Aleix Comas-Vives

Cu–Ga nanoparticles are promising catalysts for thermochemical and electrochemical CO2 conversion, with their catalytic behavior determined by their surface composition. Here, we combine density functional theory (DFT) calculations with machine-learned interatomic potentials (MLIPs) to determine how Ga distributes within Cu–Ga nanoparticles depending on composition, size, and temperature. DFT results show that Ga forms a homogeneous solid solution within the bulk Cu lattice, whereas Ga-rich terminations are favored for surfaces. Molecular dynamics (MD) simulations at the nanosecond scale for particles (within 2 nm to 4.5 nm) and Ga contents up to 24 % reveal progressive Ga migration toward the surface. The extent and rate of this redistribution increase with temperature and Ga content, accompanied by increasing sphericity and structural disorder, until Ga surface enrichment saturates. The surface Ga fraction remains similar across particle sizes, whereas the fraction of total Ga at the surface decreases as particle size increases. A mean-field model captures this surface-to-interior partitioning, reproducing the surface compositions for the MD simulations at high temperatures. At lower temperatures and for larger nanoparticles, Ga redistribution is kinetically limited within the timescale of the simulations. Calculated apparent diffusion barriers for Cu and Ga obtained at high temperatures support this interpretation.

Mixed metal oxide catalysts are highly active for CO2 hydrogenation to methanol and typically more selective than classical Cu-based catalysts. Adsorption energies of key reaction intermediates can serve as activity descriptors for high-throughput screening of metal oxide-dopant combinations, but this requires numerous computationally demanding density functional theory (DFT) calculations. Machine-learning interatomic potentials (MLIPs) and adsorption energy prediction models offer faster alternatives, yet current architectures struggle with the structural and electronic complexity of mixed metal oxides introduced by dopants, oxygen vacancies, and diverse adsorption motifs. In this work, a graph-based Gaussian process regression (GPR) model is developed, combining graph representations of adsorption structures with node-level electronic descriptors. Trained on 638 adsorption structures of CO and all formate-pathway intermediates on pristine and Zn-doped ZrO2 and HfO2, as well as pristine In2O3 surfaces, the model achieves a test-set mean absolute error (MAE) of 0.16 eV. Leave-one-feature-out analysis identifies the d-band center, surface work function, and adsorbate LUMO energy as the dominant descriptors governing the prediction accuracy. Coupling the model with UMA small v1.2-D3 establishes an efficient workflow for rapid structural relaxation and accurate adsorption energy prediction, enabling high-throughput screening of mixed metal oxide catalysts for CO2 hydrogenation to methanol.

Mixed metal oxides, such as ZnZrO _x , have attracted considerable interest as CO ₂ hydrogenation to methanol catalysts due to their high methanol selectivity (> 70%) and catalytic stability at elevated reaction temperatures (> 300°C). In this work, we introduce a novel ZnHfO _x catalyst that exceeds the intrinsic methanol formation rate of the reference ZnZrO _x at a Zn content of 20 mol% ( r _{MeOH,20ZnZrOx} = 0.59 mol _MeOH (mol _cat h) ⁻¹ , r _{MeOH,20ZnHfOx} = 0.68 mol _MeOH (mol _cat h) ⁻¹ ). Remarkably and in contrast to ZnZrO _x , the ZnHfO _x ‐based catalysts exhibit a high methanol selectivity (> 70%) up to Zn contents of 99.5 mol% despite the segregation of ZnO. Operando spectroscopy, in combination with computational analysis, identifies the Zn–V _O –Hf motif as the active site for methanol formation that proceeds via the formate‐methoxy pathway. Such active sites are not only present in solid solution‐type ZnHfO _x catalysts (≤ 35 mol% Zn), but also in the form of isolated HfO _x clusters on segregated ZnO surfaces (for high Zn contents of > 35 mol%), explaining the high selectivity (and activity) over a wide range of Zn contents.

Mixed metal oxides, such as ZnZrO _x , have attracted considerable interest as CO ₂ hydrogenation to methanol catalysts due to their high methanol selectivity (> 70%) and catalytic stability at elevated reaction temperatures (> 300°C). In this work, we introduce a novel ZnHfO _x catalyst that exceeds the intrinsic methanol formation rate of the reference ZnZrO _x at a Zn content of 20 mol% ( r _{MeOH,20ZnZrOx} = 0.59 mol _MeOH (mol _cat h) ⁻¹ , r _{MeOH,20ZnHfOx} = 0.68 mol _MeOH (mol _cat h) ⁻¹ ). Remarkably and in contrast to ZnZrO _x , the ZnHfO _x ‐based catalysts exhibit a high methanol selectivity (> 70%) up to Zn contents of 99.5 mol% despite the segregation of ZnO. Operando spectroscopy, in combination with computational analysis, identifies the Zn–V _O –Hf motif as the active site for methanol formation that proceeds via the formate‐methoxy pathway. Such active sites are not only present in solid solution‐type ZnHfO _x catalysts (≤ 35 mol% Zn), but also in the form of isolated HfO _x clusters on segregated ZnO surfaces (for high Zn contents of > 35 mol%), explaining the high selectivity (and activity) over a wide range of Zn contents.

Copper is a highly promising catalyst for the electrochemical CO₂ reduction reaction (CO2RR) since it is the only pure metal that can form highly added-value products such as ethylene and ethanol. Since the CO2RR takes place in aqueous solution, the detailed atomic structure of the water–copper interface is essential for unraveling the key reaction mechanisms. In this study, we investigate copper–water interfaces exhibiting nanometer-scale roughnesses. We introduce two molecular dynamics protocols to create rough copper surfaces, which are subsequently brought into contact with water. From these interfaces, we sample additional training configurations from machine-learning-interatomic-potential-driven molecular dynamics simulations containing hundreds of thousands of atoms. An active learning workflow is developed to identify regions with high spatially resolved uncertainty and convert them into DFT-feasible cells through a modified amorphous matrix embedding approach. Finally, we analyze the local environments at the interface using unsupervised machine-learning techniques. Unique environments emerge on the rough copper surfaces absent from model systems, including stacking-fault-induced configurations and undercoordinated corner atoms. Notably, corner atoms consistently feature chemisorbed water molecules in our simulations, indicating their potential importance in catalytic processes.

ConspectusCatalytic CO₂ hydrogenation to methanol is among the most attractive routes in CO₂ conversion, as methanol is a chemical feedstock and a relevant energy carrier for the sustainable methanol economy. Cu-based catalysts are the typical choice for this reaction, and Cu-ZnO-Al₂O₃, the industrial reference material for hydrogenating CO (in the presence of CO₂), has also been shown to perform well for CO₂ hydrogenation. Adding other elements to Cu NPs as promoters (Zn, Ga, In, etc.) and using specific supports (ZrO₂, Al₂O₃) enhance the catalytic activity and selectivity of Cu toward methanol, often minimizing the undesired competitive Reverse Water-Gas Shift and methanation reactions. However, these materials are complex, showing a delicate interplay between metal-metal and metal-support interactions in driving the overall selectivity toward methanol. Besides, the reactive gas-phase atmosphere (CO/CO₂/H₂/H₂O in various ratios), in other words, the chemical potential, significantly affects the catalyst states, in terms of both structures and dynamics; this additional complexity often precludes the identification of the active sites, hampering the design of better catalytic materials based on structure-activity relationships derived from simple descriptors.In this Account, we show how combining experiments and atomistic calculations provides detailed information on how interfaces, alloying, and dynamics play a crucial role in CO₂ hydrogenation by stabilizing specific adsorbates from CO₂ to key reaction intermediates, i.e., formate or methoxy. Specifically, we discuss the role played by metal/oxide interfaces and alloying/dealloying processes in driving catalytic activity (and selectivity); we also highlight how reaction conditions that define the chemical potential alter the stability and dynamics of the reactive states of catalysts. All of these aspects are crucial and interconnected, hence a challenge for both experimental and theoretical approaches.This Account discusses these challenges and exemplifies their importance, focusing on the following: (i) How benchmarking catalytic models against experimental data is crucial in obtaining reliable computational models of the active sites; (ii) How specific surfaces/interfaces are particularly suited to stabilize key catalytic intermediates such as activated CO₂, formate, and methoxy species; (iii) How dynamic changes in the systems can be accounted for via ab initio molecular dynamics combined with metadynamics, confronted with in situ X-ray absorption spectroscopy; (iv) How the "oxygen chemical potential" defined by the applied reaction conditions (e.g., H₂/CO₂ ratio) may affect the nature and stability of catalysts by using ab initio atomistic thermodynamics.Finally, we provide an outlook on ongoing methodological developments that are needed to refine our understanding of the properties of these fascinating and dynamic materials.

Cu-based catalysts are considered references and “gold” standards in the selective hydrogenation of CO2 to methanol, a chemical feedstock and an intermediate to sustainable fuels. Improving catalyst performances, i.e., higher activity and methanol selectivity, relies on the use of promoters such as Ga and Zn, which are known to alloy with Cu. Nevertheless, establishing detailed, molecular-level relationships between structure and activity remains challenging as multi-metallic catalysts readily and dynamically respond to reaction conditions associated with specific chemical potentials. Notably, CO2 hydrogenation reaction conditions affect the catalyst structure by being reducing for Cu, but oxidizing for Ga or Zn, hence the possible dynamics. Furthermore, these systems always contain a support, which is used to keep the active phase dispersed, but also generates additional interfaces and complexity. Here, we evaluate the structures and dynamics of CuGa and CuZn alloys using ab initio molecular dynamics combined with metadynamics, focusing in particular on i) alloying stability for CuGa and CuZn NPs across promoter loadings (10.5 – 52.6 %), ii) the effect of the support, here silica, on the alloy structure, and iii) how CO2 hydrogenation conditions favor GaOx and ZnOx phases, to understand and rationalize the catalytic activity of these materials. Remarkably, both promoters (Ga and Zn) prefer to locate at the Cu particle surface, with increasing promoter loadings, and Zn atoms readily desorb from the particle. CuZn particles interact significantly more strongly with amorphous SiO2 than CuGa ones. Under CO2 hydrogenation conditions, considering the addition of O adatoms, Ga completely oxidizes to GaIII when enough oxygen is present. In contrast, Zn only partially oxidizes, and a small fraction remains metallic and alloyed with Cu. These differences explain why Zn exhibits better tolerance to high promoter concentrations compared to Ga, and likely why Zn is preferred over Ga at the industrial level.

Adsorption energies are key catalytic descriptors that reveal adsorbate-site interactions on heterogeneous catalysts. However, their computation via DFT is time-consuming, limiting high-throughput screening. This work presents a machine learning (ML) methodology based on graph representations of local adsorption sites, using a Graph Neural Network (GNN) with per-atom local descriptors derived from accessible physicochemical properties. The approach is evaluated on two bimetallic datasets. The first includes AB-type bimetallic flat surfaces with varying A:B ratios, predicting binding energies for small monodentate adsorbates (C, N, O, S, H) with MSEs of 0.073/0.181 eV (train/test). The second dataset comprises reaction energies of key intermediates for CO2 hydrogenation on Ni-Ga-based surfaces. The GNN model achieves an impressive performance (MSE: 0.001/0.002 (train/test) eV) on complex atomic configurations, even bidentate ones. Beyond predictive performance, clustering analysis provides an explainable framework, showing how structural and electronic descriptors can rationally guide catalyst design and deepen understanding of adsorbate-metal interactions.

In this work, we assess how the structure of SiO₂-supported, Ni–Ga alloys determines their activity and selectivity for the hydrogenation of CO₂ to methanol. Using a hydrothermal deposition-precipitation approach followed by activation at 700 °C in H₂, we synthesize catalysts containing α-Ni, α-Ni₉Ga, α’-Ni₃Ga, or δ-Ni₅Ga₃ phases supported on amorphous SiO₂. Operando X-ray pair distribution function analysis and X-ray absorption spectroscopy confirm unequivocally the structure of all phases and their stability under reaction conditions; additionally, all catalysts contain GaO _x species in varying amounts. We observe that the catalysts α’-Ni₃Ga/SiO₂ and δ-Ni₅Ga₃/SiO₂ exhibit high methanol formation rates (∼0.8 mmol_MeOH mol_Ni ^–1 s^–1), which are 27 times greater than those of α-Ni₉Ga/SiO₂ and α-Ni/SiO₂. Notably, α’-Ni₃Ga/SiO₂ shows the highest selectivity for methanol at 71%, compared to 55% for δ-Ni₅Ga₃/SiO₂ and 11% for α-Ni₉Ga/SiO₂, which challenges the conventional view of α’-Ni₃Ga being a poor catalyst for methanol synthesis. To explain the high methanol selectivity and productivity of α’-Ni₃Ga/SiO₂ compared to the other alloy phases, DFT calculations were performed. It was found that the Ni-rich step sites in α’-Ni₃Ga effectively stabilize key reaction intermediates (HCOO* and CH₃O*) for the formation of methanol. However, such Ni-rich step sites in α’-Ni₃Ga also favor CO* dissociation, which could facilitate methane formation, yet the presence of GaO _x decreases the stability of CO* on α’-Ni₃Ga, explaining ultimately the promotion of HCOO* formation. This study highlights the importance of Ga species (both metallic and oxidic) in modulating the electronic properties of heterogeneous catalysts, providing a versatile toolbox to stabilize key reaction intermediates, leading ultimately to high product selectivity.