Spin polarized transport in magnetic tunnel junctions

Spin polarized transport in magnetic tunnel junctions (MTJs) has recently aroused enormous interest due to its fertile physics and potential application in non-volatile random access memories and next generation magnetic field sensors. However, mysteries still remain underlying the physical mechanism of MTJs.; One mystery is the bias dependence of tunneling magnetoresistance (TMR). We have investigated this issue both experimentally and theoretically. We observed large spin dependent conductance minimum shift in asymmetric MTJs. To explain the observed conductance anomaly, we proposed a modified Brinkman model by incorporating the voltage-dependent density of states (DOS) of the ferromagnetic (FM) electrodes. This model suggests that a reasonable variation of the effective DOS is required to fit the rapid decrease of resistance and TMR under large voltage bias, which indicates that the most significant contribution of the bias dependence comes from the electronic structure of the FM electrodes. In addition, we have also revealed another mechanism for bias dependence that is based on electron wavefunction matching at the interfaces on either side of the barrier. In our study of inversed TMR effect, negative TMR was observed in MTJs with AlO_x/ZrO_x and ZrO_x/AlO _x barriers. Energy dependent DOS alone is not sufficient to provide coherent explanation of all the observed phenomena. Inspired by Slonczewski model, we considered the barrier shape effect on the electron wavefunction matching at high bias. In this extended Slonczewski model, both energy dependent DOS and barrier height were found to influence the bias dependence of TMR.; Another important issue in spin polarized tunneling is the tunneling characteristic length. It is commonly believed that the interface contribution dominates the tunneling behavior. We have carefully designed experiments to investigate the thickness effect of FM electrode and NM electrode. In the first experiment, we have demonstrated that up to 10 monolayers of CoFe is required to saturate TMR. After structure related origins have been carefully excluded, this slow saturation of TMR strongly indicates the existence of a bulk contribution. To further justify our point, we also carefully designed a MTJ structure with a composite FeNi/Cu/FeNi electrode to investigate the NM layer thickness dependence of TMR behavior. We found, if we keep the top FM layer (above Cu and adjacent to the barrier) in the composite electrode very thin (no more than several nm), a much thicker Cu layer is needed to fully suppress the TMR, which is consistent with our bulk-like contribution argument in the spin polarized tunneling.