All-Optical Synapses Based on a Mechanoluminescent Material.

Neuromorphic computing systems hold promises to overcome the inefficiencies of conventional von Neumann architecture, which are constrained by data transfer bottlenecks. However, conventional electrically modulated synapses face inherent limitations such as limited switching speed, elevated power consumption, and substantial interconnection loss. Optical signaling offers a transformative alternative, leveraging ultrafast transmission, high bandwidth, and minimal crosstalk. Here, an all-optical synapse based on a mechanoluminescent material of Li0.1Na0.9NbO3:Pr3+ (LNN:Pr3+) is presented, which emulates biological synapses, including homologous and heterologous synaptic behaviors, through optical signal processing. The engineered trap depth distribution of LNN:Pr3+ enables multi-stimuli response to UV light, mechanical force, and thermal input, replicating diverse synaptic functionalities such as short-term potentiation (STP), long-term potentiation (LTP), paired-pulse facilitation (PPF), and learning-experience behavioral adaptation. Furthermore, its utility is showcased in hardware-level denoising and multimode-fused perception, achieving spatiotemporal feature extraction in dynamic environments. This work not only sheds light into designing fully optical synapses but also bridges mechanoluminescence (ML) with neuromorphic engineering, advancing energy-efficient, light-driven artificial intelligence technologies.