During the second reporting period, the NIMFEIA consortium made significant progress towards achieving these objectives. A summary of this progress is given here with a more de-tailed description following in the explanation of the work carried out per work package.
Objective 1: We established a quantitative, experimentally validated understanding of nonline-ar magnon interactions in vortex-state reservoirs, including three-magnon scattering, Floquet-engineered frequency combs, and GHz-MHz coupled dynamics. Extensive laboratory-scale experiments, supported by micromagnetic modelling and analytical theory, showed high-quality pattern recognition, Mackey-Glass prediction, and short-term memory performance using mag-non scattering, firmly demonstrating the reservoir-computing principle in reciprocal space. We also clarified the role of device geometry, material choice, and driven Floquet states in shaping computational capacity.
Objective 2: We realized the first electrical read-out for magnon reservoirs by integrating NiFe vortex disks with industrial MRAM technology, fabricating fully functional MTJ-based magnon sensors at the CMOS wafer level. Additionally, we demonstrated direct electrical read-out in vortex-based MTJs. We optimized device architectures (antenna geometries, layer stacks, bit-line thinning, nanocontacts) and demonstrated electrical detection of nonlinear magnon pro-cesses, including three-magnon splitting and frequency combs, directly from MTJs. This establishes a scalable path toward industrially compatible magnonic hardware.
Objective 3: We achieved the first experimental demonstration of gesture recognition using a magnetic reservoir, classifying thousands of real radar traces via a single-skyrmion Brownian reservoir with performance competitive with software models but at unprecedentedly low ener-gy cost. Although vortex-based reservoirs are better matched to scalar time-series than multi-dimensional radar data, we achieved high prediction accuracy for chaotic time-series (Mackey-Glass), validating the magnon reservoir’s real-world computational utility.
Objective 4: We identified and simulated self-induced Floquet mechanisms in synthetic antifer-romagnets, enabling frequency-comb physics far beyond the GHz range. We developed theo-retical tools for ultrafast magnon-pair excitation and coherent THz spin-current generation, extended experimental infrastructure up to 170 GHz, and demonstrated high-frequency AFM magnon modes in candidate materials. Work on THz-GHz down-conversion, neural-network quantum-state modelling, and alternative THz-capable spin textures (SAFs, skyrmions, sMTJs) further advances the early-stage scaling pathway toward true THz magnon reservoirs.