Advertisement

Predicting Chemical Reaction Barriers with a Machine Learning Model

  • Aayush R. Singh
  • Brian A. Rohr
  • Joseph A. Gauthier
  • Jens K. NørskovEmail author
Article
  • 1 Downloads

Abstract

In the past few decades, tremendous advances have been made in the understanding of catalysis at solid surfaces. Despite this, most discoveries of materials for improved catalytic performance are made by a slow trial and error process in an experimental laboratory. Computational simulations have begun to provide a way to rationally design materials for optimizing catalytic performance, but due to the high computational expense of calculating transition state energies, simulations cannot adequately screen the phase space of materials. In this work, we attempt to mitigate this expense by using a machine learning approach to predict the most expensive and most important parameter in a catalyst’s affinity for a reaction: the reaction barrier. Previous methods which used the step reaction energy as the only parameter in a linear regression had a mean absolute error (MAE) on the order of 0.4 eV, too high to be used predictively. In our work, we achieve a MAE of about 0.22 eV, a marked improvement towards the goal of computational prediction of catalytic activity.

Graphical Abstract

Keywords

Density functional theory Machine learning Kinetic modeling 

Notes

Acknowledgements

Support from the U.S. Department of Energy, Office of Basic Energy Science, Chemical Sciences, Geosciences, and Biosciences Division, to the SUNCAT Center for Interface Science and Catalysis is gratefully acknowledged. B.A.R. acknowledges fellowship support from the National Science Foundation Graduate Research Fellowship (Grant No. DGE-114747). This work was supported by a research grant (9455) from VILLUM FONDEN.

Compliance with Ethical Standards

Conflict of interest

The authors have no conflict of interest to declare.

Supplementary material

10562_2019_2705_MOESM1_ESM.csv (11 kb)
Supplementary material 1 (CSV 11 KB)

References

  1. 1.
    Henkelman G, Jónsson H (1999) A dimer method for finding saddle points on high dimensional potential surfaces using only first derivatives. J Chem Phys 111:7010–7022.  https://doi.org/10.1063/1.480097 CrossRefGoogle Scholar
  2. 2.
    Henkelman G, Uberuaga BP, Jónsson H (2000) Climbing image nudged elastic band method for finding saddle points and minimum energy paths. J Chem Phys 113:9901–9904.  https://doi.org/10.1063/1.1329672 CrossRefGoogle Scholar
  3. 3.
    Henkelman G, Jónsson H (2000) Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J Chem Phys 113:9978–9985.  https://doi.org/10.1063/1.1323224 CrossRefGoogle Scholar
  4. 4.
    Samuelson P, Thorp EO, Kassouf ST (1968) Beat the Market: A Scientific Stock Market System. J Am Stat Assoc 63:1049.  https://doi.org/10.1039/TF9363201333 Google Scholar
  5. 5.
    Wang S, Petzold V, Tripkovic V et al (2011) Universal transition state scaling relations for (de)hydrogenation over transition metals. Phys Chem Chem Phys 13:20760.  https://doi.org/10.1039/c1cp20547a CrossRefGoogle Scholar
  6. 6.
    Hammer B, Nørskov JK (2000) Theoretical surface science and catalysis—calculations and concepts. Adv Catal 45:71–129.  https://doi.org/10.1016/S0360-0564(02)45013-4 Google Scholar
  7. 7.
    Andersson MP, Bligaard T, Kustov A et al (2006) Toward computational screening in heterogeneous catalysis: Pareto-optimal methanation catalysts. J Catal 239:501–506.  https://doi.org/10.1016/j.jcat.2006.02.016 CrossRefGoogle Scholar
  8. 8.
    Peterson AA, Christensen R, Khorshidi A (2017) Addressing uncertainty in atomistic machine learning. Phys Chem Chem Phys 19:10978–10985.  https://doi.org/10.1039/c7cp00375g CrossRefGoogle Scholar
  9. 9.
    Khorshidi A, Peterson AA (2016) Amp: A modular approach to machine learning in atomistic simulations. Comput Phys Commun 207:310–324.  https://doi.org/10.1016/j.cpc.2016.05.010 CrossRefGoogle Scholar
  10. 10.
    Jørgensen PB, Jacobsen KW, Schmidt MN (2018) Neural message passing with edge updates for predicting properties of molecules and materialsGoogle Scholar
  11. 11.
    Jørgensen PB, Mesta M, Shil S et al (2018) Machine learning-based screening of complex molecules for polymer solar cells. J Chem Phys doi.  https://doi.org/10.1063/1.5023563 Google Scholar
  12. 12.
    Ulissi ZW, Singh AR, Tsai C, Nørskov JK (2016) Automated discovery and construction of surface phase diagrams using machine learning. J Phys Chem Lett 7:3931–3935.  https://doi.org/10.1021/acs.jpclett.6b01254 CrossRefGoogle Scholar
  13. 13.
    Brockherde F, Vogt L, Li L et al (2016) By-passing the Kohn–Sham equations with machine learning.  https://doi.org/10.1038/s41467-017-00839-3
  14. 14.
    Hansen K, Montavon G, Biegler F et al (2013) Assessment and validation of machine learning methods for predicting molecular atomization energies. J Chem Theory Comput 9:3404–3419.  https://doi.org/10.1021/ct400195d CrossRefGoogle Scholar
  15. 15.
    Schneider E, Dai L, Topper RQ et al (2017) Stochastic neural network approach for learning high-dimensional free energy surfaces. Phys Rev Lett 119:150601.  https://doi.org/10.1103/PhysRevLett.119.150601 CrossRefGoogle Scholar
  16. 16.
    Ulissi ZW, Tang MT, Xiao J et al (2017) Machine-learning methods enable exhaustive searches for active bimetallic facets and reveal active site motifs for CO2 reduction. ACS Catal 7: 6600–6608.  https://doi.org/10.1021/acscatal.7b01648 CrossRefGoogle Scholar
  17. 17.
    Ulissi ZW, Medford AJ, Bligaard T et al (2017) To address surface reaction network complexity using scaling relations machine learning and DFT calculations. Nat Commun 8:14621.  https://doi.org/10.1038/ncomms14621 CrossRefGoogle Scholar
  18. 18.
    Ma X, Li Z, Achenie LEK, Xin H (2015) Machine-learning-augmented chemisorption model for CO2 electroreduction catalyst screening. J Phys Chem Lett 6:3528–3533.  https://doi.org/10.1021/acs.jpclett.5b01660 CrossRefGoogle Scholar
  19. 19.
    Vanderbilt D (1990) Soft self-consistent pseudopotentials in a generalized eigenvalue formalism. Phys Rev B 41:7892–7895.  https://doi.org/10.1103/PhysRevB.41.7892 CrossRefGoogle Scholar
  20. 20.
    Ouyang R, Curtarolo S, Ahmetcik E et al (2018) SISSO: A compressed-sensing method for identifying the best low-dimensional descriptor in an immensity of offered candidates. Phys Rev Mater doi.  https://doi.org/10.1103/PhysRevMaterials.2.083802 Google Scholar
  21. 21.
    Pedregosa F, Varoquaux G, Gramfort A et al (2011) Scikit-learn: machine learning in python. J Mach Learn Res 12:2825–2830Google Scholar

Copyright information

© Springer Science+Business Media, LLC, part of Springer Nature 2019

Authors and Affiliations

  • Aayush R. Singh
    • 1
    • 2
  • Brian A. Rohr
    • 1
    • 2
  • Joseph A. Gauthier
    • 1
    • 2
  • Jens K. Nørskov
    • 1
    • 2
    • 3
    Email author
  1. 1.Department of Chemical EngineeringSUNCAT Center for Interface Science and Catalysis, Stanford UniversityStanfordUSA
  2. 2.SLAC National Accelerator LaboratorySUNCAT Center for Interface Science and CatalysisMenlo ParkUSA
  3. 3.Department of PhysicsDenmark Technical UniversityKgs. LyngbyDenmark

Personalised recommendations