The interfacial Dzyaloshinskii-Moriya interaction and spin-orbit torque driven domain wall motion in Co/Ni multi-layers with a Pt_$x$(Ir,Au)_1−$x$ seedlayer Open Access

The union of strong spin-orbit coupling in 5d heavy metals and conventional ferromagnetism in 3d transition metals has lead to new spin phenomena in complex magnetic multi-layers. This includes the interfacial Dzyaloshinskii-Moriya interaction, which is now well-established to stabilize chiral magnetic domain walls and Skyrmions with the Néel-type configuration,^1,2 and the spin Hall effect^3,4 that enables their efficient manipulation with electric current.^5–8 Collectively, these effects are referred to as spin-orbit torques when used to manipulate magnetic objects although the precise mechanism by which this occurs remains a topic of some controversy.⁹ Regardless, the emergence of such spin-orbit coupled systems has sparked excitement about their prospect for use in future spintronic applications, which motivates the present work.¹⁰ 

Here, we measure the interfacial DMI and current-induced domain wall motion (CIDWM) in Co/Ni multi-layers as a function of composition in a Pt_x(Ir,Au)_1−x alloy buffer layer. Until recently, the impact of 5d heavy metal alloying on spin-orbit coupling phenomena had only been explored theoretically.¹¹ We have previously have shown that alloying a Pt buffer layer with Ir leads to a monotonic reduction of the interfacial DMI vector measured from the asymmetric expansion of magnetic bubble domains.¹² That experimental data was fit to a modified dispersive stiffness model,^12,13 which gave values for the DMI vector that matched strikingly well with independent measurements performed using Brillouin light scattering (BLS). Having validated the experimental methodology, this work serves to expand the heavy metal composition space for controlling interfacial DMI and directly measure its impact on the efficiency of current-induced domain wall motion.

Magnetic multi-layers were grown by DC magnetron sputtering onto thermally oxidized Si wafers with a working pressure of 2.5 mTorr argon with an additional 0.5 mTorr N₂ in the case of insulating TaN deposition. The target diameters were 5″ with a working distance fixed at 4.75″. The film stacks deposited were Substrate/TaN(3)/Pt(3.5)/Pt_x(Ir,Au)_1−x(1.2)/[Co(0.2)/Ni(0.6)]₂/Co(0.2)/Ta(0.8)/TaN(6) with units in nm. The Pt_x(Ir,Au)_1−x buffer layer was deposited using the alternating wedge growth technique to mimic the cosputtering process as in Ref. 12. M − H loops measured using alternating gradient field magnetometry (AGFM) and vibrating sample magnetometry (VSM) across the composition gradient indicate a saturation magnetization, M_s ∼ 645kA/m, and strong in-plane anisotropy with μ₀H_k = 1 − 1.3T. Details on the structural characterization of similar Co/Ni multi-layer films can be found in Ref. 14. The interfacial DMI vector was determined using the bubble domain expansion technique operating in the creep regime.^{2,12,13,15–17} Selected areas of the sample were damaged using a focused ion beam (FIB) to serve as bubble nucleation sites as in Lau et al.¹⁷ Growth velocities in the presence of an in-plane magnetic field were determined from the domain wall displacement following a 1-20msec perpendicular field pulse. The values of DMI were extracted from experimental measurements of velocity vs field using the augmented dispersive stiffness model, which accounts for contributions from a chiral-damping like effect.¹² This model has previously shown excellent agreement with independent measurements performed by Brillouin lights scattering (BLS) for the case or Pt_xIr_1−x buffer layers. Measurements of current driven domain wall velocity were performed on nanowires patterned by a combination of photolithography and ion milling. Additional details on nanowire fabrication can be found in Bromberg et al.¹⁰ A series of 100ns current pulses were injected through a coplanar waveguide with velocities determined based on the displacement of the wall over a range of current. Current densities were determined from normalizing current by the full metal layer thickness, which does not vary among the samples examined here.

Kerr microscopy images of asymmetric bubble growth in the case of Pt_xAu_1−x is shown in figure 1. As with prior studies on Co/Ni multi-layers we note a marked asymmetry to the curve including a critical field where the velocities of the left and right domain walls cross. This behavior is fit well to the augmented dispersive stiffness model across the spectrum. In the case of both Pt-Ir¹² and Pt-Au alloy buffer layers, there is a monotonic decrease in the interfacial DMI vector as the composition deviates from Pt (figure 2). We note that five separate samples with different buffer layers (Ir, Pt_xIr_1−x, Pt, Pt_xAu_1−x, and Au) were compared to produce this data. It is likely that the abrupt change associated with a small amount of Au is explained in part by variation in sample quality rather than an intrinsic impact of Au on DMI. Regardless, the trend seems clear across the composition spectrum that a peak occurs for the case of pure Pt. This result is largely consistent with first principles predictions performed recently by Hanke et al.¹¹ 

Having established a significant dependence of interfacial DMI on buffer layer composition, we now turn to its impact on current-driven domain wall velocities. One intent of the present work is to decouple the impact of the spin Hall angle from that of the interfacial DMI vector. Although the changing buffer layer composition will have some impact on the effective spin Hall angle, we speculate that it will remain largely governed by the 3.5nm Pt seedlayer. Conversely, it has previously been shown that the DMI vector is determined predominantly by the interface and the impact of the Pt seedlayer should be minimal in this regard. Current-driven DW velocity as a function of current density is shown in figure 3 for the case of Ir and Au alloying. Although the threshold velocities are comparable, it is evidence that Au leads to a rapid drop in the efficiency of CIDWM. Combining the results of figures 2 and 3, we show the CIDWM as a function of measured interfacial DMI in figure 4. The blue line is approximately where we expect the DW to transition into a purely Néel-type DW. In early work in this area, it was speculated that the wall chirality could govern the velocity of the wall. It is evident that a direct correlation exists between CIDWM and DMI that extends beyond this value confirming that it is the strength of the DMI rather than the wall chirality that should be considered in determining the efficiency of CIDWM.

We have shown that the addition of either Ir or Au to a Pt buffer layer in Co/Ni multi-layers decreases the interfacial DMI as measured by asymmetric domain expansion. This result is consistent with first principles predictions of DMI in related systems.¹¹ The measured value of DMI is directly correlated to the current-induced domain wall velocity confirming that increasing DMI beyond the critical value to stabilize Néel DWs continues to improve device performance.

This work is financially supported by the Defense Advanced Research Project Agency (DARPA) program on Topological Excitations in Electronics (TEE) under grant number D18AP00011 and also funded (in part) by the Dowd Fellowship from the College of Engineering at Carnegie Mellon University. The authors would like to thank Philip and Marsha Dowd for their financial support and encouragement.