Neuromorphic machine vision has attracted extensive attention on wide fields. However, both current and emerging strategies still suffer from power/time inefficiency, and/or low compatibility, complex device structure. Here we demonstrate a driving-voltage-free optoelectronic synaptic device using non-volatile reconfigurable photovoltaic effect based on MoTe₂/α-In₂Se₃ ferroelectric p-n junctions. This function comes from the non-volatile reconfigurable built-in potential in the p-n junction that is related to the ferroelectric polarization in α-In₂Se₃. Reconfigurable rectification behavior and photovoltaic effect are demonstrated firstly. Notably, the figure-of-merits for photovoltaic effect like photoelectrical conversion efficiency non-volatilely increases more than one order. Based on this, retina synapse-like vision functions are mimicked. Optoelectronic short-term and long-term plasticity, as well as basic neuromorphic learning and memory rule are achieved without applying driving voltage. Our work highlights the potential of ferroelectric p-n junctions for enhanced solar cell and low-power optoelectronic synaptic device for neuromorphic machine vision.

神经形态机器视觉在广泛的领域引起了广泛的关注。然而，当前和新兴的策略仍然存在功率/时间效率低下和/或兼容性低、设备结构复杂的问题。在这里，我们展示了一种使用基于 MoTe ₂ /α-In ₂ Se ₃铁电 pn 结的非易失性可重构光伏效应的无驱动电压光电突触装置。该功能来自与 α-In ₂ Se _{3 中}的铁电极化相关的 pn 结中的非易失性可重构内置电位. 首先展示了可重构整流行为和光伏效应。值得注意的是，诸如光电转换效率之类的光伏效应的品质因数非易失性地增加了不止一个数量级。基于此，模拟了视网膜突触样视觉功能。光电短期和长期可塑性，以及基本的神经形态学习和记忆规则是在不施加驱动电压的情况下实现的。我们的工作突出了铁电 pn 结用于增强型太阳能电池和用于神经形态机器视觉的低功率光电突触装置的潜力。