The appeal of model catalysts lies in their simplicity. Industrial ethylene epoxidation catalysts consist of silver particles dispersed on porous alumina, often promoted with chlorine, cesium, and other additives. That complexity is a nightmare for mechanistic studies: signals from the support, the promoters, and the metal all overlap in spectroscopic measurements, and the roughness of the porous support frustrates high-resolution imaging. A model catalyst replaces this tangled system with a flat, chemically well-defined substrate carrying metal particles whose size, shape, and density are controlled by the experimenter. HOPG, with its atomically flat and chemically inert basal planes, has long been a favorite support for such studies. The catch is that silver does not wet graphite well, so depositing enough metal to cover a useful fraction of the surface without producing a continuous film is a genuine materials challenge.

The Novosibirsk group, led by Anna Nartova together with coauthors including Alexandra Ananina, Aleksey Dmitrachkov, Ren Kvon, Maxim Panafidin, Boris Andryushechkin, and Valerii Bukhtiyarov, set out to prepare samples with large silver particles and high surface coverage, precisely the combination needed for X-ray photoelectron spectroscopy, or XPS, studies of the epoxidation reaction. XPS is exquisitely sensitive to the chemical state of surface atoms, but it needs a substantial amount of the element of interest to produce strong, interpretable signals. If silver particles are too small or too sparse, the spectrum is dominated by the graphite support and the oxygen-bearing species on the silver cannot be tracked reliably. Large particles, in contrast, generate robust Ag 3d and O 1s signals, but growing them uniformly across a flat support is easier said than done.

The researchers compared several preparation routes, and the results reveal how sensitively the final catalyst morphology depends on the details of the deposition process. One approach involved damaging the graphite basal plane by etching with argon ions before depositing silver. The logic is straightforward: defects on the otherwise inert graphite surface act as nucleation sites where silver atoms can anchor, potentially increasing the density of particles and the coverage they achieve. In practice, however, the strategy produced two unwelcome side effects. First, the dense population of closely spaced particles merged as they grew, forming a film-like silver coating rather than the discrete, well-separated particles that mechanistic studies require. Second, and perhaps more insidiously, the damaged graphite promoted encapsulation of silver particles by carbon layers, a phenomenon known as Ag-C encapsulation, in which the metal becomes partially buried beneath graphitic material.

Both effects are fatal for the intended purpose. A film-like coating no longer models the dispersed particles of a real catalyst, so any particle-size or geometry effects on reactivity are lost. Encapsulation is worse still, because it changes which silver atoms are actually exposed to the reactant gases and can create misleading spectroscopic signatures. A sample whose active surface is partially hidden under carbon cannot serve as a trustworthy platform for studying how oxygen species on silver convert ethylene to ethylene oxide. The authors conclude that argon-ion etching of the graphite basal plane, whatever its benefits for nucleation density, makes such samples unsuitable for reaction mechanistic studies, a cautionary finding for anyone tempted to use defect engineering to boost metal loading on graphitic supports.

The winning strategy was far less invasive: thermal vacuum deposition of silver using an electron-beam evaporator onto freshly peeled HOPG. By cleaving the graphite immediately before deposition, the researchers ensured an atomically clean basal plane free of adsorbed contamination, and the gentle thermal evaporation allowed silver to aggregate into discrete particles on the pristine surface. The resulting samples featured uniformly distributed, relatively flat silver particles with mean sizes that could be controlled across a range of roughly 30 to 100 nanometers. Crucially, the silver coverage reached up to 0.59 of the support surface area, meaning that nearly sixty percent of the graphite was decorated with silver particles, a loading high enough to yield strong XPS signals while still preserving the discrete-particle morphology that mechanistic work demands.

Quantifying these morphologies required more than eyeballing micrographs. The team combined scanning electron microscopy, scanning tunneling microscopy, and XPS with deep-learning image analysis, using tools developed by the group to extract particle-size distributions, the number of silver particles per unit area, and the fraction of the support covered by metal. This is a notable methodological point in its own right: manual counting of thousands of nanoparticles across multiple imaging modalities is slow and prone to operator bias, whereas convolutional neural networks can process microscopy images consistently and at scale. The deep-learning approach, which the authors have previously applied to scanning transmission electron microscopy data and to automated analysis of X-ray photoelectron spectra, allowed rigorous comparison between preparation methods on objective, quantitative footing.

The study also probed how the prepared samples behave under thermal treatment, a question of direct relevance to reaction conditions. When a sample containing 100-nanometer silver particles was annealed in vacuum at 500 degrees Celsius, the particles underwent Ostwald ripening, the classic coarsening process in which small particles shrink and disappear as their atoms migrate to feed larger neighbors. Ripening reduces the number of particles and shifts the size distribution upward, which means that any catalytic measurement performed after high-temperature treatment reflects a different surface than the one initially prepared. Documenting this behavior on the model system provides a benchmark for how much thermal history a given sample has experienced and helps researchers design experiments in which the particle population is stable, or at least well characterized, throughout the measurement.

The final validation came from the oxygen signal. Analysis of the O 1s XPS line confirmed that the prepared low-oxygen Ag/HOPG systems are suitable model catalysts for instrumental studies of ethylene epoxidation. In other words, the surfaces carry silver in a state where adsorbed and reactive oxygen species can be detected and distinguished from background oxygen on the support, opening the door to in situ experiments that track the chemical identity of the oxygen species responsible for selective epoxidation. That question, whether electrophilic oxygen species, nucleophilic oxygen, or some combination drives the selective pathway, has animated the epoxidation literature for years, and clean model systems are the key to answering it with spectroscopic rather than circumstantial evidence.

The broader significance of the work extends beyond one reaction. Ethylene epoxidation is a touchstone of heterogeneous catalysis, the rare example of a direct, selective oxidation of a hydrocarbon by molecular oxygen, and silver remains the only metal that performs it industrially. Recent studies continue to probe particle-size effects, the role of chlorine promoters, and the structure of active oxygen species on silver, and some groups are even exploring whether additives such as nickel can extend selective epoxidation beyond silver itself. All of these efforts depend on well-characterized surfaces, and the Novosibirsk study delivers a practical recipe: avoid damaging the graphite, deposit silver gently on freshly cleaved HOPG, control the mean particle size through deposition conditions, and verify the morphology with machine-learning-assisted microscopy. The result is a family of model catalysts with large, flat, uniformly distributed silver particles and coverage approaching sixty percent of the support, ready for the in situ XPS and scanning tunneling microscopy experiments that may finally pin down how silver turns ethylene into one of the chemical industry’s most valuable building blocks. The work was supported by the Russian Science Foundation under grant 24-63-00037.

Subject of Research: Preparation of silver-on-graphite model catalysts for ethylene epoxidation studies

Article Title: Model Ag/HOPG Catalysis for Ethylene Epoxidation: Comparison of Methods for the Preparation of Large Ag Particles with High Surface Coverage

Article References: Nartova, A. V., Ananina, А. A., Dmitrachkov, A. M., Kvon, R. I., Panafidin, M. A., Andryushechkin, B. V., & Bukhtiyarov, V. I. (2026). Model Ag/HOPG Catalysis for Ethylene Epoxidation: Comparison of Methods for the Preparation of Large Ag Particles with High Surface Coverage. Catalysis Letters, 156(11), Article 295. https://doi.org/10.1007/s10562-026-05538-7

Image Credits: AI Generated

DOI: 10.1007/s10562-026-05538-7

Keywords: silver catalysts, HOPG, ethylene epoxidation, XPS, scanning tunneling microscopy, scanning electron microscopy, nanoparticles, Ostwald ripening, deep learning image analysis, surface chemistry, heterogeneous catalysis, model catalysts