Proceedings of Internet Conference on Theory and Computation of Halide Perovskites (ComPer)
Publication date: 4th September 2020
Hybrid halide perovskites (HHPs) have shown very promising results for the next generation of solar cells, but the discovery of new HHPs, possibly Pb-free, is required to meet the increasing energy demand, safety standards, and durability. In this work, we combine quantum mechanical calculations, in the framework of Density Functional Theory, and machine learning methods, to estimate and predict the properties of novel HHPs. Starting from an initial dataset of 240 HHPs, we have identified useful trends to design HHPs with the desired band gap: the band gap increases with an increase of the electronegativities of the constituent species, while it reduces with an increase of the lattice constants of the system.[1] While structure-property relationships can be accurately described using DFT, these calculations are still computationally demanding, limiting their use in screening a large set of candidate structures. We thus use machine learning methods, in the framework of convolutional neural networks, applied to a dataset of around 850 HHPs, to develop a predictive model for the electronic properties of HHPs.[2] In this architecture, each neural network element has a designated role in the estimation process from predicting complex features of the perovskites such as lattice constant and octahedral till angle to narrowing down possible ranges for the values of interest, obtaining a predictive model which has very good accuracy.