Simulating complex gas-phase chemistry is a critical part of understanding and predicting the behavior of coupled chemical and physical systems, such as astrochemistry and atmospheric chemistry, combustion, and process safety. A key part of simulations are robust chemical kinetic models that accurately describe the chemical mechanisms at temperatures and pressures relevant to the system of interest. Developing these mechanisms has traditionally been a challenging and labor-intensive task.

Argonne researchers developed Frhodo to alleviate repetitive tasks—allowing optimized, complex mechanisms to be developed more efficiently. To this end, machine learning algorithms are used to optimize the mechanism under study within the constraints of researcher-specified uncertainty values. In Frhodo’s highly interactive GUI not only can the progress of the optimization be tracked, but also any calculated value (rate of reaction, enthalpy of reaction, heat release rate, etc.) can be viewed. This allows the researcher to rapidly iterate through potential solutions to find the mechanism that best fits the target data. In addition to automated optimization, the researcher can also manually adjust kinetic parameters, allowing for traditional methods of mechanism development and fine tuning of an optimized mechanism from the automated routines. These capabilities have made it considerably faster to use Frhodo to optimize chemical kinetics mechanisms while also making it possible to optimize more variables than can be done by hand.