(443h) Generalized Brønsted Evans Polanyi Relationships for Metal Surfaces from Machine Learning | AIChE

(443h) Generalized Brønsted Evans Polanyi Relationships for Metal Surfaces from Machine Learning


Göltl, F. - Presenter, University of Wisconsin-Madison
Mavrikakis, M., University of Wisconsin - Madison
In state of the art catalyst design approaches the largest cost is associated with the parameterization of Brønsted-Evans-Polanyi (BEP) relationships [1], a method to correlate kinetic parameters with thermodynamic ones, at the elementary step level. Therefore, the development of a generalized formulation of BEP correlations, i.e. BEPs that can accurately describe a wide array of reactions, would be highly desirable and would accelerate catalyst design.

Here we develop generalized BEPs for a diverse set of reactions on metal surfaces. We focus on 788 reactions from the CatApp database [2], a database that contains reaction- and activation energies for reactions on monometallic surfaces calculated using the DACAPO code and using the RPBE density functional. Our analysis builds on typical BEPs, which only use the reaction energy as a descriptor. However, to improve accuracy we introduce a set of descriptors for the type of reaction encountered derived from scaling relationships [3] and add an additional descriptor for the catalyst surface. In a second step we use three different machine-learning approaches for inter- and extrapolation. We find a significant increase in accuracy for the prediction of activation energies and our best approach leads to errors that are comparable to errors typically associated with DFT calculations. In further analysis, we find that MAE values are similar for different types of reactions and we study the impact of training set size on the performance of the approach. Finally we confirm that the reaction energy is the most important descriptor, but all the other introduced parameters collectively contribute significantly to the improved description.

[1] J.K. Nørskov, T. Bligaard, A. Logadotir, S. Bahn, L.B. Hansen, M. Bollinger, H. Bengaard, B. Hammer, Z. Sljivancanin, M. Mavrikakis, Y. Xu, S. Dahl, C.J.H. Jacobsen, J. Catal. 209, 275-278 (2002)

[2] J.S. Hummelshøj, F. Abild-Pedersen, F. Studt, T. Bligaard, J.K. Nørskov, Angew. Chemie - Int. Ed. 51, 272–274 (2012).

[3] F. Abild-Pedersen, J. Greeley, F. Studt, J. Rossmeisl, T.R. Munter, P.G. Moses, E. Skulason, T. Bligaard, J.K. Norskov, Phys. Rev. Lett. 99, 016105, 2007