Computational Methods for the Chemical Space Exploration of Metal Halide Perovskite Interfaces

Abstract

Additive engineering—the use of small molecules to simultaneously address the inherent chemical instability of metal halide perovskites and enhance the optoelectronic properties of these materials—has been demonstrated to be an effective strategy to accelerate the commercialization of solution-processable perovskite devices. However, the additive selection process for perovskite incorporation is typically based on chemical intuition, leaving a limited understanding of how the chemical, electronic, and geometric properties of additive molecules influence interactions with perovskite interfaces. In this dissertation, I focus on the usage and development of computational physics tools to systematically investigate the impact of perovskite interface functionalization with small molecules on fundamental physical behavior, facilitating the construction of structure-property relationships based on simple molecular features. In Chapter 1, I establish the foundation for discussing surface functionalization in metal halide perovskites, systematically exploring how chemical space is navigated in the search for effective surface ligands. Chapter 2 provides an in-depth overview of the methodologies employed throughout this work, including density functional theory, cheminformatics, and workflow automation. In Chapter 3, I examine the chemical and electronic effects of functionalizing methylammonium lead iodide (\ce{MAPbI3}) perovskite with 5-aminovaleric acid (5-AVA) and two metal halide salts: 5-aminovaleric acid iodide (5-AVAI) and 5-aminovaleric acid chloride (5-AVACl). The film quality, as assessed by X-ray diffraction (XRD) and photoluminescence (PL) spectrophotometry, improves with the inclusion of all additives. Scanning electron microscopy (SEM) and atomic force microscopy (AFM) reveal an increase in grain size and a decrease in film roughness with the incorporation of 5-AVAI and 5-AVACl. DFT calculations, along with X-ray photoelectron spectroscopy, reveal that 5-AVAI and 5-AVACl exhibit stronger surface interactions than 5-AVA with both the \ce{PbI2}-rich and \ce{MAI}-rich surfaces of \ce{MAPbI3}, primarily due to hydrogen bonding, revealing the mechanism behind the improvement in film quality. Furthermore, I reveal that 5-AVACl can spontaneously decompose into 5-AVA and HCl, leading to similarities in film properties between 5-AVA and 5-AVACl after annealing. In chapter 4, I employ a selection of twenty-five heterocyclic molecules to model the effects of heteroatomic species (N, O, S, Se, and P) and heteroatom electron delocalization on the chemical and electronic properties of \ce{MAPbI3} perovskite using DFT calculations. I demonstrate that the interaction strength of each heterocycle increases as the heteroatom electron delocalization (i.e., degree of unsaturation) decreases. We observe that adsorption energies are strongest for N-donors and weakest for O-donors, with P-, S-, and Se-donors exhibiting intermediate adsorption energies. The electronegativity of the heteroatom plays a crucial role in governing surface charge transfer from the adsorbate to the perovskite surface. Higher electronegativity is correlated with a reduced extent of Pb reduction—or potential oxidation in the case of O-donors—highlighting its impact on interfacial electronic interactions. Heteroatom electronegativity also serves as a predictor for surface band gap shifts, with more electronegative donors leading to increased surface band gaps. Conversely, the adsorption of low-electronegativity P-donors generally results in surface band gap reductions. In chapter 5, I introduce an advanced framework for probing the surface functionalization of any perovskite composition using DFT. By combining several open-source Python codes (PyMatGen, ASE, Fireworks), I demonstrate that the analysis performed in chapters 3 and 4 can be automated, simultaneously improving the repeatability and physicality of the DFT predictions. This automated workflow removes the inherent bias of the operator by identifying adsorption sites on any surface using Delaunay triangulation and stochastically places small molecules on these sites using Open Catalyst Project (OCP) subroutines from Meta FAIR Chemistry. I demonstrate the capabilities of this workflow by constructing a subset of 200 small molecules from the ZINC 20 database with fewer than ten heavy atoms, selected to maximize chemical diversity through Tanimoto similarity. The ground-state geometries of these molecules on the \ce{PbI2}-rich surface of formamidinium lead iodide (\ce{FAPbI3}) perovskite were identified using the high-throughput workflow. In this work, I demonstrate that cheminformatics descriptors enable the construction of structure–property relationships linking molecular structure to DFT-calculated adsorption energies and surface charge transfer. Notably, we found that the Wildman–Crippen molar refractivity of the selected additive molecules exhibited a strong correlation with adsorption energy, while the extended topological atom (ETA) index was highly correlated with interfacial charge transfer. These findings suggest the potential of cheminformatics descriptors as predictive metrics for assessing the doping ability of molecular additives in MHPs.