Nanoscale imaging of magnetization reversal driven by spin-orbit torque

We use scanning electron microscopy with polarization analysis to image deterministic, spin-orbit torque-driven magnetization reversal of in-plane magnetized CoFeB rectangles in zero applied magnetic field. The spin-orbit torque is generated by running a current through heavy metal microstrips, either Pt or Ta, upon which the CoFeB rectangles are deposited. We image the CoFeB magnetization before and after a current pulse to see the effect of spin-orbit torque on the magnetic nanostructure. The observed changes in magnetic structure can be complex, deviating significantly from a simple macrospin approximation, especially in larger elements. Overall, however, the directions of the magnetization reversal in the Pt and Ta devices are opposite, consistent with the opposite signs of the spin Hall angles of these materials. Our results elucidate the effects of current density, geometry, and magnetic domain structure on magnetization switching driven by spin-orbit torque.

Figure 1

(a) A scanning electron micrograph of one of the devices studied in this paper. The bright regions in the upper right and lower left corners are Au contact pads. Between the contact pads are $4-\mu \mathrm{m}-\mathrm{wide}$ Ta strips with transverse CoFeB (lighter contrast) rectangles on top. (b) Sample geometry for direct imaging of in-plane magnetization reversal driven by spin-orbit torque. Rectangles of CoFeB (shaded black or white depending on the magnetization orientation as in SEMPA images) are fabricated on top of strips of Pt or Ta. When a current pulse is sent through a strip, a transverse spin current (represented by the dashed red lines and arrows) is generated by the spin-orbit torque of the heavy metal. The spin currents generated in Pt and Ta are opposite: A current pulse in the $+x$ direction will rotate the CoFeB magnetization in the $–y$ direction for the Pt sample, but the same pulse in the Ta sample will rotate the CoFeB magnetization in the $+y$ direction. Panels (c) and (d) show in-plane and out-of-plane vibrating sample magnetometry hysteresis loops for unpatterned witness films of Pt(6)/CoFeB(2)/Pt(2) and Ta(6)/CoFeB(2)/Pt(2), respectively, demonstrating the in-plane easy axis. Uncertainties are derived from the uncertainty from the VSM and are smaller than the data markers.

Figure 2

SEMPA images of in-plane magnetization reversal driven by the spin-orbit torques of Pt and Ta. The magnetization direction is indicated by the color wheel inset in (a) and by small black arrows, and the edges of the heavy metal strips are indicated by the dashed lines. The top two panels show the magnetization of CoFeB rectangles on top of Pt strips before (a) and after (b) a $0.3-\mu \mathrm{s}(3\times {10}^{12})-\mathrm{A}/{\mathrm{m}}^2$ current pulse in the $+x$ direction. The magnetization is switched from the $+y$ direction to the $-y$ direction [axes indicated in (a)]. The bottom two panels show the magnetization of CoFeB rectangles on top of Ta strips before (c) and after (d) a $0.3-\mu \mathrm{s}(8\times {10}^{11})-\mathrm{A}/{\mathrm{m}}^2$ current pulse in the $+x$ direction. In this case, the magnetization is switched from the $–y$ direction to the $+y$ direction, opposite the direction of the switching seen with the Pt strips.

Figure 3

Simultaneous measurement of magnetization reversal in 30 different CoFeB rectangles (on Ta strips). These figures are difference images generated by subtracting subsequent SEMPA images of one of the magnetization components from each other. White contrast represents magnetization that has reversed from $–y(–x)$ to $+y(+x)$, black contrast represents magnetization that has reversed from $+y(+x)$ to $–y(–x)$, and gray contrast represents nonmagnetic areas as well as regions in which the magnetization did not change between images. Panels (a) and (b) show changes in the $y$-component magnetization after $(8\times {10}^{11})-\mathrm{A}/{\mathrm{m}}^2$ current pulses in the $+x$ and $–x$ directions, respectively. Panels (c) and (d) show the corresponding changes in the $x$-component magnetization. Panels (e) and (f) show the $y$-component magnetization after slightly smaller $(6\times {10}^{11})-\mathrm{A}/{\mathrm{m}}^2$ current pulses in the $+x$ and $–x$ directions, respectively. Here the spin-orbit torque is weaker and can only reverse some of the magnetization in the wider CoFeB rectangles.

Figure 4

(a) The fraction of the $y$ component of CoFeB magnetization reversed $\left(f_y\right)$ vs current density for the $(2.0\times 4.0)-\mu \mathrm{m}$ CoFeB rectangles on Pt strips. The open loop shows clear deterministic and reversible spin-orbit torque-driven magnetization reversal in zero applied magnetic field. (b) The magnitude of $f_y$ vs current density for all three sizes of CoFeB rectangles (averaged over switching from $+y$ to $-y$ and vice versa). Filled symbols indicate switching with the symmetry expected from the spin-orbit torque and Oersted field (which are the same for the case of Pt), and open symbols indicate switching with the opposite symmetry. The critical currents extracted from these data are presented in Supplemental Material Tables 1 and 2 [34]. Panels (c) and (d) show analogous data for the Ta/CoFeB sample. In this case, the spin-orbit torque and the Oersted field have opposite symmetries, and the observed switching symmetry is consistent with the spin-orbit torque (open symbols) and not the Oersted field (filled symbols). Vertical bars indicate single standard uncertainties determined by the standard deviations of all ten elements measured for each size.