Multiterminal Tunable InZnO Dendritic Transistor-Based Reservoir Computing for Chaotic Time Series Prediction
Li Qiang Zhu, Feng Zhang, Pengyu Chen, Xiang Wan, Chee Leong Tan, Huabin Sun, Yong Ping Xu, Shancheng Yan, Leisheng Jin, Zhihao Yu · IEEE Transactions on Electron Devices · 2025
To address the challenges of complex time series signal processing in neuromorphic systems, the multigate dendritic transistor, with its multiphysics coupling mechanism, provides a hardware foundation for biomimetic dendritic spatiotemporal information processing and multistate reservoir computing (RC) systems. In this article, we present a tunable multiterminal InZnO dendritic transistor based on chitosan electrolyte. The device utilizes the ionic dynamics of chitosan and the parallel control capability of multiple gates to achieve cooperative modulation. It demonstrates tunable synaptic behaviors, including adjustable excitatory postsynaptic current (EPSC) and multipulse facilitation. Moreover, it emulates dendritic direction-selective responses to temporally and spatially correlated electrical stimuli, allowing for discrimination of input sequence order. Finally, coplanar-gate modulation extends the reservoir state space by tuning the relaxation time of the channel current. As a result, the RC system based on this device achieves high accuracy in Lorenz chaotic time series prediction, with a coefficient of determination (R2)up to 0.99.