ABSTRACT Reservoir computing (RC) is a neuromorphic computing paradigm for efficient temporal data processing. However, its hardware implementation is hindered by challenges such as high operating bias and energy consumption, as well as the functional heterogeneity of the components involved. Herein, we present an ion‐gated synaptic transistor with an iodonium salt covalently functionalized Ti 3 C 2 T x (f‐Ti 3 C 2 T x ) channel that modulates both the band structure and ion dynamics through strong electron‐withdrawing groups. The device can achieve an ultralow‐energy consumption (0.77 fJ per spike) at a low operating voltage (0.6 V). The dynamic configuration between volatile and nonvolatile modes via tunable synaptic characteristics facilitates integration into a homogeneous RC system, thereby reducing process complexity and hardware costs. The f‐Ti 3 C 2 T x synaptic transistor with nonlinear short‐term memory exhibits adaptive time characteristics, serving as a physical reservoir, while the nonvolatile device simulates the synaptic behavior of the readout network, demonstrating linear weight updates. The homogeneous RC system can perform handwritten digit recognition with an accuracy of 97.9% and practical time‐series prediction, offering high hardware robustness and lower training costs than conventional neural networks. This work provides a framework for the design and application of Ti 3 C 2 T x ‐based synaptic transistors as reliable bioinspired hardware platforms for advancing neuromorphic computing.
Xuan et al. (Sat,) studied this question.