Machine learning-guided exploration of carbon-based photothermal materials for solar evaporation

Currently, solar thermal desalination stands out as a potential solution to address global freshwater shortage issue. However, the relationship between performances and structures of photothermal materials for solar evaporation as well as operational conditions remains unclear, hindering traditional screening of high-performance materials. Machine learning (ML) as a useful artificial intelligence tool offers a high promise to explore highly efficient photothermal materials by predicting their performances via their structural characteristics in data-driven fashion, thereby saving time and resources. Herein, four ML models are used for predicting the evaporation rate and efficiency of carbon-based photothermal materials, and among which the Voting Regressor model demonstrates the best performance, achieving lowest mean square errors of 0.2551 kg² m⁻⁴ h⁻² for evaporation rate and 0.0048 for evaporation efficiency, respectively. By utilizing Shapley Additive Explanation (SHAP) analysis, the influence of material properties and operational conditions are revealed on evaporation performance. Furthermore, the reliability of the ML approach is confirmed through water evaporation tests using biochar derived from the pyrolysis of loofah. This study opens a new start by introducing ML strategy in exploring high-performance photothermal materials, accelerating the development of solar evaporation technology.