Two‐Terminal Perovskite Optoelectronic Synapse for Rapid Trained Neuromorphic Computation with High Accuracy

Abstract

Emerging neural morphological vision sensors inspired by biological systems that integrate image perception, memory, and information computing are expected to transform the landscape of machine vision and artificial intelligence. However, stable and reconfigurable light‐induced synaptic behavior always relies on independent gateport modulation. Despite its potential, the limitations of uncontrollable defects and ionic characteristics have led to simpler, smaller, and more integration‐friendly two‐terminal devices being used as sidelines. In this work, the synergy between ion migration barriers and readout voltage was proven to be the key to realizing stable, reconfigurable, and precisely controllable postsynaptic current in two‐terminal devices. Following the same mechanism, optical and electrical signal synchronous triggering is proposed to serve as a preprocessing method to achieve a recognition accuracy of 96.5%. Impressively, the gradual ion accumulation during the training process induces photocurrent evolution, serving as a reference for the dynamic learning rate and boosting accuracy to 97.8% in just 10 epochs. The PSC modulation potential under short optical pulse of 20 ns is also revealed. This optoelectronic device with perception, memory, and computation capabilities can promote the development of new devices for future photonic neural morphological circuits and artificial vision.This article is protected by copyright. All rights reserved

In-sensor neuromorphic computing using perovskites and transition metal dichalcogenides

“…Compared to other materials (e.g. metal oxide [47][48][49], organic material [50][51][52], graphene [53] and gallium arsenide [54][55][56]), perovskites and TMDs exhibit adjustable band gap and high quantum efficiency [57][58][59][60][61][62]. In addition, by changing the halogen species in perovskites and constructing the TMD based heterostructures, the photoelectric properties of materials and devices can be modulated and optimized [63].…”

“…In addition, by changing the halogen species in perovskites and constructing the TMD based heterostructures, the photoelectric properties of materials and devices can be modulated and optimized [63]. Therefore, perovskites and TMDs have been widely utilized for developing optoelectrical devices [57,60,[64][65][66]. In this section, we will introduce two important optoelectronic materials, perovskites and TMDs, including their structural and optoelectronic characteristics.…”