All-photonic synapse based on iron-doped lithium niobate double metal-cladding waveguides

In an artificial neural network composed of neuromorphic computing units, the resistance of the memristor can be modulated dynamically and repeatedly under external stimuli, such as electric fields, magnetic fields, and light illumination, leading to variations in local conductivity and memory effects. Here we show, using an all-photonic memristor, that the memristance arises naturally in an optical system in which solid-state ionic transport is realized under low-power external incident light. These results show an extremely stable multilevel storage weight and a high signal-to-noise ratio. In all-photonic memristors, the light signal is employed as the extra stimuli of the memristor devices to ensure a large memory window and a variation margin of multiple storage levels.