What the study reports
Xiao and colleagues describe a machine-learning approach to thermal emitter design that combines material selection with three-dimensional structural parameters. Its aim is to target both broad emission spectra and selected spectral bands using sparse data. [1]
What this does not establish
This brief is based on the accessible abstract, not an independent assessment of the full experimental record. It does not quantify cooling performance, fabrication yield or durability. Spectral design alone is insufficient to predict cooling in a particular environment.
Our engineering reading
The useful methodological shift is to search geometry and material together. Engineering adoption needs measured spectra, fabrication constraints and an energy balance under relevant operating conditions.
Editorial interpretation; not an additional measured result.Ultrabroadband and band-selective thermal meta-emitters by machine learning
Chengyu Xiao et al. · Nature · 2 July 2025
DOI: 10.1038/s41586-025-09102-ySource access: Publisher abstract and publication metadata. Checked 3 October 2026.