Temperature-dependent behavior of VO2-based artificial neurons

Abstract

Temperature serves as a pivotal factor influencing information transmission and computational capacity in neurons, significantly affecting the function and efficiency of neural networks. However, the temperature dependence of VO2-based artificial neuron, which is one of the highly promising artificial neurons, has been hardly reported to date. Here, high-performance VO2 devices with NDR features are prepared by rapid annealing and electroforming processes. We constructed VO2-based artificial neurons with output properties similar to those of biological neurons on the basis of the Pearson–Anson oscillation circuit. The temperature-dependent behavior of VO2 neurons was fully investigated. Increasing temperature leads to a decrease in the peak-to-peak value of the output spikes of VO2 neurons. The spike period of VO2 neurons remains relatively stable near room temperature, but it decreases as the temperature reaches above 26 °C. These temperature-dependent features of VO2 neurons are similar to the ones of biological neurons, suggesting a natural advantage of VO2-based artificial neurons in mimicking biological neural activity. These findings contribute toward comprehending and regulating the temperature-dependent behavior of artificial neurons based on Mott memristor.