Femtosecond luminescence measurements of the intersubband scattering rate in Al x Ga 1 x As/GaAs quantum wells under selective excitation M. Hartig, S. Haacke, and B. Deveaud Laboratorie d’Optoe ´lectronique Quantique, Institute for Micro and Optoelectronics, E ´ cole Polytechnique Fe ´de ´rale de Lausanne, CH-1015 Lausanne, Switzerland L. Rota Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, United Kingdom Received 6 May 1996 We have investigated the intersubband scattering of electrons in GaAs quantum wells using luminescence up-conversion with 100-fs resolution. The decay time of the n =2 electron-to-heavy-hole transition ( e ,hh) 2 depends both on the excess energy of the charge carriers and on the excitation density. A Monte Carlo simulation allows us to reproduce the experimental data with high accuracy. The intrinsic LO-phonon scatter- ing rate is found to be 2.010 12 s -1 for 80-meV subband separation. We show the wave-vector dependence and explain the density and excess energy dependence. S0163-18299652340-4 In quantum wells, LO-phonon emission is the most im- portant energy relaxation channel for electrons. The prob- ability of interaction is proportional to 1/| q | 2 q is the wave vector of the emitted phonon, 1 and to the overlap of the wave functions of the initial and final electron states. 2 If the intersubband separation is larger than the LO-phonon en- ergy, the intersubband scattering is expected to be subpicosecond. 1 If, on the contrary, the subband separation is smaller than 36 meV, the scattering time is due to acoustical phonons and carrier-carrier CCscattering. In this case and for carrier densities 10 11 cm -2 the decay rate should be greatly reduced. 3 Femtosecond dynamics in quantum wells have been in- vestigated using different techniques: Raman scattering, pump-probe, and photoluminescence PLup-conversion. 3–5 A coherent picture emerges for the cooling and capture rates. 6–8 For intersubband scattering, if a consensus now ex- ists on the theoretical description of phonons, 9,10 the experi- mental situation is not clear. Different techniques have de- livered different values: this is true both for the case of an intersubband spacing smaller than 36 meV Refs. 4, 11, and 12and when it is larger than 36 meV. 13–16 A resurgence of the interest in this subject has developed recently due to the demonstrated feasibility of infrared inter- subband quantum-well lasers. 17 An exact knowledge of the intersubband scattering processes is fundamental in order to understand how these lasers work. Our results are obtained by PL up-conversion, a technique which allows the direct and sensitive determination of the dynamics of the photoge- nerated charge carriers in time- and energy-resolved mea- surements down to very low densities, keeping a 100-fs time resolution. In this paper, we will show how the results of the experiment may be obscured by different mechanisms de- pending on the excitation density and on the photon energy, and we will give final results that are in very good agreement with a Monte Carlo simulation of the experiment including all the relevant scattering processes. A resolution of 100 fs is obtained by up-conversion in a LiIO 3 crystal. Two different laser sources have been used, a synchronously pumped dye laser at 600 nm and a Ti:sapphire laser at 705 nm. Multiple quantum-well samples 25 wells with different well widths have been studied, all with 70-Å barriers of either AlAs or Al 0.45 Ga 0.65 As. Our samples were grown by molecular beam epitaxy and the sample parameters were checked by a combination of PLE measurements and x-ray-diffraction studies. We concentrate on the results from similar samples, with well widths between 90 and 135 Å. The energy spacing between the first and the second confined electron levels is about 100 meV. Differences in confinement energy arise from the changes in the Al concentration, as well as in the width of the well. The excitation density was varied between 5 10 10 cm -2 and 10 12 cm -2 and the lattice temperature was kept at 77 K, allowing a reasonable thermal population in the second heavy-hole subband. Figure 1 shows energy resolved spectra of a 135-Å 92-Å well sample, with an Al x Ga 1 -x As AlAsbarrier, under dif- ferent excitation conditions density and wavelength. The spectra are dominated by the ( e ,hh 1 transition at about 1.55 eV. Since the PL intensity is proportional to the joint density of states and to the product of the Fermi distribution func- tions of electrons and holes, the spectra yield a variety of information concerning the carrier thermalization and the de- gree of thermal equilibrium between the different subbands. At this point it is important to emphasize that all dynamical processes concerning the hole population in the QW occur beyond the resolution of our experiments. The effective hole thermalization occurs on a time scale less than 50 fs, due to the small excess energy of the holes and the effectiveness of hole-hole scattering. 18 As the holes have a small excess en- ergy and as we keep the lattice temperature above 50 K, their cooling has a negligible influence on the spectra. The electron temperature manifests itself as an exponen- tial slope on the high-energy side of the luminescence. 19 If the electrons in subbands e 1 and e 2 are in thermal equilib- rium i.e., same chemical potential, the PL spectra should show a steplike increase at the onset of the ( e ,hh 2 transition by a factor of 2 corresponding to the underlying change in the density of states. Such a situation is observed in the up- per spectrum of Fig. 1a. On the contrary, nonequilibrium is evidenced in other spectra displayed in Fig. 1. The lower PHYSICAL REVIEW B 15 NOVEMBER 1996-II VOLUME 54, NUMBER 20 54 0163-1829/96/5420/142694/$10.00 14 269 © 1996 The American Physical Society