Optically controlled spin-polarization memory effect on Mn delta-doped heterostructures

We investigated the dynamics of the interaction between spin-polarized photo-created carriers and Mn ions on InGaAs/GaAs: Mn structures. The carriers are confined in an InGaAs quantum well and the Mn ions come from a Mn delta-layer grown at the GaAs barrier close to the well. Even though the carriers and the Mn ions are spatially separated, the interaction between them is demonstrated by time-resolved spin-polarized photoluminescence measurements. Using a pre-pulse laser excitation with an opposite circular-polarization clearly reduces the polarization degree of the quantum-well emission for samples where a strong magnetic interaction is observed. The results demonstrate that the Mn ions act as a spin-memory that can be optically controlled by the polarization of the photocreated carriers. On the other hand, the spin-polarized Mn ions also affect the spin-polarization of the subsequently created carriers as observed by their spin relaxation time. These effects fade away with increasing time delays between the pulses as well as with increasing temperatures.

refer to the first and second set of samples as MN-and CMN-series, respectively. All samples were grown using a hybrid system combining metal-organic chemical vapor and pulsed-laser ablation depositions. First, an undoped GaAs buffer layer, the In 0.16 Ga 0.84 As QW (10 nm) and a GaAs spacer layer (d S ) were grown by MOCVD at a high temperatures (~600 °C). The precursors were trimethylgallium, trimethylindium and arsine. On the CMN-series, carbon tetrachloride doping was used to grow the δ C separated by a 10 nm GaAs layer from the InGaAs QW. On the second stage, we have used a Q-switched YAG: Nd laser ablation system with Mn and GaAs targets at temperatures T Mn for growing the Mn delta-doping layer and the GaAs capping layer (d C ), respectively. All the growth was performed in the same reactor. Further details of the growth can be found in ref. 13. A complete list of the growth parameters from the investigated samples is presented in Table 1.
Time-resolved photoluminescence (PL) measurements were performed using a fs Ti:Sa laser and a streak-camera system (time resolution ~50 ps). The laser wavelength was tuned for resonant QW excitation. The right-(σ + ) and left-(σ − ) circular-polarized components of the excitation beams and the optical emission were selected with appropriated optics. The circular polarization of each beam can be selected independently. The time delay Δt between the pulses from the two beams was controlled by changing the optical path of one of the beams. From now on we refer to the pulses of the beam that arrive a time Δt before the pulses of the other beam as the pre-pulses. The results presented here correspond to the condition where the pre-pulses are σ − polarized, and the following pulses from the second beam are σ + polarized. Measurements with opposite polarizations were also performed and gave equivalent results.
The degree of polarization of the PL emission is defined as: where I σ+/σ− is the intensity of the σ +/− emission component.   Table 1 presents the PL decay time (τ ) and the spin-relaxation time (τ s ) obtained by fitting the time-resolved PL transients using only one excitation beam by simple exponential decays. We point out that all samples present similar values of τ of order of 0.2 ns, which is consistent with the results obtained for similar QWs 15 and somehow larger than our experimental resolution. Therefore we are not very confident on doing a detailed analysis based on the small variations observed for τ . However, τ s ranges from less than 1 ns up to more than 2 ns, which we will discuss below. Figure 2 shows typical time-resolved measurements from sample MN1 (see Table 1) using two excitation beams. The excitation energy is ~30 meV above than the PL peak (1.383 eV) with an averaged power of 10 mW. We point out that during the laser pulse the excitation power is ~10 4 stronger than the measured averaged power as estimated considering the ratio between the pulse duration (~100 fs) and the laser repetition (~12 ns). The experiment was performed using a time delay of Δt = 500 ps between the two excitation beams as indicated by the diagram of Fig. 2. The streak camera images were obtained by measuring the σ + and σ − circularly-polarized components of the PL emission. The aim of our experiment is to use the pre-pulse to induce a particular spin-polarization of the Mn ions, and to use the following pulse after a time Δt, to probe the effects of the Mn polarization on the polarization degree of the PL emission from the QW. Figure 3a,b show the transients of σ − and σ + components of the integrated QW PL emission from sample MN1 using one and two excitation beams. With a single σ + excitation beam, the PL transients show an expected dominance of the σ + emission giving rise to an initial polarization degree of Pol ~50% immediately after the laser pulse, as shown in Fig. 3c. When the second excitation beam is turned on giving rise to σ − pre-pulses with Δt = 500 ps, the polarization of the PL emission is initially negative, around − 50%, as shown in Fig. 3c. Thus, when the σ + excitation pulse hits the structure there is a residual PL emission with σ − dominance. The σ + excitation pulse inverts this state, so that the PL polarization degree Pol changes from negative to positive. Figure 3c reveals a rather interesting effect. As pointed out, the magnitude of the polarization degree immediately after a laser pulse for a single beam excitation condition is ~50% for sample MN1. However, when the pre-pulse is applied, the polarization degree immediately after the σ + excitation pulse becomes surprisingly smaller (~25%). In fact, a correction to take into account the presence of the residual carriers from the pre-pulse is necessary. The corrected-polarization degree is also shown in Fig. 3c, which was calculated by subtracting the estimated PL intensity created by the pre-pulse considering a mono-exponential decay (Fig. 3b) from the PL intensity measured after the second pulse. As shown in Fig. 3c, by doing this correction the initial polarization degree immediately after the σ + excitation pulse becomes ~30%, which is still significantly smaller than the ~50% value from the measurements without a pre-pulse.

Results and Discussion
We define here a parameter, ΔPol, as the difference between the modulus of the initial polarization degree when the sample is excited with a single excitation beam (Pol 0 ) and the modulus of the initial polarization degree under the presence of a pre-pulse, considering the correction discussed above concerning the subtraction of the residual PL intensity generated by the pre-pulse. This parameter is directly shown in Fig. 3c for the sample MN1. As the value of Pol 0 can vary from sample to sample, we will analyze here the relative effect of the pre-pulse through the ratio (ΔPol/Pol 0 ). This measured ratio as a function of the time separation Δt is presented in Fig. 4 for all investigated samples. We observe two main results. First, the effect of the pre-pulse decreases with Δt for all samples. Second, we notice that the two series of samples show a rather distinct behavior. The effect of the pre-pulse on the CMN samples is significantly smaller as compared to the samples from the MN samples. A reference sample, without Mn, shows no effect at all, as expected.
We believe that these results are an indication of an optical control of the Mn spin polarization. In this interpretation, the spin-down polarized carriers created by the σ − pre-pulse interact with the Mn ions giving rise to an effective spin-down polarization of the magnetic ions. Conversely, the polarization of the Mn ions affects the spin-polarization of the carriers created by the following σ + pulse. This fast process occurs during the rising of the PL emission so that it cannot be resolved by our experimental set-up, but it gives rise to a reduced initial polarization degree associated to the pre-pulse. As the Mn ions should present relatively long spin times  compared to the spin of the carriers, they must act as an effective spin reservoir. As the delay time between the pulses is increased, the Mn ions lose its spin polarization, becoming less effective in reversing the spins of the photocreated carriers, which explains the decreasing of (ΔPol/Pol 0 ) with increasing Δt as shown in Fig. 4. A quantitative analysis of the complex relation between (ΔPol/Pol 0 ) and the effective polarization of the Mn ion is prevented by the fact that we do not have access to the polarization dynamics during the laser pulse. However, the non-zero values of (ΔPol/Pol 0 ) obtained for Δt = 1.5 ns indicate that the Mn spin lifetimes are longer than 1 ns in our structures. We point out that we have also performed measurements using the same circular polarization degree for both the pulse and the pre-pulse, not shown here. We observed that under this condition (ΔPol/Pol 0 ) becomes essentially zero. This excludes heating effects caused by the pre-pulse.
The polarization effect is stronger for sample MN1 as compared to sample MN2, which is consistent with the larger separation between the δ Mn layer and the QW for MN2. This result supports our interpretation that the observed effect originates from the spin coupling between Mn ions and confined carriers, so that it is reduced when the overlap between these entities decreases. We also remark that (ΔPol/Pol 0 ) is significantly smaller for the CMN samples as compared to the MN samples and is null for the reference sample without Mn. We interpret the reduced effect on the CMN series as an indication of a reduced interaction between confined carriers and Mn ions on these samples. This conclusion is supported by previous independent results that also indicated a reduced overlap on those samples 10,13 . In addition to provide additional holes to the QW, the C delta-doping layer modifies the self-consistent potential profile of the structure, which changes the wave-function overlap 10 . Besides, CMN samples were grown in a slightly different temperature. This should affect the Mn distribution and may also contribute to the reduced overlap on these samples.
We also point out that even for measurements with a single excitation beam, the Mn ions should act as a spin reservoir. In this case, the carriers with a well defined spin-polarization created by one excitation pulse should interact with the Mn ions during its transient. This interaction should result in an effective polarization of the Mn ions, which in turn, should contribute to a longer spin polarization time of the photocreated carriers. Thus, samples with stronger interaction between the confined carriers and the Mn ions should present longer spin times, which is indeed observed as we compare the results from Table 1 and Fig. 4. This effect explains the relatively larger spin relaxation times (τ s ) obtained for the MN-samples as compared to the CMN samples. The correlation is also consistent when we compare samples MN1 and MN2, as the first one presents the larger values of (ΔPol/ Pol 0 ) and the longer τ s .
Finally, we studied the temperature dependence of τ s and (ΔPol/Pol 0 ) for a constant Δt = 0.6 ns for the sample MN1. Both parameters decrease with increasing temperatures as shown in Fig. 5. Albeit reduced, the Mn spin memory effect is observed on temperatures up to 100 K. We do expect that the scattering mechanisms by phonons become more efficient with increasing temperatures, resulting in faster spin-relaxation of the carriers via the Elliot-Yafet mechanism, which is consistent with the reducing spin lifetimes observed in Fig. 5. At the same time, (ΔPol/Pol 0 ) also decreases with the temperature. Two correlated effects might be attributed to this behavior. On one hand, the decreasing of τ s due to scattering mechanisms must reduce the efficiency of the spin-polarized photocreated carriers to orientate the Mn spins, thus reducing the effective magnetization of the Mn ions and the resulting value of (ΔPol/Pol 0 ). On the other hand, increasing temperatures should also reduce the magnetization of the Mn ions per se, and therefore, diminishing (ΔPol/Pol 0 ). Furthermore, the decrease of the spin polarization time of the Mn ions with temperature can also contribute to the decreasing of τ s due to the reduction of the spin reservoir 16 .

Summary
In conclusion, we observed a clear effect of spin memory for samples where the QW confined carriers present a significant interaction with the Mn ions from a nearby δ Mn layer. For those samples, we demonstrated that by applying a pre-pulse from an additional excitation beam with an opposite circular polarization gives rise to a reduced polarization degree of the PL emission as compared to the results without the pre-pulse. We propose that the pre-pulse writes the spin information on Mn ions that act as a spin reservoir, and this memory is read by the second pulse as a reduced polarization degree. Furthermore, the Mn spins also act as a spin memory that increases the spin relaxation time of the photocreated carriers on measurements without the pre-pulse. The results demonstrate that despite the relatively large spatial separation between the confined carriers and the Mn ions in our samples, their magnetic interaction persists and gives rise to the possibility to manipulate the spins of the Mn ions optically.