(337d) Accurately Predicting Products and Rates of Complex Reacting Systems

The complexities of chemical reactions, particularly in systems where many reactions are occurring simultaneously, have long challenged chemical engineers. If we could predict how the product distribution and conversion depends on the feed composition and reaction conditions, we could rationally design and optimize the system on the computer. The first step is to identify all the important species and reactions; this can be done using automatic mechanism generation software such as Genesys, NetGen, or RMG. Unfortunately, we seldom have enough experimental data to determine the rate parameters for all the important reactions. During the last two decades, it has become practical to use quantum chemistry to supplement the experimental data for many systems, sometimes allowing predictions even before any experiments are done. Recently machine learning has made it practical to make more reliable predictions based on much larger experimental and quantum chemistry datasets. In this talk I summarize the current status of predictive chemistry, some recent successes, and some of the remaining challenges.