Self-Powered Flexible Triboelectric-Gated Ion-Gel Transistor for Neuromorphic Tactile Sensing and Human Activity Recognition.

Abstract

Wearable and flexible neuromorphic devices capable of accurately emulating synaptic behaviors while autonomously responding to mechanical stimuli hold great promise for intelligent sensing and bio-inspired computing. Here, we present a fully flexible synaptic device in which a graphene-channel ion-gel-gated transistor (g-IGT) fabricated on a plastic substrate is directly driven by a poly(vinylidene fluoride-co-trifluoroethylene)-based triboelectric nanogenerator (TENG), enabling self-powered tactile sensing and analog weight storage. The device emulates the multistore memory hierarchy through sensory-short-term-long-term memory transitions, characterized by synaptic decay times of ∼70 ms (sensory), 0.2-0.45 s (short-term), and >2.0 s (long-term). Furthermore, rate-coded learning is demonstrated through spike-rate-dependent plasticity, enabling frequency-selective potentiation and depression. When the experimentally measured weight-update profiles are implemented in a single-layer perceptron for human activity recognition, the neural network system achieves >88% classification accuracy, even under 1.1 MPa bending stress. Furthermore, the network system maintains relatively stable performance (>75%) even under the extreme environment with a high noise level. These results establish the TENG-driven g-IGT as a viable route toward mechanically compliant, battery-free neuromorphic platforms capable of sensing, learning, and adapting in situ.