In the nonlocal spin valve (NLSV) geometry, four-terminal electrical Hanle effect measurements have the potential to provide a particularly simple determination of the lifetime $\left({\tau }_s\right)$ and diffusion length $\left({\lambda }_{\mathrm{N}}\right)$ of spins injected into nonmagnetic (N) materials. Recent papers, however, have demonstrated that traditional models typically used to fit such data provide an inaccurate measurement of ${\tau }_s$ in ferromagnet (FM)/N metal devices with low interface resistance, particularly when the separation of the source and detector contacts is small. In the transparent limit, this shortcoming is due to the back diffusion and subsequent relaxation of spins within the FM contacts, which is not properly accounted for in standard models of the Hanle effect. Here we have used the separation dependence of the spin accumulation signal in NLSVs with multiple FM/N combinations, and interfaces in the diffusive limit, to determine ${\lambda }_{\mathrm{N}}$ in traditional spin valve measurements. We then compare these results to Hanle measurements as analyzed using models that either include or exclude spin sinking. We demonstrate that differences between the spin valve and Hanle measurements of ${\lambda }_{\mathrm{N}}$ can be quantitatively modelled provided that both the FM contact-induced isotropic spin sinking and the full three-dimensional geometry of the devices, which is particularly important at small contact separations, are accounted for. We find, however, that considerable difficulties persist, in particular due to the sensitivity of fitting to the contact interface resistance and the FM contact magnetization rotation, in precisely determining ${\lambda }_{\mathrm{N}}$ with the Hanle technique alone, particularly at small contact separations.

• Received 25 August 2015
• Revised 25 July 2016

DOI:https://doi.org/10.1103/PhysRevB.94.094431