Researchers from the Changchun Institute of Optics, Fine Mechanics and Physics of the Chinese Academy of Sciences, together with colleagues, developed a design strategy for a nanostructured tungsten surface that selectively absorbs and emits light. The work was published in the journal Applied Thermal Engineering.

The new surface, called a metasurface (an artificial structure with periodic elements smaller than the wavelength of light, which gives it unusual optical properties), is capable of efficiently absorbing visible and near-infrared light, as well as controlling thermal radiation. This could be useful in solar thermophotovoltaic systems: they first convert sunlight into heat, and then the thermal radiation is converted into electricity using photovoltaic cells.

Conventional absorbers often have limited spectral selectivity, are sensitive to the angle of incidence of light and lose efficiency under complex conditions. Creating nanostructures that simultaneously solve these problems is difficult, because optimizing their optical properties requires numerous numerical calculations.

To simplify the task, the scientists proposed a tungsten metasurface with a "metal-insulator-metal" structure and ordered nanoscale resonators. Tungsten was chosen for its good light absorption and resistance to high temperatures.

To study the influence of structural parameters on absorption, the researchers combined numerical modeling by the finite-difference time-domain method (a method for calculating the propagation of electromagnetic waves) with machine learning. They compared six algorithms and recognized XGBoost (a machine learning method that builds an ensemble of decision trees) as the most accurate for predicting absorption efficiency and optimizing the structure.

Analysis of the electromagnetic field distribution showed that the combination of nanodisks and nanorings provides strong interaction of light with the structure over a wide range of wavelengths. The influence of the angle of incidence and polarization of light was also assessed.

According to the modeling results, the optimized metasurface achieved an average absorptance of 98.3% in the wavelength range from 350 to 1000 nanometers, and in the 450-700 nanometer range absorption was 99.8%. At angles of incidence up to 60 degrees and both main polarizations, absorption remained above 90%. As a selective thermal emitter, the structure showed a calculated spectral conversion efficiency of 99.4% at a temperature of 2000 K, which indicates potential for directing thermal radiation into wavelengths useful for silicon photovoltaic cells.

The authors note that combining machine learning and metasurface design can improve the performance of solar thermophotovoltaic systems. In addition, the approach could find application in photodetectors, optical sensors and other devices requiring precise control of light. However, these are modeling results, not a finished commercial product.

Source: Tech Xplore