(702a) Recent Advances in Predictive Kinetics

Many important reaction systems involve quite complicated reaction networks. In the past the kinetics were usually modeled by empirical fits (e.g. to power law models). But now it is becoming practical to compute thermochemistry and rates for thousands of reaction intermediates using quantum chemistry, and these large data sets can be used to train fast models for predicting a range of molecular and reaction properties. This suggests one could build kinetic models a priori, before doing experiments, and also use model structures with much higher fidelity to the true underlying chemistry. However, building, checking, and using these complicated models introduces a number of new challenges. In this talk we report some of our recent work on overcoming these challenges, including automated workflows for computing sensitive Keq’s and k(T)’s using quantum chemistry, methods for coupling reaction and transport, methods for handling macromolecules, training models using a mix of quantum and experimental data, and methods for estimating solvent effects.