Metasurface Absorber Hits 98.3% Light Absorption
Researchers used machine learning to design a tungsten-based structure that selectively absorbs and emits light across a broad wavelength range.
Updated on Nov. 16, 2026 in Energy

Researchers at the Changchun Institute of Optics, Fine Mechanics and Physics of the Chinese Academy of Sciences designed a tungsten-based metasurface absorber and emitter using machine learning. The optimized structure reports an average absorptivity of 98.3% across the 350-1,000-nanometer range, as published in Applied Thermal Engineering.
Why it matters
Conventional absorbers have limited spectral selectivity and lose performance at steep incident angles or under demanding conditions. A design that holds high absorption across angles and temperatures could improve devices that depend on precise control of light and heat.
The study rests on finite-difference time-domain simulations paired with comparison of six machine learning algorithms, with XGBoost identified as the strongest predictor of absorption. The reported 99.4% conversion efficiency at 2,000 K is a calculated value, and the extraction does not indicate whether the design has been physically built and tested.
The players
Changchun Institute of Optics, Fine Mechanics and Physics
A research institute under the Chinese Academy of Sciences that conducts work in optics and photonics.
The details
The team used electromagnetic field analysis and structural-component contributions to explain the enhanced optical response of a metal-insulator-metal design with nanoscale resonators. Tungsten was chosen for its favorable optical absorption and suitability for high-temperature operation.
Timeline
Article describing the study published November 16, 2026.
The Big Picture
The work extends a long effort in metasurface photonics by using data-driven search to tune nanoscale absorbers rather than relying on manual parameter sweeps. Pairing machine learning with electromagnetic simulation points toward faster design cycles for selective thermal emitters.
Advances like this could eventually feed into photodetectors, optical sensors, and solar thermophotovoltaic systems that rely on capturing sunlight and managing heat. Commercial use would depend on experimental confirmation of the simulated performance and on manufacturing the nanoscale structure at scale.
The takeaway
Machine learning can speed up the search for optical structures with precise light control. Watch for experimental fabrication results to confirm the simulated performance before expecting practical devices.
Further reading
Explore more on Energy developments in light absorption and thermal conversion technology.



