Skyrmion dynamics for unconventional computing

• Applicants: Professor Dr. Mathias Kläui; Professor Dr. Peter Virnau
• Subject Area: Experimental Condensed Matter Physics ; Theoretical Condensed Matter Physics
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 568748825

In this combined theory-experiment project, we will develop and then use quantitative finite temperature quasi-particle simulations of magnetic skyrmion spin structure dynamics in conjunction with real-time magnetic imaging to explore, develop and analyze selected unconventional computing approaches that lend themselves to exploiting thermally induced stochastic skyrmion dynamics. Unconventional computing approaches using thermal activation promise particularly low power logic and magnetic skyrmion spin structures are promising due to their topologically enhanced stability combined with strongly non-linear dynamics and tuneable stochastic dynamics. Experimental skyrmion systems are studied on micrometer lengthscales and timescales of seconds, and this is orders of magnitude outside the realm of conventional micromagnetic simulations. So to model these systems, we will as a first step develop a quantitative numerical model based on the Thiele ansatz that captures the real experimental dynamics. A key component which is still missing in this framework is the mapping of simulation time units to real time. Once this problem is addressed, we will be able to cover experimental time- and length-scales in numerical simulations which will enable a parallel in-silico scan of parameter space and experimental conditions to identify suitable scenarios for experiments. These tools will then be used to model and analyze thermal skyrmion dynamics in experimental structures and understand the interplay of non-flat energy landscapes, skyrmion dynamics and correlations. We will go beyond the naturally occurring variations in the skyrmion energy and engineer the skyrmion energy landscape to tune the dynamics. As a second step, we will explore selected unconventional computing paradigms with skyrmions. Based on our previous realization of Brownian Reservoir Computing, we will realize time-multiplexed reservoir computing in systems optimized with the help of in-silico predictions. We will then explore skyrmions for random node perturbation for Constrained Parameter Inference Learning algorithms. Finally, we will exploit in conjunction with theoretical modelling the non-flat energy landscape to map the dynamics onto a Markov model and eventually analyze the dynamics in terms of an Ising machine. We will use the previously established approaches of static and dynamic tuning of the energy landscape to understand what Hamiltonians are naturally realized and what Hamiltonians can be realized in these systems by tuning of the interactions.

• DFG Programme: Research Grants