Influence of MgO Tunnel Barrier thickness in 3-terminal Spin Hall Nano-Oscillators

The Spin Hall Effect (SHE) can be used to generate pure spin currents, capable of exerting a spin transfer torque (STT) that induces oscillations in a ferromagnetic layer[1–3]. Until now, reported publications concern the STT effect induced from a spin Hall current on a fixed MgO barrier [1], [4]. However, the influence of $\mathrm {R}\times \mathrm {A}$ on a performance of spin Hall current induced STT oscillations not been studied yet. To this end, we study the effect of spin Hall current induced STT on variable (wedge) MgO thickness. In this work, a 3-terminal MTJ stack incorporating MgO-wedge was deposited on a 200 mm, Si/100 nm Al 2O3 wafer in a Timaris Singulus PVD deposition system, leading to a variable $R \times A$ over the wafer from $1 \Omega \mu \mathrm {m}^{2}$ up to $70 \Omega \mu \mathrm {m}^{2}$. The deposited stack was consisted of the following materials: 15 Ta/1.4 Co 0.4 Fe $_{0.4} \mathrm {B}_{0.2} /$[Wedged] MgO /2.2 Co 0.4 Fe $_{0.4} \mathrm {B}_{0.2} /0.85$ Ru/2.0 Co 0.7 Fe $_{0.3} /20$ Ir 0.2 Mn $_{0.8} /5$ Ru (thickness in nanometer). The stack was subsequently patterned into 30 different circular and ellipse-shaped nanopillars. The nanopillars patterning were done using electron beam lithography followed by etching in ion beam milling. The Hall bar (Ta layer $( 24 \mu \mathrm {m}$ long and $1 \mu \mathrm {m}$ wide)) geometry was engineered targeting a small DC current through the Ta line $( I_{Ta})$ should stimulate a magnetization dynamic effects caused by the SHE. Subsequent of nanofabrication, the nanopillars were measured in an automated 4-point geometry for statistical measurement. The devices were then measured in high-frequency setup (3-terminal device geometry) for pure spin Hall nano-oscillator measurement. Fig. 1(a) shows the measured TMR (%) distribution as a function of $R \times A$ in the final devices. The $R \times A$ ranging from $20 \Omega \mu \mathrm {m}^{2}$ to $100 \Omega \mu \mathrm {m}^{2}$, the change of TMR (%) is relatively small (160 % to 200 %) and exponentially decreases as the $R \times A$ decreases below $20 \Omega \mu \mathrm {m}^{2}$. The exponential decrease in TMR (%) in low $R \times A$ region can be explained by the barrier imperfection (pin holes) due to thin MgO barrier. The 10% devices TMR (%) found to be below 40%, this is due to the incomplete planarization process of the sample. However, the TMR ratio of 90% devices was TMR (%) of above 80% clearly indicates that the nanofabrication process was successful. The Fig. 1(b) represents four TC for $R \times A$ values of 55, 11, 5 and $1.8 \Omega \mu \mathrm {m}^{2}$ of nanopillar size: 200 nm. The plotted curves clearly show that for a high $\mathrm {R}\times \mathrm {A}$ value $( 55 \mu \mathrm {m}^{2})$ the characteristic carve (TC) is in the centre with high TMR ratio of 195% while decreasing the $\mathrm {R}\times \mathrm {A}$ values causes a certain decrease in TMR (%) and there are significantly shifted towards negative field values, which indicates ferromagnetic coupling $( H_{F})$ between the free and reference layer of the multilayer stack[5]. This shift in $H_{F}$ in low $R \times A$ can be a result from either lateral magnetic flux arising from the synthetic antiferromagnet (SAF) that leads to stray fields and couple with the free layer or Néel coupling induced by the roughness of the MgO/CoFeB interface. In contrast, The $H_{F}$ decreases with higher $R \times A$ as the thickness of the barrier increases and is due to reduced Néel coupling in the thicker barriers. To get the deeper insight role of the $H_{F}$ dependence as a function of $R \times A$ values, for two nanopillar diameter 100 nm and 200 nm are plotted as shown in Fig. 1(c). Where 200 nm diameter nanopillars show the transition from positive to negative $H_{F}$ as the $R \times A$ decreases. While for the 100 nm nanopillars, $H_{F}$ dependence is not quite visible except the low $R \times A$ region. The reason behind this behavior is not completely understood. Although the possible explanation would be that the roughness on the nanopillars edges contributed to having multi-domain behavior which leads to this behavior. After the statistical and magnetic analysis of the devices, a set of 100 nm nanopillar diameter with different $R \times A$ was measured in a high-frequency measurement setup. The magnetic field $H_{app} \quad = -150$ Oe is applied to set the MTJ nanopillar in the anti-parallel state while keeping the tunneling current $( I^{tunneling})$ at a low fixed bias of $- 50 \mu \mathrm {A}$ during the measurement to transduce the magnetic oscillations. The power spectral density (PSD) measurements were done as a function of the spin Hall current $( I^{spin\, Hall})$ with an alternative sign of ±5 mA in the step of 0.5 mA. The fact that only one polarity $I^{spin\, Hall}$ of was shown oscillations excludes the possibility of having thermal STT on the device. Fig. 2(a) shows the $P_{matched}$ and TMR (%) distribution as a function of $R \times A($ each dots and square represents a single device). All the $P_{matched}$ values were acquired from the highest (negative) values of $I^{spin\, Hall}$. From the acquired results, it can be confirmed that the increasing the thickness of the MgO barrier (increasing $R \times A)$ leads to increases the TMR (%) and in turn increases the $P_{matched}$ of the device. To quantify the obtained results, $P_{matched}$ and $f$ were extracted and plotted as a function of spin Hall current density $(J^{spin\, Hall})$ as shown in Fig. 2(b-c) for a particular device. The total $P_{matched}$ of 12 nW is obtained. The observed $f$ indicates Red-shift behavior as increases which is a signature sign of steady state oscillations.