(435h) Automatic Discovery of Reaction Paths and Ab Initio/Rrkm-ME Calculations Towards Pyrolysis and Combustion Mechanisms | AIChE

(435h) Automatic Discovery of Reaction Paths and Ab Initio/Rrkm-ME Calculations Towards Pyrolysis and Combustion Mechanisms

Authors 

Grambow, C., Massachusetts Institute of Technology
Green, W., Massachusetts Institute of Technology
Automatic kinetic model generators are instrumental in describing systems with many chemical species and elementary reaction steps, such as combustion, atmospheric, environmental, and astrochemical processes. Modeling software, such as Reaction Mechanism Generator (RMG), use a rate-based algorithm to keep the number of species and reactions tractable by excluding unimportant reactions based on flux, and growing the model based on known or estimated parameters for important reactions. The thermochemical and kinetic parameters can be estimated from group additivity and by using a database of known rate rules for these reaction types, respectively, allowing the inclusion of species and reaction steps as needed. Thus, unknown reaction types are omitted in the models. The discovery of unknown reaction types had to occur with human intuition of a saddle-point structure connecting reactant(s) to product(s), where some insight to atom rearrangement had to be implemented as an initial guess. Then, if successfully located, rate coefficients could be calculated and included into models to improve their reliability and predictions. Here, we introduce a strategy for the automatic discovery of unknown reaction steps (and types) without a priori knowledge by employing bond electron (BE) matrices combined with double-ended chain-of-state methods. This strategy allows an automatic sampling of the chemical space in converting reactant(s) to product(s), and arriving at a reasonable saddle-point structure without the need of human intuition. In BE matrices, products are generated from bond breaking and bond forming of the atoms from a given reactant. Then, a double-ended chain-of-state method attempts to connect the reactant structure to product structure by atom rearrangements passing through the saddle-point structure. This estimated saddle-point is then refined using a quasi-Newton optimization technique, followed by intrinsic reaction coordinate (IRC) calculations to confirm that the saddle-point structure indeed connects reactants to products.

The automatic discovery of reaction steps allows an unbiased exploration of the potential energy surface (PES) where novel saddle-points connecting to reactants and products can occur through unconsidered atom rearrangements or possibly through an unknown reaction type altogether. Several applications of this method to discover unexpected reaction pathways of species important in combustion and pyrolysis chemistry are reported.