All‐Inorganic Lead‐Free Cs₂AgBiBr₆/ZnO Artificial Retina Synapse Based on Photoelectric Synergistic Dual‐Mechanism for Neuromorphic Computing

Adaptive learning capability of optoelectronic synaptic hardware holds promising application prospects in next generation artificial intelligence, and the development of biometric retina perception is sternly hampered by three crucial issues, including well-balance between excitatory and inhibitory, non-volatile multi-state storage, and optimal energy consumption. In this work, a novel Cs2AgBiBr6/ZnO non-volatile optoelectronic synapse is proposed and successfully programmed with optical excitatory and electronic inhibitory in the light of dual-mechanism: Lead-free perovskite Cs2AgBiBr6 guarantees abundant photogenerated carrier concentration, and the process of carrier capture and release occurs in ZnO layer, which can collaboratively modulate various synaptic plasticity behaviors depending on distinct stimulus. Consequently, multi-bit storage is attained with the dual-mechanism non-volatile memory (DNVM) as a function of consecutive light spikes. The energy consumption of the DNVM is 1.85 nJ at a single light spike, and an ultra-low one of 13.8 fJ is triggered with a single electrical pulse, which approximatively meets the requirement of the biological synaptic event energy consumption. The performance of the DNVM is further evaluated with the Pavlov's classical conditioning experiment and visual hardware system, offering an exciting paradigm for implementing on-chip adaptive visual perception and neuromorphic computing.