Hirshfeld Analysis and Perovskite Stability

Perovskites are a promising, low cost and highly efficient resource in photovoltaics. Since the first perovskite solar cells were developed, in 2009, the technology has been rapidly improving, reaching power conversion efficiency above 20% in 2016. Some of the key attractive qualities that HOIPs present are their long carrier lifetime, long diffusion length and crystalline defect tolerance, broad absorption range and high carrier mobility and light absorption capability. However, fast degradation and lead dependency remain as challenges for the development of HOIP photovoltaic devices and their commercialization.

This research project started as one of my doctoral lines of research, with the initial motivation of accelerating the discovery of stable, non-toxic hybrid organic-inorganic perovskites (HOIP) by exploring structural and qualitative properties of HOIP classes through Hirshfeld surface analysis and fingerprint analysis tools, simultaneously correlating the material features to crystal structure and bonding characteristics.

T. Stona de Almeida. Doctoral thesis. ProQuest (2021)

Hirshfeld Surface image 1 Abstract: A computational quantum chemistry framework is coupled with traditional approaches, presenting a new protocol to assess perovskite stability both visually and statistically. A ranking of these metrics is defined and contextualized. The thesis provides new contributions to the understanding of the chemical bonding interpretation to the role of the tolerance/octahedral factor in perovskite systems. Results also include a database of circa 20k perovskites, containing ionic radii and tolerance factors, plus chemical features - eg: bandgap, volume, thermodynamic properties; the first large scale Hirshfeld surface quantum library for perovskite structures, consisting of circa 1500 perovskite compounds; a protocol for unsupervised classification of Hirshfeld statistics and their employment as a predictive feature in perovskite design; a dimensionality reduction analysis of the large database, and other insights.

Keywords: Machine learning; Hirshfeld Surfaces; QSPR; Perovskites; Materials Informatics.

T. Stona de Almeida. Abstract, ISBN: 978-625-00-8356-7

Hirshfeld Surface image 2 Abstract: The Hirshfeld surface is, at first sight, an intuitive object: the geometric locus that equally splits the electronic density contribution of a subset of atoms - the promolecule - within a larger molecule or crystal cell - the procrystal. The result is a closed surface that is not necessarily convex but still easy to imagine. Or is it? The application of the Hirshfeld surface is often qualitative and visual, being used as a supporting argument or to gather intuitive clues to guide a more analytical approach in chemical characterization. For this, the surface is computationally estimated and plotted in 3D, where a color-coded scalar property can be added to represent a 4th dimension e.g. curvature. Finally, the Hirshfeld surface fingerprint is calculated, a color-coded 2-dimensional bivariate histogram of internal and external distances between the surface and the internal and external atomic nuclei relative to the promolecule. Questions arise regarding the numerical method employed to compute the surface mesh, which often contains multiple vertices where only one should be present; the accuracy of the fingerprint choice of bin size and its consequences; the loss of information that could be preserved by other methods. Are there other more appropriate scalar properties of the surface to be explored? This work is an exercise of critical thinking about the Hirshfeld surface method, a dissection of its geometry, its interpretation, and applications.

Keywords: Hirshfeld surface analysis; molecular geometry; mathematical chemistry.


T. Stona de Almeida. Cambridge Open Engage (2023). doi.org/10.33774/coe-2023-7nqjm (Preferred source)
T. Stona de Almeida. Zenodo (2023) doi.org/10.5281/zenodo.10112561

Link to the talk at Ronin Institute's channel on YouTube: Ronin Institute Lightning Talks, October 2023

Abstract of talk: The Hirshfeld surface was introduced by Spackman and Byrom in 1997, based on the Hirshfeld partitioning scheme for electronic contribution in molecular crystals. The Hirshfeld surface analysis and its fingerprint have become since then popular tools for visualization of intermolecular interactions and crystal packing. In this talk, I would like to go over some aspects of the Hirshfeld surface protocol and raise some questions about the method.

Example of Hirshfeld surface for Ethylammonium Germanium Fluoride: