Description
Magnetic skyrmions, topologically protected, particle-like spin textures, are highly promising candidates for energy efficient spintronic devices, particularly as artificial synapses in neuromorphic hardware [1-3]. Their key advantages include full electrical controllability, non-volatility and stability at room temperature, and the ability to represent synaptic weights through the countable number of skyrmions confined at a specific location in a device. Because skyrmions behave as discrete entities that can be nucleated, moved, summed, and electrically read, they offer a natural physical substrate for implementing weighted-sum operations central to neural-network computation [3].
This project aims to demonstrate such synaptic behaviour in micrometer-scale devices patterned in a magnetic tunnel junction (MTJ) stack. The MTJ incorporates a free CoFeB layer magnetically coupled to a Co/Ru/Pt skyrmion-hosting multilayer (MML). By enabling electrically controlled skyrmion nucleation in the MML and exploiting the magnetic imprint transferred to the CoFeB free layer, the device can encode and read non-volatile synaptic weights through tunnel magnetoresistance.
Micromagnetic simulations are used to model the full stack, skyrmionic multilayer, spacer, and MTJ free layer—and are used to optimise material thicknesses and multilayer repetition to obtain stable skyrmions with desired diameters and strong imprint onto the free CoFeB layer for efficient readout. For experimental validation, we fabricate MTJs integrating the skyrmionic free layer and demonstrate electrical nucleation and synaptic-like weight updates. Building on this concept, we aim to design small arrays of MTJs capable of performing weighted-sum operations using multiple skyrmion-based synapses.
Together, the results establish a clear route toward electrically operated, skyrmion-based synaptic elements and scalable spintronic neuromorphic hardware.
[1] J. Grollier et al., Nature Electronics 3 (2020).
[2] K. M. Song et al., Nature Electronics 3 (2020).
[3] T. da Câmara Santa Clara Gomes et al., Nature Electronics 8 (2025).
| Field of Research/Work | Condensed Matter and Materials |
|---|
