Scientists Watch Single Platinum Atoms Rearrange During Catalysis, Revealing Two Kinds of Active Sites at Once

For nearly a century, chemists have puzzled over the exact location on a catalyst’s surface where molecules transform. Now a team of researchers in Belgium, Spain and the Netherlands has managed something remarkable: they watched individual platinum atoms move, disappear and reappear on the surface of single nanoparticles while a real chemical reaction was taking place, and in doing so they caught two different classes of active sites living side by side on the same particle.

The study, published in Nature Catalysis, focuses on carbon monoxide oxidation over platinum nanoparticles, a textbook reaction that underpins technologies from automotive catalytic converters to air purification. Despite decades of investigation, fundamental questions about which atomic sites actually drive the reaction have remained unresolved, largely because no analytical technique could observe a working catalyst at the atomic level while gases flowed over it and while its activity was being measured simultaneously.

The researchers, led by K. Jenkinson, T. Stoops, S. Van Aert, Bert Weckhuysen and Sara Bals, combined operando transmission electron microscopy with quantitative image analysis to build three-dimensional atomic models of individual platinum nanoparticles at different stages of a reaction cycle. Rather than capturing a single snapshot, they followed the particles through activation, deactivation and reactivation, tracking how each atom’s local environment changed as the gas atmosphere switched between conditions that favored the reaction and conditions that suppressed it.

The technical achievement rests on several methodological pillars. Using high-angle annular dark-field scanning transmission electron microscopy, or HAADF-STEM, the team acquired images in which the intensity of each atomic column depends strongly on the atomic number of the elements present, making heavy platinum atoms appear as bright, countable features. Statistical model-based quantification methods, developed over years by the Antwerp group, then allowed the researchers to count atoms in each column and estimate a full three-dimensional atomic structure from limited two-dimensional projections. Deep convolutional neural networks were employed to restore single-shot microscopy images, compensating for scan noise and drift, while careful control of the electron dose limited knock-on damage to the delicate nanoparticles under study.

From these reconstructed models, the team computed the coordination number of every surface atom, meaning the count of nearest neighbors each atom possesses. Atoms with fewer neighbors than their bulk counterparts, known as coordinatively unsaturated or undercoordinated atoms, have long been suspected as the seats of catalytic activity. By correlating the abundance of these undercoordinated atoms with the catalytic performance measured during the same experiment, the researchers established, for the first time in a realistic catalyst under working conditions, a direct experimental link between the population of low-coordination sites and reaction behavior.

The central discovery concerns the coexistence of two distinct kinds of active sites. Taylor-type sites, named after the British scientist Hugh Taylor, are associated with low-coordination atoms such as steps, edges and corners, where the electronic structure of the metal is perturbed in ways that facilitate bond activation. Langmuir-type sites, in contrast, relate to the ideal flat terraces contemplated in Irving Langmuir’s adsorption framework, where binding and reaction follow simpler, site-uniform assumptions. Surface science experiments on single crystals have historically treated these two regimes as separate and often mutually exclusive descriptions of catalytic behavior.

What the operando tracking revealed is that, within a single realistic platinum nanoparticle during carbon monoxide oxidation, both types of sites exist at the same time and interconvert dynamically. As the reaction atmosphere changed, surface atoms exchanged positions, facets grew and shrank, and the balance between flat terrace-like regions and stepped, undercoordinated regions shifted continuously. The abundance of each site type responded directly to the gas environment, demonstrating that the active surface is not a static landscape but a fluid, adaptive interface whose atomic architecture is dictated by the reaction itself.

This finding carries significant implications for how chemists conceptualize heterogeneous catalysis. The dominant paradigm of structure sensitivity, which seeks to rationalize reaction rates by counting nearest neighbors and assigning activity to specific geometric motifs, remains valid, but the new results show that any static classification is incomplete. A catalyst particle operating in a reactor is constantly restructuring, and the sites responsible for catalysis may migrate across the surface over the course of a single reaction cycle. Models that assume a fixed population of active sites therefore risk misrepresenting the true nature of the working catalyst.

The broader significance extends beyond the specific platinum-carbon monoxide system. The methodology, which enables quantitative tracking of facets and undercoordinated atoms in catalytic nanoparticles under reactive atmospheres, opens the door to experimental investigations of active-site roles in many other catalytic processes that were previously accessible only through computational modeling or idealized surface science. Combined with molecular dynamics simulation tools and complementary spectroscopic measurements, operando atomic-resolution electron microscopy of this kind could help close the persistent gap between model systems and industrial catalysts, where pressures, temperatures and complex gas mixtures conspire to obscure atomic-scale behavior.

For the catalyst design community, the work offers both a caution and an opportunity. The caution is that single-site descriptions of catalysts may be fundamentally misleading, since a working nanoparticle can host multiple coexisting site types whose relative importance fluctuates with reaction conditions. The opportunity is that, if the interplay between Taylor and Langmuir sites can be understood and ultimately engineered, it may become possible to design nanoparticles whose surfaces dynamically favor the most productive site configurations under operating conditions. The three-dimensional datasets, mass spectrometry records and microscopy images generated during the study have been made available through the Zenodo repository, allowing other researchers to scrutinize and build upon this rare atom-by-atom view of catalysis in action.

Subject of Research: Operando atomic-resolution tracking of active sites on platinum nanoparticles during catalytic carbon monoxide oxidation

Article Title: Operando single-atom tracking in individual Pt nanoparticles detects the coexistence of Langmuir and Taylor active sites

Article References: Operando single-atom tracking in individual Pt nanoparticles detects the coexistence of Langmuir and Taylor active sites. (n.d.). https://doi.org/10.1038/s41929-026-01612-w

Image Credits: AI Generated

DOI: 10.1038/s41929-026-01612-w

Keywords: heterogeneous catalysis, platinum nanoparticles, active sites, operando electron microscopy, CO oxidation, Taylor sites, Langmuir sites, HAADF-STEM, coordination number, 3D atomic imaging, catalyst structure, surface reconstruction