Importance of a Nonlocal Description of Electron–Electron Interactions in Modeling the Dissociative Adsorption of H₂ on Cu(100)

Abstract

Describing the dissociation of molecules on metal surfaces is of major importance in understanding processes such as corrosion or heterogeneous catalysis. Because of its high efficiency and reasonable accuracy, density functional theory (DFT) is the method of choice in this field nowadays, but it can be expected that the semilocal treatment of the exchange-correlation introduces fundamental errors. To understand those errors, it is possible to reduce the complexity of the problem to, for example, describe the dissociation of H 2 on Cu(100). In this work, we model this process for an embedded, metallic cluster comparing coupled cluster, configuration interaction, and semilocal DFT calculations. We calculate a two-dimensional potential energy surface (PES) and identify the minimum energy pathway (MEP) for dissociation, the transition state, and also the dissociation minimum. Compared to DFT calculations, the interactions between the molecule and the surface are far more long ranged in explicitly correlated methods, and the dissociation is also far more exothermic. To accurately describe dissociation, it is necessary to take the multideterminant nature of the wave function in embedded MRCI+Q calculations into account. For them, we find excellent agreement with experiment. On the basis of these calculations, we discuss the reasons for shortcomings of semilocal density functionals described in recent literature. These results clearly show that an accurate nonlocal treatment of the electron−electron interactions qualitatively changes the potential energy surface for molecules interacting with extended systems and possibly improves agreement with experiments. It will be interesting to see how they can be transferred to and affect more complex problems such as modeling entire reaction pathways.

Quantum and classical dynamics of reactive scattering of H2 from metal surfaces

“…A new trend is to use a novel implementation 11 of the SRP-DFT approach, 214,215 in which a weighted average of two GGA DFs is used in a semi-empirical manner in order to achieve a chemically accurate description of experiments on a specific system. New developments concern the application of embedded correlated wave function approaches 79,220,226 and of the quantum Monte-Carlo method 239,240 to H 2 -metal surface interactions.…”

“…In this approach, a 12-atom gold nanocluster was carved out from an ideal Au surface. Sautet and co-workers 226 have also performed calculations with a correlated wave function approach, on H 2 + Cu(100). The correlated wave function methods tested were CASSCF, 222 CASPT2, 223,224 and CIS (configuration interaction singles, CI singles 225 ).…”

Anchoring High-Concentration Oxygen Vacancies at Interfaces of CeO2–x/Cu toward Enhanced Activity for Preferential CO Oxidation

“…6,58 In the meantime, the competitive molecules H 2 may dissociate on Cu 0 at noninterface sites and thus could not be oxidized due to the absence of active oxygen species nearby, which conversely favors the preferential CO oxidation. 59,60 Thus, the way that the CeO 2−x /Cu catalyst works could be attributed to the synergetic effect of Ce 3+ −V O coupled with Cu + . In essence, the interfacial oxygen vacancies anchored offer the active oxygen species that give rise to the great performance in preferential CO oxidation, even at high temperature range.…”

Dissociative chemisorption of methane on Ni(111) using a chemically accurate fifteen dimensional potential energy surface

“…Even for single point calculations on molecule-metal surface interactions, stochastic approaches like the quantum-Monte Carlo (QMC) 5 and correlated wave function approaches with density embedding, [6][7][8] while in principle accurate, still face convergency issues. Recent QMC calculations were able to reproduce the semi-empirical reference value of the H 2 + Cu(111) barrier height to within no better than 1.6 kcal/mol, as a result of limitations imposed by fixed node, locality, and time step errors.…”