A First-Principles Study on the Role of an Al2O3 ... - ACS Publications

Oct 30, 2015 - Ofer Neufeld,. †. Natav Yatom,. ‡ and Maytal Caspary Toroker*,‡. †. The Nancy and Stephen Grand Technion Energy Program and. â€...
1 downloads 0 Views 3MB Size
Research Article pubs.acs.org/acscatalysis

A First-Principles Study on the Role of an Al2O3 Overlayer on Fe2O3 for Water Splitting Ofer Neufeld,† Natav Yatom,‡ and Maytal Caspary Toroker*,‡ †

The Nancy and Stephen Grand Technion Energy Program and ‡The Department of Materials Science and Engineering, Technion-Israel Institute of Technology, Haifa 32000, Israel S Supporting Information *

ABSTRACT: Understanding the role of an overlayer material on a catalyst is crucial for improving catalytic activity. Iron(III) oxide (α-Fe2O3) is a widely studied catalyst commonly used for solar water splitting. Recently, the water splitting efficiency with α-Fe2O3 was enhanced by deposition of an α-Al2O3 overlayer. In order to understand the origin of this improvement, we perform first-principles calculations with density functional theory + U on the αFe2O3(0001) surface with an α-Al2O3 surface overlayer. We find catalysis is unfavorable directly over α-Al2O3 and rather takes place over α-Fe2O3 exposed areas. In agreement with experiment, we find that α-Al2O3 coverage decreases the overpotential required for water oxidation on α-Fe2O3. We explain this improvement through the decrease in the work function of α-Fe2O3 upon αAl2O3 coverage that aids in extracting electrons during the water oxidation reaction. We suggest that selecting an overlayer with a smaller work function than that of the catalyst as a strategy for future development of better catalysts. KEYWORDS: water splitting, density functional theory, DFT+U, iron oxides, surface states

I. INTRODUCTION The role of an overlayer material above a catalyst is fundamental for understanding how to improve catalysis in general and photocatalysis and photoelectrochemical cells in particular.1 In a photoelectrochemical cell, photons are absorbed at the cathode and/or anode electrodes. The excited electron and hole then diffuse through the electrodes toward their surfaces. At the cathode and anode surfaces, the electrons and holes participate in reduction and oxidation reactions, respectively. Since an overlayer changes the surface on which catalysis occurs, the overlayer plays a central role in photoelectrochemical cell operation. Remarkable advances have been developed for understanding and improving water oxidation with the iron(III) oxide catalyst (α-Fe2O3; α will be dropped from this point on),2−10 a widely studied anode material in photoelectrochemical cells. Some design strategies include nanostructuring,9,11,12 doping,10,13,14 coating,15 and assembly in multilayer junctions,16 but further improvements in conversion efficiency are still needed.14,17−23 The most recent breakthrough in designing Fe2O3 catalysis is the addition of overlayers.24,25 In recent work the overpotential for water oxidation was successfully reduced by a surface treatment using atomic layer deposition (ALD) for Fe2O3 with thin overlayers of materials such as α-Al2O3, Ga2O3, and In2O3.24,26−28 Here we focus on surface modification via a thin aluminum(III) oxide (α-Al2O3, sapphire; α will be dropped from this point on) overlayer that was found to favorably reduce the required overpotential for water oxidation by 0.1 V.26,27 © XXXX American Chemical Society

In this paper, we use density functional theory (DFT) + U calculations in order to shed light on the role of an Al2O3 overlayer for water oxidation with Fe2O3. This work is a continuation of previous efforts that model the surface and water oxidation on Fe2O3(0001),29−33 including our own previous work.34,35 This is the first theoretical study where we model water oxidation catalysis over Al2O3 and over Fe2O3 fully or partially covered with Al2O3. We find that the Al2O3 overlayer does not participate in the surface catalytic activity. Furthermore, we find that partial coverage with some uncovered Fe2O3 active areas requires a lower overpotential than completely uncovered Fe2O3, in agreement with recent experiments.26,27

II. COMPUTATIONAL METHODS This section includes the following computational details: (1) the theoretical and computational framework, (2) the slab models for Fe2O3 with and without an Al2O3 overlayer, and (3) details on water oxidation reaction free energy calculations. Further computational details and relaxed unit cell structures are given in the Supporting Information. We used the Vienna ab initio simulation package (VASP)36,37 for spin-polarized DFT+U calculations according to the formalism of Dudarev et al.38 DFT+U accounts for high electronic correlation effects that regular DFT exchangeReceived: May 31, 2015 Revised: October 27, 2015

7237

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis

Figure 1. Fully hydroxylated slabs of (a) pure Al2O3, (b) Al2O3 monolayer covered Fe2O3, (c) Al2O3 bilayer covered Fe2O3, and (d) 2 × 2 Fe2O3 (0001) slab with 75% Al2O3 monolayer surface termination. Top views are given in the insets. The atom colors red, blue, gold, and white denote O, Al, Fe, and H atoms, respectively. Active sites are highlighted with a black circle in each cell. In the inset in (d), blue shading represents the 25% Fe2O3 surface area and green shading the Al2O3 surface area. The figure was created with VESTA visualizing software.52

k-grid for all bulk and slab calculations. Bulk Al2O3 and 1 × 1 slab calculations converged with a 3 × 3 × 1 k-grid and 2 × 2 supercell slab calculations with a 2 × 2 × 1 grid. Atomic geometry relaxation was performed up to a tolerance of 0.03 eV/Å for atomic forces. Bulk Fe2O3 was calculated with a hexagonal 30-atom unit cell with long-range antiferromagnetic ordering according to the information given in ref 40. A relaxed geometry with lattice parameters of a = 5.097 Å and c = 13.873 Å was obtained, in good agreement with experimental measurements of a = 5.034 Å and c = 13.747 Å.49 Bulk Al2O3 calculations were performed in the hexagonal 30-atom cell. We obtained a relaxed cell with lattice vectors a = b = 4.804 Å and c = 13.108 Å, consistent with previous theoretical work42 and experimental measurements of a = b = 4.758 Å and c = 12.991 Å.50 The Fe2O3 and Fe2O3 covered with Al2O3 slabs were cleaved from the Fe2O3 bulk relaxed hexagonal unit cells with (0001) termination, since Al2O3 is expected to adopt the lattice constant of the Fe2O3 substrate upon thin deposition of Al2O3. The Al2O3 pure slab was cleaved from the Al2O3 bulk relaxed hexagonal unit cell with (0001) termination. We chose to focus on this facet, since this is one of the stable facets of Fe2O3 under aqueous conditions30,51 and since we could compare to previous literature on this facet.29 The initial position of terminating hydrogen atoms was placed according to the

correlation (XC) functional approximations fail to represent for Fe2O3.39 We chose a U value of 4.3 eV that was derived ab initio40 for Fe atoms, which was found useful for modeling the ground state of Fe2O3 as well as water oxidation with Fe2O3.29,34,35,40 All other atoms were not given explicit orbital corrections, that is, a U value of 0. Calculations were performed with the Perdue−Burke− Ernzerhof (PBE)41 XC functional, since this functional was previously shown to give a correct description of ground state geometrical and electronic properties for both Fe2O340 and Al2O3,42,43 and since using hybrid functionals gives similar results.44 Projected augmented plane wave (PAW) potentials were used to represent the frozen core electrons and nuclei of each atom: [Ne]3s2 shells in Fe atoms, [He] shells in O atoms, and [Ne] shells in Al atoms.45,46 We solved the Kohn−Sham equations to self-consistency to a tolerance of 10−5 eV in total energy (the 2 × 2 slab calculations gave the same results with a harsher SCF convergence criterion of 10−6 eV). All calculations used a plane-wave basis set, while symmetry was not imposed. k-space integration was performed via the tetrahedron method with Blöchl corrections.47,48 We converged the plane wave energy cutoff to a tolerance of 1 meV/atom in total energy in all calculations, resulting in a 750 eV energy cutoff. k-grids converged up to 1 meV/atom in total energy with a Γ-centered 7238

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis information in ref 29. All slab calculations were converged with a 10 Å vacuum separating the slab from its periodic image. Coating of Al2O3 over Fe2O3 was modeled with four slab structures (see Figure 1):52 The first structure (Figure 1a) is a pure Al2O3 1 × 1 slab with 1/3 ML (monolayer) reactive site coverage. This geometry represents a water oxidation reaction over a thick Al2O3 coating, where the Fe2O3 is sufficiently far from the active site to not participate in the reaction (>1 nm). Since this slab geometry represents a thick Al2O3 overlayer, we assume that the Al2O3 geometry is best described by the coating (and not substrate) lattice parameters. The second structure (Figure 1b) is “thin coating” with a monolayer-coated Fe2O3, where Al atoms replaced Fe atoms in the topmost layer. We model Fe2O3 with an Al2O3 overlayer with the orientation relationship Fe2O3(0001),[1000]∥(0001), [1000]Al2O3, since the two materials have identical lattice structures (corundum) with relatively close lattice parameters (5.5% miss-fit53). The Fe2O3 thickness converged the overpotential with a tolerance of 0.04 eV (without any coating this thickness converged at a tolerance of 0.01 eV). The third structure (Figure 1c) is a “thin coating” with a bilayer-coated Fe2O3 where Al atoms replaced Fe atoms in the two topmost layers. For the bilayer we used a thicker slab with more Fe2O3 layers (six stoichiometric units thick) to keep the bulk-like geometry of Fe2O3 with the previously converged number of Fe2O3 layers and not have the coating thicker than the substrate. In the last structure (Figure 1d), we considered a 2 × 2 slab with 75% of the surface having a monolayer of Al2O3 coating. 1/12 ML reactive site coverage was modeled, where the active sites resided on an Fe2O3 exposed area. Slab shape and volume were held fixed to bulk Fe2O3 structure, ionic positions were allowed to relax, and slab thickness was chosen to be similar to the previously converged thickness in the 1 × 1 monolayer covered case. We add a specific note regarding the *O intermediate 2 × 2 slabs (defined in the next section). Since for this intermediate the electronic self-consistency is particularly hard to converge to a spin symmetric ground state, spin and geometrical constraints along with high numerical tolerance parameters (10−8 eV SCF convergence criterion, with geometry relaxation performed up to a tolerance of 10−3 eV/Å for atomic forces) were used to find the true system ground state. Water oxidation was modeled over pure Al2O3(0001) and Al2O3(0001)-covered Fe2O3(0001) surface slabs with the following mechanism as previously suggested (seen in Figure 2):54 H 2O + * → *OH 2

(i)

*OH 2 → *OH + H+ + e−

(ii)

+

*OH → *O + H + e



Figure 2. Catalytic water splitting cycle as demonstrated for a 75% Al2O3 monolayer partially covered Fe2O3. The atom colors red, gold, blue, and white denote O, Fe, Al, and H atoms, respectively. Numbers 1−5 indicate the reaction steps. Arrows indicate entering and leaving reactants and products. Highlighted in yellow is the active site on the intermediates. The upper half of slabs is displayed. This figure was created using VESTA.52

overall neutral. The reaction free energies were calculated by subtracting the total energies of reactants and products while including zero point energy (ZPE) and entropic contributions. The free energies are calculated at zero potential with respect to the normal hydrogen electrode (NHE) and at pH 0. This does not limit our analysis, since overpotential differences between covered and uncovered Fe2O3 remain the same with the addition of constant terms associated with bias and pH55 in the expressions for the free energies. The corresponding equations for the free energies are55

H 2O + *O → *OOH + H + e

*OOH → O2 + * + H + + e−



⎛1 ⎞ 1 E∗OH2 − ⎜ E∗vac + E H2O⎟ + Δξ1 − T ΔS1 ⎝2 ⎠ 2

(1)

ΔG2 =

1 1 (E OH + E H2) − E∗OH2 + Δξ2 − T ΔS2 * 2 2

(2)

ΔG3 =

1 1 (E∗O + E H2) − E∗OH + Δξ3 − T ΔS3 2 2

(3)

ΔG4 =

⎛1 ⎞ 1 (E OOH + E H2) − ⎜ E∗O + E H2O⎟ + Δξ4 * ⎝2 ⎠ 2 − T ΔS4

ΔG5 =

(4)

⎛1 ⎞ 1 1 ⎜ E + E H2 + EO2⎟ − E∗OOH + Δξ5 ⎝ 2 ∗vac ⎠ 2 2 − T ΔS5

(iii) +

ΔG1 =

(5)

E*OOH, for instance, stands for the total energy of the slab with an *OOH adsorbate (on both sides), and EH2 is the total energy of the H2 molecule. Values for the total energies of molecules were obtained by calculations at the same level of theory: −6.77, −9.87, and −14.23 eV for H2, O2, and H2O, respectively. Δξi and TΔSi represent the ZPE differences and entropic differences for each reaction step i. We obtained ZPE corrections for all of our slabs from vibrational frequencies derived from finite difference Hessians (using the central difference approximation) of molecules in vacuum and

(iv) (v)

where * denotes an O atom vacancy site on the surface and *OH2, for example, is the water adsorbate in the vacant site (see Figure 2 for partially covered Fe2O3). All slabs are not charged, since these reaction intermediates are obtained after an electron transfer process has finished such that the system is 7239

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis

Table 1. Free Energy Differences for the Intermediate Reaction Steps of Water Oxidation and Their Estimated Overpotentials and Valence Band Edge Positions for the Fully Hydroxylated Intermediate

1. pure Fe2O3 (1/3 ML) 2. pure Fe2O3 (1/12 ML) 3. thick Al2O3 overlayer (1/3 ML) 4. thin monolayer Al2O3 (1/3 ML) 5. Thin bilayer Al2O3 (1/3 ML) 6. 25% Fe2O3 near 75% Al2O3 monolayer covered surface (1/12 ML)

ΔG1 (eV)

ΔG2 (eV)

ΔG3 (eV)

ΔG4 (eV)

ΔG5 (eV)

overpotential Φ (V)

valence band edge position (eV)

0.06 0.21 −0.77 0.39 0.64 0.24

−0.03 −0.15 −1.60 −0.58 −0.96 −0.32

1.85 1.89 2.30 2.32 2.26 1.73

1.65 1.55 0.95 1.22 1.22 1.75

0.92 0.97 3.58 1.10 1.29 1.05

0.73 0.77 2.46 1.21 1.14 0.64

−5.34 −5.32 −5.26 −4.79 −4.90 −5.01

Figure 3. Cumulative free energy plots for water oxidation on Fe2O3 with an Al2O3 overlayer: (a) 1/3 ML active sites with thick Al2O3 coverage, thin monolayer Al2O3 coverage, and thin bilayer Al2O3 coverage; (b) 1/12 ML active sites with partial 75% Al2O3 coverage.

the potential for water oxidation equals the experimental value of 1.23 eV,58 and this changes ΔG5 by an amount of 0.49 eV, which does not affect our conclusions. Charges were calculated with the Bader charge analysis scheme.59 The valence band edges of slabs were calculated relative to vacuum by aligning the energy of the last occupied orbital with the vacuum region potential energy.60 A water monolayer was not included in the calculations, since we only compare trends between covered and uncovered surfaces. Any associated shifts in the energy levels as a result of electrolyte are expected to be similar for both surfaces. Our assumption is based on previous work in ref 29, which shows that adding a water layer did not significantly change the reaction free energies and valence band edge positions.

adsorbates (further details can be found in the Supporting Information). Entropic corrections for gaseous molecules were obtained from thermodynamic tables.56 The entropic correction for water was evaluated as the entropic contribution to gasphase water minus the condensation energy at 298.15 K. Entropic corrections for the surface species were found in ref 29 to be negligible and were omitted from the analysis.29 The overpotential is calculated by subtracting the largest free energy difference of an intermediate reaction step and the final cumulative free energy per electron transfer: Φ = max{ΔGi} −

1 4

∑ ΔGi i

1 = max{ΔGi} − (2E H2 + EO2 − 2E H2O) 4

(6)

III. RESULTS AND DISCUSSION We were motivated to study Fe2O3 with an Al2O3 overlayer, since recent experiments show a favorable decrease of ∼0.1 V in overpotential when a thin layer of Al2O3 is deposited over the surface of Fe2O3.26,27 We find water oxidation is not favorable over pure Al2O3 (see slab in Figure 1a). This slab of pure Al2O3 represents a relatively thick Al2O3 overlayer (>1 nm) on top of Fe2O3. As seen in Table 1 (rows 1 and 3) and in Figure 3a, the overpotential required for catalysis over Al2O3 is much larger than that required over a pure Fe2O3 slab. This is explained by

This is a lower limit for the overpotential, since the geometries of the transition states are unknown, but is a good approximation, since in highly endergonic reactions the two are close. In recent work the energy barrier for a single water dehydration over Fe2O3 was calculated to be 0.93 eV,57 and in comparison with the thermodynamic limit of 1.82 V for the reaction potential,29 these highly endergonic reaction kinetic corrections to the overpotential are expected to be small. Another point is that one can calculate reaction free energies by setting the total energy of an O2 molecule such that 7240

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis

Figure 4. Schematic illustration of water splitting under anodic bias in a PEC cell with water oxidation over: (top) pure Fe2O3; (bottom) Al2O3 thin overlayer covered Fe2O3 with exposed Fe2O3 sites. Arrows indicate entering and leaving reactants, products, and charge carriers, Φ is the externally applied bias, ω is the solar photon frequency, and Δ is the increase in valence band edge position upon Al2O3 deposition.

Finally, we find that partial surface coverage of Fe2O3 with a thin Al2O3 layer reduces the overpotential for water oxidation. We covered 75% of the surface with a monolayer of Al2O3 using a 2 × 2 slab with 1/12 ML reactive sites (see Figures 1d and 2) and let the reaction occur over an Fe2O3 exposed surface (since we previously found it to be unfavorable over Al2O3). On comparison of rows 2 and 6 in Table 1 (we compare 2 × 2 slabs that have the same concentration of reactive sites, since the overpotential has a slight dependence on active site concentration and since the lower concentration of oxygen vacancy active sites is more realistic), and the cumulative free energies in Figure 3b, we see that in fact a small improvement is achieved upon partial Al2O3 coverage: the overpotential is reduced by 0.13 V upon partial surface coverage, in agreement with the experimental value of 0.1 V.26,27 This improvement can be explained by the higher valence band edge position at 75% coverage for the *OH intermediate that favors electron extraction during dehydrogenation and results in a lower free energy (ΔG3). As we have shown in previous work,60,61 the band edge positions of oxides depend on the positions of their constituent parent materials and play a key role in determining the feasibility of water splitting. Here the slab is composed of Al2O3, which has a higher valence band edge position than Fe2O3 by 0.1 eV (as seen in rows 1 and 3 in Table 1) and therefore the valence band edge of partially covered Fe2O3 is higher than that of pure Fe2O3 (by 0.3 eV, as seen by comparing rows 2 and 6 in Table 1). We calculate the projected density of states (not shown) and find that the valence band edge states are located on surface O atoms. This may allow holes to be driven to the active site with a much lower applied bias (as schematically described in Figure 4), which by definition lowers the overpotential, as we calculate. In the *O intermediate of partially covered Fe2O3, we find that the hole is indeed localized on an O atom of the Fe2O3 surface. That is, there is a missing electron (and proton) due to dehydrogenation (the Bader charge on O is −0.91e, less than the −1.21e of O ions in the bulk, and the Bader charge of Fe is

the fact that, as opposed to Fe, Al has only one dominant stable oxidation state: Al(+III). That is, in pure Fe2O3, the oxidation state of Fe can change from Fe(+III) to Fe(+II) easily, and therefore an electron localizes on the surface Fe atom in the *vac intermediate. The corresponding Bader charges of the Fe atoms are 1.81e and 1.38e (compared to Fe2O3 bulk 1.84e) in the *OOH and *vac intermediates, respectively. On the other hand, in pure Al2O3, the corresponding Bader charges on Al atoms are 2.47e and 2.17e (in comparison to an Al2O3 bulk Al3+ charge of 2.48e) in the *OOH and *vac intermediates, respectively. This results in a high free energy for reaction 5 (ΔG5 in Table 1, row 3) that corresponds to a transition from *OOH to *vac intermediates, which requires accepting an additional electron to the surface. Thus, water oxidation will not occur over areas with a thick coating of Al2O3. Next, we find the reaction is not favorable even when the Al2O3 deposition is thin. In previous experiments,26,27 measurements were performed on Fe2O3 covered by an Al2O3 overlayer with several thicknesses, some as thin as 0.1 nm (equivalent to a single monolayer). The samples were synthesized using a single cycle of atomic layer deposition (ALD). To model this thin coverage, we considered both monolayer and bilayer coverage of Al2O3 over Fe2O3 (see slabs in Figure 1b,c). The results, as seen in Table 1 (rows 4 and 5) and Figure 3a, have similar trends in either of the thin Al2O3 layers or the thick Al2O3 layer: the overpotential increases quite drastically in comparison to that for the pure Fe2O3. In the thin (mono- and bilayer) coating, the reaction is limited by the high energy (>1 eV above the conduction band edge) of Al 3p and 3s states that are unwilling to occupy electrons, and instead excess electrons in the *vac intermediate (the slab that contains an oxygen vacancy) reside on Fe atoms below Al2O3 (Bader charge is 1.42e over Fe2+ in comparison to the bulk 1.84e over Fe3+ ions). Furthermore, holes accumulated during dehydrogenation are also located at Fe atoms. Therefore, water oxidation is not likely to take place directly over Al2O3 but rather over the Fe2O3 surface exposed to water. 7241

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis 1.81e, similar to the 1.84e of Fe3+ ions in the bulk). Previous studies suggest choosing coating materials with different band edge positions in order to drive charge carriers toward the surface.61 Here we directly model water oxidation and show that the addition of an Al2O3 overlayer with a higher valence band edge position helps in extracting an electron during the second proton-coupled electron transfer reaction. The band edge positions are of course not the only factor determining functionality, since the active site must also include atoms that can readily change oxidation states, as in pure Fe2O3.

152/11). O.N. acknowledges the Irwin and Joan Jacobs graduate school scholarship, the Leonard and Diane Sherman Interdisciplinary Graduate School Fellowship for excellence, and the Rieger Foundation, Marshal-Tulin fellowship.



IV. CONCLUSIONS We modeled surface coverage of Fe2O3 with Al2O3 and found that the overpotential for water oxidation over Al2O3 is high: >1 V. Hence, we deduce that water does not oxidize directly on top of Al2O3 but would rather be oxidized over an Fe2O3 exposed surface. In addition, we find a decrease of 0.13 V in overpotential for water oxidation over Fe2O3 partially covered with Al2O3, in good agreement with previous experiments.26,27 The decrease in overpotential has been thought to arise from passivation of surface defects by the overlayer.26,27 In this paper we provide an additional explanation that contributes to the decrease in overpotential. We find that the higher valence band edge of the overlayer relative to the interior material could assist in catalysis of reactions that couple electron extraction (hole transfer). This is extremely important in the case of Fe2O3, since a large bias is required in order to drive a sufficient amount of holes to the surface and motivate deprotonation. The external required bias could be reduced by raising the valence band edge position with surface treatment. Further support for our model is given by experimentally measured decreases in overpotentials in In2O3 and Ga2O3,24,28 which also have a higher valence band edge than Fe2O3.62,63 Therefore, we suggest covering catalysts with materials that have different band edge positions in order to reduce the overpotential. Of course, surface coverage needs to be partial so that some catalytic sites are exposed. For water oxidation with Fe2O3, we offer several potential alternative overlayers that have higher valence band edge positions relative to Fe2O3: CuWO4, Co3O4, BiFeO3, and BiVO4.64 We anticipate that this strategy will be useful for designing better catalysts.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.5b01748. ZPE corrections and geometries obtained for pure Al2O3 in a hexagonal bulk structure and all other slab geometries for pure Fe2O3 and Al2O3-covered Fe2O3 (PDF)



REFERENCES

(1) Li, J.; Wu, N. Catal. Sci. Technol. 2015, 5, 1360−1384. (2) Barroso, M.; Cowan, A. J.; Pendlebury, S. R.; Grätzel, M.; Klug, D. R.; Durrant, J. R. J. Am. Chem. Soc. 2011, 133, 14868−14871. (3) Upul Wijayantha, K. G.; Saremi-Yarahmadi, S.; Peter, L. M. Phys. Chem. Chem. Phys. 2011, 13, 5264−5270. (4) Sivula, K.; Le Formal, F.; Grätzel, M. ChemSusChem 2011, 4, 432−449. (5) Lin, Y.; Yuan, G.; Sheehan, S.; Zhou, S.; Wang, D. Energy Environ. Sci. 2011, 4, 4862−4869. (6) Liao, P.; Carter, E. A. J. Phys. Chem. C 2011, 115, 20795−20805. (7) Yang, Y.; Ratner, M. A.; Schatz, G. C. J. Phys. Chem. C 2014, 118, 29196−29208. (8) Yang, Y.; Ratner, M. A.; Schatz, G. C. J. Phys. Chem. C 2013, 117, 21706−21717. (9) Townsend, T. K.; Sabio, E. M.; Browning, N. D.; Osterloh, F. E. Energy Environ. Sci. 2011, 4, 4270−4275. (10) Kronawitter, C. X.; et al. Energy Environ. Sci. 2014, 7, 3100− 3121. (11) Lin, Y.; Yuan, G.; Liu, R.; Zhou, S.; Sheehan, S. W.; Wang, D. Chem. Phys. Lett. 2011, 507, 209−215. (12) Kim, J. Y.; Magesh, G.; Youn, D. H.; Jang, J.-W.; Kubota, J.; Domen, K.; Lee, J. S. Sci. Rep. 2013, 3, 2681. (13) Ling, Y.; Wang, G.; Wheeler, D. A.; Zhang, J. Z.; Li, Y. Nano Lett. 2011, 11, 2119−2125. (14) Liao, P.; Caspary Toroker, M.; Carter, E. A. Nano Lett. 2011, 11, 1775−1781. (15) Deng, J.; Lv, X.; Gao, J.; Pu, A.; Li, M.; Sun, X.; Zhong, J. Energy Environ. Sci. 2013, 6, 1965−1970. (16) Li, J.; Meng, F.; Suri, S.; Ding, W.; Huang, F.; Wu, N. Chem. Commun. 2012, 48, 8213−8215. (17) Barroso, M.; Pendlebury, S. R.; Cowan, A. J.; Durrant, J. R. Chem. Sci. 2013, 4, 2724−2734. (18) Barroso, M.; Mesa, C. A.; Pendlebury, S. R.; Cowan, A. J.; Hisatomi, T.; Sivula, K.; Grätzel, M.; Klug, D. R.; Durrant, J. R. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 15640−15645. (19) Rosso, K. M.; Smith, D. M. A.; Dupuis, M. J. Chem. Phys. 2003, 118, 6455−6466. (20) Young, K. M. H.; Klahr, B. M.; Zandi, O.; Hamann, T. W. Catal. Sci. Technol. 2013, 3, 1660−1671. (21) Cummings, C. Y.; Marken, F.; Peter, L. M.; Tahir, A. A.; Wijayantha, K. G. U. Chem. Commun. 2012, 48, 2027−2029. (22) Pendlebury, S. R.; Barroso, M.; Cowan, A. J.; Sivula, K.; Tang, J.; Gratzel, M.; Klug, D.; Durrant, J. R. Chem. Commun. 2011, 47, 716− 718. (23) Pendlebury, S. R.; Cowan, A. J.; Barroso, M.; Sivula, K.; Ye, J.; Gratzel, M.; Klug, D. R.; Tang, J.; Durrant, J. R. Energy Environ. Sci. 2012, 5, 6304−6312. (24) Steier, L.; Herraiz-Cardona, I.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Tilley, S. D.; Grätzel, M. Adv. Funct. Mater. 2014, 24, 7681−7688. (25) Klahr, B.; Gimenez, S.; Fabregat-Santiago, F.; Bisquert, J.; Hamann, T. W. J. Am. Chem. Soc. 2012, 134, 16693−16700. (26) Le Formal, F.; Tétreault, N.; Cornuz, M.; Moehl, T.; Grätzel, M.; Sivula, K. Chem. Sci. 2011, 2, 737−743. (27) Klahr, B.; Hamann, T. J. Phys. Chem. C 2014, 118, 10393− 10399. (28) Hisatomi, T.; Le Formal, F.; Cornuz, M.; Brillet, J.; Tetreault, N.; Sivula, K.; Gratzel, M. Energy Environ. Sci. 2011, 4, 2512−2515. (29) Liao, P.; Keith, J. A.; Carter, E. A. J. Am. Chem. Soc. 2012, 134, 13296−13309. (30) Nguyen, M.-T.; Seriani, N.; Piccinin, S.; Gebauer, R. J. Chem. Phys. 2014, 140, 064703.

AUTHOR INFORMATION

Corresponding Author

*E-mail for M.C.T.: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This research was supported by the Morantz Energy Research Fund, the Nancy and Stephen Grand Technion Energy Program, the I-CORE Program of the Planning and Budgeting Committee, and The Israel Science Foundation (Grant No. 7242

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243

Research Article

ACS Catalysis (31) Hellman, A.; Pala, R. G. S. J. Phys. Chem. C 2011, 115, 12901− 12907. (32) Trainor, T. P.; Chaka, A. M.; Eng, P. J.; Newville, M.; Waychunas, G. A.; Catalano, J. G.; Brown, G. E., Jr Surf. Sci. 2004, 573, 204−224. (33) Nguyen, M.-T.; Piccinin, S.; Seriani, N.; Gebauer, R. ACS Catal. 2015, 5, 715−721. (34) Toroker, M. C. J. Phys. Chem. C 2014, 118, 23162−23167. (35) Neufeld, O.; Toroker, M. C. J. Phys. Chem. C 2015, 119, 5836− 5847. (36) Kresse, G.; Hafner, J. Phys. Rev. B: Condens. Matter Mater. Phys. 1993, 47, 558−561. (37) Kresse, G.; Furthmüller, J. Comput. Mater. Sci. 1996, 6, 15−50. (38) Dudarev, S. L.; Botton, G. A.; Savrasov, S. Y.; Humphreys, C. J.; Sutton, A. P. Phys. Rev. B: Condens. Matter Mater. Phys. 1998, 57, 1505−1509. (39) Anisimov, V. I.; Aryasetiawan, F.; Lichtenstein, A. I. J. Phys.: Condens. Matter 1997, 9, 767. (40) Mosey, N. J.; Liao, P.; Carter, E. A. J. Chem. Phys. 2008, 129, 014103. (41) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. Rev. Lett. 1996, 77, 3865−3868. (42) Ilyasov, V. V.; Ershov, I. V. Phys. Solid State 2012, 54, 2335− 2343. (43) Benam, M. R.; Rahnamaye Aliabad, H. A.; Hosseini, S. M. Phys. Status Solidi A 2006, 203, 2223−2228. (44) Liao, P.; Carter, E. A. Phys. Chem. Chem. Phys. 2011, 13, 15189− 15199. (45) Blöchl, P. E. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50, 17953−17979. (46) Kresse, G.; Joubert, D. Phys. Rev. B: Condens. Matter Mater. Phys. 1999, 59, 1758−1775. (47) Blöchl, P. E.; Jepsen, O.; Andersen, O. K. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 49, 16223−16233. (48) Lehmann, G.; Taut, M. Phys. Status Solidi B 1972, 54, 469−477. (49) Finger, L. W.; Hazen, R. M. J. Appl. Phys. 1980, 51, 5362−5367. (50) Lee, W. E.; Lagerlof, K. P. D. J. Electron Microsc. Tech. 1985, 2, 247−258. (51) Lad, R. J.; Henrich, V. E. Surf. Sci. 1988, 193, 81−93. (52) Momma, K.; Izumi, F. J. Appl. Crystallogr. 2011, 44, 1272−1276. (53) Bolding, B. C.; Carter, E. A. Mol. Simul. 1992, 9, 269−283. (54) Rossmeisl, J.; Qu, Z. W.; Zhu, H.; Kroes, G. J.; Nørskov, J. K. J. Electroanal. Chem. 2007, 607, 83−89. (55) Bajdich, M.; García-Mota, M.; Vojvodic, A.; Nørskov, J. K.; Bell, A. T. J. Am. Chem. Soc. 2013, 135, 13521−13530. (56) CRC Handbook of Chemistry and Physics, 77th ed.; CRC Press: Boca Raton, FL, 1996−1997. (57) Pan, H.; Meng, X.; Qin, G. Phys. Chem. Chem. Phys. 2014, 16, 25442−25448. (58) Valdes, A.; Qu, Z.-W.; Kroes, G.-J.; Rossmeisl, J.; Nørskov, J. K. J. Phys. Chem. C 2008, 112, 9872−9879. (59) Henkelman, G.; Arnaldsson, A.; Jónsson, H. Comput. Mater. Sci. 2006, 36, 354−360. (60) Toroker, M. C.; Kanan, D. K.; Alidoust, N.; Isseroff, L. Y.; Liao, P.; Carter, E. A. Phys. Chem. Chem. Phys. 2011, 13, 16644−16654. (61) Toroker, M. C.; Carter, E. A. J. Mater. Chem. A 2013, 1, 2474− 2484. (62) Pan, C. A.; Ma, T. P. Appl. Phys. Lett. 1980, 37, 714−716. (63) Fu, L.; Liu, Y.; Hu, P. a.; Xiao, K.; Yu, G.; Zhu, D. Chem. Mater. 2003, 15, 4287−4291. (64) Chen, S.; Wang, L.-W. Chem. Mater. 2012, 24, 3659−3666.

7243

DOI: 10.1021/acscatal.5b01748 ACS Catal. 2015, 5, 7237−7243