Exploration of Double Perovskite Material Space via Machine Learning for Tandem Solar Cells

Abstract

The extensive compositional landscape of double perovskite materials (A₂BB'X₆) provides a promising basis for the design of tandem perovskite solar cells. However, the vast combinatorial space and multiple phase spaces also pose significant challenges in the engineering of efficient and stable tandem perovskite solar cells. In this work, machine learning is used to search for double perovskite materials suitable for tandem perovskite solar cells in a wide dataset containing over  1 569 248  hypothetical double perovskite materials.  Finally,  through  first-principles  calculations,  it was found that double perovskite materials with two different space groups (i.e.,  Fm3m-Rb₂BiAgI₆ and I4/m-K₂CrErBr₆) with stable and suitable bandgaps could serve as promising absorbers for tandem perovskite solar cells. Utilizing semiconductor device simulations, the tandem Rb₂BiAgI₆ solar cell and tandem K₂CrErBr₆ solar cell demonstrate power conversion efficiency of 24.00% and 42.78%, respectively. This research not only provides valuable insights into the compositional engineering of double-perovskite materials but also offers critical guidance for the development of high-efficiency tandem perovskite solar cells.