Spin–orbit torque switching of the antiferromagnetic state in polycrystalline Mn₃Sn/Cu/heavy metal heterostructures

Electrical control of magnetization through spin-transfer torques (STT) or spin-orbit torques (SOT) has enabled significant innovations in the development of non-volatile memories as well as in basic spintronic research using ferromagnets.^1–7 In the advent of big data, Internet of Things (IoT) and artificial intelligence (AI), there are growing demands for the higher density and faster operation speed of the memory devices. With this perspective, antiferromagnets have attracted significant attention because their small stray fields could prevent the perturbation between neighboring cells and the spin dynamics in much faster in antiferromagnets than ferromagnets.^8–11 The recent intensive studies have led to the breakthrough of the electric-current control of antiferromagnetic (AF) sublattices detected by the anisotropic magnetoresistance (AMR).^12–14 Integrating these emerging technologies with existing spintronics is essential to find further advancement to induce electrical switching in an antiferromagnet via STT or SOT by injecting spin current through the interface with other materials.^13,15–18 Besides, it would be more beneficial to use linear responses to magnetization for the readout such as magneto-conductance, anomalous Hall effects (AHE), and Faraday/Kerr effects, than the currently available quadratic responses to magnetization (e.g. AMR, quadratic magneto-optical effects).^9–11,19,20 For example, such a linear-response may allow a polycrystalline form of an antiferromagnet to serve as a switching device.

Meanwhile, recent development in the understanding of topological characteristics in electronic band structure^11,21–23 has led to the discovery of the AHE in chiral antiferromagnets Mn₃X (X = Sn, Ge, Ga, Ir, Pt, and Rh). This has shown that a large linear transverse response such as AHE and anomalous Nernst effect (ANE) may exist in the antiferromagnets in the absence of magnetization due to the Berry curvature in momentum space^11,24–29 Moreover, a successful thin-film growth has enabled the electrical detection of the AF state by the Hall resistance even in a polycrystalline sample.³⁰ 

Recently, magnetic SOT switching of the antiferromagnet Mn₃Sn has been performed by applying electrical current in Ru/Mn₃Sn/Pt or W heterostructures,³¹ where the spin Hall angle of Pt and W layer determine the switching direction. On the other hand, in addition to the spin Hall effect, interfacial effects at the ferromagnets/heavy metals interface can also generate notable field-like torque on the magnetization.³² To estimate the possible interfacial effect in the SOT switching at Mn₃Sn/heavy metals heterostructures, here we perform the electrical switching measurement of Ru/Mn₃Sn/Pt or W and Ru/Mn₃Sn/Cu/Pt or W and compare the volume fraction of the switched Mn₃Sn domain in each sample. Mn₃Sn is one of the best studied among various kagome-based metals with nontrivial topology of band structure.^11,33,34 The hexagonal D0₁₉ structure of Mn₃Sn has the ABAB-stacking of a (0001)-kagome layer of Mn, and the associated geometrical frustration leads to a three-sublattice non-collinear AF ordering of Mn spins below the Néel temperature T_N ∼ 430 K^35,36 (Fig. 1(a)). The AF spin texture consist a ferroic ordering of a cluster magnetic octupole,³⁷ which leads to the large linear responses, such as AHE, ANE, and magneto-optical Kerr effect (MOKE), instead of a tiny uncompensated magnetization M ∼ 0.006 µ_B/f.u. induced by the spin canting within the (0001)-plane.^11,38,39 Moreover, the material hosts a topological magnetic Weyl semimetal state,^29,38–40 leading to the large responses robust against disorder and thermal fluctuations.

To exert SOT on Mn₃Sn by the spin Hall effect from the heavy metals, we prepare the thin films with heavy metals Pt or W in the heterostructures: Ru(2)/Mn₃Sn(40)/Pt(7.2) or W(7.2)/AlO_x(5) (in nm) deposited on a thermally oxidized Si substrate; in addition, we prepare the stacks with Cu insertion layer: Ru(2)/Mn₃Sn(40)/Cu(5)/Pt or W(7.2)/AlO_x(5) (Fig. 1(b)). Ru and Mn₃Sn are deposited at room temperature using a dc magnetron sputtering machine under the pressure ∼5 × 10⁻⁷ Pa. After the fabrication of the Mn₃Sn layer, the stacks are annealed at 450 °C for 0.5 h. After cooling to room temperature, we grow the Cu, Pt, and/or W, and AlO_x layer by molecular beam epitaxy (MBE) with the base pressure of below 2 × 10^-8 Pa. The pressure keeps in situ conditions in the transfer from the sputter chamber to the MBE chamber. The composition of the Mn₃Sn layer is Mn_3.01(2)Sn_0.99(2), determined by scanning electron microscopy-energy dispersive X-ray spectrometry (SEM-EDX).

The multilayer thin films are fabricated into a 16 μm × 96 μm Hall bar structure by photolithography and contacted with Ti/Au electrodes (Fig. 1(c)). To measure the AHE, 0.2 mA read current I_read is applied in x-direction and the Hall voltage is detected in y-direction. In the field switching measurement, an out-of-plane magnetic field H_z is applied in z-direction, while in electrical switching measurement a bias field H_x and a write current I_write is applied in x-direction. Figure 1(d) shows the measurement sequence of the electrical switching measurement. The temperature in all measurements is at room temperature ∼293K. A bias field μ₀H_x = 0.1T is applied in the whole electrical switching measurement. Firstly, a 100 ms write current I_write is applied and followed by a 0.2 mA read current I_read, and then the Hall voltage V_H is measured after a wait time of 600 ms to cool down the Mn₃Sn sample to the room temperature.³¹ 

Firstly, we measure the anomalous Hall resistance R_H as a function of the out-of-plane magnetic field H_z to estimate the population of switchable domains at the out-of-plane direction (Fig. 2). A clear hysteresis of R_H is observed with the zero-field change of ΔR_H^Field (= R_H(+H_z→0) R_H(H_z→0)) in four different multilayer devices: (a) Mn₃Sn(40)/Pt(7.2), (b) Mn₃Sn(40)/W(7.2), (c) Mn₃Sn(40)/Cu/(5)/Pt(7.2), and (d) Mn₃Sn(40)/Cu/(5)/W(7.2). The results indicate that the +z-polarized(-z-polarized) magnetic octupoles of Mn₃Sn correspond to a negative(positive) Hall resistance, consistent with previous report.^11,30 Compared to the Pt or W devices without Cu layer, devices with Cu insertion layer show ∼5 times smaller Hall resistance. Calculated from the total resistance of the devices with and without Cu insertion layer measured by a two-probe method, the current flowing in Cu layer is estimated to be ∼60% of the total current. Using the simple model for the shunting effect on AHE in the multilayer system by ignoring the interface resistance,^41,42 the AHE signal in devices with Cu insertion layer are estimated to be ∼16% of the devices without Cu layer , which is close to the 5 times difference observed in our experiments.

Figures 3(a) and 3(b) show the results of the electrical switching of R_H in Mn₃Sn(40)/Pt(7.2) and Mn₃Sn(40)/W(7.2) devices. Here we define ΔR_H^Current = R_H(+I_write→0) R_H(I_write→0). We find that sign of ΔR_H^Current and the corresponding switching direction are opposite in the Pt and W devices. This can be explained by the different sign of spin Hall angle in Pt (θ_SH > 0) and W (θ_SH < 0), being consistent with previous report.³¹ The estimated switching current density in Pt or W layer is ∼10¹¹ A/m² being comparable to the SOT switching in ferromagnets and antiferromagnets.^6,7,12–14 The temperature of the devices under the similar write current density in Mn₃Sn/Pt or W devices are reported to be ∼50K in our previous study using the samples prepared in the same method.³¹ On the other hand, in addition to the SOT generated from spin Hall effect, it has been reported that the Rashba spin-orbit coupling at the interface of ferromagnets/heavy metal can also generate a field-like torque on the magnetic moments.³² To check if such interfacial effect at Mn₃Sn/Pt or W interface has notable contribution to the SOT switching, we perform the electrical switching in Pt and W devices with a Cu insertion layer (Fig. 3(c) and 3(d)). A clear electrical switching is observed in both devices, while the shunting effect of Cu layer and the smaller resistance of device drastically decrease ΔR_H^Current, which is similar to the decrement of ΔR_H^Field. Since the shunting effect should be the same when applying read current in both field and current switching measurement, we compare the volume fraction of the switched domain of Mn₃Sn by the ratio between the electrical switching and field switching Hall resistance signal, |ΔR_H^Current/ΔR_H^Field| (Table I). This ratio is 30% and 27% in Mn₃Sn/Pt and Mn₃Sn/W devices, respectively, consistent with previous report.³¹ We found that |ΔR_H^Current/ΔR_H^Field| in devices with Cu insertion layer are only slightly smaller than that devices without Cu. This result indicates that the spin current generated from Pt or W layer can diffuse to the Mn₃Sn layer and exert SOT on Mn₃Sn efficiently. The decay of spin current in Cu layer is neglectable because the spin diffusion length of Cu at room temperature is 1∼2 order larger than 5nm Cu thickness.⁴³ The similar ratios of |ΔR_H^Current/ΔR_H^Field| in with and without Cu layer devices indicate that the spin Hall effect in heavy metals contribute dominantly to the SOT switching mechanism in Mn₃Sn compared to other interfacial effects.

To summarize, we demonstrate the SOT switching of the Hall resistance in polycrystalline Mn₃Sn/Pt, Mn₃Sn/W and Mn₃Sn/Cu/Pt, Mn₃Sn/Cu/W heterostructures. The polarity of the magnetic switching in both devices with and without a Cu insertion layer are determined by the sign of the spin Hall angle of the heavy metals (Pt (θ_SH > 0) and W (θ_SH < 0)), being consistent with the SOT mechanism. To estimate the volume fraction switched by electrical current, we calculate the ratio between current switching Hall resistance ΔR_H^Current and field switching Hall resistance ΔR_H^Field, and find that the |ΔR_H^Current/ΔR_H^Field| values are nearly the same in the Pt and W devices with and without Cu layer. Our result indicates that the spin current generated by the spin Hall effect of Pt or W layer contributes dominantly to the SOT compared to possible interfacial effects at Mn₃Sn/heavy metal interface.

We thank D. Nishio-Hamane for SEM-EDX measurements. This work is partially supported by CREST(JPMJCR18T3), Japan Science and Technology Agency (JST), by Grants-in-Aids for Scientific Research on Innovative Areas (15H05882, 15H05883) from the Ministry of Education, Culture, Sports, Science, and Technology of Japan, and by Grants-in-Aid for Scientific Research (16H06345, 18H03880, 19H00650).

H.T. and T.H. contributed equally to this work.

The data that support the findings of this study are available from the corresponding author upon reasonable request.