Optically Detected Spin-Orbit Torque Ferromagnetic Resonance in an In-Plane Magnetized Ellipse

Abstract

Spin-orbit torque (SOT) is currently the subject of intense research activity due to its potential for switching storage elements in magnetic random access memory[1], generating microwave frequency spin wave auto-oscillations in spin Hall nano-oscillators (SHNOs)[2], and manipulating magnetic textures, e.g. skyrmions[3] in magnetic logic devices. Device operation is underpinned by the spin Hall effect, in which a spin current can be efficiently generated at the interface between a ferromagnetic metal (FM) and a current carrying heavy metal (HM) underlayer due to strong spin-orbit coupling in the HM. The spin current may then exert a SOT on the magnetization of the FM, exciting magnetization dynamics central to device function.A recent study of nano-contacted SHNOs[2] revealed a spatial dependence of the SOTs due to the current distribution within the HM of the device. Moreover, micron and sub-micron scale ferromagnetic elements are known to support non-uniform equilibrium states resulting from strong demagnetizing fields[4]. Since the in-plane (Slonczewski) spin-torque term of the Landau-Lifshitz equation of motion depends upon the relative local orientation of the magnetization and the injected spin polarization, the resulting SOT can be spatially dependent. At the same time the Oersted (Oe) field generated by the DC and RF currents (IDC, IRF) in the HM will also lead to a spatially dependent DC deflection and RF excitation of the magnetization respectively. Therefore, to explore the influence of non-uniform SOTs at the nanoscale, a spatially resolved probe of the resulting magnetization dynamics is required.In this work time-resolved scanning Kerr microscopy (TRSKM) was used to acquire spatiotemporally resolved SOT ferromagnetic resonance (FMR) spectra from an in-plane magnetized Co40Fe40B20(2 nm) ellipse of size 2×0.8 μm2 positioned at the centre of a Pt(6 nm) Hall cross[5]. DC and RF currents were combined using a bias tee and passed through the 2 μm wide Pt lead perpendicular to the long axis of the ellipse, such that the resulting spin polarisation (σ(t)) and Oe-fields (h(t)) were parallel to the long axis of the ellipse, Figure 1(a). Preliminary time-resolved polar Kerr measurements in response to a broadband current pulse allowed the resonance frequency (~2.7 GHz) to be determined for an in-plane bias magnetic field (HB) of 200 Oe, while the time-resolved measurements at HB = 200 Oe in response to an RF current with frequency (fRF) of 2.72 GHz allowed the time delay of node and antinodes of precession at resonance to be determined. Sweeping the magnetic field with the time delay set to a node or an antinode allowed the real or imaginary components of the dynamic susceptibility to be probed respectively. Time-resolved polar Kerr images, Figs. 1(b) and 1(c) revealed that the dynamics at resonance were quasi-uniform at the center of the ellipse, while weaker signal at the edge was consistent with the spatial resolution of ~400 nm. SOT-FMR spectra were therefore acquired from the center of the ellipse for a range of field angles θB and IDC, while fRF was set to 3.2 GHz so that the resonance field was well separated from the field at which switching of the equilibrium magnetization occurred.Optically-detected SOT-FMR spectra acquired with HB applied along the HA (θB = 90°), and IDC = 0 mA showed negligible influence of SOT, i.e. symmetric FMR peaks with respect to field with no broadening or narrowing when magnetized in opposite HA directions. A macrospin calculation reproduced the peaks and a damping parameter α of 0.03 was extracted, similar to the previously reported value for these devices[1]. The coefficient of the in-plane torque term (ST) was assumed to be zero in the HA calculations with IDC = 0 mA. The value of IDC was then increased from 1 mA to 10 mA and spectra measured with HB along the HA, which again showed symmetric FMR peaks. However, when the bias field was applied 30° to either side of the HA, marked asymmetry in both amplitude and linewidth were observed, in addition to a combined one- and two-fold angular dependence of the resonance field due to the DC Oe-field and shape anisotropy respectively. For a particular field history, the linewidth exhibited a marked crossover from broad to narrow as θB was changed from +30° to -30° from the HA. These observations were reproduced by the macrospin calculations, but failed to reproduce the linewidth narrowing/broadening for a single value of ST with α = 0.03, while a smaller value of α = 0.025 reproduced the linewidth of all FMR peaks with ST = (6.75±0.75)×10-7 Oe A-1 cm2 .This work paves the way for spatially resolved characterization of SOT devices at the deep nanoscale, e.g. using near-field magneto-optical techniques[6,7], that is needed to achieve greater understanding of SOT-induced dynamics. **