Understanding chemical processes with entropic sampling

Kinetic models are widely used in simulating the relationship between the input space and the outcome space of a chemical process. Ignoring the computational cost, complete profiling, i.e., performing simulation at all grid points in the input space, would be the best way to understand the model, because it provides us a complete picture of inter-variable relationships. Optimization methods that sample favorable input points can only provide narrower views. In this paper, we employ entropic sampling, a statistical physics method, to approximate complete profiling. It is cost effective and provides a holistic picture of the model, where one can perform post-hoc exploratory analyses across any regions of the outcome space. Using a kinetic model of the nucleophilic aromatic substitution reaction, we analyze how the failure rate is related to process parameters and elucidate different ways to achieve low failure rates.