Predictive Kinetics for Gas and Liquid Systems

For most technologically important systems, including combustion, pyrolysis, and atmospheric oxidation of organic compounds, it is very difficult to construct a reliable kinetic model. There are typically hundreds of reaction intermediates, and only a small fraction of the rate parameters are known experimentally. It is usually impossible to measure the concentrations of all the kinetically significant chemical species. Until recently, it was not feasible to numerically solve the large systems of differential equations that describe these systems, even if one did manage to measure or obtain good estimates of all the rate parameters. However, advances in computational chemistry, numerical methods, and computer hardware have now made it possible to construct and solve kinetic models capable of reliably modeling these complicated systems. We are now beginning to extend these techniques to model spatially inhomogeneous reacting flows. Since 2004, our group has been developing the open-source software package Reaction Mechanism Generator (RMG) for automatically generating kinetic mechanisms.

We use state-of-the art quantum chemistry techniques to estimate thermochemistry, barriers and transition state properties for specific reactions; these are then coupled with transition state theory, experimental data, and functional group additivity ideas to provide rate estimates for broad classes of reactions. The computer is used to generate and solve the large reaction scheme implied by these rate estimates. The results are then validated by comparison with experiment, and predictions can be made for situations where no experimental data are available. In one recent case, the computer considered the impact of more than 100,000 reactions before selecting the few hundred kinetically dominant reactions needed to develop a numerically accurate kinetic scheme for an ethane steam-cracker.

Special treatment is required for reactions in solvents, on surfaces, and for fast gas-phase reactions (which do not thermalize). To account for liquid-phase kinetics, we compute thermochemical and kinetic corrections based on the solvents and solutes of interest. The basis for these corrections are quantum-chemical solvation model calculations and extensive databases of experimental data that we have curated. These corrections are then applied on top of the gas-phase rate estimates to predict the rates of reactions in pure solvents. We are currently extending our results to multi-component mixed solvent systems and further developing the implementation of these methods in RMG.

The model predictions are tested against experimental data from the literature, and measurements made in our laboratory or by our collaborators. We have utilized laser techniques to probe free radicals in the gas-phase and in solution, and a variety of analytical techniques to measure product yields in liquid phase and catalytic oxidations.

Uncertainties enter all along the process of constructing and experimentally validating a kinetic model, from the microscopic electronic structure calculations through the macroscopic measurements that are typically made on a complicated product stream. We are quantifying these uncertainties and developing improved methods for the steps, which introduce the largest errors. For example, we are currently benchmarking the use of different density functionals for quantum calculations, and we are examining methods for minimizing the number of chemical species, which must be treated exactly in a simulation.