Probing the Coverage Dependence of Site and Adsorbate Configurational Correlations on (111) Surfaces of Late Transition Metals
Zhongnan Xu and John R. Kitchin, “Probing the Coverage Dependence of Site and Adsorbate Configurational Correlations on (111) Surfaces of Late Transition Metals”, The Journal of Physical Chemistry C 118(44), 25597-25602 (2014). https://doi.org/10.1021/jp508805h
This summary was written by a language model from the paper and has not been checked line by line. Rely on the published paper for details.
Note on the source text: the available extraction is the “Just Accepted” manuscript, and the figure pages (and some text that shares those pages) are missing, so several paragraphs begin mid-sentence. This summary covers only what is readable.
Question
Linear scaling relations between adsorption energies let a few DFT calculations stand in for many, and they underpin activity and selectivity analyses for many reactions. Most of them, however, were built at a single low coverage on a single facet, whereas real catalysts operate at high coverage on surfaces with many kinds of sites. The paper asks two related questions: can adsorption on one site predict adsorption on a geometrically different site, and do correlations between different adsorbates survive, unchanged, at higher coverage (Xu 2014).
Approach
The authors computed 896 adsorption energies with VASP and the PBE functional (not spin polarized, so Ni and Co were excluded). The adsorbates were eight atoms, H, C, N, O, F, S, Cl and Br, on the (111) surfaces of seven fcc metals, Cu, Rh, Pd, Ag, Ir, Pt and Au, at 0.25, 0.5, 0.75 and 1.0 ML in fcc, hcp, bridge and ontop sites. Structures that reconstructed substantially or lost adsorbates were excluded. Energies are per adsorbate relative to the isolated gas-phase atom. Only atomic adsorbates were used, to focus on coverage effects transmitted through the surface electronic structure rather than direct through-space interactions; sample DFT+D calculations indicated that dispersion did not change the conclusions. Correlations were assessed with linear regression and 95% confidence intervals on the fitted slopes at each coverage (Xu 2014).
Key findings
Site correlations depend strongly on geometric similarity. Energies at fcc and hcp sites, which differ only in the second metal layer, are strongly correlated for every metal, adsorbate and coverage. Correlations of fcc energies with bridge and ontop energies are weaker, and coverage can shift ontop energies in the opposite direction from fcc energies, notably for some C, N, O and S systems with strongly negative adsorption energies. The authors conclude that higher coverage is a key driver of the breakdown of correlations between geometrically different sites (Xu 2014).
Correlations between adsorbates are coverage dependent in almost every case. Among the halogens (Group 17) the correlations are strong at low coverage but weaken at higher coverage, with wider confidence intervals, and for no adsorbate pair does a single slope fall inside the 95% confidence intervals at all coverages. S and O remain strongly correlated at all coverages, but their slopes diverge at high coverage. Among the Period 2 elements, C-O correlations have somewhat wider confidence intervals than N-C or N-O, which the authors relate to the two-electron valence difference between C and O. Slopes of correlations involving C change monotonically with coverage, and only the N-O correlation admits one common slope within the confidence intervals at all coverages; correlations at 0.5 to 1.0 ML are all shifted from those at 0.25 ML. The authors suggest that halogen correlations break down at high coverage because repulsion from these electronegative species becomes less systematic, in line with known exceptions to the d-band model (Xu 2014).
To connect to scaling relations for hydrogenated species, the authors recast the Abild-Pedersen bond-counting rule for the slope as a ratio of electrons needed to complete an octet, (8 minus the valence electrons of one element) over (8 minus those of the other). At low coverage this predicts the S-O, C-O and N-O slopes very well and the halogen and C-N slopes reasonably well, but the disagreement generally grows as coverage increases (Xu 2014).
Significance
The results imply that activity peaks, activity ceilings and selectivity boundaries derived from low-coverage scaling relations will probably move at realistic coverages, and that poor correlations between dissimilar sites mean particle shape and size could also shift scaling. Because correlations still exist at high coverage (apart from the halogens) and the slope seems to change roughly linearly with coverage, the authors propose building models that predict how scaling relations change with coverage, using tools such as cluster expansions, semi-empirical electronic structure models or Monte Carlo simulation (Xu 2014).