Enhanced spin–orbit torque via interface engineering in Pt/CoFeB/MgO heterostructures

Spin–orbit torque facilitates efficient magnetisation switching via an in-plane current in perpendicularly magnetised heavy-metal/ferromagnet heterostructures. The efficiency of spin–orbit-torque-induced switching is determined by the charge-to-spin conversion arising from either bulk or interfacial spin–orbit interactions or both. Here, we demonstrate that the spin–orbit torque and the resultant switching efficiency in Pt/CoFeB systems are significantly enhanced by an interfacial modification involving Ti insertion between the Pt and CoFeB layers. Spin pumping and X-ray magnetic circular dichroism experiments reveal that this enhancement is due to an additional interface-generated spin current of the non-magnetic interface and/or improved spin transparency achieved by suppressing the proximity-induced moment in the Pt layer. Our results demonstrate that interface engineering affords an effective approach to improve spin–orbit torque and thereby magnetisation switching efficiency. © 2019 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/1.5084201 The electrical manipulation of magnetisation in magnetic nanostructures has opened up new avenues for the further development of spintronic devices because this approach affords simple device miniaturisation and the potential for large-scale integration.1–3 Conventionally, spin-transfer torque (STT) has been used to control the magnetisation in magnetic multilayer structures,3,4 particularly for STT-magnetic random access memory (STT-MRAM), in which a spin-polarised current is injected in the direction perpendicular to the film plane. Recently, another type of spin-torque realised by spin–orbit coupling, the so-called spin–orbit torque (SOT),5,6 has been widely investigated because it allows for efficient manipulation of the magnetisation using in-plane current, particularly during magnetisation switching5,6 domain-wall and skyrmion motion.7–10 SOT in heavy metal (HM)/ferromagnet (FM)/oxide heterostructures arises from the spin current induced by a charge current via the spin Hall effect (SHE) in the HM and/or the interfacial spin–orbit coupling (ISOC) effect at HM/FM interfaces. This spin current exerts a torque T on the local magnetisation as T = τDLm̂ × (ŷ × m̂) + τFLm̂ × ŷ,11,12 where τDL (τFL) denotes the dampinglike torque (field-like torque), m̂ denotes the unit vector along the magnetisation direction, and ŷ denotes the unit vector along the APL Mater. 7, 031110 (2019); doi: 10.1063/1.5084201 7, 031110-1

The electrical manipulation of magnetisation in magnetic nanostructures has opened up new avenues for the further development of spintronic devices because this approach affords simple device miniaturisation and the potential for large-scale integration. [1][2][3] Conventionally, spin-transfer torque (STT) has been used to control the magnetisation in magnetic multilayer structures, 3,4 particularly for STT-magnetic random access memory (STT-MRAM), in which a spin-polarised current is injected in the direction perpendicular to the film plane. Recently, another type of spin-torque realised by spin-orbit coupling, the so-called spin-orbit torque (SOT), 5,6 has been widely investigated because it allows for efficient manipulation of the magnetisation using in-plane current, particularly during magnetisation switching 5,6 domain-wall and skyrmion motion. [7][8][9][10] SOT in heavy metal (HM)/ferromagnet (FM)/oxide heterostructures arises from the spin current induced by a charge current via the spin Hall effect (SHE) in the HM and/or the interfacial spin-orbit coupling (ISOC) effect at HM/FM interfaces. This spin current exerts a torque T on the local magnetisation as T = τ DLm × (ŷ ×m) + τ FLm ×ŷ, 11,12 where τ DL (τ FL ) denotes the dampinglike torque (field-like torque),m denotes the unit vector along the magnetisation direction, andŷ denotes the unit vector along the ARTICLE scitation.org/journal/apm direction perpendicular to both directions of charge current (x) and inversion symmetry breaking (ẑ). As τ DL governing SOT-induced magnetisation switching is known to arise mainly due to SHE in HMs, 6,13 most studies have focused on finding HM materials with a large effective spin Hall angle θ eff SH 14-24 for the realisation of energyefficient SOT-based spintronic devices. On the other hand, it has recently been reported that the HM/FM interface also strongly influences τ DL . [25][26][27][28][29][30][31][32][33] There are three examples; first, the magnitude and the sign of τ DL are changed by interface modifications, 26,27 which is attributed to the contribution of the ISOC effect to τ DL . Second, θ eff SH of the HM/FM bilayer strongly depends on the spin memory loss 28 or the spin transparency 29 of its interface, thereby indicating that material engineering of HM/FM bilayers can improve the SOT efficiency. Third, the interface itself generates a spin current and thus contributes to the SOT. 31,32 These results suggest that interface engineering could enable further enhancement in the SOT-induced magnetisation switching efficiency.
Among various HM/FM bilayers, a Pt/CoFeB bilayer is a promising candidate for spintronic device application for the following reasons: Pt is a highly conductive HM 33 when compared with other HMs such as β-W 34 or β-Ta, 35 thus reducing Joule heating and thereby power consumption. 33 A CoFeB alloy is a widely used FM in spintronic devices because of its large spin polarisation in conjunction with a crystalline MgO, 36 and a large tunnel magnetoresistance of up to 600% at room temperature has been achieved in CoFeB/MgO-based tunnel junctions. 37 Moreover, Pt/CoFeB/MgO structures have been widely used for voltage-driven spin devices, domain wall, and skyrmion motion devices. [38][39][40][41] However, θ eff SH of the CoFeB/Pt bilayer of 0.07 25 is significantly smaller than the highest reported value among Pt/FM bilayers: ∼0.2 for Pt/Co samples; 29 thus, it is of interest to further enhance the spin Hall angle in this structure.
In this study, we report significant SOT enhancement in Pt/CoFeB/MgO structures via modification of the Pt/CoFeB interface with the insertion of a thin Ti layer. We observe that the SOTinduced effective fields and magnetisation switching efficiency are doubled upon introducing a 1-nm-thick Ti layer at the Pt/CoFeB interface. The enhancement in SOT or θ eff SH is qualitatively consistent with ferromagnetic resonance (FMR) spin-pumping experiments, in which the inverse SHE voltage of the Pt/CoFeB structure increases by twofold upon Ti insertion. Furthermore, we find that the Ti interfacial layer markedly reduces the magnetic damping constant and the Pt proximity effect. Our results demonstrate the existence of a significant interfacial contribution and suggest that interface engineering can form an efficient approach to improve SOT-induced magnetisation switching.
Samples of Pt(5 nm)/Ti(t Ti )/Co 32 Fe 48 B 20 (CoFeB 1 nm)/ MgO(1.6 nm) structures were prepared by magnetron sputtering on thermally oxidised Si substrates with a base pressure of <4.0 × 10 −6 Pa (3.0 × 10 −8 Torr) at room temperature. The Ti thickness t Ti was varied from 0 nm to 3 nm, and the CoFeB thickness used in this study ranged from 0.9 to 1 nm, which range guarantees strong perpendicular magnetic anisotropy (PMA). Samples were annealed at 150 ○ C for 40 min under vacuum conditions to induce the PMA. The Hall bar structures with a width of 5 µm were fabricated by photolithography followed by Ar-ion-beam etching. All measurements were carried out at room temperature. For the spin-pumping measurements, the barbell-shaped samples were placed on a coplanar waveguide which generated an alternating magnetic field with a frequency ranging from 8 GHz to 14 GHz. The inverse spin Hall voltage V ISHE was measured as a function of the external magnetic field applied along the sample plane and normal to the barbell shape. 42 The soft x-ray absorption spectroscopy (XAS) at both the Co and Fe L edges was measured using the total electron yield method with the application of 96% circularly polarised incident X-rays under an applied magnetic field of 1.9 T. The X-ray propagation direction was parallel to the film normal and the 10 ○ -tilted magnetic field. The measurements were performed at the BL25SU beam-line at SPring8. Detailed experimental information is available in the literature (including hard XMCD measurements). 43,44 The hard XMCD measurements were performed at the BL39XU beam-line SPring8. A transmission-type diamond X-ray phase retarder with a thickness of 1.4 mm was used to achieve a high degree of circular polarisation (>95%) of X-rays. The X-ray fluorescence yield mode was used to record the X-ray absorption spectra under a magnetic field of 1.9 T. The X-ray propagation direction was parallel to the film normal and the magnetic field. During scanning around the Pt L 3 edge, the helicity of the X-rays was reversed at 0.5 Hz.
We first study the effect of the interfacial modification upon Ti insertion on SOT-induced magnetisation switching using Pt(5 nm)/Ti(t Ti )/CoFeB(1 nm)/MgO Hall bar structures, where t Ti is varied from 0 to 3 nm [ Fig. 1(a)]. Here, we remark that Ti has a weak spin-orbit coupling 42,45 and that all films considered in this study have perpendicular magnetic anisotropy (PMA). To perform SOT-induced switching experiments, we sweep an in-plane current pulse with a duration of 10 µs while measuring the anomalous Hall resistance RH between each pulse to detect the magnetisation direction. During the experiment, a magnetic field of 10 mT is applied along the current (+x) direction to achieve deterministic switching. 5,46 Figure 1(b) shows the SOT-induced switching results, from which we can infer two points of note. One is the same switching polarity for all samples irrespective of t Ti : under a positive magnetic field, positive (negative) current favours downward (upward) magnetisation. This switching polarity corresponds to the positive θ eff SH of Pt, 5,46 thereby indicating that the Ti insertion layer does not affect the sign of θ eff SH . The second point is that the critical switching current I C reduces by half when 1 nm of Ti is inserted at the Pt/CoFeB interface. On the other hand, I C becomes larger for samples with a thicker t Ti , which is attributed to increase in current shunting through the Ti layer with a small θ eff SH . Figure 1(c) depicts the reciprocals of the critical switching current (1/I C ) and SOT switching efficiency represented by the critical current density normalised by the anisotropy field (J C /B k ) −1 as a function of t Ti . 47 This confirms that the SOT switching efficiency improves in the regime where t Ti lies between 0.8 and 1.5 nm. We note that the SOT switching efficiency increases by 30% with ∼1 nm-Ti insertion, indicating that the reduction in Ic is substantially contributed by the decrease in B k . Here, we consider that the current is uniformly distributed in the Pt/Ti bilayers since the conductivities of the Pt and the Ti layers are similar to each other, but an order of magnitude larger than that of CoFeB. 22,48 The B k values are in the supplementary material. There has been a report about effect of Ta insertion between CoFeB and Pt on SOTdriven magnetization switching. 48 However, unlike in our study, these authors did not show an enhancement of the SOT effect with such a heavy-metal insertion layer. In order to quantify the effect of the Ti insertion on the SOT, we measure the SOT-induced effective magnetic fields of Pt/Ti(t Ti )/CoFeB/MgO structures using the harmonic lock-in technique. 12,13 As illustrated in Fig. 2(a), the application of an ac current with a frequency of ω generates the first-harmonic Hall voltage (V 1ω ), which indicates the z-component of magnetisation (Mz), and the second-harmonic Hall voltage (V 2ω ), which represents the oscillation of magnetisation (∆Mz) due to SOT-induced effective fields: damping-like (∆B DL ) and field-like (∆B FL ) effective fields. We obtain voltages V 1ω,2ω x and V 1ω,2ω y when a magnetic field is applied longitudinally (B = Bx) and transversely (B = By) with respect to the current direction, respectively. Figures 2(b) and 2(c) display the V 2ω x and V 2ω y values, respectively, for the Pt/Ti(t Ti )/CoFeB/MgO samples with t Ti = 1, 2, and 3 nm, while Fig. 2

(d) presents V 2ω
x (closed symbols) and V 2ω y (open symbols) as functions of B x(y) for the Pt/CoFeB/MgO structures without the Ti layer for a current density Je of 1 × 10 8 A/cm 2 . The insets depict V 1ω , which required to estimate the SOT-induced effective fields (∆B DL and ∆B FL ). ∆B DL and ∆B FL values for all samples are obtained using both the 1st and the 2nd harmonic data with the narrow field regime method (from −100 to 100 mT) which has been suggested by Kim, et al. 11 Here, planar Hall effect and thermal effect were also taken into consideration. 12,46 From this SOT analysis, we found that the magnitudes of both ∆B DL and ∆B FL in the film with t Ti = 1 nm are enhanced by a factor of 2 when compared with those of the film without Ti. This result is consistent with the trend of the switching data presented in Fig. 1.
We estimate θ eff SH using the conventional spin-transfer theory: 49,6 where ∆B DL (0) represents the zeroth order of ∆B DL , 12 ̵ h represents the reduced Planck constant, e represents the elemental charge of an electron, Ms represents the saturation magnetisation of the CoFeB layers, and Je represents the current density. The estimated value of the inserted 1-nm-thick Ti film is 0.19 ± 0.03, which is larger than that of the Pt/CoFeB film without Ti (0.13 ± 0.01). This result is consistent with the enhancement in the switching efficiency; however, it is in contrast to previous results, 25,29 wherein θ eff SH is reported to generally decrease upon insertion of an interfacial layer of a 3d metal.
We next examine the origin of the enhancement by performing spin-pumping experiments using ferromagnetic resonance (FMR) 46 for Pt (t Pt )/Ti (0, 1 nm)/CoFeB/MgO films, where the Pt thickness t Pt ranges from 2.5 to 16 nm. As schematically illustrated in Fig. 3(a), when the magnetisation is in resonance, a spin current is injected from the CoFeB layer into the Pt layer, which leads to increase in the effective damping constant ∆α eff in the CoFeB layer along with the generation of a transverse electric voltage via inverse SHE (V ISHE ) in the Pt layer. The former is related to the total spin current dissipated in the CoFeB layer, while the latter is due to the spin current injected into the Pt layer. A discrepancy between the two values, ∆α eff and V ISHE , can be induced if there is generation or extinction of spin current at the interface. 28 The V ISHE value is measured as a function of the external magnetic field, and ∆α eff is obtained from the variation in the line width of measured V ISHE spectra as a function of the FMR frequency [see Fig. 3

(b)]. Figures 3(c) and 3(d) show
the Pt thickness dependence of normalized V ISHE by sample resistance and ∆α eff , respectively. When the 1-nm-thick Ti interfacial layer is introduced, V ISHE increases, but ∆α eff significantly decreases. The doubled magnitude of V ISHE in the Ti-inserted samples demonstrates an increase in θ eff SH of the Pt/Ti/CoFeB structure, which is consistent with the enhanced SOT shown in Figs. 1 and 2. Moreover, the concurrent reduction in ∆α eff by insertion of 1-nm-thick Ti demonstrates that the increase in the precession angle is an origin of the enhanced V ISHE . 50,51 The decrease in ∆α eff indicates that the Ti insertion reduces a source of spin current depolarization at the Pt/CoFeB interface that is known to increase the magnetic damping. While the quantitative analysis requires further study, this result suggests that the Ti insertion improves the interfacial spin transparency of the spin current from HM to FM and thus increase θ eff SH or resultant SOT.
We next discuss the possible origins of the enhancement in θ eff SH by the interfacial modification: interface-induced spin current and modified spin transparency of the FM/HM interface. We first consider the modification of the ISOC effect since the Pt/CoFeB interface is replaced by Pt/Ti and Ti/CoFeB interfaces when the Ti layer is introduced. This insertion can enhance the SOT if the ISOC effect of the Pt/CoFeB interface is of opposite sign to the bulk SHE 26 or if the newly generated interface of Pt/Ti provides a positive contribution. 31 Here, we rule out the contribution of the Ti/CoFeB interface because our previous study showed a negligible SOT in the Ti/CoFeB/MgO structure. 27 To verify the modification of the ISOC effect by the Ti layer, we carry out X-ray magnetic circular dichroism (XMCD) measurements at the Fe and Co L 2,3 edges [ Fig. 4(a)]. The orbital-to-spin magnetic moment ratio (mo/ms eff ) of the Pt/Ti(t Ti )/CoFeB/MgO samples, which reflects the magnitude of the ISOC effect of the HM/FM interface, 52,53 is estimated using the sum rule. 54,55 We find that the ratio slightly decreases with increasing t Ti , thus indicating that the ISOC effect cannot be the reason for the enhanced SOT, as shown in Fig. 4(b). Here, we remark that in this respect, the Pt/Ti interface can be a source of spin current and the SOT; 31 however, further studies are required to clarify the spin current generation from the non-magnetic interface. Another possible origin is the improvement of spin current transmission at interfaces which is related to the spin memory loss 28 or spin transparency. 29 As the proximity effect in Pt is a source of spincurrent depolarisation, the insertion of the Ti layer may enhance the SOT by suppressing the induced moment in Pt. Thus, we investigate the effect of the Ti interfacial layer on the induced moment in Pt using hard XMCD analysis. Figure 4(c) shows XMCD spectra at the Pt L 3 edges, wherein a clear XMCD signal is observed, thus indicating a finite magnetic moment induced in the Pt 5d orbit for the samples without the Ti layer; however, this signal is completely eliminated by insertion of the 1-nm Ti layer. Thus, the insertion of the Ti layer can enhance the spin transparency and resultant SOT. We observed proximity-induced magnetic moment in the Ti insertion layer, which however is much smaller than that in the Pt layer. 47 The removal of the proximity-induced moment can also explain the reduction in ∆α eff upon insertion of the Ti interfacial layer [ Fig. 3(d)]. 56 This is also supported by results indicating that the SOT monotonously decays in Ta (5 nm)/Ti(t Ti )/CoFeB/MgO structures with increase in Ti interfacial layer thickness, where the proximity effect in Ta is negligible [ Fig. 4(d)].
In summary, we demonstrated a large enhancement in the SOT and SOT-induced switching efficiency in a Pt/CoFeB system ARTICLE scitation.org/journal/apm by means of interfacial modification involving Ti insertion. The FMR spin-pumping experiments together with XMCD investigations revealed that the enhancement can be attributed to an additional interface-generated spin current and/or improved spin transparency arising due to suppression of the induced moment in the Pt layer. While a quantitative analysis of the effect of the interfacial modification on the SOT requires further experimental and theoretical studies, our experimental results suggest that interface engineering is a promising approach to boost the efficiency of current-induced switching in SOT-based spintronic devices.
See supplementary material for information about the resistivity of each layer, anisotropy field of each film, quantitative SOT analysis, and the proximity effect in the Ti layer.