Estimating Bulk-Composition-Dependent H2 Adsorption Energies on CuxPd1-x Alloy (111) Surfaces
J. R. Boes, G. Gumuslu, J. B. Miller, A. J. Gellman and J. R. Kitchin, “Estimating Bulk-Composition-Dependent H2 Adsorption Energies on CuxPd1-x Alloy (111) Surfaces”, ACS Catalysis 5(2), 1020-1026 (2015). https://doi.org/10.1021/cs501585k
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.
Question
DFT can readily predict how strongly a well-defined site binds an adsorbate, but a real alloy surface presents a distribution of sites of different local composition, and its surface composition differs from the bulk because of segregation, which itself can change in the reacting gas. Rigorous treatments (ab initio thermodynamics, or cluster expansions with Monte Carlo) are expensive. The paper asks whether a simpler model can predict the effective dissociative H2 adsorption energy of a CuxPd1-x alloy as a continuous function of bulk composition, in a form that can be compared directly with measurements on a composition spread alloy film (Boes et al. 2015).
Approach
On the experimental side, H2-D2 exchange kinetics were measured at 100 points of a CuPd composition spread alloy film (about 100 nm thick, x from 0.3 to 1.0 on a Mo substrate, annealed at 800 K) in a multichannel microreactor between 300 and 600 K. A two-step microkinetic model gave adsorption and desorption barriers, and their difference was taken as the H2 adsorption energy, using a linear fit of the desorption barrier because it is poorly determined above x of about 0.8.
The model has four ingredients. First, a basis of active sites: hydrogen prefers the fcc hollow on both pure metals, so the authors take the four distinct three-atom fcc sites on an fcc(111) surface (Pd3, Pd2Cu, PdCu2, Cu3), neglecting the B2 phase that is stable near x = 0.51 to 0.68 at 800 K. Second, each site is embedded in a four-layer pure Cu slab and in a pure Pd slab (VASP, PBE) to bracket the ligand effect, and the slab lattice constant is varied; Vegard’s law, which a cluster expansion of CuPd ground states confirms, maps lattice constant to bulk composition, and a quadratic fit of adsorption energy versus lattice constant gives each site’s energy as a function of x. Third, a random surface gives the probability of each site at a surface Cu fraction y, with mixed sites weighted by their three rotations. Fourth, y is found from the Langmuir-McLean form of the Gibbs isotherm, using experimental vacuum segregation energies from low energy ion scattering plus an adsorbate term equal to the hydrogen coverage (from a Langmuir isotherm over the site basis) times the difference in H adsorption energy between Cu(111) and Pd(111) (Boes et al. 2015).
Key findings
The measured adsorption energy was roughly constant near -0.3 eV on Cu-rich films (with large uncertainty), became more negative as Pd content rose, and then turned upward again below x of about 0.6. In the calculations, embedding sites in Pd rather than Cu made adsorption more favorable, but by less than about 0.05 eV, so the trends with lattice constant were nearly the same for both hosts. Under ultrahigh vacuum the experimental segregation data put more Cu in the top layer than in the bulk at all compositions, whereas at 800 K and 1 atm H2 the model predicts a substantial enrichment of Pd at the surface, driven by the stronger binding of hydrogen to Pd-rich sites.
The resulting effective adsorption energies agree with experiment for x > 0.5, where the data lie within a predicted band about 0.08 eV wide; the Pd-embedded estimates match better, suggesting Pd ligand effects matter. Neither a single site nor a linear interpolation between the pure metals describes the data, and the model without segregation also fails, which the authors take as evidence for adsorbate-induced segregation. For x < 0.4 the data leave the predicted band. The authors attribute this to a Pd hydride phase: the measured value for alpha-PdH (-0.56 eV/atom) falls inside the predicted bounds, while for beta-PdH (measured -0.3 eV/atom) their calculations on a Pd-terminated rock salt slab give -0.327 eV/atom (hcp) and -0.283 eV/atom (fcc), and the experimental trend at high Pd content moves toward this weaker value. They note the segregation driving force is the simplest possible form, ignores strain differences between sites, and therefore overpredicts favorable adsorption; coverage effects above 0.25 ML could weaken adsorption by 0.05 to 0.1 eV (Boes et al. 2015).
Significance
The work offers an inexpensive route from a handful of DFT calculations to composition-dependent alloy reactivity that includes segregation under reaction conditions, and it pairs that prediction with high-throughput experiments across the same composition range. Disagreement between model and measurement is also used diagnostically, here to point to hydride formation in Pd-rich alloys (Boes et al. 2015).