Compositional engineering of perovskites with machine learning

Abstract

Perovskites are promising materials candidates for optoelectronics, but their commercialization is hindered by toxicity and materials instability. While compositional engineering can mitigate these problems by tuning perovskite properties, the enormous complexity of the perovskite materials space aggravates the search for an optimal optoelectronic material. We conducted compositional space exploration through Monte Carlo (MC) convex hull sampling, which we made tractable with machine learning (ML). The ML model learns from density functional theory calculations of perovskite atomic structures, and can be used for quick predictions of energies, atomic forces, and stresses. We employed it in structural relaxations combined with MC sampling to gain access to low-energy structures and compute the convex hull for CsPb(Br 1−x Cl x ) 3 . The trained ML model achieves an energy prediction accuracy of 0.1 meV per atom. The resulting convex hull exhibits two stable mixing concentrations at 1/6 and 1/3 Cl contents. Our data-driven approach offers a pathway towards studies of more complex perovskites and other alloy materials with quantum mechanical accuracy.

Mixed-Halide Perovskite Alloys CsPb(I1–xBrx)3 and CsPb(Br1–xClx)3: New Insight of Configurational Entropy Effect from First-Principles and Phase Diagrams

“…To this end, fast (and precise) evaluation of the energy of each model system would be needed, such as machine learning. 31 In this work, we study the thermodynamic properties by performing a DFT traversal of 20-atomic models of binary mixed-halide perovskites CsPb(I Br )…”

“…For small systems, these configurations can be accessed in a traverse manner . Otherwise, techniques beyond DFT, such as cluster expansion and machine learning, have been employed in combination with some energy-minimizing algorithms, e.g., Monte Carlo simulated annealing. − When using the minimal alloy formation energy as the criterion of thermodynamic stability, the results of these studies − ,, generally exhibit very shallow (a few meV per perovskite unit) convex hulls on which only a few compositions are located. This corresponds to two consequences: (a) most of the alloy compositions are unstable and will spontaneously decompose into some specific compositions, and (b) it is practically impossible to synthesize perovskite alloys of targeted compositions, especially at finite temperatures; instead, a mixture of many different compositions crossing the whole alloy space coexist in the product.…”

“…x x 1 3 as in ref., 31 naturally due to the smaller computational model in this paper. As illustrated by Figure 4c, at most of the x values of both alloys, the DOS is distributed within a large range of energy, which is well above the minimum.…”

Structural disorder by octahedral tilting in inorganic halide perovskites: New insight with Bayesian optimization

“…BOSS has already been applied to solve problems such as conformer search for organic molecules [54,55] and adsorption of organic molecules at semicon-ductor surfaces [54], to resolve different organic adsorbate types and film growth at metallic surfaces [56][57][58], and to identify the interface geometries of inorganic or organic materials at perovskite surfaces [59][60][61]. The application of BO in materials science is not limited to structure search, e.g., it was used to train the force fields to study phase transition in hybrid perovskites [19], to select optimal ML hyperparameters [62,63], or to efficiently solicit experimental data [64,65].…”

Machine Learning-Based Screening for Potential Singlet Fission Chromophores: The Challenge of Imbalanced Data Sets

“…Nowadays, we live in the data science era and new evidence is collected daily for the effectiveness of machine learning (ML) algorithms in the discovery of new advanced materials. − However, the scientific papers combining ML and SF are intermittent. In 2019, Schröder et al used ML to explore the quantum dynamics in pentacene dimers.…”