NeuMag · Harnessing Nonlinear Wave Dynamics in Magnets for Energy-Efficient Neuromorphic Computing

We live in a technological world where faster, smaller, and energy-efficient information devices are central to modern life. Continued progress in computing requires new paradigms for data-storage and processing that move beyond the limitations of semiconductor-based technologies. Magnonics, which harnesses spin waves (magnons) as information carriers, offers an attractive path toward low-power, wave-based computing. Yet, next-generation magnonic devices remain unrealized, in part due to the difficulty of achieving strong nonlinear interactions and scalable device integration. NeuMag addresses this challenge by establishing synthetic antiferromagnets (SyAFs) as a nonlinear-wave platform for neuromorphic computing. In SyAFs, strong interlayer exchange coupling enhances nonlinear dynamics, while coupling to surface acoustic waves provides an additional route to hybrid magnon–phonon interactions. Together, these effects enable frequency mixing, and harmonic generation, which can be exploited for physical reservoir computing. This project will experimentally demonstrate this concept by fabricating SyAF/SAW devices, characterizing their nonlinear dynamics using GHz spectroscopy, and applying them to benchmark machine-learning tasks such as digit and spoken-word recognition. Unlike conventional approaches that rely on CMOS electronics or linear spin-wave propagation, NeuMag proposes a new paradigm: encoding information into wave excitations and harnessing nonlinear spin dynamics to transform inputs into high-dimensional feature spaces. This will open opportunities for neuromorphic computing and establish a foundation for future spintronic processors. Moreover, NeuMag provides interdisciplinary training in nanofabrication, and machine learning, supporting the fellow’s growth into an independent, globally connected researcher. In doing so, the project will advance the frontier of nonlinear magnonics and contribute to energy-efficient hardware for artificial intelligence.