Assessing the Performance of Density Functional Theory Methods on the Prediction of Low-Frequency Vibrational Spectra

Abstract

The low-frequency (terahertz) dynamics of condensed phase materials provide valuable insight into numerous bulk phenomena. However, the assignment and interpretation of experimental results require computational methods due to the complex mode types that depend on weak intermolecular forces. Solid-state density functional theory has been used in this regard with great success, yet the selection of specific computational parameters, namely the chosen basis set and density functional, has a profound influence on the accuracy of predicted spectra. In this work, the role of these two parameters is investigated in a series of organic molecular crystals, in order to assess the ability of various methods to reproduce intermolecular forces, and subsequently experimental terahertz spectra. Specifically, naphthalene, oxalic acid, and thymine were chosen based on the varied intermolecular interactions present in each material. The results highlight that unconstrained geometry optimizations can be used as an initial proxy for the accuracy of interatomic forces, with errors in the calculated geometries indicative of subsequent errors in the calculated low-frequency vibrational spectra, providing a powerful metric for the validation of theoretical results. Finally, the origins of the observed shortcomings are analyzed, providing a basic framework for further studies on related materials.