(134e) Moving from Postdictive to Predictive Kinetics in Reaction Engineering

Accurate quantitative chemical kinetic models are useful in many applications, ranging from design of chemical processes to building a consensus for international treaties. If the feed is simple and the chemistry is very specific, it can be accurately represented using a simple rate law, and just a few laboratory experiments might be sufficient to determine the small number of rate parameters. For these ‘textbook systems’ the reaction engineering community knows exactly how to proceed: first do the experiments, so we know what chemistry occurs and have a rate law, then ‘predict’ how that same chemistry will occur in a larger scale reactor. Perhaps ‘postdict’ is more accurate since typically the model calculations are completed after all the experiments are completed.

However, in other cases the feed is complicated and the chemistry is unselective, so that there are a large number of reactions occurring simultaneously in the system, including secondary reactions of the primary products. In cases like this we are often unsure which molecules and reactive intermediates are present, and we might not know which reactions are occurring. Even if we could build a kinetic model it would have too many parameters for us to determine them experimentally. In this situation we need a way to actually predict the chemistry.

Here we explain how reaction chemistry can be predicted, by combining calculations from first principles with generalizations derived from different systems. Predictive methods depend on (a) knowing which reactions are possible, (b) reliable fast estimates of rate parameters, and (c) very accurate computations of the rate-controlling parameters (which can be k’s or Keq’s). Failures in any of these steps can lead to serious discrepancies between model and experiment. Challenges in extending the range of systems that can be accurately predicted will be outlined, with suggestions for a path forward.