Water molecules found to gate the door on carbon dioxide electrolysis

Water, long treated as a passive backdrop in electrochemistry, has been caught actively controlling who gets to the electrode and who does not. In a study published in Nature Catalysis, researchers led by F. Pelayo García de Arquer of ICFO in Barcelona, working with Núria López’s computational group at ICIQ, have watched the carbon dioxide reduction reaction unfold in real time inside a working membrane electrode assembly, at current densities of up to one ampere per square centimetre, the regime that industrial CO2 electrolysers actually operate in. What they saw overturns the conventional picture: the ordered layer of water molecules hugging the copper catalyst behaves like a gatekeeper, determining whether carbon species can reach the interface or whether the current is wasted making hydrogen instead.

The stakes are high. Electrochemical CO2 reduction promises to convert captured carbon dioxide into valuable fuels and chemicals such as carbon monoxide, methane, ethylene and ethanol, offering a route to decarbonise chemical manufacturing and transport. Copper remains the most promising electrocatalyst for these products because its selectivity can be tuned, and specific facets and ensembles of copper atoms enable the carbon-carbon coupling steps that build multicarbon molecules. Yet decades of mechanistic studies have relied on idealised static surfaces and simplified adsorbate models, and the discrepancies between different in situ spectroscopy studies have persistently frustrated efforts to build a unified picture. Most past experiments were also performed at current densities far below those relevant to deployment, in cells whose stagnant electrolytes and non-uniform electric fields do not resemble a real electrolyser.

The Barcelona team tackled this problem on two fronts. First, they used a three-electrode open flow cell with 1 M potassium bicarbonate electrolyte, which supplies a continuous stream of CO2 and decouples thermodynamic driving forces from mass transport limits. Then they built a custom operando Raman spectroscopy platform integrated directly into a membrane electrode assembly, achieving optical access to the buried cathodic interface through the anion exchange membrane and the iridium oxide-coated titanium anode. This allowed them to record spectra during genuine continuous CO2 electrolysis across a broad range of current densities, all the way to one ampere per square centimetre, while product analysis proceeded in parallel through gas chromatography and high-performance liquid chromatography.

The key to their analysis lies in how water organises itself at an electrified interface. Liquid water is best viewed as a continuously evolving, disordered hydrogen-bond network, but individual molecules can be classified phenomenologically by how many hydrogen bonds they make: tetrahedral water with four hydrogen bonds, trihedral water with three, and isolated free water. These populations shift with potential, current density, surface coverage and local ionic environment, and they can be quantified from the hydrogen-oxygen stretching band of the Raman spectrum. The researchers deconvoluted the stretching envelope into Gaussian components centred near 3,200 and 3,400 reciprocal centimetres for the four- and three-hydrogen-bond waters respectively, and tracked how these populations evolved as current was applied.

What emerged was a strikingly non-monotonic story. At rest, interfacial water sat in a random orientation with a dominant four-hydrogen-bond character. As current density rose toward 0.1 A per square centimetre, the three-hydrogen-bond fraction climbed from about 45 percent to 84 percent, with molecules flipping into an H-down orientation that aligns their hydrogens toward the negatively charged surface. This population of three-hydrogen-bond water turned out to be the primary proton source for CO2 reduction, and its rise coincided with growing selectivity toward multicarbon products. The four-hydrogen-bond fraction fell correspondingly, dipping to a minimum of about 16 percent. Between 0.2 and 0.4 A per square centimetre the picture shifted again, and beyond 0.4 A per square centimetre both water types flipped to H-down configurations as the increasingly strong interfacial electric field took over.

To understand why, the team turned to ab initio molecular dynamics simulations run on copper surfaces with varying coverages of potassium cations and adsorbed species. The simulations revealed a division of labour: adsorbed potassium ions act as hydrogen-bond breakers, dismantling the interfacial water network, whereas adsorbed hydroxide, carbon monoxide and carbonate species act as hydrogen-bond makers, extending and stiffening the network. Increasing the number of potassium cations in the double layer dropped the four-hydrogen-bond water fraction from 0.70 to 0.45, while adding hydroxide at that same cation coverage restored it to 0.73. This explained the experimental oscillation: as the potential grows more negative, anion coverage falls and cation concentration rises, disrupting water order, until the accumulation of polar adsorbates such as hydroxide and carbonate reorders the network once more, making the interfacial layer increasingly ice-like and rigid.

The coupling between water structure and reaction intermediates was mapped with two-dimensional correlated Raman spectroscopy, a technique that compares intensity fluctuations across many spectra to reveal which vibrational modes change together and in what order. The analysis traced a sequential pathway: adsorbed carbon monoxide forms first, shifting from a low-frequency to a high-frequency band; adsorbed hydroxide then appears and stabilises the high-frequency carbon monoxide species; and that intermediate converts into a carbonaceous intermediate implicated in carbon-carbon coupling. A strong positive correlation between the three-hydrogen-bond water signal and the high-frequency carbon monoxide band confirmed that interfacial water physically rearranges to accommodate and stabilise the species needed for multicarbon products.

Then came the gating mechanism itself. Beyond roughly 0.2 A per square centimetre, hydroxide generation falls behind its removal through the anion exchange membrane, carbonate accumulates, and the four-hydrogen-bond water fraction rebounds. This denser, more ordered water layer restricts access of carbon dioxide and carbon species to the interface while providing abundant protons for the competing hydrogen evolution reaction. A control experiment feeding argon instead of CO2 produced nearly 90 percent faradaic efficiency for hydrogen with a similar four-hydrogen-bond water content, confirming that the ordered water layer gates access to the catalyst surface regardless of which reactants are present. Electrochemical performance tracked the spectroscopy closely: carbon dioxide selectivity peaked, multicarbon selectivity reached 45 percent and the hydrogen efficiency bottomed out at 0.2 A per square centimetre, precisely where hydroxide coverage and three-hydrogen-bond water were maximal and carbonate and four-hydrogen-bond water were minimal.

The practical implications are considerable. If the interfacial hydrogen-bond network is a design variable rather than background noise, it can be engineered. The authors suggest that surface functionalisation with hydroxyl, amine, imidazole or sulfonate groups, hydrophobic terminations that modulate water orientation, tailored electrolytes that control cation concentration, and selective water-exchange membranes could all be used to hold the interface in the productive regime. Carbon monoxide reduction studies, which eliminate the carbonate signals that complicate spectroscopy, could refine the picture further, since surface-bound hydrogen donors and intermediates could then be tracked cleanly.

What makes the work resonate beyond electrochemistry is the demonstration that at industrial current densities, the hardest problems may not be about the catalyst at all, but about water management at a scale of a few molecular layers. The same molecules that deliver protons for synthesis can also cement themselves into a barrier that chokes off the reaction. Seeing that gate open and close, molecule by molecule, while the electrolyser runs, gives catalyst and system designers for the first time a molecular target to aim for: not merely a better copper surface, but a better-ordered dance of water above it.

Subject of Research: Interfacial water structure and its role in gating CO2 electroreduction on copper catalysts at high current density

Article Title: Operando resolution of CO2 electrolysis interfaces at application-relevant current density reveals water gating

Article References: Guha, A., Seemakurthi, R. R., Polesso, B., Haldar, P., Ram, R., Pinilla-Sánchez, A., Mondal, I., Kernan, D., López, N., & García de Arquer, F. P. (2026). Operando resolution of CO2 electrolysis interfaces at application-relevant current density reveals water gating. Nature Catalysis. https://doi.org/10.1038/s41929-026-01620-w

Image Credits: AI Generated

DOI: 10.1038/s41929-026-01620-w

Keywords: CO2 reduction, electrocatalysis, copper catalyst, interfacial water, Raman spectroscopy, molecular dynamics, hydrogen evolution, membrane electrode assembly, hydrogen-bond network, multicarbon products, electrolyser, water gating