Neural Network‐Assisted Angle‐Detecting Photodetector Based on Microstructure Optimization

ABSTRACT

The functionality of conventional photodetectors is confined to measuring the intensity of incident light, lacking the capability to discern its angle of incidence—a limitation that constrains their application scope. To address this, we present a novel photodetector architecture designed for direct angle detection. This is achieved by integrating 3D microstructures that optically modulate the incident light, thereby encoding the incidence angle into a variable photocurrent signal. A neural network is subsequently employed to decode this relationship, enabling the precise inversion of the incident angle from normalized photocurrent readings. Experimental results demonstrate that an optimized device featuring parallel dual‐cylinder microstructures achieves an average angle detection error of 4.62°. Furthermore, it attains high detection accuracies of 82% and 88% across two orthogonal angular dimensions at a 5° resolution. This work offers a new avenue of thought for research in the field of optical angle detection, while the integration of neural networks with micro/nano‐devices paves the way for the development and exploration of next‐generation intelligent sensors.