Abstract
<p>Ionic Liquids (ILs) have emerged as environmentally friendly organic solvents in recent years with many unparalleled advanced chemical and physical properties compared with conventional solvents. Deep eutectic solvents (DESs) are a class of ILs also recognized as alternative solvents capable of reducing wastes and pollutions. As innovative fluids, ILs are typically combined using a large asymmetric organic cation and small inorganic anion. Given their potential significant number of varieties, they have been further categorized into different types. Among them, the imidazolium-based ILs have emerged as one of the most popular IL classes. Despite the many advantages of ILs, a fundamental understanding of the origin of their enhanced performance is still lacking. Molecular dynamics (MD) and Monte Carlo (MC) simulations are great tools for elucidating the chemical understanding of ILs; however, the accuracy of MD and MC simulations is highly dependent on a set of empirical parameters that reproduce chemical and physical solvent properties, i.e., force fields (FFs). Herein this thesis, three sets of nonpolarizable FFs, e.g., OPLS-2009IL, OPLS-VSIL, and OPLS-DES, were developed and validated.<br />
Machine learning (ML) tools, i.e., artificial neural networks (ANNs), are currently under intense development to accurately and efficiently reproduce computationally expensive quantum mechanical (QM) calculations. Eventual implementation of ML, such as ANNs, into MD simulations could eliminate the need of FFs, as well as increase the level of accuracy that is comparable to high level QM methods. In this thesis, ML is briefly discussed on its methods and potential issues in the last chapter, as well as its usage for FF development. Future applications are also discussed.</p>