Linking magnetic memory and computational capacity in artificial spin ice

Artificial spin ice (ASI) is a metamaterial composed of nanomagnetic islands that are used as functional platforms for neuromorphic computing. The global field amplitude during minor field loops has been used to input information to the microstate of ASI and has been used to demonstrate reservoir computing in both simulation and experiment [1,2]. The memory effects emergent in ASI are crucial to understanding the computational capacity. In previous studies, ASI has been shown to exhibit return point memory (RPM) after several ‘training’ field loops [3]. We explore, using dipolar simulations, the temporal harmonics and subharmonics that emerge from these field loops as we vary the interaction strength of the system and the quenched disorder. Linking these physical memory effects to the ASI’s computational potential via the memory capacity and nonlinearity.

[1] Jensen J. H. et al. (2020) ALIFE 2020: The 2020 Conference on Artificial Life, Jul., (376–383).

[2] Gartside J. C. et al. (2022). Nat Nanotechnology, 17 (460-469).

[3] Gilbert, I. et al. (2015). Physical Review B, 92 (10).