Chemical Environment Effects on the Atomic Oxygen Absorption into

17388

J. Phys. Chem. C 2007, 111, 17388-17396

Chemical Environment Effects on the Atomic Oxygen Absorption into Pt(111) Subsurfaces Zhihui Gu and Perla B. Balbuena* Department of Chemical Engineering, Texas A&M UniVersity, College Station, Texas 77843 ReceiVed: June 25, 2007; In Final Form: September 10, 2007

Density functional theory is used to study chemical environment effects such as surface coverage of oxygenated species and presence of adsorbed oxygen reduction intermediate OH and water on the atomic oxygen absorption process into Pt(111) subsurfaces. The results show that increasing the on-surface oxygen coverage can energetically and kinetically stabilize subsurface atomic oxygen present at 0.11 and 0.25 monolayer subsurface coverage. At constant subsurface coverage, the absorption energy barrier increases with increasing on-surface coverage. Adsorbed on-surface OH has similar effects on the atomic oxygen absorption process as those caused by on-surface adsorbed atomic oxygen. A monolayer of water over the on-surface oxygen does not significantly alter the thermodynamics and kinetics of the oxygen absorption process.

1. Introduction Electrocatalyst durability is one of the challenges for longterm application of polymer electrolyte membrane fuel cells (PEMFC). Performance decrease of PEMFCs has been attributed in part to the loss of active surface of the platinum cathode electrocatalyst under steady-state and cycling conditions.1-3 One of the proposed mechanisms for active surface loss of PEMFC cathode catalysts is the oxidation and dissolution of platinum in acidic environment at a certain potential range,4,5 where the equilibrium concentration of dissolved species can be monitored as a function of voltage and temperature.6 Such measurements have confirmed the presence of dissolved Pt in the cathode outlet water.1 The dissolved platinum species can redeposit on the Pt particles to form larger particles following an Ostwald ripening process1,7 and also can diffuse out of the membrane electrode assembly, becoming deposited on the polymer membrane.2,7-9 Our previous thermodynamic investigation showed that the dissolution reactions are electrochemical reactions, and the same study yielded the influence of the alloy composition on the relative stabilities against dissolution of platinum for platinumbased alloys.10 The observation of an apparent connection between oxide growth stages detected approximately between 0.8 and 1.2 V (measured with respect to the standard hydrogen electrode,SHE, potential) and the dissolved Pt concentration for a Pt electrode in acid medium has been reported.6,11 It was shown that the measured equilibrium concentration of dissolved platinum in perchloric acid electrolyte increases for potentials greater than 0.6 V versus SHE and decreases for potentials greater than 1.1 V under potentiostatic conditions.11 In the range of potentials from 0.85 to 1.15 V, half a monolayer of oxygen atoms is adsorbed on the cathode catalyst,12,13 and from 1.15 to 1.2 V, a second half oxygen atom monolayer is formed on the catalyst surface; it is suggested that a place exchange of adsorbed oxygen atoms with the surface platinum atoms takes place with further formation of an adsorbed oxygen layer.12,14 This place exchange would make the oxygen atoms diffuse into the subsurface, leaving the platinum atoms exposed to the electrolyte for further oxidation and dissolution. It has been reported that * Corresponding author: e-mail [email protected].

atomic oxygen can be found in the subsurface of transition metals during surface reactions in various experimental environments15-20 and that subsurface oxygen can affect the reactivity of a catalytic surface.21-23 Theoretical calculations of subsurface oxygen underneath different transition metals were reported recently.24-26 Subsurface oxygen may further be considered as the initial stage in the corrosion and growth of metal oxides.27 Under electrochemical conditions, subsurface oxygen can be formed in the oxidation of polycrystalline platinum, and the concentration of subsurface oxygen has been found to depend on the cycling rate by cyclic voltammetry experiments.28,29 In our previous report we characterized the preferred adsorption sites of an oxygen atom on Pt(111) and Pt(100) surfaces as a function of coverage, the preferred sites for absorption of atomic oxygen inside Pt(111) and Pt(100) subsurfaces, and the migration process of an oxygen atom into the subsurface at low on-surface coverage.30 Summarizing, spinpolarized density functional theory (DFT) calculations showed that the most stable adsorption site on the Pt(111) surface is the face-centered cubic (fcc) hollow, and the most stable Pt(111) subsurface site is the tetrahedral (tetra-II) site located underneath a hexagonal close-packed (hcp) hollow site.30 Atomic oxygen endothermic absorption into the Pt(111) subsurface at 0.25 ML coverage follows this minimum energy path: it migrates from an fcc site to the nearest hcp site with an energy barrier of 0.63 eV where the transition-state geometry has the oxygen placed at a bridge site; from the hcp site, the oxygen atom then diffuses into the subsurface tetra-II site with an energy barrier of 2.22 eV and a corresponding reverse barrier of 0.70 eV, with the atomic oxygen in hcp site as the transition state.30 At lower coverage of 0.11 ML, the absorption energy barrier on Pt(111) decreases; however, the reverse energy barrier is close to zero.30 In this work we investigate chemical environment effects on the atomic oxygen absorption process into the subsurface, including different on-surface oxygen and OH coverage and the presence of a water monolayer over the on-surface adsorbed oxygen. It should be noted that the systems in this study resemble single-crystal surfaces exposed to gas-phase atomic oxygen and not to an electrochemical environment. Nonetheless, we expect that insights derived from this investigation of atomic

10.1021/jp074948s CCC: $37.00 © 2007 American Chemical Society Published on Web 10/25/2007

Atomic Oxygen Absorption into Pt(111) Subsurfaces

J. Phys. Chem. C, Vol. 111, No. 46, 2007 17389

Figure 1. Energy difference between systems with and without subsurface oxygen (∆E, in electronvolts) as a function of total coverage. Subsurface oxygen coverage was (blue squares) 0.11 ML or (pink triangles) 0.25 ML. Symbols represent calculated points; lines have been drawn to help visualization.

oxygen absorption into the subsurface may help to clarify the initial stages of oxide growth and their relation to the metal dissolution process, in the same way that calculations on singlecrystal surfaces have proved to be fundamental to understand oxygen reactivity and to guide design of electrocatalysts.31,32 2. Computational Details Spin-polarized density functional theory (DFT) was employed for all calculations within the generalized gradient approximation (GGA) by use of the PBE functional33 for electron exchange and correlation. The Vienna Ab Initio Simulation Package (VASP)34,35 was used to solve the Kohn-Sham equations in a simulation cell with periodic boundary conditions using planewave basis sets in which the Blo¨chl’s all-electron projector augmented (PAW) method36 with frozen core approximation is implemented.37 An energy cutoff of 400 eV is used for all calculations, the convergence criterion for wave function optimization is 10-4 eV, and the convergence criterion for structure relaxation is that the force acting on each atom becomes less than 0.01 eV/Å. The Monkhorst-Pack scheme38 was employed for k-point sampling, and the calculated bulk equilibrium lattice constant of fcc Pt is 3.977 Å with converged k-mesh of 14 × 14 × 14, which agrees within 1.45% with the experimental value of 3.92.39 The platinum surfaces are modeled in the supercell approach by using a five-layer Pt slab in which the top three layers are allowed to relax and the bottom two layers are fixed at the calculated bulk positions. A vacuum space approximately equal to 12 Å is left above the top layer in all cases. The surface species were adsorbed on the most stable fcc hollow site, and subsurface oxygen was lodged in the tetrahedral site underneath the Pt(111) hcp hollow site, based on our previous investigation.30 The climbing image nudged elastic band method (cNEB)40-42 was employed to investigate the saddle points and minimum energy paths (MEPs) for oxygen absorption into the Pt subsurface from surfaces at different total coverage. In this report (2 × 2) and (3 × 3) cells corresponding to 0.25 and 0.11 monolayer (ML) subsurface oxygen coverage were used to study the minimum energy paths and transition states of oxygen

absorption into the subsurface, in which Monkhorst-Pack grids of 5 × 5 × 1 and 3 × 3 × 1 were used for calculations of 0.25 and 0.11 ML, respectively. 3. Results and Discussion 3.1. Atomic Oxygen Coverage Effect on Oxygen Absorption. As discussed in the Introduction, we had found that the diffusion process from surface to subsurface yielding 0.11 and 0.25 ML subsurface coverage at zero on-surface atomic oxygen is energetically unfavorable, and the energy barrier for such a migration process is lower at the lowest subsurface coverage.30 In this work we calculated (Figure 1) the surface coverage effect on the energetics of the absorption process at the same subsurface coverage of 0.11 and 0.25 ML. We define ∆E as the energy difference between two systems with the same total oxygen coverage: the first system (energy Esub) has a certain oxygen coverage in the subsurface and the difference (total coverage - subsurface coverage) on the surface, and the second system (energy Eon) has all the oxygen coverage on the surface, ∆E ) Esub - Eon. Figure 1 (data in Table 6 of Supporting Information) shows that ∆E decreases with increasing on-surface coverage (total coverage - subsurface coverage) for both 0.11 and 0.25 ML subsurface coverage, which indicates that increasing on-surface oxygen coverage stabilizes the subsurface oxygen to some extent. The slopes of the curves in Figure 1, however, show different behaviors for the two subsurface coverage values. At subsurface coverage of 0.11 ML, ∆E drops fast from 1.19 to 0.39 eV, going from total coverage 0.67 to 0.78 ML, and it becomes negative at 1.00 ML total coverage (on-surface coverage 0.89 ML). At total coverage smaller than 0.67 ML, the decrease of ∆E is less pronounced than that between 0.67 and 0.78 ML. Therefore, at 0.11 ML subsurface coverage, the on-surface oxygen stabilizing effect on subsurface oxygen starts to become apparent when the total oxygen coverage is g0.78 ML (on-surface oxygen coverage g0.67 ML). In contrast, at 0.25 ML subsurface oxygen, ∆E decreases faster at small total coverage than at comparatively higher coverage (a drop of 1.2 eV from 0.25 to 0.50 ML compared to a decrease of 0.68 eV from 0.50 to 0.75 ML); thus the on-

17390 J. Phys. Chem. C, Vol. 111, No. 46, 2007 TABLE 1: Vertical Distance between Subsurface Oxygen Atoms and Pt Atoms in the Top Layera coverageb/ML 0.00/0.11 0.11/0.22 0.22/0.33 0.33/0.44 0.44/0.55 DPt-O (Å) 0.37 0.40 0.59 0.86 0.88 a With 0.11 ML oxygen in the subsurface at various values of total atomic oxygen coverage. b Surface coverage/total coverage.

surface oxygen stabilizing effect on subsurface oxygen increases faster at total coverage smaller than 0.50 ML. The stabilizing effect on subsurface oxygen at 0.25 ML (Figure 1) becomes clear with the surface coverage increasing, as indicated by the decreasing ∆Es, and it is found that at a total coverage of 1 ML (0.75 surface coverage) the absorption process becomes energetically favorable. Upon comparison of ∆E for the two values of subsurface oxygen coverage (Figure 1) at similar onsurface oxygen coverage, it is found that the subsurface oxygen is more stable at 0.25 ML than at 0.11 ML; for example, ∆E is -0.35 eV with 0.75 ML on-surface oxygen at 0.25 ML subsurface oxygen, whereas the ∆E value is positive (0.39 eV) at 0.11 ML subsurface coverage and 0.78 ML on-surface coverage. In summary, the on-surface oxygen stabilizes the subsurface oxygen energetically and the effect becomes more marked at higher on-surface oxygen coverage. The rate of increase of such stabilizing effect is faster at 0.25 ML than at 0.11ML subsurface coverage. Our calculated energy difference (∆E) shown in Figure 1 (and Table 6 of Supporting Information) between the 0.75/0.25 (total coverage/subsurface coverage) and 0.5/0.0 systems is 0.13 eV. This value can be compared with the reported value of 0.24 eV in ref 24, calculated on the basis of the binding energies reported in Table 1 and by application of eq 1 of ref 24 to the 0.75/0.25 and 0.5/0.0 coverage conditions. Additional comparison with Legare’s work was done by calculating the binding energies relative to O2 as done in ref 24. We found binding energies of 1.24 eV for 0.25 ML on-surface oxygen, 0.85 eV for 0.5 ML on-surface oxygen, and 0.46 eV for 0.5 ML on-surface and 0.25 ML subsurface oxygen. In the first two cases, the binding energies are higher than Legare’s results of 1.03 and 0.73 eV, respectively, whereas the third binding energy is lower than the value of 0.65 eV reported in ref 24. These differences between our results and those from ref 24 may originate from the different functionals used in our work (PBE) and in Legare’s work (PW91); the different performance of these functionals was recently discussed by Mattsson et al.43 Figure 2 shows the minimum energy paths for 0.25 ML oxygen migration into the Pt(111) subsurface at two different total oxygen coverage valuess: 0.25 and 0.5 ML. As reported earlier,30 the oxygen first migrates from the fcc hollow site (point A) to the hcp hollow site (point B) through a bridge site, and then the oxygen atom begins to diffuse into the subsurface site (tetrahedral site beneath hcp hollow site, point C). A first transition state is located as the atomic oxygen migrates to the hcp hollow site (Figure 2). With the total oxygen coverage increasing from 0.25 to 0.5 ML layer, the energy barrier of onsurface diffusion from fcc to hcp site also increased from 0.61 to 0.73 eV, whereas the reverse energy barrier decreased from 0.20 to 0.14 eV. The barrier for diffusion from fcc to hcp is in good agreement with earlier reports.44 Thus, with the total coverage increasing, the on-surface diffusion energy barrier increased and the reverse energy barrier decreased, which indicates that higher total oxygen coverage makes the oxygen diffusion to the hcp site more difficult and unstable from kinetic and energetic points of view. For the oxygen migration from the hcp site into the subsurface tetrahedral site, the energy barrier at 0.50 ML total oxygen coverage is 3.18 eV, which is much

Gu and Balbuena higher than that at 0.25 ML total oxygen coverage (2.22 eV). The reverse energy barrier of absorption at 0.50 ML total coverage is 2.98 eV, much larger than that at 0.25 ML total coverage (0.70 eV). From these observations, increasing the total oxygen coverage produces a huge increase of the energy barrier for migration from hcp to the subsurface. On the other hand, the reverse energy barrier increased with increasing total oxygen coverage so that the subsurface oxygen could be kinetically more stable at higher coverage. The increase of the energy barrier for the forward reaction (from hcp site B in Figure 2 to the transition state between B and C) at higher total oxygen coverage can be explained by the cost associated with a larger lattice deformation (see Table 1 of Supporting Information) required at higher total coverage. For example, at 0.50 ML total coverage, the change in each Pt-Pt bond length for the Pt atoms coordinated to O at the hcp and at the transition state sites is 0.362 Å; that is,oin average the bond lengths change from 2.794 to 3.156 Å, whereas the equivalent change at 0.25 ML is 0.333 Å. Further analysis shows additional deformation of the vertical distance between the top Pt layer and the second layer. Table 1 of Supporting Information shows a larger deformation of 0.341 Å between the top and second layers at 0.5 ML than that of 0.201 Å at 0.25 ML. Therefore, the energy barrier increase also results from the high energy cost associated with the deformation of the interlayer distance between the top and second layers. Substantial lattice deformation is also required for the reverse process, diffusion of O from the subsurface site C to the transition state in Figure 2: the change in Pt-Pt bond length at 0.50 ML is 0.297 Å, compared with 0.091 Å at 0.25 ML. Also, at 0.50 ML the oxygen lodges deeper in the subsurface (1.293 Å from the Pt top layer) than at 0.25 ML (1.215 Å); this fact together with on-surface increased repulsive O-O interactions will make the reverse process more difficult than at the lower 0.25 ML total coverage. Figure 3 shows the MEPs for migration of oxygen at 0.11, 0.22, 0.33, 0.44 and 0.55 ML total coverage in all cases yielding 0.11 ML oxygen absorption into the Pt(111) subsurface tetrahedral site. For 0.11 and 0.22 ML total coverage, the MEPs and the energy barriers for on-surface oxygen diffusion and absorption into the subsurface are almost identical; this indicates that a low total oxygen coverage,