Light-Stimulated Artificial Synapses with Accelerating Photopic Adaption Based on III-Nitride Heterojunction Transistor

Le résumé fourni par la source

Neuromorphic computing, inspired by biological synapses, has emerged as a pivotal approach to overcome the limitations of von Neumann architecture. In this work, we employ a three-terminal III-nitride metal-oxide-semiconductor field-effect transistor (MOSFET) structure based on the GaN-on-silicon platform, characterized by ultralow power consumption, to emulate artificial synapses through light stimulation. Remarkably, our device exhibits a minimal power consumption of 7 fJ per synaptic event, surpassing that of biological synapses (∼10 fJ). Notably, we observe that the paired pulse facilitation (PPF) index can be modulated by the external light intensity, pulse width, and light pulse interval. By manipulating the light intensity and pulse width, we achieve a PPF index exceeding 300% in our device. Furthermore, our devices demonstrate gate-tunable synaptic plasticity, enabling electric/light cooperative control and increasing the tuning freedom. Finally, acceleration of photopic adaption and a 91.28% handwritten digit recognition accuracy are achieved, which provide strong support in potential visual sensory applications.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.