VIS and NIR LED based Voltage-Tunable Sensor for Material Classification

This study introduces a cost-effective and energyefficient material classification system designed to enhance recycling and waste management efficiency. The system employs a voltage-tunable germanium-on-silicon (Ge-on-Si) photodetector, operating across the $400-1600 \mathrm{~nm}$ wavelength range, to distinguish seven material types, including plastics, aluminum, glass, and paper. To ensure broad spectral coverage and stable operation, the system integrates three lowpower LED light sources: one in the visible (VIS) range and two in the short wavelength infrared (SWIR) range. The simplified optical setup and compact sensor holder reduce mechanical complexity, enabling easy integration into automated sorting platforms. Classification is achieved by analyzing photocurrent responses under varying bias voltages using machine learning (ML) algorithms. Empirical results demonstrate rapid and accurate material identification with minimal power consumption, highlighting the system’s potential for scalable and sustainable deployment in modern waste sorting infrastructure.