Permissions for Reuse

Greater understanding of defect formation is needed to further increase the Curie temperature (T_C) of ferromagnetic semiconductors, such as Ga_1−xMn_xAs, which hold promise for magnetoelectronic devices.¹ In typical molecular beam epitaxy (MBE) growth of GaMnAs, low substrate temperatures (T_sub) of 250  °C are required to suppress precipitation of secondary crystal phases (e.g., MnAs) and incorporate a large fraction of Mn atoms as substitutional (Mn_Ga) acceptors in GaAs. Mn atoms also form double-donor interstitials (Mn_i) that compensate holes. Annealing films in air at temperatures of 180  °C removes Mn_i,² thus increasing the hole density (p) and T_C, in agreement with the Zener model of carrier-mediated ferromagnetism.^3,4 This model also predicts that T_C increases linearly with Mn_Ga, which is supported by studies over the past decade with 0.02<x<0.08.⁴ More recently, MBE growth of GaMnAs with x0.1 was demonstrated in attempts to increase T_C.^5,6 These films require reduced T_sub (180  °C) and are limited to 10  nm thickness; T_C's after annealing reach the previous limit of 170  K.⁴ In these studies, the effect of arsenic flux was not explored, though it should be a critical parameter at these conditions.

GaAs films grown at T_sub250  °C can contain double-donor arsenic antisites at concentrations As_Ga10²⁰  cm⁻³ (1% of Ga sites).^7,8 For a fixed As:Ga flux ratio, As_Ga in low-temperature (LT) GaAs increases exponentially as T_sub is decreased below 300  °C.^9,10 Extended defects are prominent below 200  °C, leading to polycrystalline films.^11,12 Previously, we developed combinatorial (nonrotated) growth to continuously vary As:Ga and, therefore As_Ga, in GaMnAs films near the paramagnetic/ferromagnetic transition (1% Mn) with T_sub=250  °C.^13,14 There, excess arsenic flux of 1% suppresses ferromagnetism, an unfeasible flux precision for rotated growths. Here we extend this technique to T_sub150  °C, where the metastable solubility limit of Mn in GaAs is increased beyond 10%. As in LT GaAs, increased As_Ga and extended defects occur unless As:Ga is precisely balanced to achieve stoichiometry. We discuss systematic variations in magnetotransport, ferromagnetism, and crystal structure of GaMnAs (x0.1) as As:Ga is varied along the [110] direction [Fig. 1(a) inset]. The highest T_C's are obtained in a narrow band of stoichiometric material. Stoichiometric, 100  nm thick films exhibit excellent crystalline quality and square magnetic hysteresis up to 165  K. Stoichiometric films grown with varying x (>0.1), T_sub, and growth rate reach the same T_C (150–170  K) after annealing; increasing Mn_Ga does not increase T_C as the Zener model predicts. We find that the linear dependence of T_C on Mn_Ga is limited to 0.02<x<0.1.

Figure 1.

Samples are grown on 2  in., semi-insulating (001) GaAs wafers by MBE as described previously.¹³ The total fraction of Mn in the films (x) is calibrated from growth rate calibrations of MnAs and GaAs reflection high-energy electron diffraction (RHEED) intensity oscillations. After growth of the buffer layer, T_sub is dropped to 150  °C and substrate rotation is stopped. Below 350  °C, the arsenic shutter is closed; it is opened for 10  s prior to LT growth to ensure an arsenic-terminated surface. T_sub is measured in real time during growth using band-edge thermometry with an accuracy of ±2  °C.¹⁵ During the first 5–15  nm of growth, T_sub increases by 10  °C due to radiation from the heated Ga and Mn sources; T_sub quickly stabilizes after the heater power is dropped. We observe a two-dimensional (streaky) 1×2 RHEED pattern on As-rich material transitioning to a three-dimensional (spotty) pattern on Ga-rich material, observed by eye as a mirror finish (As-rich) transitioning to haze from Ga droplets on the surface.¹³ Wafers are cleaved into 3×3  mm² pieces along [110], yielding 17 samples with systematically varying As:Ga.

Magnetometry and magnetotransport are measured on sample pairs with equal As:Ga (adjacent in the [10] direction). Magnetotransport is measured in the van der Pauw geometry from 2  to  380  K with out-of-plane (hard axis) fields up to 14  T. T_C is determined from superconducting quantum interference device (SQUID) magnetometry (Quantum Design magnetic properties measurement system SQUID vibrating sample magnetometer) while warming in 50  Oe ([110] in plane) after cooling from 350  K in 1  T. Saturation moment (M_sat) is determined from hysteresis loops (<0.5  T). We anneal sample pairs in air at 180  °C until conductivity is maximized [Fig. 1(c) inset].² High-resolution x-ray diffraction (HRXRD) is performed on half-wafers by using a triple-axis (Philips X'Pert MRD PRO) diffractometer with a 0.25° aperture for 1  mm resolution along [110]. Lattice constant (a_GaMnAs) is determined from (004) -2 scans.

Electronic, magnetic, and structural properties all depend critically on As:Ga. Figure 1 plots data from a single 100  nm thick Ga_0.84Mn_0.16As film grown with T_sub=150  °C at 0.852  Å/s. Room temperature longitudinal (_xx) and Hall (_xy) conductivities [Fig. 1(a)], T_C and M_sat at 5  K [Fig. 1(b)], and a_GaMnAs [Fig. 1(c)] all maximize at the same As:Ga where stoichiometry is achieved. _xx increases with annealing [Fig. 1(c) inset] due to outdiffusion of Mn_i from the bulk to the surface.² Examples of -2 scans in Fig. 1(d) reveal high-quality XRD epilayer peaks with thickness fringes observed at stoichiometry (black points) which match dynamical simulations (line).¹⁶ The postannealing epilayer shift (red) shows a reduction in a_GaMnAs from removal of Mn_i. This corroborates the model of Masek et al.¹⁷ and results of Sadowski and Domagala¹⁸ demonstrating that the main contributor to a_GaMnAs expansion is Mn_i. It follows that the largest number of Mn_i occur for stoichiometric material, the same As:Ga at which T_C, M_sat, _xx, and _xy are largest.

As-rich material displays a broad XRD peak without thickness fringes, indicating low crystal quality [blue, Fig. 1(d)]. At this As:Ga, ferromagnetic MnAs particles (NiAs structure) are detected by magnetometry (T_C=320  K). No remnant magnetization (M_rem) is observed at temperatures above the GaMnAs T_C in the stoichiometric region, but a M_rem of 0.1µ_B/Mn from MnAs persists above the GaMnAs T_C in the film with As:Ga=13.5 (35% greater than As:Ga=10, the stoichiometric condition). Excess arsenic might promote MnAs precipitation by creating extended defects that act as nucleation sites. MnAs is not observed in thinner As-rich films (7.5  nm), which could be below the critical thickness for defect formation. For our As-rich material, we reproduce previously observed limits^5,6 to T_C, _xx, and film thickness; these limits are overcome by achieving stoichiometry at precise As:Ga.

Excess arsenic and Mn_i significantly alter magnetic and electronic properties such as T_C, M_sat, coercive field (H_C), resistivity (_xx), and magnetoresistance (MR). Figure 2(a) shows the T_C increase to 165  K after minimizing defects, and we observe square hysteresis at this high T_C [Fig. 2(b)]. Excess arsenic reduces the H_C and M_sat at 5  K [Fig. 2(c)]. _xx and MR increase by orders of magnitude at 10  K in samples with Mn_i and excess arsenic [Fig. 2(d)]. Mn_i increases MR (at 14  T) from 3% to 44%, and excess arsenic (As:Ga=12) increases it to 87%. _xy switching occurs in all samples due to the anomalous Hall effect [Fig. 2(d)].

Figure 2.

The growth parameter space of heavily alloyed GaMnAs is systematically explored with 16 nonrotated wafer growths with 0.1<x<0.22, 120<T_sub<160  °C, and growth rate varied from 0.198  to  1.41  Å/s. Figure 3(a) plots T_C against x for the stoichiometric samples from each heavily alloyed wafer and also lightly alloyed wafers (x<0.1, 200<T_sub<250  °C). We observe the expected linear increase in T_C with x (x<0.1),^3,4 but it does not extend to the heavily alloyed regime, where T_C=150–165  K postannealing for all x (maximum observed at x=0.16) despite the wide range of growth parameters. In contrast, a_GaMnAs increases linearly up to x0.16 for both as-grown and annealed films [Fig. 3(b)] and is fitted using the model of Masek et al.^17,18 Assuming As_Ga0 in the stoichiometric region, a_GaMnAs=a_GaAs+0.02(x−z)+1.05z, where z is the atomic fraction of Mn_i and (x−z) is the atomic fraction of Mn_Ga. Linear fits, plotted as lines in Fig. 3(b), demonstrate a constant fraction of interstitials (z/x) of 0.24 in as-grown samples and a reduction of this fraction to 0.13 after annealing. Recent theoretical work suggests that the coefficient for Mn_Ga is larger than 0.02 (0.09), which would decrease our estimates of z/x.¹⁹ The linear behavior indicates that Mn_Ga and Mn_i increase proportionally with x, up to x0.16. Contrary to a prediction of the Zener model, T_C is independent of x for x>0.1 even though Mn_i compensation maintains the same proportions in this heavily alloyed regime.

Figure 3.

In conclusion, heavily alloyed, 100  nm thick GaMnAs films (0.1<x<0.22) with reproducible, high magnetic (T_C165  K), and structural quality are grown by utilizing a combinatorial technique to achieve stoichiometry. Magnetotransport, ferromagnetism, and lattice constant are critically dependent on As:Ga for low T_sub, with stoichiometric, annealed material displaying optimal properties. While structural and magnetic data indicate a linear increase in Mn_Ga up to x0.16, we do not observe the predicted T_C increase,³ suggesting that the Zener model may not be applicable to the heavily alloyed regime. Application of this combinatorial technique provides a reproducible method for obtaining high-T_C GaMnAs and allows systematic exploration of the growth parameter space.

We thank J. H. English and A. W. Jackson for MBE advice and M. A. Scarpulla for helpful discussions. This work was supported by ONR, MURI, and AFOSR. The authors used MRL Central Facilities supported by the MRSEC Program of NSF Contract No. (DMR05-20415). S.M. acknowledges support by the DoD through the NDSEG Fellowship Program.