Bimetallic Surfaces
Bimetallic surfaces combine two metals at or near the surface, for example as a monolayer of one metal supported on another, a surface or subsurface alloy, or a sandwich structure with a different metal in the subsurface layer. Their adsorption and catalytic properties can differ strongly from those of either parent metal, which makes them a route to tuning catalysts. The changes are commonly explained through strain and ligand effects acting on the surface d band (see d-band model).
Active Ni/Pt(111) surface and its weak metal-hydrogen bond (2003)
Kitchin, Khan, Barteau, Chen, Yakshinskiy and Madey combined hydrogen TPD, STM, LEIS, Auger and angle-resolved XPS with DFT to identify the active Ni/Pt(111) surface. Ni deposited at 300 K stays on top and binds hydrogen about as strongly as a thick Ni film (H2 desorption near 353 K), whereas Ni deposited at or annealed to about 600 K diffuses below a Pt-enriched surface, and those surfaces release H2 near 234 to 241 K, below clean Pt(111) at 285 K. DFT showed that subsurface Ni (a Pt-Ni-Pt sandwich) is 0.65 eV per Ni atom more stable than a Ni overlayer and binds H more weakly than either Pt or Ni, explaining the experiment (Kitchin et al. 2003). Summary: kitchin2003.
Pt(111) with subsurface 3d metals (2004)
Kitchin, Nørskov, Barteau and Chen modeled Pt-skin structures, a Pt(111) surface layer above a layer of Ti, V, Cr, Mn, Fe, Co or Ni, as an idealization of Pt-rich Pt-3d alloy surfaces. DFT showed that the subsurface metal lowers the surface d-band center and weakens dissociative H2 and O2 adsorption, with the largest changes for early 3d metals. They discussed the consequences for PEM fuel cells: lowering the d-band center can improve CO tolerance at the anode but may make H2 dissociation unfavorable, while weaker O and OH binding at the cathode may help explain reported oxygen reduction enhancements on Pt-Fe, Pt-Ni, Pt3Co and Pt3Ni (Kitchin et al. 2004). Summary: kitchin2004.
Supported Ni, Pd and Pt monolayers (2004)
Kitchin, Nørskov, Barteau and Chen used DFT on pseudomorphic Ni, Pd and Pt monolayers on close-packed facets of Nb, Ta, Mo, W, Re, Ru, Ir, Os, Ni and Pt, together with surface and subsurface alloys and Pt-3d sandwich structures. They found negligible charge transfer (d-band filling changes below about 2%), a d-band width set by a nearest-neighbor matrix element, and a correlation between the d-band center and the dissociative adsorption energy of H2. They proposed these correlations as a way to screen bimetallic systems for desired properties without full first-principles calculations (Kitchin et al. 2004). Summary: kitchin2004a.
Pt skins with weaker oxygen binding for oxygen reduction (2004)
The oxygen reduction volcano of Nørskov, Rossmeisl, Kitchin and coworkers places Pt slightly on the strong-binding side, so surfaces that bind oxygen somewhat more weakly than Pt should be more active. DFT calculations for a Pt monolayer over a monolayer of Ni, Co or Fe on Pt give oxygen binding weakened by roughly 0.3 to 0.5 eV relative to Pt(111), with smaller changes for OH. Together with the subsurface 3d metal results of (Kitchin et al. 2004), this was used to rationalize the higher activity measured for Pt skins on Pt alloys with 3d metals (Nørskov et al. 2004). Summary: norskov2004; see also Oxygen reduction reaction.
O binding on Ag3Pd(111) is not an average of the parent metals (2008)
At 0.25 ML, O binding on Ag3Pd(111) configurations does not interpolate linearly between Ag(111) and Pd(111): some Pd-rich alloy surfaces bind O more strongly than Pd(111) and some Ag-rich ones more weakly than Ag(111), and second-layer composition alone shifts the average binding by about 0.3 to 0.6 eV per O (Kitchin et al. 2008). Third-layer composition matters much less (up to 0.1 eV), so variations in the top two layers capture the essential physics. Summary: kitchin2008.
Isolated Pd atoms in Cu and Au hosts (2009)
Tierney, Baber, Kitchin and Sykes combined STM and DFT to compare single Pd atoms alloyed into Cu(111) and Au(111). Pd alloys exothermically into both surfaces and stays isolated at low coverage, yet the same Pd loading dissociates H2 and spills H onto Cu but shows no H uptake in Au. The authors argue that the host metal can change the thermodynamics and kinetics of a reaction pathway far more than a simple dilution, electronic or geometric effect would suggest (Tierney et al. 2009). Summary: tierney2009. See also Single-atom alloys and Hydrogen spillover.
Which second metals stay at step and kink sites (2009)
Han, Kitchin and Sholl used DFT segregation energies on flat, stepped and chiral kinked fcc surfaces, together with a correlation based on coordination number, bond strength difference and impurity-induced d-band center shift, to predict whether an isolated second metal stays at undercoordinated kink sites or dissolves into the host. Of 347 3d/4d/5d impurity/host pairs with a firm prediction, 206 favor kink decoration. The authors suggest the same reasoning applies to bimetallic nanoparticles, which expose many undercoordinated sites (Han et al. 2009). Summary: han2009; see also Surface segregation.
Screening overlayer and impurity alloys with a d-band width model (2010)
İnoğlu and Kitchin used a DFT-parameterized tight-binding model to estimate surface d-band widths for Pd(111) with Fe, Co, Ni, Ru, Rh, Ag, Ir, Pt or Au surface impurities (average error below 2.5% against DFT) and for all overlayer/host combinations of Fe, Co, Ni, Cu, Ru, Rh, Pd, Ag, Ir, Pt and Au. For a fixed host, overlayers further right or down the periodic table gave wider bands and lower d-band centers, implying weaker binding, with Ag overlayers widest. Alloy thermodynamic stability was not considered, and the approach was proposed for rapid screening, for example of sulfur-tolerant Pd-based membranes (İnoğlu 2010). Summary: inouglu2010.
Large-scale screening of bimetallic motifs with a d-band width model (2011)
This paper applied the solid-state table to four bimetallic motifs on fcc(111) (overlayer, subsurface underlayer, and 0.25 ML surface and subsurface dopants) for all binary pairs of Fe, Ni, Cu, Ru, Rh, Pd, Ag, Ir, Pt and Au. Estimated d-band widths matched DFT for 110 overlayers with mean absolute error below 4.5%, failing mainly for Ag and Au overlayers and reconstructed surfaces (Ir@Fe, Ru@Ni, Ag@Cu) where host-dictated unrelaxed distances are a poor approximation. Combined with segregation energies in a surface free energy, the model flags which alloys are both stable against segregation and more S-tolerant than the parent metal (İnoğlu 2011). Summary: inoglu2011.
Effective adsorption energies from a basis of alloy sites (2015)
Boes, Gumuslu, Miller, Gellman and Kitchin modeled a random CuxPd1-x(111) surface as a weighted mix of the four three-atom fcc hollow sites (Pd3, Pd2Cu, PdCu2, Cu3), each computed with DFT as a function of bulk composition via Vegard’s law and weighted by its probability at the segregated surface composition. The effective H2 adsorption energy matched composition spread film measurements within a band of about 0.08 eV for x > 0.5, whereas neither a single site nor a linear interpolation between the pure metals did; deviations for Pd-rich alloys were attributed to Pd hydride formation (Boes et al. 2015). Summary: boes2015a.
Ordering and core level shifts in Cu-Pd films (2015)
Boes et al. combined EBSD, spatially resolved XPS and DFT on Cu-Pd composition spread films to show that the bulk phase (disordered FCC versus ordered B2) leaves a fingerprint in the Cu 2p3/2 core level shift (Boes et al. 2015). Electron backscatter diffraction confirmed that the roughly 100 nm films follow the bulk phase diagram, and calculations attributed the anomalous shift in the B2 range to fewer Cu-Cu nearest neighbors in the ordered phase. This makes XPS a phase-sensitive probe of the near-surface region of alloy films relevant to Pd-based hydrogen membranes. Summary: boes2015.
Alloy catalysis across continuous composition (2016)
Kitchin and Gellman review how continuous composition libraries expose features of alloy reactivity that a few discrete samples can miss (Kitchin 2016). On CuxPd1-x films, a dip in the Cu core level shift for 0.33 < xPd < 0.55 marked the ordered B2 phase, and the measured H2 adsorption barrier fell with Pd content up to xPd = 0.40 and then became composition independent, possibly due to Pd hydride formation. Electrocatalysis examples from other groups include PdAu films with peak HER/HOR activity near 70 at% Au, attributed to Pd dimer and trimer ensembles, and PdBi films whose intermetallic phase matched Pd activity without forming a hydride. Summary: kitchin2016.
Pt monolayers on fcc(111) substrates analyzed by interpretable deep learning (2024)
Huang, Wang, Kamanuru, Achenie, Kitchin and Xin predicted d-band centers of sites on alloyed, strained and differently terminated fcc metal surfaces with a theory-infused graph neural network and examined Pt monolayers on fcc(111) metal substrates as a subgroup. Strain is the dominant contribution in this series, and a resonance contribution tied to the substrate’s bulk d-band center also appears; Pt on Re shows the most skewed d band among 5d substrates (Huang et al. 2024). Summary: huang2024.