Skyrmions Diffusion enabling stochastic computing*

Skyrmions have been suggested for a variety of applications including memory but in particular also non-conventional computing such as probabilistic computing [1]. A key problem for probabilistic computing is that cascading gates propagate undesired correlations. Therefore one needs to reshuffle the signals to keep them uncorrelated. While for many non-conventional computing approaches non-magnetic implementations are most promising, for building a “reshuffler”, skyrmions might be ideally suited due to the low footprint and low power compared to e.g. CMOS implementations [1].
We have studied low skyrmion pinning materials where we can stabilize skyrmions and controllably nucleate and displace them by current pulses due to spin-orbit torques [2]. We find topologically non-trivial N=1 skyrmions that move in the direction of the current pulses . At zero applied current, we find thermally activated skyrmion motion. We track the trajectories of skyrmions and from the dependence of their mean-square-displacement (MSD) on time, we can identify motion by diffusion and obtain the diffusion constant [2]. There is a strong dependence of the skyrmion diffusion parameter on temperature and the skyrmion size. Finally we patterning the reshuffler geometry and ascertain its performance.
Beyond ferromagnetic skyrmions, we are evaluating the efficiency of skyrmion spin structures in antiferromagnets, which have major advantages. In addition to the absence of a skyrmion Hall effect [3], also antiskyrmions can be stabilized for which first indications have been found.
[1] D. Pinna et al., Phys. Rev. Appl. 9, 064018 (2018)
[2] J. Zazvorka et al., Nature Nano. 14, 658 (2019)
[3] K. Litzius et al. Nature Phys. 13, 170 (2017)
[4] J. Barker et al., Phys. Rev. Lett. 116, 147203 (2016)