(38b) Towards Automated Discovery of Plausible Reaction Paths in Complex Catalytic Systems Using Network Generation and Optimization

The application of state-of-the-art computational tools, viz. density functional theory (DFT) to calculate the energetics of every species and transition state and microkinetic modeling to obtain reaction rates and yields is computationally intractable for complex reaction networks comprising of thousands of reactions and intermediates. In this talk, we present a computationally tractable method to identify dominant reaction pathways for such complex heterogenous catalytic reaction systems. The method: (i) constructs the reaction network in an automated manner using the knowledge of potential elementary steps, (ii) uses simple data-driven models derived from DFT calculations that can then be applied to the rest of the network, (iii) employs a mixed-integer optimization problem to “screen” for the dominant pathways based on energetic feasibility metrics or maximum reaction rate estimates, (iv) rank orders potential alternative solutions of the optimal pathways within the errors of density functional theory and the data-driven model, and (v) generates volcano plots based on known linear scaling correlations for catalyst screening.

The details of the proposed method will be discussed in the context of demonstrative examples comprising of oxygenate decomposition and dehydrogenation of organic hydrogen carriers on transition metal catalysts.