Optical Charge Trap Memory Based on Graphene/ZnO Heterostructures for Long‐Term Retention and Adaptive Learning

Abstract Optoelectronic neuromorphic devices promise real‐time processing of unstructured biometric data, yet challenges remain in achieving material non‐toxicity and long‐term synaptic retention, along with a clear understanding of the underlying mechanisms. Here, an optical charge trap memory (CTM) is reported based on a graphene/ZnO nanoparticle (GZO) heterostructure that addresses these limitations through its large photoconductive gain and abundant charge trap sites. Distinct from previous nanoparticle/graphene photodetectors, the work leverages interfacial trap‐mediated processes to implement stable optoelectronic neuromorphic functionalities, including long‐term potentiation, retention, and efficient relearning. The GZO CTM exhibits robust charge trapping characteristics, with charge retention exceeding 54 h at the e −1 loss point, attributed to interfacial trap states and a substantial energy barrier at the GZO interface. These trapped charges enable stable potentiation and efficient memory reprogramming, requiring significantly fewer pulses after partial memory loss. Furthermore, artificial neural network simulations based on GZO CTM characteristics demonstrate rapid convergence and near‐unity accuracy in handwriting recognition tasks within 20 training epochs, even under noise conditions. This study highlights the potential of trap‐engineered GZO heterostructures as scalable, energy‐efficient platforms for biocompatible neuromorphic computing, particularly in wearable systems requiring stable optical memory and real‐time signal processing.