Accurate Simulations of the Reaction of H₂on a Curved Pt Crystal through Machine Learning

Abstract

Theoretical studies on molecule–metal surface reactions have so far been limited to small surface unit cells due to computational costs. Here, for the first time molecular dynamics simulations on very large surface unit cells at the level of density functional theory are performed, allowing a direct comparison to experiments performed on a curved crystal. Specifically, the reaction of D 2 on a curved Pt crystal is investigated with a neural network potential (NNP). The developed NNP is also accurate for surface unit cells considerably larger than those that have been included in the training data, allowing dynamical simulations on very large surface unit cells that otherwise would have been intractable. Important and complex aspects of the reaction mechanism are discovered such as diffusion and a shadow effect of the step. Furthermore, conclusions from simulations on smaller surface unit cells cannot always be transfered to larger surface unit cells, limiting the applicability of theoretical studies of smaller surface unit cells to heterogeneous catalysts with small defect densities.

SBH17: Benchmark Database of Barrier Heights for Dissociative Chemisorption on Transition Metal Surfaces

“…98 Furthermore, recent work has shown that this DF can reproduce experiments of D 2 on chemically related curved Pt crystals with (111) terraces and (100) steps with chemical accuracy. 105 The barrier height reported for PBEα57-vdW-DF2 was −0.008 eV. We retain this value as the reference value (see Table 2), even though it was set to 0.0 eV in the previous SBH10 database.…”

Quantum and quasi-classical dynamics of the C(3P) + O2(3Σ−g) → CO(1Σ+) + O(1D) reaction on its electronic ground state

“…Given the ease with which QCT simulations can be carried out the present work suggests that they are a meaningful approach for understanding the dynamics of atom plus diatom reactions and for larger systems as well. QCT-based approaches have also been used for the reaction of H 2 on platinum surfaces, 75 the formation of CO 2 on amorphous solid water, 70,71 CO adsorbed on NaCl, 76 and for investigating photodissociation reactions of larger molecules such as acetaldehyde or the syn-Criegee intermediate. 77,78…”

Micron-scale heterogeneous catalysis with Bayesian force fields from first principles and active learning

“…For complex surface structures with a wide variety of dynamic reactive sites, it is highly nontrivial to replace ab initio methods with cheaper and more scalable methods. Previously, cluster expansions [5] and empirical [6] / machine-learned (ML) [7] force fields have been employed. These methods rely only on the atomic positions in local environments and do not explicitly consider the electronic structure, thereby scaling linearly with the number of atoms 𝑂 (𝑁 atom ) and providing several orders-of-magnitude acceleration compared to DFT.…”

“…On noble metals such as Pt, an H 2 molecule approaches the Pt surface and dissociates to form two chemisorbed H atoms, which diffuse and eventually recombine and desorb from the surface as molecular H 2 (Figure 1) [10]. A variety of force field-based methods have been employed to model H/Pt, most notably ReaxFF [11], embedded-atom method (EAM) [12], tight-binding [13], low-dimensional models [14], and ML force fields [7]. Empirical models such as ReaxFF and EAM have been shown to be inaccurate and unreliable for high-temperature long-timescale simulations [15].…”

“…Specifically for the H/Pt heterogeneous systems, previous force fields are limited to describing low-pressure and low-temperature conditions. Latest models include a ReaxFF model from Gai et al [11] and a HDNNP from Gerrits [7]. Both models were optimized to match DFT calculations of H-covered Pt surfaces.…”