At low temperatures (<400 °C), single atoms of Rh supported on rutile TiO2 (rTiO2) are responsible for the formation of CO during the reverse water gas shift (RWGS), while methane production is associated with the Rh-TiO2 interface due to the observed correlation between methane formation rates and the volume-averaged Rh nanoparticle diameter. As the temperature is increased to >540 °C, there is a notable increase in CO selectivity as the methane production rates tend towards zero. The time to reach zero depends on the temperature but is independent of the initial Rh structure (single atoms and/or nanoparticles), which is controlled by the catalyst preparation method (wetness impregnation versus colloidal nanoparticles). At 600 °C and >4 h time on stream, the catalytic behaviour becomes completely agnostic to the initial Rh structure as well as weight loading, and the catalysts are highly selective for the RWGS reaction. Post-reaction HR-TEM image analysis confirms Rh nanoparticles crystallize/order during the reaction; at 400 °C, most of the Rh particles are disordered, while at 600 °C, they are more ordered (i.e., there is the development of defined faceting). Infrared spectroscopy of CO adsorption on Rh nanoparticles confirms the appearance of defined facets after annealing in nitrogen at high temperatures. Annealing the Rh/rTiO2 catalysts prior to the RWGS reaction demonstrates the structural transformation of Rh depends only on time and temperature and not on reactant or product fugacity. Sites responsible for stabilizing Rh single atoms are no longer competent at higher temperatures, enabling single atom integration into existent nanoparticles. As the reaction temperature is increased to temperatures >540 °C, the dominant Rh structure for CO production evolves from single atoms to extended surfaces.

While in situ high-resolution transmission electron microscopy (HRTEM) allows atomic-resolution observation of dynamic processes such as chemical reactions and phase transformations, quantitative analysis of such data is often hindered by its complexity and volume that make manual labeling or traditional segmentation methods ineffective. Here, we demonstrate a data-efficient deep learning workflow using a standard U-Net architecture to segment atomic positions from in situ HRTEM images of copper oxides on copper substrates and gas interfaces-a system with challenging heterogeneous backgrounds and complex interfaces. The workflow is optimized for small datasets, achieving high accuracy after being trained on fewer than 50 manually labeled images augmented with rotation, scaling, and Gaussian filtering. The resulting U-Net model achieved test accuracies of approximately 92% for bulk atoms and 63% for interfacial atoms on images similar to the training dataset, with most misclassifications occurring at the complex interfacial regions. Moreover, our approach offers an approximately 300,000-fold improvement in labeling speed compared to manual labeling. This work presents a practical and adaptable framework for accelerating quantitative analysis of experimental systems for which acquiring large, annotated datasets is prohibitive, demonstrating the effective application of U-Net for a common challenge in electron microscopy.

Autoencoder-transformer model applicable to both simulated and experimental XAS spectra is developed to predict the oxidation state of copper.

The macroscopic properties of oxide films are profoundly influenced by their formation during metal oxidation, yet the atomic-scale mechanisms governing these processes remain elusive. Using in situ environmental transmission electron microscopy (ETEM), we observe the real-time transitions from lateral oxide growth on bare metal surfaces to inward growth along the oxide-metal interface, demonstrating that atomic-scale surface steps critically regulate oxide film growth. Positive Ni steps, located above the NiO-Ni interface, facilitate lateral NiO growth by promoting surface adatom transport. Conversely, negative Ni steps, situated below the interface, redirect growth inward along the NiO-Ni interface. Atomistic simulations illuminate that positive step edges create asymmetric energy barriers favoring surface diffusion for lateral growth, whereas negative steps restrict adatom diffusion, driving inward interfacial growth. These insights highlight the critical influence of surface steps on oxidation mechanisms, offering a pathway for engineering metal surfaces to control oxide film formation and tailor macroscopic properties.

X-ray absorption spectroscopy (XAS) and electron energy-loss spectroscopy (EELS) produce detailed information about oxidation state, bonding, and coordination, making them essential for quantitative studies of redox and structure in functional materials. However, high-throughput quantitative analysis of these spectra, especially for mixed valence materials, remains challenging as diverse experimental conditions introduce noise, misalignment, broadening of the spectral features. We address this challenge by training a machine learning model consisting of an autoencoder to standardize the spectra and a transformer model to predict both Cu oxidation state and Bader charge directly from L-edge spectra. The model is trained on a large dataset of FEFF-simulated spectra and evaluates model performance on both simulated and experimental data. The results of the machine learning model exhibit highly accurate prediction across the domains of simulated and experimental XAS as well as experimental EELS. These advances enable future quantitative analysis of Cu redox processes under in situ and operando conditions.

Solid-melt interfaces govern metal-mediated epitaxy of gallium nitride (GaN) and related III-nitride semiconductors, yet a quantitative connection between interfacial structure and growth kinetics has remained elusive. Using atomistic simulations based on machine-learning interatomic potentials, we show that crystal orientation determines the local atomic arrangement and electrostatic character at the interface, inducing facet-dependent ordering in the adjacent liquid melt. This ordering reshapes the N-transport free-energy landscape, substantially lowering migration barriers relative to pristine surfaces. Kinetic Monte Carlo simulations reveal that under realistic Ga-mediated molecular beam epitaxy (MBE) conditions, GaN grows in a diffusion-controlled, layer-by-layer regime, where the interface-normal adsorption free-energy landscape sets the dominant kinetic resistance. Combining facet-specific free-energy profiles with N diffusivity and solubility in liquid Ga, we develop a fitting-free model predicting growth rates of 0.01–0.04 nm/s, consistent with MBE-grown GaN nanoparticles. Similar ordering in other III-nitride systems indicates a generalizable route from atomistic energetics to anisotropic crystal growth kinetics.

The transition to hydrogen as a green reductant in metal production is critical for decarbonizing the metallurgical industry, yet atomic-scale mechanisms governing reduction pathways and phase evolution remain unresolved. Using in situ environmental transmission electron microscopy, we identify a hidden pathway that reveals dynamic formation of amorphous metallic iron (Fe) during the hydrogen-driven reduction of ferrous oxides of Fe3O4 and FeO. Real-time imaging uncovers three coexisting transformation routes: (i) Fe3O4 → FeO, (ii) Fe3O4 → amorphous Fe, and (iii) FeO → amorphous Fe. The resulting amorphous Fe exhibits fluid-like mobility, enabling its rapid aggregation and crystallization into core-shell nanostructures, with a crystalline core enveloped by an amorphous shell. Complementary ab initio molecular dynamics simulations trace the amorphous Fe formation to interfacial strain at the metal/oxide interfaces, where large lattice mismatches destabilize the metal lattice during initial metallization. This interplay between thermodynamics and kinetics governs phase evolution: thermodynamics favors a self-limiting amorphous Fe overlayer, while rapid oxide reduction kinetics drives amorphous overgrowth. Our findings demonstrate that amorphous intermediates bypass rate-limiting crystalline steps, providing mechanistic insights to optimize H2-based processes for sustainable steelmaking. These insights bridge the gap between macroscopic process engineering and atomic-scale dynamics, with broader implications for catalysis and nanostructured material synthesis, where oxide reduction pathways critically shape functional phases and microstructures.

Titanium dioxide (TiO2) is widely used as a catalyst support due to its stability, tunable electronic properties, and surface oxygen vacancies, which are crucial for catalytic processes such as the reverse water-gas shift (RWGS) reaction. Reduced TiO2 surfaces undergo complex surface reconstructions that endow unique properties but are computationally challenging to describe. In this study, we utilize machine-learning interatomic potentials (MLIPs) integrated with an active-learning workflow to efficiently explore reduced rutile TiO2 surfaces. This approach enabled the prediction of a phase diagram as a function of oxygen chemical potential, revealing a variety of reconstructed phases, including a previously unreported subsurface shear plane structure. We further investigate the electronic properties of these surfaces and validate our results by comparing experimental and theoretical high-resolution transmission electron microscopy (HRTEM). Our findings provide new insights into how extreme surface reductions influence the structural and electronic properties of TiO2, with potential implications for catalyst design.

The electrocatalytic processes of a copper catalyst during nitrate electroreduction are unveiled by correlated operando microscopy and spectroscopy.

Alloying plays a crucial role in tuning the surface properties of metals, but the atomic-level mechanisms by which alloying elements influence surface structure dynamics under reactive conditions remain elusive. Using Cu(Au) in oxidizing environments as a model system, we reveal a dynamic oxygen-induced transformation of the topmost atomic layer into a periodically hill-and-valley morphology with reversible switching between undulated and flattened surface states. These interconversions are driven by the retreat of surface Au to the subsurface during oxygen adsorption and its resegregation to the surface upon oxygen desorption. This cyclical mobility establishes a feedback loop, allowing the surface to dynamically reconfigure in response to changes in the oxygen pressure. The results offer a broadly applicable framework for understanding atomic-scale surface restructuring in alloy systems, where differences in the chemical reactivity of alloying elements drive dynamic redistribution between surface and subsurface regions. This dynamic coupling has practical implications for designing corrosion-resistant coatings and metastable nanostructures with tunable catalytic properties.

The breaking of translational symmetry at oxide surfaces gives rise to coordinatively unsaturated cations/anions and surface restructuring-key factors that govern surface reactivity. Using direct in situ environmental transmission electron microscopy (TEM) observations along with atomistic modeling, we report oscillatory redox behavior in CuO under H2, where cyclic surface reconstruction and reactivity modulation occur via the Mars-van Krevelen (MvK) mechanism. We observe self-switching between oxygen-rich and oxygen-deficient surface reconstructions, alternately activating and deactivating the surface for H2O formation. During periods of chemical inactivity, the oxygen-deficient surface undergoes slow reoxidation via lattice oxygen diffusing from subsurface and bulk reservoirs, restoring the active oxygen-rich surface termination. The inherent disparity in chemical activity among undercoordinated surface ions, along with sluggish subsurface-to-surface oxygen replenishment, drives this oscillatory redox cycle, modulating H2-induced loss of lattice oxygen at the surface and its delayed replenishment from the subsurface. This creates spatiotemporally separated redox steps at the oxide surface. The phenomena and atomistic insights presented here have significant implications for manipulating the surface reactivity of oxides by tuning the separation of these redox steps.

Characterizing catalyst stability by identifying the predominant mechanisms, time scales, and driving forces of catalyst reconstruction under relevant reaction conditions is necessary for the design and commercialization of emerging catalysts. Here, we study Rh/TiO2 catalysts under CO2 hydrogenation conditions (773 K, 75% H-2, 25% CO2) at 10-50% initial CO2 conversion and utilize reactivity studies along with ex situ and in situ spectroscopy and microscopy to characterize changes in catalyst activity and structure as a function of time on stream and the initial Rh domain structure. Rh/TiO2 is a prototypical catalyst for CO2 hydrogenation where Rh structure and Rh-TiO2 interactions have been proposed to explain reactivity, selectivity (between CO and CH4 formation), and catalyst stability. Here, the influences of the initial Rh structure (varying from Rh single atoms to Rh nanoparticles), support reconstruction, regeneration and pretreatment, and the composition of the reaction environment on reaction selectivity and catalyst stability were explored. Product selectivity between CO and CH4 was determined to be dependent on the relative fraction of Rh single atoms and Rh nanoparticle-TiO2 interfacial sites under the reaction conditions, each exhibiting distinct stability under the explored reaction conditions. Surprisingly, Rh single atom active sites were stable for the duration of 90 h reactivity measurements, even at high Rh surface density (>= 1.8 Rh atoms/nm(2)) on the support, while Rh nanoparticles sintered. All catalysts exhibited increasing selectivity to CO with time on stream (>10 h). We conclude that the distribution of Rh structures evolved under reaction conditions through three distinct mechanisms (Rh particle fragmentation, Ostwald ripening, and particle migration and coalescence) that occurred on varying time scales and that changes in catalyst reactivity on the similar to 90 h time scale were primarily controlled by the distribution and density of initial Rh structures.

Surface characterization at the atomic scale is essential for understanding the catalytic properties of supported metal nanoparticles. Secondary electron (SE) imaging in scanning transmission electron microscopy (STEM) provides three-dimensional surface topographic information, enabling the characterization of the size, morphology, and distribution of supported nanoparticles. Furthermore, real-time observation of catalyst materials in a gaseous environment would enhance the understanding of catalyst dynamics under operational conditions. Ongoing technical developments in SE-STEM and advancements in computational methods are expected to facilitate atomic-scale surface observations and enable more quantitative and statistical analyses. This progress will not only elucidate fundamental mechanisms at the atomic level but also provide comprehensive and universal insights into catalyst performances. This minireview showcases the recent advancements and research findings in surface-sensitive SE imaging in STEM for the characterization of active catalyst materials.

Solid-melt interfaces play a pivotal role in governing crystal growth and metal-mediated epitaxy of gallium nitride (GaN) and other semiconductor materials. Using atomistic simulations based on machine-learning interatomic potentials (MLIPs), we uncover that multiple layers of Ga atoms at the GaN-Ga melt interface form structurally ordered and electronically charged configurations that are critical for the growth kinetics of GaN. These ordered layers modulate the free energy landscape (FEL) for N adsorption and substantially reduce the migration barriers for N at the interface compared to a clean GaN surface. Leveraging these interfacial energetics, kinetic Monte Carlo (KMC) simulations reveal that GaN growth follows a diffusion-controlled, layer-by-layer mechanism, with the FEL for N adsorption emerging as the rate-limiting factor. By incorporating facet-specific FELs and the diffusivity/solubility of N in Ga melt, we develop a predictive, fitting-free transport model that estimates facet-dependent growth rates in the range of 0.01 to 0.04 nm/s, in agreement with experimental growth rates observed in GaN nanoparticles synthesized by Ga-mediated molecular beam epitaxy (MBE). This multiscale framework offers a generalizable and quantitative approach to link atomic-scale ordering and interfacial energetics to macroscopic phenomena, providing actionable insights for the rational design of metal-mediated epitaxial processes.