(174s) How to Model Local Solvation Environments in Pure Liquids and Complicated Mixtures without Dynamics

Understanding solvation environments at a molecular level is often necessary in many areas of chemistry, biology, and engineering. We first report a practical means of modeling the local solvation of ions and small molecules in water. We use a global optimization algorithm to identify low energy intermediates with different numbers of water molecules. Next, we use the Smooth Overlap of Atomic Positions (SOAP) kernel and sketchmap analyses as unsupervised machine learning algorithms to identify inputs that are closely related to each other and reduce the dimensionality of the feature vectors that represent the inputs. After studying the geometries of different solvent environments, we use Quasi Chemical Theory (QCT) and thermodynamic cycles to compute the solvation energies. Our analysis using different sizes of microsolvated clusters is found to result in different solvation energy scales. Small clusters are found to model the TATB solvation scale, however larger clusters are found to model experimental data using the Cluster Pair based (CPB) approximation. After understanding the importance of local solvent arrangements, we move to model complex solvation environments like water-methanol mixtures. We use water-methanol mixtures with different proportions to model solvation energies following a similar procedure. We also try to model CO2 diffusion in different molten salts in light of understanding sustainable CO2 chemistry. This talk will show the applicability and limitations of this procedure to different solvated systems and thus show a practical procedure for modeling reaction mechanisms in complicated solvation environments without computationally laborious Born-Oppenheimer molecular dynamics simulations.