Temperature Evolution of Spin Relaxation in a $\mathrm{NiFe}/\mathrm{Cu}$ Lateral Spin Valve

Temperature dependence of spin relaxation process in a Cu wire has been studied by means of nonlocal spin-valve measurements. The spin-diffusion length of the Cu wire is found to take maximum at the characteristic temperature, below which the spin-diffusion length is reduced. The mechanism of the reduction can be explained by considering the spin-flip scattering due to the oxidized surface of the Cu wire. The thickness dependence of the characteristic temperature supports the interpretation with the surface oxidation.

Figure 1

(a) Schematic illustration of the fabrication procedure of a lateral spin valve by means of shadow evaporation technique. Py is evaporated at an oblique angle. The Cu is evaporated perpendicular to the substrate. (b) Scanning-electron-microscope image of the fabricated lateral spin valve consisting of two Py wires bridged by a Cu wire together with the probe configuration for the nonlocal spin-valve measurement. The location of Py wires are indicated by the dotted lines. The magnetic field direction is indicated by a solid arrow.

Figure 2

(a) Temperature dependence of the spin signal in the $\mathrm{Py}/\mathrm{Cu}$ lateral spin valve with the separation distance 250 nm. The inset shows the field dependence of the nonlocal spin-valve signal measured at 20 K. (b) Temperature dependences of the spin signals for different separation distances.

Figure 3

(a) Temperature dependence of the spin-diffusion length of the Cu wire deduced from the distance dependence of the spin signal. The inset shows the spin signal as a function of the separation distance $d$ measured at 290 K and 10 K. The solid and open dots are the experimental data at 10 K and 290 K, respectively. The solid and dotted lines are calculated evolutions of the spin signal using Eq. (1) at 10 K and RT, respectively. (b) Temperature dependence of the resistivity of the Cu wire. (c) Calculated spin signal as a function of the temperature for $d=250\mathrm{nm}$ and 1650 nm.

Figure 4

(a) Temperature dependences of the spin signal for various Cu thicknesses. In all the devices, the separation distance is fixed to 250 nm. (b) Thickness dependence of the characteristic temperature $T_m$ of the $\Delta R_S$ maximum. (c) Thickness dependence of the spin signal $\Delta R_S$ measured at 10 K.