Understanding the Liquid Phase with Molecular Dynamics Simulations, Vibrational Spectroscopy, and Uncertainty Quantification

Abstract

Many chemical processes that significantly affect human life, such as energy storage, catalysis, and biochemistry, involve the liquid phase. Consequently, understanding liquids at the molecular level has been a major focus of scientific research for decades. With the advent of modern algorithms and hardware, the field of <em>in silico</em> chemistry has matured, and computational tools have become indispensable components of a chemist's methodological arsenal. Among these tools, molecular dynamics simulations are particularly well suited for investigating the liquid state, as they enable the dynamic modeling of large liquid systems over long timescales at affordable computational cost. Various structural and dynamic quantities can be derived from such simulations, providing detailed insight at atomistic resolution and linking microscopic properties to macroscopic observables. As the accurate computation of these properties often requires sophisticated numerical and statistical procedures, the molecular dynamics simulation community has recently begun adopting methods from uncertainty quantification to improve the accuracy and reliability of computational predictions. <br/> This dissertation combines state-of-the-art molecular dynamics simulations with vibrational spectroscopy and uncertainty quantification to advance the understanding of the liquid phase. It comprises eight published and peer-reviewed works that address different aspects of liquid-phase systems and together span the overarching topic introduced above. The development and application of free and open-source software tools for the analysis of molecular dynamics simulations are central themes of this dissertation, which is divided into two topical parts. <br/> The first part focuses on the structure of ionic liquids. By introducing a new cluster analysis tool into the TRAVIS software, the morphology of an ionic liquid family is unraveled. Then, a polarizable force field for an ionic liquid/graphite interface is developed. Finally, for the first time, the induction of vibrational optical activity in an ionic liquid anion by a chiral probe molecule is demonstrated through vibrational circular dichroism spectroscopy and traced back to a distortion of its conformational equilibrium. <br/> The second part addresses uncertainty quantification, primarily in the context of dynamic properties. Two works provide tutorials and best-practice workflows for uncertainty-aware modeling strategies and introduce the MSDIFF software for computing diffusion coefficients and ionic conductivities. In addition, a multifidelity Monte Carlo framework for the simulation of diffusion coefficients is presented using liquid water as an example, achieving significant computational speedups compared to traditional approaches. The dissertation concludes with a comprehensive review of the emergence of uncertainty quantification in the field of <em>in silico</em> chemistry.