Fig. 1. Known limitations in supported metal particles and reactive gas modulation possibilities described in this work. (a) Subnanometer-sized metal clusters and surface single metal atoms are highly reactive, but the materials are unstable, expensive and difficult to obtain. (b) Under ambient/reaction conditions, small-sized particles can irreversibly aggregate into large crystallites, which are more stable but no longer retain the unique catalytic properties. (c) In this work, we have shown that simple and inexpensive treatment with gas transforms low-activity large metal particles into highly active clusters and atoms. (d) Gas molecules were classified on the basis of their ability to increase/decrease/stabilize nanoparticles of different metals.

Fig. 2. Computational modeling at the atomic level. Calculated migration barriers for Pd atoms with adsorbed molecules for Path-1 (a) and Path-2 (b). Schematic illustration of Path-1 and Path-2 for adsorbed NO, CO, H2 (c), and O2, H2O and pure Pd (d). Path-1 is shown in green, and Path-2 is shown in blue. (e) Visualized charge redistribution for Pd atoms with adsorbed molecules on graphene. The yellow color of the density corresponds to an excess of electrons, whereas the blue color corresponds to electron deficiency. The iso-value is 0.025 e/?3. (f) Calculated COHP diagrams for the studied systems. (g) Atomic structures of Pd38 nanoparticles with adsorbed CO, NO, H2, O2, and H2O molecules on the graphene surface together with charge redistribution between graphene and Pd38. The yellow color of the density corresponds to an excess of electrons, whereas the blue color corresponds to electron deficiency. The iso-value is 0.01 e/?3. (h) Illustration of the migration pathway of Pd38 nanoparticles on graphene. (i) Calculated migration barriers for Pd38 nanoparticles over graphene. (j) Dependence of the migration barrier on the NO concentration adsorbed on the Pd38 nanoparticle.

Fig. 3. Modeling the local environment effect. (a,b) Illustration of the local environment of the (100) (c) and (111) (d) Pd surfaces for which the ICOHP was calculated. Dependence of integrated COHP (a) and bond distance (b) on the type of adsorbed molecule.

Fig. 4. Computational analysis pipeline for palladium nanoparticles on graphite. An automated algorithm for analyzing images obtained with an electron microscope is exemplified with an argon atmosphere. A side-by-side comparison of all images for CO, H2, H2O, NO and O2 is presented in the supplementary information. (a) Overview of the pipeline, including image matching and neural network analysis steps. (b) Examples of two aligned images and matched nanoparticles. (c) Distribution of nanoparticle sizes at different thresholds, showing an average increase in nanoparticle size across the entire threshold range. (d) Relationship between the binarized neural network output and threshold, highlighting common problems in nanoparticle size determination. (e) Dimensionality reduction plot showing how images move within the image vector space during the reaction. (f) Comparison of nanoparticle radii and shifts in two directions. (g) Analysis of nanoparticle distribution throughout the dataset; error bars indicate standard deviations along each axis. (h) Distribution of median sizes across all thresholds in the example images from 'c'.

Fig. 5 Catalytic experiments comparing the activity of the initial and gaseous NO-treated Pd/C catalysts. (a) Modification of a nanoscale catalyst by treatment with gaseous NO. (b) HAADF-STEM image of initial Pd/C (C - graphite); Pd single atoms was not found. (c) HAADF-STEM images of Pd/CNO which shows a large number of single palladium atoms as well as the presence of small clusters. (d) HAADF-STEM image of initial Pd/SiO2; Pd single atoms was not found. (e) HAADF-STEM images of Pd/SiO2NO which shows a large number of single palladium atoms as well as the presence of small clusters. HAADF-STEM images before and after NO treatment were obtained from different areas. (f) Scope and yield comparison of the products of Suzuki-Miyaura cross-coupling reactions catalyzed by Pd/C systems with/without NO treatment. The catalytic activity results revealed an increase in the activity of the NO-treated catalyst in the C?C bond formation reaction. (g) Study of reaction kinetics with Pd/C and Pd/CNO catalysts on the example of 4-nitro-1,1'-biphenyl synthesis.

Fig. 6. Surface dynamic behavior. Atomization/dispersion, motion and disintegration of various metallic nanoparticles in an atmosphere of nitrogen monoxide (a) and nitrosyl chloride (b).