Threshold interface magnetization required to induce magnetic proximity effect

Proximity-induced magnetization (PIM) has broad implications across interface-driven spintronics applications employing spin currents. We directly determine the scaling between PIM in Pt and the temperature-dependent interface magnetization in an adjacent ferromagnet (FM) using depth-resolved magnetometry. The magnetization due to PIM does not follow the generally expected linear scaling with the FM interface magnetization, as a function of temperature. Instead, it vanishes while the FM interface magnetization remains. The effective magnetic susceptibilities of heavy-metal (HM) layers are shown to give rise to the previously unexplained asymmetric PIM found in HM/FM/HM trilayers.

Figure 1

Temperature-dependent magnetization (main) for trilayer Pt/CFTB/Pt (open markers), showing three transition temperatures $T_{\mathrm{C}1-3}$ indicated with arrows. Filled markers are extracted from PNR measurements. Magnetization hysteresis loops show two-step magnetization reversal at temperatures (b) between $T_{\mathrm{C}3}$ and $T_{\mathrm{C}2}$, and (a) single-step reversal both below $T_{\mathrm{C}3}$ and (c) between $T_{\mathrm{C}2}$ and $T_{\mathrm{C}1}$. Inset schematics show (left) the scattering geometry for PNR and XRMR measurements, and (right) the sample layer structure and tantalum distribution profile through the CFTB layer.

Figure 2

Example reflectivity, and spin-asymmetry at various temperatures, for Pt/CFTB/Pt from (left) PNR and (right) XRMR. The horizontal dashed lines show zero spin-asymmetry, and vertical dashed lines in (d) and (e) highlight that the peaks in the XRMR spin-asymmetry are strongest at Kiessig fringe minima in the reflectivity. The solid lines show best fits to the reflectivity and spin-asymmetry, corresponding to the mSLD profiles in (c) and (f). The sample layer structure is indicated against the mSLD profiles; slight structural differences, primarily in the lower Pt layer thickness, may result from the different sampling volumes (beam spot sizes) for PNR ($\sim 3\mathrm{cm}\times 3\mathrm{cm}$) and XRMR ($\sim 200\mu \mathrm{m}\times 300\mu \mathrm{m}$), however, layer thicknesses extracted from unconstrained fitting of PNR and XRMR data are consistent within standard error. Apparent overlap of the PNR mSLD profile into Pt layers, and XRMR mSLD profile into CFTB, is a result of the error-function profile used to model the interface roughnesses.

Figure 3

Parametric scaling of PIM in Pt (XRMR mSLD) against CFTB interface magnetization (PNR mSLD) for lower Pt/CFTB (filled markers) and upper CFTB/Pt interfaces as a function of temperature. Solid lines show the linear scaling between PIM and interface magnetization over the temperature range where each interface is locally ferromagnetic. The conventional linear description of PIM, Eq. (1), necessarily passes through the origin.