Temperature-dependent perpendicular anisotropy and Gilbert damping of $L{1}_0\text{−}\mathrm{FePd}$ films: Role of noble-metal buffer layers

The moderate bulk perpendicular magnetic anisotropy (PMA, $K_{\mathrm{u}}\approx 1\mathrm{MJ}/{\mathrm{m}}^3$) and low Gilbert damping (α < 0.01) make $L{1}_0$-FePd a promising candidate for energy-efficient and nonvolatile spintronic devices with large areal densities (down to 5-nm pitch sizes or even lower). Existing applications subject spintronic devices to a wide range of operating temperatures (e.g., −55 to 150 °C). To better address the technological viability of FePd for spintronic applications, it is of utmost importance to evaluate the material performance of $L{1}_0$-FePd (e.g., anisotropy strength and Gilbert damping) at elevated temperatures. In this work, we systematically investigate the effect of buffer layers (Cr/Pt, Cr/Ru, Cr/Rh, Cr/Ir, and Ir) on the PMA and Gilbert damping of $L{1}_0$-FePd from room temperature (RT, 25 °C) to 150 °C using the time-resolved magneto-optical Kerr effect metrology. It is found that the effective anisotropy field ($H_{\mathrm{k},\mathrm{eff}}$) of FePd decreases with the testing temperature ($T_{\mathrm{test}}$) and the ratio of $H_{\mathrm{k},\mathrm{eff}}$(150 °C)/$H_{\mathrm{k},\mathrm{eff}}$(25 °C) is positively correlated to the degree of $L{1}_0$ phase ordering. The Gilbert damping of $L{1}_0$-FePd either increases with $T_{\mathrm{test}}$ or stays nearly constant over the $T_{\mathrm{test}}$ range. We attribute the temperature dependence of Gilbert damping to the spin diffusion length of the metallic buffer layer (λ), presumably through the spin pumping effect. Results of this work provide guidance to tailor $L{1}_0$-FePd properties through buffer layer engineering for applications in spintronic devices over wide operating temperature ranges.

Figure 1

(a) A HAADF STEM image of MgO/Ir(11)/FePd(8) sample. (b) A magnified Fig. 1 around the Ir/FePd interface. (c),(d) Hysteresis loops of Cr/Pt/FePd(8) sample at (c) 25 °C (RT) and (d) 125 °C. The insets show the out-of-plane $M-H$ loops near $H_{\mathrm{ext}}=0$. (e) Magnetization ratio $M_{\mathrm{s}}\left(T\right)/M_{\mathrm{s}}$(25 °C) as a function of $T_{\mathrm{test}}$ from RT to 125 °C, for all samples. For better visualization, one representative error bar is shown given that all data points (except for data at 25 °C, which do not have uncertainties) have similar error bars (∼±2%). (f) Magnetization ratio $M_{\mathrm{s}}$(125 °C)/$M_{\mathrm{s}}$(25 °C) vs $S$. The dashed line shows the linear fitting. (e),(f) share the same figure legends.

Figure 2

(a) The schematics of sample stack and TR-MOKE measurement configurations (${\theta }_H\approx {80}^{\circ }$). (b) TR-MOKE signals (symbols) measured on Cr/Pt/FePd(8) under varying $H_{\mathrm{ext}}$ at RT and their fitting curves (solid lines). (c),(d) An example of fitting $f$ vs $H_{\mathrm{ext}}$ and 1/τ vs $H_{\mathrm{ext}}$ to extract $H_{\mathrm{k},\mathrm{eff}}$ and α from Cr/Pt/FePd(8). Circles and curves represent experimental data and modeling fitting, respectively.

Figure 3

(a) $H_{\mathrm{k},\mathrm{eff}}$ vs $T_{\mathrm{test}}$ for all FePd films deposited on different buffer layers. (b) The ratio of $H_{\mathrm{k},\mathrm{eff}}$ at 150 °C to its RT value as a function of $S$. The dashed line guides the general trend.

Figure 4

Gilbert damping as a function of $T_{\mathrm{test}}$ for (a) 8-nm FePd samples and (b) Pt-seeded FePd samples. In panels (a) and (b), symbols represent experimental data and dashed lines are corresponding linear fittings that are used to calculate the change of α (denoted as Δα) in Fig. 5.

Figure 5

(a) The change of α vs the spin-diffusion length λ normalized to the buffer-layer thickness $t [\mathrm{\Delta }\alpha =\alpha$(150 °C)$-\alpha$(25 °C)]. (b) Comparison of the theoretically predicted (dashed line) and experimentally measured (symbols) trend for $\left(\mathrm{\Delta }\alpha /g_{\uparrow \downarrow }\right)/\left(\mathrm{\Delta }{\alpha }_2/g_{\uparrow \downarrow ,2}\right)$ vs $\lambda /t$.