Currently, artificial neural networks (ANNs) have become a powerful tool for tackling a wide range of complex problems in machine learning, with exceptional capabilities in areas such as image recognition, natural language processing, and autonomous systems [1]. However, while the theoretical advancements of ANNs have been substantial, translating these models into efficient and scalable hardware remains a challenging task. Conventional digital computing systems, which are predominantly based on CMOS technologies, are limited by their high energy consumption. One alternative approach is wave-based computing, where information processing is achieved through the propagation and interference of waves within a physical medium. In these systems, the wave interference patterns can naturally facilitate complex, all-to-all connectivity, which is a fundamental requirement for dense and adaptive neural architectures [2].
In this study, we investigate the use of spin waves as a novel platform for implementing neural network functionalities. Spin waves are collective oscillations of electron spins in magnetic materials, operating at microwave frequencies with remarkably short wavelengths, making them ideal for developing high-speed signal processing devices. Based on the micromagnetic simulations, we propose a novel spin wave-based architecture, where spin waves serve in the dynamical coupling of magnetic molecules, which are deposited on the low-damping magnetic substrate, specifically yttrium iron garnet (YIG). By analyzing the dispersion relations and transmission spectra of spin waves in YIG, we confirm the feasibility of establishing dynamic interconnections between these molecular nodes. This mechanism can enable magnetic molecules to function as artificial neurons, interconnected via propagating spin waves. Therefore, such a framework can pave the way for new directions in molecular-based neuromorphic computing.
[1] Marković, D., Mizrahi, A., Querlioz, D., & Grollier, J. (2020). Physics for neuromorphic computing. Nature Reviews Physics, 2(9), 499-510.
[2] Papp, Á., Porod, W., & Csaba, G. (2021). Nanoscale neural network using non-linear spin-wave interference. Nature communications, 12(1), 6422.
