(307g) Automated Reaction Mechanism Generation for Heterogeneous Catalisys

Authors: 
West, R. H., Massachusetts Institute of Technology
Goldsmith, C. F., Brown University
Accurate prediction of the reactivity and selectivity of catalytic materials under industrially relevant conditions requires a detailed list of elementary surface reactions, or a microkinetic mechanism. The development of reliable microkinetic mechanisms is time consuming and error prone. For a given catalytic system the relevant elementary reactions must be determined, the rate coefficients must be estimated accurately yet efficiently, and the resulting mechanism must be made thermodynamically consistent. These problems are compounded because the mechanism grows exponentially with the size of the reactants, and because real-world catalysts have multiple crystalline facets with distinct kinetics.

In this talk, we present recently developed software that can generate microkinetic mechanisms for heterogeneous catalysis automatically. For a given catalyst, the user provides the initial conditions (e.g. temperature, pressure, and gas-phase composition), and the computer automatically determines which reactions are important, obtains parameterizations of the thermodynamic properties and rate coefficients, and solves the governing equations â?? without subsequent human intervention.

The software is based upon the open-source software Reaction Mechanism Generator (RMG) [1,2], which was originally developed for gas-phase combustion and pyrolysis. The new code is referred to as RMG-Cat. By building our software on preexisting code for automatic mechanism generation, we can take advantage of RMGâ??s ability to solve several key problems in computer-generated chemical kinetics. RMG-Cat can automatically:

  • Recognize when two or more species in the mechanism are equivalent

  • Predict all the possible elementary reactions for each species and pair of species

  • Determine which of the possible reactions are actually important

  • Estimate all the necessary thermodynamic and kinetic parameters

  • Ensure that the mechanism is thermodynamically consistent

  • Include flexibility for new reactants on novel materials

  • Accomplish all of the above more quickly than a human

An automatic mechanism generator like RMG-Cat requires four main components to function: a unique representation of chemical species; templates and rules to generate reactions; an algorithm to choose which reactions to include in the final mechanism; and methods to estimate all the thermodynamic and kinetic parameters in the mechanism. We describe these four aspects of the software and an example of its application to syngas production from methane on nickel.

Species representation: RMG-Cat represents adsorbates using chemical graph theory, with atoms as nodes and bonds as edges. Efficient graph theory algorithms for (sub)graph isomorphism are used throughout the code for matching reaction templates, identifying functional groups for parameter estimation, and recognizing when species are chemically equivalent.

Generation of Reactions: Chemical reactions are classified according to â??reaction familiesâ?. Each family contains: a template to identify when the reaction can occur, a recipe to specify the changes to be made to the molecular graphs, and rules to determine the reaction rate coefficient.

Reaction Mechanism Expansion: RMG-Cat considers all possible reactions, but to avoid the exponential explosion in mechanism size and to prevent the inclusion of reactions and intermediates that are kinetically irrelevant, it includes only those reactions that have a sufficiently high rate of reaction at the conditions at which the model will be used. The iterative, rate-based expansion algorithm is described in detail in reference [1].

Parameter Estimation: Although the final chemical kinetic mechanism may be sensitive to only a few pathways, which pathways these are is seldom known a priori. The number of intermediates and reactions to be considered is vast, so the parameter estimation process must be very fast. In all cases, whether seeking an enthalpy of formation, a binding energy, or a rate of reaction, RMG-Cat first checks a database of known values for an exact match. If no match is found in the database, then the properties must be estimated. For most parameters RMG uses estimation rules based on functional groups, arranged in hierarchical tree-structured databases. This allows efficient selection of the most specific rule for which data are available, falling back to broader generalizations when necessary. The rules are derived from the databases of known parameter values.

Test Case:

As a proof-of-concept, we apply RMG-Cat to the partial oxidation of methane on nickel. Three catalytic processes are considered:





steam reforming:

CH4 + H2O

â??

CO + 3 H2

dry reforming:

CH4 + CO2

â??

2 CO + 2 H2

partial oxidation:

CH4 + ½ O2

â??

CO + 2 H2

Finally, we will present future directions for RMG-Cat and how the code will be expanded to create microkinetic mechanisms for a much broader array of reactants and catalysts.

[1] C.W. Gao, J.W. Allen, W.H. Green, R.H. West. Reaction Mechanism Generator: Automatic construction of chemical kinetic mechanisms. Computer Physics Communications, 203, 212-225, (2016) http://dx.doi.org/10.1016/j.cpc.2016.02.013.

[2] RMG - Reaction Mechanism Generator, open-source software, RMG-Py version 1.0.4 http://reactionmechanismgenerator.github.io