Research Article pubs.acs.org/acscatalysis
A Unified Picture of Water Oxidation on Bare and Gallium OxideCovered Hematite from Density Functional Theory Kanchan Ulman,*,† Manh-Thuong Nguyen,‡ Nicola Seriani,† Simone Piccinin,§ and Ralph Gebauer† †
The Abdus Salam International Centre for Theoretical Physics, Strada Costiera 11, 34151 Trieste, Italy Center for Computational Physics, Institute of Physics, Vietnam Academy of Science and Technology, 10 Dao Tan Street, Hanoi, Vietnam § CNR-IOM DEMOCRITOS, c/o SISSA, Via Bonomea 265, 34136 Trieste, Italy ‡
S Supporting Information *
ABSTRACT: Hematite is a promising catalyst for the photoelectrochemical water oxidation reaction, which however displays a low overall efficiency. To improve it, a systematic understanding of the underlying photocatalytic mechanisms is desirable but difficult to obtain by experimental techniques alone. Here, we have investigated the oxidation of water on the most stable terminations of the bare hematite, as well as in the presence of a monolayer of Ga2O3, by first-principles density functional theory-based methods. Although several surface terminations are very close in surface energy, they all yield a very similar overpotential of ∼0.8 V on the bare surface and ∼0.95 V on the Ga2O3-covered surface. Moreover, on all the relevant terminations, the overpotential-determining reaction step is the same, involving the dehydrogenation of a surface-adsorbed hydroxyl species. The reaction mechanism is largely independent of the atomistic details of the surface termination and crucially involves the formation of reaction intermediates involving lattice oxygen bound to adsorbed oxygen from water (O*−Os). It seems likely not only that different surface terminations coexist but also that they transform into one another during reaction conditions, even at the steady state. We also shed light on the important role of midgap states of hematite in the water oxidation cycle. In presence of the O*−Os species, midgap states that localize on this species can act as long-lived hole traps and fall prey to parasitic recombination processes. Without the O*−Os species, the surface states of hematite compete with the states on the active water oxidation species to attract the holes. The Ga2O3 layer passivates these competing surface states, making holes available at the active sites for water oxidation. Though Ga2O3-covered hematite has an overpotential higher than that of bare hematite, it nevertheless plays a very important role in making the holes available at the active site. KEYWORDS: photocatalysis, water oxidation, hematite, proton-coupled electron transfer, surface trap states, oxygen evolution reaction
1. INTRODUCTION
endothermic under standard conditions and becomes energetically favorable at an applied bias (E0) of 1.23 V with respect to the standard hydrogen electrode (SHE). This reaction can however take place only in the presence of a (photo)catalyst, and much of the current research is focused on the search for efficient photocatalysts. In fact, in a photoelectrochemical cell, water splitting is performed in two half-reactions, the oxygen evolution reaction (OER) taking place at the photoanode and the hydrogen evolution reaction (HER) taking place at the cathode. The two reactions can be written as 2H2O → O2 + 4H+ + 4e− and 4H+ + 4e− → 2H2, respectively. The iron oxide α-Fe2O3 (hematite) is a promising candidate material as a photoanode for the PEC water splitting reaction. It has a favorable band gap of ∼2.1 eV, therefore lying in the visible range, is stable in aqueous environments, and is abundant in
The efficient exploitation of solar energy is essential for securing the sustainable development of our world. Today, electricity production through photovoltaic devices is a standard procedure that is technically feasible and economically viable, and in fact, it is spreading steadily. However, there are two challenges that need to be addressed. First, the sun provides energy in an irregular manner, as irradiation often changes quickly, for example, because of clouds. Second, electricity is not an optimal carrier for many applications, for example, in transportation, where liquid and gaseous hydrocarbon fuels dominate. Both issues could be solved if solar energy were employed to produce fuels through chemical reactions,1 as hydrocarbons are easy to store and use on demand. In this context, one key reaction is the production of hydrogen from water through photoelectrochemical (PEC) water splitting. Hydrogen can then be used as fuel or as a reactant for the hydrogenation of carbon dioxide to hydrocarbons. The water splitting reaction (2H2O → O2 + 2H2) is © 2017 American Chemical Society
Received: November 5, 2016 Revised: January 20, 2017 Published: January 24, 2017 1793
DOI: 10.1021/acscatal.6b03162 ACS Catal. 2017, 7, 1793−1804
Research Article
ACS Catalysis nature.2 However, hematite has some limitations; e.g., it exhibits a low hole mobility3,4 and has high recombination rates and low lifetimes of the photogenerated electron−hole pairs.5 Several attempts have been made to mitigate these problems; for example, nanostructuring helps to reduce the mean free path lengths for holes and leads to reduced recombination rates for the photogenerated electron−hole pairs.6−11 Strategies like n-doping of hematite with Ti, Si, Ge, etc.,6,12 can help increase the carrier mobility of hematite. Another problem of hematite is the position of the electronic bands in the material: while the alignment of the valence band of hematite with the redox levels of H2O/O2 allows the OER to take place at the surface of hematite without any applied bias, the level alignment for H2/H+ is higher in energy than the bottom of the conduction band of hematite by ∼0.2−0.4 eV.2,13 Thus, the HER at the counter electrode requires an external bias. To solve this problem, surface modification of Ti-doped hematite with fluoride can be employed as it realigns the bands so that the entire water splitting reaction, i.e., both the HER and the OER, can take place without any applied bias.14 Despite these advances, the kinetics of the OER at the hematite surface remains slow, with even the best ultrathin films of hematite exhibiting a large overpotential of 0.5−0.6 V for the OER.2,6,15 These problems call for a more systematic understanding of the overall photocatalytic reaction, which could lead to better approaches for overcoming the limitations of this material. It is therefore very important to gain insight into the details of the water oxidation reaction at the hematite surface, to understand the origin of this overpotential as a prerequisite toward efforts to reduce it. From an experimental point of view, it is however very difficult to obtain information about the reaction mechanism at the atomistic level in operando, because of the complexity of the system, with an interface between a nanostructured solid and a liquid. For this reason, many ab initio studies have been conducted to provide microscopic information about the process, concentrating on the most common (0001) surface. These studies have employed the Nørskov approach16 to study (photo)electrochemical reactions by ab initio methods. A number of results are in agreement with experiments. Trainor et al. showed that the hematite (0001) surface prefers to be completely hydroxylated in an aqueous environment,17 while Carter et al. estimated an overpotential of 0.77 V for the water oxidation reaction at the hydroxylated surface, again in agreement with experiments.18 That study also shows that the water oxidation reaction proceeds via the formation of a hydroperoxo (−OOH) reaction intermediate. However, as mentioned by Hellmann and Pala19 and later confirmed20 by some of the authors of the work presented here, considering a more realistic PEC environment, i.e., in water under illumination and in the presence of a suitable cathode electrically connected to the hematite anode, the completely hydroxylated termination is not the most stable, but the oxygen-terminated surface becomes thermodynamically favored. Nguyen et al.20 also estimated the overpotential and found it to have a relatively high value of 1.22 V for the most stable termination, while the second most stable termination, also an oxygen-rich termination, yields an overpotential of 0.84 V, in better agreement with experiments. We should however note here that an unspoken assumption in the previous work was that only oxygen coming directly from water would participate in the process and build all intermediates.
In this work, we show that, including the possibility that lattice oxygen participates in the formation of some intermediates, the picture that arises is much more consistent with itself and with experimental findings. In particular, we identify a low-energy intermediate that is crucial for having a low overpotential also on the thermodynamically stable termination. This intermediate is stabilized by the formation of a weak bond between the surface oxygen atom and a nearby substrate oxygen atom, and the overpotential-determining step involves the dehydroxylation of a surface hydroxyl group to an oxygen atom at the surface. This leads to a unified, complete, and fully consistent picture of the reaction on this surface: four terminations, described in detail below, are very near in surface energy and therefore might coexist at the surface even at thermodynamic equilibrium; on all these terminations, the overpotential is very similar (0.79−0.8 V), with similar intermediates. We expect therefore the macroscopic properties of the reaction (overpotential and photocurrent) to be largely independent of the details of the surface termination. Moreover, the overpotential-determining step identified here is the proton-coupled electron transfer involving a hydroxyl species at the surface. This is in agreement with recent X-ray absorption spectroscopy measurements21 that indicate that the hydroxyl group concentration increases as the applied bias increases and then starts to decrease at the bias for which gas evolution at the photoanode starts to be observed. With this insight, we have also investigated the water oxidation reaction in the presence of a monolayer of gallium oxide at the hematite surface. Recent experiments have shown that overlayers of noncatalytic oxides like Al2O3, Ga2O3, and In2O3,22,23 grown on hematite, reduce the onset potential for the reaction under illumination, but the mechanism is unclear. Expectedly, theoretical studies of bare Al2O3 and Ga2O3 surfaces show that these oxides are noncatalytic; i.e., they display a much larger overpotential.24 For this reason, it has been proposed that the effect of these oxide layers is rather to passivate surface states at the hematite surface that otherwise act as recombination centers for electron−hole pairs. Still, it could be that, as an ultrathin oxide film, they might become catalytically active. Using the atomic structure of the Ga2O3 film we determined previously,25 here we calculate the overpotential and indeed show that, even in the case of an ultrathin oxide film at a hematite surface, Ga2O3 has an overpotential higher than that of hematite, although it is substantially lower than that of pure Ga2O3. Also in this case, the overpotential-determining step involves dehydroxylation of a surface hydroxyl group.
2. CALCULATION METHODOLOGY Spin-polarized density funtional theory (DFT) calculations were performed using the Quantum ESPRESSO package,26 which uses a plane-wave basis set to expand the electronic wave functions. Ultrasoft pseudopotentials27 were used to describe the interaction between the valence electrons and the ionic core, with plane-wave cutoffs of 40 Ry for the wave function and 320 Ry for the charge density. The Perdew−Burke− Ernzerhoff (PBE)28 form of the generalized gradient approximation (GGA) was used for the exchange-correlation functionals; an additional Coulombic repulsion, as in the PBE+U framework,29,30 was used to correct for the self-interaction of the electrons in the Fe 3d states. The effective Coulombic repulsion parameter U = 4.2 eV was chosen, which gives an electronic band gap of ∼2.0 eV for bulk hematite, in agreement with earlier studies.31−34 To accelerate the convergence of the 1794
DOI: 10.1021/acscatal.6b03162 ACS Catal. 2017, 7, 1793−1804
Research Article
ACS Catalysis
Figure 1. Various surface terminations for α-Fe2O3 (0001) and for one overlayer of α-Ga2O3 on the α-Fe2O3 (0001) surface: (a) S1 (−O3FeFe), (b) S2 (−O3FeFeO), (c) S3 (−FeFeO2), (d) S4 (−FeFeO3), (e) S5 (−FeFeO3Fe), (f) T1 (−O3FeGa), (g) T2 (−O3FeGaO), (h) T3 (−FeGaO2), (i) T4 (−FeGaO3), and (j) T5 (−FeGaO3Ga). Color scheme for atomic spheres: gold for Fe, red for O, and blue for Ga.
self-consistent field calculations, a Gaussian smearing of 0.02 Ry was also used. Bulk calculations for α-Fe2O3 and α-Ga2O3 were performed using a 4 × 4 × 2 k-point mesh, for sampling the first Brillouin zone. Isolated molecule calculations for H2O and H2 molecules were performed by placing each molecule in a large cubic box of ∼21 Å, using Γ-point sampling. All the surface and overlayer calculations were performed using a 4 × 4 × 1 k-point mesh. A denser 8 × 8 × 1 k-point mesh was used to perform the postprocessing calculations. All geometries were allowed to relax, until the forces on each atom in each direction were
