Eseroglou, C., Erhard, L., & Comas-Vives, A. (2026, June 2). Active learning for the study of CuGa nanoparticles [Poster Presentation]. TCH Science Days 2026, Wien, Austria. https://doi.org/10.34726/12944
CO₂ hydrogenation to methanol is a key reaction for sustainable chemical production. CuGa nanoparticles exhibit high methanol selectivity and a characteristic volcano-type dependence on Ga content, peaking at ~10% Ga and declining at higher loadings; however, the atomistic origin of this behaviour remains unclear. Here, we employ molecular dynamics simulations based on machine learning interatomic potentials (MACE) to investigate the structural dynamics of Cu-Ga nanoparticles (3-6 nm) across compositions (0-25% Ga), temperatures (100-1000 K), and nanosecond timescales, quantifying Ga segregation using a geometric descriptor.
We find that Ga atoms segregate to the nanoparticle surface, reaching a temperature- and size-dependent equilibrium. Introducing oxygen (O/Ga = 0.5-1.5) from partially oxidative CO₂ hydrogenation conditions further enhances this segregation, stabilizing surface Ga through Ga-O motifs and leading, at high Ga contents, to the emergence of fully oxidized GaOₓ surface layers. Using ab initio thermodynamics, we show that fully oxidized surface phases are favored at high Ga concentrations, whereas partially oxidized configurations are preferred at low Ga content under methanol synthesis conditions.
Based on these results, we propose a mechanistic explanation for the observed volcano behaviour: at low Ga concentrations, surface-segregated Ga modifies the electronic environment of Cu sites and enhances reactant binding at the CuGaOₓ interface, improving catalytic activity, while at higher Ga contents, the formation of stable GaOₓ phases progressively blocks Cu sites, leading to catalyst poisoning. The balance between these competing effects naturally yields an optimal Ga concentration.