Origin of temperature dependence in tunneling magnetoresistance

– We present detailed measurements of the differential resistance (dV/dI) of stateof-the-art FM/AlOx/FM magnetic tunnel junctions (MTJ) as a function of applied bias and temperature. Temperature effects are particularly significant in physical quantities involving narrow features such as those at low-voltage bias. We show that the temperature evolution of the tunneling characteristics and, in particular, the pronounced rounding of the dV/dI curves with increasing temperature can be well explained by thermal smearing of the tunneling electron energy distribution. Tunneling magnetoresistance [1] (TMR) in magnetic tunnel junctions (MTJ) continues to receive increasing attention [2], since high TMR values [3, 4] allow for applications such as magnetic random access memory [5] (MRAM) and magnetic read-heads [6]. Although the low-temperature TMR can reach the optimum values expected from Julliere’s model [1], even high-quality MTJs suffer a significant loss of TMR with increasing temperature. It was suggested that emission of surface magnons by tunneling electrons lead to extra current channels through the tunneling barrier. The contribution from these channels grows with the number of thermally excited magnons, more so for the antiparallel (AP) configuration, with the consequent overall increase in conductivity and corresponding loss in TMR with increasing temperature [7,8]. In a different approach, Julliere’s model was extended [9] to include both a temperature-dependent surface spin polarization, and a spin-independent current channel, assumed to arise from hopping within the barrier [10–12]. Other proposed mechanisms include barrier defect and/or magnetic impurity scattering [13–16] and band structure effects [17]. Although thermal smearing is known to dominate the temperature dependence of superconducting tunneling [18], the same mechanism has been largely ignored in MTJs, based on the conclusions in Simmons’ original theoretical work [19] where thermal effects are found to be very small for elastic normal-state tunneling through an ideal trapezoidal barrier. In this work, we show that thermal smearing plays a much more important role in determining the temperature dependence of TMR than previously believed. Inclusion of thermal smearing in the analysis of experimental data enables us to explain a number of experimental observations, such as the much stronger T -dependence of TMR at zero bias compared to finite