Current/Voltage Dual-Modal Hybrid Ionotronic Oxide Dendrite Transistor for Neuromorphic Computing.

Current and voltage are fundamental variables in modern electronic technology. The developments of voltage-driven neuromorphic devices have led to the breakthrough of the von Neumann bottleneck, which has attracted significant attention. As a comparison, implementing synaptic functions on current-driven neuromorphic devices still poses challenges. In this work, a current/voltage dual-modal hybrid ionotronic oxide dendrite transistor (HIODT) is proposed. The HIODT exhibits good electrical performance and rich ion dynamics. Basic synaptic functions and "learning-experience" behaviors were mimicked under current/voltage dual-modal modulation. Interestingly, an effective linear synaptic weight updating strategy is implemented using current and voltage spike schemes. Thus, excellent recognition accuracies of >90% for small digits and of ∼80% for the Fashion-MNIST (Modified National Institute of Standards and Technology) database are achieved by employing a three-layer artificial neural network. Moreover, the device can emulate key features of pain perceptual nociceptors (PPN), including sensitization and desensitization. Remarkably, spatiotemporal dendritic integration and associative learning behavior have also been mimicked under current/voltage dual-modal modulation. The present work provides unique insights for the current/voltage dual-modal spiking strategy for functional neuromorphic devices.