SkyNeu: Energy Efficient Antiferromagnetic Skyrmion Based Artificial Neuron for Neuromorphic Computing

Neuromorphic computing emulates the structural and functional principles of biological neural networks to achieve enhanced efficiency, flexibility, and intelligence in tasks like pattern recognition, sensory data processing, and autonomous decision-making. CMOS-based neuromorphic computing faces challenges in energy efficiency and packaging density due to the mismatch with biological computing and scaling limitations in transistor design. To overcome these limitations, antiferromagnetic (AFM) skyrmion-based spintronic devices present a promising solution for neuromorphic computing due to their unique features, such as strong exchange interactions, interfacial Dzyaloshinskii-Moriya interaction (DMI), minimal stray fields, low sensitivity to external magnetic fields, and inherent non-volatility. The linear motion of AFM skyrmions prevent edge annihilation in nanoscale racetracks, enabling the design of high-speed and energy-efficient neuromorphic computing devices. In this work, AFM skyrmion dynamics under DMI gradient is studied at 0K and 300K. It is observed that creating the DMI gradient is an efficient way to drive AFM skyrmions with a high longitudinal speed in the order of ∼ 1000 m/s with no skyrmion Hall effect (SkHE). Furthermore, a leaky integrate fire with self-reset (LIFSR) artificial neuron device is proposed at room temperature based on DMI gradient with low energy consumption of 33.25fJ and 34.27 fJ per neuron spike at 0K and 300K, respectively. This presents a novel approach to manipulating AFM skyrmions, paving the way for the development of energy-efficient AFM skyrmion-based devices in neuromorphic computing.