Permissions for Reuse

Recent developments in semiconductor-based spintronics and quantum computation^1,2 have generated intense interest in examining the spin dynamics in a wide range of semiconductor material systems at telecom wavelengths^3,4,5,6 due to their potential applications in spin-dependent optical modulators and switches for optical telecommunications.^6,7,8,9 The optical wavelengths of 1.31 and 1.55  µm allow a minimum dispersion and signal loss in standard silica-based fiber-optic networks. Spin-dependent all-optical switching in semiconductor quantum well etalons⁸ and spin-dependent ultrafast optical gain modulation in microcavity lasers have been recently demonstrated.⁹ In addition, differential quantum efficiency as high as 64% is attainable for all-epitaxially grown telecom-wavelength vertical cavity surface emitting lasers (VCSELs) with InP lattice-matched AlInGaAs alloys for quantum well active regions and AlGaAsSb alloys for highly reflective distributed Bragg reflectors operating above room temperature.^10,11 The possibility of integrating laser sources, optical modulators, and optical switches on a single photonic integrated chip with added functionalities based on spin degrees of freedom makes AlInGaAs alloys ideal candidates for spintronics.

In this letter, we present room temperature measurements of the electron spin coherence times and effective g factors in AlInGaAs multiple quantum wells, which are optimized active regions of VCSELs with wavelengths in the range of 1.26–1.53  µm designed for optical telecommunication applications.^10,11 We characterize the samples using photoluminescence and measure the electron spin dynamics using pump-probe optical spectroscopy. The electron effective g factors and spin coherence times are found to vary with aluminum alloy concentration from –2.3 to –1.1 and from 50  to  100  ps, respectively.

Our samples are comprised of a set of quaternary alloy AlInGaAs multiple quantum wells grown using digital alloy molecular beam epitaxy. These samples are composed of digital alloys of ternaries (Ga_0.47In_0.53As and Al_0.48In_0.52As, referred to as GaInAs and AlInAs, respectively) lattice matched to InP.¹⁰ The sample structure is of the form AlInAs/(AlIn)_r(GaIn)_1–rAs/5×[(AlIn)_y(GaIn)_1–yAs/(AlIn)_x  (GaIn)_1–x  As]  /  (AlIn)_y  (GaIn)_1–y  As  /  (AlIn)_r  (GaIn)_1–rAs/AlInAs/(AlIn)_s(GaIn)_1–sAs/AlInAs/InP substrate with r=0.75 and s=0.82, shown schematically in Fig. 1(a). The average aluminum concentrations of the quantum wells and barriers are _w=0.48x and _b=0.48y, respectively. The typical values of compressive strain in the quantum wells and tensile strain in the barriers along the growth (in-plane) direction are ~1.5% (~0.5%) and ~0.6% (~0.2%), respectively,^10,11 consistent with results calculated using a generalized 14-band K·p envelope-function theory.^12,13 All samples are undoped. The photoluminescence (PL) spectra of these samples are excited nonresonantly with 2.5  mW of 800  nm light from a tunable mode-locked Ti:sapphire laser and measured using a liquid nitrogen cooled InGaAs photodiode array detector [Fig. 1(b)]. The PL peaks of the samples at room temperature occur at 1.53  µm (sample A), 1.34  µm (sample B), 1.30  µm (sample C), and 1.26  µm (sample D), and the corresponding energies are 0.811, 0.925, 0.954, and and 0.984  eV, respectively. The PL intensity decreases with increasing aluminum concentration _w due to Al incorporated impurities, such as oxygen, and lower confinement of the electron wave functions arising from lower conduction band offsets. The PL spectrum of sample C consists of two broad peaks (1.28 and 1.30  µm), and the 1.28  µm peak might be due to sample inhomogeneity. We calculate the effective band gaps of the quantum wells using a generalized 14-band K·p envelope-function theory,^12,13 and the calculated effective band gaps for samples A–D are 1.55, 1.35, 1.30, and 1.25  µm, respectively.

Figure 1.

Time-resolved Kerr rotation (TRKR), an optical pump-probe spectroscopic technique,¹⁴ is used to probe the electron spin dynamics. An optical parametric amplifier pumped by a regeneratively amplified Ti:sapphire mode-locked laser produces ~200  fs duration pump and probe pulses tunable from 1.1  to  1.65  µm with a repetition rate of 100  kHz, whose relative delay is adjusted by a mechanical delay line. The helicity of the pump beam for spin injection is modulated with a photoelastic modulator at 42  kHz and the linearly polarized probe beam for spin detection is mechanically chopped at a frequency of 800  Hz. The pump and probe beams are focused to a spot size of ~50  µm on the sample, which is mounted between the two poles of an electromagnet, which provides a magnetic field B up to 0.6  T and perpendicular to the direction of the pump and probe beams. The circularly polarized pump beam excites electron and hole spins which precess at the Larmor frequency around the transverse applied magnetic field. The typical pump and probe powers used in the measurements are 1 and 0.1  mW, respectively. The rotation of the linear polarization axis of the probe beam is proportional to the net electron spin polarization along the probe's normal incidence. Using a balanced photodiode bridge combined with a lock-in detection technique, the electron spin precession is measured with subpicosecond temporal resolution.¹⁴

In Fig. 2, we show representative TRKR scans for sample A (8% Al) and Sample D (18% Al) at room temperature with applied magnetic fields B=0 and B=0.6  T. The electron spin magnetization precesses in the plane perpendicular to the applied magnetic field, and the Kerr rotation angle as a function of time delay t can be described by an exponentially decaying cosine: (t)=₀  exp(–t/T)cos(2_Lt), where ₀ is the initial amplitude of the electron spin polarization, _L is the Larmor precession frequency _L=g^*µ_BB/h, where g^* is the effective in-plane electron g factor of electrons in the quantum wells, µ_B is the Bohr magneton, h is Planck's constant, and T is the electron spin coherence time. The data are fitted by adjusting the parameters ₀, T, and g^*. The lack of additional frequency components and the agreement of g^* with the expected value^15,16 for electrons indicate that hole spin precession is not observable.

Figure 2.

The magnetic field dependence of T at room temperature for all four samples is plotted in Fig. 3(a). All the samples exhibit weak magnetic field dependence. The corresponding Larmor frequencies _L of all samples have a linear magnetic field dependence [Fig. 3(b)], yielding |g^*|=2.31 (sample A), |g^*|=1.49 (sample B), |g^*|=1.47 (sample C), and |g^*|=1.07 (sample D). Although the sign of g^* cannot be determined from our measurements, the fact that bulk InAs has a large and negative g factor while bulk AlAs has a small positive g factor indicates that g^* is negative.^15,16 The electron effective g factor as a function of _w at room temperature is plotted in Fig. 4(a), showing that g^* increases with increasing _w consistent with the expected trends.^15,16

Figure 3.

Figure 4.

The electron spin coherence time T as a function of _w at room temperature with applied magnetic field B=0.6  T is displayed in Fig. 4(b). The measured electron spin coherence times are in the range from 50  to  100  ps. T increases with increasing _w except for sample C. The observed nonlinear _w dependence of T is due to the combined effects of orbital scattering and internal effective magnetic field in the quantum wells.^12,13

Electron spin relaxation near room temperature is dominated by the D'yakonov-Perel' mechanism,^13,17 and the spin relaxation rate is T²₀, where is the momentum-dependent precession frequency about the internal effective magnetic field of the quantum wells and ₀ is the orbital coherence time, which is proportional to the electron mobility µ. As _w increases, increases due to band structure effects on the effective internal magnetic field, causing T₂ to decrease with increasing _w. To illustrate these effects, we calculate the spin coherence times T₂ as a function of _w for samples A–D assuming a constant mobility for all samples using a generalized 14-band K·p envelope-function theory,^12,13,18 and the results are shown in Fig. 4(c). The calculated T₂ does not show a smooth trend with _w because also depends on _b and the digital alloy layered structure of the samples. However, the observed trends in T might be partly due to the nonlinear _w dependence of the electron mobility in these samples. As _w increases, ₀ decreases due to orbital scattering, causing T₂ to increase with increasing _w. Finally, imperfection in crystal growth may lead to structural inversion asymmetry in the quantum wells, which generates an additional internal effective magnetic field contributing to electron spin decoherence.¹²

We have measured the electron spin coherence times and electron spin precession frequencies in telecom-wavelength quaternary multiple quantum wells and observed the electron spin coherence times of ~100  ps at room temperature. Our measurements reveal that the electron effective g factor can be tuned by varying the aluminum alloy concentration. These results demonstrate the possibility of optical modulation using spin manipulation in this material system.

The authors wish to acknowledge the support of Intel and MARCO. One of the authors (N.P.S.) acknowledges the support of the Hertz Foundation.