Researchers at the University of California Davis have developed an AI-enhanced spectrometer-on-a-chip capable of analyzing light and chemical signatures using a device smaller than a grain of sand. Published in the journal Advanced Photonics, the study describes a silicon-based sensor that could dramatically reduce the size and cost of spectroscopy systems used in medicine, food inspection, and environmental monitoring. [1]
Traditional spectrometers rely on bulky optical components such as prisms or gratings to separate light into different wavelengths. The new chip replaces that hardware with artificial intelligence. Instead of spreading light physically, the system uses 16 specially engineered silicon detectors that each respond differently to incoming light. A neural network then reconstructs the original spectrum from the encoded signals with roughly 8-nanometer resolution.
The researchers also introduced photon-trapping surface textures that allow standard silicon sensors to detect near-infrared wavelengths up to 1100 nanometers—a range important for biomedical imaging because it penetrates human tissue more effectively than visible light.
According to the paper, the chip occupies just 0.4 square millimeters while maintaining strong noise resistance and ultrafast response times. The team says the technology could pave the way for portable diagnostic tools, wearable health monitors, pollution sensors, and compact hyperspectral imaging systems.