Terahertz four-wave mixing spectroscopy for study of ultrafast dynamics in a semiconductor optical amplifier Jianhui Zhou, Namkyoo Park, Jay W. Dawson, and Kerry J. Vahala Department of&lied Physics, Mail Stop 128-95, California Institute of Technology, Pasadena, Calijornia YI ID Michael A. Newkirk and Barry 1. Miller A Tk T Bell Laboratories, Holmdel, New Jersey 07733 (Received 2 April 1993; accepted for publication 18 June 1993) Ultrafast dynamics in a 1Spm tensile-strained quantum-well optical amplifier has been studied by highly nondegenerate four-wave mixing at detuning frequencies up to 1.7 THz. Frequency response data indicate the presence of two ultrafast physical processes with characteristic relaxation lifetimes of 650 fs and < 100 fs. The longer time constant is believed to be associated with the dynamic carrier heating effect. This is in agreement with previous time-domain pump-probe measurements using ultrashort optical pulses. Understanding the ultrafast intraband dynamics of the range of several hundred microwatts. This is in con- semiconductor active layers is of considerable importance trast to time-domain pump-probe measurements, where since, in addition to purely fundamental considerations, short pulse peak power is usually in the range of several the physical processes involved influence the modulation hundred milliwatts or higher.“S5 The information revealed response and spectral properties of semiconductor lasers in TFWM measurements is therefore readily applicable to and can induce crosstalk between multiplexed signals in analysis of modulation and spectral properties of semicon- semiconductor optical amplifiers. ductor lasers, where small-signal data are often needed. Ilnt.il recently, time-domain pump-probe measurement Among other advantages of TFWM are the simplicity of using ultrashort optical pulse~,‘-~ pioneered by Ippen and the experimental setup and better accuracy in measure- co-workers,1’.-3 has been the only direct technique for study ment of lifetime data since short lifetimes are mapped to a of ultrafdst intraband dynamics. These measurements have wide frequency span. In addition to providing a frequency- revealed characteristic time constants of 550 and 200 fs in domain tool for investigation of intraband dynamics, 1.5 ~i-rmInGaAsP bulk optical amplifiers’ and 700 and 250 TFWM measurements provide signal spectra that give am- fs in 1.5 /ltm multiple quantum-well optical amplifiers.” The plifier crosstalk strength for widely separated optical chan- longer time constants were attributed to a heated carrier nels as well as efficiency for possible use of TWAs as para- distribution cooling back to an equilibrium temperature metric channel converters. and is referred to as dynamic carrier heating. The shorter The TWA used in this study was a tensile-strained time constants were recently found to be caused by a com- InGaAs/InGaAsP multiple quantum-well TW.A operating bination of effects, including delay in the carrier heating, at 1.5 pm. Details on the structure appear in Refs. 10 and and spectral hole burning.” In addition, measurements on 11. The specific device was operated at a 100 mA bias multiple quantum-well optical amplifiers by Weiss et aZ.” current where the TM mode small-signal gain was mea- have resolved time constants of 2-7 ps, which were attrib- sured to be 20 dB for the wavelengths investigated, and the uted to carrier diffusion in the barrier layer and identified gain ripple less than 0.1 dB. The pump, probe, and local as the carrier capture time. oscillators used in the measurements are three single- Highly nondegenerate four-wave mixing due to ultra- frequency, tunable, Er-doped fiber ring lasers.” These la- fast intraband dynamics in semiconductor gain media was sers, having output powers of about 1 mW, linewidths less first proposed by Agrawal.’ Recently, highly nondegener- than 4 kHz, *’ and low intensity noise levels, I-( provide ideal ate four-wave mixing measurements in semiconductor sources for the TFWM experiment. traveling-wave optical amplifiers (TWAs) have been used As shown in Fig. 1, the optical outputs from the pump as a frequency-domain technique for studies of ultrafast and probe lasers, having optical frequencies fp and fq, dynamics in semiconductors. To date, however, all the were combined using a 3 dB fiber coupler2 and then colli- measurements have been limited to a detuning range of mated and coupled into the TWA using microscope objec- approximately 500 GHz.~-~ Here, we report measurements tives. The optical power coupled into the TWA was about of four-wave mixing in a tensile-strained multiple 250 pc1-w.Two fiber polarization controllers were employed quantum-well TWA at detuning frequencies up to 1.7 to match the polarization state of input pump and probe to THz, which corresponds to an equivalent temporal resolu- the TWA TM mode. Two optical isolators providing 60 dB tion of 94 fs (r= 1/2?rf). To our knowledge, this repre- isolation were placed before and after the TWA to elimi- sents the first report of terahertz four-wave mixing nate feedback caused by the facets of the input and output (TFWM) spectroscopy in semiconductor gain media. fibers, and to also serve as polarization selectors. TFWM is a small-signal technique since semiconduc- During the measurements, the pump laser (f’) was tor dynamics is probed using optical powers typically in ftxed at 1534 nm, while the probe ( fq) was tuned to vary 1179 Appl. Phys. Lett. 63 (9), 30 August 1993 0003-6951/93/63(9)/1179/3/$6.00 @ 1993 American Institute of Physics 1179 Downloaded 20 Dec 2005 to 131.215.225.171. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp FIG. 1. Experimental setup for the TFWM measurement. Detuning Frequency (-GHz) the detuning frequency fP- fq. The four-wave mixing sig- nal at 2fp- fq was heterodyne detected by mixing the sig- FIG. 3. Normalized four-wave mixing signal power vs negative dctuning nal with the fiber-laser local oscillator and tuning the local frequency showing theoretical fits. oscillator near the signal frequency. The mixing frequency was maintained at 4.00 GHz throughout the measurements have extended the simplified formula to include the possi- to eliminate the frequency dependence of detection elec- bility of several physical processes, tronics. The detected photocurrent was amplified by a mi- crowave amplifier with 40 dB gain and then measured us- ing a spectrum analyzer. The detuning frequencies were determined using an opt.ical spectrum analyzer. The optical powers of pump, probe, and local oscillator were As.0 mea- where l&vM, EP, and I$ are amplitudes of the signal, sured, so that the measured four-wave mixing signal could pump, and probe fields, f IS the detuning frequency, and Tj be normalized. and cj are the lifetimes and the complex coupling coeffi- The four-wave mixing signal was measured for detun- cients of the various relaxation processes. Physically, the ing frequencies from - 1.7 to 1.5 THz, limited only by the real and imaginary parts of the coupling coefficient repre- tunability of the fiber lasers used in the experiment. The sent the strengths of the dynamic gain and index gratings signal level at the maximum detuning frequency was still created by the specific relaxation process. Therefore, the about 20 dB above the noise floor, owing to the narrow coupling coefficients can serve, in some cases, as a criterion linewidth and low intensity noise of the fiber lasers. The for identifying the responsible physical mechanisms. normalized signal power as a function of detuning fre- The dashed line in Figs. 2 and 3 represents the re- quency is illustrated in Figs. 2 and 3. sponse accounting for carrier density modulation (inter- In prior work at lower frequencies, we have been able band recombination) only. The deviation of the measured to model these effects using uncoupled rate equations char- signal data from this one time constant fit indicates the acterized by a lifetime and a complex coupling coefi- presence of the ultrafast relaxation mechanisms. In addi- cient.“‘” To study the new ultrahigh-frequency data we tion, the asymmetric nature of the signal spectrum is be- lieved to be caused by phase interferences which occur between various mechanisms present. Since the positive detuning data show constructive interferences, we esti- mated a second time constant (i.e., in addition to the in- -10 0 Bpsnmmtal oat.4 ci- - - , Trw ConrtanlF,l E - * mr comtant Fit terband recombination time constant) using the positive $ - 2 me Constant6, signal spectrum. However, a two time constant response is i: -20 a still insufficient to provide a good fit for the whole signal P spectrum as shown in the figures. This indicates the exist- .o -30 m ence of an even faster process. A third term having a life- P ,N time of < 100 fs was required to obtain a good fit. -40 : Using the three-term formula, excellent fits were ob- 2 tained for both positive and negative detuning data with -50 '. ‘\ the same parameter set (see Figs. 2 and 3:-3 time con- ‘l . -50:o , . a&- . ..( . ,A stant fit.). The fitting paramet.ers are as follows: cl 100 1000 10000 = 5.6%“.““, ~,=0,0642r”.~~, ~;=O.O113e”~~“, .rl =200 ps, Detuning Frequency (GHz) r,= 650 fs, and ~~= 50 fs. Here we have used a value of 3.6 for the linewidth enhancement factor. It is important to FIG. 2. Normalized four-wave mixing signal power vs positive detuning note that the fits are quite sensitive to the longer intraband frrquency showing theoretical fits. relaxation time constant 650 fs, but are relatively insensi- 1180 Appl. Phys. Lett., Vol. 63, No. 9, 30 August 1993 Zhou et a/. 1180 Downloaded 20 Dec 2005 to 131.215.225.171. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp tive to the shorter time constant, because the equivalent With improvements to the fiber lasers, the detuning fre- temporal resolution of this measurement system is still lim- quency could be further extended to provide even better ited.