| This thesis describes the effects of surfaces and interfaces on spin-dependent electron transport in metallic ferromagnetic-normal metal nanostructures. Bulk spin-dependent transport properties of metals can be understood in terms of the charge transport properties for devices larger than the characteristic diffusion length for spin polarized electrons. For devices with reduced cross-sectional dimensions, approaching the electronic mean-free-path, the surfaces of the device begin to dominate the spin dependent transport. Additionally, the interplay of the transport properties of the ferromagnet and normal metal collectively determine the overall properties in concert with the effect of the surfaces.;A process to fabricate lateral metallic spin transport devices was developed using electron beam lithography to pattern two-angle shadow masks for deposition and lift-off. Ferromagnetic metals (FM), NiFe or Co, and normal metals (N), Cu or Al, were evaporated from high purity sources in ultra-high vacuum, without breaking vacuum, to minimize interfacial resistance. The temperature-dependent magnitude of the non-equilibrium spin accumulation was measured in a non-local geometry to obtain information on spin injection and relaxation. Further, the reduction in spin accumulation as the source-detector separation was increased allowed for a measurement of the effective spin diffusion length in the nanoscopic devices, measured here to be less than 600 nm. The application of experimentally-constrained analytical 1D models of spin transport returns not only information about the spin-dependent properties of the N, but also that of the FM. By measuring four different combinations of N and FM metals, the contributions from the N and FM properties can be separated from those of the interfaces. Additionally, modifying the cross-sectional geometry of the device gave information about the contributions of boundaries to spin relaxation. A complementary measurement of these properties can be attained through electrical Hanle effect measurements, although quantitative analysis is possible only by developing models that include both diffusive FM/N interfaces and surfaces. Elliot-Yafet theory (EY) predicts spin diffusion lengths greater than 1000 nm for bulk materials, much larger than measured here. EY relaxation was invoked in numerical simulations of the full lateral device in three dimensions; the simulations were conducted to model the temperature and spatial dependence of the spin accumulation. Through these simulations we demonstrate that enhanced spin relaxation at surfaces can reproduce the general experimental observations, including the dependence on cross-sectional geometry. |
