Floating-gate synaptic transistors with bending resistance and strong charge storage capability via interfacial redispersion and reconfiguration

Abstract

The development of artificial neural networks based on neuromorphic devices is expected to break through the von Neumann bottleneck and promote the further development of artificial intelligence technology. However, there is still much room for improvement in the charge storage capacity and bending resistance of the current neuromorphic devices. In this work, we leveraged the interfacial redispersion and reconstruction phenomena occurring during the preparation of stacked films using the same dispersant to improve the carrier interlayer transport barrier and the interlayer bonding strength of the films. The constructed floating-gate synaptic transistor exhibits a charge storage window of up to 25 V (@V_G: −20 to 20 V) and maintains excellent performance after 10 000 bending cycles with a radius of 4 mm. The device successfully simulates typical behaviors of biological synapses, such as long-term plasticity (LTP), paired-pulse facilitation/depression (PPF/PPF), and spike-timing-dependent plasticity (STDP). In addition, the device demonstrates excellent photoelectric response characteristics and photocurrent retention capability both before and after bending, enabling photoelectric bimodal-response. Importantly, simulations conducted with a simulator based on a three-layer artificial neural network (ANN) show that the device achieves high recognition accuracy (97.1%, 94.4%, and 91.9%), which is close to the ideal values (97.8%, 94.5%, and 93.1%). This indicates that the device has great application potential in flexible memory devices, electronic skin, and neural network simulation and computing.