Bloembergen Problem based magnonic physical reservoir computing

Abstract

In this work, we propose a concept of physical reservoir computing based on the three-wave decay of magnetostatic surface spin waves (MSSW) in a thin yttrium iron garnet film and numerically model the operation of a spin-wave active ring resonator (SWARR) under the influence of these three-wave processes as a physical reservoir. To ensure the reliability of our modeling results, we first confirm that, with some modifications, the mathematical model of the Bloembergen Problem for the nonlinear interaction of three waves is in good qualitative agreement with experimental data from the literature on the three-wave decay of short pulses of MSSW. Based on the modified Bloembergen Problem, we then construct a numerical model to simulate the operation of an SWARR under three-wave decay and perform numerical simulations using the developed program code. We find that the performance characteristics of the modeled SWARR-based reservoir are similar to those experimentally measured earlier for an SWARR operating under conditions where three-wave processes are prohibited, and only four-wave processes are allowed. Since the threshold for the three-wave decay of MSSW is considerably lower than that for the four-wave processes, this suggests a significant reduction in power consumption for an SWARR-based physical reservoir if it operates in the mode of three-wave decay of the primary MSSW. We also demonstrate that the proposed physical reservoir network can be used to denoise data sequences. We show this using the example of the Lorenz96 chaotic time series.