Density functional theory study of trends in water dissociation on oxygen-preadsorbed and pure transition metal surfaces

Oxygen and water are the most reactive gases of the ambient air. The adsorption of both molecules on transition metal surfaces have been studied extensively, but mostly separately. However, water and oxygen usually co-exist, and therefore realistic systems need to take into consideration both simultaneously. As these adsorption reactions are so common, state-of-the-art results are beneficial as they capture large trends as accurately as possible. A comprehensive study of oxygen and water co-adsorption and dissociation on Ag(111)-, Au(111)-, Pd(111)-, Pt(111)-, Rh(111)- and Ni(111)-surfaces have been performed using density functional theory. We present a very strong general trend, where dissociated oxygen systematically lowers the activation energy of water dissociation on transition metal surfaces. This makes the oxygen dissociation the rate-determining step of the water dissociation reaction. The effect is caused by the additional pathway that the dissociated oxygen enables for the dissociation of water molecule.


Introduction
Water and oxygen are the most reactive gases present in ambient air. The water vapor of ambient air adsorbs to all surfaces. However, this adsorbing water does not dissociate easily, and in chemical processes that require water dissociation, metal catalysts can be used to lower the energy barrier of the reaction. Because this water splitting is not always conducted under vacuum, it is important to consider how the presence of oxygen affects this process, as many metals have the ability to cleave oxygen.
The oxygen dissociation reaction on different metal surfaces has been studied extensively using density functional theory (DFT). For example, Yan et al. [1] studied water-promoted oxygen dissociation on transition metal surfaces including cobalt, rhodium, iridium, nickel, palladium, platinum, copper, silver, and gold. Previous research also includes oxygen adsorption on silver surface [2], and the kinetics of oxygen dissociation on rhodium, silver, and gold surfaces [3].
surfaces [11]. Experimental work has also been done on these water dissociation reactions, one example being water dissociation on palladium [12]. However, more computational studies are still required to capture trends in water dissociation in the presence of oxygen, due to the additional pathways that the coadsorbing oxygen can provide for the dissociating water. This type of computational research is quite rare [13], and more often contains only one or two metals in a single study, examples including reactions on gold and silver surfaces [14], rhodium and nickel surfaces [15], and on a platinum surface [16]. The effect of oxygen on water dissociation has been experimentally studied for many metals, such as silver [17], nickel [18], palladium [19], platinum [20] and rhodium [21]. When both oxygen and water are taken into consideration, this research can also support the study of the effects of humid air in other catalytic processes, that take place on these metal catalysts.
Computational methods in chemistry continue to advance in a rapid manner. More reactions on larger amounts of different metal surfaces can be covered, and thereby more general trends can be discovered. In addition, these advancements allow the use of relatively large unit cells in computations. An increased size of the unit cell lowers the risk of inaccurate results caused by the repulsion of atoms that are too close https://doi.org/10.1016/j.susc.2023.122305 Received 1 December 2022; Received in revised form 3 April 2023; Accepted 10 April 2023 to each other, and prevent the discovery of the true global minimum of the reaction.
In this work, we present trends between oxygen dissociation and water dissociation on oxygen-preadsorbed and clean metal surfaces. These metal surfaces include Au(111)-, Ag(111)-, Pd(111)-, Pt(111)-, Rh(111)-and Ni(111)-surfaces. In addition, we present a computational model for the interaction between the metal surface and both oxygen and water molecules. Our results have been compared to both experimental and computational studies to solidify our results and the general trends presented in this study.

O 2 and H 2 O dissociation on metal surfaces
The chemical reactions under study were oxygen dissociation on clean metal surfaces, and water dissociation on both clean and oxygenpreadsorbed metal surfaces. After dissociation, atoms and ions adsorb to either a three-fold coordinated hollow site, a two-fold bridge site, or onto a one-fold top site of the metal surface.
The oxygen dissociation reaction can be presented with Eq. (1), where (ads) represents an adsorbed species.
Water dissociation on clean metals can be presented with an Eq. (2). In this reaction, a water molecule that has adsorbed molecularly on the metal surface dissociates into a hydrogen atom and a hydroxyl group.
The water dissociation reaction has been shown to, both thermodynamically and kinetically, benefit from pre-adsorbed oxygen on various metal surfaces [4,6]. When the water dissociation reaction uses the oxygen of the oxygen-preadsorbed metal surface to create adsorbed hydroxyl, it can be presented with Eq. (3). The reaction energy and the activation energy were calculated for reactions (1-3) computationally using details presented in Section 2.
We can calculate the energy of the molecular adsorption . . of the oxygen or water molecule on the metal surface with Eq. (4), where is the total energy of the whole molecular adsorption system, is the energy of a pure metal slab, and is the total energy of an independent molecule, which was either water or oxygen. In the case of water dissociation on oxygen-preadsorbed metal surfaces, consists of the energy of the metal slab with an added oxygen atom adsorbed at the hollow site.
The reaction energy for the dissociation reactions (1), (2) and (3) can be calculated with Eq. (5), where is the energy of the dissociatively adsorbed state, which corresponds to the final state of the reaction.
The activation energy can now be described using the energy of the highest energy state along the minimum energy path from the initial state to the final state of the dissociation reaction, according to Eq. (6).
Activation energy and reaction energy can be used in conjunction to create a linear, empirical relationship between reactions that are similar to each other. For example, oxygen dissociation on different metals can follow this Bell-Evans-Polanyi (BEP) relationship. The BEP-relationship can be used to estimate the kinetics of the reaction based solely on the thermodynamic values of the reaction [22]. The BEP-relationship is presented in Eq. (7), where 0 corresponds to the intercept, and to the slope of this fitted line.

Computational methods
The modeling of the reaction pathways was conducted using the GPAW program [23]. Density functional theory was utilized with Perdew-Burke-Ernzerhof (PBE) exchange and correlation functional [24]. Real-space grid spacing was 0.2 Å, and the sampling of the k-space was carried out with a 3 × 3 × 1 Monkhorst-Pack grid. The criterion for convergence was the force smaller than 0.06 eV/Å on all individual atoms. All of the reaction barriers were calculated using nudged elastic band -method with climbing image (CI-NEB) [25], with a total of 10 images.
Initial structures were constructed using Atomic Simulation Environment (ASE) [26]. All of the metal (111)-surfaces were modeled using a slab with 3 × 3-surface and five layers of atoms, and with 8 Å vacuum on both top and bottom side of the slab. We chose the 3 × 3-surface as it has reasonably good low coverage limit, since 3 × 2and 2 × 2-surfaces can suffer from quite strong adsorbent-adsorbent interactions. The oxygen and water molecules from G2-database as implemented in ASE were used, and their structures were optimized in a sufficiently large lattice to prevent repulsion between molecules. The oxygen-preadsorbed surfaces had one adsorbed oxygen atom in their three-fold hollow site, which made the oxygen coverage on these surfaces 1/9 ML. In the water dissociation calculations, the two bottom atomic layers were fixed. Magnetization of ferromagnetic nickel was taken into account via relaxation of magnetic moments using spin polarization and thus spin polarization was employed on calculations using a Ni(111)-surface. Dispersion correction was secluded from the calculations, due to reported absence of changes in both relative structure stabilities and adsorption sites, and only minute differences in adsorption geometries of water adsorption [27]. These minute differences were verified with oxygen and water dissociation reactions on Pd(111)-and Pt(111)-surface. The energy differences were highest for the oxygen dissociation on Pt(111)-surface with 0.06 eV energy difference in activation energy and 0.05 eV difference in reaction energy. For other reactions, the energy differences varied between −0.02 and 0.02 eV.
Vibrational frequency calculations, with real-space grid spacing of 0.15 Å, and the sampling of the k-space with a 4 × 4 × 1 Monkhorst-Pack grid, were employed to calculate vibrational modes and zero point energy corrections. The spin polarization was not taken into account for the final state of water dissociation reaction on Ni(111)-surface due to poor convergence. The transition states were verified with a single imaginary mode along the reaction coordinate.

Oxygen dissociation
Minimum reaction paths for the oxygen dissociation reactions on the studied metals were discovered as follows. Different configurations for the initial state of the reaction was tested. In all metal surfaces in this work, oxygen molecule, in the initial state of the dissociation reaction, was found to adsorb molecularly at the top-bridge site of the metal surface. The most stable adsorption site for an oxygen atom was found to be a hollow site, and therefore dissociation reaction of the oxygen molecule leads to oxygen atoms being adsorbed on neighboring hollow sites. Once the initial and final states of the reaction were found, an initial approximation for the reaction path was created by guessing the transition state, and then presuming a linear path from the initial state to the transition state, and from the transition state into the final state. This approximation for the reaction path was then corrected using the CI-NEB algorithm. During CI-NEB calculations, no configurations that had lower energies than initial or final states were found, thus indicating that the initial and final states found before the CI-NEB calculations correspond to the true initial and final states for the reaction. The reaction paths for the oxygen dissociation on   . different metals varied only slightly from each other. As an example of the reaction path of the oxygen dissociation reaction, the oxygen dissociation on Pd(111)-surface is presented in Fig. 1. The reaction parameters for all of the studied oxygen dissociation reactions, along with corresponding references, are presented in Table 1.

Water dissociation on pure metals
Water dissociation reaction was studied on nickel, rhodium, platinum, gold, silver and palladium surfaces. As an example of such a reaction, the reaction path of the water dissociation on palladium surface is presented in Fig. 2. On all studied metals, the initial physisorption site of water was found to be the top site. Used in the CI-NEB algorithm, the initial guess of the minimum energy path on all studied metals transitioned hydrogen atom of the water molecule into a nearby hollow site, leaving a chemisorbed hydroxyl ion on the top site of the metal surface. However, this initial guess was found not to be accurate for all of the studied metals. During the dissociation reaction on all metals but platinum, the hydrogen migrated to a hollow site. But on a platinum surface, the water dissociation lead to hydrogen being adsorbed at the top site. The adsorption site of the hydroxyl group on studied metals was divided between the top site (Rh, Pt, Au and Pd), and the bridge site (Ni and Ag). On both of the adsorption sites, water has a planar configuration. Table 2 Adsorption sites of molecules and ions present in the initial and final states of the water dissociation reaction, along with reference sites presented by other research groups. When comparing adsorption sites, one should also consider migration presented in Table 5. The adsorption sites of dissociating water molecule at the initial and final states of the dissociation reaction are presented in the Table 2 with reference sites. Once the minimum energy path between these Hydrogen is colored white, oxygen is colored red and palladium green. Initial state is on the left, transition state in the middle, and final state is on the right.

Table 3
Reaction parameters for the water dissociation on studied metals, along with reference values presented by other research groups.
Au (111) Ag (111) Pd (111) Pt (111) Rh (111) Ni (111) . adsorption sites was found using the CI-NEB algorithm, the molecular adsorption energy along with the reaction and activation energies of water dissociation on studied metals were calculated, and are presented in the Table 3.

Water dissociation on oxygen-preadsorbed metals
Water dissociation was also studied on oxygen-preadsorbed metal surfaces, in order to discover the effect the oxygen has on the pathway of the dissociation reaction. In all of the studied surfaces, an oxygen atom was found to adsorb on a hollow site. Before its dissociation, water molecule adsorbed on the top site of all studied metals in a planar configuration, as was the case in the clean metal surfaces. The initial guess for the minimum energy path presumed the water molecule on top site to donate its hydrogen to oxygen atom, thus producing two hydroxyl groups on the metal surface. During the CI-NEB-calculations, this presumption was found to be true. The presence of oxygen reduced the activation energy of all reactions.
As an example of the reaction pathway, the water dissociation on oxygen-preadsorbed Pd(111)-surface is presented in the Fig. 3. The oxygen atom receiving the hydrogen was found to migrate to a bridge site. The hydroxyl group, that followed the donation of hydrogen from the water molecule, remained on the top site of the metal surface on all metal surfaces except on nickel surface, where the other hydroxyl group migrated onto a hollow site. As this atomic configuration leaves unsymmetrical adsorption sites for the hydroxyl groups, the migration of this hydroxyl group is discussed in more detail in the next chapter.
Wang and Bu [13] report similar initial and final states as presented in our work, in addition to the migration of the hydroxyl group after the water dissociation reaction. However, they report hydroxyl groups preference to migrate to a hollow site, whereas our results indicate this only on a Ni(111)-surface. Additionally, Pozzo et al. [15] reported similar adsorption sites for both initial and final states of this reaction on Rh(111)-surface. Reaction parameters for the water dissociation on oxygen-preadsorbed metals is presented in the Table 4.

Diffusion of the hydroxyl group
The water dissociation reaction on a oxygen-preadsorbed surface, presented in Fig. 3, leaves hydroxyl group into a top site, thus leading into a structure, where the two hydroxyl groups are adsorbed into different sites. CI-NEB calculations on the palladium and silver surfaces revealed, that the diffusion of a lone hydroxyl group from a top site into a bridge site has a negligible barrier. This enables a spontaneous diffusion of the hydroxyl group, which can hold true also for the migration of hydroxyl groups on other metal surfaces. The diffusion of a lone hydroxyl group was therefore investigated for all the studied metal surfaces, so that the energy gain from symmetrical adsorption sites for the hydroxyl groups can be discovered. The energy differences between top-and bridge-sites for the adsorption of the hydroxyl group are presented in the Table 5. For a Ni(111)-surface, the diffusion from bridge site to a hollow site is presented in the Table 5, due to the varying diffusion sites in contrast to other metal surfaces.

BEP-relationship
By gathering reactions' activation and reaction energies from Tables 1, 3 and 4, BEP-relationship can be constructed for studied oxygen and water dissociation reactions. These BEP-plots are presented in Fig. 4. BEP-relationship is a useful tool, as it provides a shortcut to acquiring activation energies within this group of materials, which can be time-consuming and computationally expensive. In addition, a good fit of the linear plot is usually an indicator of similar reaction paths and transition states of the reactions presented within the BEP-line.

Comparison to work by other groups
Computational results of this work are compiled to Fig. 5. When results for oxygen dissociation are compared to other groups' work presented in the Table 1, it can be seen that the activation and reaction energies differ, especially on Au(111)-surface. Initial adsorption site for the molecular oxygen adsorption was seen to be quite peculiar, as position of other oxygen atom is close to a top site and of other close to a bridge site. However, this type of structure minimizes the distance traveled along the surface, and allows oxygen atoms to move into different hollow sites without needing to approach the top site of the metal atoms, while preventing the steric hindrance between oxygen atoms to rise during the migration process. This molecular adsorption site was also reported by Yin et al. [1], whereas Fajin et al. [28] reported a bridge site configuration for molecular oxygen adsorption on Au(111)-and Ag(111)-surfaces, where both oxygen atoms are located between a top and a bridge site. In the final state of the reaction, oxygen atoms were adsorbed to a hollow site in all three studies.
As can be seen from the Table 3, our results regarding water dissociation without oxygen were in a good agreement with other groups' work. However, differences are quite large when comparing water dissociation with oxygen on all surfaces but Pd(111) and Ni(111), when comparing results on Table 4.

The computational model for water dissociation
Computational results presented in this work provide a very clear trend. Water dissociation on pure metal surfaces has a relatively high barrier on all studied surfaces. This high barrier can be significantly lowered by introducing a small concentration (1/9 ML) of oxygen on the surface. Upon dissociation, this oxygen creates an additional reaction path for the dissociation of the water molecule, as the proton can now be donated to a oxygen atom thus forming a hydroxyl group. As the barrier to create these hydroxyl groups is rather low, and the energy differences between both states are rather small, a chemical equilibrium will form between hydroxyl group and water formation. As this dissociation of water, presented in Eq. (8), is so rapid, the ratedetermining step of the hydroxyl formation, is actually the dissociation of oxygen molecules.
Reactions affecting the equilibrium of the reaction (8) are formation of oxygen molecule from oxygen atoms, and the desorption of water from the metal surface. According to results presented in the Table 1, on all metals but gold, the oxygen dissociation is an exothermic reaction with a rather high barrier. Conversely, this makes the formation of a oxygen molecule from oxygen atoms a rare event. As stated in the Table 3, water desorption from metal surface is an endothermic reaction. This means that once water is releasing from the surface, more adsorbed water will be created to reach equilibrium, and oxygen atoms will be available to cleave water to create more hydroxyl groups. Thus the concentration of hydroxyl groups on the metal surface should stay rather constant.

Model limitations
Model presented in this study has some restrictions. First, it presumes that there is no spontaneous oxidization on any of the surfaces. As can be seen from results in the Table 1, the oxygen dissociates very easily on rhodium and nickel surfaces, which indicates that there is a real possibility for the oxide structure to occur. There is also a requirement for constant source of oxygen, as oxygen is consumed in the reaction with water. As all of the reactions are presumed to happen in ambient air, the solid surface is presumed to contain a relatively low amount of oxygen.
In addition, it is presumed that there is no side reactions occurring in the surface. The hydroxyl-groups could form hydrogen peroxide. Our own yet-unpublished results indicate that this reaction would be very endothermic reaction, and would most likely lead to spontaneous and rapid dissociation of the formed hydrogen peroxide. A side reaction between hydroxyl group and a water molecule would only lead to products being the same as the starting materials. There is also a possibility for hydroxyl groups to dissociate into hydrogen and oxygen atoms. Whether presence of oxygen or water would catalyze this dissociation of hydroxyl group is not investigated in this study.

BEP-relationship
Reaction paths for all of the studied oxygen dissociation reactions were very similar to each other. Sequentially, the BEP-relationship is very linear, with R 2 -value of 0.95. If nickel is not taken into consideration, the R 2 -value drops a little to 0.91. Conversely, when nickel is taken into consideration on water dissociation reactions on oxygenpreadsorbed and pure metal surfaces, the linearity of the plot drops significantly: on oxygen-preadsorbed metal surfaces the R 2 -value drops from 0.98 to 0.45, and in pure metals from 0.92 to 0.78. This indicates that when considering water dissociation reactions, the group that the other metals belong to does not include nickel, the only magnetic period 4 element within the here-studied set of transition metals. The spin-polarization used for the Ni(111)-surfaces is not the cause for this departure from other metals, as the effect of spin polarization was rather small on all Ni(111)-surfaces. For activation and reaction energies of water dissociation reactions on oxygen-preadsorbed Ni(111)-surface, the difference was −0,1 eV. On pure Ni(111)-surface the effect was 0,13 eV for activation energy, and 0,16 eV for reaction energy. For oxygen dissociation reactions on Ni(111)-surface, the effect of spin polarization was −0,10 eV for activation energy, and 0,16 eV for reaction energy.

Experimental results by other groups
Our computational data supports the experimental results of other groups extremely well, even results in higher coverages and with different oxygen/water coverage ratios. Shavorskiy et al. studied the dissociation of water on oxygen-covered Rh(111) [31]. Using X-ray photoelectron spectroscopy (XPS) experiments along with low-energy electron diffraction (LEED) and near edge X-ray absorption fine structure, Shavorskiy et al. discovered that in oxygen concentrations below 0.15 ML with 160 K temperature, O is consumed, and OH is created about the twice the amount of preadsorbed oxygen coverage. This leads to water and hydroxyl containing surface with a ratio of 3:2 (H 2 O:OH), thus forming a stable layer. With rising of an oxygen concentration, the conversion of oxygen decreases linearly until at 0.3 ML, where conversion of oxygen stops completely. Also on oxygen concentrations between 0.17 and 0.3 ML, the sum of O, OH and H 2 O stays approximately constant with a maximum value around 0.6 ML. Water on clean Rh(111)-surface at temperatures below 154 K was reported to form an intact layer of chemisorbed water. Dissociative adsorption of oxygen on Rh(111) was stated to happen on temperatures higher than 300 K. This same, stable OH/H 2 O-layer was shown to be formed on Pt(111) by Karlberg and Wahnström using temperature programmed desorption (TPD) [20]. The OH/H 2 O-layer was successfully created with O 2 deposition at 90 K, then heated to 160 K to dissociate O 2 molecule to atomic oxygen. Then, temperature was hold at constant 163 K, just above the desorption temperature of water, to create OH/H 2 O-layer on the Pt(111) surface.
By using XPS and LEED, Gladys et al. [32] studied the same interaction on Pd(111). When water is dosed at temperature of 160 K, adsorbed oxygen with a concentration of 0.2 ML reacts, and thus OH/H 2 O-layer is created with a ratio of 1:2. The same outcome was reported, when water adsorbs at lower temperature after which the temperature was rose to 155-175 K. On a lower 130 K temperature, no hydroxyl groups were formed. A lone water was reported not to dissociate spontaneously on clean Pd(111) surface.
Ojifinni et al. [33] studied interactions between water and oxygen on Au(111)-surface. They reported that pure Au(111)-surface will not dissociate water molecules even at temperature of nearly 535 K. However, when Au(111) contains adsorbed oxygen (0.18 ML of 18 O) on its surface in near 155 K temperature, oxygen reacts with all of deuterated water (0.08 ML of D 2 16 O), thus producing 16 O 18 O. They suggest that hydroxyl groups are formed at the surface, but are promptly reacting to form water upon heating, leaving oxygen on the Au(111) surface. The reaction between oxygen and water is proven by oxygen scrambling, where the different isotopes of oxygen have been mixed between species.
Nakamura et al. [34] used surface X-ray diffraction, and infrared reflection adsorption spectroscopy at temperatures 25 and 140 K to study formation of hydroxyl groups on Ni(111) surface. Molecular adsorption of water molecule was stabilized due to hydrogen bonds between oxygen and water, but dissociation of water was not reported on Ni(111)-surface in 25 K temperature. Increase of the temperature to 140 K caused the molecular water to migrate to a top site of the Ni(111)-surface. However, hydroxyl group formation was discovered to happen on higher temperatures by another group. Shan et al. [35] used high resolution electron energy loss (HREEL) spectroscopy, Auger electron spectroscopy and TPD to study the formation of hydroxyl groups on a nickel surface. They identified adsorbed hydroxyl species on Ni(111) from the HREEL and TPD spectra. Hydroxyl groups could be formed either from adsorbed water and preadsorbed atomic oxygen in temperature of 170 K. Pure water layers were shown not to dissociate spontaneously, and exposure to 20 eV electrons at temperatures below 120 K were required to dissociate water thus creating hydroxyl groups without preadsorbed oxygen atoms on the Ni(111) surface. The recombination of the hydroxyl groups is observed in temperatures of approximately 180-240 K, thus creating desorbing water and leaving adsorbed oxygen atoms on the surface.
Bao et al. [36] studied the adsorption of water and oxygen on Ag(111)-surface. They found a similar behavior on Ag(111) using reflection electron microscopy and in situ Raman spectroscopy. They discovered that the interaction of H 2 O with oxygenated Ag(111) in room temperature led to stable OH-species adsorbed on the surface. In temperatures approximately 600 K and above, hydroxyl groups react to form gaseous water leaving oxygen on the Ag(111)-surface.

Conclusions
In this work, we provided a comprehensive computational model for the behavior of oxygen and water of ambient air on Ag(111)-, Au(111)-, Pd(111)-, Pt(111)-, Rh(111)-and Ni(111)-surfaces. The dissociation of water is more energetically favored when oxygen dissociates first to the surface, creating oxygen atoms, which further react to hydroxyl groups in the reaction with water. This behavior was found true for all studied metal surfaces. The computational model was in an excellent agreement with the experimental work done by other groups.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
The xyz coordinates of all the studied systems are included in the supplementary material. With these coordinates the systems can be visualized, and with an appropriate DFT code our results can be reproduced.