Trends in the Chemical Properties of Early Transition Metal Carbide Surfaces: A Density Functional Study
John R. Kitchin, Jens K. Nørskov, Mark A. Barteau, and Jingguang G. Chen, “Trends in the chemical properties of early transition metal carbide surfaces: A density functional study”, Catalysis Today 105(1), 66-73 (2005). https://doi.org/10.1016/j.cattod.2005.04.008
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
Early transition metal carbides are extremely hard, refractory materials that have attracted interest as catalysts since Levy and Boudart reported “platinum-like” behavior of tungsten carbide, which they attributed to carbon altering the electron distribution of the metal. This paper asks how incorporating carbon changes the electronic structure of the parent metal surface, whether that change makes the metal more Pt-like, and whether the d-band center model that the authors had used for H adsorption on bimetallic surfaces also describes carbide surfaces (Kitchin et al. 2005).
Approach
All calculations used DACAPO with ultrasoft pseudopotentials and the PW91 functional. Bulk β-Mo2C (orthorhombic, roughly an hcp Mo sublattice with carbon in half of the octahedral holes) was optimized at a 350 eV cutoff, and its closest-packed “(0001)” surface was modeled with Mo- and C-terminated slabs two unit cells thick with a dipole correction. The rocksalt monocarbides TiC, VC, NbC and TaC were treated at 400 eV, with lattice constants and bulk moduli from a Murnaghan equation of state, and their (111) surfaces modeled as four metal-carbon layers in metal- and carbon-terminated forms. Atom-projected densities of states were computed with two projection radii: a short 1 Å cutoff and an unrestricted (“infinite”) one. Atomic hydrogen binding energies (HBE) on threefold sites served as the probe of surface reactivity (Kitchin et al. 2005). Note that minus signs appear to have been lost in the extracted text (for example in band energies and interlayer relaxations), and a few table entries are garbled, so only clearly readable numbers are reported here.
Key findings
Computed β-Mo2C lattice parameters were 1.7 to 1.8% larger than experiment, similar to the errors for pure Mo and Pt. Rocksalt carbide lattice constants (for example 4.332 Å versus 4.328 Å for TiC) agreed closely with experiment, bulk moduli were typically within about 5% (the VC comparison, against a polycrystalline estimate, was off by 13%), and the trend in heats of formation was reproduced. On the metal-terminated (111) surfaces the contraction of the first interlayer spacing was always predicted qualitatively, and quantitatively for NbC and TaC, while it was underestimated for TiC and overestimated for VC; the authors attribute the VC discrepancy to the substoichiometric VC0.80 sample used experimentally (Kitchin et al. 2005).
The Mo d band of both β-Mo2C terminations is broader than that of Mo(110), with new low-lying features traced to hybridization of Mo d states with carbon s states (from subsurface and, for the C-terminated surface, surface carbon) and a shoulder from Mo d and carbon p mixing. The integrated d-electron count per surface Mo atom was essentially unchanged: 6.47 for Mo(110), 6.45 for Mo-terminated and 6.69 for C-terminated β-Mo2C, compared with 9.26 for Pt(111). Carbon therefore does not make Mo Pt-like by donating electrons to the d band, at least within this orbital population analysis (Kitchin et al. 2005).
Contrary to the authors’ expectation, H bound more strongly to Mo-terminated β-Mo2C (3.19 to 3.32 eV depending on site) than to Mo(110) (3.04 eV) or Pt(111) (2.66 eV), and sites with a carbon directly below were unstable. The C-terminated surface bound H more weakly (2.83 to 3.00 eV), but one site converted to a C-H species bound by 3.36 eV. On the rocksalt carbides, H likewise bound more strongly to the metal-terminated surfaces than to the pure metal, except for VC, which the authors could not explain. They attribute the stronger binding to tensile strain: metal-metal distances in the carbides are 0.17 to 0.33 Å longer than in the parent metals (for example 2.89 Å in Ti versus 3.06 Å in TiC). On every C-terminated rocksalt surface, H formed a very stable C-H bond (3.92 eV on TaC(111)) rather than occupying threefold sites (Kitchin et al. 2005).
The HBE correlated poorly with the d-band center from the infinite projection, which includes deep carbon-derived states that pull the center down without affecting H bonding. With the 1 Å projection, HBE followed the d-band center trend, though with more scatter than for bimetallic surfaces, which the authors ascribe to band-shape distortions neglected by the simplest d-band model and to possible direct H-C interactions (Kitchin et al. 2005).
Significance
The study shows that DFT reproduces bulk and surface structural properties of carbides well, and that their chemistry is not simply Pt-like: metal-terminated carbide surfaces are more reactive than the parent metals, and only surface carbon makes them less so. This points to a passivating role of surface carbon, to the importance of defect chemistry (different levels of carbon coverage at defects), and to surface carbon atoms as possible active participants rather than mere site blockers. It also extends the d-band model beyond bimetallics while flagging its sensitivity to how the density of states is partitioned (Kitchin et al. 2005).