Ultrahigh efficient spin orbit torque magnetization switching in fully sputtered topological insulator and ferromagnet multilayers

Spin orbit torque (SOT) magnetization switching of ferromagnets with large perpendicular magnetic anisotropy has a great potential for the next generation non-volatile magnetoresistive random-access memory (MRAM). It requires a high performance pure spin current source with a large spin Hall angle and high electrical conductivity, which can be fabricated by a mass production technique. In this work, we demonstrate ultrahigh efficient and robust SOT magnetization switching in fully sputtered BiSb topological insulator and perpendicularly magnetized Co/Pt multilayers. Despite fabricated by the magnetron sputtering instead of the laboratory molecular beam epitaxy, the topological insulator layer, BiSb, shows a large spin Hall angle of θSH = 10.7 and high electrical conductivity of σ = 1.5 × 105 Ω−1 m−1. Our results demonstrate the feasibility of BiSb topological insulator for implementation of ultralow power SOT-MRAM and other SOT-based spintronic devices.


Integration of topological insulator to SOT-MRAM
In the first scenario, BiSb is directly deposited on top of CoFeB. This scenario has some disadvantages and should be avoided. Since there is only one CoFeB/MgO interface, ∆ ~ 40 is small. Furthermore, we have shown that deposition of BiSb on a ferromagnetic layer can damage the ferromagnetic layer due to the large kinetic energy of Bi/Sb atoms, which reduces the effective spin Hall angle [Sci. Rep. 10, 12185 (2020)] and possibly reduce the TMR ratio. In the second scenario, we couple CoFeB to (Co/Pt)n multilayers ferromagnetically or antiferromagnetically with a middle Ta or Ru layer, then deposit BiSb on top of the (Co/Pt)n multilayers. This scenario has many advantages. First, the (Co/Pt)n multilayers add an extra ∆ (~ 40 for n = 2 as demonstrated in this work) so that the total ∆ ~ 80 can be achieved. It has been shown that it is possible to increase ∆ by this way without increasing the switching current density [Saito et al., Appl. Phys. Lett. 101, 022414 (2012)]. Furthermore, we can increase the number of (Co/Pt) pairs to keep ∆ > 60 when the diameter is further reduced. Secondly, the (Co/Pt)n multilayers protect the CoFeB layer from diffusion of Bi/Sb atoms during BiSb deposition. Finally, we have demonstrated in this work that a large spin Hall angle larger than 10 can be achieved with the BiSb/(Co/Pt)n interface. Thus, the BiSb/(Co/Pt)n/Ta(Ru)/CoFeB/MgO will be the better structure than BiSb/CoFeB/MgO for the free layer to utilize the giant spin Hall effect of BiSb for realistic SOT-MRAM. BiSb-based SOT-MRAM stack with CoFeB/BiSb (left) and CoFeB/Ta(Ru)/(Co/Pt)n/BiSb (right) free layer. The right structure has many advantages over the left structure.

Structure analysis
We used X-ray diffraction (XRD) and transmission electron microscopy (TEM) to characterize the structure of the (Co/Pt)n/BiSb multilayers. Figure S3(a) and S3(b) show a XRD θ-2θ spectrum and a cross-sectional TEM image of the (Co/Pt)n/BiSb multilayers, which indicate that the deposited BiSb thin film has a dominant (110) orientation.

Perpendicular uniaxial anisotropy field after Hall bar fabrication
To estimate the perpendicular uniaxial anisotropy field of the (Co/Pt) multilayers after Hall bar fabrication, we measured the anomalous Hall resistance as function of the inplane magnetic field. Figure S4 shows the anomalous Hall resistance as a function of the in-plane magnetic field. By fitting to RH = RH(0) � 1 − � k � 2 , we obtained Hk = 5.2 kOe.  Figure S5 shows the second harmonic Hall resistance data for estimation of the antidamping-like HAD as a function of J BiSb , which are summarized in Fig. 2(c)

Self spin-orbit torque in the (Co/Pt) multilayers
Recently, it was reported that (Co/Pt) multilayers can generate a "self" spin-orbit torque [Jinnai et al., Appl. Phys. Lett. 111, 102402 (2017)]. To evaluate the contribution of the self-SOT effect in the (Co/Pt) multilayer, we performed control experiments on stand-alone [Co(0.4)/Pt(0.4)]2 multilayers. We fabricated a 50 μm × 25 μm Hall bar similar to that in Fig. 1(b) of the manuscript. Figure S6 We first attempt to switch the magnetization of the (Co/Pt)2 multilayers by the self-SOT effect. We applied the same current density to the (Co/Pt)2 multilayers as that flowed into the [Co/Pt]2 multilayers in the [Co/Pt]2/BiSb heterostructure. In the first experiment shown in Fig. S6(b), we applied a DC current up to ±1.38×10 7 Acm -2 under an in-plane bias field of 2.75 kOe. We observed no switching but Joule heating. Next, we attempt self-SOT switching by 1 ms and 0.1 ms pulse currents ramped up to ±2.5×10 7 Acm -2 and ±2.75×10 7 Acm -2 , respectively. Again, we observed no switching as shown in Fig. S6(c) and S6(d).
Next, we performed second harmonic measurements to evaluate the "self" spinorbit torque and "self" spin Hall angle in the [Co(0.4)/Pt(0.4)]2 multilayers. The results are shown in Fig. S7. By fitting the high field data in the RH 2ω -Hx curve (Figs. S7(a) -S7(e)) to the equation (1) in the manuscript, we obtained HAD at each current density.
Figure S7(f) shows HAD as a function of J CoPt , from which we evaluate that "self" the spin Hall angle of [Co(0.4)/Pt(0.4)]2 is 0.26. This value is consistent with those observed in Pt/(Co/Pt)n by Jinnai et al. Appl. Phys. Lett. 111, 102402 (2017), which shows a maximum effective spin Hall angle of 0.30 for the underneath Pt layer. Considering the current distribution in [Co/Pt]2/BiSb heterostructure, we calculated that this "self" spinorbit torque contributes to 13% of the total spin-orbit torque, and thereby not the main source of the observed SOT switching. By subtracting the contribution of the self-SOT from the raw data, we evaluate θSH of BiSb to be 10.7.

On the asymmetric magnon scattering
Recently, it was observed that asymmetric magnon scattering in (BiSb)2Te3/(CrBiSb)2Te3 heterostructure [Phys. Rev. Lett. 119, 137204 (2017)] can results in a similar second harmonic signals to that SOT. Because the magnetic layer (CrBiSb)2Te3 itself is a topological insulator, it has surface states with spin-momentum locking. The electron spins on the surface states of (CrBiSb)2Te3 are scattered by the magnons of Cr atoms in an asymmetric way due to spin-momentum locking. This is a very rare case and observed so far only in (CrBiSb)2Te3. Meanwhile, the magnetic layer in our study is just ferromagnetic metal (Co/Pt)n multilayers with no such topological surface states. Thus, asymmetric magnon scattering is unlikely in our case.
However, there might be a situation when the electron spins on the surface of BiSb are scattered by magnon in (Co/Pt)2. Even if such a situation might be possible, we show three experimental evidences indicating that the magnon scattering is negligible in our [Co/Pt]2/BiSb multilayers.

Evidence 1: Absence of the J 3 component
First, it is well known that magnon scattering rate is a non-linear function of the current density, and can be described by aJ + bJ 3 . This is because the spin density generated by spin-momentum locking is proportional to J, while the magnon population increased by   Fig. S9(b).

Instead, the observed is a text-book SOT curve: sharply increases when
M approaches the in-plane x-direction, then drops at high fields following ~ 1/(Hx -Hu). i.e. when the magnetic field was applied in plane along the y direction. Figure S10 shows the 2 measured with a magnetic applied in plane along y direction. We observed no , which disagrees with the asymmetric magnon scattering mechanism.
From these above observations, we conclude that asymmetric magnon scattering is negligible in (Co/Pt)2/BiSb.