Publication date: 28th August 2024
Perovskites are a class of materials with a distinctive crystal structure ABX3. They have
garnered significant attention as promising candidates for solar cell applications. These
materials, typically consisting of a hybrid organic-inorganic lead or tin halide-based
structure, offer exceptional light absorption, charge-carrier mobility, and tunable band
gaps. The ease of fabrication and potential for low-cost production, combined with rapidly
improving power conversion efficiencies that now rival traditional silicon-based solar cells,
underscore their potential in the photovoltaic industry. Challenges remain, particularly
concerning stability and toxicity. In our current study simple and hybrid perovskites and
their properties have been studied through machine learning (ML) models, density
functional theory (DFT) calculations, and synthesis. Addressing stability challenges, as a
first step we concentrated on lead containing organic-inorganic perovskites, trying to find
stable compositions, while the challenging toxicity is our next goal. We started with data
mining, then used 7 ML algorithms to predict band gap energies for generated perovskites
compositions. Based on the gathered data the best model for the prediction was chosen
Linear Regression model. Due to it’s band gap energy FAPbBr1.125I1.875 was chosen as one
of the best compositions for solar cell applications. The ML results were confirmed through
DFT calculations and experiments. Band gap energies were theoretically calculated
through DFT and hybrid methods. Hartree–Fock method is used for hybrid functionals. The
synthesis was implemented thought solvothermal method and the uniform thin films were
obtained trough spin coating.
The authors thank funding of Science Committee RA in frame of scientific project N22rl-012.