Nonlinear Electrical Spin Conversion in a Biased Ferromagnetic Tunnel Contact

Abstract

The conversion of spin information into electrical signals is indispensable for spintronic technologies. Spin-to-charge conversion in ferromagnetic tunnel contacts is well-described using linear (spin-)transport equations, provided that there is no applied bias, as in nonlocal spin detection. It is shown here that in a biased ferromagnetic tunnel contact, spin detection is strongly nonlinear. As a result, the spin-detection efficiency is not equal to the tunnel spin polarization. In silicon-based 4-terminal spin-transport devices, even a small bias (tens of mV) across the Fe/MgO detector contact enhances the spin-detection efficiency to values up to 140 % (spin extraction bias) or, for spin injection bias, reduces it to almost zero, while, parenthetically, the charge current remains highly spin polarized. Calculations reveal that the nonlinearity originates from the energy dispersion of the tunnel transmission and the resulting nonuniform energy distribution of the tunnel current, offering a route to engineer spin conversion. Taking nonlinear spin detection into account is also shown to explain a multitude of peculiar and puzzling spin signals in structures with a biased detector, including two-and three-terminal devices, and provides a unified, consistent and quantitative description of spin signals in devices with a biased and unbiased detector.

On the origin of the giant spin detection efficiency in tunnel barrier based electrical spin detector

“…While this reasoning seems to qualitatively capture the trend observed in most experimental results, there still exists some features that remain unexplained [19]. Notably in their explication of the origin of the non-linearity, the fact that the increase of applied voltage needed to compensate the loss of current after spin precession (i. e., the spin voltage) becomes sensitive to the increase of junction bias, leading to a P det independent on the bias.…”

“…Although the linear theory for spin injection and detection captures the essential mechanisms of these processes, deviations from the linear model have been systematically reported in experiments based on 3-terminal (3T) Hanle devices and in 4T devices where both injector and detector are under bias [8,12]. Recently, Jansen et al [19] experimentally observed that the spin detection efficiency P det at a tunnel junction strongly depends on the applied bias, offering a way to magnify the detected spin accumulation. In their work, the authors pointed out that the spin signal amplification is inherent to the nonlinear transport occurring at the tunnel junction which arises from the dependence of the transmission probability with the energy of injected carriers.…”

“…1(a)). This model has already been studied previously [19], however, in the present work the FM quasi Fermi level is set as the reference electrode and the oxide barrier deformation with applied voltage is considered.…”

“…Recently, it has been experimentally highlighted that the discrepancies between experiments and theoretical calculations are due to the energy dependence of the carrier transmission probability in the tunnel junction. This suggests that a new description based on a non-linear transport of spin should be invoked [19]. Although preliminary ideas in this direction were already advanced some ten years ago, it is not until recently that nonlinearities have been recognized as an essential ingredient for the understanding of FM tunnel junctions.…”

Study of Spin Transport and Magnetoresistance Effect in Silicon-based Lateral Spin Devices for Spin-MOSFET Applications

“…2 shows the trend in the reported values of P for Sibased devices. [22][23][24][25][26][27][28][29][30][31] We find that relatively high P values have been reported year by year by using ferromagnet/MgO tunnel barrier contacts. On the other hand, the reported MR ratios measured by local twoterminal measurements were quite small less than 0.01% at room temperature, 15,22,24,30,32) largely inconsistent with the expected values from Eq.…”