Departures from the Adsorption Energy Scaling Relations for Metal

Article pubs.acs.org/JPCC

Departures from the Adsorption Energy Scaling Relations for Metal Carbide Catalysts Ronald Michalsky,† Yin-Jia Zhang,‡ Andrew J. Medford,§ and Andrew A. Peterson*,† †

School of Engineering and ‡Department of Chemistry, Brown University, Providence, Rhode Island 02912, United States § Department of Chemical Engineering, Stanford University, Stanford, California 94305, United States S Supporting Information *

ABSTRACT: The activity of heterogeneous catalysts is often limited by a strong correlation between the chemisorption energies of reaction intermediates described by the “scaling relations” among the transition metals. We present electronic structure calculations that suggest that metal carbides do not in general follow the transition-metal scaling relations and tend to exhibit a carbophobic departure relative to the transition metals, meaning they tend to bind carbon-bound species weakly compared to oxygen-bound species. This contrasts with the oxophobic departure exhibited by Pt and Pd. Relative to the parent metals, carbides tend to bind carbon and oxygen more weakly and hydrogen more strongly. The departures are rationalized with the adsorbate−surface valence configuration and the energy of the metal sp-states. We employ these general trends to aid in the understanding of various catalytic properties such as the high activity of iron carbides for Fischer−Tropsch synthesis and Pt-group catalysts for partial oxidation of methane. These conclusions are shown to extend beyond atomic probe adsorbates to molecular fragments of relevance to catalysis, making these concepts generally useful for the theory-based design of catalytic materials.

■

requirement,20−22 as even the best Pt-group metals miss perfect catalytic activity by ∼230 mV.20 Correlated adsorption energies are suggested to also limit the synthesis of methanol23 and ammonia24 on monometallic catalysts and have motivated the design of bimetallic22,25 or metal compound5,24,26 catalysts. The metal carbides are an interesting class of catalytic materials27−41 on which the scaling relations are relatively unexplored, as the scaling relations were originally developed for materials that tend to follow the d-band theory of adsorption.42 Carbide catalysts have received considerable attention starting with the suggestion that the carbided form of otherwise non-noble metals such as Mo and W exhibits certain catalytic activities comparable to noble metals;27 in particular, tungsten carbide has been studied as an alternative to Pt due to its H2 dissociation characteristics.27,28,43 Similarly, insertion of carbon into Fe or Co yields carbide phases with increased or decreased activity, respectively, for Fischer− Tropsch synthesis,29 which may be due to effects such as the experimentally confirmed weakening of CO binding relative to the parent metal30 or the enhancement of CO dissociation over CH4 liberation on vacancy versus surface carbon sites.31 Such characteristics explain the interest in carbide catalysts for the water−gas-shift reaction,32 the reduction of CO2 or CO,33,34 Mars−van Krevelen-like Fischer−Tropsch synthesis (with carbon vacancy formation and replenishment),29,35 fuel hydroprocessing,36 or steam reforming of methane.38

INTRODUCTION The catalytic activity of a surface is largely dictated by the forming and breaking of chemical bonds between the surface and the reaction intermediates. The strength of such bonds, quantified as adsorption energies, can be estimated experimentally with surface science techniques1 or computationally via electronic structure methods.2,3 In practice, adsorption energies of similar adsorbates tend to be highly correlated across surfaces, a phenomenon often referred to as “the scaling relations”.3−5 In fact, correlated adsorption energies between reaction intermediatesand of transition states through Brønsted-Evans−Polanyi relationships6−9are believed to limit the activity of a number of industrially interesting catalysts.10−12 We can readily see an example in the electrochemical reduction of CO2 to CO. If Pt is used as an electrocatalyst, the reaction intermediates COOH and CO are bound strongly, enabling the low-overpotential activation of CO2 to adsorbed CO; however, the CO binds too strongly to the Pt surface, slowing desorption. To make CO liberation more facile, the electrocatalyst can be changed to Au, which binds CO weakly. However, because the binding energies are correlated, Au also binds COOH relatively weakly; this results in the requirement of an overpotential to produce the weakly bound COOH.13−15 Similarly, in the electrochemical reduction of CO2 to CH4 in which the protonation of adsorbed CO is suggested to be the potential-limiting step,16−19 the CO/CHO correlation is believed to be responsible for the dearth of efficient electrocatalysts.18 In the oxygen reduction and evolution reactions, the correlation of O/OH or OH/OOH adsorption energies is considered the origin of the overpotential © 2014 American Chemical Society

Received: April 16, 2014 Revised: May 22, 2014 Published: June 5, 2014 13026

dx.doi.org/10.1021/jp503756g | J. Phys. Chem. C 2014, 118, 13026−13034

The Journal of Physical Chemistry C

Article

Table 1. Calculated and Experimental Lattice Constants calculatedb

a

a

carbide

crystal structure

a (Å)

TiC TaC Fe3C Mo2C WC

cubic (225) cubic (225) orthorhombic (62) orthorhombic (60) hexagonal (187)

4.351 4.488 5.119 4.826 2.931

b (Å)

6.774 6.160

experimental c (Å)

a (Å)

4.547 5.317 2.857

4.328 4.454 5.090 4.729 2.906

b (Å)

6.748 6.028

c (Å)

ref

4.523 5.197 2.837

49 49 57 58 59

Space group in parentheses. bLattice constants optimized via GGA are comparable to previous calculations.34,49

Figure 1. Transition-metal carbide surfaces (stable at ambient conditions; surface energies in J m−2; cn is the average coordination number of a surface metal). Geometric details are given in the Supporting Information.

pseudopotential code DACAPO with atomic manipulations handled in the Atomic Simulation Environment (ASE).51−55 Exchange-correlation interactions were treated by the revised Perdew−Burke−Ernzerhof (RPBE) functional of Hammer, Hansen, and Nørskov56 derived in the generalized gradient approximation (GGA). A k-point sampling of 4 × 4 × 4 (for bulk calculations with periodic boundary conditions in all directions) or 4 × 4 × 1 (for surfaces periodically repeated in the directions parallel to the surface) was used for sampling of the Brillouin zone. A Fermi−Dirac smearing of 0.1 eV was used to achieve convergence, and results were extrapolated to 0 K. Calculations for Fe3C were spin-polarized. The linesearch BFGS algorithm within ASE was employed for relaxing atomic geometries until the maximum force on any unconstrained atom was less than 0.05 eV Å−1. The lattice parameters of the studied carbides were calculated by minimizing the total energy of the bulk carbides as a function of lattice constant. The results are shown in Table 1, which are within 2.5% of the experimental values, similar to the accuracy standardly seen for DFT-calculated transition metal lattice constants. All surface calculations were performed on periodically repeated slabs with 2 × 2 × 4 metal (i.e., noncarbon) atoms in the x, y, or z direction (except Fe3C with 2 × 3 × 4 metal atoms). To avoid reminiscent stress in the adsorption calculations, the lower two layers were constrained to the DFT-calculated bulk geometry, while the atoms in the upper two layers were unconstrained. Twelve carbide surfaces with

Motivated by the distinct adsorption trends calculated on Mo2C relative to the transition metals,34 in the current work we present an electronic structure analysis of a class of representative carbide surfaces to understand adsorption trends on carbides as well as their relationship to adsorption energies on pure transition metals. While many phenomena, such as reconstruction, surface reduction and oxidation, spectator species, and size effects, will affect real applications, we strive to capture the trends under realistic conditions by examining adsorption trends for both atomic and molecular probe adsorbates on a number of different stoichiometric carbide surfaces as well as under expected reactive conditions, such as with carbon vacancies,29,35 with oxygen adatoms,34,44,45 and as oxycarbides.32 The utility of these trends is demonstrated for understanding adsorption trends of oxygenated or hydrogenated molecular fragments that are relevant to the heterogeneous33,46or electrochemical13,14,18,47,48 reduction of CO2, including strengthened adsorption of CHO relative to CO, a useful design principle for the development of electrocatalysts for CO2 reduction. The origin of these chemisorption trends49,50 is discussed in terms of surface geometry (bond valence) and electronic structure (splitting surface metal sp states).

■

COMPUTATIONAL METHODS The electronic structure calculations were performed in density functional theory (DFT) with the open-source planewave/ 13027



Article

metal- or mixed-termination were selected based on experimental and theoretical work confirming their stability near ambient conditions;31,34,43,49,60 these structures are shown in Figure 1. The surfaces were modeled using 10 Å of vacuum in the direction perpendicular to the surface, and the electrostatic interaction between the periodic slabs was decoupled by introducing a dipole layer in the vacuum between the slabs. To directly compare the properties of metal carbide surfaces to transition-metal surfaces, the adsorption energy, EB[i], of the adsorbate i was calculated as ΔE B[i] ≡ E[i + surface] − E[surface] − Eref [i]

(1)

where E[i+surface], E[surface], and Eref[i] are the total electronic energy of the surface with the adsorbate at the favored adsorption site, the surface without the adsorbate, and an element-specific reference for the adsorbate in the gas phase, respectively (detailed in the Supporting Information). The optimized adsorption energies were determined from calculations of adsorption at all nonsymmetric adsorption sites for a given surface and given surface conditions. The computational methods were verified by reproducing the adsorption energies of C or O on W(210) or Mo(210) from the literature3,34,61 within an uncertainty of ±0.05 eV. Utilizing eq 1, the adsorption energies of 1/4 monolayer (ML, defined as number of adsorbates per surface metal atom) or 1/6 ML for Fe3C(001) H, C, and O were computed on stoichiometric carbide surfaces in the presence or absence of 25−50% lattice carbon vacancies (VC) or 1/4 to 3/4 ML O adatoms (MLO). To determine the sensitivity of molecular adsorption energies to the adsorption energy of the atomic constituent that is bonding to the surface, the adsorption energies of CH3, CH, CO, CHO (formyl radical), COOH (carboxyl radical), OCHO (formate), OH, and H2O were calculated as well. Details on local optimization procedures, reference energies, surface coverage, reference metal surfaces and a comprehensive collection of the computed adsorption energies, optimized geometries, and electronic structure are given in the Supporting Information.

Figure 2. Details of the binding energies of carbon and oxygen on several carbide surfaces and their comparison to transition-metal surfaces. The yellow symbols indicate the stoichiometric carbide surface, whereas the red symbols connected to this indicate the oxidized (typically weaker binding) and reduced (typically stronger binding) surfaces. The open symbols represent bonding on transition metal close-packed (diamonds) and stepped (circles) surfaces, taken from the literature.3,34,61

coadsorbates can be deposited on the surface.34 Specifically, the surface may be reducedresulting in carbon vacanciesor oxidizedresulting in adatoms, which we take to be atomic oxygen, given the oxophilic nature seen above. To understand if the adsorption trends are an artifact of the choice of the stoichiometric surface, data for several nonstoichiometric surfaces are also shown in Figure 2. The same general trend holds for oxycarbide surfaces, which we show in Figures S3 and S7 of the Supporting Information. We see the expected trend that the reduced species, which are higher in vacancies and have a lower effective coordination number at the remaining metal surface, are more reactive. We also see the expected trend that the oxidized surfaces, which are partially covered in oxygen adsorbates, tend to be less reactive. However, the oxophilic chemisorption trend remains intact, as the reduced and oxidized carbides stay almost exclusively to the carbophobic region of the figure. On a given surface, adsorption energies between C and O typically scale monotonically as their oxidized/reduced state changes, as can be seen in Figure 2. We show the complete C, H, and O adsorption energy trends for the 12 modeled surfaces in Figure 3, and we see that a carbophobic departure is preserved. We additionally see that the carbophobic nature is intact when comparing the binding energies of carbon and hydrogen, shown in Figure 3B. We can notice a number of other trends within this data. The activities of given facets on early transition-metal carbides (e.g., TiC or TaC) are more sensitive to the surface condition than those on mid or late transition-metal carbides (e.g., Mo2C or Fe3C). Analogously, the activities of the early transition metal carbides are highly anisotropic,63 i.e., dependent on the distribution of specific facets at the carbide surface. As would be expected, significant deviation from linearity is found on surfaces that promote the reaction of adsorbates or the lattice carbon into molecular adsorbates (e.g., formation of C3 compounds on TiC(310), shown in Figure S4, Supporting Information). In Figure 4, we compare the C, H, and O binding energy of a subset of the carbide surfaces to that of their parent metal surfaces; for example, WC vs W. Atomic hydrogen tends to bind stronger on metal carbides than on the parent metal surfaces, as reported previously for early transition-metal carbides64 and Mo2C;34 here, we can see this trend typically applies across carbides in Figure 4. However, we can also see

■

RESULTS AND DISCUSSION Probe Adsorbates. The simplest adsorbates of relevance to catalysis are the single atoms, from which we have chosen C, O, and H as probes. On the pure transition metals, strong correlations exist between the binding energies of these adsorbates. For example, we can see this as the white (open) symbols in Figure 2; the black line indicates the best fit to this data, including both the close-packed (e.g., fcc(111)) and stepped (e.g., fcc(211)) surfaces. Outliers include Pt and Pd, which fall significantly above the line and tend to have interesting catalytic properties. Because Pt and Pd bind oxygen relatively weakly compared to their affinity for carbon, we refer to these surfaces as “oxophobic”. This departure may help explain the unique activity of these catalysts, for example, in the partial oxidation of methane, which may be limited by too strong binding of oxygen on more reactive metals such as Rh.62 Also present in Figure 2 are the binding energies of several carbide surfaces, shown as the yellow symbols. It can be seen that a deviation opposite to that of Pt and Pd exists for these surfaces; namely, they bind carbon weakly compared to their affinity for oxygen, in what we might term an “oxophilic” or “carbophobic” departure. Carbide surfaces can be expected to be dynamic in catalytic operation: vacancies can be created and destroyed, and 13028



Article

Figure 3. Adsorption energies of oxygen (A) or hydrogen (B) vs carbon. Open circles show correlations for stepped metal surfaces and open diamonds for close-packed metal surfaces, all taken from the literature3,34,61 except for Ti(112̅2) that was added for this study (see Supporting Information). The stepped surfaces include fcc(211) data and stepped Mo(210), W(210), and Ti(112̅2). Adsorption energies on stoichiometric carbide surfaces are shown with yellow symbols and numbered as follows: 1: WC(0001), 2: WC(1011̅), 3: Mo2C(001), 4: Mo2C(101), 5: Mo2C(011), 6: Fe3C(001), 7: TiC(001), 8: TiC(111), 9: TiC(310), 10: TaC(001), 11: TaC(111), or 12: TaC(110). The presence of 25−50% carbon vacancies or of approximately 1/4 to 3/4 monolayer oxygen adatoms simulates reducing or oxidizing reaction conditions, as shown in the legend. All raw data are available in the Supporting Information, such that these graphs can be reproduced at any desired level of detail. The present version emphasizes the oxophilic and oxophobic departures with respect to stoichiometric carbide surfaces (shown with yellow symbols).

that carbides typically bind both C and O weaker relative to the parent metal. More precisely, metal carbides show significantly weaker carbon binding but only moderately weaker oxygen binding, relative to their parent metal, shedding more light onto the nature of the carbophobic departure. On some metal carbide surfaces with or without oxygen adatoms the hydrogen binding energies are relatively close to those of Pt (Figure 3B), the single pure metal near the top of the HER volcano (i.e., with near-optimum hydrogen binding).65 While hydrogen coverage effects are not accounted for in this work, weakening of hydrogen bonding due to hydrogen coverage41 may partly justify Pt-like characteristics. This includes the ability to dissociate H2 in the presence of H2O at ambient conditions predicted for instance for WC27 as well as the experimentally reported higher activity of WC for catalyzing the HER, relative to pure W.66

Figure 4. Comparison of the binding energy of C, H, and O on carbide surfaces versus on their parent (carbon-free) metals. In most cases, carbon and oxygen bind more weakly on the carbides, while hydrogen binds more strongly on the carbides. Carbides: 1: WC(0001), 2: WC(101̅1), 3: Mo2C(001), 4: Mo2C(101), 5: Mo2C(011), 6: Fe3C(001), 7: TiC(001), 8: TiC(111), 9: TiC(310). Comparison is with the stepped parent metal surface, and I is H, C, or O.

Figure 5. Chemisorption energies of molecular adsorbates that are limiting or relevant to CO2 electroreduction, i.e., (A) CO, (B) CHO (formyl), and (C) COOH (carboxyl) and OCHO (formate) on stepped metal (empty circles, black line)3,8,18,61 or close-packed metal carbide surfaces (i.e., WC(0001), Fe3C(001), Mo2C(001), TaC(111), and TiC(111); filled circles, red line), respectively. Linear fits are to guide the eye. The lower left panel shows the site preference of CO on carbide surfaces (carbon shown in gray, metals in brown) and the distance between the O of CO and the next-closest surface metal atom, d(M−OCO) in Å, that is, continuously increasing with weaker CO chemisorption, as shown schematically on the lower right panel. 13029



Article

Molecular Adsorbates. The different adsorption behavior of carbides as compared to transition metals suggests differing catalytic trends from those which may limit the reactivity of the transition metal catalysts. In particular, the carbophobic nature may suggest a relatively weaker binding of adsorbates that interact with the surface through carbon and relatively stronger binding through oxygen. Similarly, highly hydrogenated molecular fragments may bind stronger at the metal carbide relative to the parent metal surface. With the aid of Figure 5, we discuss how these principles apply to oxygenated adsorbates relevant to catalytic CO2 reduction, CO hydrogenation (Fischer−Tropsch), or deoxygenation of biomass-derived oxygenates. Similar trends for some hydrogenated molecular fragments, such as OH, CH, and CH3 are provided in the Supporting Information. Of particular importance in catalysis, CO binds with its carbon and sometimes also with its oxygen atom to the catalyst surface. Relative to the parent metal, CO binding may strengthen or weaken on carbides: CO binding is about 0.49 eV stronger on Mo2C(001) vs Mo(210) or 0.16 eV stronger on Fe3C(001) vs Fe(210) but 0.41 eV weaker on WC(0001) vs W(210), as shown in Figure 5A. As compared to the transition metals, we find CO binding on the carbides to be about three times more sensitive to the carbon adsorption energies, suggesting additional interaction of the oxygen constituent with the surface. A highly carbophobic surface such as WC(0001) may weaken the metal−carbon (M−C) bond and strengthen the C−O bond in the CO molecule leading to the end-on (α-state) geometry of CO* on WC(0001)67 and Fe3C(001) with a relatively large distance between a surface metal and the oxygen of CO (see Figure 5). On the other hand, Mo2C(001) may strengthen the interaction between the metal and oxygen, weakening the C−O bond and resulting in the inclined (β-state) of CO*30 with decreased M−O distance. The nature of carbides may be attractive for the reduction of CO2, CO, or complex oxygenates into hydrocarbonsas has long been demonstrated in thermochemical applications.33,46 For example, the hydrogenation of CO has been proposed to be the potential-limiting step in the electroreduction of CO218 and the rate-limiting step in Fischer−Tropsch synthesis.35 Figure 5B shows the relationship between the calculated CO and CHO binding energies on many transition metals and carbides; we can see that the scaling observed in the transition metals (white circles) is not necessarily intact on the carbides (red circles). In fact, incorporation of these binding energies into a free energy model for CO electroreduction suggests this step to be exergonic on the majority of the studied carbide surfaces. Also, the relatively strong adsorption of formate (Figure 5C) and the presence of oxygen adatoms discussed above suggest possibilities of metal carbide catalysts in formic acid formation, reactions where the abstraction of oxygen from the adsorbate is desirable, and in oxygen-assisted catalysis.44,45 In the context of Mars−van Krevelen (MvK) type reaction mechanisms in Fischer−Tropsch synthesis on iron carbide,29,35 although much debate exists about the mechanism, a recent study has suggested the two most endothermic surface reactions are the liberation of lattice carbon (ClatticeH* to CH2*; the asterisk denotes a surface-bound species) and the hydrogenation of lattice-bound carbon monoxide (ClatticeO* to ClatticeOH*) resulting in the breaking of the CO bond to replenish the lattice carbon (Clattice* + OH*).35 We can use our results to qualitatively understand the desirability of the carbophobic nature of iron carbide catalysts within this

mechanistic framework. The ideal MvK catalyst for Fischer− Tropsch synthesis must bind carbon weakly enough (carbophobic) to liberate the lattice carbon, while it has to bind O or OH strongly enough (oxophilic) to break the CO bond to replenish the defect. Understanding Adsorption in the Context of Materials Properties. The departures of the carbides from the carbon/oxygen scaling was established above with atomic probe adsorbates, which differ in the degree of undercoordination i.e., 2-fold undercoordinated oxygen (two valence electrons in O p orbitals) vs 4-fold undercoordinated carbon (four electrons in C sp3 orbitals) vs 1-fold undercoordinated hydrogen (one electron in the H s orbital). We find much of this trend to be conserved with the equally undercoordinated OH and CH3 adsorbates (shown in Figure S2, Supporting Information), which indicates it is not solely an artifact of the choice of probe adsorbates. However, depending on its valence configuration,4 a metal surface can be expected to reconstruct to accommodate adsorbates of different valences. The presence of carbon in the metal crystal lattice can alter or restrict this geometric rearrangement, resulting in the relatively weak adsorption of highly undercoordinated adsorbates. The surface−adsorbate bond is analyzed via representative charge density differences in Figure 6. On Mo2C(001), with 33 mol % lattice carbon, both

Figure 6. Charge density differences (in units of the elementary charge per Å3) of stoichiometric Mo2C(001) and TaC(001) with 25% lattice carbon vacancies (VC) or the bcc(110) parent metal surfaces. A) The volumetric data are given at adsorbate height, i.e., slab cross sections at the z coordinate of the adsorbate or at “surface height” (indicated), i.e., 0.76 or 0.66 Å (the covalent radii of C or O) closer to the slab. 13030



Article

success in describing adsorption trends on transition metals. While these trends are seen to some degreefor example, stronger CO binding on TiC(111) or TaC(111) vs weaker CO binding on Mo2C(001) or Fe3C(001))the atomic adsorption energies on metal carbides show little to no correlation with the d-band center, either across the stoichiometric surfaces or for a given surface at various reaction conditions (Figures S9 and S10, Supporting Information). Although refinements to the dband model and other understandings have been discussed,5,50,64,68−70 a low predictive power of the (unrefined) d-band theory for metal carbides may be expected since carbides violate the model’s core assumption of approximate independence of the metal sp states, which is seen in Figure 8.

oxygen and carbon are stabilized with three metal−adsorbate bonds at equivalent 3-fold metal sites. However, Mo(110) reconstructs during the adsorption of carbon: the two more distant Mo atoms (horizontal in the figure) move 0.102 Å closer, while the two closer Mo atoms (vertical) move 0.144 Å further apart to facilitate the adsorption of carbon at a 4-fold site. The lattice carbon of metal carbides gradually restricts this surface reconstruction. On Fe3C(001) (Figure S12, Supporting Information) with only 25 mol % lattice carbon, adsorption of carbon at 4-fold metal sites is conserved in principle, but the biaxial symmetry of the adsorption site on Fe(110) is broken on Fe3C(001). We note that similar arguments may explain the tendency for formation of C−C bonds on surfaces that lack 4fold or 3-fold M sites (Figure S13, Supporting Information). In consequence, the oxophilic departure from the transition-metal scaling for adsorbates with low and equal valence, such as OH/ CH3 (Figure S2, Supporting Information), is less pronounced relative to the O/C scaling but in principle preserved. If lattice carbon is removed from a carbide resulting in an accessible surface vacancy (shown for TaC, Figure 6), then carbon adsorbates tend energetically to fill these vacancies, while oxygen and hydrogen tend to form two metal−adsorbate bonds at metal−metal bridge sites. We note that the atomic hydrogen probe adsorbate has only one unpaired electron, increasing the “degree of freedom” in finding adsorption sites, and we tend to see more variability in its preferred binding site. The geometric preference of the adsorbates on most carbide surfaces varies only slightly upon addition of oxygen adatoms which indicates that the passivating effect of these adatoms is predominantly of an electronic nature. Since the reactivity of metal carbide surfaces cannot be explained alone with bond valence arguments, a distinct difference in the electronic structure between transition-metal and metal carbide surfaces is plausible. Understanding such electronic structure details is useful when employing the abovedescribed principles for the rational design of metal compound catalysts. Figure 7 decomposes the adsorbate−surface bond

Figure 8. (A) Correlation between the adsorption energies of carbon with the metal-projected sp-band center of the stoichiometric metal carbide surfaces (round symbols, numbering identical to Figure 3) and among surface conditions for Mo2C (triangles). (B) Correlation between the adsorption energies of oxygen and representative metal carbide surfaces at various surface conditions for Mo2C and TiC. A complete data set is provided in the Supporting Information. Linear fits are a guide to the eye.

On transition metal surfaces, the metal sp states are expected to contribute the larger fraction of the bonding between a surface and an adsorbate.2,3,42 This can be expected to hold for metal carbides as well. However, unlike the transition metals, the metal sp-band center at a carbide surface shifts significantly with the parent metal. We find that this shift correlates strongly with the adsorption energies on carbide surfaces with varying parent metal (Figure 8A) and varying reaction conditions (Figure 8B). Emptying the metal sp statesshifting the spband center up via, e.g., formation of carbon vacancies correlates with stronger adsorption energies, which is evidence for the emptying of antibonding states. This correlation is strong on most close-packed surfaces and describes the adsorption of carbon generally better than the adsorption of oxygen. Also, the appearance of a dominant splitting (see Figure S11, Supporting Information) of the M sp states into bonding states around −10 eV (hybridized with M d and C s states) and antibonding states near the Fermi level (that couple to the M d t2g states) may correlate with the carbophobic nature of the carbide surfaces. Although we note a correlation between the sp-band center and the binding energies of O and C, we note that this is perhaps less predictive than that of the d band model, which may be expected as we observe much less of a correlation between oxygen and carbon binding energies on the carbides than on the transition metals.

Figure 7. Representative Mo-projected density of states (DOS) of stoichiometric (A) Mo(110) and (B) Mo2C(001) (occupied states below the Fermi level are colored gray or maroon, respectively) and (C) Mo2C(001) at various surface conditions.

■

into metal-projected (M) s, p, and d orbitals to identify the formation of M sp states that are absent at the parent metal surface but present and active in the adsorption process on metal carbide surfaces. The d-band model,42 which correlates adsorption energies to the central moment of the metal d band, has seen tremendous

CONCLUSIONS The chemisorption energies of atomic and molecular adsorbates on 12 representative model transition-metal carbide surfaces depart from the adsorption energy correlations on 13031



Article

(3) Abild-Pedersen, F.; Greeley, J.; Studt, F.; Rossmeisl, J.; Munter, T.; Moses, P.; Skúlason, E.; Bligaard, T.; Nørskov, J. Scaling Properties of Adsorption Energies for Hydrogen-containing Molecules on Transition-metal Surfaces. Phys. Rev. Lett. 2007, 99, 16105. (4) Calle-Vallejo, F.; Martínez, J. I.; García-Lastra, J. M.; Rossmeisl, J.; Koper, M. T. M. Physical and Chemical Nature of the Scaling Relations between Adsorption Energies of Atoms on Metal Surfaces. Phys. Rev. Lett. 2012, 108, 116103. (5) Fernández, E.; Moses, P.; Toftelund, A.; Hansen, H.; Martínez, J.; Abild-Pedersen, F.; Kleis, J.; Hinnemann, B.; Rossmeisl, J.; Bligaard, T.; Nørskov, J. Scaling Relationships for Adsorption Energies on Transition Metal Oxide, Sulfide, and Nitride Surfaces. Angew. Chem., Int. Ed. 2008, 47, 4683−4686. (6) Pallassana, V.; Neurock, M. Electronic Factors Governing Ethylene Hydrogenation and Dehydrogenation Activity of Pseudomorphic PdML/Re(0001), PdML/Ru(0001), Pd(111), and PdML/ Au(111) Surfaces. J. Catal. 2000, 191, 301−317. (7) Michaelides, A.; Liu, Z. P.; Zhang, C. J.; Alavi, A.; King, D. A.; Hu, P. Identification of General Linear Relationships between Activation Energies and Enthalpy Changes for Dissociation Reactions at Surfaces. J. Am. Chem. Soc. 2003, 125, 3704−3705. (8) Bligaard, T.; Nørskov, J.; Dahl, S.; Matthiesen, J.; Christensen, C.; Sehested, J. The Brønsted−Evans−Polanyi Relation and the Volcano Curve in Heterogeneous Catalysis. J. Catal. 2004, 224, 206−217. (9) Viñes, F.; Vojvodic, A.; Abild-Pedersen, F.; Illas, F. Brönsted− Evans−Polanyi Relationship for Transition Metal Carbide and Transition Metal Oxide Surfaces. J. Phys. Chem. C 2013, 117, 4168− 4171. (10) Nørskov, J. K.; Abild-Pedersen, F.; Studt, F.; Bligaard, T. Density Functional Theory in Surface Chemistry and Catalysis. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 937−943. (11) Kozuch, S.; Shaik, S. A Combined Kinetic-Quantum Mechanical Model for Assessment of Catalytic Cycles: Application to CrossCoupling and Heck Reactions. J. Am. Chem. Soc. 2006, 128, 3355− 3365. (12) Stegelmann, C.; Andreasen, A.; Campbell, C. T. Degree of Rate Control: How Much the Energies of Intermediates and Transition States Control Rates. J. Am. Chem. Soc. 2009, 131, 8077−8082. (13) Hori, Y. Electrochemical CO2 Reduction on Metal Electrodes. In Modern Aspects of Electrochemistry; Springer: New York, 2008; Vol. 42, Chapter 3, pp 89−189. (14) Hansen, H. A.; Varley, J. B.; Peterson, A. A.; Nørskov, J. K. Understanding Trends in the Electrocatalytic Activity of Metals and Enzymes for CO2 Reduction to CO. J. Phys. Chem. Lett. 2013, 4, 388− 392. (15) Zhu, W.; Michalsky, R.; Metin, Ö .; Lv, H.; Guo, S.; Wright, C. J.; Sun, X.; Peterson, A. A.; Sun, S. Monodisperse Au Nanoparticles for Selective Electrocatalytic Reduction of CO2 to CO. J. Am. Chem. Soc. 2013, 135, 16833−16836. (16) Peterson, A. A.; Abild-Pedersen, F.; Studt, F.; Rossmeisl, J.; Nørskov, J. K. How Copper Catalyzes the Electroreduction of Carbon Dioxide into Hydrocarbon Fuels. Energy Environ. Sci. 2010, 3, 1311− 1315. (17) Ou, L. H.; Chen, S. L. DFT Study of CO2 Reduction to Hydrocarbons on Cu Surfaces. J. Electrochem. 2011, 17, 155. (18) Peterson, A. A.; Nørskov, J. K. Activity Descriptors for CO2 Electroreduction to Methane on Transition-Metal Catalysts. J. Phys. Chem. Lett. 2012, 3, 251−258. (19) Nie, X.; Esopi, M. R.; Janik, M. J.; Asthagiri, A. Selectivity of CO2 Reduction on Copper Electrodes: The Role of the Kinetics of Elementary Steps. Angew. Chem., Int. Ed. 2013, 52, 2459−2462. (20) Marković, N. M.; Ross, P. N. Surface Science Studies of Model Fuel Cell Electrocatalysts. Surf. Sci. Rep. 2002, 45, 117−229. (21) Nørskov, J. K.; Rossmeisl, J.; Logadottir, A.; Lindqvist, L.; Kitchin, J. R.; Bligaard, T.; Jonsson, H. Origin of the Overpotential for Oxygen Reduction at a Fuel-Cell Cathode. J. Phys. Chem. B 2004, 108, 17886−17892. (22) Friebel, D.; Viswanathan, V.; Miller, D. J.; Anniyev, T.; Ogasawara, H.; Larsen, A. H.; O’Grady, C. P.; Nørskov, J. K.; Nilsson,

transition-metal surfaces. These oxophilic and carbophobic departures may help to explain the reactivity of carbide catalysts and may aid in the design of catalysts which intrinsically avoid the rate- and potential-limiting steps of many chemical reactions on heterogeneous catalysts and electrocatalysts, such as shown here for the hydrogenation of CO which limits the electrocatalytic reduction of CO2 into renewable hydrocarbon fuels. The relations of atomic and molecular adsorption energies can be utilized to map metal and metal carbide surfaces in the catalytic activity space that is described by the strength of carbon, oxygen, and hydrogen binding at the surface. Generally, metal carbides tend to bind oxygen relatively strongly, relative to how strongly carbon is bound, which is an opposite departure from the behavior observed for Pt and Pd. This may guide the choice of these catalysts in applications such as CO2 reduction or deoxygenation reactions. The provided data of adsorption energies can be employed to establish Brønsted− Evans−Polanyi relations to predict catalytic activities of a given metal carbide catalyst for selected catalytic applications. The surface reactivity of metal carbides can be partially rationalized with the valence configuration of the catalyst surface as well as the contributions of the metal-projected sp states to the adsorbate−surface bond. These origins of surface activity provide the beginning of a basis for the understanding of these complex catalyst materials and outline possible factors that modify the catalytic activity of metal carbide surfaces at the atomic scale, i.e., formation of carbon vacancies, partial surface oxidation, and filling or emptying of the metal d and sp band (e.g., when applying surface strain).

■

ASSOCIATED CONTENT

S Supporting Information *

Detailed computational methods; database of adsorption and reference energies; OH/CH3 scaling relation; data for oxycarbide surfaces; adsorption geometries; surface energy correlations; electronic structure details; DOS analyses for metal and metal carbide surfaces; charge density details; C−C bond formation trends. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS We are grateful to Aleksandra Vojvodic for fruitful discussions. This work was supported financially by the Office of Naval Research under Young Investigator Award N00014-12-1-0851. Electronic structure calculations were conducted at the Center for Computation and Visualization, Brown University.

■

REFERENCES

(1) Brown, W. A.; Kose, R.; King, D. A. Femtomole Adsorption Calorimetry on Single-Crystal Surfaces. Chem. Rev. 1998, 98, 797− 831. (2) Hammer, B.; Nørskov, J. Theoretical Surface Science and Catalysis-Calculations and Concepts. In Bruce, H.K., Gates, C., Ed.; Impact of Surface Science on Catalysis; Advances in Catalysis; Academic Press: New York, 2000; Vol. 45, pp 71−129. 13032



Article

(44) Holz, M. C.; Kähler, K.; Tölle, K.; van Veen, A. C.; Muhler, M. Gas-phase Oxidation of 2-propanol over Au/TiO2 Catalysts to Probe Metal support Interactions. Physica Status Solidi (B) 2013, 250, 1094− 1106. (45) Siler, C. G. F.; Xu, B.; Madix, R. J.; Friend, C. M. Role of Surface-Bound Intermediates in the Oxygen-Assisted Synthesis of Amides by Metallic Silver and Gold. J. Am. Chem. Soc. 2012, 134, 12604−12610. (46) Kojima, I.; Miyazaki, E.; Yasumori, I. Synthesis of Hydrocarbons from CO and H2 over Metal Carbide Catalysts. J. Chem. Soc., Chem. Commun. 1980, 212, 573−574. (47) Kuhl, K. P.; Cave, E. R.; Abram, D. N.; Jaramillo, T. F. New Insights into the Electrochemical Reduction of Carbon Dioxide on Metallic Copper Surfaces. Energy Environ. Sci. 2012, 5, 7050−7059. (48) Rosen, B. A.; Salehi-Khojin, A.; Thorson, M. R.; Zhu, W.; Whipple, D. T.; Kenis, P. J. A.; Masel, R. I. Ionic Liquid-Mediated Selective Conversion of CO2 to CO at Low Overpotentials. Science 2011, 334, 643−644. (49) Viñes, F.; Sousa, C.; Liu, P.; Rodriguez, J. A.; Illas, F. A Systematic Density Functional Theory Study of the Electronic Structure of Bulk and (001) Surface of Transition-metals Carbides. J. Chem. Phys. 2005, 122, 174709. (50) Vojvodic, A.; Hellman, A.; Ruberto, C.; Lundqvist, B. I. From Electronic Structure to Catalytic Activity: A Single Descriptor for Adsorption and Reactivity on Transition-Metal Carbides. Phys. Rev. Lett. 2009, 103, 146103. (51) Bahn, S. R.; Jacobsen, K. W. An Object-oriented Scripting Interface to a Legacy Electronic Structure Code. Comput. Sci. Eng. 2002, 4, 56−66. (52) Both Dacapo and ASE are open-source codes that are freely available from the Department of Physics at the Technical University of Denmark. (53) Vanderbilt, D. Soft Self-consistent Pseudopotentials in a Generalized Eigenvalue Formalism. Phys. Rev. B 1990, 41, 7892−7895. (54) Payne, M. C.; Teter, M. P.; Allan, D. C.; Arias, T. A.; Joannopoulos, J. D. Iterative Minimization Techniques for Ab Initio Total-energy Calculations: Molecular Dynamics and Conjugate Gradients. Rev. Mod. Phys. 1992, 64, 1045−1097. (55) Kresse, G.; Furthmller, J. Efficiency of Ab-initio Total Energy Calculations for Metals and Semiconductors Using a Plane-wave Basis Set. Comput. Mater. Sci. 1996, 6, 15−50. (56) Hammer, B.; Hansen, L. B.; Nørskov, J. K. Improved Adsorption Energetics within Density-functional Theory Using Revised Perdew-Burke-Ernzerhof Functionals. Phys. Rev. B 1999, 59, 7413−7421. (57) Herbstein, F. H.; Smuts, J. Comparison of X-ray and Neutrondiffraction Refinements of the Structure of Cementite Fe3C. Acta Crystallogr. 1964, 17, 1331−1332. (58) Otani, S.; Ishizawa, Y. Preparation of Mo2C Single Crystals by the Floating Zone Method. J. Cryst. Growth 1995, 154, 202−204. (59) Kurlov, A. S.; Gusev, A. I. Phase Equilibria in the W-C System and Tungsten Carbides. Russ. Chem. Rev. 2006, 75, 617. (60) Wu, S. Y.; Ho, J. J. Adsorption, Dissociation, and Hydrogenation of CO2 on WC(0001) and WC-Co Alloy Surfaces Investigated with Theoretical Calculations. J. Phys. Chem. C 2012, 116, 13202−13209. (61) Wang, S.; Petzold, V.; Tripkovic, V.; Kleis, J.; Howalt, J. G.; Skúlason, E.; Fernández, E. M.; Hvolbæk, B.; Jones, G.; Toftelund, A.; Falsig, H.; Björketun, M.; Studt, F.; Abild-Pedersen, F.; Rossmeisl, J.; Nørskov, J. K.; Bligaard, T. Universal Transition State Scaling Relations for (De)hydrogenation over Transition Metals. Phys. Chem. Chem. Phys. 2011, 13, 20760−20765. (62) Enger, B. C.; Lødeng, R.; Holmen, A. A Review of Catalytic Partial Oxidation of Methane to Synthesis Gas with Emphasis on Reaction Mechanisms over Transition Metal Catalysts. Appl. Catal. A: Gen. 2008, 346, 1−27. (63) Lowekamp, J. B.; Rohrer, G. S.; Morris Hotsenpiller, P. A.; Bolt, J. D.; Farneth, W. E. Anisotropic Photochemical Reactivity of Bulk TiO2 Crystals. J. Phys. Chem. B 1998, 102, 7323−7327.

A. Balance of Nanostructure and Bimetallic Interactions in Pt Model Fuel Cell Catalysts: In Situ XAS and DFT Study. J. Am. Chem. Soc. 2012, 134, 9664−9671. (23) Studt, F.; Abild-Pedersen, F.; Wu, Q.; Jensen, A. D.; Temel, B.; Grunwaldt, J. D.; Nørskov, J. K. CO Hydrogenation To Methanol On CuNi Catalysts: Theory And Experiment. J. Catal. 2012, 293, 51−60. (24) Jacobsen, C. J. H.; Dahl, S.; Clausen, B. S.; Bahn, S.; Logadottir, A.; Nørskov, J. K. Catalyst Design by Interpolation in the Periodic Table: Bimetallic Ammonia Synthesis Catalysts. J. Am. Chem. Soc. 2001, 123, 8404−8405. (25) Studt, F.; Abild-Pedersen, F.; Bligaard, T.; Sørensen, R. Z.; Christensen, C. H.; Nørskov, J. K. Identification of Non-Precious Metal Alloy Catalysts for Selective Hydrogenation of Acetylene. Science 2008, 320, 1320−1322. (26) Suntivich, J.; May, K. J.; Gasteiger, H. A.; Goodenough, J. B.; Shao-Horn, Y. A Perovskite Oxide Optimized for Oxygen Evolution Catalysis from Molecular Orbital Principles. Science 2011, 334, 1383− 1385. (27) Levy, R. B.; Boudart, M. Platinum-Like Behavior of Tungsten Carbide in Surface Catalysis. Science 1973, 181, 547−549. (28) Weigert, E. C.; Stottlemyer, A. L.; Zellner, M. B.; Chen, J. G. Tungsten Monocarbide as Potential Replacement of Platinum for Methanol Electrooxidation. J. Phys. Chem. C 2007, 111, 14617−14620. (29) Cheng, J.; Hu, P.; Ellis, P.; French, S.; Kelly, G.; Lok, C. M. Density Functional Theory Study of Iron and Cobalt Carbides for Fischer−Tropsch Synthesis. J. Phys. Chem. C 2010, 114, 1085−1093. (30) Ko, E.; Madix, R. Effects of Adsorbed Carbon and Oxygen on the Chemisorption of H2 and CO on Mo(100). Surf. Sci. 1981, 109, 221−238. (31) Huo, C. F.; Li, Y. W.; Wang, J.; Jiao, H. Insight into CH4 Formation in Iron-Catalyzed Fischer−Tropsch Synthesis. J. Am. Chem. Soc. 2009, 131, 14713−14721. (32) Liu, P.; Rodriguez, J. A. Water-Gas-Shift Reaction on Molybdenum Carbide Surfaces: Essential Role of the Oxycarbide. J. Phys. Chem. B 2006, 110, 19418−19425. (33) Dubois, J. L.; Sayama, K.; Arakawa, H. CO2 Hydrogenation over Carbide Catalysts. Chem. Lett. 1992, 21, 5−8. (34) Medford, A. J.; Vojvodic, A.; Studt, F.; Abild-Pedersen, F.; Nørskov, J. K. Elementary Steps of Syngas Reactions on Mo2C(001): Adsorption Thermochemistry and Bond Dissociation. J. Catal. 2012, 290, 108−117. (35) Gracia, J.; Prinsloo, F.; Niemantsverdriet, J. Mars−van Krevelenlike Mechanism of CO Hydrogenation on an Iron Carbide Surface. Catal. Lett. 2009, 133, 257−261. (36) Dhandapani, B.; Clair, T. S.; Oyama, S. Simultaneous Hydrodesulfurization, Hydrodeoxygenation, and Hydrogenation with Molybdenum Carbide. Appl. Catal. A: Gen. 1998, 168, 219−228. (37) Ren, H.; Hansgen, D. A.; Stottlemyer, A. L.; Kelly, T. G.; Chen, J. G. Replacing Platinum with Tungsten Carbide (WC) for Reforming Reactions: Similarities in Ethanol Decomposition on Ni/Pt and Ni/ WC Surfaces. ACS Catal. 2011, 1, 390−398. (38) Vojvodic, A. Steam Reforming on Transition-Metal Carbides from Density-Functional Theory. Catal. Lett. 2012, 142, 728−735. (39) Rodriguez, J. A.; Illas, F. Activation of Noble Metals on Metalcarbide Surfaces: Novel Catalysts For CO Oxidation, Desulfurization and Hydrogenation Reactions. Phys. Chem. Chem. Phys. 2012, 14, 427−438. (40) Hunt, S. T.; Nimmanwudipong, T.; Romn-Leshkov, Y. Engineering Non-sintered, Metal-Terminated Tungsten Carbide Nanoparticles for Catalysis. Angew. Chem., Int. Ed. 2014, 53, 5131− 5136. (41) Michalsky, R.; Zhang, Y. J.; Peterson, A. A. Trends in the Hydrogen Evolution Activity of Metal Carbide Catalysts. ACS Catal. 2014, 4, 1274−1278. (42) Hammer, B.; Nørskov, J. K. Electronic Factors Determining the Reactivity of Metal Surfaces. Surf. Sci. 1995, 343, 211−220. (43) Hwu, H. H.; Chen, J. G. Surface Chemistry of Transition Metal Carbides. Chem. Rev. 2005, 105, 185−212. 13033



Article

(64) Kitchin, J. R.; Nørskov, J. K.; Barteau, M. A.; Chen, J. G. Trends in the Chemical Properties of Early Transition Metal Carbide Surfaces: A Density Functional Study. Catal. Today 2005, 105, 66−73. (65) Greeley, J. Computational Studies of Trends in Electrocatalysis. ECS Trans. 2012, 45, 85−95. (66) Harnisch, F.; Sievers, G.; Schröder, U. Tungsten Carbide as Electrocatalyst for the Hydrogen Evolution Reaction in pH Neutral Electrolyte Solutions. Appl. Catal., B 2009, 89, 455−458. (67) Friend, C. M.; Serafin, J. G.; Baldwin, E. K.; Stevens, P. A.; Madix, R. J. Bonding and Adsorption Structure of CO on W(100)(5 × 1)C. J. Chem. Phys. 1987, 87, 1847−1850. (68) Xin, H.; Vojvodic, A.; Voss, J.; Nørskov, J. K.; Abild-Pedersen, F. Effects of d-band Shape on the Surface Reactivity of Transition-metal Alloys. Phys. Rev. B 2014, 89, 115114. (69) Vojvodic, A.; Ruberto, C.; Lundqvist, B. I. Atomic and Molecular Adsorption on Transition-metal Carbide (111) Surfaces from Density-functional Theory: A Trend Study of Surface Electronic Factors. J. Phys.: Condens. Matter 2010, 22, 375504. (70) Ruberto, C.; Lundqvist, B. I. Nature of Adsorption On TiC(111) Investigated with Density-functional Calculations. Phys. Rev. B 2007, 75, 235438.

13034


Departures from the Adsorption Energy Scaling Relations for Metal

Recommend Documents