
Assignment CdTe (CdZn)Te
X 1.5954 1.6145
(A,X) 1.5892 1.608
X-LO 1.5747 1.5938
(A,X)-LO 1.5683 1.5866
(A,X)-2LO; (e,A) 1.548 1.565
DAP 1.538 1.557
Eg 1.606 1.625
Eg -(e,A) 0.058 0.060
EA1 (fit) 0.0590 0.0586
the PL maximum connected with free exciton recombination
with one optical phonon emission (optical phonon replica of
X); (A, X)-LO is the optical phonon replica of (A,X); (e,A)
corresponds to the recombination of the conduction electron
with the hole on the shallow acceptor; the DAP points to
shallow-donor–shallow-acceptor recombination. Eg = X +
10.6 meV is the energy gap determined from free exciton
luminescence. Then, Eg − (e,A) gives the energy of the
acceptor level estimated from the luminescence, while EA1 is
this parameter obtained from the fit of α. The well-resolved
excitonic spectra show that the samples are of good quality.
Unfortunately, at least three components are superimposed in
the range of the (e,A) band, namely (e,A), (A,X)-2LO and
the DAP. So, a precise determination of the acceptor energy
is difficult. The bound exciton luminescence (A,X) with the
maximum at 1.5892 eV coincides with the value given for
Na-doped CdTe by [21]. Sodium is one of the most common
impurities in undoped CdTe and we consider that, in our case
also, Na creates the observed acceptor level EA1 .
A detailed analysis of the results of the Monte Carlo
simulations of the (CdZn)Te sample is shown in figure 4,
where the theoretical α, background absorption and the density
of states G of the impurity bands are presented. We fit the
background α by linear and exponential functions, which fit
properly the peripheral parts of the investigated interval. The
course of G for both levels is the same due to the same character
of Coulomb interaction in our classical approach with the
linewidth of FWHM = 5.5 meV. The best fit was obtained
with the compensating donor density ND = 9 × 1014 cm−3 ,
whose variations influence slightly the shape of G.
An excellent fit was obtained within our very simple
model of two shallow acceptors and one compensating donor.
If the additional charges are included in the model, the IB
broadens and the start of the absorption is smoothed. The
influence of charged levels to α is demonstrated in figure 5,
where new fully charged donor and acceptor levels with the
same density are supplemented in the model calculations
and the compensation of the sample is increased. The
finding implies that the sample is pure without a substantial
concentration of other invisible charged defects. In impure
samples with a defect density above 1016 cm−3 , the square-
root-like shape of α is smoothed and merges with the
fundamental absorption edge. The reported procedure is thus
hardly applicable to an absorption study of shallow high-
density states. The situation could be complicated by the
warping of the conduction band discussed in the theory. If G
is narrowed twice, the best fit is obtained with the invisible
1285
R Grill et al

4 10

Density of states (1016 cm -3eV-1)

3 1
7

Absorption (cm-1) 6
2
2 5

4
3

3
1
2

0 0
1.54 1.56 1.58 1.60
Energy(eV)

Figure 4. Theoretical absorption of the (CdZn)Te sample (full line, 1), background absorption (dashed line, 2) and the absorption connected
with acceptor–conduction band transition (dash-dotted line, 3). Line 1 is the sum of lines 2 and 3. The peaks show the impurity band density
of states calculated by Monte Carlo simulations on a cubic cell containing 4000 acceptors.

Experiment
Fit
2.0 5.10 14
1.10 15
Absorption (cm-1)

2.10 15
5.10 15

1.8

1.6

1.4
1.55 1.56 1.57 1.58
Energy(eV)

Figure 5. The influence of the background doping on the shape of the absorption. The legend indicates the concentration of the additional
donor ND and acceptor NA charged levels (ND = NA ).

charge density ∼4 × 1015 cm−3 . A careful study of the CB
edge in disordered systems is thus desirable.
The interpretation of the second level EA2 is, however,
not clear. The corresponding DAP is not observed in
the PL and the activation energy difference EA1 − EA2 =
20.4 meV is close to the LO phonon energy 21.2 meV.
This finding points to a possibility of interpreting this effect
as a transition from the deeper acceptor EA1 to the CB
accompanied by an emission of an optical phonon (phonon-
assisted absorption, PAA); see figure 3. In our samples, the
differentiation of both mechanisms is difficult. Contrary to
the phonon-assisted band-to-band absorption in indirect gap
semiconductors, where α ∝ (h̄ω − Eg )2 , the PAA connected
with the transition from the acceptor localized state to the
CB is characterized by the dependence on h̄ω similar to
equation (1). The PAA with phonon emission also does
not manifest distinct thermal evolution at low T. The only
argument against PAA is the magnitude of the relevant α.
Though all phonon-assisted processes in PL are damped
(figures 1 and 2), the magnitude of PAA exceeds about 1.6
times the basic acceptor-to-CB profile. We conclude that a
final decision about the nature of this transition cannot be
made based on the available experimental data. Evidently,
another plausible explanation could be accepted connecting
the effect with the complicated structure of the band edge,
which is responsible for the strong increase of α above
1.60 eV.
Comparing the shape of the fundamental absorption edge
in both samples, we observe the start of a strong increase of
α at similar energies relative to Eg . In CdTe, the fundamental
absorption starts at about 20 meV lower energy relative to Eg

1286

and corrupts partly the shape of α in the region of defect-
induced absorption. A seeming contradiction between the
FWHM of G, which implies a higher purity of the CdTe
sample, and the opposite result obtained from the shape of the
absorption edge could be explained by a higher concentration
of neutral defects in the CdTe sample. Neutral defects do not
manifest in the Coulomb broadening of G; they can, however,
influence the band edge.
It is a well-known fact that deep levels are hardly
characterized electrically and an alternative model with a deep
acceptor can also be valid. In such a case, the shallow and
deep acceptors are compensated by shallow donors located
near the conduction band. The model with a deep acceptor
level would have needed to describe the shallow donors using
an additional parameter, the shallow donor density ND , which
is unknown in our samples. We roughly estimate ND to be
1014 –1015 cm−3 . This is the reason why we use the model
with a deep donor in this paper. The presented theoretical
findings concerning the properties of shallow acceptor levels
(ionization energy and density) are not influenced significantly
by the selected model.
5. Conclusion
The measurement of the absorption constant below the gap
energy was used to obtain the density and activation energy of
shallow acceptors in CdTe and Cd0.96Zn0.04Te samples. The
width of the impurity band has enabled us to estimate the total
concentration of charged impurities in the sample including
the deep defects, which are difficult to observe by other
analytical methods. The concentration of charged defects in
both samples well below 1016 cm−3 is much lower than can
be deduced from the 6N purity starting elements. With this
result, we conclude that foreign impurities are mostly neutral
or form neutral complexes in our samples.
Acknowledgment
This work is a part of the research programme
MSM113200002 financed by the Ministry of Education of
Defect-induced optical transitions in CdTe and Cd0.96Zn0.04Te
the Czech Republic and supported by the grant agency of the
Czech Republic under contract 106/01/0648.
References
[1] Molva E, Pautrat J L, Saminaydar K, Milchberg G and
Magnea N 1984 Phys. Rev. B 30 3344
[2] Ossau W, Kuhn T A and Bicknell-Tassius R N 1990 J. Cryst.
Growth 101 135
[3] Stadler W, Hofmann D M, Alt H C, Muschik T, Meyer B K,
Weigel E, Müller-Vogt G, Salk M, Rupp E and Benz K W
1995 Phys. Rev. B 51 10619
[4] Bajaj J, Shin S H, Newman P R, Hoffman J E and
Stafelbroek M G 1986 J. Vac. Sci. Technol. A 4 2051
[5] Barnett Davis Ch, Allred D A, Reyes-Mena A,
Gonzáles Hernandéz J, Gonzáles O, Hess B C and
Allred W P 1993 Phys. Rev. B 47 13363
[6] Marple D T F 1966 Phys. Rev. 150 728
[7] Lorenz M R and Segal B 1963 Phys. Lett. 7 18
[8] Marple D T F and Aven M 1967 Proc. Conf. on II–VI
Semiconducting Compounds (Providence, 1967) p 315
[9] Höschl P, Ivanov Yu M, Belas E, Franc J, Grill R, Hlı́dek P,
Moravec P, Zvára M, Sitter H and Toth A L 1998 J. Cryst.
Growth 184/185 1039
[10] Callaway J 1974 Quantum Theory of the Solid State
(New York: Academic) p 524
[11] Čápek V, Zimmerman K, Koňák Č, Popova M and Polı́vka P
1973 Phys. Status Solidi b 56 739
[12] Weiler M H 1981 Semiconductors and Semimetals vol 16,
ed R K Willardson and A C Beer (New York: Academic)
p 119
[13] Rzepka E, Marfaing Y, Cuniot M and Triboulet R 1993 Mater.
Sci. Eng. B 16 262
[14] Metropolis N, Rosenbluth A W, Rosenbluth M N, Teller A H
and Teller E 1953 J. Chem. Phys. 21 1087
[15] Grill R 1995 J. Phys.: Condens. Matter 7 3565
[16] Reboredo F A 1995 Phys. Rev. B 51 5089
[17] Soltani M, Certier M, Evrard R and Kartheuser E 1995 J. Appl.
Phys. 78 5626
[18] Selim F A and Kröger F A 1977 J. Electrochem. Soc. 124 401
[19] Magnea N and Pautrat J L 1994 Luminescence of CdTe,
CdZnTe and CdSeTe EMIS Datareviews Series No 10 Part
B3.4 ed P Capper (London: INSPEC) pp 441
[20] Cooper D E and Newman P R 1989 Phys. Rev. B 39 7431
[21] Molva E, Chamonal J P and Pautrat J L 1982 Phys. Status
Solidi b 109 635
1287