Among multiple alternatives, Cu has emerged as the most promising candidate for the hydrogenation of CO_X (CO₂, CO, etc.) to methanol, mainly due to the absence of the Sabatier reaction (thermodynamically favoured overreduction to methane). But pure Cu-catalysts show low activity and selectivity, and this can only be improved by using Lewis acidic supports (Al₂O₃, ZrO₂), or by adding promoters such as Ga or Zn. Especially the latter are of high interest, since these systems consistently outperform the catalysts with an irreducible support (higher activity and selectivity). [1] But understanding the active structure of these systems proved to be extremely challenging: The gas atmosphere strongly interacts with the catalyst surface, forming a dynamic equilibrium, which is impossible to understand using classical calculations, or state-of-the-art in-situ spectroscopy. [2] This requires us to look beyond static calculations with slab models and understand the dynamics of the system on a microscopic, fundamental level. Herein, we used ab initio molecular dynamics (AIMD) in combination with Metadynamics (MTD) [3] to explore structurally district configuration of CuGa nanoparticles (NPs) and evaluate their free energy. With this approach, we could rationalise the structure of these NPs with different concentrations of Ga (~10% to ~50 %), both under oxidising conditions (CO₂ hydrogenation or air) and reducing conditions (CO hydrogenation or H₂). For low Ga-loadings in vacuum, Ga-islands are formed on the particle surface, which grow when the Ga-content increases until a full overlayer is obtained. If these particle are exposed to a more oxidising atmosphere, the Ga-islands/overlayers are oxidised to a suboxide, partially or fully covering the Cu NPs.

These findings allowed us to re-examine previously measured XAS data of CuGa NPs with different loadings of Ga, which showed some irregularities with regards to the EXAFS coordination number, especially when considering the observed particle size. Using the data obtained by MTD, we could construct a new model with a partial core-shell structure, which not only resolves the conflict from the XAS data, but also rationalises why Ga acts as a promoter for low concentrations while becoming a poison at higher concentrations. This shows that by combining in-situ XAS and MTD, the limitation of pure experimental setups as well as classical slab calculations can be overcome and the catalytic systems can be understood on a more fundamental level.

[1] E. Lam, et al., J. Catal., 2021, 394, 266-272.
[2] A. Müller, et al., submitted manuscript, 2022.
[3] A. Barducci, et al., WIREs Comput Mol Sci, 2011, 1, 826-843.