Binary Approach to Ternary Cluster Expansions: NO–O–Vacancy System on Pt(111)

Bajpai, Anshumaan; Frey, Kurt; Schneider, William F.

doi:10.1021/acs.jpcc.7b00914

Cited by 25 publications

(59 citation statements)

References 78 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…kMC has proven its potential in this respect having been successfully applied to a large variety of catalytic systems. [24][25][26][27][28][29][30][31][32][33][34][35][36][37] A recent kMC framework is the graph-theoretical Zacros code of Stamatakis et al 38,39 which is particularly suited for applications in catalysis, since it is able to deal with multidentate species, complex elementary steps (e.g. involving more than two sites in specific geometric arrangements), and reaction barriers changing in the presence of spectator species that exert lateral interactions.…”

mentioning

confidence: 99%

A machine learning approach to graph-theoretical cluster expansions of the energy of adsorbate layers

Vignola

Steinmann

Vandegehuchte

et al. 2017

The Journal of Chemical Physics

View full text Add to dashboard Cite

_________________________________________The accurate description of the energy of adsorbate layers is crucial for the understanding of chemistry at interfaces. For heterogeneous catalysis not only the interaction of the adsorbate with the surface, but also adsorbate-adsorbate lateral interactions significantly affect activation energies of reactions. Modeling the interactions of adsorbates with the catalyst surface and with each other can be efficiently achieved in the cluster expansion Hamiltonian formalism, which has recently been implemented in a graph-theoretical kinetic Monte Carlo (kMC) scheme to describe multi-dentate species. Automating the development of the cluster expansion Hamiltonians for catalytic systems is challenging and requires the mapping of adsorbate configurations for extended adsorbates onto a graphical lattice. The current work adopts machine learning methods to reach this goal. Clusters are automatically detected based on formalized, but intuitive chemical concepts. The corresponding energy coefficients for the cluster expansion are calculated by an inversion scheme. The potential of this method is demonstrated for the example of ethylene adsorption on Pd (111), for which we propose several expansions, depending on the graphical lattice. It turns out that for this system the best description is obtained as a combination of single molecule patterns and a few coupling terms accounting for lateral interactions. _______________________________________

show abstract

mentioning

confidence: 99%

A machine learning approach to graph-theoretical cluster expansions of the energy of adsorbate layers

Vignola

Steinmann

Vandegehuchte

et al. 2017

The Journal of Chemical Physics

View full text Add to dashboard Cite

show abstract

“…While the energy is represented by only a single term in interaction-free models, current CE models for catalytic reactions can have 20 and more terms. [7,16,[18][19][20] This means that the computational effort of every single evaluation of the Hamiltonian can be 20 times higher (or more) in CE-based models compared to models without lateral interactions. II.…”

Section: Slow Evaluation Of the Cluster Expansion Hamiltonian (Ceh)mentioning

confidence: 99%

“…CE models are typically fitted to first‐principles calculations, although simple CEs have been successfully fitted to experimental temperature‐programmed desorption spectra and adsorption isotherms . The CE is, in principle, exact, but is always truncated in practical application due to constraints in the number of parameters that can be determined (and meaningfully fitted) on a first‐principles level. In the past years, significant effort has been spent on reducing the computational effort involved in determining the CE parameters by applying machine learning and other sophisticated techniques and making such models more reliable …”

Section: Introductionmentioning

confidence: 99%

“…[51] The CE is, in principle, exact, but is always truncated in practical application due to constraints in the number of parameters that can be determined (and meaningfully fitted [20] ) on a first-principles level. In the past years, significant effort has been spent on reducing the computational effort involved in determining the CE parameters by applying machine learning [52,53] and other sophisticated techniques [20,54] and making such models more reliable. [20,53,54] At this point, complex CE models for catalytic co-adsorbate systems with several intermediates, three-and four-body interactions, and interactions "beyond nearest-neighbors" [7,16,[18][19][20][21][22]49,[52][53][54] are already a reality, which makes complex CE models almost readyto-use for practical application in the modeling of heterogeneous catalysis.…”

Section: Introductionmentioning

confidence: 99%

“…In the past years, significant effort has been spent on reducing the computational effort involved in determining the CE parameters by applying machine learning [52,53] and other sophisticated techniques [20,54] and making such models more reliable. [20,53,54] At this point, complex CE models for catalytic co-adsorbate systems with several intermediates, three-and four-body interactions, and interactions "beyond nearest-neighbors" [7,16,[18][19][20][21][22]49,[52][53][54] are already a reality, which makes complex CE models almost readyto-use for practical application in the modeling of heterogeneous catalysis. However, their incorporation into KMC models is currently impeded by severe performance issues: current benchmark of two popular KMC codes report an increase in computing time per KMC step by up to four orders of magnitude after introduction of lateral interactions into a test model.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Efficient Implementation of Cluster Expansion Models in Surface Kinetic Monte Carlo Simulations with Lateral Interactions: Subtraction Schemes, Supersites, and the Supercluster Contraction

Heß

2019

J Comput Chem

View full text Add to dashboard Cite

While lateral interaction models for reactions at surfaces have steadily gained popularity and grown in terms of complexity, their use in chemical kinetics has been impeded by the low performance of current kinetic Monte Carlo (KMC) algorithms. The origins of the additional computational cost in KMC simulations with lateral interactions are traced back to the more elaborate cluster expansion Hamiltonian, the more extensive rate updating, and to the impracticality of rate‐catalog‐based algorithms for interacting adsorbate systems. Favoring instead site‐based algorithms, we propose three ways to reduce the cost of KMC simulations: (1) representing the lattice energy by a smaller Supercluster Hamiltonian without loss of accuracy, (2) employing the subtraction schemes for updating key quantities in the simulation that undergo only small, local changes during a reaction event, and (3) applying efficient search algorithms from a set of established methods (supersite approach). The cost of the resulting algorithm is fixed with respect to the number of lattice sites for practical lattice sizes and scales with the square of the range of lateral interactions. The overall added cost of including a complex lateral interaction model amounts to less than a factor 3. Practical issues in implementation due to finite numerical accuracy are discussed in detail, and further suggestions for treating long‐range lateral interactions are made. We conclude that, while KMC simulations with complex lateral interaction models are challenging, these challenges can be overcome by modifying the established variable step‐size method by employing the supercluster, subtraction, and supersite algorithms (SSS‐VSSM). © 2019 Wiley Periodicals, Inc.

show abstract

Scalable approach to high coverages on oxides via iterative training of a machine‐learning algorithm

2020

View full text Add to dashboard Cite

Understanding the interaction of multiple types of adsorbate molecules on solid surfaces is crucial to establishing the stability of catalysts under various chemical environments. Computational studies on the high coverage and mixed coverages of reaction intermediates are still challenging, especially for transition‐metal compounds. In this work, we present a framework to predict differential adsorption energies and identify low‐energy structures under high‐ and mixed‐adsorbate coverages on oxide materials. The approach uses Gaussian process machine‐learning models with quantified uncertainty in conjunction with an iterative training algorithm to actively identify the training set. The framework is demonstrated for the mixed adsorption of CHx, NHx and OHx species on the oxygen vacancy and pristine rutile TiO2(110) surface sites. The results indicate that the proposed algorithm is highly efficient at identifying the most valuable training data, and is able to predict differential adsorption energies with a mean absolute error of ∼0.3 eV based on <25 % of the total DFT data. The algorithm is also used to identify 76 % of the low‐energy structures based on <30 % of the total DFT data, enabling construction of surface phase diagrams that account for high and mixed coverage as a function of the chemical potential of C, H, O, and N. Furthermore, the computational scaling indicates the algorithm scales nearly linearly (N1.12) as the number of adsorbates increases. This framework can be directly extended to metals, metal oxides, and other materials, providing a practical route toward the investigation of the behavior of catalysts under high‐coverage conditions.

show abstract

Binary Approach to Ternary Cluster Expansions: NO–O–Vacancy System on Pt(111)

Cited by 25 publications

References 78 publications

A machine learning approach to graph-theoretical cluster expansions of the energy of adsorbate layers

A machine learning approach to graph-theoretical cluster expansions of the energy of adsorbate layers

Efficient Implementation of Cluster Expansion Models in Surface Kinetic Monte Carlo Simulations with Lateral Interactions: Subtraction Schemes, Supersites, and the Supercluster Contraction

Scalable approach to high coverages on oxides via iterative training of a machine‐learning algorithm

Contact Info

Product

Resources

About