ИСТИНА

Pd supported catalysts are widely accepted as the most active for combustion of methane [1-2] and low temperature oxidation of CO [3]. The catalytic cycle involving a Pd/PdO pair involves a quick Pd oxidation step forming an inert Pd(O*) species which transforms slowly to more active PdO bulk oxide [4]. However, the consequent Pd oxidation mechanism and its transformation into PdO bulk oxide remains a subject of discussions [3, 4]. Herein, we studied the atomic O diffusion from surface to bulk starting with Pd(100) models which were arranged as slabs containing from 2 to 6 metal layers over a γ-Al2O3 support (Fig. 1a). The computations at the PAW-PBE level with VASP [5-6] demonstrated that the average atomic charge in the metallic layers above the first layer in contact with the oxide surface quickly converges to nearly zero values and are not sensitive to the surface γ-Al2O3 defects studied earlier [7] whose interaction can charge positively or negatively the first Pd layer. The 3-layer Pd slab model (dashed line in Fig. 1a) without the oxide slab and first contact Pd layer (with “frozen” lowest layer) was thus used as a reasonable compromise (Fig. 1b, c) which reflects the structural perturbation produced by the oxide at the external Pd surface. This allowed less costly computations of the complex supported catalysts. Both isolated O atoms and peroxo-species O2 (4, 9, and 16 atoms in total) were deposited at the external Pd(100) slab with a surface area of the unit cell ~100 Å2 . Cases of adsorbed (Fig. 1c) and chemisorbed O2 molecules were studied nearby the O surface atoms. The presence of both types of adsorbates facilitates the transition of O atoms from the surface to the first Pd layer, i.e., the change of enthalpy along the O-jump becomes negative: -0.411 (adsorbed O2) and -0.615 eV (chemisorbed O2) versus 0.555 eV without additional O2. In order to study the O jump to the second layer (for three layers in total), the full Pd(O*)/γ-Al2O3 (Fig. 1a) was further considered to avoid energy losses owing to “frozen” lowest layer in the small Pd(O*) model. The results are compared to known data relative to O diffusion in Pd [8]. 1. W.R. Schwartz, L.D. Pfefferle, J. Phys. Chem. C 116 (2012) 8571−857; 2. P. Stefanov, S. Todorova, et al. Chem. Eng. J. 266 (2015) 329–338; 3. Z. Duan, G. Henkelman, ACS Catal. 4 (2014) 3435−3443; 4. Y.-H. Chin, M. Garcia-Dieguez, E. Iglesia, J. Phys. Chem. C. 120 (2016) 1446–1460; 5. G. Kresse, J. Hafner, Phys. Rev. B. 47 (1993) 558–561; 6. G. Kresse, J. Furthmüller, Phys. Rev. B. 54 (1996) 11169–11186; 7. A.A. Rybakov, A.V. Larin, et al. Theor. Chem. Acc., 135 (2016) 152; 8. J. Gegner, G. Hörz, R. Kirchheim, J. Mater. Sci. 44 (2009) 2198–2205; 9. V. Sadovnichy, A. Tikhonravov, V. Voevodin, V. Opanasenko, “Lomonosov”: Supercomputing at Moscow State University. In Contemporary High Performance Computing: From Petascale toward Exascale; Vetter, J. S., Ed.; CRC Press: Boca Raton, USA, 2013; 283–307.