More Accurate Approach for Adsorbate Thermophysical Properties

Microkinetic modeling has become increasingly popular for predicting reaction networks in heterogeneous catalysis. The accuracy of adsorbate thermophysical properties is an important aspect of a microkinetic mechanism, as these properties need to be thermodynamically consistent, and the reaction equilibrium constants require accurate free energies of adsorbates. As we transition into the exascale, advanced methods that require a lot of computational power become relevant and useful tools to considerably improve chemical process predictions.
We propose a classical Monte Carlo phase space integration (PSI) approach for obtaining accurate thermophysical properties of adsorbates, and initially assess the method here on a H atom on a Cu(111) surface. We produce training data for the potential energy surface of the system by performing ensemble DFT calculations on Argonne National Laboratory’s Theta system. Two different sampling schema for training data generation are presented. A minima-preserving neural network (MP-NN) is constructed from the training data and used within the PSI routine to obtain the partition function, from which the thermodynamic properties are directly derived. The results are benchmarked against direct state counting results, acquired with discrete variable representation.
We found that contributions of translational anharmonicity are significant for H on Cu(111). The PSI approach is in excellent agreement with a quantum state counting benchmark over the temperature range of interest, especially compared to the most commonly used models, the harmonic oscillator (HO) and the free translator (FT). The free energy of dissociative adsorption of H2 was calculated from the results, as well as the Langmuir isotherms at 400, 800 and 1200 K in a partial pressure range of 0-1 bar. It shows that surface coverages depend heavily on the model used for obtaining the free energy, as the anharmonic effects lead to significantly higher predicted surface site fractions of hydrogen.
Our method has been implemented in our open-source phase space integration code, AdTherm. The method is to be extended for multiatomic adsorbates and will be a part of the Sandia National Laboratories’ ECC computational framework for automated chemistry, a broader effort to provide accurate microkinetic mechanisms.