Skyrmionium-Based Leaky Integrate and Fire Neuron

Neuromorphic computing is an intriguing paradigm for the development of low-power devices. The incorporation of spintronic devices offers nonlinearity and short-term memory effects that are the crucial functions to facilitate the neuromorphic devices with enhanced energy efficiency. A skyrmionium can be considered as one of the promising spin textures for these applications, since it follows straight trajectory along the nanotrack owing to the zero net topological charge. This work proposes a neuron based on the skyrmionium and utilizes the gradient of perpendicular magnetic anisotropy (PMA) on the nanotrack to achieve the leaky-integrate-fire (LIF) functionality. This gradient generates the different energy states and is responsible for inducing the leaky behaviour in the device by moving the skyrmionium in a direction to minimize the energy. It is reported that the suggested device is 15.5% more energy-efficient than the AFM skyrmion-based LIF neuron, dissipating 3.74 fJ of energy per LIF operation. This proposed artificial neuron potentially enable the development of a power-efficient and dense spiking neuromorphic computing system with a straightforward single-device implementation.