Electrical manipulation of ferromagnetic NiFe by antiferromagnetic IrMn

We demonstrate that an antiferromagnet can be employed for a highly efficient electrical manipulation of a ferromagnet. In our study, we use an electrical detection technique of the ferromagnetic resonance driven by an in-plane ac current in a NiFe/IrMn bilayer. At room temperature, we observe antidampinglike spin torque acting on the NiFe ferromagnet, generated by an in-plane current driven through the IrMn antiferromagnet. A large enhancement of the torque, characterized by an effective spin-Hall angle exceeding most heavy transition metals, correlates with the presence of the exchange-bias field at the NiFe/IrMn interface. It highlights that, in addition to the strong spin-orbit coupling, the antiferromagnetic order in IrMn governs the observed phenomenon.

Figure 1

Spin-orbit FMR experiment. (a) Schematic representation of the measurement technique. MW current-induced effective field $\mathbf{h}(h_x,h_y,h_z)$ drives magnetization precession around the total field ${\mathbf{H}}_{\mathrm{eff}}$. Precessing magnetization results in oscillating resistance due to AMR. This mixes with the oscillating current of the same frequency resulting in a measurable dc voltage. (b) Resonance curve decomposed into symmetric and antisymmetric components measured in a bar with 2-nm IrMn at frequency of 17.9 GHz.

Figure 2

AFM-induced torque and its symmetries. (a) Comparison of resonance curves measured in samples with and without the IrMn layer. Both measurements are performed at 17.9 GHz, $\theta ={45}^{\circ }$. Antisymmetric components are normalized to 1 $\mu \mathrm{V}$. (b) Symmetries of $V_{\mathrm{sym}}$ and $V_{\mathrm{asy}}$ for the sample with 2-nm IrMn. Solid lines are fits to Eqs. (1) and (2).

Figure 3

AFM thickness dependence of current-induced fields and anisotropies. (a) $h_z, h_y$, and calculated Oersted field $h_{\mathrm{Oe}}$ for 1.8-$\mu \mathrm{m}$ wide bars with different IrMn thicknesses, as well as the sample with the 2-nm Cu spacer layer. The results are normalized to a current density of ${10}^7\mathrm{A}/{\mathrm{cm}}^2$ in IrMn. The shaded area around $h_{\mathrm{Oe}}$ is the error due to uncertainties in layer resistivities, whereas the error bars of $h_z$ and $h_y$ are due to the standard errors from the fitting of the symmetries, AMR and MW current. The systematic uncertainties in layer resistivities have not been included in the error bars of $h_z$ and $h_y$, however, this uncertainty, which is approximately 20%, is included in the values of effective spin-Hall angles in the main text. The dotted line is the estimated spin-Hall effect contribution to $h_z$ for ${\lambda }_{sd}=1$ nm. (b) Angle dependencies of resonance field for the samples with 2- and 4-nm IrMn thicknesses, as well as the sample with the 2-nm Cu spacer layer. Solid lines are fits taking into account unidirectional, uniaxial and rotational anisotropies. (c) IrMn thickness dependence of the exchange bias and the rotational anisotropy extracted from the fits in (a). (d) IrMn thickness dependence of exchange bias and coercivity extracted from hysteresis loops measured using MOKE and AMR switching.

Figure 4

DC bias dependence of the FMR linewidth. (a) Change of FMR linewidth with dc current for the IrMn(4)/Cu(2)/NiFe(4) structure measured at $\omega /2\pi =8$ GHz and (b) IrMn(2)/NiFe(4) structure measured at $\omega /2\pi =14.1$ GHz, for two different directions of magnetization with respect to the current. The data points are extracted using the linewidth difference between positive and negative bias currents.

Figure 5

Independence on the exchange bias direction. Angle dependence of the resonance field for two different 4-nm IrMn bars (top) and corresponding symmetric components of the measured dc voltage (bottom). Although the exchange bias is substantial and has different directions for the two bars, the angle dependence of the symmetric component of the Lorentzian, corresponding to $h_z$, is not affected.

Figure 6

Temperature dependence of current-induced fields and anisotropies. (a) Temperature dependence of the $h_z/h_y$ ratio for the sample with 2-nm IrMn. The inset shows the temperature dependence of the AMR and total resistance of the bar. (b) Temperature dependence of the exchange bias and the coercivity for the same sample extracted from AMR switching measurements.

Figure 7

AFM-induced torque and Gilbert damping. (a) Frequency dependence of the FMR linewidth for the sample with 6-nm IrMn. The slope of the linear fit allows us to extract the effective Gilbert damping $\alpha =\gamma {\mu }_0(\partial \mathrm{\Delta }H/\partial \omega )$, where $\gamma /2\pi$ = 28 GHz/T. (b) Current-induced out-of-plane field $h_z$ plotted against the effective Gilbert damping for samples with different IrMn thicknesses. (c) Gilbert damping is proportional to the square of the exchange bias, suggesting that one of the main damping mechanisms in our samples is the two-magnon scattering due to the inhomogeneity of the field at the FM/AFM interface induced by the exchange anisotropy.

Figure 8

(a) Comparison of resistance change due to heating caused by dc (left) and MW (right) currents. The dc measurement is symmetric with respect to 0 current. (b) MW current vs square root of applied MW power obtained from the heating calibration. The solid line is a linear fit to the data (circles).

Figure 9

(a) Resistances of bars with different length/width ratios. The fit to a line gives the average contact resistance. (b) Resistances of bars with the same dimension ratio 60 but different IrMn thicknesses. Resistances of 3–8-nm samples are fitted to Eq. (B1) (solid like). (c) Resistances of bars with different Cu spacer thicknesses fitted to Eq. (B1). (d) The magnitude of intrinsic AMR of NiFe in samples with different IrMn thicknesses, extracted using Eq. (B5). In the inset, we show a typical measurement for extracting the total AMR $\mathrm{\Delta }{R}_{\mathrm{tot}}$.

Figure 10

Schematic representation of the Oersted fields induced by the current in the multilayer.

Figure 11

(a) A resonance curve measured in a Ru(3)Py(4) bar at 17.9 GHz, decomposed into symmetric and antisymmetric Lorentzians. (b) Resonances measured in the 3-nm IrMn sample at 5 K and in the 2-nm IrMn sample at room temperature (295 K), at 16.5 and 18.6 GHz microwave frequencies, respectively. The antisymmetric components are normalized to 1 $\mu \mathrm{V}$ (not shown), and the symmetric components are shown with dotted lines.

Figure 12

(a) Dependence of $V_{\mathrm{sym}}$ and $V_{\mathrm{asy}}$ and their ratio on applied microwave power for the 2-nm IrMn sample. Solid lines are linear fits. (b) Frequency dependence of the $h_z/h_y$ ratio for the 3-nm IrMn sample. (c) $h_z/h_y$ ratio for 2-nm IrMn samples with different dimensions and measured in two different setups. The bar dimensions are 1.8 $\mu \mathrm{m}\times 38\mu \mathrm{m}$, 4 $\mu \mathrm{m}\times 240\mu \mathrm{m}$, 1 $\mu \mathrm{m}\times 10\mu \mathrm{m}$, and $500\mathrm{nm}\times 5\mu \mathrm{m}$.