Reconfigurable Bio-Plausible Spiking Neurons Based on Antiferromagnetic Skyrmions Utilizing Anisotropy Gradient

Biological neurons exhibit diverse spiking behaviors, such as fast, phasic, and bursting modes, which are essential for efficient neural communication. However, most spintronic neuromorphic systems remain limited to simplified leaky-integrate-and-fire (LIF) models due to the difficulty of realizing complex neuronal dynamics in hardware. In this work, we develop a reconfigurable spiking neuron based on the intrinsic dynamics of antiferromagnetic (AFM) skyrmions driven by an anisotropy gradient in structured nanotracks. By engineering an anisotropy gradient along a trapezoidal nanotrack, the device achieves multimodal LIF functionality without external current injection or additional hardware overhead. The nanotrack geometry naturally enables leaky integration, while controlled anisotropy modulation allows dynamic reconfiguration between fast, phasic, and bursting spiking modes. An artificial neural network constructed from these AFM skyrmions as neurons demonstrates improved Iris flower classification performance compared to conventional LIF-based models, highlighting their potential as an energy-efficient and bio-plausible neuromorphic computing platform.