Reconfigurable Neurotransistors Based on Wide-Bandgap Semiconductors for Adaptive Reservoir Computing
Tao Chen, Zheyang Zheng, Sirui Feng, Lining Zhang, Yan Cheng, Yat Hon Ng, Kevin Jing Chen · 2024
We demonstrate a reconfigurable neurotransistor based on wide-bandgap (WBG) semiconductors with tunable nonlinear information processing and memory functions for physical reservoir computing (RC). The neurotransistor, which utilizes gallium nitride (GaN) as the channel material, can be configured as volatile memory (VM) or non-volatile memory (NVM). As a VM, the device can perform adaptive nonlinear information processing with the gate-tunable nonlinear functions and short-term memory. As an NVM, it exhibits multistate storage capability with fast and linear weight updating, long retention time, and high endurance. We utilize the GaN-based neurotransistor to demonstrate a highly adaptive RC system featuring enriched nonlinear dynamics and tunable temporal responses. The RC system is capable of long-term forecasting of chaotic time series across different timescales.