For decades, the ability to visualize the chemical composition of materials, whether for diagnosing a disease, assessing food quality, or analyzing pollution, depended on large, expensive laboratory instruments called spectrometers. These devices work by taking light, spreading it out into a rainbow using a prism or grating, and measuring the intensity of each color. The problem is that spreading light requires a long physical path, making the device inherently bulky.
A recent study from the University of California Davis (UC Davis), reported in Advanced Photonics, tackles the challenge of miniaturization, aiming to shrink a lab-grade spectrometer down to the size of a grain of sand, a tiny spectrometer-on-a-chip that can be integrated into portable devices. The traditional approach of spatially spreading light is abandoned in favor of a reconstructive method. Instead of physically separating each color, the new chip uses only 16 distinct silicon detectors, each engineered to respond slightly differently to incoming light. This is analogous to giving a handful of specialized sensors a mixed drink, with each sensor sampling a different aspect of the drink. The key to deciphering the original recipe is the second part of the invention: artificial intelligence (AI).
The heart of this innovation lies in two technological breakthroughs. First, the team engineered the surfaces of standard silicon photodiodes with specialized photon-trapping surface textures (PTSTs). Silicon is typically effective at sensing visible light but is notoriously poor at sensing near-infrared (NIR) light (wavelengths up to 1100 nm), which is critical for many applications, such as biomedical imaging, because it penetrates human tissue more deeply than visible light. The PTST surface acts like a cleverly designed texture that forces NIR photons to scatter within the thin silicon layer instead of passing straight through. This dramatically increases the likelihood that the silicon absorbs light, making the entire chip sensitive across a broad spectral range.
Beyond simple color detection, the architecture employs high-speed sensors to provide an inherent, ultra-fast capability for measuring photon lifetime. This temporal precision allows the device to capture fleeting light–matter interactions that are invisible to traditional instruments.
Second, the chip uses a powerful fully connected neural network (AI). Since the 16 unique detectors only capture encoded, noisy signals, the AI is trained on thousands of examples to learn the complex, hidden relationship between the detectors' raw outputs and the original, pure light spectrum. The AI addresses this "inverse problem," reconstructing the light spectrum with high accuracy (around 8 nm resolution). This computational method completely removes the need for bulky optics.
The final result is a system with a minimal footprint (0.4 square mm), high sensitivity, and strong noise resistance. The AI-augmented chip can maintain signal clarity even in the presence of significant electrical interference, a major challenge in portable, low-cost electronics. By extending the sensing range of silicon into the crucial NIR spectrum while enabling high performance through machine learning, this technology establishes a pathway for truly integrated, real-time hyperspectral sensing across applications ranging from advanced medical diagnostics to environmental remote sensing.
