Incorporation of Linear Scaling Relations into Automatic Mechanism Generation for Heterogeneous Catalysis

To predict the selectivity and reactivity of novel catalysts at industrially relevant conditions requires a detailed microkinetic mechanism, comprising many elementary reactions. Recent advances such as such as the work of Ulissi et al [5] use a combination of scaling relations, machine learning, and DFT calculations, to gradually refine a microkinetic model until the rate limiting steps have been calculated with sufficient accuracy to be confident that they are correctly identified. However, such a system requires as input a comprehensive kinetic model containing all the possible pathways. Our recently developed Reaction Mechanism Generator for Heterogeneous Catalysis (RMG-Cat) [3], built upon the open-source RMG software primarily used for gas-phase pyrolysis and combustion [1,2], can provide such mechanisms ab inito: the user supplies just the initial conditions (eg. reactant composition, temperature, pressure) and the software predicts all the possible reactions, estimates the thermochemical and kinetic parameters, solves the governing equations, and decides which reaction pathways to include and explore further. RMG-Cat makes its decisions regarding which pathways to explore and which to ignore, using the estimated parameters, so it is important that the estimates are reasonable, even if the important parameters will be refined with more accurate calculations later in the model development process.

Linear Scaling Relations (LSRs) can provide reasonable estimates of adsorption energies in a very computationally efficient manner [4-6]. We have now implemented linear scaling relationships for the estimation of adsorption energies in the RMG-Cat software. Our database of parameters is organized in a hierarchical tree structure, enabling detailed functional group descriptions to be used when data are available and more general descriptions to be used when necessary. We include parameters to describe many adsorbates on a range of metal surfaces, and a framework to re-train the parameters whenever new data are available.

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

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

[3] Goldsmith, C. F., West, R. H. “Automatic Generation of Microkinetic Mechanisms for Heterogeneous Catalysis” J. Phys. Chem. C. http://doi.org/10.1021/acs.jpcc.7b02133

[4] Medford, A. J., Lausche, A. C., Abild-Pedersen, F., Temel, B., Schjødt, N. C., Nørskov, J. K., Studt, F. “Activity and Selectivity Trends in Synthesis Gas Conversion to Higher Alcohols” Topics in Catalysis (2013) 57, 135-142

[5] Ulissi, Z. W., Medford, A. J., Bligaard, T., Nørskov, J. K. “To address surface reaction network complexity using scaling relations machine learning and DFT calculations” Nature Comm. (2017) 8, 14621-14627

[6] Hummelshøj, J. S., Abild-Pedersen, F., Studt, F., Bligaard, T., Nørskov, J. K. “CatApp: A Web Application for Surface Chemistry and Heterogeneous Catalysis” Angewandte Chemie International Edition (2011) 51, 272-274