S. Tsujino et al.: Quantum Control of Electron Transfer
391
phys. stat. sol. (b) 221, 391 (2000) Subject classification: 73.20.Dx; 78.30.Fs; 78.47.+p; 78.55.Cr; 78.66.Fd; S7.12
Quantum Control of Electron Transfer S. Tsujino1 (a), M. RuÈfenacht (a), P. Miranda (b), S. J. Allen (a), P. Tamborenea (a), W. Schoenfeld (c), G. Herold (d), G. Lupke (d), T. Lundstrom (c), P. Petroff (c), H. Metiu (a), and D. Moses (b) (a) Quantum Institute, University of California, Santa Barbara, CA 93106, USA (b) Institute for Polymers and Organic Solids, University of California, Santa Barbara, CA 93106, USA (c) Materials Department, University of California, Santa Barbara, CA 93106, USA (d) Free Electron Laser Center, Vanderbilt University, Nashville, TN 37235, USA (Received April 10, 2000) We explore electron transfer in double quantum well structures induced by femtosecond mid-infrared intersubband excitation. Spatial transfer of electrons from one quantum well to its hole filled neighbor is detected by recombination luminescence. The process results in upconversion of the mid-infrared exciting light to near-infrared luminescence. Two mid-infrared pulses with variable time delay allow us to display the field and intensity autocorrelation function for the upconverted signal and measure the electron transfer dynamics. Electron transfer between two GaAs quantum wells separated by 300 nm can be saturated and the intensity autocorrelation function exhibits a slow 18 ps recovery. Transfer between wells separated by only 25 nm is coherently controlled by the phase of the two collinear infrared pulses.
Coherent control of charge excitations is potentially important for quantum information storage and processing. Unlike spins, dephasing of charge excitations can be very short and femtosecond pump and probe is required to explore phase coherent processes. In semiconductor nanostructures, coherent creation and destruction of excitons by femtosecond light pulses [1], coherent control of the photocurrent [2] and charge oscillations which arise from time-varying quantum interference between levels [3, 4] have been demonstrated. Here we optically inject and spatially separate electrons and holes, storing them for relatively long times in charge transfer double quantum wells (CTDQWs) (Fig. 1a) then explore the dynamics of intersubband excitations and charge transfer between these double quantum well structures by femtosecond mid-infrared (MIR) autocorrelation spectroscopy. A CTDQW is an uncoupled double quantum well separated by a low energy middle barrier and surrounded by high energy external barriers. The possibility of coherent transfer of electrons by femtosecond intersubband excitations in CTDQWs has recently been proposed [5, 6]. Further it has been previously shown that spatially separated and long lifetime electron±hole pairs can be created by visible light excitation, and electrons transferred from one QW to the other by MIR intersubband excitation. The transfer is detected by subsequent near-infrared (NIR) upconverted emission [7]. 1
) Corresponding author: e-mail:
[email protected]
392
S. Tsujino et al.
We use the upconversion process to measure electron transfer and dephasing of intersubband excitation in two different CTDQW structures, a fat (300 nm) CTDQW and a narrow (25 nm) CTDQW. The excited states of the fat CTDQW are well described as nearly free classical particles; the quantized states are very closely spaced. In the narrow QW the excited states are quantized. The samples are prepared by molecular beam epitaxy on (100) semi-insulating GaAs substrate. The fat CTDQW is, from top to bottom: a 15 nm thick GaAs cap layer, 350 nm thick Alx Ga1 x As barriers with x 0:43, 5.9 nm thick GaAs QW, 300 nm thick Alx Ga1 x As with x 0:23, 5.9 nm thick GaAs QW, 200 nm thick Alx Ga1 x As barriers (x 0:43) and 300 nm thick n -GaAs contact layer. The narrow GaAs CTDQWs are designed to allow participation by only one subband; the sample consists of the following, from top to bottom: 20 nm thick GaAs cap layer, 100 nm thick AlGaAs superlattice (SL) barrier, 16 periods of CTDQWs, a 102 nm thick SL barrier and 50 nm thick n -GaAs layer. The structure of each CTDQW is 8 nm thick GaAs, 25 nm thick Alx Ga1 x As (x 0:2) and 6 nm thick GaAs. The CTDQWs are each surrounded and separated from each other by a 17.8 nm thick SL barrier. The SL barrier is composed of 3 nm thick AlAs and 0.7 nm thick GaAs layers. The samples are processed to 2 mm square mesa diodes with an Ohmic contact to the n -GaAs and a semi-transparent Schottky electrode at the surface. The edge of the sample is polished at 45 degree to couple well with the MIR radiation. Samples are cooled to 4.5 K, and dc bias or pulsed bias is applied to the sample. We first excite the sample by visible light (Ar laser or HeNe laser) with a 250 ms duration to supply spatially separated electron±hole pairs. At a time, TIR , after the visible light excitation is terminated, we irradiate the device with the TM-polarized MIR light perpendicular to a 45 degree facet. NIR luminescence through the semi-transparent Schottky electrode is collected, dispersed through a monochromator and detected by a photomultiplier. Nearly Fourier transform limited 200 fs MIR pulses originate from an optical parametric amplifier (OPA) and pass through a Michelson interferometer. In this way we create collinear double pulses with variable time delay and of equal intensity. For single pulse experiment, only one of the pulses is used to excite the sample. Fat 300 nm CTDQW Figure 1b shows the luminescence from the QW in the CTDQW as a function of time. The sample is biased at 0V dc. When the Ar laser light is turned on, the created electron±hole pairs are separated by the external electric field Fex (< 0, including the built-in field), and relax to different GaAs layers: electrons to QWR in the substrate side, and holes to QWL in the surface side of the CTDQW, until the electric field Fe h
> 0 created by trapped electrons and holes compensates the external field, Fex . While the light excites the sample, photoluminescence (PL) is observed from both GaAs QWs in the CTDQW structure. 350 ns after Ar laser is terminated, the MIR pulse with photon energy 142.5 meV irradiates the sample, and induces the interband luminescence from the hole-rich QWL by transferring electrons 300 nm from QWR to QWL [7]. As expected from the symmetry of the sample (inset of Fig. 1b), the spectrum of the upconverted luminescence is similar to the PL spectrum under steady Ar laser pump. By changing MIR photon energy, we found upconversion is induced by intersubband excitation centered at 147 meV with the full width at the half maximum 45 meV. The
Quantum Control of Electron Transfer
393 Fig. 1. a) Sample structure. b) Photoluminescence (PL) and mid-infrared light induced upconverted luminescence of fat CTDQW sample at dc 0V as a function of time. The inset shows the comparison between PL spectrum and the upconverted luminescence
breadth of the spectrum shows that electron in the GaAs QW is excited to the quasi-continuum of the fat CTDQW. Efficient transfer across 300 nm is possible when Fe h cancels Fex making the internal field F ( Fe h Fex ) small. A slight decrease of jFe h j leads to an increasing field degrading the transfer efficiency. Pulse biasing is very effective in increasing the upconversion responsivity
Fig. 2. Upconversion responsivity of fat CTDQW sample as a function of the pulse height VH when the sample is pulse biased at various delay conditions
394
S. Tsujino et al.
after a long delay TIR. The pulse sequence is as follows: First we apply voltage VL during the visible light excitation. Next, the bias is switched to VH at time TV after the visible light is stopped. Finally, infrared light excites the sample while the bias is VH. Figure 2 shows the upconversion responsivity measured as a function of VH for several delays. VL is fixed at 3:0 V. As VH is increased from 3 V, the responsivity slowly rises then suddenly increases by more than two orders of magnitudes (Fig. 3). At this bias the electron transfer efficiency to the hole QW increases while keeping the stored carrier concentration unchanged. We interpret the threshold at 2:3 V to the voltage where the sign of F is changed. Because of the large electron±hole separation, the lifetime te h of the electrons and holes is long. te h = 417 ms is found from TIR dependence of upconversion intensity at
Fig. 3. Upconversion intensity of fat QW sample when the sample is excited by collinear and equal intensity double infrared light pulse as a function of the time delay between pulses. The inset is the interferogram of the excitation light measured by an energy detector (upper) and the interferogram of the upconversion signal (lower) for small time delay
Quantum Control of Electron Transfer
395
dc 0V. However, since the upconversion is sensitive to the small changes in F, the dc bias responsivity decreases by a factor 103 at TIR 750 ms. Figure 2 shows that the decrease of the maximum upconversion responsivity is increased by a factor 10 at TIR 750 ms when the sample is pulse biased, showing that pulsed bias is important for large TIR. It is interesting that the upconversion signal is still observed at a dc bias and for VH smaller than the threshold voltage. This indicates that the electrons excited from QWR may be launched toward QWL as a moving wavepacket [5], but then redistribute over the 300 nm of the CTDQW. Under these conditions the relaxation time of the excited electron can be slow. To probe this effect, we performed double MIR pulse upconversion. The energy of the double pulse created by the Michelson interferometer and measured by a linear detector as a function of the time delay exhibits the expected interference within 200 fs around zero delay (upper figure of the inset in Fig. 3). The lower figure of the inset in Fig. 3 is the interferogram of the upconversion signal for biases VL 2:5 V and VH 2:4 V, and delay times TIR 6:2 ms and TV 5:2 ms. We found that at this bias condition, the upper half of the interferogram is flattened. This shows that the upconversion signal is highly saturated. Since the saturation is reduced by increasing VH (not shown), the saturation is ascribed to the drop of the transfer efficiency due to the dynamical decrease of F and caused by the distribution of the excited electron over the bottom of the fat CTDQW. We do not see a recovery of the saturation in 1 ps time scale. Instead, we found that the saturation relaxed exponentially with an 18 ps time constant. The observed long relaxation time is consistent to the previously reported slow relaxation of electrons excited into barrier layers in modulation doped QWs [8, 9]. Narrow 25 nm CTDQW Here the intersubband transition occurs between quantized subbands. The same sample was used for upconversion experiment using CO2 laser [7], and showed that it has an excitation energy at 117 meV and increases to 124 meV when negative bias is increased. We found that photon energy at 147.5 meV can also efficiently induce upconversion, probably via higher excited levels. Figure 4 shows the result of a double pulse experiment. Figure 4a is the correlation of the infrared light itself (same as Fig. 3a). Figures 4b and c show the upconversion interferogram: Fig. 4b is measured with HeNe laser (8 mWcm 2 ) excitation at 1:5 V dc biased condition, and Fig. 4c is measured with Ar laser (27 mWcm 2 ) at 2:0 V dc. An extended interference oscillation is observed for delay time larger than 0.2 ps, which is not seen in the interferogram of the infrared light itself (Fig. 4a) or in that of the fat CTDQW sample. Since the overlap between two pulses is negligible for longer delay than 0.2 ps, the extended interference shows that the intersubband excitation, created by the first pulse, is coherently enhanced or destroyed by the second pulse, depending on the relative phase. The decay with delay measures the dephasing rate. In Fig. 4b, the dephasing time is estimated to be 200 fs [6], similar to the previously reported value [10]. Figure 4c shows an about a factor 2 longer time. In conclusion, we have explored the dynamics of electron transfer by intersubband excitation using upconversion. We found saturation and slow recovery (18 ps) of the transfer in fat CTDQW. In narrow CTDQW, the extended interferograms indicate that charge transfer can be coherently controlled out to 400 fs.
396
S. Tsujino et al.: Quantum Control of Electron Transfer Fig. 4. Upconversion responsivity of narrow QW sample in double infrared pulse excitation as a function of the time delay between pulses: a) infrared light, b) upconversion for HeNe laser excitation and dc 1:5 V, and c) upconversion for Ar laser excitation and dc 2:0 V
Acknowledgements One of the authors (S.T.) acknowledges the help of Dr. J. Kosub in the first experiment, and to Prof. A. Heeger for making his OPA system available. This work was supported by National Science Foundation Science and Technology Center for Quantized Electronic Structures (QUEST), DMR 91-20007, and the International Joint Research Grant Program of the New Energy and Industrial Technology Development Organization (NEDO), Japan.
References [1] A.P. Heberle, J.J. Baumberg, and K. Kohler, Phys. Rev. Lett. 75, 2598 (1995). [2] E. Dupont, P.B. Corkum, H.C. Liu, M. Buchanan, and Z.R. Wasilewski, Phys. Rev. Lett. 74, 3596 (1995). [3] I. Brener, P.C.M. Planken, M.C. Nuss, M.S.C. Luo, S.L. Chuang, L. Pfeiffer, D.E. Leaird, and A.M. Weiner, J. Opt. Soc. Amer. B 11, 2457 (1994). [4] A. Bonvalet, J. Nagle, V. Berger, A. Migus, J.-L. Martin, and M. Joffre, Phys. Rev. Lett. 76, 4392 (1996). [5] P.I. Tamborenea and H. Metiu, Phys. Lett. A 240, 265 (1998). [6] M. RuÈfenacht, S. Tsujino, S.J. Allen, W. Schoenfeld, and P. Petroff, phys. stat. sol. (b) 221, 407 (2000) (this issue). [7] M. RuÈfenacht, S. Tsujino, Y. Ohno, and H. Sakaki, Appl. Phys. Lett. 70, 1128 (1997). [8] J. Baier, I.M. Bayanov, U. Plodereder, and A. Seilmeier, Superlattices Microstruct. 19, 9 (1996). [9] S. Tsujino, C. Metzner, T. Noda, and H. Sakaki, phys. stat. sol. (b) 204, 162 (1997). [10] R.A. Kaindl, S. Lutgen, M. Woerner, T. Elsaesser, B. Nottelmann, V.M. Axt, V.M., T. Kuhn, A. Hase, and H. Kunzel, Phys. Rev. Lett. 80, 3575 (1998).
