Machine learning has revolutionized the search for efficient solar materials, particularly in the realm of perovskites. These materials are gaining popularity in the solar industry due to their lower costs and simpler manufacturing processes compared to traditional silicon solar cells.
A recent study conducted by researchers from EPFL, in collaboration with partners in Shanghai and Louvain-La-Neuve, has utilized advanced computational techniques and machine learning to identify optimal perovskite materials for photovoltaic applications. By focusing on the critical factor of band gap – the specific energy range that determines a material’s ability to absorb sunlight and convert it into electricity effectively – the team aims to improve the efficiency and cost-effectiveness of solar panels.
The researchers first compiled a comprehensive dataset of band-gap values for 246 perovskite materials using cutting-edge calculations based on hybrid functionals. These sophisticated computational methods allowed for a more accurate prediction of band gaps compared to conventional approaches like Density Functional Theory, especially for materials like perovskites where electron interaction and polarization effects play a crucial role.
By training a machine-learning model on this dataset and applying it to a database of around 15,000 candidate materials, the team identified 14 new perovskites with promising band gaps and high energetic stability, making them excellent candidates for high-efficiency solar cells.
This groundbreaking research, published in the Journal of the American Chemical Society, paves the way for the development of next-generation solar technologies that are both efficient and cost-effective. With machine learning accelerating the discovery of optimal perovskite materials, the solar industry is poised for a significant transformation towards sustainable energy solutions.
