Effects of high temperature annealing on single crystal ZnO and ZnO devices
W. Mtangi, F. D. Auret, M. Diale, W. E. Meyer, A. Chawanda, H. de Meyer,
P. J. Janse van Rensburg, and J. M. Nel
Department of Physics, University of Pretoria, Private Bag X20, Hatfield 0028, South Africa
(Received 17 January 2012; accepted 2 March 2012; published online 16 April 2012)
We have systematically investigated the effects of high-temperature annealing on ZnO and ZnO
devices using current voltage, deep level transient spectroscopy (DLTS) and Laplace DLTS
measurements. Current–voltage measurements reveal the decrease in the quality of devices fabricated
on the annealed samples, with the high-temperature annealed samples yielding devices with low
barrier heights and high reverse currents. DLTS results indicate the presence of three prominent
defects in the as-received samples. Annealing the ZnO samples at 300
C, 500
C, and 600
CinAr
results in an increase in reverse leakage current of the Schottky contacts and an introduction of a new
broad peak. After 700
C annealing, the broad peak is no longer present, but a new defect with an
activation enthalpy of 0.18 eV is observed. Further annealing of the samples in oxygen after Ar
annealing causes an increase in intensity of the broad peak. High-resolution Laplace DLTS has been
successfully employed to resolve the closely spaced energy levels.
V
C
2012 American Institute of
Physics.[http://dx.doi.org/10.1063/1.3700186]
INTRODUCTION
ZnO is a wide and direct bandgap semiconductor with
experimental bandgap energy of 3.4 eV. Its high exciton bind-
ing energy of 60 meV at room temperature and resistance to
radiation damage gives it advantages and makes it a good can-
didate for fabrication of devices that can operate within the
ultraviolet region and space applications compared to other
wide bandgap materials, e.g., GaN. ZnO exists mostly as
n-type, whose conductivity is supposedly due to the existence
of native point defects, such as oxygen vacancies and Zn
interstitials.
1,2
Some other impurities, such as hydrogen, are
also claimed to contribute to the n-type conductivity in
ZnO.
3–5
As in many wide bandgap semiconductors, there
exists a so-called doping asymmetry,
6
i.e., it is easy to get n-
type ZnO, but rather difficult to produce p-type ZnO. The use
of ZnO in the optoelectronic industry for fabrication of high
performance solar cells and ultraviolet detectors requires the
material to be of high quality, i.e., with minimal defects. The
knowledge of the origins, identity, quality, and stability of
defects in ZnO is of vital importance, as defects often affect
the electrical and optical properties of the material. Defects
can also reduce device lifetime and decrease the light emis-
sion efficiency.
6
The existence of native defects in ZnO has
hindered success in the fabrication of UV-emitting diodes,
because of their self-compensation behavior.
7,3
ZnO also pos-
sesses deep-level emission bands that emit all colors in the
visible range with good color-rendering properties.
8
An
understanding of the origins of emissions related to deep-level
defects in ZnO for development of high efficiency optoelec-
tronic devices is also required. There is a need to have control
over the conductivity and defects in ZnO, i.e., by removing
them if unwanted or introducing them when required. Anneal-
ing of crystals is one possible way of recovering defects or
removing them. The study of defects in ZnO through anneal-
ing has been performed using the Hall effect,
9–11
positron
annihilation spectroscopy,
12
x ray diffraction, Rutherford
backscattering spectroscopy (RBS), admittance spectros-
copy,
13
cathodoluminescence, and photoluminescence.
14
In
this paper, we study the effects of annealing on ZnO and the
electrical properties of the devices formed on the annealed
material. The quality of the devices fabricated on the annealed
samples is characterized using IV measurements, while deep
level transient spectroscopy and Laplace DLTS are used to
study the deep-level defects introduced in the annealed
material.
EXPERIMENT
Single-crystal ZnO samples obtained from Cermet Inc.
were used in this study. In this experiment, five sets consist-
ing of two samples per set with the same specifications were
used. All the sets were annealed in argon ambient for 1 h at
different temperatures. The first set was annealed at 300
C,
the second set at 400
C, the third set at 500
C, the fourth
set at 600
C, and the fifth set at 700
C. A sample from the
same as-received wafers was also used as the control. One
sample from each annealed set was used for the fabrication
of Schottky contacts in the first part of the experiment. Prior
to Schottky contact and ohmic contact fabrication, the
annealed and control samples were degreased in acetone fol-
lowed by methanol for five minutes in an ultrasonic bath.
The five-minute degrease in methanol was followed by a
three-minute boiling in hydrogen peroxide at a temperature
of 100
C. The samples were then blown dry using nitrogen
gas. Al/Au ohmic contacts were then fabricated on the
O-polar face using the thermal evaporation technique. Pd
Schottky contacts with a diameter of 0.6 mm were then fabri-
cated on the Zn-polar face using the resistive evaporation
technique to ensure that no process-induced defects were
introduced into the material, as has been reported.
15
In the
second part of the experiment, the remaining samples from
each set were each annealed at the same temperature at
which they were annealed before, but this time in oxygen
0021-8979/2012/111(8)/084503/6/$30.00
V
C
2012 American Institute of Physics111, 084503-1
JOURNAL OF APPLIED PHYSICS 111, 084503 (2012)
Downloaded 05 Jun 2012 to 137.215.6.53. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
ambient. The cleaning of the samples was the same as the
one outlined in the first part of the experiment. Pd Schottky
contacts and Al/Au ohmic contacts with the same thick-
nesses were then fabricated, as in the first part. Current-
voltage measurements were then performed in the dark at a
temperature of 296 K. DLTS and Laplace DLTS measure-
ments were finally performed to characterize the defects
induced by annealing ZnO single crystals.
RESULTS AND DISCUSSIONS
Figure 1(a) shows the semi-logarithmic IV characteris-
tics obtained from the Schottky contacts fabricated on the
Ar-annealed and un-annealed samples, while Fig. 1(b) shows
the IV characteristics obtained on the Ar and then
O
2
-annealed samples. From these characteristics, it can be
observed that the annealed samples produce contacts with
high reverse currents compared to the un-annealed sample.
The quality of the contacts deteriorates after annealing the
Ar-annealed samples further in O
2
. Of particular interest is
the leakage current that has been obtained on the 400
C-
and 500
C-annealed samples. This is possibly due to the
effects of annealing and the defects introduced in the mate-
rial. This is going to be discussed in detail in the sections to
follow. The IV characteristics of the contacts have been
analyzed by fitting the linear part of the forward IV charac-
teristics to the pure thermionic emission model.
Table I shows the parameters that were obtained from
the IV characteristics of Figs. 1(a) and 1(b). The 500
C-
annealed samples reveal the lowest barrier heights after
annealing in Ar and also Ar and then O
2
. High barrier
heights have been measured for the 400
C-annealed sam-
ples. There has been an increase in series resistance after Ar
and then O
2
annealing for all the other temperatures exam-
ined. This could possibly be due to the formation of an insu-
lative oxide layer that might have been formed during
oxygen annealing.
Figures 2(a) and 2(b) show the DLTS spectra for the
Ar-annealed and Ar- and then O
2
-annealed samples, respec-
tively. These spectra were obtained in the 30–330-K temper-
ature range, at a reverse bias of –2.0 V, filling pulse width of
2.0 ms, and filling pulse of 0.3 V. The annealed samples and
the un-annealed samples show three prominent defects, E1,
E2, and E3, that have been observed in ZnO and reported
before.
16–19
The 300
C Ar annealing of the ZnO samples
introduces a new peak E4
300
C
, as has also been reported by
Ref. 20. After annealing at 400
C, no new peak is intro-
duced, but only the intensity of E1 increases. After 500
C
annealing, a broad peak E4
500
C
is also observed. A broad
peak E4
600
C
is also observed after 600
C annealing. After
700
C annealing, E4
T
is not observed, but a new peak Ex is
observed. (N.B.: T refers to the annealing temperature at
which the peak is introduced). The formation of new defects
affects the IV characteristics and parameters of the Schottky
contacts. As has been mentioned before, the 400
C-annealed
samples have very low leakage current and high barrier
heights, since there are no new defects introduced. It can be
stated that the defects introduced in ZnO are the cause of
high leakage currents and low barrier heights, as they modify
the current transport mechanisms by which carriers crossover
the barrier. These mechanisms could include all forms of
tunneling through the barrier.
We have annealed the Ar annealed samples in oxygen
ambient in a bid to see if the Ar annealing-induced defects
could be removed by oxygen annealing. This is because, in
our previous results,
20
we have observed that oxygen anneal-
ing does not introduce E4. From Fig. 2(b), it can be clearly
seen that Ar annealing followed by O
2
annealing does not
FIG. 1. (a) IV characteristics for the Ar-annealed samples obtained at 296 K.
(b) IV characteristics for the Ar þ O
2
-annealed samples obtained at 296 K.
TABLE I. Values of Schottky barrier height (SBH), ideality factor n, and
series resistance R
s
for the Ar and Ar þ O
2
-annealed for the best selected
contacts.
SBH (eV) Ideality factor Series resistance (X)
Annealing
temperature (
C) Ar Ar þ O
2
Ar Ar þ O
2
Ar Ar þ O
2
300 0.70 0.64 1.25 1.45 206 936
400 0.81 0.75 1.37 1.15 312 29k
500 0.63 0.48 1.72 1.54 391 411
600 0.74 0.64 1.43 1.55 110 100
700 0.73 0.74 1.17 1.14 8.4k 12k
Un-annealed 0.74 1.67 190
084503-2 Mtangi et al. J. Appl. Phys. 111, 084503 (2012)
Downloaded 05 Jun 2012 to 137.215.6.53. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
remove E4
T
, but rather causes an increase in the intensity of
E4
T
and Ex. This means that, once E4
T
has been introduced,
it is difficult to anneal it out at the same temperature at which
it forms. The E4
T
peak is very broad, indicating that it might
consist of two or more closely spaced energy levels.
Figure 3 shows the Arrhenius plots for the E1, E2,
Ex, and E3 defects. The identities of E1 ¼ E
c
–0.12denotes
that E1 is located at an energy level of 0.12 eV below the min-
imum of the conduction band. The same applies to E2 and E3.
E1 ¼ E
c
– 0.12, E2 ¼ E
c
– 0.10, and E3 ¼ E
c
– 0.30 are still
not clear, though others have tried to identify them using their
annealing behaviors to O
i
or V
zn
;forE1,
20
transition metal
related; for E3.
19
Ex has been estimated to have an activation
enthalpy of 0.18 eV. Its identity is not known yet.
High-resolution Laplace DLTS
21,22
has been employed
to separate the closely spaced energy levels in the E4 peak
intheAr-andthenO
2
-annealed samples. Figure 4(a) shows
theLaplacespectraforthe300
C-annealed sample. Lap-
lace has managed to split E4
300
C
into three peaks. The
Arrhenius plots are given in Fig. 4(b). Figure 5(a) shows
the Laplace spectra for the 500
C-annealed samples, whose
Arrhenius plots are given in Fig. 5(b). Figure 6(a) shows
the Laplace spectra obtained from the 600
C-annealed
samples, and the Arrhenius plots are presented in Fig. 6(b).
The 600
C-annealed samples have only two peaks. The
associated energy levels and apparent capture cross sections
are presented in Table II. It must be noted that the
annealing-induced defect at 300
C might not be the same
as the one introduced at 500
C and also 600
C. The
FIG. 3. Arrhenius plots obtained from the Ar-annealed samples.
FIG. 2. (a) DLTS spectra obtained from the Ar-annealed samples at a quies-
cent reverse bias of 2.0 V, V
p
¼ 0.3 V, filling pulse width of 2 ms, and rate
window of 100 s
1
. (b) DLTS spectra obtained from the Ar þ O
2
-annealed
samples at a quiescent reverse bias of 2.0 V, V
p
¼ 0.3 V, filling pulse width
of 2 ms, and rate window of 100 s
1
.
FIG. 4. (a) Laplace DLTS spectra obtained from the 300
CArþ O
2
-
annealed samples. These were measured at a reverse bias of 2.0 V, filling
pulse height of 0.30 V, and width of 2 ms. (b) Arrhenius plots obtained from
the 300
CArþ O
2
-annealed samples.
084503-3 Mtangi et al. J. Appl. Phys. 111, 084503 (2012)
Downloaded 05 Jun 2012 to 137.215.6.53. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
difference in the activation enthalpy and capture cross-
sectional area of the annealing-induced defects for different
annealing temperatures indicates the possible transforma-
tion of defects with annealing. The 300
C-induced defect
appears to be neutral, judging from its small capture cross-
section. Since, at 400
C, the defect is not observed, it can
be mentioned that it is mobile at a temperature greater than
300
C, but less than 400
C. This is in close agreement
with the first principles calculations of Janotti et al.,
23
which suggest that the V
o
2 þ
anneals out at a temperature of
655 K. This annealing out of a defect at 400
Cwasalso
observed by Vlasenko and Watkins
24
from the Optically
Detected Electron Paramagnetic resonance measurements,
where they reported the stability of the L3 peak up to a tem-
perature of 400
C, after which it disappears.
The physical meaning of the capture cross-sections of
the 500
C-annealed and 600
C-annealed contacts is not
clear yet. From the Laplace Arrhenius plots of the 500
C-
annealed samples, E4a has an apparent capture cross section,
which might suggest it to be an intermediate defect. The
600
C anneal produces a defect whose capture cross-section
is too large. This value of the capture cross-section is too
large to have any physical meaning. The error in this value is
FIG. 5. (a) Laplace DLTS spectra obtained from the 500
CArþ O
2
-
annealed samples. These were measured at a reverse bias of 2.0 V, filling
pulse height of 0.30 V, and width of 2 ms. (b) Arrhenius plots obtained from
the 500
CArþ O
2
-annealed samples.
FIG. 6. (a) Laplace DLTS spectra obtained from the 600
CArþ O
2
-
annealed samples. These were measured at a reverse bias of 2.0 V, filling
pulse height of 0.30 V, and width of 2 ms. (b) Arrhenius plots obtained from
the 600
CArþ O
2
-annealed samples.
TABLE II. Values of energy level and apparent capture cross section for the E4 defect observed after Ar þ O
2
annealing, obtained using Laplace DLTS.
E4a (T
a
) E4b (T
a
) E4c (T
a
)
Annealing
temperature (
C)
Energy
level (eV)
Capture cross
section (cm
2
)
Energy
level (eV)
Capture cross
section (cm
2
)
Energy
level (eV)
Capture cross
section (cm
2
)
300 0.49 3.9e15 0.52 6.4e14 0.53 2.3e13
500 0.62 3.0e13 0.47 2.8e15 0.44 2.1e15
600 0.88 1.5e8 0.64 3.2e12
a
T denotes the temperature at which the sample was annealed.
084503-4 Mtangi et al. J. Appl. Phys. 111, 084503 (2012)
Downloaded 05 Jun 2012 to 137.215.6.53. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
68 10
9
cm
2
. Further investigations have to be performed
to try and understand the reasons behind this large value. At
a temperature between 600
C and 700
C, this defect
anneals out, as it is not observed at 700
C. This defect might
possibly be related to the V
o
0
. This is because the V
o
0
anneals out at a temperature of 909 K,
23
which falls within
the 600
C to 700
C temperature range.
Figures 7(a)–7(c) show the depth profiles of the compo-
nents of the E4
T
peak for the ZnO samples annealed in Ar
and then O
2
at 300
C, 500
C, and 600
C, respectively. This
has been obtained at a constant reverse bias of –2.0 V and
increasing pulses in steps of 0.1 V. From the variation of the
trap concentration with depth, i.e., a decrease in concentra-
tion as we move from the interface into the ZnO bulk, it can
clearly be observed that annealing of ZnO creates defects
whose concentration decreases as we move deeper into the
semiconductor bulk. Similar observations have been noted
and reported from the Hall effect measurements.
25
CONCLUSION
As-received ZnO has three prominent defects, which are
native to the material, as observed using conventional DLTS.
Deep-level defects in ZnO can be recovered by annealing the
material at high temperatures in Ar ambient. The 300
C,
500
C, and 600
C Ar annealings introduce a broad peak
consisting of energy levels which are closely spaced and can
be separated by the high-resolution Laplace DLTS. Anneal-
ing single-crystal ZnO samples at 400
C in Ar does not
introduce a new defect, but causes an increase in intensity of
the E1 peak. The 700
C annealing does not introduce the
E4
T
, but rather introduces a level closer to the minimum
of the conduction band, Ex with an activation enthalpy of
0.18 eV. Further annealing of the samples in O
2
causes an
increase in intensity of the annealing-induced defect peaks.
From the IV characteristics of the devices fabricated on the
as-received and annealed samples, it can be observed that
the annealed samples yield poor quality devices, i.e., with
high reverse currents and low barrier heights compared to
the as-received samples. Annealing single crystal ZnO sam-
ples at high temperatures introduces defects with high con-
centrations closer to the metal/semiconductor interface that
affect device operation by modifying the current transport
mechanism from pure thermionic emission to other mecha-
nisms that include all forms of tunneling.
ACKNOWLEDGMENTS
We would like to thank the South African National
Research Foundation (NRF) for financial support. The Lap-
lace DLTS software and hardware used in the research was
kindly provided by A. R. Peaker (Centre for Electronic
Materials Devices and Nanostructures, University of Man-
chester) and L. Dobaczewski (Institute of Physics, Polish
Academy of Sciences).
1
Y. Natsume and H. Sakata, Mater. Chem. Phys. 78, 170, (2002).
2
D. C. Look, J. W. Hemsky, and J. R. Sizelove, Phys. Rev. Lett. 82, 2552,
(1999).
3
F. A. Selim, M. H. Weber, D. Solodovnikov, and K. G. Lynn, “Nature of
native defects in ZnO,” Phys. Rev. Lett. 99, 085502 (2007).
4
C. G. Van de Walle, Phys. Rev. Lett. 85, 1012, (2000).
5
D. M. Hofmann et al., Phys. Rev. Lett. 88, 045504 (2002).
6
Z. Q. Chen, S. Yamamoto, M. Maekawa, A. Kawasuso, X. L. Yuan, and
T. Sekiguchi, J. Appl. Phys. 94(8), 4807 (2003).
7
Y. Tsur and I. Riess, Phys. Rev. B 60, 8138 (1999).
FIG. 7. (a) Depth profile obtained from the 300
CArþ O
2
-annealed sam-
ples. This was determined at a constant reverse bias of 2.0 V with increasing
filling pulse height in steps of 0.1 V. (b) Depth profile obtained from the
500
CArþ O
2
-annealed samples. This was determined at a constant reverse
bias of 2.0 V with increasing filling pulse height in steps of 0.1 V. (c) Depth
profile obtained from the 600
CArþ O
2
-annealed samples. This was deter-
mined at a constant reverse bias of 2.0 V with increasing filling pulse height
in steps of 0.1 V.
084503-5 Mtangi et al. J. Appl. Phys. 111, 084503 (2012)
Downloaded 05 Jun 2012 to 137.215.6.53. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions