Conduction electron scattering and spin-flipping at sputtered Al/Cu Interfaces

Abstract

We use two different techniques to derive the two parameters describing conduction electron scattering and spin-flipping at sputtered Al/Cu interfaces in the current-perpendicular-to-plane (CPP) geometry. These parameters are: 2ARAl/Cu, twice the interface specific resistance, where A is the area through which the CPP current flows; and δAl/Cu, which gives the probability P of spin-flipping from P = 1 – exp(−δ). A technique involving simple multilayers, and sample temperature not exceeding room temperature, gives 2ARAl/Cu = 2.3 ± 0.2 fΩm2. A technique involving exchange-biased spin-valves (EBSVs), where the sample is annealed briefly to 453 K, gives 2ARAl/Cu = 2.0 ± 0.15 fΩm2. Averaging the two values, but increasing the uncertainty for reasons explained, gives the best estimate of 2ARAl/Cu = 2.15 ± 0.4 fΩm2. This average is comparable to, but smaller than, the published value of 2ARAl/Cu = 3.6 ± 1 fΩm2 derived from thermal conductance measurements, and larger than our calculated values for interface thicknesses up to 6 monolayers (ML). However, it is similar to our calculated values for an interface thickness of 8 ML. Combining extrapolation of higher temperature bulk diffusion data for Al in Cu and vice-versa, with x-ray and transmission electron microscope (TEM) studies of similarly sputtered multilayers, indicates that such interface thicknesses are possible, especially for annealed multilayers. CPP-magnetoresistance (MR) measurements of the EBSV samples give only very small spin-flipping at the Al/Cu interface—δAl/Cu = 0.05−0.05+0.02. Such a small value is consistent with expected small spin-orbit interactions in both Al and Cu. Supplementary studies of CPP-MR of Permalloy (Py)-based EBSVs containing [Cu/Al/Cu] trilayers, show unusual behavior when the central Al layer is at least 10 nm thick, giving a CPP-MR like that for Py/Al, independent of Cu layer thicknesses from 0 to 10 nm. MR, x-ray, and TEM results give some clues as to the origins of this behavior, but a completely satisfactory explanation is not yet available.