Physical mechanisms leading to controllable and stable magnetization switching are considered as the core of spintronic devices with high performance.^[1–4] In recent years, researchers have found that, by applying a current through the cross-section of a non-magnetic metal (NM)/ferromagnetic metal (FM) bilayer structure, the spin–orbit interactions (SOIs) can provide strong spin torques to switch the adjacent magnetic layer via the spin–orbit torque (SOT) effect.^[5–7] This method serves as a more efficient way to switch the magnetization than the spin-transfer torque (STT) effect and is expected to inspire a variety of high-speed and multifunctional spintronic devices.^[8–13]

The most common spin generators in NM/FM bilayers are heavy metals (HMs), such as Ta, W, Pt, Hf, since bulk SOIs in heavier elements are much more significant.^[14,15] However, the effective spin Hall angle of HM is limited, leading to a relatively high critical magnetization switching current density j_sw on the order of 10¹¹–10¹² A/m.^[5–7,15] Furthermore, with regard to certain high-performance bulk magnets, such as ferrimagnetic alloys,^[8] Heusler alloys,^[16] magnetic insulators,^[17] featuring a larger thickness, j_sw is expected to be further increased to get over the augmenting anisotropic energy barrier. These facts seriously hinder the realization of low power consumption. Therefore, numerous studies have been conducted to investigate novel NM materials with larger such as topological insulators (TIs),^[18–20] two-dimensional electron gas (2DEG),^[21,22] oxidized HMs,^[23,24] and so on. Among those materials, TIs (including Bi₂Te₃, Bi₂Se₃, Bi_0.9Sb_0.1, (BiSb)₂Te₃) exhibit great potential for application, thanks to the large SOTs induced by the spin–momentum locking on their surfaces at room temperature. For instance, a several-nm-thick magnetic layer can be switched by SOTs from Bi₂Se₃ or Bi_0.9Sb_0.1 with a j_sw of 10¹⁰ A/m,^[19,20,25] while with (BiSb)₂Te₃, j_sw could even be reduced to the order of 10⁹ A/m.^[26] Note that all those reported TI samples are grown on single-crystal substrates with an ultra-precise growing method of molecular beam epitaxy (MBE) to form excellent topological surface states (TSS). This method is acceptable at laboratory level, but incompatible with industrial memory or logic fabrication process. To resolve this obstacle, sputtered Bi₂Se₃ films on oxidized silicon substrate are recently reported to possess a great and can efficiently switch the adjacent ferromagnet at room temperature.^[27,28] Magnetron sputtering is commonly used in semiconductor industry and therefore offers a possibility to introduce typical TI-related chalcogenides with enhanced SOT into large-scale application. Nevertheless, few similar materials are reported so far to maintain large through this method.

In this paper, we report that Bi₂Te₃ films sputtered on oxidized silicon substrate can exhibit a remarkable SOT effect. By applying harmonic Hall measurement in Bi₂Te₃/CoTb bilayer, in this system is determined as large as 3.3± 0.3 at room temperature. Then, current-induced switching of a 6-nm-thick perpendicular magnetic anisotropy (PMA) CoTb ferrimagnetic alloy is also realized in the mentioned structure. Compared with traditional HMs, j_sw as well as the power consumption to switch the magnetic layer can be reduced by more than one order.

The samples in our work are deposited on thermally oxidized Si substrates using DC or RF magnetron sputtering under a base pressure better than 2×10⁻⁸ Torr. We first deposited a series of substrate/MgO (2 nm)/Bi₂Te₃ (t nm)/MgO (1 nm) stacks with t varying from 5 nm to 40 nm for the four-point resistivity measurements. Figure 1(a) shows that the resistivity of the sputtered Bi₂Te₃ has a downward trend versus its thickness t. This trend is similar to those in previously reported TI materials^[20] and can be explained by the combined effect of surface conduct and bulk conduct properties in Bi₂Te₃.^[29] To guarantee a relatively good as well as enough passing current through the cross-section, 8-nm-thick Bi₂Te₃ layer with a resistivity of 3011 μΩ⋅cm is chosen for the SOT experiments in our work. This Bi₂Te₃ layer should have a polycrystalline structure, and the roughness characterized by atomic force microscope (AFM) is shown in the inset of Fig. 1(a). The root mean square value of the surface roughness is 0.35 nm, which is smooth enough for the following heterostructure deposition.

Fig. 1.

	Figure Option ViewDownloadNew Window
	Fig. 1. (a) Thickness-dependent resistivity change measured by four-point method in MgO/Bi2Te3/MgO trilayers. Inset is an AFM image of Bi2Te3 (8 nm) sample. (b) Stack schematic of the investigated Bi2Te3/CoTb heterostructure. (c) Dependence of the magnetization direction of CoTb layer on x. When x is in the range of 6–8 nm, CoTb layers show excellent PMA properties.

In addition, a ferrimagnetic rare-earth-transition–metal alloy CoTb, in which Co and Tb sub-lattices are coupled anti-ferromagnetically, is used as the adjacent magnetic layer for two reasons: 1) CoTb alloy possesses robust bulk PMA within a large thickness range, and bulk PMA magnets present better switching performance and thermal stability in real device applications;^[19,30] 2) CoTb alloy has a relatively high resistivity (206 μΩ⋅cm), allowing more current to pass through the Bi₂Te₃ layer. Here, CoTb is deposited by the co-sputtering of Co and Tb targets. By tuning the chemical composition or temperature, CoTb could approach angular momentum compensation point or magnetization compensation point, and exhibits plenty of charming magnetic properties.^[30–32] However, few evidences yet support that CoTb approaching compensation point could contribute to increase in the whole system. Therefore, we use a Co-rich chemical composition, i.e., Co_0.83Tb_0.17, throughout our whole work, which is estimated by the independent deposition speed of Co and Tb. As shown in Fig. 1(b), a series of substrate/MgO (2 nm)/Bi₂Te₃ (8 nm)/CoTb (x nm)/Ta (1.5 nm) stacks are deposited. It can be found that the capping Ta layers are fully oxidized and not involved in the SOT switching process.^[33] Vibration sample magnetometry is utilized to determine the magnetic anisotropies of all the samples. As illustrated in Fig. 1(c), when x ranges from 3 nm to 12 nm, the magnetization direction of the CoTb varies from in-plane to out-of-plane, and tends to go slightly back to the in-plane state when x reaches 12 nm.

For electrical measurements, we choose the 6-nm-thick CoTb sample, since it possesses good PMA when the magnetic layer remains relatively thin. The net magnetization M_s of this 6-nm-thick CoTb layer is 215 emu/cc. By typical lithography and ion beam etching, the sample is fabricated into 5-μm-width Hall bar device. Keithley 6221 current source and 2182 voltage-meter are used to implement AHE and SOT switching experiments. To measure the SOT efficiency, harmonics Hall measurement is then implemented by injecting an AC current of 133.33 Hz along x axis. Two SR830 lock-in amplifiers are used to detect the harmonic signals.

Figure 2(a) shows the fabricated device with gold electrodes and the basic electrical measurement schematic. The anomalous Hall effect (AHE) resistance versus the applied out-of-plane magnetic field is plotted in Fig. 2(b). Squared loop and similar coercive field confirm that the stack maintains its as-deposited magnetic properties. Figure 2(c) plots the curves of the first and the second harmonic voltages, i.e., R_1ω and R_{x(y), 2ω} versus the in-plane magnetic field H_x(y). A ratio coefficient B_x(y) could be firstly calculated with the respective curvature of R_1ω and the slope of V_x(y), 2ω^[34,35]

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (a) Image of 5-μm-width Hall bar device with electrical measurement setup. (b) Anomalous Hall resistance curve of Bi2Te3 (8 nm)/CoTb (6 nm) sample. (c) Harmonics resistance curve with applied in-plane magnetic field. Hx-induced damping-like signal (red points) is much larger than Hy-induced field-like signals (blue circles, amplified by 3 times). (d) Effective damping-like field as a linear function of current density in Bi2Te3 layer.

Figure 2(d) shows the calculated H_DL as a function of applied charge current density j_c in the Bi₂Te₃ layer. Here, j_c is calculated by a parallel circuit model of Bi₂Te₃/CoTb bilayer, which can be written as

Current-induced magnetization switching at room temperature is demonstrated as well and plotted in Figs. 3(a) and 3(d). Here, a series of 0.1-ms-width current pulses are applied to switch the magnet, while a 400 Oe in-plane magnetic field is used to break the magnetization precession symmetry. The j_sw to switch 6-nm-thick CoTb is about 9.7×10⁹ A/m². This value is one or two orders lower than that in typical HM-based heterostructure,^[5–8] confirming again the existence of strong SOTs in the sputtered Bi₂Te₃. When the direction of the applied in-plane field is reversed, the switching polarity of the curve changes as expected in the theoretical SOT switching framework. We also notice that the variation of SOT-driven R_AHE (∼ 1.6 Ω) is smaller than that of magnetic-field-driven R_AHE (∼ 2.1 Ω), revealing a partial switching phenomenon (76 % magnetization). This difference is also observed in previous switching results with MBE-grown TIs and can be explained by the Joule-heating-caused demagnetization in CoTb layer.^[26] We are convinced that reducing shunting current in CoTb layer by reasonable device design could efficiently eliminate the problem.

Fig. 3.

	Figure Option ViewDownloadNew Window
	Fig. 3. Current-induced switching curves of Bi2Te3/CoTb under an in-plane magnetic field Hx = 400 Oe (a) and –400 Oe (e). (b)–(c) The AHE curves of Pt/CoTb, Ta/CoTb, and W/CoTb samples, respectively. (f)–(h) The current-induced switching curves of Pt/CoTb, Ta/CoTb, and W/CoTb samples under Hx = 400 Oe. The different switching polarities reveal different spin Hall angle signs.

To better evaluate the SOT performance of the sputtered Bi₂Te₃ sample, some control samples based on conventional sputtered SOT sources (Pt, Ta, W) are also studied with a stack structure of HM (5 nm)/CoTb (4 nm). While maintaining the same chemical composition and M_s, the thickness of CoTb is reduced in these samples, since SOTs originating from common HMs can hardly switch magnets with large thickness. As shown in Figs. 3(b)–3(d), all the control samples have squared R_AHE loops versus perpendicular magnetic field, indicating that they possess good PMA properties. Besides, because of the different shunting effects of Bi₂Te₃, Pt, Ta, and W, R_AHE in the four samples are not consistent. Figures 3(f)–3(h) illustrate the current-induced magnetization switching curves of Pt/CoTb, Ta/CoTb, and W/CoTb samples. In these experiments, the applied in-plane magnetic field is fixed to + 400 Oe to guarantee a more reasonable comparison. We observe similar R_AHE in current-induced switching and field-induced switching for each sample, revealing that the aforementioned partial switching phenomenon barely exists in HM based samples. Regarding the calculation of current density, the resistivity of Pt, Ta, and W is measured to be 27 μΩ⋅cm, 112 μΩ⋅cm, and 151 μΩ⋅cm, respectively. We find that the values of j_sw to switch 4-nm-CoTb for Pt, Ta, and W are approximately 5.5×10¹¹ A/m², 2.3 × 10¹¹ A/m², and 1.7×10¹¹ A/m². All those critical current densities are consistent with previously reported values^[5–8] and are obviously much larger than those in our Bi₂Te₃/CoTb system, although CoTb in Bi₂Te₃ based sample is even thicker, i.e., 6 nm. Note that, under an in-plane field with fixed direction, the switching polarity in Bi₂Te₃ sample is in line with that in Ta or W samples, which is opposite to that in Pt sample. This SOT polarity characteristic indicates that our sputtered Bi₂Te₃ possesses the same spin Hall angle sign as Ta and W, but is opposite to Pt. Nevertheless, this phenomenon is in contradiction with SOTs in previous-reported Bi-based materials: both MBE-grown TIs and sputtered Bi₂Se₃ films are demonstrated to have the same sign as Pt.^{[19,20,25–27]} The origin of this difference will be discussed subsequently. In our work, values of of Pt, Ta, and W in Fig. 4(a) are determined as 0.07± 0.007, 0.14± 0.01, and 0.18± 0.02, through previously mentioned harmonic measurement and Eq. (4). In accordance with the varying j_sw values in those SOT materials, these calculated values are quite reliable and accepted in our following discussion.

Fig. 4.

	Figure Option ViewDownloadNew Window
	Fig. 4. (a) Effective spin Hall angles of sputtered Pt, Ta, W, and Bi2Te3. (b) Normalized power consumption to switch the magnetization of CoTb per unit volume by different SOT source materials. The black bars and orange bars correspond respectively to the calculation formulas related to and jsw.

Essentially, improving the magnetization switching energy efficiency is the eventual goal of obtaining higher . Considering that the resistivity of our sputtered Bi₂Te₃ is similar to that of MBE-grown TIs and can be dozens, even hundreds of times higher than that of conventional HMs, we think that it is significant to figure out whether the sputtered Bi₂Te₃ indeed offers energy superiority by calculating the switching power consumption P. Basically, P to switch the magnetization of CoTb per unit volume with different SOT sources can be analyzed by a heat dissipation formula^[27]

Last but not least, we will discuss about the origin of the strong SOTs in the sputtered Bi₂Te₃ films. In MBE-grown TIs, SOTs basically come from the spin–momentum locking in TSS. Thus, great surface quality is the basic precondition to gain large . However, good TSS definitely will not appear in the deposited Bi₂Te₃ films, since magnetron sputtering is a rapid and rough growing method. Generally, there are two possible reasons to answer this question: 1) sputtered Bi chalcogenides could possibly possess nanoscale grain structure. Some ab initio calculation results suggest that this reduced dimensionality may contribute to the non-equilibrium spin accumulation driven by intraband Edelstein effect.^[27] 2) It is also possible that remarkable SOIs already exist in the sputtered Bi₂Te₃ films with relatively high resistivity and heavy elements. To further analyze the possible origin, an important point is that, the observed spin Hall angle sign of our sputtered Bi₂Te₃ films is opposite to that in MBE-grown TIs (same as Pt). Another recent work also reports a similar result: sputtered W_xTe_{1 – x} films possess opposite spin Hall angle sign to WTe₂ single crystals.^[38–40] However, for related Se-based material, the sputtered Bi₂Se₃ films have the same spin Hall angle to the MBE-grown Bi₂Se₃.^[27,28] Those evidences indicate that sputtered Te-based films probably exhibit different SOT mechanisms to sputtered Se-based films, although they both belong to the Bi chalcogenide class. Besides, the opposite spin Hall angle signs of sputtered Bi₂Te₃ and MBE-grown Bi₂Te₃ also exclude the possibility of major origin from TSS. Therefore, considering that both sputtered Te-based materials^[38] and Bi-based chalcogenides^[27] are reported to have obvious thickness-dependent spin Hall angle, the second reason attached to intrinsic bulk spin Hall effect is more convincing from our perspective.

In summary, we investigate SOTs in Bi₂Te₃ films grown by an industry-compatible deposition method magnetron sputtering. A high SOT efficiency χ (8.7± 0.9 Oe/(10⁹ A/m²)) and a remarkable (3.3± 0.3) are obtained in Bi₂Te₃/CoTb heterostructure as shown by harmonic Hall measurement. In addition, an ultra-low current switching current density (9.7×10⁹ A/m² to switch 6-nm-thick PMA CoTb) is achieved as well. Compared with other sputtered SOT sources, the sputtered Bi₂Te₃ in our work shows much higher energy efficiency, indicating that it is a promising candidate in future SOT spintronic devices. Our work may also provide an alternative route to introduce more laboratory-level high-performance chalcogenides in industry-level spintronic applications.