Crystallinity-engineered heterojunction interfaces for Robust visual memory in optoelectronic artificial synapses

A biomimetic optoelectronic synaptic device based on WO₃/InWZnO heterojunction transistor is proposed to emulate human visual signal processing. By tailoring the crystallinity of the WO₃ layer through thermal annealing, the devices exhibit significantly enhanced optoelectronic performance, featuring a responsivity (R) of 18.8 A W^–1, a signal-to-noise ratio (SNR) of 4.5 × 10⁷, and a specific detectivity (D*) of 1.7 × 10¹² Jones under 650 nm optical stimulation. The devices also demonstrate diverse synaptic plasticity behaviors, including paired-pulse facilitation (PPF) and post-tetanic potentiation (PTP), highlighting their capability to mimic complex neural functions. In addition, the retention characteristics are also significantly improved, further supporting persistent information retention in neuromorphic vision systems. The synaptic parameters extracted from the proposed devices were implemented in a multilayer perceptron (MLP) model to assess their system-level applicability. Devices incorporating crystalline WO₃ demonstrated improved recognition performance, achieving validation accuracies of 95.1 % and 85.8 % under Gaussian and pepper and salt noise conditions, respectively. Furthermore, under a 2 × 2 optoelectronic synapse array configuration, the proposed devices exhibit superior nonvolatile characteristics, retaining 61 % of their initial current after 200 s. These characteristics underscore the potential of the proposed devices for high-performance, noise-resilient memory-based neuromorphic computing applications.