Full text of DOI:10.1007/BF02665841

Photoluminescence Measurements on Undoped
Journal of ELECTRONIC MATERIALS, Vol. 30, No. 6, 2001
Special Issue P6a0pe3r
CdZnTe Grown by the High-Pressure Bridgman Method
Photoluminescence Measurements on Undoped
CdZnTe Grown by the High-Pressure Bridgman Method
1,4
2
1
1
3
3
K. SUZUKI, S. SETO, T. SAWADA, K. IMAI, M. ADACHI, and K. INABE
1.—Hokkaido Institute of Technology, Sapporo 006-8585, Japan. 2.—Ishikawa National College
of Technology, Ishikawa 929-0392, Japan. 3.—Kanazawa University, Kanazawa 920-8667,
Japan. 4.— e-mail: suzukik@hit.ac.jp
The low temperature photoluminescence of Cd0.91Zn0.09Te grown by the high-
pressure Bridgman (HPB) method exhibits a neutral donor bound exciton
emission (D0X) at 1.65603 eV with its excited state (D0X*) at 1.65798 eV and
neutral acceptor bound exciton emissions (A0X) at 1.64566 eV and 1.65201 eV.
Assuming a direct generation and subsequent relaxation of excitons at the D0X*
state, we demonstrate that the temporal evolution of the above emission bands
is well reproduced by a set of rate equations. The resultant radiative-lifetime of
1.4 ns for the D0X and 1.5 and 2.0 ns for the A0Xs are compared with various
CdZnTe’s (CZTs) grown by the other methods to demonstrate the particular
nature of the HPB CZT.
Key words: CdZnTe, time-resolved photoluminescence, rate equation, bound
exciton lifetime
INTRODUCTION
ing the compensation mechanism of the HPB CZT.
Photoluminescence measurements, especially the
bound exciton luminescence, are often utilized for the
survey of residual impurities in high purity com-
pound semiconductors since several of the residual
impurities show very sharp emission lines at their
particular energy positions. In ternary compounds,
however, the bound exciton emissions are broadened
due to the effect of potential fluctuation by composi-
tional disorder (alloy broadening).4 Sometimes sev-
eral emissions are mixed up to give a broad, feature-
less emission. In ternary compounds it is, therefore,
not straightforward to obtain as much information
about residual impurities only by time-integrated
luminescencemeasurementsasinbinarycompounds.
Thetime-resolvedPLisapowerfultooltocompensate
the missing information due to the broadening of the
emissions.Theradiative-lifetimeoftheboundexciton
(BE) has been reported so far in many bulk materials
such as GaN,5 InP,6 CdS,7 CdSe,8 ZnSe,9 and ZnTe10
orinCdTe/CdMnTequantumwells.11 Insomeofthese
papers,6–9 the experimental lifetimes were discussed
based on the theoretical recombination model pro-
posed by Rashba and Gurgenishvili (RG).12 In spite of
a simple assumption for the interaction between a
neutralimpurityandanexciton, theirtheorygivesan
excellent agreement on the relationship between the
excitonic radiative lifetime (t) and the exciton local-
ization energy (Ex) as t µ EX1.5 for a variety of semicon-
In spite of its development as room temperature
nuclear detectors, the compensation mechanism of
undoped CdZnTe (CZT) grown by the high-pressure
Bridgman (HPB) method is not well understood yet.
Fiederle et al.1 pointed out the existence of a shallow
donor (Cl) and acceptors (Cu, Na, As) from the photo-
luminescence (PL) measurements and showed the
compensation model including a shallow acceptor, a
shallow donor, and a deep donor. The existence of Cl
as a shallow donor implies some similarity for the
compensation mechanism with that of semi-insulat-
ing CdTe:Cl grown by the traveling heater method
(THM). However, the growth of CdTe:Cl starts from a
Te-rich condition to provide Cd vacancy for the com-
pensation of the shallow donor to attain high resistiv-
ity. On the other hand, the growth of HPB CZT is
conducted under high pressure to suppress the loss of
Cd during growth, leading to a better stoichiometric
condition. Further, it is reported that the electron
transport, especially the electron drift mobility of
HPBCZTismainlylimitedbyscattering2 ratherthan
by trapping.3 These suggest that the compensation
mechanism is different from that of Cl-doped CdTe.
The presence of Cl in this material, therefore, is
controversial. Understanding the fundamental prop-
erties of donors and acceptors is important for reveal-
(Received November 21, 2000; accepted January 23, 2001)
603

604
Suzuki, Seto, Sawada, Imai, Adachi, and Inabe
Table I. Transition Energy of Bound Excitons in CZT
Zn
(%)
Energy
(eV)
EX
(meV)
Sample
Transition
HPB
9
D0X*
D0X
1.65798
1.65603
1.65201
1.64566
1.65476
1.64933
2.2
4.1
A0XH
A0XL
D0X
8.1
14.5
4.1
9.5
Te-solution
(In doped)
9
A0X
Fig. 1. Logarithm of time-integrated spectra of HPB CZT measured at
three temperatures of 11, 18, and 28 K. Note that the spectra are
shifted with respect to each other.
ductors.7,9,13 The time-resolved PL of CZT was first
reported by Yom et al.13 In the case of highly mixed
ternary compounds, however, the effect of composi-
tional disorder on the exciton dynamics should be
taken into account.14 We have investigated time-
resolved, as well as time-integrated, photolumines-
cence of bound excitons in undoped Cd0.91Zn0.09Te
grown by HPB and are compared with that of CZT
grown from Te-solution.
EXPERIMENT
Thesampleswerecounter-gradeCd0.91Zn0.09Tecrys-
tals (5 ¥ 5 ¥ 1 mm3) from eV PRODUCTS and CZT
crystals with several different compositions grown
from Te-rich solution (either In-doped or undoped).
The time-integrated photoluminescence was mea-
sured by using an Ar+ laser (488 nm) as an excitation
source. Themeasurementswereperformedunderthe
temperaturerangefrom11to52K.Thetime-resolved
PL was measured at 4.2 K. The frequency-doubled
light of a mode-locked titanium-sapphire laser run-
ning at an 82 MHz repetition rate with a pulse
duration of 200 fs was employed as an excitation
source. The temporal evolution of the PL emissions
from the sample was detected with a streak camera
through a 25-cm single monochromator. The system
response to the excitation laser pulse itself shows an
FWHM of 100 ps.
Fig. 2. Logarithm of integrated intensity for the D0X and the A0XL as a
function of inverse temperature. The solid lines are calculated results
by Eq. 1 for the D0X and Eq. 2 for the A0XL.
showsaprominentneutraldonorboundexciton(D0X)
emission at 1.65603 eV, a neutral acceptor bound
exciton (A0XL) emission at 1.64566 eV and its longitu-
dinal optical (LO)-phonon replica. In addition, an
emission due to the free exciton (FE) recombination is
observed at 1.66014 eV. Judging from these BE posi-
tions,4 the Zn concentration of about 0.09 is esti-
mated. At longer wavelength regions, free-to-bound
and donor-acceptor pair emissions are observable.
The energy separation between these emissions com-
pare favorably with earlier data,1 although there
exists a small systematic shift to high energy due
probably to a slight Zn-rich condition of the sample
used for our measurements. In addition to the above-
mentioned emissions, we observe an excited state of
the neutral donor bound exciton (D0X*) at 1.65798 eV
and a small structure at the high-energy shoulder
RESULTS AND DISCUSSION
Figure 1 shows the time-integrated PL spectrum of
the HPB CZT measured at three different tempera-
tures of 11, 18, and 28 K. Although the crystal was
grown nominally undoped, the spectrum at 11 K

Photoluminescence Measurements on Undoped
CdZnTe Grown by the High-Pressure Bridgman Method
605
Fig. 3. Temporal evolutions measured at each of the energy positions denoted in the figure. The dashed lines are the results of the rate equation
model convolved with the system response approximated by a Gaussian with s = 36 ps.
(1.65201 eV) of the A0XL. Although no phonon replica
is accompanied for this newly observed structure, we
tentatively assigned it as another neutral acceptor
bound exciton (A0XH). The emergence of the fine
structure for the D0X emission indicates that the
spectrum does not suffer much from the effect of
compositional disorder (inhomogeneous broadening).
The localization energy (Ex), as defined by the energy
difference between the free exciton peak and the
correspondingboundexciton,peaksfortheD0X,A0XH,
and A0XL are 4.2, 8.1, and 14.5 meV, respectively. The
localization energy of the A0XL is large as compared
with other CZT crystals measured in the present
experiment (Ex £ 10 meV), and the value is compa-
rable to the observed exciton dissociation energy of
this system.15 As is seen from the temperature depen-
dence of PL spectra, both A0Xs quench at a relatively
low temperature, namely, we do not see any trace of
the structure at T=18K for the A0XH and at T = 30 K
for the A0XL and its 1-LO replica emissions. At higher
temperatures, the phonon replica of the free exciton
emission (~1.64 eV) and a free-to-bound emission
(~1.61 eV) dominate the corresponding spectral re-
gions. Judging from the similarity of the quenching
temperature and the localization energy to those of
Cu in CdTe, the acceptor responsible for the A0XH is
assigned to be residual Cu. Qualitatively, a similar
spectrum, but with only one A0X, is obtained for In
doped Cd0.91Zn0.09Te grown from a Te solution where
the doped In acts as a shallow hydrogenic donor. The
observed peak positions of the excitonic emissions are
summarized in Table I.
The logarithm of the integrated luminescence in-
tensity for the D0X and the A0XL are plotted versus
inverse temperature in Fig. 2. Often, bound exciton
luminescence intensity exhibits a two-step-quench-
ing process as described by the following expression16
2 /kT
I(T) / I0 = (1+ C1e-E /kT + C2e-E
)
-1
(1)
1
The temperature dependence of the D0X is well ex-
pressedbythisformulawiththeparameters, C1 =8.2,
E1 = 1.4±0.2 meV, C2 = 16751, E2 = 25±6 meV. The
fitted activation energies are compared with those
determined by spectroscopic observation to identify
the dominant non-radiative recombination channels.
The energy difference of the D0X and D0X* measured
from the spectrum is 1.95 meV which is close to E1.
Further, as can be seen from Fig. 1, the intensity ratio
of D0X* to D0X increases with increasing tempera-
ture. Therefore, the dominant non-radiative channel
for the D0X exciton is a thermal excitation to the D0X*
and subsequent emission to a free state for tempera-
tures lower than 20 K. At higher temperatures, the
dissociation of the exciton is a dominant non-radia-
tivechannelprovidingthatthevalueoftheactivation
energy, E2, corresponds to the exciton dissociation
energy. Contrary to the D0X, the intensity of the A0XL
stays constant or even increases until about 18 K. As
the localization energy measured from the spectrum
and the dissociation energy for the A0XL exciton are
comparable, the temperature dependence for the
emission can be obtained by putting E1 = E2 in Eq. 1,
and then we obtain a modified expression
–1
I¢(T)/I0 = (1+C¢ e–E¢/kT
)
(2)

606
Suzuki, Seto, Sawada, Imai, Adachi, and Inabe
Table II. The Fitting Parameters Used in Figure 3 and Corresponding Theoretical Lifetime
3
Sample
Transition
N (1/cm )
Cn (cm3/ps)
en (1/ps)
Lifetime (ns) Exp.
Theory
HPB
D0X*
D0X
4 ¥ 10-19
G = 3 ¥ 10–3 (1/ps)
3 ¥ 10–20
5 ¥ 10-3
8 ¥ 10–4
—
1.3
1.4
1.5
2.0
2 ¥ 1016
8 ¥ 1015
0.39
1.1
2.5
A0XH
A0XL
1.9 ¥ 1016
2 ¥ 10–20
—
A fairly good fit is obtained with E¢ = 15.2 meV, the
value is in close agreement with the exciton dissocia-
tion energy.15 This indicates that the thermal run
away of A0XL exciton to any higher excited states or to
the free exciton state is not necessary to take into
account for the analysis of the dynamics at low tem-
peratures.
ThetemporalevolutionsoftheFE,D0X*,D0X,A0XH,
and A0XL are shown in Fig. 3. As is seen from the
figure, aqualitativelysimilarbehaviorisobservedfor
the FE and the D0X* decay. Namely, the intensity of
these emissions shows a steep rise at the onset and a
rapid decrease followed by a slower decay thereafter
(double exponential decay). On the other hand, those
for the D0X, A0XH, and A0XL show a slower rise reach-
ing the peak well after the excitation (~500 ps).
Obviously, the slow rise of these emissions indicates
the capture process of the free exciton while the steep
rise of the D0X* indicates the direct exciton genera-
tionattheneutraldonorsitebylaserexcitation. After
the maximum intensity is reached, both A0Xs show a
simple exponential decay, while a dispersive decay is
observed for the D0X at times longer than 4 ns. Based
on these observations and what we have learned from
the temperature dependence of time-integrated PL, a
set of rate equations to describe the exciton dynamics
is proposed as follows:
Fig. 4. Bound exciton lifetimes versus the localization energy for the
CZT samples with various Zn concentrations as denoted in the inset.
The dashed line indicates t µ EX1.5 relation. (a) Ref. 12, (b) CdTe:Cl
grown by THM.
dnexn
nex
(N
dt
- n)
*
*
ex AH
= -
- CDnex (ND - nD - nD )
t
ex
(3)
(4)
-CAHnex (NAH - nAH
)
*
*
-CALnex (NAL - nAL ) + eDnD
where ND,, NAH, and NAL are the total concentrations
ofaneutraldonor, andneutralacceptors. nD*, nD, nAH,
and nAL are the number of the occupied sites by
excitons. CD*, CAH, and CAL are capture coefficients of
excitons to each center. t represents a radiative life-
time of each exciton state. eD and eD* are the thermal
emission rate from the neutral donor and its excited
state. GD is the relaxation rate of excitons from the
D0X* to the D0X. The free excitons are captured by the
D0X*, A0XH, and A0XL states. The D0X state, however,
is formed only by the relaxation of D0X*. It is interest-
ing to note that capture and re-emission of excitons
viaanintermediate(excited)stateareproposedforan
A0X transition in Zn-rich CZT.12 Re-emission of the
bound excitons at A0X states is not taken into account
based on the observed quenching behavior as de-
scribed previously. On the other hand, we assume
thermal release of the donor bound exciton to the free
statethroughitsexcitedstate.Thedashedlinesinthe
dn*D
dt
n*D
= -
+ C*Dnex (ND - nD - n*D)
tD *
-e*Dn*D - GDn*D + eDnD
dnD
dt
nD
tD
+ GDn*D = eDnD
(5)
(6)
(7)
= -
dnAH
dt
nAH
t AH
= -
= -
+ CAHnex (NAH - nAH)
dnAL
dt
nAL
t AL
+ CALnex (NAL - nAL
)

Photoluminescence Measurements on Undoped
CdZnTe Grown by the High-Pressure Bridgman Method
607
figure indicate the calculated results using the pa-
rameters summarized in Table II. The donor and
acceptor concentration of the order of 1016 cm–3 is
assumed based on the result reported by Fiedelre et
al.1 In addition, 10% of the initial excitons are as-
sumedtobegeneratedattheD0X*bylaserexcitation.
ExceptforthedispersivebehavioroftheD0Xatlonger
times, the intensity change over three orders of mag-
nitude is reproduced quite well for each of the emis-
sions. The similarity of the resultant capture coeffi-
cient and decay time constant for the A0XH with the
A0XL supports the identification of the A0XH as a
recombination of the neutral acceptor bound exciton.
The lifetime for BE emissions, especially that for the
D0X, is extremely long for its localization energy.
The theoretical bound exciton lifetime is given by17
tions and thus better crystal quality in general. How-
ever, the D0X lifetime for the HPB sample is even
muchlongerthanthatofthebinaryCdTe:Clgrownby
▲
THM( inFig.4). Thislongerlifetimeandadeparture
from the power law are demonstrated to be a particu-
lar nature of HPB CZT.
CONCLUSIONS
The time-resolved as well as time-integrated PL of
HPB CZT has been investigated. A neutral donor
bound exciton, its excited state, and two neutral
acceptor bound exciton emissions are specified in
time-integrated PL. From the temperature depen-
dence of integrated intensity, a dominant non-radia-
tive recombination channel at low temperatures is
attributed to thermal excitation of excitons back to
the free state through the excited for the D0X; while
for the A0XL, thermal excitation is not possible due to
its deep localization. The temporal evolution over
three orders of magnitude of the free exciton and all
of the bound exciton emissions including an excited
state are well reproduced by a set of rate equations.
The long lifetime of the D0X and a departure from the
power law for the lifetime versus localization energy
are particularly observed in HPB CZT.
3m0c3
2ne2w fBE
2
l
(8)
tBE
=
= 4.5
2
nfBE
where n is the refractive index of the sample; w is the
angular transition frequency; fBE is the oscillator
strength;l isthetransitionwavelengthincm;andm0,
c, earethefundamentalphysicalconstantswiththeir
usualmeanings.Theboundexcitonoscillatorstrength
is
REFERENCES
fBE 8pa3EX
1. M. Fiederle, T. Feltgen, J. Meinhardt, M. Rogalla, and K.W.
Benz, J. Cryst. Growth 197 635 (1999).
2. K. Suzuki, S. Seto, A. Iwata, M. Bingo, T. Sawada, and K.
Imai, J. Electron. Mater. 29 704 (2000).
(9)
=
fFE
W
3. K. Zanio, Semiconductors and Semimetals (New York: Aca-
demic Press, 1978), p. 103.
4. K. Oettinger, D.M. Hoffman, Ai.L. Efros, B.K. Meyer, M.
Salk, and K.W. Benz, J. Appl. Phys. 71 4523 (1992).
5. G. Pozia, J.P. Bergman, T. Paskova, and B. Monemar, Phys.
Stat. Sol. (B) 216, 45 (1999).
6. U. Heim, Phys. Stat. Sol. (B) 48, 626 (1971).
7. C.H. Henry and K. Nassau, Phys. Rev. B 1, 1628 (1970).
8. F. Minami and K. Era, Solid State Commun. 53, 187 (1985).
9. T. Steiner, M.L.W. Thewalt, and R.N. Bhargava, Solid State
Commun. 56, 933 (1985).
where W is the volume of one Cd(Zn)Te molecule, and
aEX is the characteristic radius of the bound exciton.
The following (RG) model,12 aEX is estimated by as-
suming a d-function potential for the bound-exciton
states, and is given by
h
(10)
aEX
=
2mEX
10. W. Schmid and P.J. Dean, Phys. Stat. Sol. (B) 110, 591
(1982).
11. M.O’Neill,P.Harrison,M.Oestreich,andD.E.Ashenford,J.
Appl. Phys. 78, 451 (1995).
12. S.S. Yom, S. Perkowitz, P.M. Amirtharaj, and J.J. Kennedy,
Solid State Commun. 65, 1055 (1988).
13. E.I. Rashba and G.E. Gurgenishvili, Soviet Phys.—Solid
State 4, 759 (1962).
14. K. Suzuki, S. Seto, K. Imai, T. Sawada, U. Neukirch, and J.
Gutowski, J. Electron. Mater. 28, 785 (1999).
15. E. Cohen, R.A. Street, and A. Muranevich, Phys. Rev. B 28,
7115 (1983).
16. D. Bimberg, M. Sondergeld, and E. Grobe, Phys. Rev. B 4,
3451 (1971) .
17. D.L. Dexter, Solid State Physics, Vol. 6, ed. F. Seitz and D.
Turnbull (New York: Academic, 1958), p. 361.
18. B. Segall and D.T.F. Marple, Physics and Chemistry of II-VI
Compounds, ed. M. Aven and J. S. Prener (Amsterdam:
North-Holland, 1967), p. 336.
where m is the translational mass of an exciton, and
EX is the localization energy of bound exciton. Com-
bining Eqs. 8–10, the power law t µ EX1.5 is obtained.
In addition, assuming that the liner interpolation
between CdTe and ZnTe is valid for W, m, fFE,18 and n,
we calculated the theoretical lifetime for HPB CZT as
summarized in Table II.
In Fig. 4, the experimental lifetimes for CZTs of
various compositions are summarized as a function of
exciton localization energy. Obviously, the power law
is well established for most of the CZT samples,
however, it does not hold for the HPB sample. Experi-
mentally, thelifetimeisdeterminedbythecompeting
rates of radiative and non-radiative recombinations.
The longer experimental lifetimes of bound excitons
indicate less possibility of non-radiative recombina-