JOURNAL OF APPLIED PHYSICS

VOLUME 95, NUMBER 10

15 MAY 2004

Impact of the k-linear term on nonlinear optical response

of the C-exciton manifold in ZnO
K. Hazu,a) K. Torii,b) and T. Sotac)
Department of Electrical, Electronics, and Computer Engineering, Waseda University, Shinjuku,
Tokyo 169-8555, Japan

S. Adachid)
Department of Applied Physics, Hokkaido University, Kitaku, Sapporo 060-8628, Japan

SF. Chichibue)
Institute of Applied Physics, Graduate School of Pure and Applied Sciences, University of Tsukuba,
1-1-1 Tennodai, Tsukuba, Ibaraki 305-8573, Japan

G. Cantwell
Director of Research Zn-Technology, Inc. 910 Columbia St. Brea, California 92821

D. C. Reynolds and C. W. Litton

Materials and Manufacturing Directorate, Air Force Research Laboratory,
Wright-Patterson Air Force Base, Ohio 45433

Received 27 October 2003; accepted 27 February 2004

Time-integrated and spectrally resolved four-wave mixing FWM has been applied to study the
impact of the k-linear term on nonlinear optical response of the C-exciton manifold in ZnO. From
the excitation wavelength dependence of the FWM signal, we have found possible evidence that the
mixing of the 1 and 5L states leads to a three-branch excitonic polariton dispersion but the
oscillator strength transfer is small. The energy separation between the lower polariton branch and
the extra branch due to the k-linear term has been found to be 3.5 meV. In addition, a binding energy
of biexcitons consisting of two C excitons of 1.4 meV has been obtained. 2004 American
Institute of Physics. DOI: 10.1063/1.1710727

ing to circumstances.1,6 8 Since the conduction band belongs

to the irreducible representation of 7 , the irreducible representation of the ground state C-exciton manifold is 1 7
7 1 2 5 . Here 1 on the left hand side represents
the symmetry of the exciton envelope function. Excitons
with the 1 and 5 symmetry are dipole-allowed for the
electric field polarized parallel and perpendicular to the c
axis, i.e., E c and Ec, respectively. Polarized reflectance
spectra912 reveal that the C-exciton manifold is isolated
from B exciton (n1) one by about 38.8 meV and located
still in the band gap, while the C-exciton resonance energies
are close to the n2 resonance energy of the A and B excitons. The oscillator strength of the 1 exciton is much larger
than that of the 5 exciton. However, there is little knowledge about the impact of the k-linear term on nonlinear optical response of the C-exciton manifold and the binding energy of the biexciton consisting of two C-hole excitons,
XX CC .
Four-wave mixing FWM signals are known to reflect
the dynamics of many-body Coulomb correlations such as
biexciton and two-pair scattering continuum resonances even
in a low excitation limit. In addition, FWM signals are background free emission signals due to coherent polarization
and, thus, are much more sensitive to small energy splitting
in comparison with signals recorded by the linear spectroscopic method.13
In this article, we have experimentally studied timeintegrated and spectrally resolved FWM signals from the

I. INTRODUCTION

The intrinsic exciton transitions in hexagonal semiconductors show interesting properties. The optical properties
near the band gap are determined by three exciton species.
They arise from the threefold splitting of the highest valence
bands due to the crystal-field and spin-orbit interactions. The
three valence bands are called A-, B-, and C-valence bands
and the resultant excitons are A, B, and C excitons. For typical hexagonal semiconductors such as CdS and GaN, it is
known that the order of the valence bands is A- 9 , B- 7 ,
and C- 7 . 15
In zinc oxide ZnO, there is a consensus that the
C-valence band has 7 symmetry. The valence bands with
symmetry of 7 gives rise to a term linear in k in the dispersion relation for wave vector perpendicular to the c axis, i.e.,
kc. The k-linear term is known to lead to an extra dispersion curve in the excitonic polariton dispersion curve accorda

Electronic mail: kouji@moegi.waseda.jp

Present address: Institute of Industrial Science, University of Tokyo, Meguro, Tokyo 153-8505, Japan.
c
Also at: Integrative Bioscience and Biomedical Engineering, Graduate
School of Science and Engineering, Waseda University, Shinjuku,
Tokyo 169-8555, Japan; electronic mail: tkyksota@waseda.jp
d
Also at: CREST, Japan Science and Technology Agency.
e
Also at: NICP, ERATO, Japan Science and Technology Agency JST,
2-4-6 Fujimi, Chiyodaku, Tokyo 102-0071, Japan; Photodynamics Research Center, RIKEN Institute of Physical and Chemical Research,
Aoba, Sendai, Miyagi 980-0845, Japan.
b

0021-8979/2004/95(10)/5498/4/$22.00

5498

2004 American Institute of Physics

Downloaded 20 Feb 2006 to 130.158.130.96. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

J. Appl. Phys., Vol. 95, No. 10, 15 May 2004

Hazu et al.

5499

C-exciton manifold in a bulk ZnO. FWM signals were measured from a 110 surface. From the dependence of FWM
signals on the excitation wavelength and the spectral decomposition analysis of them, we have found that the coupling
between the 1 state and 5L state leads to the three-branch
excitonic polariton dispersion for kc and E c but the oscillator strength transfer is small. The energy separation between the lower polariton branch and the extra one due to the
k-linear term has been found to be 3.5 meV. The binding
energy of the CC biexciton, XX CC , is found to be 1.4 meV.
II. EXPERIMENT

The sample studied herein was a c-face bulk ZnO crystal

grown by the seeded vapor transport method. The sample
size was 2 mm3 mm500 m. Sample characterization
by means of linear and nonlinear spectroscopy can be found
elsewhere.12,14 A frequency-doubled, mode-locked Ti:sapphire laser was used as an excitation source in FWM experiments. The fundamental laser produced pulses of 80 fs at
repetition rate of 80 MHz. The frequency-doubled output
was divided into two beams with equal intensity. These
pulses with wave vectors k1 and k2 , respectively, were superimposed on the sample surface at angle of 6. The FWM
was performed in the 2k2 k1 backward geometry using a
degenerate two-pulse configuration. The delay time 12 between the two incident pulses was defined to be positive
when the k1 pulse precedes the k2 pulse. FWM signals were
detected, time integrated and spectrally resolved, by a combination of a 0.35 m focal length monochromator with a
2400 grooves/mm grating and a multichannel charge coupled
device detector as a function of 12 . The spectral resolution
was about 0.45 meV. According to circumstances, signals
were recorded phase sensitively by a pin photodiode. The
excited exciton density was of the order of 1016 cm3 , which
was much smaller than the Mott density (51018 cm3 ).
Throughout the measurements, the sample was held in a
closed-cycle helium cryostat and its temperature was maintained to be T10 K. Signals were measured from a 110
cleaved surface, to satisfy the condition kic (i1,2). A
neutral density filter was used in recording signals because
the signal intensity due to the 1 exciton was very strong.
III. RESULTS AND DISCUSSION

Figure 1 shows the polarization dependence of FWM

signals in the time domain for kic (i1,2). The signals
have been measured by a pin photodiode with the central
excitation photon energy of 3.4181 eV. Polarization conditions are as follows. Solid curve labeled as 0:
(E1 c,E2 c), dot-dashed curve labeled as 0:
(E1c,E2c), dashed curve labeled as 90:
(E1 c,E2c), and dotted curve labeled as 90:
(E1c,E2 c). Note that the dashed curve and dot-dashed
curve are multiplied by a factor of 50 and 10, respectively.
The two data for 0 directly demonstrate that excitons
with 1 and 5 symmetry are involved in the C-exciton
manifold as predicted by group theory and difference of their
oscillator strength is very large. Figure 1 also shows the following interesting feature. For the case of 90, FWM

FIG. 1. Polarization dependence of FWM signals in the time domain at T

10 K. The central photon enrgy of excitation is 3.4181 eV. The wave
vectors k1 and k2 are perpendicular to the c axis. denotes an angle between E1 and E2 , i.e., the polarization vectors of the k1 and k2 pulses. Even
in the case of 90, FWM signals remain finite. The dashed curve and the
dot-dashed curve are multiplied by a factor of 50 and 10, respectively.

signals remain finite within experimental accuracy. Thus, we

may conclude that 1 and 5 excitons interact each other to
produce FWM signals though they are independently excited
with a time lag and they have extremely different oscillator
strength.
Figure 2 shows FWM spectra solid curves recorded for
colinear polarization condition of E1 c, E2 c, and 12
0.1 ps in varying the excitation wavelength. The polarization condition ensures that the 1 exciton is primarily
excited. The central photon energy is a 3.4234, b 3.4164,
and c 3.4136 eV. In panels ac, the relatively broad
FWM bands appear in the photon energy region where the
reflection spectrum polarized for kc and E c11,12 show a
large reflection structure assigned as the 1 exciton. When
the two-branch excitonic polariton model is applied as usual
to interpret the observed bands, the number of FWM emission bands appearing in the region ranging from the transverse (T) to the longitudinal (L) 1 -exciton resonance energies is only 2. However, the decomposition procedure with
the use of Gaussian line shape functions has indicated that
the FWM signals ac cannot be reasonably reproduced in
using two components alone. At least four components are
required to reproduce them consistently. Obtained components labeled as c 1 c 4 are shown by dashed curves in panels
ac. The central energy of each band is tabulated in
Table I.
Here we would like to comment on the decomposition
procedure. FWM emission itself is a coherent emission.
However, there exist inevitable factors making the emission
spectra broaden such as inhomogeneous exciton linewidth,
the corresponding broadening of excitonexciton scattering
continuum state, and so on. Taking account of these effects,
it is reasonable to use the Gaussian line shape functions for
reproducing the observed spectra.
The following is a possible reason why not two but four
components are required to reproduce the FWM bands. The
5 representation is two dimensional. For kc, which corre-

Downloaded 20 Feb 2006 to 130.158.130.96. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

5500

Hazu et al.

J. Appl. Phys., Vol. 95, No. 10, 15 May 2004

TABLE I. Peak energies of decomposed bands and their possible origins.

Band peak energy in eV transitions through which the corresponding
FWM emission band arises
c1
c2
c3
c4

FIG. 2. FWM spectra solid curves in the C exciton region recorded at T

10 K for colinear polarization. The central photon energy of excitation is
a 3.4234, b 3.4164, and c 3.4136 eV. Dashed curves represent possible
decomposed components and dotted curves denote fitted curves to the experimental data. With respect to origins of them, see the text and Table I.

sponds to the experimental condition, the 5 -state splits into

the longitudinal and transverse states, 5L and 5T , due to
long-range Coulomb interaction.6 The polarized reflectance
spectra12 and the theoretical fitting procedure to them15 indicate that the oscillator strength of the 5 state is so small that
the small L T splitting energy is resulted. Thus, the energy
separation between the 5L state and the 1 state at k0 is
considered to be small in comparison with the off-diagonal
matrix element linear in k . 6 8 Consequently, the k-linear
term mixes the 5L and 1 states and the mixing transfers a
part of the oscillator strength of the 1 state to the 5L state,
to lead a three-branch excitonic polariton dispersion for kc

3.4226
3.4261
3.4302
3.4327

* X LP , XX EPLP
* X EP
X LPG, XX LPLP
* X LP , XX EPEP
* X EP , XX EPUP
* X UP
X EPG, XX LPEP
* X LP
XX EPEP
* X EP , XX UPUP
* X UP
X UPG, XX EPUP

and E c. Since the oscillator strength of the 1 state is very

large, the extra branch may contribute to the FWM signal
generation process. However, an anomalous structure was
not found in the polarized reflectance spectrum.1012 This
may be explained as follows: The transferred oscillator
strength is not so large because of the small but finite energy
separation between the 5L and 1 states. Note that both
hybrid states consisting of the 5T and 2 states for kc and
Ec are considered to have equal oscillator strength in the
case of CdS.8
We try to assign the possible origins of each FWM emission band obtained through the decomposition procedure.
Let X LP , X EP , and X UP be, respectively, polaritons on the
lower, extra, and upper polariton branches. Bearing it in
mind that two-photon coherence excited by the two-photon
absorption plays an important role particularly for 120,
the bands labeled as c 1 , c 2 , c 3 , and c 4 may be assigned as
bands arising from transitions summarized in Table I, where
* denotes a two-pair
G represents the ground state and XX
scattering continuum consisting of X and X . Note that the
relation of 2E X EPE X LPE X UP holds, see Table I, and this
ensures the earlier assignment. There exist another transitions except for those summarized in Table I. This gives a
reason why there remains discrepancy between the experimental data and theoretical fittings dotted curves. The energy separation between the lower polariton branch and the
extra one has been found to be 3.5 meV from Table I. The
L T splitting of the 1 -state due to the light and exciton
interaction may be estimated from the energy separation between the c 1 and c 4 bands to be 10.1 meV. The full width at
half maximum of four decomposed bands is ranging from
6.0 to 7.5 meV at most. Thus, it may be concluded that there
exists the signature of polariton in the observed FWM spectra. This guarantees our discussion.
In tuning the excitation laser wavelength to offresonance condition the central photon energy of 3.4086
eV, the FWM spectrum shown by the solid curve in Fig. 3
has been obtained at the same delay time. The spectrum
should be compared with that shown in Fig. 2c, dotted
curve labeled as c, in the inset. Note that the spectrum
drawn by the solid curve in the inset is enlarged by 25 times.
The emission peak at 3.4212 eV disappears for 120. This
allows us to assign it as the emission due to the transition
from XX CC level to X LP level. The biexciton binding energy
is estimated as 1.4 meV from the energy separation between
* X LP (c 1 band.
the transition of XX CC X LP and XX LPLP
It is natural to consider that transitions via XX CC contribute in part to generate the FWM spectrum shown in Fig. 3.
To identify the relevant transitions, the spectral decomposi-

Downloaded 20 Feb 2006 to 130.158.130.96. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp

Hazu et al.

J. Appl. Phys., Vol. 95, No. 10, 15 May 2004

5501

IV. SUMMARY

We have experimentally studied the impact of the

k-linear term on the C-exciton manifold in ZnO by means of
time-integrated and spectrally resolved FWM. The spectral
decomposition analysis has led the following results: The
coupling between the 1 state and 5L state leads the threebranch excitonic polariton dispersion for kc and E c but
the transfer of the oscillator strength from the former to the
latter is small. The energy separation between the lower polariton branch and the extra one due to the k-linear term is
3.5 meV. The binding energy of XX CC is 1.4 meV. We believe that resonant Brillouin and/or Raman scattering is a
suitable experimental method to confirm the three-mode nature of the polariton dispersion.
FIG. 3. FWM spectrum solid curve recorded at T10 K for colinear
polarization. The central photon energy of excitation is 3.4086 eV. Dashed
curves are possible decomposed components and the dotted curve denotes a
fiiting curve to the experimental data. The inset compares the present spectrum with that shown in Fig. 2c dotted curve, to demonstrate that the
component labeled m 1 should be clearly distinguished from the component
labeled c 1 . Note that the spectrum drawn by the solid curve is enlarged by
25 times. Origins of the components labeled m 1 m 4 are summarized in
Table II.

ACKNOWLEDGMENT

The work at the University of Tsukuba was supported in

part by the 21st Century COE Center of Excellence program under MEXT Ministry of Education, Culture, Sports,
Science and Technology, Promotion of Creative Interdisciplinary Materials Science for Novel Functions.
1

tion has been carried out. The result is shown by dashed

curves in Fig. 3. Additional four components labeled as
m 1 m 4 except for the components labeled as c 1 c 3 are necessary to reproduce the measured spectrum. The peak energy
and possible origin of the additional components are summarized in Table II. The band m 1 (m 4 ) is sometimes called as
the M T (M L ) band or line.16,17 The small intensity of the
band m 2 is consistent with the earlier-mentioned interpretation that the transferred oscillator strength from the 1 state
to the 5L state is small.
TABLE II. Peak energies of additional decomposed bands and their possible
origins.
Band peak energy in eV transitions through which the corresponding
FWM emission band arises
m1
m2
m3
m4

3.4212
3.4180
3.4125
3.4111

X CCX LP
* X UP , XX LPLP
* X EP , XX CCX EP , XX LPEP
* X UP
XX EPEP
* X UP
XX LPLP
X CCX UP

For a review, see, for example, B. Honerlage, R. Levy, J. B. Grun, C.

Klingshirn, and K. Bohnert, Phys. Rep. 124, 161 1985.
2
For a review, see, for example, S. Strite and H. Morkc, J. Vac. Sci. Technol. B 10, 1237 1992.
3
S. Chichibu, T. Azuhata, T. Sota, and S. Nakamura, J. Appl. Phys. 79,
2784 1996.
4
S. Chichibu, A. Shikanai, T. Azuhata, T. Sota, A. Kuramata, K. Horino,
and S. Nakamura, Appl. Phys. Lett. 68, 3766 1996.
5
A. Shikanai, T. Azuhata, T. Sota, S. Chichibu, A. Kuramata, K. Horino,
and S. Nakamura, J. Appl. Phys. 81, 417 1997.
6
G. D. Mahan and J. J. Hopfield, Phys. Rev. 135, A428 1964.
7
K. Cho, Phys. Rev. B 14, 4463 1976.
8
E. S. Koteles and G. Winterlling, Phys. Rev. Lett. 44, 984 1980.
9
Y. S. Park, C. W. Litton, T. C. Collins, and D. C. Reynolds, Phys. Rev.
143, 512 1966.
10
J. Lagois, Phys. Rev. B 23, 5511 1981.
11
D. C. Reynolds, D. C. Look, B. Jogai, C. W. Litton, G. Cantwell, and W.
C. Harsch, Phys. Rev. B 60, 2340 1999.
12
S. F. Chichibu, T. Sota, G. Cantwell, D. B. Eason, and C. W. Litton, J.
Appl. Phys. 93, 756 2003.
13
D. Weber, W. Petri, U. Woggon, C. F. Klingshirn, S. Shenel, and E. O.
Gobel, Phys. Rev. B 55, 12848 1997.
14
K. Hazu et al., Phys. Rev. B 68, 033205 2003.
15
K. Torii, T. Sota, and S. F. Chichibu unpublished.
16
H. Akiyama, T. Kuga, M. Matsuoka, and M. Kuwata-Gonokami, Phys.
Rev. B 42, 5621 1990.
17
For example, see, E. Tokunaga, K. Kurihara, M. Baba, Y. Masumoto, and
M. Matsuoka, Phys. Rev. B 64, 045209 2001, and references therein.

Downloaded 20 Feb 2006 to 130.158.130.96. Redistribution subject to AIP license or copyright, see http://jap.aip.org/jap/copyright.jsp