Two-terminal photoelectric synapses for attojoule neuromorphic vision.

Neuromorphic vision is reaching an important transition from proof-of-concept device demonstrations toward quantitative energy benchmarks and scalable hardware integration. Yet, a chemistry-centered understanding of what attojoule-scale neuromorphic vision requires from materials and interfaces remains underdeveloped. Two-terminal (2T) photoelectric synapses are particularly well positioned for this transition by integrating sensing, memory, and computation within compact device architectures, while their performance, stability, scalability, and sustainability are fundamentally governed by materials selection and interfacial chemistry. This review therefore examines 2T photoelectric synapses from a coupled materials-device-system perspective. Starting from biological synapses and the human visual pathway, key biological concepts are translated into measurable device metrics and learning rules, followed by a comparison of the structural and operational characteristics of 2T and three-terminal (3T) architectures. The principal physical mechanisms governing photoelectric synaptic plasticity are then systematically discussed. Building on this mechanistic framework, materials-by-design strategies are examined across oxides, organic semiconductors, halide perovskites, and two-dimensional materials, with particular attention to their roles in device performance, stability, and energy efficiency. Representative neuromorphic vision functions, ranging from retina-inspired preprocessing to image recognition, are further reviewed together with device- and system-level energy considerations. Finally, safe-and-sustainable-by-design (SSbD) principles are incorporated to connect materials development with long-term technological and environmental requirements. Overall, this review highlights how coordinated materials, device, and system design can guide the development of energy-efficient, scalable, stable, and sustainable 2T photoelectric synapses for next-generation neuromorphic vision.