Giant Switchable Remanent Polarization and Photocurrent in Ferroelectric Thin Film Photomemristor for In Situ Training

Ferroelectric thin film photomemristors with a stable and tunable short-circuit current offer a promising solution for developing machine vision systems with in situ training. However, previously reported devices exhibit inadequate remanent polarization and limited short-circuit currents, posing significant challenges to the efficient implementation of in situ training in complex environments. In this work, we realized the in situ training in a machine vision system through the Pt/SrRuO₃/PbZr_0.4Ti_0.6O₃/SrRuO₃/SrTiO₃ device. Induced by the bottom electrode-induced stress, the device shows a large remanent polarization of 94 µC cm^-2 and an ultra-high short-circuit current of 183.35 nA. Furthermore, the stable and tunable multilevel photocurrents of this device enable image edge processing and in situ multifunctional neural signal training for miniature vehicles in machine vision systems. This work demonstrates the immense potential of ultra-high-performance ferroelectric thin film photomemristors for advancing machine vision systems research.