Crystal structures predictions from first principle for rigid and flexible molecules
Date
2025
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Crystal structure predictions (CSPs) remains one of the most challenging problems in the physical sciences. Identifying the correct polymorph–a distinct crystalline form of a molecule that can exhibit different physical and chemical properties–is critical, as it can prevent substantial financial losses. This underscores the value of CSP methods in guiding experimental efforts. CSP generates a polymorph energy landscape comprising thousands to millions of potential polymorphs. Identifying the experimentally observed crystal within this landscape is computationally expensive when using modern electronic structure methods directly. To mitigate this, initial screening is performed using empirical force fields (FFs), which provide a computationally efficient means of filtering candidates. However, these empirical FFs are often unreliable in terms of accuracy and do not work well for novel molecules that were not included in their parameterization, limiting their applicability. To address this limitation, tailormade FFs can be developed for molecules of interest by fitting to results from electronic structure calculations. This work focuses on establishing robust protocols for accurate CSP using FFs for both rigid and flexible molecules. For rigid-monomer CSP, our protocol successfully identified the experimentally observed polymorph within the top 20 ranked structures. Further refinement of the ranking was achieved by performing periodic boundary condition electronic structure calculations on these top polymorphs. To avoid this latter step, a protocol was developed to improve FF performance for rigid monomers by incorporating dimers from high-ranking polymorphs generated by the initial FF. The effectiveness of this approach was demonstrated in the seventh blind test (7BT), where several experimentally known polymorphs were blindly predicted with high accuracy in the structure ranking stage. The 7BT also highlighted the limitations of CSP protocols for flexible monomers. To overcome these challenges, a novel CSP protocol was developed for flexible molecules, successfully identifying the experimental polymorph for a test monomer with six torsional degrees of freedom at rank 2. These highly accurate, tailor-made FFs were also employed to explore strategies for enhancing the properties of hydrates by replacing water with hydrogen peroxide.
Description
Keywords
Crystal structure predictions, Density functional theory, Flexible crystal structure predictions, Symmetry adapted perturbation theory, Tailor fitted force field