Self-Powered Phototriggered Memristor Array with pW-Level Computing for Monolithic in-Sensor Vision

2D Ruddlesden-Popper (RP) hybrid perovskites are widely used in the field of optoelectronic devices due to their high carrier mobility and absorption coefficient. Herein, BA₂MAPb₂Br₇ (BMPB) single crystals were synthesized via the quasi-static cooling method, with a thorough investigation of the nucleation mechanism and the intrinsic relationship between rectangular and hexagonal morphologies. Remarkably, BMPB exhibits outstanding memristive performance, featuring an ultrahigh on/off ratio (∼10⁵), ultralow power consumption of ∼82.8 pW, and a long retention time (>22 000 s). Leveraging its inherent ferroelectric polarization and optoelectronic properties, we demonstrate a monolithic 5×5 optoelectronic memristor array that unifies sensing, memory, and computing functions. The array device exhibits self-powered performance under 410 nm illumination, with responsivity of 0.96 A/W and detectivity of 2.89×10⁸ Jones. Through digital logic circuit design and Vivado verification, we successfully implement solar-tracking time prediction and sunflower growth-stage monitoring. Critically, this integrated architecture enables hardware-level in-sensor computing for real-time applications. Our work provides valuable insights for next-generation in-sensor computing devices, providing a material-level solution based on 2D RP perovskites for energy-efficient AI and IoT systems.