Controllable Skyrmion Nucleation and Transition in a Confined Nanodisk for the Binarized Neuron Network of 2D Ferromagnet FeGeTe.

The integration of skyrmions in 2D ferromagnetic materials and advanced spintronic devices enables energy-efficient computing for artificial synapses in neuromorphic architecture. Nucleation of skyrmions plays a critical role in ensuring high reliability and low energy consumption. However, the key challenge lies in selectively generating skyrmions, excluded from trivial bubbles and their controllable transition. Here, as a concept example of 2D ferromagnets, we demonstrate the evolution of the skyrmion in FeGeTe excluded from the trivial bubbles within the geometrically confined 1 μm nanodisk. The skyrmions selectively nucleate from stripe domains, driven by the competition between dipolar-dipolar interactions and Zeeman energy, captured in real-time/space using in situ Lorentz TEM. Furthermore, we reveal a reversible skyrmion-to-trivial bubble transition in the 500 nm nanodisk induced by an in-plane magnetic field, confirmed by the micromagnetic simulation. The reversible switching of skyrmions and trivial bubbles enables the implementation of a binary-state neuromorphic computing framework. The skyrmion-based artificial synapses demonstrate over 85% handwriting recognition accuracy. The findings bridging fundamental skyrmion physics with practical applications offer key insights for designing next-generation 2D materials-based devices.