Retina-Inspired 2D Semiconductor NIR Sensor with PRO Architecture for Photodetection

Light detection and ranging (LiDAR) technology is critical for autonomous driving, which relies on abundant near-infrared (NIR) photodetectors to accurately capture three-dimensional spatial information. CMOS-based LiDAR detection systems are less efficient with respect to computational efficiency, inherently limited by their von Neumann architecture, leading to high latency and significant computational demands. Retina-inspired neuromorphic sensors integrated from sensing and computation present a promising alternative but need additional analog-to-digital converters (ADCs). Here we present the photosensitive ring oscillator (PRO) based visual afferent neuro-biosensors for LiDAR-based sensing applications. The PRO employs monolayer MoS₂ as channels decorated by Nd³⁺/Yb³⁺/Er³⁺ tridoped NaYF₄ up-conversion nanoparticles (UCNPs) to achieve near-infrared (NIR)-triggered oscillation frequency modulation. The PRO architecture avoids the need for ADCs, offering enhanced noise immunity and system simplicity. The simulated VoxelNet neural network effectively preprocesses images by extracting environmental information, achieving high recognition accuracy especially in low-light conditions. This work presents a new paradigm for developing a real-time, high-accuracy LiDAR sensing system.