Complementary Metal-Oxide Semiconductor-Compatible Phototransistor Platform for Neuromorphic Broadband Vision

Abstract

Photodetection and neuromorphic computing promises sensing at the edge, yet most optoelectronic synapses lack spectral breadth or scalable silicon integration. Here we report a dual-gate synaptic transistor integrating colloidal quantum dots (CQDs) with a silicon-on-insulator (SOI) channel in a complementary metal-oxide semiconductor (CMOS)-compatible platform. This architecture synergistically combines the SOI channel for efficient charge transport and signal amplification with the CQD layer serving a dual role as a broadband absorber (300–1700 nm) and a charge-trap medium. The device exhibits a responsivity of 639 A W–1 and specific detectivity exceeding 3 × 1012 Jones at 1550 nm, parallelly emulating short- and long-term plasticity, multilevel conductance tuning, and highly programmable synaptic dynamics. The top gate enables reversible potentiation and depression, and self-powered operation supports low-power sensing and light-driven learning. We demonstrate system-level utility in a synaptic array that achieves pattern recognition with peak accuracy exceeding 96%, establishing a pathway to silicon-integrated, energy-efficient neuromorphic vision.