▎ 摘　　要

Spin transport experiments in graphene, a single layer of carbon atoms ordered in a honeycomb lattice, indicate spin-relaxation times that are significantly shorter than the theoretical predictions. We investigate experimentally whether these short spin-relaxation times are due to extrinsic factors, such as spin relaxation caused by low impedance contacts, enhanced spin-flip processes at the device edges, or the presence of an aluminum oxide layer on top of graphene in some samples. Lateral spin valve devices using a field-effect transistor geometry allowed for the investigation of the spin relaxation as a function of the charge density, going continuously from metallic hole to electron conduction (charge densities of n similar to 10(12) cm(-2)) via the Dirac charge neutrality point (n similar to 0). The results are quantitatively described by a one-dimensional spin-diffusion model where the spin relaxation via the contacts is taken into account. Spin valve experiments for various injector-detector separations and spin precession experiments reveal that the longitudinal (T-1) and the transversal (T-2) relaxation times are similar. The anisotropy of the spin-relaxation times tau and tau(perpendicular to), when the spins are injected parallel or perpendicular to the graphene plane, indicates that the effective spin-orbit fields do not lie exclusively in the two-dimensional graphene plane. Furthermore, the proportionality between the spin-relaxation time and the momentum-relaxation time indicates that the spin-relaxation mechanism is of the Elliott-Yafet type. For carrier mobilities of 2x10(3)-5x10(3) cm(2)/V s and for graphene flakes of 0.1-2 mu m in width, we found spin-relaxation times on the order of 50-200 ps, times which appear not to be determined by the extrinsic factors mentioned above.