Reconfigurable Organic Electrochemical Transistors with High Dynamic Ranges for Fully Integrated Physical Reservoir Computing

Abstract Reconfigurable organic electrochemical transistors (r‐OECTs) are considered a promising platform for fully integrated physical reservoir computing (RC) owing to their dual switching modes, which are suitable for both the reservoir and readout layers. However, their restricted dynamic ranges (DR) have constrained their nonlinearity, high‐dimensional mapping capacity, and the reconfigurability of neural networks for optimal physical RC. In this study, r‐OECTs with a modified PEDOT:PSS channel and solid electrolyte is designed and fabricated, achieving dual‐modal essential synaptic functions (nonvolatile and volatile) through ethylene glycol (EG) content adjustment. The r‐OECTs with the EG of 680 and 1000 µL exhibit exceptional DR of 1.12 × 10⁵ in nonvolatile mode and 1.20 × 10³ volatile mode, respectively, representing a significant improvement over previously reported OECTs. The nonvolatile mode exhibits long‐term memory with robust and gradual long‐term potentiation (LTP) and long‐term depression (LTD), while the volatile mode demonstrates short‐term memory for extracting reservoir states. The r‐OECTs, functioning in two modes as the reservoir and readout layers with high DRs, enable accurate classification of human activities such as jogging (J), brushing teeth (B), and folding clothes (F) with an accuracy of 84.38% in a fully integrated physical RC system.