JOURNAL OF APPLIED PHYSICS 103, 124312 共2008兲

Spectral investigation of carrier recombination processes in ZnO nanorods

at high temperatures
A. Gupta, M. Omari, and N. Kouklina兲
Department of Electrical Engineering and Computer Science, University of Wisconsin-Milwaukee,
Milwaukee, Wisconsin 53211, USA
共Received 22 February 2008; accepted 20 April 2008; published online 27 June 2008兲
The mechanism of near-band-edge 共NBE兲 emission from crystalline ZnO 共c-ZnO兲 nanorods grown
on c-Si by a catalyst-assisted vapor-liquid-solid method has been investigated by performing
temperature-, power-, and time-dependent photoluminescence 共PL兲 measurements at a temperature
共T兲 range of 143– 503 K. In contrast to previous reports, we find that the NBE PL is primarily
associated with free exciton emission, whereas the contribution of band-to-band and free-to-bound
radiative recombinations remains negligible up to the highest T studied. A spectral evolution of the
NBE band with T was further analyzed within the framework of a three-parameter model, proposed
recently, with the results presented and discussed. Finally, the ratio of excitonic-to-defect
luminescence intensity has been observed to change nonmonotonically with T, which is explained
based on the difference in the quenching mechanisms of exciton and defect PL. © 2008 American
Institute of Physics. 关DOI: 10.1063/1.2948938兴

I. INTRODUCTION ter are expected to contribute to near-band-edge 共NBE兲 emis-

Wide-band-gap semiconductors have been and still re-
main a subject of intense research owing to their high poten-
tial for application in advanced radiation hard, high power,
and optical devices. ZnO, a II–VI compound, has a large
direct-band gap 共⬃3.3– 3.4 eV兲 that makes this material ad-
ditionally promising for developing solid-state light emitting
and sensing devices that operate in the ultraviolet 共UV兲 part
of the electromagnetic spectrum. Compared with GaN, ZnO
demonstrates a much larger free exciton binding energy EB
of ⬃60 meV 共⬎RT⬃ 25 meV兲, which ensures stable high-
efficiency light emission even at increased temperatures.
Nowadays, as the device sizes fast approach a critical nano-
scale range, the focus of ZnO research has also shifted from
bulk samples to nanowires/nanorods; the latter can now be
synthesized economically and controllably in the form of
highly faceted single crystals on a variety of substrates, in-
cluding optically inefficient Si 共Refs. 1–4兲 via, for example,
a catalyst-assisted vapor-liquid-solid 共VLS兲 route.5
The photoluminescence 共PL兲 characteristics of ZnO
nanowires/nanorods were investigated and reported.1,2,6,7,4 In
general, as-grown undoped samples typically reveal a strong
luminescence at both UV and visible parts of the spectrum,
attributed to exciton and surface defect emissions, and origi-
nate from physically separate regions: the crystalline core
and the surface layer, respectively.8 At temperature of a few
tens of kelvins, excitons remain localized on internal crystal
defects and/or impurities, which can be unintentionally intro-
duced into the core during the growth. The complexes, which
are also known as acceptor and donor bound excitons, con-
trol the UV luminescence characteristics of nanostructured
ZnO at this temperature range.9,10 As T increases, the defect
bound excitons dissociate into free excitons, whereas the lat-
a兲
Author to whom correspondence should be addressed. Electronic mail:
nkoulin@uwm.edu.
sion up to T of ⬃EB / kB 共kB is Boltzman constant兲, which
approximates to 600– 700 K in ZnO crystals with a size
larger than the exciton Bohr diameter in this semiconductor
共i.e., ⬃1.8 nm兲. As the dissociation of free excitons is inevi-
tably accompanied by the release of free charge carriers,
NBE emission can also originate from the radiative recom-
bination of free carriers. The consideration of these effects
remains central to many applications of ZnO nanowires/
nanorods in future optoelectronic nanodevices; however, the
behavior has not been elucidated in detail, especially for a
high temperature regime, whereas many existing literature
reports also seem to be controversial. Several authors have
performed T-dependent PL studies that involved nanostruc-
tured ZnO, including nanoneedles,11 nanowires,10 and
nanotubes,12 as well as bulk and thin films of ZnO.13,14 Most
of these studies were carried out at reduced and near room
temperatures, and while the results are greatly important to
understanding the general physical mechanisms of light
emission in low-dimensional systems, the findings cannot be
readily used to infer high-T PL behavior in nanowires and
nanorods.
More recently, Chen et al.15 investigated the behavior of
NBE PL in bulk ZnO crystals at elevated T. The authors
found that the PL intensity I increased with the excitation
power, P as I ⬃ P␣, where an exponent ␣ reportedly changed
from ⬃1.5 to 1.7 as T is increased from ⬃300 to 823 K. The
results suggest that the UV emission involves a radiative
recombination of both photogenerated free carriers and exci-
tons. Conversely, we find that the high-T NBE emission from
c-ZnO nanorods originates from a radiative recombination of
free excitons, which function as a dominant radiative carrier
recombination channel. The absence of any emission from
the free carrier recombination is in part attributed to the trap-
ping of the photogenerated free carriers by thermally ionized
surface defect states. In this work, the ratio of excitonic-to-

0021-8979/2008/103共12兲/124312/5/$23.00 103, 124312-1 © 2008 American Institute of Physics

Downloaded 03 Sep 2011 to 129.89.24.43. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
 124312-2 Gupta, Omari, and Kouklin
defect integral intensity was additionally investigated and
found to change nonmonotonically. The ratio exhibits an ex-
ponential growth with T for T ⬎ 400 K, suggesting a much
stronger quenching of defect versus excitonic PL, which is
also discussed. Among other findings, we also report that a
blue emission band with a peak at ⬃420 nm, which was
observed in the case of ZnO nanoneedle array,11 was found
to originate from a lens autofluorescence effect instead. Its
contribution was avoided with the help of a spectral decom-
position, the results of which are presented and discussed, as
well.
II. GROWTH OF ZNO NANORODS
The growth of ultradense freestanding arrays of mono-
disperse nanowires, which are particularly advantageous to
the PL measurements performed herein, was significantly fa-
cilitated with the use of hexagonally shaped self-assembled
alumina templates as shadow nanomasks for the site-specific
deposition of a single diameter two-dimensional Au-nanodot
array. The deposited Au dots act as nucleation and growth
sites of ZnO nanowires/nanorods; the approach is called a
second-order self-assembly technique, which has been eluci-
dated in more detail elsewhere.2 The procedure was used
here to produce ZnO nanorods atop of the 共100兲 c-Si in a
horizontally oriented quartz tube furnace. High purity ZnAs2
powders were heated to ⬃450 ° C to generate Zn vapors.
According to a scanning electron microscope 共SEM兲 analy-
sis, most of the as-grown structures revealed a hexagonal
shape and had a Au-nanodot on top, which is a characteristic
of a VLS growth mechanism.
The average diameter of the nanorods is ⬃90 nm and
their areal density approaches ⬃1012 cm−2. For the growth
time of ⬃20 min, the length of the nanorods typically ap-
proaches ⬃500 nm. A transmission electron microscope
analysis was routinely performed and the results revealed
that nanorods consist of a crystalline core region surrounded
by a small surface-defect layer, which is typical for
nanorods/nanowires grown by this route. The formed surface
states have electronic configurations different from those of
the bulk counterparts and are known to provide a source of
bright visible-range luminescence.8
III. T-DEPENDENT PL MEASUREMENTS
The luminescence from the nanorods was excited with a
spectrally filtered 320 nm line generated by a xenon lamp.
The light was focused into an ⬃10 ␮m diameter spot with an
Olympus 51X microscope, whereas the emitted radiation
was collected and analyzed using a double grating mono-
chromator. A photomultiplier tube operating in a photon
counting mode served as a detector. The temperature of the
samples was controlled using an Instec HCS302 heating/
cooling stage that uses liquid nitrogen and a resistive heating
element for cooling and heating, respectively. To minimize
the effect of moisture condensation, a positive pressure was
created at all times by purging the sample compartment with
an argon gas. The spectral resolution of the instrument was
limited to ⬃2 nm, with the detector integration time set not
to exceed 0.5 s for each spectral acquisition. To avoid pro-
Downloaded 03 Sep 2011 to 129.89.24.43. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys. 103, 124312 共2008兲
FIG. 1. 共Color online兲 PL spectra obtained for the temperature range of
143– 503 K from ZnO nanorods. The spectra are offset vertically for clarity.
The inset shows a result of the decomposition of the PL spectrum 共143 K兲
by multipeak Loretzian fitting. The second inset shows a SEM image of
ZnO nanorods. The scale bar is 400 nm.
longed exposure of the samples to an argon ambient under
high T, the spectra were recorded at the intervals of ⬃20 K
for T ⬍ 300 K and of ⬃60 K for T ⬎ 300 K.
IV. RESULTS AND DISCUSSIONS
Figure 1 shows PL spectra obtained on ZnO nanorods at
varied T range. At low T, the spectrum is dominated by three
bands; when T exceeds ⬃460 K, only two bands can be
clearly identified. We note that a middle band 共having a peak
position at ⬃420 nm兲, which becomes more distinct at
higher T, is an instrument artifact as the emission originates
from a lens autofluorescence effect. A narrow linewidth UV
peak is attributed to NBE emission, whereas a broad line-
width low-energy peak is attributed to defect emission. To
investigate the mechanism of the NBE emission, the spectra
were decomposed by performing a Lorentzian fit to the data;
the key fitting results are presented in Fig. 2.
The mechanism of NBE emission is T-dependent and
involves a radiative recombination of free and defect bound
excitons at low T and that of free excitons and free carriers
when T is raised to and above RT. As it will be discussed
further, defect bound excitons must be ruled out as a possible
origin of the luminescence. To differentiate between the two
other mechanisms, excitation-power dependent PL spectra
were acquired at three different temperatures: 143, 303, and
423 K. As shown in Fig. 2 共inset兲, the obtained dependences
were linear independent of temperature. These findings are
analyzed in detail based on the following discussions.
Under steady-state cw-illumination conditions, a linear
relationship is normally expected when 共i兲 the emission is
due to free electrons and holes combined with a strong ra-
diative recombination mechanism, and/or 共ii兲 the emission is
due to free excitons independent of the strength of a radiative
recombination. In contrast, a quadratic relationship is ex-
pected when the luminescence is produced by a free carrier
 124312-3 Gupta, Omari, and Kouklin
FIG. 2. 共Color online兲 Excitonic PL spectra obtained by Lorentzian peak
decomposition and shown for the 143– 503 K temperature range. The inset
shows integral intensity of UV band as a function excitation intensity at
different temperatures of 303 and 423 K.
recombination combined with a strong nonradiative recom-
bination mechanism, while any intermediate situation should
apparently yield a power law dependence with an exponent
falling somewhere in between one and two.16,17
A strong decrease in the NBE emission is taken as evi-
dence of the increase in the role of nonradiative recombina-
tion processes with the onset of high T. At the same time,
since no shift in the exponent was observed as T grows 共Fig.
2 inset兲, based on the above statements we conclude that the
NBE emission can originate only from free exciton recom-
bination independent of temperature, conversely to what has
been experimentally obtained by Chen et al. in the case of
bulk ZnO. A special care, however, must be taken with re-
gard to a free carrier density in ZnO nanowires. In particular,
we note that 共for low excitation-power densities兲 PL intensity
is also expected to scale linearly with excitation powers if an
equilibrium 共dark兲 free carrier density remains large, such as
in the case of strong doping effects, which are typical for
ZnO nanowires.18 In this connection, we performed a set of
similar photoconduction tests on individual nanowires under
RT. When excited with UV radiation, the nanowires exhib-
ited a marked photoresponse that changed linearly with the
excitation powers 共for low applied biases兲. The results sug-
gest that strong rather than weak injection conditions pre-
vailed in our case; thus, we confirmed the validity of the
model used.
In addition, time-resolved PL measurements were per-
formed by exciting the samples with a 340 nm light emitting
diode, whereas a time decay constant obtained was signifi-
cantly less than ⬃1 ns, a characteristic of excitonic emission
in ZnO crystals.13 As the exponent exceeds one in our case,
the findings rule out the possibility that emission might origi-
nate from free-to-bound transitions19 such as those involving
shallow donor and acceptor states ubiquitously present in
ZnO nanowires/nanorods. In fact, a strong influence of As-
dopants, which form acceptor states on the PL characteristics
Downloaded 03 Sep 2011 to 129.89.24.43. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys. 103, 124312 共2008兲
FIG. 3. 共Color online兲 Shows the excitonic peak position as a function of T
共squares兲 and results of fitting with a three-parameter model 共rhombuses兲.
of ZnO nanowires, was recently observed by Lee et al.6 In
the course of their T-dependent PL measurements performed
at T of 13– 290 K, the authors confirmed the appearance of a
new UV-emission band that has been attributed to a donor-
acceptor pair 共DAP兲 recombination. The band further
showed a gradual broadening and transformation into free-
to-bound emission band for T approaching ⬃100 K, whereas
the emission eventually vanished at near RT. While the na-
norods arrays in our case have grown in an As-rich atmo-
sphere, neither elemental analysis nor PL measurements con-
firmed the presence of As dopants as it was reported in Ref.
6. We attribute this in part to a low-T growth regime, as this
can easily limit the incorporation of As-dopants into the na-
norods.
A strong evolution of the excitonic band with tempera-
ture can be further seen in Figs. 1 and 2, with two key trends:
共i兲 a pronounced decrease in excitonic intensity and 共ii兲 a
redshift in the peak position as T increases. The shift in the
peak position is attributed to the effect of the band-gap renor-
malization, similar to what is known for both low-
dimensional and bulk samples of ZnO.11,6 In our case, since
the emission mechanism stays unchanged across the whole
range of T studied, a highly accurate modeling of the band-
gap evolution with temperature can be readily performed. To
verify the mechanism responsible for the shift, we refer to a
physically more appropriate three-parameter model that was
developed and offered to replace a highly empirical Varshni
formula.20 In the framework of the improved model, a band-
gap dependence on T can be described with the following
expression:
Egap = Egap共0兲 − Sប␻ coth冋 冉 冊 册 ប␻
2kT
−1 ,
where ប␻, S, and Egap共0兲 are used as fitting parameters and
represent average phonon energy, dimensionless electron-
phonon coupling constant, and the band-gap energy at T
⬃ 0 K, respectively. The results of fitting the experimental
data with the above equation are presented in Fig. 3. The
band-gap energy of ZnO nanowires at T ⬃ 0 K is ⬃3.37 eV,
which is similar to what has been reported before 共see, for
 124312-4 Gupta, Omari, and Kouklin
FIG. 4. 共Color online兲 Shows a semilog plot of integral excitonic intensity
as a function of inverse temperature. The obtained exciton binding energy,
EB is 59 meV as discussed in the text.
instance, Ref. 15兲, where the obtained average phonon en-
ergy of ⬃58 meV is closely comparable to E1共TO兲 and
E2-high phonon energies 共being ⬃50 and 54 meV, respec-
tively兲 in bulk ZnO.21,22 At the same time, the obtained value
of the electron-phonon coupling constant S ⬃ 5.08 is rela-
tively large compared with, for instance, GaAs and GaP 共for
which S is ⬃3.00 and 3.35, respectively兲,20 suggesting a
strong role of electron-phonon interaction behind T-induced
band-gap change in ZnO crystals. More studies are required
to investigate the exact role of E1共TO兲 and E2-high phonon
modes behind a band-gap shrinkage in ZnO nanorods; thus,
the investigations are postponed until future work.
A decrease in the PL intensity, which can be clearly ob-
served in Figs. 1 and 2, is prescribed to a thermal dissocia-
tion of free excitons at the onset of high T. The change is
typically accompanied by a phonon-induced broadening of
the spectrum linewidth, which also is presented in Fig. 2.
The thermal quenching has been revealed for many different
materials systems and can be universally described by the
following equation:
I0
冉 冊
I= , served can be qualitatively understood by separately consid-
Ea
1 + B exp −
k BT
where I0 is the intensity at T = 0 K, B is a constant, and Ea
stands for the activation energy. The experimental results for
the low-T part were used to extract the activation energy by
fitting the data with the above equation with the fitting re-
sults presented in Fig. 4. The obtained value of Ea
⬃ 59 meV closely corresponds to the free exciton binding
energy EB of bulk ZnO. As a result, we also must rule out
defect bound9 and surface excitons23 as a possible origin of
NBE luminescence in our case. The result is in good agree-
ment with the above discussions.
The absence of any free carrier emission can be attrib-
uted in part to a reduced strength of screening effects in low
one-dimensional wide-band-gap semiconductor systems,
Downloaded 03 Sep 2011 to 129.89.24.43. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
J. Appl. Phys. 103, 124312 共2008兲
FIG. 5. Shows a ratio of excitonic-to-defect 共integral兲 intensity as a function
of temperature. The inset shows semilog plot of the ratio for higher T range
of ⬃400– 600 K.
which promote a partial trapping of free carriers onto ionized
surface defect states. These states can easily participate in the
transport, for example, via interstate hopping or tunneling, as
the barrier associated with the traps can be effectively re-
duced under the application of an external bias; this possibly
explains why a relatively strong photoconduction response
can be readily observed in nanorods and nanowires, as we
mentioned above and in, for instance, Refs. 24 and 25. In
addition, we also performed PL-excitation studies and the
results suggest the process 共of the carrier localization兲 is
likely instead to lead to a defect-assisted 共i.e., sub-band-gap兲
emission via DAP mechanism.
Finally, we investigated the evolution of the 共integral兲
excitonic-to-defect intensity ratio as a function of T. As
shown in Fig. 5, the ratio initially decreases while it remains
almost unchanged for intermediate T, and it eventually ex-
hibits an exponential growth 共Fig. 5, inset兲, pointing onto a
much stronger thermal quenching of defect versus exciton
emission at T ⬎ 400 K. While the process of exciton disso-
ciation and subsequent capture of the released free carriers
by the defect states is expected to generally increase the de-
fect luminescence intensity, the nonmonotonic behavior ob-
ering key mechanisms responsible for exciton and defect PL
quenching. In particular, since the rate of the ionization of
deep defect states remains initially lower compared with that
of exciton dissociation, the ratio first tends to decrease with
T. As T is raised further, the ratio experiences a small
growth, as both rates effectively counterbalance each other.
Finally, at the regime of very high T, a progressively larger
number of deep defect states additionally become affected.
Under overall low illumination powers, the repopulation of
the deep ionized states then cannot be sufficient anymore to
fully compensate for the ionization losses.
V. CONCLUSION
In conclusion, the mechanism underlying high-T NBE
emission from crystalline one-dimensional ZnO nanostruc-
 124312-5 Gupta, Omari, and Kouklin J. Appl. Phys. 103, 124312 共2008兲

tures has been investigated based on the example of VLS- 15, 1152 共2004兲.
5
R. Wagner and W. Ellis, Appl. Phys. Lett. 4, 89 共1964兲.
grown ZnO nanorods. In contrast to bulk ZnO, the results 6
W. Lee, M. Jeong, and J. Myoung, Appl. Phys. Lett. 85, 6167 共2004兲.
clearly confirm that the processes involving the radiative re- 7
M. Zhao, A. Li, A. Jin, A. Zhang, A. Hu, A. Ahmad, A. McCartney, and
combination of free excitons have a dominant role behind A. Zhu, J. Phys.: Conf. Ser. 61, 703 共2007兲.
8
light generation in c-ZnO nanorods/nanowires. The spectral I. Shalish, H. Temkin, and V. Narayanamurti, Phys. Rev. B 69, 245401
evolution of the excitonic peak position with T has been 共2004兲.
9
B. Meyer, J. Sann, D. Hofmann, C. Neumann, and A. Zeuner, Semicond.
modeled with a three-parameter model, with the findings Sci. Technol. 20, S62 共2005兲.
confirming the spectral change in the NBE peak position 10
Q. Zhao, M. Willander, R. Morjan, Q. Hu, and E. Campbell, Appl. Phys.
originating from a band-gap renormalization effect. At the Lett. 83, 165 共2003兲.
11
same time, the results also point to an increased role of B. Cao, W. Cai, and H. Zeng, Appl. Phys. Lett. 88, 161101 共2006兲.
12
W. Xu, Z. Ye, D. W. Ma, H. M. Lu, L. P. Zhu, B. H. Zhao, X. D. Yang, and
electron-phonon coupling behind a band gap change in ZnO Z. Y. Xu, Appl. Phys. Lett. 87, 093110 共2005兲.
nanorods with T. An onset of a strong quenching of the de- 13
B. Guo, Appl. Phys. Lett. 82, 2290 共2003兲.
fect PL also was observed to start at T ⬃ 400 K and attributed 14
W. Shan, W. Walukiewicz, A. Ager, K. Yu, H. Yuan, H. Xin, G. Cantwell,
to a net increase in a thermal ionization rate of deep defect and J. Song, Appl. Phys. Lett. 86, 191911 共2005兲.
15
X. Chen, J. Huso, J. Morrison, and L. Bergman, J. Appl. Phys. 102,
states. Given a nonmonotonic T-evolution of the excitonic- 116105 共2007兲.
to-defect intensity ratio, we finally note that the ratio cannot 16
J. Fouquet and A. Siegman, Appl. Phys. Lett. 46, 280 共1984兲.
17
be used as a single indicator of a crystal quality of ZnO N. Kouklin, J. Liang, M. Tzolov, and J. Xu, J. Phys. D: Appl. Phys. 36,
nanostructures. 2634 共2003兲.
18
Multiple donor and acceptor states exist in ZnO, for nanowires, see, for
example, C. X. Xu, X. W. Sun, X. H. Zhang, L. Ke, and S. J. Chua,
ACKNOWLEDGMENT Nanotechnology 15, 856 共2004兲.
19
T. Schmidt, K. Lischka, and W. Zulehner, Phys. Rev. B 45, 8989 共1992兲.
The study was in part supported by NSF Grant No. 20
K. P. O. Donnell, and X. Chen, Appl. Phys. Lett. 58, 2924 共1991兲.
ECCS 0621919. 21
C. Arguello, D. L. Rousseau, and S. P. Porto, Phys. Rev. 181, 1351 共1969兲.
22
N. Ashkenov, B. N. Mbenkum, C. Bundesmann, V. Riede, M. Lorenz, D.
1
Z. Fan, P. Chang, J. Lu, E. Walter, R. Penner, C. Lin, and H. Lee, Appl. Spemann, E. M. Kaidashev, A. Kasic, M. Schubert, and M. Grundmann, J.
Phys. Lett. 85, 6128 共2004兲. Appl. Phys. 93, 126 共2003兲.
2 23
H. Chik, J. Liang, S. Cloutier, N. Kouklin, and J. Xu, Appl. Phys. Lett. 84, L. Wischmeier, T. Voss, I. Ruckmann, and J. Gutowski, Nanotechnology
3376 共2004兲. 19, 135705 共2008兲.
24
3
K. Keem, H. Kim, G.-T. Kim, J. S. Lee, B. Min, K. Cho, M.-Y. Sung, and H. Kind, H. Yan, B. Messer, M. Law, and P. Yang, Adv. Mater. 共Wein-
S. Kim, Appl. Phys. Lett. 84, 4376 共2004兲. heim, Ger.兲 14, 158 共2002兲.
25
4
S. Zhou, X. Zhang, X. Meng, K. Zou, X. Fan, and S. Wu, Nanotechnology N. Kouklin, Adv. Mater. 共Weinheim, Ger.兲 20, 2190 共2008兲.

Downloaded 03 Sep 2011 to 129.89.24.43. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions