Marshaling the Resources of First Principles Theory and High Performance Computing to Predict the Chemistry of Combustion

A quantitative description of the conversion of a fuel into products and pollutants requires a chemical model with ~10³ species and ~10⁴ reactions. Notably, many low flux/rare event reactions are of central importance to this chemistry. For example, ignition behavior is largely determined by the sequential addition of two O₂ molecules to a radical. Similarly, soot formation involves growth pathways that are secondary to the oxidative conversions, while NOx formation generally involves high barriers. The key intermediates in these reaction sequences generally have peak mole fractions of 10^-4 or less.

Highly descriptive chemical models facilitate explorations of the response of pollutant concentrations and/or ignition properties to changing engine design parameters, or to fuel additives. Such models are most useful when they accurately reproduce the true conversion routes. Recent advances in theoretical kinetics methods allow for the prediction of the underlying thermochemical kinetic parameters with accuracies that rival those of experimental determinations. At that level of accuracy, the overall chemical models should yield meaningful system response analyses.

We will describe our recent efforts at developing a suite of codes (https://github.com/Auto-Mech) that couples state-of-the art kinetic theory with high performance computing in order to provide high fidelity first-principles based combustion mechanisms. The focus of the effort involves the development of effective procedures for automatically predicting the kinetic properties of large and arbitrary sets of chemical reactions. It involves a combination of chemical physics method and software developments. We will illustrate our progress through a review of recent calculations of more than 2000 rate constants for the combustion of a dual-fuel mixture of isooctane and n-dodecane.