Femtosecond dynamics of exciton bleaching in bulk GaN
at room temperature
Yin-Chieh Huang, Gia-Wei Chern, Kung-Hsuan Lin, Jian-Chin Liang, and Chi-Kuang Suna)
Department of Electrical Engineering and Graduate Institute of Electro-Optical Engineering, National Taiwan University, Taipei, Taiwan, 10617 Republic of China
Chia-Chen Hsu
Department of Physics, National Chung Cheng University, Chia-Yi, Taiwan, Republic of China
Stacia Keller and Steven P. DenBaars
Department of Electrical and Computer Engineering, University of California, Santa Barbara, California 93106
共Received 16 October 2001; accepted for publication 13 May 2002兲
Femtosecond transient transmission pump–probe technique was used to investigate exciton dynamics in a nominally undoped GaN thin film at room temperature. An exciton ionization time of 100–250 femtoseconds was observed by the time-resolved pump–probe measurement. A comparison experiment with pre-excited free carriers also confirmed the observation of the exciton ionization process in bulk GaN. © 2002 American Institute of Physics.
关DOI: 10.1063/1.1491296兴
GaN-based semiconductors have received an ever-increasing interest for optoelectronic applications in the blue and ultraviolet spectral region.1It is well known that the near band-edge low-temperature optical properties in III–V semi-conductors are dominated by the many-body Coulomb electron–hole correlations, or the dynamics of excitons.2 With increasing temperature, the exciton linewidth broadens due to scattering processes with longitudinal optical 共LO兲 phonons and the excitons become thermally ionized.3Knox et al.4had observed an ionization time of⬃300 fs for reso-nantly excited excitons in GaAs quantum wells at room tem-perature. Wegener et al.5used a simple rate-equation model for exciton ionization to fit the measured pump–probe trans-mission trace for InGaAs quantum wells. An exciton ioniza-tion time of ⬃200 fs was measured.5 A similar procedure was adopted by Becker et al.6 in the measurement of CdZnTe quantum wells, and a fast ionization time of⬃110 fs was obtained. GaN with a hexagonal crystal structure has a direct band gap of 3.42 eV at room temperature. Free exci-tons in GaN are composed of three bands labeled as EXA, EXB, and EXC.7The binding energy (Eex) and the effective Bohr radius (aB) of the A exciton have been reported to be ⬃21 meV and ⬃29 Å.7 Because wide-gap semiconductors
have a large Eex value on the order of the thermal energy kBT at room temperature, excitonic resonances can not be easily ionized. However, exciton ionization has been con-vincingly observed up to room temperature through optical absorption measurement in high-quality GaN epilayers recently.8,9 By using transmission-type pump–probe measurement10,11around the excitonic transition energy, here we report the direct measurement of exciton ionization pro-cess in bulk wurzite GaN.
The 2.5m-thick GaN film used in our study was grown
by metalorganic chemical vapor deposition on c-plane sapphire.7 The crystal structure is wurtzite. In Fig. 1, the solid circles shows the measured low-intensity absorption spectrum of the nominally undoped GaN sample at room temperature. The A-exciton peak 共3.41 eV, 363 nm兲 and the Coulomb enhancement of the continuum absorption can still be clearly observed.
The time-resolved experiments were performed with a mode-locked Ti:Sapphire laser. The output laser pulses were frequency doubled in a beta barium borate crystal to reach the exciton energy. The frequency-doubled pulses had a pulsewidth of⬃150 fs at a wavelength around the A-exciton resonant peak. The full width half maximum of the output spectral bandwidth was 2.5 nm. One tenth of the UV beam was reflected by a beam splitter to be used as the probe beam while the rest passed through the beam splitter was used as the pump beam. We rotated the polarization of the pump
a兲Author to whom correspondence should be addressed; electronic mail:
FIG. 1. Absorption spectrum for an unintentionally doped 2.5m GaN thin film. A strong exciton resonance around 363 nm can be observed.
APPLIED PHYSICS LETTERS VOLUME 81, NUMBER 1 1 JULY 2002
85
beam by using a half-wave plate so that the polarizations of the pump 共s兲 and the probe 共p兲 beams were orthogonal to each other. The pump and probe beams were focused onto the bulk GaN sample by a focusing objective. After the sample, we measured the transmitted and reflected probe in-tensity by using a photodetector. A polarizer 共p polarized兲 was placed in front of the photodetector to remove any possible pump scattering light. We chopped the pump beam while the intensity of the reflected and transmitted probe beams were recorded as a function of the temporal delay between the pump and probe beams through a lock-in amplifier.
Figure 2 shows a typically measured probe transmission change ⌬T/T of the nominally undoped GaN thin film as a function of probe delay. The pump/probe photon energy was 3.42 eV共362.1 nm兲 corresponding to the exciton resonance. By measuring the incident pump power共6.9 mW, before the sample兲, reflected and transmitted pump powers, and the fo-cused pump/probe beam diameter共14m兲, an effective ab-sorption coefficient of 1.9⫻104 cm⫺1 was obtained. The corresponding average carrier density is 3.5⫻1017 cm⫺3 with a maximum carrier density of 1.8⫻1018cm⫺3 at the incident surface. The probe transmission changes consisted of four components corresponding to two-photon absorption 共TPA兲, exciton dynamics, and carrier dynamics 共described next兲. We have continuously varied our incident pump power from 6.9 down to 0.4 mW with the maximum surface carrier density ranged from 1.8⫻1018down to 1⫻1017 cm⫺3
共cor-responding average carrier density 2.4⫻1016cm⫺3兲. For the
whole experimental range, the TPA and exciton-dynamics components showed a linear behavior while the slow com-ponents showed a slight magnitude saturation under high ex-citation. We have also performed time-resolved reflection measurements to investigate the contribution of carrier-induced reflection changes to the measured transient trans-mission signals. It was found that at wavelengths shorter than 371 nm, the measured transmission changes are domi-nated by the carrier-induced absorption changes rather than reflection modulations. In order to analyze the measured transmission curve quantitatively, various phenomenological response functions were used in a convolution fitting proce-dure to extract the corresponding contributions and the
mea-sured time constants. The solid line in Fig. 2 is the probe transmission change obtained directly from our experiment. The dashed line is a convolution-fitting result consisting of four contributions of the corresponding carrier dynamics. A 240 fs hyperbolic-secant-squared profile was used as input pulse autocorrelation in the fitting. At zero-time delay, a negative transmission change with a width of the pump– probe autocorrelation was observed, which is usually as-cribed to the so-called ‘‘TPA peak.’’12 The TPA peak mag-nitude was linearly proportional to the pump intensity within our experimental range. Assuming the nonlinear absorption is small (I0Ⰶ␣, which is our case兲, where  is the TPA
coefficient,13I0is the incident light intensity at the input end,
and␣is the linear absorption coefficient.can then be cal-culated from the extracted transmission change. We found that the recovered TPA term exhibits a resonant feature around 362 nm corresponding to the A-exciton position, which can be attributed to the excitonic共Coulomb兲 enhance-ment of the TPA process 关shown in Fig. 3共a兲兴.14 This reso-nant TPA feature also confirms the existence of the hydro-genlike exciton line in the pure GaN film.
After the TPA peak, a fast exponential decay component 共positive兲 with time constant of 100–250 fs followed by a slower one共positive兲 of ⬃600–800 fs time constant can be observed and extracted from the pump–probe trace by using fitting processes. We choose 200 fs for a fast component in the specific fitting shown in Fig. 2. We have also included a negative step function in our convolution fitting to represent the band gap-renormalization response. As demonstrated in previously reported results in bulk InGaN,10,11 the slower time constant is attributed to the thermalization process of the photoexcited electron–hole pairs into lattice temperature. It is thus plausible to ascribe the fast decay time constant to the ionization of the resonantly created excitons into FIG. 2. Measured transient response of the unintentional doped GaN thin
film with a central wavelength of 362.1 nm. The solid line is the measured probe transmission change. The dashed line is a convolution fitting result with four fitting components:共i兲 fast exponential decay with time constant 100–250 fs,共ii兲 slow exponential decay with time constant 600–800 fs, 共iii兲 TPA, and共iv兲 negative step function.
FIG. 3. Magnitude of共a兲 the TPA coefficient, 共b兲 fast time constant compo-nent共open circle兲 and slow time constant component 共solid square兲 vs laser excitation wavelength. Time constants of 200 fs and 750 fs were chosen for the fast and slow components in the fitting. The excited average carrier density was kept fixed for different wavelength experiments.
electron–hole pairs by the interaction with LO phonons. The prolonged thermalization time constant of free carriers, com-pared to above band gap excitation, also confirms the ioniza-tion process. It is interesting to notice that we can observe this exciton ionization process even under our high excita-tion condiexcita-tion, as shown in Fig. 2, with a maximum surface photocarrier density close to the Mott-transition density.15 This is probably due to the fact that the measured signals are not just contributed from the near surface response, but from the whole sample. However, in a recent low-temperature共10 K兲 experiment, a remarkable persistence of the excitonic resonances in GaN at carrier densities well above the Mott density at early time delays has already been observed.16
Experiments on the nominally undoped GaN by tuning the laser wavelength around excitonic absorption peak from 361.3 to 365.7 nm共center wavelength兲 were also performed. Incident pump powers were varied from low 共0.4 mW兲 to high 共6.9 mW兲 excitation levels for all experimental wave-lengths. The resonance of the ionization component to the excitonic peak wavelength is shown in Fig. 3共b兲 共open circle兲, where we plot the magnitude of fast time constant component共fitting with 200 fs兲 versus the laser wavelength with a fixed photoexcited carrier density. However, the fol-lowing slower time constant component does not possess the resonant feature around the exciton peak as shown in Fig. 3共b兲 共solid square兲. This is because the thermalization of electrons and holes is mainly due to intraband carrier scat-tering and is not related to exciton dynamics.
We also performed a comparison experiment to study the dependence of the fast decay time constant on background carrier density. An additional pump beam was used to pro-duce a large amount of background carriers in the GaN sample before the transmission pump–probe measurement. These pre-excited background carriers 共lead time 5 ps兲 will influence the following exciton dynamics due to the
screen-ing effect. The correspondscreen-ing probe transmission changes as a function of time delay were shown in Fig. 4. The laser central wavelength was tuned to 364.8 nm around the exci-ton absorption peak. The retrieved transmission changes are listed in Table I. With a pre-excited carrier density of 1.1 ⫻1018 cm⫺3, the strength of the fast ionization component 共200 fs兲 decreases significantly with respect to other terms. Since the carrier density is close to Mott transition, excitons are unstable against the background electron–hole pairs. As a result, the initial formation of excitons is inhibited. This ex-plains the decreased strength of the fast time constant.
In conclusion, the exciton dynamics in bulk GaN thin film were investigated by UV femtosecond pump–probe measurement. We found that in addition to the thermalization process of free carriers, another fast relaxation process domi-nates the initial bleaching dynamics in bulk GaN. The corre-sponding strength of this fast time constant has a resonance around the excitonic absorption, and is attributed to the ion-ization of excitons into electron–hole pairs. Another com-parison experiment with pre-excited background carriers also confirms this interpretation.
This project is sponsored by National Science Council of Taiwan, ROC under Grant No. NSC 90-2112-M-002-051.
1S. Nakamura and G. Fasol, The Blue Laser Diode: GaN Based Light
Emitters and Lasers共Springer, Berlin, 1997兲.
2J. Shah, R. F. Leheny, and W. Wiegmann, Phys. Rev. B 16, 1577共1977兲. 3
J. P. Lowenau, F. M. Reich, and E. Gornik, Phys. Rev. B 51, 4159共1995兲.
4
W. H. Knox, R. L. Fork, M. C. Downer, D. A. B. Miller, D. S. Chemla, C. V. Shank, A. C. Gossard, and W. Wiegmann, Phys. Rev. Lett. 54, 1306
共1985兲.
5M. Wegener, I. Bar-Joseph, G. Sucha, M. N. Islam, N. Sauer, T. Y. Chang,
and D. S. Chemla, Phys. Rev. B 39, 12794共1989兲.
6P. C. Becker, D. Lee, A. M. Johnson, A. G. Prosser, R. D. Feldman, R. F.
Austin, and R. E. Behringer, Phys. Rev. Lett. 68, 1876共1992兲.
7J. F. Muth, J. H. Lee, I. K. Shmagin, R. M. Kolbas, H. C. Casey, Jr., B. P.
Keller, U. K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 71, 2572
共1997兲.
8A. J. Fischer, W. Shan, J. J. Song, Y. C. Chang, R. Horning, and B.
Goldenberg, Appl. Phys. Lett. 71, 1981共1997兲.
9S. Chichibu, T. Azuhata, T. Sota, and S. Nakamura, J. Appl. Phys. 79,
2784共1996兲.
10C.-K. Sun, F. Vallee, S. Keller, J. E. Bowers, and S. P. DenBaars, Appl.
Phys. Lett. 70, 2004共1997兲.
11C.-K. Sun, Y.-L. Huang, S. Keller, U. K. Mishra, and S. P. DenBaars,
Phys. Rev. B 59, 13535共1999兲.
12
K. L. Hall, J. Mark, E. P. Ippen, and G. Eisenstein, Appl. Phys. Lett. 56, 1740共1990兲.
13C.-K. Sun, J.-C. Liang, J.-C. Wang, F.-J. Kao, S. Keller, M. P. Mack, U.
Mishra, and S. P. DenBaars, Appl. Phys. Lett. 76, 439共2000兲.
14
A. Obeidat and J. Khurgin, J. Opt. Soc. Am. B 12, 1222共1995兲.
15
F. Binet, J. Y. Duboz, J. Off, and F. Scholz, Phys. Rev. B 60, 4715共1999兲.
16C. K. Choi, Y. H. Kwon, J. S. Krasinski, G. H. Park, G. Setlur, J. J. Song,
and Y. C. Chang, Phys. Rev. B 63, 115315共2001兲.
FIG. 4. Measured transient transmission response of the unintentionally doped GaN thin film with and without pre-excited carriers. The laser exci-tation wavelength was 364.8 nm. The solid lines are the measured probe transmission changes. The dotted lines are convolution fittings with four fitting components accounted for fast exponential decay of time constant 200 fs, slow exponential decay of time constant 750 fs, TPA peak, and negative step function. The corresponding retrieved transmission changes are listed in Table I.
TABLE I. Retrieved relative probe transmission changes with and without a pre-excited background free carriers.
⌬T/T Without pre-excitation With pre-excitation
TPA ⫺0.034 ⫺0.018
200 fs 0.049 0.009
750 fs 0.043 0.030
87
