MESO-Neuron: A low power and ultrafast spin neuron for neuromorphic computing

In this paper, a low power and ultrafast spin neuron for mimicking the biological neuron based on magnetoelectric spin orbit (MESO) neuron is proposed. Firstly, the physical model of MESO neuron based on Landau-Lifshitz-Gilbert (LLG) equation at room temperature is built for investigating the characteristics. By utilizing these characteristics of MESO device, current pulse is used to induce the stochastic switching behaviors. We successfully mimic the behavior of biological neuron with single activation time down to 0.8 ns. Secondly, using model derived device parameters, we further simulate a three-layer full connected neural network using MESO neurons. For a Mixed National Institute of Standards and Technology database (MNIST) handwritten pattern dataset, our system achieves a recognition accuracy of 98%. In addition, the influence of pulse width and amplitude on activation functions of MESO neuron is researched by HSPICE tools. The results show that as pulse width and amplitude are increasing, the power consumption and computing time increase while energy consumption decreases. Specifically, the power consumption performance of MESO neuron is about 10 W and improved approximately 3 orders of magnitude compared to the 45 nm CMOS neuron.