Effect of Charges on the Interaction of a Water ... - ACS Publications

Article pubs.acs.org/JPCC

Effect of Charges on the Interaction of a Water Molecule with the Fe2O3(0001) Surface Fabio R. Negreiros,* Luana S. Pedroza, and Gustavo M. Dalpian* Centro de Ciências Naturais e Humanas, Universidade Federal do ABC, Santo André, 09210-580 SP, Brazil S Supporting Information *

ABSTRACT: Electronic charges can play a significant role in interaction between water and oxide surfaces, but not much is known about it. In this work, the interaction of a single water molecule with the Fe2O3(0001) surface was studied by DFT +U calculations. To simulate the charged slabs we have used two different methodologies: directly changing the total number of electrons in the supercell, while using a background charge to keep the whole slab neutral; and including guest atoms that act as donors/acceptors. We find that both approaches give similar qualitative and quantitative results, with the added electron being localized on a single Fe atom, while the hole is delocalized mainly on the surface oxygen layer. In addition, we obtain that a water molecule binds more strongly to the negatively charged surfaces when compared to the neutral case. This reduces the energy barrier and increases the enthalpy gains for water dissociation, while a hole reduces this interaction’s strength and also inhibits water dissociation.

■

studied this structure,18 including defects on hematite, which leads to an even larger photocatalytic activity. Being hematite is an insulator, it can be doped either p- or ntype. Usually the grown samples are n-type, leading to an excess of electrons. Systematic studies of defects in this system have been performed in the literature,5,21 and in particular, Joohee Lee et al. have shown that the Fe vacancies, Fe interstitials, and/or O vacancies can be very stable in this system depending on the O/Fe chemical potentials.5 Also hematite is known to exhibit negatively charged polarons that can lead to n-type doping.22,23 However, during photoexcitation holes are created by the photons, and there is a strong tendency for the holes to be localized at the surface.24 Consequently, it is natural to suppose that these surfaces are never neutral: electrons or holes will actively participate in the chemical reactions taking place on these surfaces. In addition, real temperature effects are also important due to the complex properties of water. These important features altogether, however, have been considered in only a few computational works,25 and for hematite, they have not yet been addressed to the best of our knowledge. In this article we study the interaction of a single water molecule with the (0001) surface of hematite at the DFT+U level. We include positive and negative charges on the system and show that they strongly influence this interaction. Negative charges lead to a stronger binding between surface oxygen and hydrogen, whereas positive charges weaken this bond. We also

INTRODUCTION

A great deal of attention has been devoted to the study of photocatalytic effects on the interface between insulating oxide materials and water.1 It has been widely demonstrated for a variety of materials that water splitting can occur at these surfaces under certain specific conditions. The choice of the best material includes a balance between cost and performance. Among the desired properties, we need a specific band gap and band offset in order for the oxidation and reduction energies of water to be placed inside the band gap. Not many materials show this kind of behavior. In between them there has been considerable interest in hematite, due to its low cost and good photocatalytic activity. A considerable amount of theoretical and experimental works characterizing this structure can be found in the literature.2−11 In order to deeply understand the processes that occur at this surface, it is of fundamental importance to understand how water interacts with it. It is already known that, at the molecular level, the most stable binding site for an adsorbed water molecule is the surface Fe4+ ion,12−15 where a strong bond between the water molecule and Fe is formed while keeping a weaker interaction between water hydrogen and surface oxygen. Enthalpy and kinetics of partial dissociation of a water molecule were also evaluated in some theoretical works for different terminations and environments.15−18 Liquid water was also taken into account,19,20 for which it was obtained that at room temperature some dissociation is already found at the liquidwater/hematite interface after a few picoseconds of molecular dynamics simulation. In general, the (0001) termination is found to be one of the most interesting surfaces, showing good stability and performance for this process. Toroker has also © XXXX American Chemical Society

Received: February 20, 2016 Revised: May 17, 2016

A

DOI: 10.1021/acs.jpcc.6b01743 J. Phys. Chem. C XXXX, XXX, XXX−XXX

Article

The Journal of Physical Chemistry C show that charges can significantly reduce the energy barrier for water dissociation on this surface.

■

COMPUTATIONAL DETAILS All calculations were performed with the CP2K open source software26 at the DFT27 level using the Perdew−Burke− Ernzerhof (PBE)28 exchange-correlation approximation. Since this PBE functional fails to correctly reproduce the more strongly correlated Fe 3d electrons (in the same way as all local density and generalized gradient functionals), we adopted the “+U” Hubbard correction that includes a different Hubbard Hamiltonian specifically for the chosen localized electrons. Using a U value of 4.0 eV, the predicted gap for the bulkhematite improved from 0.9 to 2.1 eV (the experimental value is around 2.2 eV7) and excess electrons were more correctly localized on Fe sites instead of being delocalized on many iron atoms. The pseudopotentials of Goedecker, Teter, and Hutter29−31 with 1/6/16/13/7 electrons for H/O/Fe/Nb/F were used, respectively. All calculations were performed at the Γ point, with cutoffs of 500 Ry/50 Ry for the finest grid/ relative grid for Gaussian mapping and using DZVP basis functions.32 With these cutoffs, the geometric properties of bulk hematite and isolated water molecule are converged within 0.005 Å/0.01°, and the total absolute energy is converged within 0.1 meV per atom. Grimme’s VDW correction D3 with BJ damping was also applied in order to give a better description of the water molecule.33 For the geometry optimizations the position of all atoms were optimized until all forces were smaller than 4.5 × 10−4 hartree/bohr. Water has a significant dipole moment that leads to a relatively long-range interaction. Furthermore, charged slabs also show these long-range interactions. Consequently, in order to study water interacting with charged slabs one needs a unit cell as large as possible, in both perpendicular and parallel directions in order to avoid spurious interaction between replicated images. At the same time, due to the richness of the potential energy surface (PES) of water molecule adsorbed on oxide surfaces, even if it is only a single one, we need to perform several structure optimizations and also molecular dynamics simulations for a thorough description of the PES. Therefore, we considered a 3 × 3 × 1 unit cell with 18 layers giving a total of 270 atoms, a slab that shows a good compromise between size and computational cost. An illustration of this 3 × 3 × 1 unit cell is shown in Figure 1. We considered an antiferromagnetic configuration where consecutive Fe layers have the same spin, which is the most stable magnetic configuration for the bulk hematite.6,14 Variable cell optimizations with different starting geometries/unit cell dimensions were performed in order to determine the equilibrium lattice parameters. The values of a = 5.003 Å/c = 13.67 Å were obtained, which are in good agreement with the experimental values8 of a = 5.03 Å/c = 13.77 Å and also with previous calculations done at the DFT level.15,17,21,34,35 The Fe2O3(0001) surface was modeled using the same 3 × 3 × 1 hexagonal unit cell that has a total of six Fe−O3−Fe layers, terminated in a layer of Fe with the same spin direction on each side (see Figure 1). This termination was shown to be very stable at a variety of temperatures and oxygen’s chemical potential values.14,15 Twenty-two angstroms of empty space separates the slab from its periodic image in the direction perpendicular to the surface. For the nudged-elastic-band (NEB) calculations36 the climbing image procedure was adopted and three intermediate

Figure 1. Three × 3 × 1 model employed in most calculations, containing 270 atoms and a total of 18 layers (6 oxygen and 12 iron layers). The unit cell vectors of this hexagonal cell are also shown. Oxygen atoms are in yellow, and Fe atoms are in cyan. The magnetic moment of each pair of iron layers is illustrated on the right.

images were used, taken into account that the reaction paths considered are very short and easy to evaluate. For the molecular dynamics (MD) simulations, a time step of 1 fs was used and the hydrogen’s atomic mass was increased by 2. The temperature was set to 300 K, and a NVT ensemble (with a canonical sampling through velocity rescaling thermostat) was performed for 5−10 ps proceeded by a NVE simulation for additional 10 ps. These last 10 ps were used to extract the bond length averages presented in this work, and we have verified that this is long enough to give accurate averages.

■

RESULTS AND DISCUSSION Clean Surface. We start characterizing the clean, neutral Fe2O3(0001) slab. This surface was optimized, and the equilibrium interlayer distances are presented in Table 1. Table 1. Interlayer Distance Variations for the Neutral Fe2O3(0001) Slab, in Percentage of the Bulk Value Shown in the Second Column, Compared to DFT Calculations from the Literature Δd12 Δd23 Δd34 Δd45 Δd56 Δd67 Δd78 Δd89

(Fe−O) (O−Fe) (Fe−Fe) (Fe−O) (O−Fe) (Fe−Fe) (Fe−O) (O−Fe)

equil. (Å)

0 (%)

ref 34

ref 13 (%)

0.84 0.84 0.60 0.84 0.84 0.60 0.84 0.84

−69 +8 −35 +16 +5 −5 +1 +1

−67 +8 −37 +17 +7 −4 +2 +2

−53 +7 −29 +13 +3 0 0 x

With respect to the equilibrium bulk values for the interlayer distances in the (0001) direction, the optimized surface shows a large 0.6 Å contraction of the first Fe−O distance, followed by a small 0.08 Å expansion of the second O−Fe layer. From the third layer on, we have consecutive contractions in the Fe−Fe interlayer distances and O−Fe/Fe−O expansions with a decreasing magnitude as we move toward the bulk. After eight layers the equilibrium bulk interlayer distances are B


Article

The Journal of Physical Chemistry C

suggests that they have a Fe2+ oxidation state since all the other bonds were unchanged with the inclusion of the negative charge. In Figure 3a the charge density difference between the

recovered. Compared to previous DFT+U results found in the literature,13,34 we observe a very good quantitative and qualitative agreement, which shows that the fact that we included van der Waals corrections does not affect much the final optimized geometry of the clean surface. The projected density of states (PDOS) is shown in Figure 2a, setting as zero

Figure 3. Charge density difference plots between a neutral slab and a charged slab with two electrons (a) and two holes (b). Red/blue means gain/loss of electrons. Cyan/yellow represents iron/oxygen atoms.

neutral and charged slabs confirms that the extra electrons are localized only in these two specific Fe sites. The strain induced by this polaron has a similar magnitude as the one found in the literature, where we find an increase in the Fe−O bond length of 0.05−0.12 Å.5,22,23 In Figure 2b the projected density of states (PDOS) are shown for this slab. Compared to the neutral case shown in Figure 2a the added electrons locate in the middle of the gap, close to the VBM. Last, the average interlayer distance variations in Table 2 shows an even larger

Figure 2. PDOS for the Fe2O3 surface in the following configurations: neutral without defects (a); with two additional electrons (b); Nbdoped, which gives two electrons to the cluster (c); with two additional holes (d). Spin up/down are above/below the y-axis. The dashed black lines show the position for the valence band maximum (VBM) of each case. The black line is the total density of states, while the red/green/blue/light blue filled curves correspond to Fe/O/Fe2+/ Nb, respectively, where the last two were multiplied by 10 for visualization purposes. The zero energy corresponds to the VBM of the neutral case, and the Fe 2s core levels of the Fe core layer (kept frozen in all optimizations) were used to align all plots with respect to the neutral case. A similar magnitude for the alignment was obtained by alternatively making the electric Hartree potential of the frozen part of the slab equal for both neutral/charged cases.

Table 2. Interlayer Distance Variations for the Optimized Slabs in Four Configurations: Charged (±2), Nb-Doped, and in the Presence of Two F Atoms Far from the Slaba Δd12 Δd23 Δd34 Δd45 Δd56 Δd67 Δd78 Δd89 a

energy the valence band maximum (VBM) of our system. We note that the states with an energy between −5 and −0.2 eV are dominated by a mix of O 2p with Fe 3d states, while the valence states around the VBM are composed mainly of O 2p states. A closer inspection of these states reveal that they belong exclusively to the outermost surface O layer. In the conduction band at ∼+2eV, we find instead mostly unoccupied Fe 3d states. We then make the slab negatively charged by adding two electrons, while applying a positive background charge to keep the whole system neutral. To locate one electron on each side of the slab, keeping the symmetry of the slab in the perpendicular direction, we selected two iron surface atoms from each side and manually stretched their Fe−O bonds by 0.02 Å before optimizing the cluster’s geometry. After relaxation we obtained that these Fe−O bonds stretched to 1.86 Å, in comparison to 1.79 Å for the neutral case. This

(Fe−O) (O−Fe) (Fe−Fe) (Fe−O) (O−Fe) (Fe−Fe) (Fe−O) (O−Fe)

2−

Nb

2+

2F

−99 +2 −35 +20 +9 −14 +5 −1

−91 +1 −36 +21 +6 −15 +6 0

−57 +8 −37 +17 +5 −5 +1 +1

−62 +8 −37 +17 +5 −6 +2 +1

All values are in percentage of the bulk value.

contraction of the first iron layer, with the outermost iron/ oxygen layers being on average at the same height as the outermost oxygen layer. However, the deeper interlayer distances (Δd34, Δd45...) are more similar to the neutral case shown in Table 1, as one would expect since the charge defect is localized only at the surface. The same procedure was repeated now positioning the polaron on Fe sites closer to the bulk, at the fourth layer. In this case we obtain that the charge can still be easily localized in this site, but at an energy cost of 0.59 eV with respect to the surface sites, so the polaron at the bulk is less stable than on the surface. This is expected due to strain release since the surface Fe atoms are less coordinated and are therefore easier to have Fe−O bonds stretched. C


Article


positively charged slab. The charge density difference plot found in the SI confirm the similarity between the two cases. Therefore, for the clean hematite surface, charge defects are located at the surface. Electrons are localized in a single Fe atom, and the slab is found to be more compressed in the normal direction when compared to the neutral slab. Holes are instead delocalized mainly in the outermost oxygen layer and are responsible for an overall expansion of the slab. Interaction with a Single Water Molecule. With the clean surface described, we proceed by studying the adsorption of one water molecule at the surface. The main purpose of this study is to have a complete understanding of how a water molecule behaves at this surface, focusing on the charge effects. This is a prototypical system and not what is observed at the experiments, where we have many water molecules interacting among themselves. However, a more detailed comprehension of a single water molecule adsorbed at the surface can provide us insights of how the real interaction is, both from a theoretical point of view as well as in comparison to experimental results. It has been shown in the literature that H2O has more than one stable minima when adsorbed in a pristine Fe2O3(0001) surface.12−15 We performed an exploration of the PES starting from some chosen configurations and also from previously minima determined in the literature. Some of these configurations are illustrated in the SI together with their adsorption energies. We observe that the most stable configurations present a strong bond between water oxygen and surface iron, but differ in the distance between the hydrogen and the surface oxygen atoms. The lowest energy minimum found is shown in Figure 4a, where we highlight also

Another way of simulating a charged surface is by adding dopants to the slab, which avoids the use of a neutralizing background charge. We tested the effect of replacing an iron atom from the middle Fe layer by all different elements from the fourth−sixth columns of the periodic table, focusing if the dopant gives/removes electrons to/from the system and if it adds unwanted states near the Fermi energy (Ef). It was found that a single Nb dopant gives similar features as the ones presented by the previously discussed negative slab. Since Nb is less electronegative than Fe, it gives away five electrons instead of three, and by doing so it becomes a closed shell ion. The charge density difference plot (see SI) confirms that the doping of a single Nb atom is responsible for the creation of two Fe2+ surface sites on each side of the slab. Furthermore, in the PDOS plot in Figure 2c we see that the Nb atom does not show any states in the Ef region, therefore validating it as a dopant that reproduces the behavior of adding electrons to the slab. For this Nb-doped slab the interlayer distances were reoptimized and are shown in Table 2, where one can see that a very similar qualitative and quantitative behavior was found compared to the negatively charged case. However, when two electrons are removed from the slab, making it positively charged (the total charge is again neutralized with a background charge), the optimized interlayer distances shown in Table 1 give a reduction from −69% to −57% of the first layer contraction, while the other distances get closer to the neutral values as we approach the bulk. Overall, in absolute terms, the positively charged slab expands 0.10 Å on each perpendicular direction. In addition we also note that the atoms from each layer have nearly the same height, so we have that the holes are found at the surface and are distributed equally through all surface atoms, instead of being localized in a single Fe atom like in the negatively charged slab. The charge density difference plot from Figure 3b confirms this picture, with the charge loss delocalized mostly in the outermost oxygen layer, with a smaller contribution coming from the remaining layers. Results from the literature5 suggest that this delocalized solution is indeed the most stable one. The PDOS in Figure 2d shows that the electrons are removed mostly from the O 2p states, with no additional significant changes. We also performed a search for a dopant that could reproduce the positively charged configuration, similarly as it was done before, focusing on replacements for both oxygen (by carbon and nitrogen) and iron, but in all cases, the added holes were delocalized in the whole slab instead of only at the surface. This is likely due to the delocalized nature of the holes that makes them more sensitive to strain defects created by the inclusion of dopants. Therefore, in order to reproduce the positively charged surface, we included two F atoms in the center of the vacuum region of the unit cell, 11 Å far from each side of the slab and ∼11 Å far from each other (see SI for an illustration of this slab). Due to their high electronegativity there is charge transfer from the surface to them, making them negative F ions, while holes are created in the slab without any strain defects. The advantage of this approach over the direct creation of holes used in the last paragraph is the larger control over the position and influence of the fluorine atoms in the calculation since we know their charge and how distant they are from the rest of the cluster. Keeping the fluorine atoms frozen and optimizing the geometry of the slab gave interlayer distances (Table 2) of similar magnitude compared to the

Figure 4. Optimized geometry of the lowest energy minima found for a H2O molecule interacting with a neutral (a), negatively (b), and positively (c) charged Fe2O3(0001) surface. The bonds between surface iron and water oxygen (Fe−Ow) and surface oxygen and water hydrogen (Os−H) are highlighted in (a), and their values are shown in Table 3. Cyan/yellow/white atoms are Fe/O/H, respectively, and dark cyan represents a Fe2+ site.

the two ionic bonds between surface iron and water oxygen (Fe-Ow), and surface oxygen and water hydrogen (Os−H), together with their equilibrium bond distances in Table 3. Comparing the energy values and bond distances with previous DFT results found in the literature, we obtained similar bond lengths: 2.1512 and 2.19 Å15 for DFe−Ow and 1.64 Å15 for DOs−H. However, the adsorption energies are different, with our values being 0.5 eV more negative due to the use of van der Waals corrections. In order to explore in a more consistent way the PES, MD simulations were performed at room temperature, using as a starting point the six configurations illustrated in the SI. The outcome of each one of these MD runs were similar; water oxygen adsorbs on top of the Fe3+ ion and points one of its hydrogens toward a surface oxygen, in a similar way as illustrated in Figure 4a. In addition, the averaged distances between the DFe−Ow and DOs−H bonds of each run were similar D


Article


Table 3. Fe−Ow and Os−H Bond Distances (See Figure 4a) for a Water Molecule in the Fe2O3(0001) Slab in the Several Configurationsa configuration 0

H2O−(Fe2O3) H2O−(Fe2O3)2− H2O−(Fe2O3)Nb H2O−(Fe2O3)2+ H2O−(Fe2O3)2F

DFe−Ow(GO) 2.14 2.11 2.11 2.11 2.12

Å Å Å Å Å

DFe−Ow(MD) 2.15 1.85 1.85 2.10 2.10

± ± ± ± ±

0.05 0.03 0.05 0.09 0.09

Å Å Å Å Å

DOs−H(GO) 1.65 1.58 1.56 1.81 1.70

Å Å Å Å Å

DOs−H(MD) 1.8 0.98 0.99 2.5 2.3

± ± ± ± ±

0.2 Å 0.02 Å 0.05 Å 0.5 Å 0.4 Å

a Neutral (H2O−(Fe2O3)0); with two additional electrons (H2O−(Fe2O3)2−); doped with Nb (H2O−(Fe2O3)Nb), that gives two electrons to the system; with two holes (H2O−(Fe2O3)2+); in the presence of two fluorine atoms (H2O−(Fe2O3)2F) located far from the slab, which extracts electrons from it. The values reported were determined from a geometry optimization (GO) and also from a 10 ps average of a molecular dynamics simulation (MD). The mean square displacement for the MD value is also shown.

which means that the H2O interaction with Fe3+ is stronger than with Fe2+, as one would expect since the interaction is of ionic nature. Furthermore, since H2O binds more strongly close to the defect than far from it, the H2O adsorption energy should be larger in the negative slab when compared to the neutral one. The optimized geometry is illustrated in Figure 4b, and the main bond values are reported in Table 3. We observe that the added electron decreases the length of both H−Os and Ow−Fe bonds, due to the expansion of the neighbor Fe2+ ion and also due to the stronger H−Os ionic interaction coming from the added electron. When we instead add a hole, we observe that the positive background created by the defect reduces the interaction between the surface and the water molecule, increasing the Os−H bond length and making the water molecule dipole to point outward, more parallel to the surface normal, as illustrated in Figure 4c. As noted before, water molecules can present a rich PES that complicates a purely static analysis, specially when defects are included. Therefore, we have also performed MD simulations starting from the previously determined optimized structures, for both charged and doped slabs. The average bond lengths were evaluated and are shown in Table 2. For the positively charged slab we obtain that the dynamics consisted mainly of a rotation of the water molecule around the surface normal, being still strongly bound to the surface iron atom. In fact, the average angle between the surface normal and the water dipole from the dynamics is 90 ± 30°, which implies a weaker interaction between surface oxygens and hydrogen atoms. The same simulation in the H2O−(Fe2O3)2F slab shows similar qualitative features and small quantitative differences (see Table 2). For the negatively charged and Nb-doped slabs we find that the water molecule partially dissociates in about 1−10 ps, giving away one hydrogen atom to a surface oxygen. The averages shown in Table 2 were performed after this dissociation occurred and reveal the formation of a stable OH bond between one hydrogen and a surface oxygen and the strengthening of the Fe3+−Ow bond, suggesting that at least part of the charge of the hydrogen donated to the surface remained in the OH radical, making it more negative (OH−) and increasing the ionic bond with the iron ion. Thus, molecular adsorptions in these negatively charged surfaces are unstable local minima, with very small energy barrier for dissociation. This barrier will be determined in the next section. In Figure 6 the H2O adsorption energy as a function of the water oxygen height Δz is shown, setting as zero the height of the bound water molecule and choosing some values for Δz in the 0 < Δz < 8 Å range. This can be rationalized by considering how the extra electron/hole distributes at the surface. The hole

inside a 0.05 Å range, which means that even though there exist different minima at 0 K, at 300 K the system is not trapped, being able to overcome the small energy barriers that separate them and explore the flat PES. In Table 3 we show the values of DFe−Ow and DOs−H for both the geometry optimization and the MD simulation, where we see that although DFe−Ow deviates very little from its equilibrium value, DOs−H deviates a lot more, suggesting a weaker interaction between the hydrogen atoms with the surface. To analyze if there is charge transfer between the water molecule and the surface, we show in Figure 5a the charge

Figure 5. Charge density difference between the adsorbed water molecule and its isolated noninteracting counterparts, the isolated molecule and the clean hematite surface. Two configurations were considered, molecular adsorption (a) and partially dissociated molecule (c). Red/blue means gain/loss of electrons due to the interaction. This charge density difference was integrated in the xy plane (for every 0.1 Å in the z direction), and the result is shown in (b) as a function of the height for both cases. The dashed line illustrates the zero-height reference.

density difference between the minima from Figure 4a and its counterparts, i.e., the isolated water molecule and the clean hematite surface. It can be seen that a large charge rearrangement occurs, and an accumulation of charge in the water oxygen ion is found, which increases the strength of the ionic Fe3+−O bond, while weakening the O−H bond. Overall there is a negligible charge transfer between the molecule and the surface as shown in the plot of the charge density difference vs height presented in Figure 5b. We proceeded by adding an electron to the minima configuration shown in Figure 4a. In general, one would expect qualitatively that the more negative/positive surface favors more the O−H/Fe−O bonds, respectively.37 Since the electron will be localized in a single surface iron atom, we localized this electron in three different sites: the one bound to the H2O molecule; the neighbor one closest to the hydrogen atom, and a site more than 5 Å far from the hydrogen. The most stable configuration was found to be the second one, E


Article


between them is and how converged our values are. For Δz = 9 Å, the change in the total energy for F−H2O = (4, 5, 6) Å is ΔE = (0.04, 0.02, 0.01) eV, respectively. Therefore, the obtained values are converged by 0.01 eV with respect to the water− fluorine interaction. Dissociation of the Water Molecule. The energetics for H2O dissociation in the neutral Fe2O3(0001) surface was already studied in the literature.15 The DFT+U methodology was used in most calculations, and it is overall agreed that this process is heterolytic (H+ + OH−), with the formation of two types of hydroxyl radicals attached to the surface. The reported enthalpies for this process, however, differ considerably, from −0.2512,13 to −0.12 eV15 and even +0.11 eV.16 Furthermore, energy barriers disagree even more, with values ranging from 0.0515 to 0.69 eV.16 One source of this apparent discrepancy is the number of minima that exist in the flat PES. Indeed, by a closer inspection on the structure of each minima found in the literature we can observe that the initial/final states considered are different among them. Even though all these different states are indeed stable minima, they make the comparison of different calculations difficult. As mentioned before, this issue is partially solved by performing molecular dynamics that even at low temperatures is able to explore a large portion of the water molecule configuration space. In fact, of the five simulations performed at room temperature, we obtained that in some of them the water molecule dissociated after 20 ps, which means that the barriers and enthalpy changes are indeed small. In addition, the differences could also be connected to the size of the unit cell since the previous calculations from the literature used a 2 × 2 or even smaller cell. Since our 3 × 3 slab is significantly wider, our results should be less affected by the long-range electrostatic interaction between replicated cells. We determined the barrier for dissociation of a water molecule in the neutral and charged slabs. The previously determined lowest energy minima shown in Figure 4 were considered as starting points, and they were connected to the local minima involving a dissociated water molecule closest in geometry to the starting point. We note that the thermodynamics and kinetics of dissociation can be more complicated than this since the dissociation might occur from other paths not involving the global minimum, or that it might lead to a more stable final configuration with a larger enthalpy gain as the one reported here. Regardless, since we are comparing similar paths, the qualitative comparison among them is relevant. In Figure 7, the energetics of this process is shown for the neutral and charged configurations. For the neutral case, we obtained the values of 0.04 eV/0.09 eV for the energy change and barrier, respectively, values that are different from the ones reported in the literature. It is worth to mention that the inclusion of VDW corrections in our calculations cause a significantly higher water adsorption energy (0.5 eV stronger), but only small differences in bond lengths (equilibrium O−H bond length is ±0.05 Å shorter, for example). The adsorbed water molecule configuration is favored with respect to the nondissociated configurations by less than 0.1 eV. Therefore, the relative energies do not change that much due to VDW corrections. For the charged cases we obtain opposite trends. Adding electrons decreases the Os−H bond length, which results in a smaller energy barrier of 0.04 eV and a significant enthalpy gain of 0.32 eV. Indeed, such a low energy barrier of 0.04 eV justifies water dissociation occurring after a few picoseconds on all of the MD runs performed at 300 K. Adding holes, instead,

Figure 6. Adsorption energies as a function of the height of the oxygen water molecule (Δz). The zero in Δz is set as the height of the water oxygen molecule when it is adsorbed.

is mainly delocalized in the oxygen surface layer, and it is responsible for the weakening of the Hw−O bond without affecting the stronger Fe3+−Ow interaction. This weakens the water molecule adsorption magnitude. However, the electron is completely localized in a neighbor Fe surface atom, changing its oxidation state and enlarging its Fe−O bonds. This weakened bond results in a large increase in the Hw−O interaction, reducing it from 1.65 to 1.40 Å and a gain in the adsorption energy of almost 0.4 eV. To check the convergence on the adsorption energies with Δz, we evaluated the energy difference between the configurations with a dipole pointing up and down (ΔE↑↓), and parallel to the surface normal, in the 5 < Δz < 9 Å range, i.e., the H2O molecule far from the surface. The main idea is that as the distance to the surface gets bigger, the long-range interaction should get smaller, and therefore, the direction of the water dipole should change less than the total energy of the system, with ΔE↑↓ converging to zero. For the neutral slab we obtained ΔE↑↓ = 0.5 meV for Δz = 8 Å. For both negative/ positive slabs the ΔE↑↓ = 40 meV at Δz = 9 Å, so there is still some interaction between the charged slabs and the water molecule at the maximum distance allowed by our unit cell. Forty millielectron volts is, however, small enough to guarantee that our main conclusions are unchanged. Since we are using periodic boundary conditions, the total energy of the charged slabs are not constant with respect to the unit cell’s dimension,38 due to the contribution of the background charge to the total energy. Therefore, in order to check the validity of the reported energy values, we repeated the convergence test previously described for the H2O− (Fe2O3)Nb and H2O−(Fe2O3)2F configurations, where no background charge corrections are applied and the total energy is nearly constant with the unit cell’s volume. The results are shown in Figure 6 and Table 3. We obtained the same qualitative behavior in both cases, with positive/negative slabs interacting more weakly/strongly, respectively. Furthermore, quantitatively the values are very similar, being different by no more than 0.04 eV. For the H2O−(Fe2O3)2F case, it is worth to mention that by changing the distance between H2O and the two F ions (F−H2O), while keeping the molecule far from the surface, one can check how much the electrostatic interaction F


Article


with a compensating background charge, and the second by introducing dopants that give or remove electrons from the slab. We found that both procedures can give the same qualitative and quantitative results as long as one guarantees that the added charge is correctly localized inside the slab.

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpcc.6b01743. Additional charge density difference plots for selected configurations. Adsorption energies and illustrations for some configurations of H 2 O adsorbed on the Fe2O3(0001) surface (PDF)

■

AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Phone: +55 11 49960012. Fax: +55 11 49960090. Figure 7. Energetics for the H2O dissociation path for the neutral and ±2 electrons slabs obtained with a Nudge-Elastic-Band approach. We illustrate the initial, transition, and final states for the neutral case.

Notes

decreases the water oxygen interaction with hydrogen, increasing the O−H bond length and the energy barriers/ enthalpy loss by 0.24 eV/0.21 eV, respectively. Even though the energy barrier is not significantly high, the fact that the final state is 0.2 eV higher in energy implies water dissociation being a slow and quickly reversible process at room temperature in these positively charged surfaces. Finally, for the dissociated water molecule we also analyzed the charge properties of the system by evaluating the charge density difference with respect to its isolated counterparts, as done in Figure 5a for the adsorbed molecule case. In Figure 5c one can see that the dissociation has a more heterolytic picture, where we observed a considerable larger accumulation/ depletion of charge in the water-O/surface-H atoms, making them closer to OH−/H+ states. We also note a large charge redistribution at the surface, with a charge transfer from the iron atom bonded to OH to the surface oxygen atom in direct contact with hydrogen.

The authors thank FAPESP and CNPq for financial support.

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS

REFERENCES

(1) Rossmeisl, J.; Qu, Z.-W.; Zhu, H.; Kroes, G.-J.; Nørskov, J. Electrolysis of Water on Oxide Surfaces. J. Electroanal. Chem. 2007, 607, 83−89. (2) Pozun, Z. D.; Henkelman, G. Hybrid Density Functional Theory Band Structure Engineering in Hematite. J. Chem. Phys. 2011, 134, 224706. (3) Peng, H.; Lany, S. Semiconducting Transition-metal Oxides based on d5 Cations: Theory for MnO and Fe2O3. Phys. Rev. B: Condens. Matter Mater. Phys. 2012, 85, 201202. (4) Kiejna, A.; Pabisiak, T. Mixed Termination of Hematite (αFe2O3)(0001) Surface. J. Phys. Chem. C 2013, 117, 24339−24344. (5) Lee, J.; Han, S. Thermodynamics of Native Point Defects in alpha-Fe2O3: an Ab Initio Study. Phys. Chem. Chem. Phys. 2013, 15, 18906−18914. (6) Rollmann, G.; Rohrbach, A.; Entel, P.; Hafner, J. First-principles Calculation of the Structure and Magnetic Phases of Hematite. Phys. Rev. B: Condens. Matter Mater. Phys. 2004, 69, 165107. (7) Gilbert, B.; Frandsen, C.; Maxey, E. R.; Sherman, D. M. Band-gap Measurements of Bulk and Nanoscale Hematite by Soft X-ray Spectroscopy. Phys. Rev. B: Condens. Matter Mater. Phys. 2009, 79, 035108. (8) Blake, R. L.; Hessevick, R. E.; Zoltai, T.; Finger, L. W. Refinement of the Hematite Structure. Am. Mineral. 1966, 51, 123− 129. (9) Carvalho-Jr, W. M.; Souza, F. L. Thermal Enhancement of Water Affinity on the Surface of Undoped Hematite Photoelectrodes. Sol. Energy Mater. Sol. Cells 2016, 144, 395−404. (10) Carvalho, W. M.; Souza, F. L. Recent Advances on Solar Water Splitting Using Hematite Nanorod Film Produced by Purpose-built Material Methods. J. Mater. Res. 2014, 29, 16−28. (11) Freitas, A. L.; Carvalho, W. M.; Souza, F. L. Enhanced Water Oxidation Efficiency of Hematite Thin Films by Oxygen-deficient Atmosphere. J. Mater. Res. 2015, 30, 3595−3604. (12) Yin, S.; Ellis, D. H2O Adsorption and Dissociation on Defective Hematite (0001) Surfaces: A DFT study. Surf. Sci. 2008, 602, 2047− 2054.

■

CONCLUSIONS We presented a detailed study of the clean and interacting with a single water molecule neutral and charged Fe2O3(0001) surfaces, by DFT+U calculations. We obtained that negatively charged slabs increase the interaction between a water molecule, and the surface by strengthening the water hydrogen and surface oxygen bond. In addition, they significantly reduce the energy required for water molecule dissociation, reducing to a few picoseconds the average time for this process at room temperature. However, positively charged slabs strengthen the interaction between water O and surface Fe, while nearly breaking the weak bond between hydrogen and surface oxygen. Because of this effect, the water dissociation barrier and enthalpy increase, greatly disfavoring the process. We used two different methodologies to simulate charges in the slab, the first by explicitly changing in the calculation the total number of electrons in the system while keeping it neutral G


Article

The Journal of Physical Chemistry C (13) Yin, S.; Ma, X.; Ellis, D. Initial Stages of H2O Adsorption and Hydroxylation of Fe-terminated α-Fe2O3(0001) Surface. Surf. Sci. 2007, 601, 2426−2437. (14) Souvi, S. M.; Badawi, M.; Paul, J.-F.; Cristol, S.; Cantrel, L. A DFT Study of the Hematite Surface State in the Presence of H2, H2O and O2. Surf. Sci. 2013, 610, 7−15. (15) Nguyen, M.-T.; Seriani, N.; Gebauer, R. Water Adsorption and Dissociation on α-Fe2O3(0001): PBE+U Calculations. J. Chem. Phys. 2013, 138, 194709. (16) Pan, H.; Meng, X.; Qin, G. Hydrogen Generation by Water Splitting on Hematite (0001) Surfaces: First-principles Calculations. Phys. Chem. Chem. Phys. 2014, 16, 25442−25448. (17) Hellman, A.; Pala, R. G. S. First-Principles Study of Photoinduced Water-Splitting on Fe2O3. J. Phys. Chem. C 2011, 115, 12901−12907. (18) Toroker, M. C. Theoretical Insights into the Mechanism of Water Oxidation on Nonstoichiometric and Titanium-Doped Fe2O3(0001). J. Phys. Chem. C 2014, 118, 23162−23167. (19) Kerisit, S. Water Structure at Hematite-water Interfaces. Geochim. Cosmochim. Acta 2011, 75, 2043−2061. (20) English, N. J.; Rahman, M.; Wadnerkar, N.; MacElroy, J. M. D. Photo-active and Dynamical Properties of Hematite (Fe2O3)-Water Interfaces: an Experimental and Theoretical Study. Phys. Chem. Chem. Phys. 2014, 16, 14445−14454. (21) Cava, C. E.; Roman, L. S.; Persson, C. Effects of Native Defects on the Structural and Magnetic Properties of Hematite α-Fe2O3. Phys. Rev. B: Condens. Matter Mater. Phys. 2013, 88, 045136. (22) Adelstein, N.; Neaton, J. B.; Asta, M.; De Jonghe, L. C. Density Functional Theory based Calculation of Small-polaron Mobility in Hematite. Phys. Rev. B: Condens. Matter Mater. Phys. 2014, 89, 245115. (23) Lany, S. Predicting Polaronic Defect States by means of Generalized Koopmans Density Functional Calculations. Phys. Status Solidi B 2011, 248, 1052−1060. (24) Jupille, J., Thornton, G., Eds. Defects at Oxide Surfaces, 1st ed.; Springer International Publishing: Cham, 2015. (25) Schnur, S.; Groß, A. Properties of Metal-water Interfaces Studied from First Principles. New J. Phys. 2009, 11, 125003. (26) Hutter, J.; Iannuzzi, M.; Schiffmann, F.; VandeVondele, J. Cp2k: Atomistic Simulations of Condensed Matter Systems. WIREs Comput. Mol. Sci. 2014, 4, 15−25. (27) Hohenberg, P.; Kohn, W. Inhomogeneous Electron Gas. Phys. Rev. 1964, 136, B864−B871. (28) Perdew, J.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868. (29) Goedecker, S.; Teter, M.; Hutter, J. Separable Dual-space Gaussian Pseudopotentials. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54, 1703−1710. (30) Hartwigsen, C.; Goedecker, S.; Hutter, J. Relativistic Separable Dual-space Gaussian Pseudopotentials from H to Rn. Phys. Rev. B: Condens. Matter Mater. Phys. 1998, 58, 3641−3662. (31) Krack, M. Pseudopotentials for H to Kr Optimized for Gradientcorrected Exchange-correlation Functionals. Theor. Chem. Acc. 2005, 114, 145−152. (32) VandeVondele, J.; Hutter, J. Gaussian Basis Sets for Accurate Calculations on Molecular Systems in Gas and Condensed Phases. J. Chem. Phys. 2007, 127, 114105. (33) Grimme, S.; Ehrlich, S.; Goerigk, L. Effect of the Damping Function in Dispersion Corrected Density Functional Theory. J. Comput. Chem. 2011, 32, 1456−1465. (34) Pabisiak, T.; Kiejna, A. Fe Adsorption on Hematite (α-Fe2O3) (0001) and Magnetite (Fe3O4) (111) Surfaces. J. Chem. Phys. 2014, 141, 134707. (35) Hellman, A.; Iandolo, B.; Wickman, B.; Grö nbeck, H.; Baltrusaitis, J. Electro-oxidation of Water on Hematite: Effects of Surface Termination and Oxygen Vacancies Investigated by Firstprinciples. Surf. Sci. 2015, 640, 45−49. (36) Henkelman, G.; Uberuaga, B. P.; Jonsson, H. A Climbing Image Nudged Elastic Band Method for Finding Saddle Points and Minimum Energy Paths. J. Chem. Phys. 2000, 113, 9901−9904.

(37) Sánchez, C. G. Molecular Reorientation of Water Adsorbed on Charged Ag(111) Surfaces. Surf. Sci. 2003, 527, 1−11. (38) Leslie, M.; Gillan, N. J. The Energy and Elastic Dipole Tensor of Defects in Ionic Crystals Calculated by the Supercell Method. J. Phys. C: Solid State Phys. 1985, 18, 973.

H


Effect of Charges on the Interaction of a Water ... - ACS Publications

Recommend Documents