Recently, Professor Wang Jinlan from the School of Physics of Southeast University put forward a new intelligent material design strategy by combining machine learning technology and density functional theory (DFT), and successfully predicted more than 5,000 potential organic-inorganic hybrid perovskites. The band gaps of materials (HOIPs), and selected a variety of environmentally stable, band gap-free lead-free HOIPs solar cell materials. The research results were published online in the Nature issue of Nature Communications. Accelerated discovery of stable lead-free hybrid organic-inorganic perovskites via machine learning.
In the context of the energy crisis, there is an urgent need for high-efficiency and non-toxic new solar cell materials to replace traditional fossil energy. However, traditional material design methods have problems such as inefficiency and serious waste of resources, especially in the face of tens of thousands of species. In the case of candidate materials, this method is even more stretched. Recently, ML technology has emerged in the field of material design. By bypassing complex quantum mechanics, ML technology can not only greatly accelerate the design of new functional materials, but also learn from material data. The basic structure-activity relationship of materials. This new material design strategy has been successfully applied in the fields of molecular organic light-emitting diodes, shape memory alloys, piezoelectric bodies, etc. However, organic-inorganic hybrid calcium-titanium has not yet been widely used in photovoltaic applications. The mine field has been effectively explored.
Professor Wang Jinlan from the School of Physics, Southeast University, based on ML technology and DFT calculation, developed a targeted driving method for finding efficient and stable lead-free HOIPs. Researchers trained ML models from 212 reported HOIPs band gap values. , successfully predicted the band gap of more than 5,000 potential HOIPs, and finally screened six orthogonal lead-free HOIPs with appropriate solar band gap and room temperature thermal stability, two of which have direct band gap in the visible region and excellent environmental stability. The researchers also used ML technology for big data mining to obtain the key factors affecting the performance of ideal HOIPs solar cells. This targeted driving method overcomes the main obstacles of traditional trial and error methods, not only can achieve DFT accuracy instantaneously, but also Suitable for small data sets. This work has greatly accelerated the design process of hybrid perovskite materials with potential for photovoltaic applications, and can be applied to the design and discovery of other functional materials. The first author of this paper is a master's degree in physics at Southeast University. Lu Shuaihua, a grade student, and teacher Zhou Yihua, a teacher of the School of Physics, is the co-first author. Professor Wang Jinlan is the sole correspondent of the paper. As a key national research and development projects funded by the program, the National Outstanding Youth Fund.