Current-driven asymmetric magnetization switching in perpendicularly magnetized CoFeB/MgO heterostructures

The flow of in-plane current through ultrathin magnetic heterostructures can cause magnetization switching or domain-wall nucleation owing to bulk and interfacial effects. Within the magnetic layer, the current can create magnetic instabilities via spin transfer torques (STT). At interface(s), spin current generated from the spin Hall effect in a neighboring layer can exert torques, referred to as the spin Hall torques, on the magnetic moments. Here, we study current-induced magnetization switching in perpendicularly magnetized CoFeB/MgO heterostructures with a heavy metal (HM) underlayer. Depending on the thickness of the HM underlayer, we find distinct differences in the in-plane field dependence of the threshold switching current. The STT is likely responsible for the magnetization reversal for the thinner underlayer films whereas the spin Hall torques cause the switching for thicker underlayer films. For the latter, we find differences in the switching current for positive and negative currents and initial magnetization directions. We find that the growth process during the film deposition introduces an anisotropy that breaks the symmetry of the system and causes the asymmetric switching. The presence of such symmetry-breaking anisotropy enables deterministic magnetization switching at zero external fields.

Figure 1

(a) Optical microscopy image of the wire used to study current-induced magnetization switching. The dark regions indicate the magnetic film, the bright regions correspond to the substrate surface and the yellow regions represent the $\mathrm{Ta}|\mathrm{Au}$ electrodes. A pulse generator is connected to the left electrode. (b) Out-of-plane hysteresis loops measured using Kerr microscopy for Sub.$|d \mathrm{TaN}\left|1 \mathrm{CoFeB}\right|2 \mathrm{MgO}|1$ Ta: $d=3.6\mathrm{nm}$ (red circles) and $d=0.5\mathrm{nm}$ (black squares). (c), (d) Magnetization switching probability as a function of pulse amplitude for initial magnetization configurations pointing along $+z$ (black squares) and $-z$ (red circles) for the two devices shown in (b). Positive and negative probability corresponds to initial magnetization direction pointing along $+z$ and $-z$, respectively. A pulse train consisting of five 100-ns-long pulses is applied. Representative Kerr images captured after the application of $\pm 50$ V (c) and $\pm 35$ V (d) are included at the corresponding corners of each panel. Results are from substrates placed in the “left” position defined in Fig. 5. No external magnetic field is applied during the current pulse application for the results shown in (c) and (d).

Figure 2

Threshold current density $\left({J_N}^C\right)$ as a function of TaN underlayer thickness. The initial magnetization direction points along $-z$ (a) and $+z$ (b). Solid and open symbols represent positive and negative ${J_N}^C$, respectively. A pulse train consisting of five 100-ns-long pulses is applied. Results are from substrates placed in the “left” position defined in Fig. 5. No external magnetic field is applied during the current pulse application.

Figure 3

In-plane field dependence of the threshold current density $\left({J_N}^C\right)$. The field direction is along (a), (c) and transverse to (b), (d) the current flow. The underlayer is TaN: its thickness is 0.5 nm (a), (b) and 6.6 nm (c), (d). Black squares and red circles represent initial magnetization direction along $+z$ and $-z$, respectively. A pulse train consisting of five 100-ns-long pulses is applied. Results are from substrates placed in the “left” position defined in Fig. 5.

Figure 4

The fieldlike $\left(\mathrm{\Delta }{H}_Y\right)$ (a) and the dampinglike $\left(\mathrm{\Delta }{H}_X\right)$ (b) components of the current-induced effective field plotted against the TaN underlayer thickness (source: [Ref. 21]). Black squares and red circles correspond to magnetization directed along $+z$ and $-z$, respectively. The effective field is normalized by the current density $J_N$ that flows through the TaN layer. (c) Ratio of the fieldlike component to the dampinglike component $b_J/a_J=-m_Z\mathrm{\Delta }{H}_Y/\mathrm{\Delta }{H}_X$ plotted against the TaN underlayer thickness. (d) TaN thickness dependence of the offset field ${H_X}^{*}$. Solid and open symbols correspond to ${H_X}^{*}$ estimated using positive and negative currents.

Figure 5

(a) Schematic illustration of inside the sputtering chamber where the relative position of the substrates and the target is shown. Three $\sim 1\times 1\mathrm{c}{\mathrm{m}}^2$ square substrates, separated by ∼0.15 cm along the $y$ direction, are placed ∼10 cm away from the target. (b), (c) Kerr images after application of ±32 V voltage pulses for devices made of Sub.$|3.6$ nm $\mathrm{TaN}\left|1 \mathrm{CoFeB}\right|2 \mathrm{MgO}|1$ Ta substrates placed at different positions: (b) “Left” position and (c) “right” position defined in (a). The top and bottom images correspond to images when positive and negative voltage pulses are applied, respectively. (d) Magnetization switching probability as a function of pulse amplitude for the two devices shown in (b) and (c). The initial magnetization direction points along $-z$. A pulse train consisting of five 100-ns-long pulses is applied for (b)–(d). No external magnetic field is applied during the current pulse application.

Figure 6

Pulse amplitude dependence of magnetization switching probability for Sub.$\left|2.9 \mathrm{TaN}\right|1 \mathrm{CoFeB}\left|2 \mathrm{MgO}\right|1$ Ta (units in nm). The patterned wires' long axis is directed along $x$ (a) and $y$ (b). Results are from substrates placed in the “left” position defined in Fig. 5. Positive and negative probability corresponds to initial magnetization direction pointing along $+z$ and $-z$, respectively. No external magnetic field is applied during the current pulse application.

Figure 7

Magnetization switching probability as a function of pulse amplitude for initial magnetization configurations pointing along $+z$ (black squares) and $-z$ (red circles) for devices with different heavy metal underlayers. The films are Sub.$|d \mathrm{X}\left|1 \mathrm{CoFeB}\right|2 \mathrm{MgO}|1$ (units in nanometers), with $\mathrm{X}=5.9\mathrm{nm}$ Hf (a) and 3.1 nm W (b). A pulse train consisting of five 100-ns-long pulses is applied. Positive and negative probability corresponds to initial magnetization direction pointing along $+z$ and $-z$, respectively. Results are from substrates placed in the “left” position defined in Fig. 5. No external magnetic field is applied during the current pulse application.

Figure 8

Magnetization switching probability as a function of pulse amplitude for Sub.$\left|2.9 \mathrm{TaN}\right|1 \mathrm{CoFeB}\left|2 \mathrm{MgO}\right|1$ Ta (units in nm). The out-of-plane field $H_Z$ is varied: $H_Z\sim -5\left(\mathrm{a}\right)$, ∼0 (b), and ∼5 Oe (c). Positive and negative probability corresponds to initial magnetization direction pointing along $+z$ and $-z$, respectively. Results are from substrates placed in the “left” position defined in Fig. 5.

Figure 9

(a) Threshold current density $\left({J_N}^C\right)$ vs pulse length $\left(t\right)$ at zero external field for Sub.$|3.6$ nm $\mathrm{TaN}\left|1 \mathrm{CoFeB}\right|2 \mathrm{MgO}|1$ Ta. A pulse train consisting of five $t$ ns-long pulses, each separated by 10 ms, is applied. Black squares and red circles show ${J_N}^C$ when the initial magnetization direction is along $+z$ and $-z$, respectively. (b) Sequences of voltage pulses applied to the wire (top panel) and the resulting Kerr contrast (ΔI) calculated from the Kerr images. The corresponding magnetic state (1: along $+z$, −1: along $-z$) is shown in the right axis. A pulse train consisting of five 100-ns-long pulses, each separated by 10 ms, is applied at each pulse shown in the top panel. Middle and bottom panels of (b) show changes in the Kerr contrast for initial magnetization pointing along $+z$ and $-z$ at the beginning of the sequence, respectively. No external magnetic field is applied during the current pulse application. Results are from substrates placed in the “left” position defined in Fig. 5.

Figure 10

(a), (b) $z$ component of the equilibrium magnetization when current and in-plane magnetic field are turned on plotted as a function of $a_J$, the dampinglike component of the spin Hall torque. The fieldlike component of the spin Hall torque $b_J$ is set to $-a_J$ (a) and $a_J$ (b). The horizontal blue dashed lines indicate $|m_Z|=0.15$, which is used to define ${a_J}^C$. (c)–(h) ${a_J}^C$ as a function of $H_X$ (c)–(e) and $H_Y$ (f)–(h). The fieldlike component $b_J$ is varied: $b_J=-a_J$ (c), (f), $b_J=0$ (d), (g), and $b_J=a_J$ (e), (h). For all plots, black squares and red circles represent calculation results when the initial magnetization direction points along $+z$ and $-z$, respectively. $H_K=528\mathrm{Oe},\alpha =0.05$, the uniaxial anisotropy axis (direction defined by a unit vector $\widehat{k}$) is tilted 2° toward the $y$ axis, i.e., $\widehat{k}=\left(0,\sin \beta ,\cos \beta \right)$ with $\beta =2\mathrm{deg}$.

Figure 11

Micromagnetic simulations of spin Hall torque driven magnetization switching. (a), (b) $a_J$ (the dampinglike component of the spin Hall torque) dependence of the $z$ component of magnetization $\left(m_Z\right)$ at the end of 1-ns pulse (a) and the switching probability calculated from the magnetic state 20 ns after the pulse is turned off (b). Black squares and red circles represent initial magnetization along $+z$ and $-z$, respectively. Parameters used in the calculations are: saturation magnetization $M_S=1250\mathrm{emu}/\mathrm{c}{\mathrm{m}}^3$, exchange constant $A=3.1\times {10}^{-6}\mathrm{erg}/\mathrm{cm}$, uniaxial magnetic anisotropy energy $K=10.15\times {10}^6\mathrm{erg}/\mathrm{c}{\mathrm{m}}^3$, Gilbert damping $\alpha =0.05$, and the fieldlike component of the spin Hall torque $b_J=a_J$. The dimension of the simulated element is $2000\times 500\times 1\mathrm{n}{\mathrm{m}}^3$ with a discretization cell of $\sim 2\times 2\times 1\mathrm{n}{\mathrm{m}}^3$. The anisotropy axis is tilted along the $yz$ plane by 1°. Inset to (a): simulated magnetization image 20 ns after a pulse of $a_J=368.6\mathrm{Oe}$ is turned off: the initial magnetization is along $-z$.