Oxygen Reduction in Alkaline Media—a Discussion

Abstract

We propose a complete reaction sequence for oxygen reduction in alkaline solutions, in which the first two steps occur in the outer sphere mode. The oxygen-oxygen bond is broken in the third step, which involves adsorption of OH, which is desorbed in the last step. We have investigated the sequence by quantum-chemical methods and determined the energies of activation. Whether the reaction follows a four- or a two-electron mechanism, depends critically on the energy of adsorption of OH. We surmise that our mechanism holds on all electrodes which interact weakly with oxygen, in particular on gold, silver, and graphite. We explain, why Au(100) is a better catalyst than Au(111), why at high overpotentials the reaction on Au(100) reverts to a two-electron mechanism, and why this does not happen on silver.

Appendix: Theoretical background

Technical Details of the DFT Calculations

Oxygen on Silver, Platinum, and Gold Periodic density functional theory (DFT) calculations were performed using the DACAPO [43] code with implemented Vanderbilt [44] ultrasoft pseudopotentials for the representation of the atomic cores. A PBE (Perdew, Burke, Ernzerhof) [45] functional and a set of plane waves with a cutoff energy of 350 eV (400 eV for the density) were chosen to describe the valence electrons. Brillouin zone integration [46] was performed using 4 k-points in the x- and y-directions respectively and 1 k-point in the z-direction. The surface was represented by 3 layers of metal atoms, and a 3×3 unit cell was employed.

OHonGold All calculations were performed using the VASP code [47, 48]. The correlation and exchange functionals were described within the generalized gradient approximation (GGA) in the Perdew, Burke and Ernzerhof (PBE) flavor [45]. The electron-ion interactions were represented through ultrasoft pseudopotentials [44], and a plane wave basis set was used to describe the valence electrons. The basis set was expanded to a kinetic energy cutoff of 500 eV. Brillouin zone integration was performed using (10x10x1) k-point MonkhorstPack [49] grid . We used a dipole-correction scheme [50] to avoid slab-slab interactions. To investigate the adsorption of OH as a function of the coverage, both surfaces Au(100) and Au(111) were modelled by a (2 x 2) supercell with four metal layers. In all the calculations 15 A of vacuum were considered. For the relaxations, the two bottom layers were fixed at the calculated nearest-neighbour distance corresponding to bulk, and all the other layers plus the OH were allowed to relax. To mimic the aqueous media, a water layer was considered on the previous systems at each coverage. In order to take into account van der Waals interactions, the DFT-D3 [51] approach of Grimme [52] was used, which consists of adding a semiempirical dispersion potential to the conventional Kohn–Sham DFT energy.

Reaction 6 The DFT calculations were performed using the b3lyp functional as implemented in the Gaussian 09 program suite [53]; the standard 6-311++g(d, p) basis set was employed to describe the O and H atoms. The spin-polarized formalism was used to treat open shell molecules. Local hydration was considered explicitly by including of several water molecules into the nearest solvation sheath of the ions, while long-range solvent effects were addressed by the Polarized Continuum Model (PCM) taking a value of 78 as static dielectric constant. The molecular geometry in initial and final states was fully optimized without any restrictions.

Breaking of the Oxygen-Oxygen Bond These calculations were performed using the Gaussian 09 suite [53]. The Au(100) surface was modeled a the metal cluster composed of two layers of gold atoms (16+8) arranged according to the fcc structure typical for gold with the nearest-neighbor distances Au-Au fixed at the experimental value of 2.88 Å, and was kept unchanged in all calculations. The H ₂O OOH ⁻ system was first optimized in the bulk solution and then placed above the Au ₂₄ cluster, as shown in Fig. 6. The potential energy surface presented in Fig. 7 was then obtained by systematically stretching the O–O bond of the OOH ⁻ ion and optimizing other parameters of the system undergoing adsorption. In the potential energy scan, some constraints were applied: the metal cluster was kept rigid and the O–O bond of the OOH ⁻ anion was kept always perpendicular to the surface at the central bridge site of the Au24 cluster. In all calculations, the PBE1PBE hybrid functional [45] was used together with the 6-31++G(d,p) basis set for the H ₂O OOH ⁻ system, the pseudopotential LANL1DZ for the metal cluster, and the polarizable continuum model (PCM) for the solvent (water).

Technical Details of the Molecular Dynamics

To calculate the PMF canonical ensemble (constant NVT) steered-molecular dynamics simulation was conducted for 1 ns at 298 K on a simulation box containing a Ag(100) slab with three metal layers (thickness 4.09 Å), an ensemble of 470 water molecules, and the O ₂ or O$_{2}^{-}$ species. Previously, an equilibration run of 700 ps was carried out. Periodic boundary conditions were set in the x and y directions, and the Ewald summation method was used to handle with long- range electrostatic interactions.

Well-known 12-6 Lennard-Jones pairwise potential was used to model the interactions between the species. For the water, we used the SPC/E (extended simple point charge) model and the corresponding parameters for the oxygen and hydrogen were taken from Yoshida et al. [54]. The Lennard-Jones parameters for silver were taken from Agrawal et al. [55] and for the O ₂ and the O$_{2}^{-}$ species the parameters were taken from Poling et al. [56] and Shen et al. [57], respectively. The cross interactions were computed through the Lorentz-Berthelot mixing rules, $\epsilon _{ij} = \sqrt {\epsilon _{ii}\epsilon _{jj}}$ , and σ _ij = (σ _ii + σ _jj )/2.

All simulations were performed using the LAMMPS (large-scale atomic/molecular massively parallel simulator) code [58] with a time step equal to 2.0 fs. The average temperature of 298 K was maintained by using a Nose-Hoover thermostat with a relaxation time of 0.1 ps.