Ordered hydroxyls on Ca3Ru2O7(001)

As complex ternary perovskite-type oxides are increasingly used in solid oxide fuel cells, electrolysis and catalysis, it is desirable to obtain a better understanding of their surface chemical properties. Here we report a pronounced ordering of hydroxyls on the cleaved (001) surface of the Ruddlesden-Popper perovskite Ca3Ru2O7 upon water adsorption at 105 K and subsequent annealing to room temperature. Density functional theory calculations predict the dissociative adsorption of a single water molecule (E ads = 1.64 eV), forming an (OH)ads group adsorbed in a Ca-Ca bridge site, with an H transferred to a neighboring surface oxygen atom, Osurf. Scanning tunneling microscopy images show a pronounced ordering of the hydroxyls with (2 × 1), c(2 × 6), (1 × 3), and (1 × 1) periodicity. The present work demonstrates the importance of octahedral rotation and tilt in perovskites, for influencing surface reactivity, which here induces the ordering of the observed OH overlayers.

C omplex ternary oxides are intensively investigated for use in electrocatalytic reactions such as water splitting or the oxygen reduction reaction 1,2 . Of particular interest are perovskite-type oxides, with the basic formula ABO 3 (where A stands for an alkali, alkaline earth, or rare earth metal, while B refers to transition metal) and variations of this structure, such as double-perovskites, A 2 BB O 6 or the Ruddlesden-Popper series A n+1 B n O 3n+1 3 . Microscopic insights into the interaction of perovskite-type oxides with relevant molecules, in particular with water, are sorely needed for progress in this area.
Surface science can deliver such insight as was shown for metal 4 and metal-oxide surfaces 5 . The very first and fundamental question is whether a molecule in direct contact with the surface will adsorb as an intact entity or dissociate. This is governed by the subtle energy differences between interaction of water with the surface-cation on the one side and H bonding to surface O atoms on the other, and was addressed for several binary transition metal oxides. Molecular adsorption was found for water on anatase TiO 2 (101) 6,7 and on FeO(111) 8 . On TiO 2 (110) 9, 10 and RuO 2 (110)/Ru(0001) 11,12 water adsorbs molecularly as well, but an equilibrium between molecular and short-lived dissociated states is under debate 12 . Mixed molecular and dissociative adsorption was observed on Fe 3 O 4 (001) 13 , Fe 3 O 4 (111) 13 and ZnO(10-10) 14 . Water was found to exclusively dissociate on the (1 × 1) and (2 × 1) surfaces of α-Fe 2 O 3 (012) 15 as well as α-Fe 2 O 3 (0001) 13 .
The AO-terminated surfaces of perovskites are closely related to the (001) surfaces of the binary alkaline earth oxides. On MgO (001) water partially dissociates and forms two stable structures with c(4 × 2) or p(3 × 2) symmetry, depending on temperature 16 . On CaO(001) mixed dissociative and molecular adsorption were experimentally observed already at very low coverages 17 . In a density functional theory (DFT) study Hu et al. 18 derived two key factors that facilitate the dissociation of water on alkaline earth oxides: (i) an increase in the lattice constant, which enhances hydrogen bonding with the substrate, and (ii) the flexibility of the substrate. In a recent study 19 on the SrO-terminated surfaces of the layered perovskites Sr 3 Ru 2 O 7 and Sr 2 RuO 4 it was shown that the water monomer adsorption also follows these key factors. Water dissociates into a (OH) ads , adsorbed on a cation-cation bridge site, and a proton that forms a surface hydroxyl with a surface oxygen atom. The (OH) ads stays trapped by the surface hydroxyl, circling it by jumping between the four adjacent Sr-Sr bridge positions. The dynamic behavior of these ion pairs was predicted by theory 20 .
The rotation and tilting of the O octahedra in perovskites have been shown to influence ferroelectricity, magnetism, and electronic structure [21][22][23][24] , and these structural elements are key for understanding the physical properties of these complex materials. There are indications that this concept influences the structure/ property relationship in solid-state chemistry as well-for example in the context of fuel cell materials, octahedral tilting and distortion was found to facilitate inter-octahedral proton transfer 25 and O diffusion 26 , respectively, in the bulk. As motion of the octahedra renders inter-atomic distances inequivalent, this should result in different adsorption geometries and strengths of adsorbed molecules, and thus also be a decisive factor in surface chemistry.
Ca 3 Ru 2 O 7 , the prototypical perovskite material considered here, is a layered perovskite of the Ruddlesden-Popper series (alternating ABO 3 and AO layers) and cleaves easily between adjacent CaO layers. The RuO 6 octahedra are rotated in the ab plane and tilted by 12.9°with respect to the c axis. This affects the position of the surface O atom with respect to the four surrounding Ca-Ca bridge positions on the CaO terminated surface, rendering these sites inequivalent for water adsorption. This paper shows that the distance between a bridge site and its neighboring surface oxygen atoms is a pivotal factor for the water adsorption. Distinct ordering of the dissociated water is observed which can be fully rationalized with the surface geometry formed by the underlying octahedra. Based on accompanying DFT calculations, detailed structural models for the (2 × 1), (1 × 3), and (1 × 1) OH overlayers are presented.

Results
The pristine surface. As a sample, Ca 3 Ru 2 O 7 was chosen. This is the n = 2 member of the  Table 2) rather than intrinsic defects that would stem from the cleaving process.
Water adsorption. First, increasing coverages of water are followed in the O1s X-ray photoelectron spectroscopy (XPS), see Fig. 2. The pristine surface exhibits an asymmetric peak at 529.2 eV binding energy (BE), broadened towards the high BE side. According to DFT calculations the O1s core level for the oxygen atoms in the CaO plane is shifted to 0.7 eV higher BE compared to the RuO 2 planes which serve as DFT reference. Keeping in mind that the higher BE O1s signal of the CaO planes contributes less to the total O1s intensity, the observation of only one asymmetric peak in the experiment is attributed to the limited resolution of the experimental setup. For all annealed water structures observed by STM (i.e., c(2 × 6), (1 × 3) and (1 × 1), see below) the spectrum shows a shoulder between 530.0 and 531.0 eV BE; i.e., roughly 1.5 eV above the as-cleaved O peak. This fits well to the calculated BE shifts of 1.57 and 1.23 eV for surface hydroxyls in the (1 × 3) and (1 × 1) OH overlayer, respectively (see Supplementary Table 3). Only after saturating the surface with hydroxyls and dosing additional water at 105 K without annealing, a peak at 533.1 eV BE is observed; i.e., 3.9 eV above the bulk oxygen peak. This fits well to the calculated BE shift of 3.63 eV for molecular water. Therefore, in agreement with the DFT results, the shoulder and the peak are attributed to dissociated and molecular water, respectively.
The water monomer. Figure 3a shows   The unit cell is marked by the rectangle. The distances from the bridge sites B1 to B4 to the surface oxygen atoms O1 and O2 differ because of the tilted RuO 6 octahedra and are listed in Table 1. The distance B1-O1 is shown as an example, other distances are constructed analogously. b Top view and c side view of the DFT model of the most preferred configuration B1O1 for one dissociated water monomer with E ads_B1O1 = 1.64 eV. The first subsurface layer of RuO 6 octahedra is shown. Water dissociates, forming an (OH) ads group adsorbed on Ca-Ca bridge site B1 with a H transferred to the neighboring surface oxygen atom, O1. The tilt angle of the hydroxylated RuO 6 octahedron increases from 12.9°to 16.3°, when compared to the pristine surface. The energetically equivalent configuration B2O2 is not shown. d According to Tersoff-Hamann simulations, STM shows the (OH) ads as a bright feature. When compared to the Ca-Ca bridge centre (dashed line), the (OH) ads sits slightly closer to the O surf H (in [100] direction) The adsorption energies for the dissociated monomer configurations decrease with increasing distortion of the hydroxylated octahedron, reflected in the decreasing O-Ru-O bond angle along the octahedron's c axis (rows 1-3   Figure 4c shows the DFT model and the TH simulation of the (2 × 1) OH overlayer, corresponding to a 0.5 ML coverage. The calculation gives an adsorption energy decrease of 20 meV per H 2 O molecule compared to the isolated water monomer, suggesting slightly repulsive interaction between the (OH) ads (see Table 2). The brackets in Fig. 4b mark spots separated by half a unit cell along [010], named bright-dark feature. The feature is tentatively assigned to two (OH) ads on neighboring bridge sites, i.e., B1O1 and B2O2, with one of the (OH) ads appearing darker than the other. Interestingly, the darker (OH) ads are all oriented toward the [0-10] direction.
Increasing the initial dose to 0.5 L at 100 K, and annealing for 1 h at room temperature leads to a c(2 × 6) OH overlayer that covers the whole surface (see Fig. 5a, b). The OH overlayer is now compressed along [010]; 7 (OH) ads are found on six unit cells, with a shift by half an OH-OH distance between adjacent lines. The (OH) ads along [010] do not form perfectly straight lines, but slightly undulate. The coverage increase from 0.34 to 7/12 = 0.58 ML (again, with respect to O surf or Ca sites) fits well to the dose increase from 0.3 to 0.5 L when compared to the results in Fig. 4b.
Exposing the c(2 × 6) OH overlayer to an additional H 2 O dose of 0.15 L at 100 K, and annealing for 1 h at room temperature results in a (1 × 3) OH overlayer corresponding to 0.67 ML, see Fig. 5c. The surface is covered by (1 × 3) domains surrounded by a mixture of the bright-dark features and single water monomers. According to the DFT model and TH simulation, the (1 × 3) OH overlayer consists of dimers of dissociated water in the most favorable configurations along [010], i.e., B1O1 and B2O2 (see Fig. 5d), or B2O2 and B1O1, with one unoccupied bridge site between them. The bright-dark features are separated by one unoccupied bridge site from either another bright-dark feature or a water monomer (along [010]). Counting the number of brightdark features and water monomers in a large scale STM image (Fig. 6a) gives an average 0.60 ML coverage for the mixture.
Further increasing the water coverage by increasing the initial dose to the c(2 × 6)OH overlayer and decreasing the annealing time results in the (1 × 3) domains covering first an increasing part and then almost the whole surface, see Fig. 6b, c, respectively. The experimental annealing procedure (i.e., placing the sample into a RT environment for 15 min) leaves considerable uncertainty regarding the exact temperature that was reached; probably it was somewhere between 160 and 300 K.
Dosing 0.4 L to the (1 × 3) covered surface leads to bright, slightly undulating lines along [010] with almost no gaps, see Fig. 7a. The continuous, undulating lines agree well with TH simulations for a (1 × 1) OH overlayer, i.e., the full monolayer with all O surf atoms hydroxylated, see inset Fig. 7a. Dosing more water to the fully hydroxylated surface without annealing leads to diffuse, white patches in STM that are assigned to molecular water moving under the tip, see Fig. 7b.
Annealing an OH covered sample for longer times at room temperature leads to partial desorption of water. Supplementary  Fig. 2a shows the sample in Fig. 6c (i.e., almost fully covered by the (1 × 3) OH overlayer) after annealing for 3 h at room temperature. Only a few (1 × 3) OH overlayer patches remain while the sample is mainly covered by a mixture of the (2 × 1) OH overlayer and bright-dark features. A different OH covered sample annealed for 20 min at 330 K was still covered by patches of the (2 × 1) OH overlayer, see Supplementary Fig. 2b. A rough estimate for the adsorption energy based on the desorption rates and temperatures is 1 to 1.26 eV. Annealing at higher temperatures was not possible because of the epoxy glue that was used to mount the sample.
Adsorption at oxygen vacancies. Oxygen vacancies (V O ) were investigated as they are a common theme in metal-oxides and especially important for mixed ionic and electronic conductors, serving as electrodes in solid oxide fuel cells 3,28 . V O s do not form spontaneously when cleaving the samples; here they were instead created by irradiating the clean surface by 1 keV electrons 29, 30 , see Fig. 8a, b. At low temperatures no interaction was observed between the water and the V O (Fig. 8c)  room temperature bright, double-lobed features are observed that sit centred on the bright substrate lines (Fig. 8d). These features are assigned to two neighboring surface hydroxyls, formed by the dissociation of water that filled the V O . The adsorption energy for dissociative adsorption at an oxygen vacancy is 2.41 eV according to the DFT calculations. The hydrogen can be desorbed by scanning with a high positive bias voltage 31 leaving behind the pristine surface (Fig. 8e, f).

Discussion
On the basis of our XPS experiments and DFT calculations it is clear that water adsorbs dissociatively on Ca 3 Ru 2 O 7 . STM images show ordering of the (OH) ads in agreement with the calculated adsorption model. Interestingly, the evolution of ordered structures, which at first sight is rather complex, can be rationalized by considering the rotation and tilt of the O octahedra. In the most preferred configuration B1O1 (Fig. 3b) Table 2 and Supplementary Fig. 1). The order of adsorption preference (B1O1, B2O1, B3O1) is in agreement with the decreasing distances between the bridge site and O1 on the pristine surface, which, in turn, are an indicator for the natural tilting direction of the octahedron (see Table 1).
With the largest bridge site to O surf distance (i.e., B1-O1 and B2-O2) providing the highest adsorption energy only these sites are filled and, at low coverages, every other bridge site along [010] is left out, leading to a local (2 × 1) ordering of individual hydroxide pairs (Fig. 4b). As the coverage increases, the (2 × 1) OH overlayer is compressed along [010] and forms the c(2 × 6) OH overlayer with 7 (OH) ads on six unit cells (0.58 ML). The undulation probably arises from the inequivalent positions of the (OH) ads with respect to the bridge site due to the compression. Additionally, the protons of the (OH) ads may be oriented differently depending on the exact positions of the oxygens. Further increasing coverage leads to fragments of the dissociated water locating at neighboring bridge sites in the (1 × 3) OH overlayer and the bright-dark features (see Fig. 5c). The ratio of surface covered by the (1 × 3) OH overlayer to surface covered by the mixture of bright-dark features and water monomers depends on the overall water coverage. The more water, the more surface is covered by the (1 × 3) OH overlayer (see Fig. 6). This trend agrees with the (1 × 3) OH overlayer being equivalent to 0.67 ML and the mixture of bright-dark features and monomers being equivalent to roughly 0.60 ML. The skipping of available sites and the monotonic decrease in adsorption energy as the coverage increases (see Table 2) suggests that the ordering is governed by electrostatic repulsion. On BaO(001) 32 the electrostatic repulsion between (OH) ads and O surf H was shown to be screened by the Ba 2+ cations. However, the polarizability 33 of Ca 2+ is four times lower than of Ba 2+ , and Ca 3 Ru 2 O 7 (001) is less symmetric than BaO(001) thus the screening may be less effective.
The bright-dark feature is suggested to be a dissociated water dimer in two adjacent bridge sites along [010], eventually leading to a full (1 × 3) OH overlayer. However, in the precursor brightdark structure the (OH) ads that is oriented toward the [0-10] direction appears darker. A possible explanation is the inequivalence of the bridge sites available for the second water once the  The diffuse, white patches are assigned to molecular water (see Fig. 2). STM parameters: T sample = 78 K, V sample = −0.8 V, I tunnel = 0.1 nA; fast scan direction is 18°clockwise from horizontal. All scale bars correspond to 3 nm first water has adsorbed and dissociated. In one case the (OH) ads is next to the unoccupied site, in the other case the O surf H. However, DFT calculations resulted in no energy differences for the two cases.
It is instructive to compare the adsorption behavior on Ca 3 Ru 2 O 7 to the related strontium ruthenate perovskite and the simpler CaO, where the surface is much more symmetric and where detailed surface measurements are available. The dissociation into an (OH) ads fragment, adsorbed on a cation bridge site, and into a proton, adsorbed on a neighboring surface oxygen, was recently observed on the SrO-terminated surface of Sr n+1 Ru n O 3n+1 (n = 1,2) 19 . There the oxygen octahedra are not tilted with respect to the [001] direction, hence the surface oxygen sublattice is square and all bridge sites are equally spaced from the O surf . The (OH) ads fragment hops between the four equivalent cation bridge sites around the O surf H fragment at liquid-nitrogen temperature. In the present work no hopping of the (OH) ads was observed. This again agrees with one of the bridge sites surrounding the O surf being clearly preferred for adsorption (see Table 2). At low coverages dimers are not formed, but every other bridge site along the [010] direction is unoccupied as discussed above. In contrast, on Sr n+1 Ru n O 3n+1 stable dimers in adjacent Sr-Sr bridge sites are preferentially formed, evolving into one-dimensional chains 19 . On the simpler binary oxide CaO(001) the dissociation of the water monomer was predicted with adsorption energies around 0.9 eV 18, 20 . The higher adsorption energy of 1.64 eV on Ca 3 Ru 2 O 7 fits well to the reported key relevance of the lattice constant, as here the Ca-Ca distance is 10.2% larger than on CaO(001) 18 . Experimentally, mixed dissociative and molecular adsorption were observed already at very low coverages on CaO(001) 17 , in contrast to the exclusively dissociative adsorption on Ca 3 Ru 2 O 7 . It is interesting that O surf H hydroxyl pairs that form when water dissociates at an oxygen vacancy and fills it also show preferential arrangement. (Here such V O s were not observed on the as-cleaved surface but artificially created by electron bombardment.) After dosing water at 105 K and annealing to room temperature bright, double-lobed features formed (Fig. 8d). The two lobes always span a bright substrate line, indicating that the split-off proton prefers to adsorb on the O surf that has the larger distance (4.5 vs. 3.2 Å) to the surface hydroxyl that formed at the location of the healed V O .
In conclusion, water adsorbs exclusively dissociatively on Ca 3 Ru 2 O 7 (001) upon water exposure at 105 K and subsequent annealing to room temperature. The resulting ordered OH overlayers are (except for the large-cell c(2 × 6) OH overlayer) fully rationalized with DFT calculations. Molecular water is not observed in any overlayer up to full monolayer coverage. The OH overlayers show a pronounced ordering of the dissociation fragments, caused by inequivalent adsorption sites in the surface unit cell. This inequivalence originates from the distorted oxygen sublattice consisting of the apical oxygen atoms of the underlying rotated and tilted RuO 6 octahedra. The present work thus demonstrates the crucial influence of octahedral rotation and tilt on the surface reactivity of perovskites, suggesting that the engineering of these structural elements (e.g., by utilizing strain) could be useful for tuning the surface chemistry of these increasingly important materials.

Methods
Experimental set-up and sample preparation. The experiments were carried out in an ultra-high vacuum (UHV) system consisting of a preparation chamber and an STM chamber with base pressures of 2 × 10 −11 and 6 × 10 −12 mbar, respectively. A low-temperature STM (commercial Omicron LT-STM) was operated at 78 K in constant-current mode using an electro-chemically etched W-tip. The bias voltage was applied to the sample; positive or negative bias voltages result in STM images of the unoccupied or occupied states, respectively. All STM images shown were corrected for distortions as described elsewhere 34 . High-quality calcium ruthenate single crystals were grown by the floating zone technique using a mirror-focused furnace 35 . The composition of the samples was determined via inductively coupled plasma mass spectroscopy using laser ablation for direct analysis of the solid samples. Before insertion into the UHV, the samples were fixed on stainless-steel sample plates with conducting silver epoxy glue (EPO-TEK H21D, Epoxy Technology Inc.), and a metal stud was glued on top with another epoxy adhesive (EPO-TEK H77, Epoxy Technology Inc.). The crystals were cleaved in the analysis chamber at 100 K by removing the metal stub with a wobble stick (see Supplementary Note 2 for comment on success rate). Deionized water (Millipore water, purified in-house) was further cleaned by several freeze-pump-thaw cycles and was dosed in the preparation chamber while keeping the sample at 105 K. Annealing was done by bringing the sample in contact with an annealing stage, held at room temperature. For creating O vacancies, the sample was bombarded by electrons from a well-outgassed electron source in the preparation chamber with the sample held at 105 K. XPS was performed in the preparation chamber using nonmonochromatized Mg Kα X-rays and a SPECS PHOIBOS 100 electron analyzer at normal emission with a pass energy of 15 eV.
Density functional theory. The calculations were performed using the Vienna Ab-initio Simulation Package. This code employs the projector augmented plane wave formalism and PAW potentials 36 with an energy cutoff of 400 eV. As a proper treatment of dispersion effects is important for the adsorption of water 37 the socalled optB86b 38,39 functional was used. The k-point meshes used were generated by the Monkhorst-Pack 40 scheme using a Gamma-centred 6 × 6 × 1 mesh for the surface calculations on the 1 × 1 unit cell and correspondingly reduced meshes for   19 . STM images were simulated using the TH 41 approach and all core level shifts were calculated in the final state approximation 42 .
Data availability. The data that support the findings of this study are available from the corresponding author on request.