High-throughput computations and machine learning for halide perovskite discovery

Abstract

Halide perovskites are materials of considerable interest for solar cells, photodiodes, LEDs, photocatalysis, and photorechargeable batteries. One of the most attractive features of this class of materials is the sheer tunability of their stability, electronic bandgaps, optical absorption behavior, and defect properties, via composition engineering, phase transformation, change in dimensionality, surface and interface engineering, and octahedral rotation and distortion. Due to the ease of simulating well-defined crystal structures and systematically investigating compositional and structural factors that affect their properties, first-principlesbased density functional theory (DFT) computations are frequently used for studying halide perovskites, leading to high-throughput data sets, screening of promising materials, and training of machine learning (ML) models for accelerated prediction and optimization. In this article, we present an overview of computational data-driven discovery of novel halide perovskites using some examples from the literature we believe best represent success in this field. Specific methods used for prediction of properties, optimization and screening across large chemical spaces, and automated design of novel structures and compositions, are highlighted. DFT-ML-based design frameworks have been instrumental in expanding the pool of stable halide perovskites with desired optoelectronic properties and will continue to inform new discovery in close synergy with targeted experiments.

A high-throughput computational dataset of halide perovskite alloys

“…These correlations help expand our design principles based purely on the fractions of different species and provide an opportunity to train predictive models for various properties. 16,22 3.4 Improving property predictions using HSE06 and spinorbit coupling It was shown in Section 3.1 that for a set of selected HaP compositions, while PBE-optimized lattice constants match well with experiments, PBE band gaps are underestimated, and HSE-PBE-SOC band gaps match better with measured values. GGA-PBE computations are generally reliable for the structure and stability (formation or decomposition energy) of both hybrid and purely inorganic HaPs, but advanced levels of theory such as the HSE06 functional or GW approximation, with the inclusion of SOC to account for the relativistic effects of heavy atoms such as Pb, are of paramount importance when it comes to electronic and optical properties.…”

“…We recently published a thorough overview of many such efforts applying DFT and/or ML towards halide perovskite discovery. 22 In this work, we report a large HT-DFT dataset of 495 chemically distinct, pseudo-cubic, halide perovskite alloys. This dataset builds upon the 229 compounds reported in prior work by Mannodi-Kanakkithodi and Chan, 16 adding more types of mixing, better property estimates, and detailed analysis of trends and correlations.…”

Machine learning in energy chemistry: introduction, challenges and perspectives

“…Searching for the stability and high PCE lead-free halide perovskites 141 is also an important downstream task of perovskite machine learning application. Using the property density distribution function (PDDF), Stanley et al constructed the features and apply in the predicted bandgap, formation energy, and convex hull distance of lead-free halide perovskite 142 .…”

A framework for materials informatics education through workshops

“…As such, we opted for a project involving the discovery of novel halide perovskite alloys for optoelectronic applications, using ML models trained upon a substantial computational data set of relevant properties. 14,22 A Jupyter notebook laying out this entire project can be found on GitHub 23 as well as on nanoHUB; 24 this includes all code necessary for reading and visualizing the data (properties + different types of descriptors), performing clustering and examining correlations, training different types of regression models, and coupling the best predictive models with genetic algorithm to discover new perovskite compositions that show a targeted set of properties. In addition, several boxes of descriptive text and images detail all the technical content, motivation, methodology, and primary results/observations of the project.…”