Benchmarking Ab Initio Computational Methods for the Quantitative Prediction of Sunlight-Driven Pollutant Degradation in Aquatic Environments

Understanding the changes in molecular electronic structure following the absorption of light is a fundamental challenge for the goal of predicting photochemical rates and mechanisms. Proposed here is a systematic benchmarking method to evaluate accuracy of a model to quantitatively predict photo-degradation of small organic molecules in aquatic environments. An overview of underlying com- putational theories relevant to understanding sunlight-driven electronic processes in organic pollutants is presented. To evaluate the optimum size of solvent sphere, molecular Dynamics and Time Dependent Density Functional Theory (MD-TD-DFT) calculations of an aniline molecule in di↵erent numbers of water molecules using CAM-B3LYP functional yielded excited state energy and oscillator strength values, which were compared with data from experimental absorption spectra. For the first time, a statistical method of comparing experimental and theoretical data is proposed. Underlying Gaussian functions of absorption spectra were deconvoluted and integrated to calculate experimental oscillator strengths. A Matlab code written by Soren Eustis was utilized to decluster MD-TD-DFT results. The model with 256 water molecules was decided to give the most accurate results with optimized com- putational cost and accuracy. MD-TD-DFT calculations were then performed on aniline, 3-F-aniline, 4-F-aniline, 3-Cl-aniline, 4-MeOacetophenone, and (1,3)-dimethoxybenzophenone with CAM-B3LYP, PBE0, M06-2X, LCBLYP, and BP86 functionals. BP86 functional was determined to be the best functional. Github repository: https://github.com/eustislab/MD_Scripts