Accelerated photonic design of coolhouse film for photosynthesis via machine learning

Controlling the suitable light, temperature, and water is essential for plant photosynthesis. While greenhouses/warm-houses are effective in cold or dry climates by creating warm, humid environments, a cool-house that provides a cool local environment with minimal energy and water consumption is highly desirable but has yet to be realized in hot, water-scarce regions. Here, using a synergistic genetic algorithm and machine learning, we propose and demonstrate a coolhouse film that regulates temperature and water for photosynthesis without requiring additional energy or water. This scalable film, selected from hundreds of potential designs, selectively and precisely transmits sunlight needed for photosynthesis while reflecting excess heat, thereby reducing thermal load and evapotranspiration. Its optical properties also exhibit weak angle dependence. In demonstrations in subtropical and arid regions, the film reduces temperatures by 5–17 °C and cuts water loss by half, resulting in more than doubled biomass yield and survival rates. It also improves crop resistance to heat and drought in greenhouse cultivation. The integration of machine learning and photonics provides a powerful toolkit for designing photonic structures and devices aimed at sustainability. This study uses machine learning to design a coolhouse film that regulates temperature and water evaporation to maximize plant photosynthesis efficiency. The film selectively transmits the sunlight needed for photosynthesis, improving crop yield and survival rates in hot, arid regions.