A research team led by Shen Guozhen and Wang Zhuoran from the School of Integrated Circuits and Electronics at the Beijing Institute of Technology (BIT), recently published a groundbreaking paper titled A symmetry-reconfigurable photodiode for sensing and computing in the prestigious journal Nature Electronics.

The study introduces an innovative symmetry-reconfigurable photodiode (SRPD) that successfully enables dual-mode operation for sensing and computing.

It has been systemically validated in two scenarios: transmission imaging and neuromorphic eye-machine interaction. This development offers a novel device solution for low-power edge visual intelligent systems.

As edge devices like micro-robots, wearable electronics, and smart terminals rapidly evolve, there is a growing demand for visual systems that offer low power consumption, high integration, and real-time processing capabilities. Traditional architectures, which separate image sensing, analog-to-digital conversion, and image processing, are versatile but suffer from high energy consumption during signal conversion, data transmission, and storage computation, making them unsuitable for ultra-low-power edge scenarios.

Processing-in-sensor (PIS) technology is seen as a key solution to this challenge, but its progress is limited by incomplete device mechanisms, insufficient scalability, and poor adaptability to complex scenarios.

Addressing these challenges, the research team designed and built the SRPD based on I-V-VI group semiconductor AgBiS₂. This device features a simple two-terminal structure and is highly compatible with thin-film transistor readout circuits, providing a solid foundation for array integration and system expansion.

The team has successfully achieved monolithic integration on a 64×64 TFT chip, creating a high-transmittance image sensor capable of infrared transmission imaging of targets like silicon wafers and ink. Furthermore, they demonstrated compatibility with MOSFET integration, showcasing its engineering potential for adaptation with silicon-based readout circuits.