Water adsorption at zirconia: from the ZrO2(111)/Pt3Zr(0001) model system to powder samples

We present a comprehensive study of water adsorption and desorption on an ultrathin trilayer zirconia film using temperature programmed desorption (TPD), X-ray photoelectron spectroscopy (XPS), as well as scanning tunneling microscopy (STM) at different temperatures. The saturation coverage is one H2O per surface Zr atom, with about 12% dissociation. The monolayer TPD peak (180 K, desorption barrier 0.57 0.04 eV) has a tail towards higher temperatures, caused by recombinative desorption from defect sites with dissociated water. STM shows that the defects with the strongest H2O adsorption are found above subsurface dislocations. Additional defect sites are created by multiple water adsorption/desorption cycles; these water-induced changes were also probed by CO2 TPD. Nevertheless, the defect density is much smaller than in previous studies of H2O/ZrO2. To validate our model system, transmission Fouriertransform infrared absorption spectroscopy (FTIR) studies at near-ambient pressures were carried out on monoclinic zirconia powder, showing comparable adsorption energies as TPD on the ultrathin film. The results are also compared with density functional theory (DFT) calculations, which suggest that sites with strong H2O adsorption contain twofold-coordinated oxygen.


Introduction
Many technological applications of zirconia (ZrO 2 ; pure or yttrium-doped, then known as yttria-stabilized zirconia YSZ) involve interaction with water.Examples are internal steam reforming in solid oxide fuel cells, 1 catalysis, 2 gas sensors, 3 or applications as biocompatible material. 4However, little is known about the interaction of water with ZrO 2 surfaces on a fundamental level, which is mostly due to a lack of suitable samples.6][7] Water adsorbs only weakly on many defect-free oxide surfaces; in ultra-high vacuum (UHV) it then desorbs below room temperature (RT), typically at 160-250 K. 8-10 H 2 O can bind more strongly to surfaces with defects, as shown for the prototypical example of rutile TiO 2 (110). 11,12In these cases, H 2 O dissociates into an OH group, which lls an oxygen vacancy, and into a hydrogen atom, which binds to surface oxygen and forms a second OH group.On TiO 2 , these OH groups are stable up to x490 K. 13 On defect-free oxide surfaces, water can bind strongly if the surface termination includes highly reactive cations.It can then readily dissociate, e.g. on a-Cr 2 O 3 (001), 14 a-Fe 2 O 3 (012), 15 and on oxides of the alkaline earths, 16 including CaO-terminated Ca 3 Ru 2 O 7 (001). 17On RuO 2 (110), 18 PdO(101), 19 and Fe 3 O 4 (001), 20 water binds to coordinatively unsaturated cations and forms partially dissociated structures stabilized by hydrogen bonds.
Such detailed studies are not yet available for water on ZrO 2 ; only few reports on H 2 O adsorption can be found in the literature.On powder materials, very high adsorption enthalpies were reported for low H 2 O coverages (T2 eV on monoclinic and z1.5 eV on tetragonal ZrO 2 ), decreasing to liquid-water like binding (0.45 eV) at coverages around 2-4 H 2 O/nm 2 . 21,22Moderate values (z0.6 eV, derived from a TPD peak at 240-250 K) were reported for H 2 O on oxidized polycrystalline Zr, i.e., a hydroxylated and, possibly, substoichiometric lm. 23To the best of our knowledge, there are no studies of H 2 O adsorption on well-dened singlecrystalline ZrO 2 surfaces so far.This is related to the fact that single crystals of pure ZrO 2 grown from the melt exhibit phase transformations upon cooling, thus only crystals of doped zirconia (e.g., YSZ) are available; these retain the hightemperature cubic phase.The surface chemistry of YSZ is much more complex than that of pure ZrO 2 , however, as shown for adsorption of H 2 O, 24 hydrogen, 25 or CO and CO 2 . 26herefore, the investigation of pure ZrO 2 , which is needed as a starting point of a well-grounded fundamental understanding, has to rely on thin lms.
Well-dened, pure zirconia surfaces can be created as thin lms, which have the additional advantage that the surface of the otherwise perfectly insulating zirconia can be probed by techniques that rely on electron transfer, such as scanning tunneling microscopy (STM).8][29][30] Temperature programmed desorption (TPD) studies on 7 ML-thick ZrO 2 lms on Pt(111) show three desorption peaks (190 K, 275 K, and 370 K with adsorption energies 0.46, 0.68, and 0.92 eV, respectively). 31,32The authors suggested that water in the low-temperature peak binds molecularly while the rest is in dissociated form.The two desorption peaks at higher temperatures were assigned to defect sites.
Ultra-thin ZrO 2 lms can also be prepared by oxidation of suitable Zr alloys, Pt 3 Zr 33,34 or Pd 3 Zr. 35This approach has the advantages of better homogeneity of the lms, especially a more uniform thickness, and is also less susceptible to impurity adsorption as these alloys are much less reactive than pure Zr.By a combination of STM, Auger and photoelectron spectroscopy, as well as DFT, it has been shown that the lms grown by oxidation of Pt 3 Zr or Pd 3 Zr consist of one trilayer (O-Zr-O), structurally equivalent to a trilayer of cubic ZrO 2 (111), but with additional distortions. 33,35,36The oxide lm is created by Zr diffusion to the surface when annealing in oxygen.In the case of Pt 3 Zr, diffusion of Zr in the alloy is slow, thus a Zr-depleted Pt region remains below the oxide.The slightly smaller interatomic distance of Pt (0.2775 nm) in comparison with Pt 3 Zr (0.281 nm) leads to a contraction of the Pt layer below the oxide lm.As a consequence, mist dislocations form between the Pt layer at the interface and the alloy below.These subsurface dislocations are visible as bright ridges in STM.The atoms in the ultra-thin ZrO 2 trilayer above the Pt layer are well ordered with an average in-plane distance of 350 pm and exhibit strong vertical buckling.These height differences are related to the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 superstructure (1.2 nm periodicity), resulting from the different lattice constants of the oxide and the underlying Pt layer.The unit cell of the superstructure includes 12 Zr atoms, which are accessible to adsorbates due to the large distance between the O atoms in the layer above, and the low interlayer distance between the O and Zr layers. 33A model of the superstructure as calculated by density functional theory (DFT) is presented in the DFT chapter below.
In this work we present an extensive study of water on ZrO 2 ultra-thin lms grown by oxidation of Pt 3 Zr(0001) single crystals.We combine TPD, X-ray photoelectron spectroscopy (XPS), and STM with DFT calculations and show how adsorbed H 2 O behaves on zirconia trilayers.Furthermore, we performed Fourier-transform infrared spectroscopy (FTIR) absorption measurements of H 2 O on monoclinic ZrO 2 powder, to test the validity of the ultra-thin oxide as a model system for technological zirconia surfaces.

Ultra-high vacuum setup
UHV experiments were conducted in three different systems.Temperature-programmed desorption (TPD) and X-ray photoelectron spectroscopy (XPS) measurements were performed in a UHV chamber described thoroughly elsewhere. 37his system features a LHe ow cryostat, which can cool the sample to 20 K.The base pressure in the chamber was below 10 À10 mbar.A Pt 3 Zr(0001) single crystal (6 mm diameter) was wrapped with Ta ribbons at the circumference; these were spotwelded to thicker Ta bars leading directly to the cryostat to ensure good thermal contact.A K-type thermocouple was spotwelded directly to the backside of the crystal for accurate temperature measurements.No sample plate was used.The chamber includes a molecular-beam setup for precise gas dosing (such as D 2 O) with a sharp top-hat prole in a circular area with a diameter of 3.2 mm. 37This allows us to perform TPD and XPS measurements with water sticking exclusively to the center of the well-prepared crystal surface; there is no inuence from water on the Ta ribbons or from the edges of the crystal.Furthermore, the molecular beam provides accurate gas doses.Since the sticking coefficient equals unity, the coverage can be given in monolayers or molecules per trilayer Zr atom.We dene one monolayer (ML) as one water molecule per surface Zr atom (9.5 nm À2 , corresponding to 12 molecules per ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 ZrO 2 trilayer unit cell).For XPS, a monochromatised Al Ka (1486.7 eV) X-ray source was used.All XPS measurements were done at 100 K and in normal emission.To reduce the inuence of the residual gas on the TPD data, experiments in this chamber (TPD, XPS) were done with D 2 O, while H 2 O was used everywhere else.
Scanning tunneling microscopy (STM) measurements were performed in two different chambers.Low-temperature STM was performed using an Omicron LT-STM in a two-chamber UHV system with a base pressure below 1.5 Â 10 À11 mbar in the STM chamber, and 4 Â 10 À11 mbar in the preparation chamber.During the measurement the sample was cooled to 78 K. Room-temperature STM measurements were preformed using an Omicron m-STM in a chamber described in ref. 33.This two-chamber system is divided into a measurement chamber (p base < 7 Â 10 À11 mbar) and a preparation chamber (p base < 10 À10 mbar).STM data were acquired in constant-current mode with electrochemically etched W tips.We report sample voltages for the STM images, thus positive voltages refer to tunneling into the unoccupied states of the surface.All STM images were corrected for creep of the piezo scanner as described in ref. 35.

In situ FTIR spectroscopy setup
Fourier-transform infrared absorption spectroscopy (FTIR) measurements of ZrO 2 powder were recorded in transmission mode on an Agilent Cary 660 spectrometer with a mid-infrared source and a deuterated triglycine sulfate (DTGS) detector.The powder samples were pressed into thin pellets using a pressure equivalent to 1.5 t on a 0.8 cm 2 area (sample diameter 10 mm, mass about 20 mg) and subsequently placed inside a homemade in situ/operando reactor cell. 38This cell provides an inert all-quartz surrounding of the sample in the heated area.In situ measurements can be performed up to 1273 K under owing and static conditions.Also, vacuum with a minimum pressure of 10 À7 mbar is obtainable.The window material, BaF 2 , provides access to wave numbers above 800 cm À1 .Experiments in owing mode were performed using He as carrier gas.In static mode, water vapor was admitted stepwise to the evacuated cell.The water reservoir was degassed beforehand to remove dissolved atmospheric gases, in particular CO 2 .All reported spectra are corrected by the spectrum of the dry pre-oxidized oxide pellet at room temperature and under vacuum, prior to exposure to water.

Sample preparation
The Pt 3 Zr(0001) single crystals used for UHV studies were grown, cut and polished by MaTecK (Germany).The cleaning procedure was based on the recipe from ref. 33, with slight changes: cycles of sputtering (2 keV Ne + ions in the XPS/TPD chamber, or Ar + in the STM chambers; 20 min, current density z 4 mA cm À2 ) and annealing (1175 K, 10 min).The last sputter cycle was applied during a linear temperature ramp from 680 K to 380 K in 20 min to reduce the density of steps and eliminate half-steps. 33The cleanliness of the sample was checked with XPS.An ultra-thin ZrO 2 trilayer (O-Zr-O) was prepared by rst annealing in oxygen (p ¼ 1 Â 10 À7 mbar, 680 K, 10 min) and then in UHV (1205 K, 30 min).The rst annealing step causes Zr diffusion to the surface and oxidation, but results in poorly ordered structures.In the second step the oxide forms a well-ordered ultra-thin O-Zr-O trilayer; three-dimensional (3D) ZrO 2 clusters disappear by spreading out and/or dissolving into the bulk.In the present study, the nal annealing temperature was set higher than in ref. 33 and 36 to reduce the amount of ZrO 2 clusters.Although both, the annealing temperature and the annealing time, were increased compared to the recipe from Antlanger et al., 33 the lm did not break up, as veried by CO TPD, which did not show any indications of a CO signal from adsorption on the substrate. 36The resulting sample is a trilayer of ZrO 2 (111) on a Zr-depleted Pt 3 Zr(0001) surface.In the following this is shortened to "ZrO 2 /Pt 3 Zr".For the LT-STM measurements, the sample could be heated to only 1160 K; nevertheless, STM showed large areas free of ZrO 2 clusters (Fig. S1 of the ESI †).The preparation procedures are summarized in the ESI (Table S1 †).
Both D 2 O (TPD and XPS measurements) and H 2 O (STM measurements) were puried via several freeze-pump-thaw cycles.All gases were checked for cleanliness with mass spectrometers.For TPD and XPS measurements, water was dosed using the molecular-beam doser (see above), for STM measurements water was dosed by back-lling of the chamber, making the given doses less accurate.
For the FTIR experiments, commercial powder of monoclinic ZrO 2 (zirconium (IV) oxide, 99.978%; Alfa Aesar) was used.To guarantee a well-dened, carbon-and water-depleted and sufficiently (for the subsequent experimental temperatures) sintered material, the sample was heated inside the in situ setup in pure, dry oxygen up to 1273 K and held for one hour prior to the water adsorption experiments.The sample was routinely checked by XRD for structural changes aer annealing and aer the water adsorption experiments.Aer the pretreatment, the surface area was determined as 2 m 2 g À1 (ZrO 2 ) by nitrogen adsorption at 77 K according to the Brunauer-Emmett-Teller (BET) method.For BET measurements, a Quantachrome Nova 2000 Surface Area and Pore Size Analyzer was used.Gases were supplied by Messer (O 2 5.0,He 5.0).The cooling trap for removing residual moisture from O 2 was set at $153 K.

Computational methods
The DFT calculations were performed with the Vienna Ab initio Simulation Package (VASP) using the projector augmented wave framework. 39Earlier work has shown that dispersion effects can play a substantial role in the correct description of both the metal-zirconia 33 and the solid-water interface. 16,40,41Therefore, the van-der-Waals corrected so-called optB86b 42,43 functional was used to properly treat dispersion effects, employing the formalism introduced by Dion et al. 44 Due to the large size of the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 model cell, a single k point at the G point was sufficient to properly describe the electronic structure of the model cell.The energy cutoff was set to 400 eV.The structures were relaxed until the residual forces were below 0.01 eV ÅÀ1 .STM simulations were calculated using the Tersoff-Hamann 45 formalism.To supplement the XPS measurements, core level shis of the oxygen 1s states were obtained in the nal state approximation. 46Dissociation barriers were calculated using the improved dimer method. 47,48Results

General aspects of water adsorption and desorption
3.1.1TPD.Temperature programmed desorption (TPD) measurements form the backbone of the analysis of the behavior of water on ZrO 2 /Pt 3 Zr.Fig. 1(a) shows three distinct regions in the TPD spectrum for D 2 O: two desorption peaks with maxima at 150 and 180 K, and a tail extending towards high temperatures.We also checked for simultaneous desorption of other species (m/z ¼ 3, 4, 18, 19, and 28) and found no peaks apart from the D 2 O cracking products.The desorption temperature of 150 K is typical for multilayer water. 9When plotting this peak in a log(I) vs. 1/T plot, see Fig. 1(b), the ascending slope (dotted line) yields a desorption energy of 0.47 eV, 49,50 which is slightly lower than expected for multilayer D 2 O ice (0.53 AE 0.02 eV; ref. 51).This may be due to the secondlayer ice not having fully developed its crystalline order.
The TPD peak with the maximum at 180 K exhibits a rstorder desorption behavior and reaches saturation at a coverage of approximately 1 ML (one D 2 O molecule per Zr atom in the oxide); it is therefore the monolayer peak.The area of the monolayer peak does not stay perfectly constant during repeated TPD measurements, see the section on water-induced changes, below.On the high-temperature side of the monolayer peak, starting at approximately 190 K, the desorption rate decreases more slowly than expected for a rst-order peak.The signal forms a long tail and vanishes below the detection limit at z540 K, see inset of Fig. 1(a) (comparison with D 2 O TPD of other surfaces shows that the tail is not an instrumental artifact).The tail may consist of a multitude of peaks that cannot be discerned from each other.The coverage in the tail corresponds to approximately 0.15 ML, while the rest of the monolayer makes up the peak at 180 K.When plotting the desorbed amount of water, taken from TPD, vs. the dosed amount, the intercept of a linear t yields a dose of 0.01 ML (Fig. 1(d)).This shows, within the error of such an analysis, that no water was present on the surface before dosing.
The differential desorption energy E des of water bound in the monolayer peak can be calculated using the inversion analysis method: 52,53 the Polanyi-Wigner formula (1), which describes the desorption rate Àdq/dt in dependence of coverage q and temperature T, is inverted under the assumptions of rst-order desorption (n ¼ 1), the prefactor n being independent of q, and a constant heating rate b: where T is the temperature and t is the time.Eqn ( 2) is fully dened by the measured TPD data except for n.To determine n and E des , we follow ref.54 in using eqn (1) to simulate TPD spectra that consist mainly of the monolayer peak.In our case of D 2 O on ZrO 2 /Pt 3 Zr, the best overlap between experimental and simulated curves is achieved at n ¼ 10 15AE1 s À1 .This is a typical value for water desorption. 8Fig. 1(c) shows E des as a function of q, derived for an initial coverage of 1 ML.The desorption energy stays almost constant between 0.55 and 0.95 ML and amounts to 0.57 AE 0.04 eV, taking the error bars of n into account.
At lower coverages (well inside the high-temperature tail), the results of the inversion analysis in Fig. 1(c) give adsorption energies of up to z1.5 eV.These energies should be treated with caution; here n might vary considerably from the value determined for the monolayer peak (e.g.due to recombinative desorption or surface restructuring during desorption).clusters, and the ZrO 2 trilayer, see Fig. 2(a).Fig. 2(b, c) shows the peak areas of these four different O 1s components aer various ash-annealing steps.The peak at 530.1 eV originates from oxygen bound in the ZrO 2 trilayer and is very close to the reported value of 529.9 eV. 36This peak does not change signicantly with coverage or temperature, except for a slight decrease when the intensity is dampened by 2 ML D 2 O. Within the accuracy of the ts, the peak attributed to 3D clusters also remains constant.(The trilayer and 3D ZrO 2 clusters have distinct O 1s and Zr 3d spectra; 36 the 3D cluster/trilayer ratio is consistent for O 1s and Zr 3d, which conrms the correctness of our peak tting.An STM image of a 3D ZrO 2 cluster can be found in the ESI.†) The O 1s peak at 533.8 eV originates from non-dissociated D 2 O. 6,55,56 When ash-annealing to 160 K, which is above the desorption temperature of the multilayer peak, yet below the desorption temperature of the monolayer peak, this peak decreases by roughly 50%, in agreement with only 1 ML remaining on the surface, see Fig. 2(c).When annealing to 180 K (maximum of the monolayer peak), the molecular D 2 O peak decreases further and nearly vanishes at 200 K (beyond the monolayer peak).When annealing to 550 K, the spectrum decreases in the region 531-532 eV, i.e., between the two main peaks.This region contains signals from oxygen in zirconia clusters (E B ¼ 531.0 eV) 36 and from hydroxy groups (E B ¼ 531.8 eV); a shi of 1.7 eV between the trilayer oxide peak and the hydroxy peak lies within the expected range. 6,55,57,58ince zirconia clusters do not change when annealed at 550 K, the decrease stems solely from hydroxy groups.The TPD hightemperature tail is therefore due to dissociated water, which recombines before desorption.When comparing the O 1s signal from OD groups and from the ultra-thin lm to XPS simulations using the SESSA program, 59 the amount of water present as hydroxy groups is estimated as 0.12 ML.A small hydroxy component appears also on the as-prepared oxide [labelled "clean" in Fig. 2(b, c)], possibly due to adsorption from the residual gas.D 2 O dissociation induced by X-ray beam damage was excluded by dosing 2 ML of D 2 O at 100 K and annealing the sample directly to 200 K.This yielded exactly the same result as shown for 200 K in Fig. 2(b, c), showing that there is no dissociation while taking several X-ray photoelectron spectra.
3.1.3Scanning tunneling microscopy.Fig. 3(a) shows 0.12 AE 0.04 ML of H 2 O (dosed via back-lling of the chamber) on a freshly-prepared ZrO 2 /Pt 3 Zr surface at 100 K.This coverage corresponds to the high-temperature TPD tail.During measurements, the sample was held at 78 K.Both temperatures lie well below all desorption peaks of water, see above.Three different regions are present in the STM image: ordered, apparently uncovered and clustered areas.The water species in the ordered areas are stable during imaging; they typically appear as 80-100 pm high protrusions.The ordered areas mostly show the same periodicity as the ZrO 2 trilayer, i.e., ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 with respect to the Pt layer below. 33The clustered areas, presumably with a higher local coverage, appear fuzzy, indicating that the water molecules are moving under the inuence of the tip.This suggests that one or two adsorption sites per unit cell are more stable than the others.Following the XPS measurements shown above, which show that the most stable adsorption sites are occupied by hydroxy groups, we attribute the ordered features to OH; this This journal is © The Royal Society of Chemistry 2018 assignment is conrmed below.Dissociation of H 2 O leads to two hydroxy groups, one (O w H, also known as terminal OH) containing the oxygen atom of the water, and the other formed by a surface oxygen atom and the split-off proton (O s H).These two OH groups should appear differently in STM.As we see only one type of distinct protrusion, we have to assume that STM shows the terminal OH, which is geometrically higher, while an adjacent O s H is invisible due to its smaller apparent height.(This is in agreement with DFT, see below.)The coverage of one dissociated H 2 O, i.e., two OH groups, per ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 unit cell was used as the input for an XPS simulation using SESSA. 59The result shows that this coverage accounts for z2/3 of the hydroxy groups seen in XPS.
When measuring H 2 O on ZrO 2 at room temperature (well inside the high-temperature tail), only a low coverage of H 2 Orelated protrusions is found even aer supplying large gas doses, see Fig. 3(b).This is in agreement with the TPD results.Again, the protrusions appear about 80-100 pm high.The p ÞR23:4 ordering can still be made out locally.Water adsorbs preferentially above the subsurface dislocation lines.As mentioned above, these stem from the mist between the Zr-depleted Pt layer between the ZrO 2 trilayer and the Pt 3 Zr substrate. 33Furthermore, water adsorbs on top of rotational domain boundaries 34 of the oxide lm (these are rather sparse; there is none in Fig. 3).Fig. 3(c) and (d) show single water species (hydroxy groups) adsorbed on a ZrO 2 trilayer, imaged with a special tip that lets these species appear as dark depressions.Assuming that the bright protrusions in the STM image (c, d) correspond to Zr as suggested for "normal" tips in ref. 33, the red lines indicate that the adsorption site of the water species is above a Zr atom.This is consistent with the site expected for the terminal O w H.

Water-induced surface changes
While repeating D 2 O TPD measurements on ZrO 2 /Pt 3 Zr, the monolayer peak changed over the course of several cycles, see Fig. 4(a).There, 18 successive TPD measurements were carried out, each starting with 2 ML D 2 O dosed at 100 K and then ramped to 550 K (with one exception for the dose as noted below).The monolayer peak area decreases by 8% from the rst to the 18 th run, with the biggest decrease during the rst cycles.At the same time, the amount of water desorbing above 185 K increases, such that the total amount of desorbed water remains constant (note that also the area of the multilayer peak stays constant within the experimental accuracy).Thus, the decreasing monolayer peak height is not due to an increasing amount of water remaining on the surface between TPD runs.These results rather show that, under the inuence of repeated water adsorption/desorption cycles, adsorption sites with low E des (T des ¼ 180 K) change to adsorption sites with higher E des .Between the 8 th and the 18 th run, eight TPD cycles with 2 ML and one with 15 ML were interposed (not shown); dosing 15 ML affects the surface in the same way as 2 ML.The changing height of the 180 K peak is also visible in Fig. 1(a).(Note that the measurements in Fig. 1 were not done in the sequence of increasing or decreasing coverage, thus the 180 K peak heights at coverages above 1 ML seem to scatter randomly.)Such changes did not appear in multiple TPD cycles with CO 2 , so they must be caused by the interaction of the surface with water, not by the heating/cooling cycles alone.
It was found that CO 2 TPD is a good indicator for the waterinduced surface change, see Fig. 4(b). 1 ML of CO 2 , corresponding to 1 CO 2 per surface Zr atom, was dosed (at 50 K) and desorbed by ramping the temperature to 300 K before the rst and aer the 18 th D 2 O TPD run.The CO 2 multilayer peak is at 78 K, 37 and the low-temperature (90-100 K) shoulder of the monolayer peak is attributed to compression of the CO 2 layer. 60,61The monolayer peak (104 K) and the multilayer peak have roughly the same area, thus approximately one CO 2 molecule per two Zr atoms forms the monolayer.Aer the water adsorption/desorption cycles, the CO 2 monolayer peak area decreases and the high-temperature tail (105-130 K) increases by approximately 0.05 ML (z10% of a CO 2 monolayer).When repeating the CO 2 TPD, there is no further change.The TPD behavior indicates that CO 2 binds more strongly to the newlycreated sites.The moderate increase of the CO 2 adsorption energy points to electrostatic interaction between OH and the quadrupole moment of CO 2 , rather than chemical bonding of CO 2 , e.g. as bicarbonate (on TiO 2 (110), the CO 2 TPD peak attributed to bicarbonate is at z213 K, ref. 62).We also note that the combined coverage of the monolayer (including the low-T shoulder) and tail slightly decreases aer the D 2 O cycles (by z5%).This reduction of the total amount of CO 2 bound in the rst layer seems to be mainly related to less compression (weaker low-T shoulder).It is likely that the newly created sites with higher CO 2 and D 2 O desorption energy pin the CO 2 layer, preventing the formation of a well-ordered compressed CO 2 layer.
We have also used STM to search for water-induced surface modications.Aer one cycle of dosing 1.5 AE 0.3 ML water and annealing to 550 K, the surface appears as shown in Fig. 5 indicating that some sites in the superstructure are preferred over others.Based on the real-space images, there must be at least two such preferred positions.This behavior is similar to H 2 O dosed at RT, see above.This observation can not be explained by H 2 O or OH simply remaining on the surface bound to very stable adsorption sites without any surface change, as the density of water species (and therefore the density of sites) increases with the number of cycles.Likely, there is some modication (deformation) of the oxide lm, leading to This journal is © The Royal Society of Chemistry 2018 stronger bonding to OH.Nevertheless, the oxide lattice remains largely intact as demonstrated by the sharp FFT spots.

Infrared measurements on monoclinic ZrO 2 powder samples
Water adsorption on a pre-sintered powder sample of monoclinic ZrO 2 was studied in isothermal (at room temperature) and isobaric experiments.For the isothermal measurements, water was dosed at pressures between 4 Â 10 À3 and 25 mbar (saturation vapor pressure).The infrared absorption spectra do not show any substantial discrete peaks but only a broad band of H-bonded species (Fig. 6(a)). 24As this band does not change shape in the pressure range of interest, the amount of adsorbed water on the sample can be approximated by plotting the relative height/absorbance of this band versus the partial pressure.This approximation is justied by comparison with gravimetric experiments up to pressures where water condensation in capillaries appears, at about 80% relative humidity. 24Since the absorption spectra show strong water gas-phase peaks at pressures above 10 À2 mbar, the absorption was determined at wavenumbers near 3160 cm À1 (vertical line in Fig. 6(a)), where the signal intensity of gaseous water is zero (Q-branch of the rst overtone of the d rotationalvibrational mode 63 ).An absolute calibration of coverage vs. absorption is not possible due to light scattering in the pressed powder. 64Based on gravimetric data, 24 the FTIR signal at saturation water pressure (p H 2 O ¼ 25 mbar) and room temperature corresponds to 20-30 molecules per nm 2 , which is a factor of 2-3 above our previous monolayer denition of 1 H 2 O per surface Zr.
Initial adsorption also induces a slight negative peak between 3700 and 3800 cm À1 , which we attribute to a small coverage of isolated species, becoming part of the H-bonded network at higher coverage.This frequency range indicates OH bound to one or two Zr atoms.Comparison of the shape of the band of H-bonded species between different pressures shows also an initial negative signal of comparable size superimposed on the broad band above 3650 cm À1 .Absorption in  this frequency range is frequently observed on monoclinic ZrO 2 and attributed to threefold-coordinated OH. 65,66 In addition, an isobaric experiment was performed.This experiment was performed in owing mode by passing helium through a water vapor saturator at room temperature (p H 2 O ¼ 25 mbar) and monitoring the infrared absorption as a function of temperature.These data are less accurate than the isothermal ones due to various instrumental factors as well as the increasing appearance of ZrO 2 lattice vibrational overtones in the spectra; nevertheless the results are valuable for the analysis (see below).
For determination of adsorption enthalpies, we make use of the fact that the chemical potentials m ad of the adsorbed and m g of the gas phase must be equal in equilibrium: where the enthalpy H g and entropy S g per molecule in the gas phase is calculated from tabulated data (assuming an ideal gas). 67E ad is the adsorption energy per atom.We dene E ad as a positive number for stable adsorption sites, as above, thus the minus sign.The volume of the adsorbed H 2 O can be neglected, i.e., its energy and enthalpy are assumed to be equal.For calculating the adsorption energy, E ad , which can be directly compared to the DFT and TPD results above, we need to know S ad , the entropy per molecule in the adsorbed state.S ad has a congurational contribution due to the multiple adsorption sites available below saturation, with an upper limit of Àk B ln q (we use the saturation coverage for q ¼ 1, but changing this to, e.g., 1 H 2 O per surface Zr atom has only a very minor inuence on the calculated E ad ).This upper limit is reached if all sites are equal and there is no interaction between adsorbates.The other contribution to S ad comes from the adsorbate's vibrational and orientational degrees of freedom.For submonolayer coverages, the substrate provides a corrugated potential-energy landscape acting as a template for the adsorbate preventing a liquid layer, see the DFT results below.Therefore, we may safely assume that the adsorbed H 2 O is bound and conned at least as strongly as in ice.Thus, we take the entropy of ice, extrapolated to temperatures above 0 C as an upper limit of its entropy: Since S extrap ice (T) is almost linear in temperature between 150 and 273 K, 68 we use a linear t, which yields S extrap ice /(eV K À1 ) ¼ (3.69 Â 10 À5 + 1.42 Â 10 À6 T/K).(We have also tried using the entropy of liquid H 2 O here.This would result in unphysical adsorption energies below the vaporization energy of H 2 O, however, supporting our choice of ice for estimating the entropy.)Fig. 6(b) shows the result of this calculation for both the isothermal (T ¼ 298 K) and the isobaric (298-573 K) experiments.The bars span the range between the two limits for the entropy, zero and S max ad (4).For low coverage, these two experiments correspond to substantially different temperatures, thus a comparison between them can be used to estimate the entropy.As expected, the agreement is better for the upper limit of the entropy, i.e., the congurational entropy of the adsorbed layer cannot be neglected and the vibrational and orientational entropies are similar to ice.Nevertheless, we think that the actual entropy must be slightly less than the upper limit since the assumption of S max ad yields a reversal of the slope of the isobaric adsorption energies at low coverages, which we consider unrealistic.Therefore, our best guess is an adsorption energy of z0.65 eV at low coverages, decreasing to values close to the binding energy in ice at 1 ML.It should be noted that similar values of the adsorption energies could be also obtained from a Langmuir adsorption isotherm (z0.62 eV over a large range of coverages, when assuming a constant prefactor of 10 15 s À1 ).This method was already used for adsorption studies on zirconia. 24,69ompared with the present approach, the disadvantage of the Langmuir model is that the entropies of the gas and the adsorbed phase are hidden in the prefactor and effects such as temperaturedependent rotational and vibrational gas-phase entropies or coverage-dependent congurational entropies of the adsorbed phase will necessarily lead to a non-constant prefactor.
As mentioned, the adsorption energies in Fig. 6(b) are based on the broad absorption band and do not include the isolated species giving rise to the negative adsorption peak discussed above.Thus, one should expect the zero of the coverage scale of Fig. 6(b) a bit further to the le than shown, with higher adsorption energies at these very low coverages.

DFT calculations
The experiments by Antlanger et al. 33 have shown that the ultrathin ZrO 2 lm forms a commensurate ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 supercell with respect to the Pt lattice, which forms the interface to the Pt 3 Zr(0001) substrate below.Each unit cell contains 12 formula units of ZrO 2 , which results in an average ZrO 2 in-plane lattice constant of 350 pm.Since accounting for the alloy as the substrate would require a very large unit cell (>500 atoms), a pure 5-layer Pt(111) slab served as the supporting structure of the ZrO 2 lm in our calculations (d Pt-Pt ¼ 277.8 pm; bottom two Pt layers xed).The viability of such a model has been shown by Meinel et al. 28 and for modeling the adsorption of metal adatoms by Choi et al. 34 3.4.1The ZrO 2 lm.Structural optimization using a simulated-annealing approach leads to a heavily distorted ZrO 2 lm, shown in Fig. 7(a).With the occurrence of twofold coordinated O atoms at the surface, this structure is similar to the one found by Puigdollers et al. 70 using a PBE+U functional, and very similar to our previous result 34 (which is only a local energy minimum according to our present study).Probably due to the at potentialenergy landscape with a large number of local minima, also the present structure does not perfectly represent the experimental one in all details.E.g., Fig. 3(c) suggests a higher symmetry than the DFT model in Fig. 7a.Nevertheless, the main structural elements should be represented well enough to capture the key features for the interaction with the water molecules.The interlayer distances between the mean layer heights, d O-Zr , d Zr-O , and d O-Pt , are 93, 72, and 223 pm, respectively, and all oxide layers are strongly buckled (125, 116, and 96 pm peak-peak for the upper O, Zr, and lower O layer).The group of six high-lying Zr atoms, marked with a dashed white triangle in Fig. 7(a), have contracted Zr-Zr distances of 323 to 347 pm.They are separated by low-lying Zr ("valleys") with Zr-Zr distances up to 389 pm.In the high Zr triangles, the interaction between the ultra-thin oxide lm and the Pt(111) substrate is dominated by the lower oxygen atoms of the trilayer, while in the valleys the Zr atoms bind strongly to the Pt surface atoms.This distortion breaks four Zr-O bonds at the surface, leading to a twofold coordination of some of the oxygen atoms at the borders of the high triangles.
3.4.2Adsorption of H 2 O on the ultrathin lm.To screen a large number of adsorption sites, 36 starting congurations with the H 2 O molecule positioned in a 6 Â 6 grid with respect to the surface unit cell were tested.Aer the relaxation, the H 2 O molecule had always moved to one of the twelve Zr surface atoms.The adsorbed H 2 O molecule does not dissociate spontaneously in any of these twelve sites.Due to the distortions of the ZrO 2 lm, these H 2 O adsorption sites have substantially different local geometries, which leads to a wide range of calculated adsorption energies, from 0.48 to 0.99 eV (see Fig. 7(a)).For isolated H 2 O molecules, adsorption at the laterally compressed "high" Zr atoms is mostly on the weaker side, (0.48-0.77eV), also reected by comparably long Zr-O w distances around 236 pm.Adsorption is stronger at valley sites (0.62-0.99 eV), where the ultra-thin ZrO 2 lm is stretched, making the Zr atoms more reactive and accessible.The strongest adsorption is found where the H 2 O molecule can form a hydrogen bond to an undercoordinated (twofold) oxygen atom at the border of a "high" Zr group (marked with energy 0.98 in Fig. 7(a); the site "(0.99)" is equivalent, aer shiing of the lm).A side view of this position with the O-H bond lengths is shown in Fig. 7(b).Due to the short H bond, dissociation via proton hopping is easy (67 meV barrier in DFT).At this site and one other (marked 0.71/0.77 in Fig. 7(a); not counting sites where the oxide lm gets substantially shied), dissociation is slightly more favored than molecular adsorption, see Fig. 7(c).
STM simulations show essentially identical images for molecular and dissociated water, see Fig. 7(d, e); the maxima are caused by the adsorbed water molecule and the terminal O w H, respectively.This is not surprising as the difference between the two structures is only a small displacement (z40 pm) of the lower of the two protons, and an even smaller displacement of the O w (z14 pm).
Upon adsorption of a full monolayer, i.e., one H 2 O molecule per Zr atom, the mean adsorption energy decreases to a value of 0.68 eV.It should be noted that the calculated ground state is not formed when each Zr atom is covered by one H 2 O molecule.Instead, the H 2 O molecules cluster above the "high" Zr atoms, connected by H 2 O chains across the valleys.However, our calculations show that the optB86 functional overestimates the formation energy of hexagonal bulk ice Ih with a calculated value of 0.74 eV compared to the experimental value (z0.58 eV,  ref. 71).This overestimation of the binding between H 2 O molecules will articially increase the tendency for the formation of water clusters.Nevertheless, our calculations show that the adsorption energy in the high-coverage limit is close to the binding energy of multilayer ice.
We have also calculated the core level binding energies of the O 1s states to conrm the identication of the different species by XPS.In the nal-state approximation, taking the average binding energy of the O 1s states of the ZrO 2 lm as a reference, the core levels shi towards higher binding energies by 1.1 eV and 3.55 eV for the dissociated and molecular adsorbates, respectively.This agrees well with the respective experimental values of 1.7 eV and 3.7 eV.Only small changes of the core level shi of about 0.1 eV are predicted for the different adsorption sites.
Finally, we have performed selected calculations for two ZrO 2 trilayers using the same lateral cell size as for one trilayer.These results should be seen with caution, however, as the lm is strained (thicker lms should approach the bulk lattice constant, which is z3% larger), and due to the large number of local energy minima.Nevertheless, as a general trend it appears that adsorption energies increase with thickness, in agreement to DFT calculations in the literature for monoclinic bulk ZrO 2 (  111) showing adsorption energies up to z1.2 eV.The DFT energies for water on the ultrathin lm agree reasonably well with experiment, especially when considering the overestimation of H bonding with the current functional.For bulk monoclinic ZrO 2 (m-ZrO 2 ), the experimental adsorption energies at low coverage are very high (T2 eV; 21,22 our FTIR study does not give any value for the adsorption energy of the species leading to the negative peak).These values are even above DFT-calculated values for at surfaces, 65,72 probably related to defects.However, at high coverages, all experiments (including our FTIR) indicate values of z0.6 eV, while DFT predicts strong H 2 O adsorption on bulk m-ZrO 2 , with typical adsorption energies in the 0.8-1.2eV range. 65,72Our own test calculations for a few H 2 O/m-ZrO 2 congurations show similar results.To some degree, the problem may be blamed on overestimation of the strength of H bonds (as mentioned previously); this mainly affects the energies at high coverages.
The density of sites with high adsorption energy ("defect sites") depends strongly on the type of samples used.Astonishingly, it is rather high for 7 ML ZrO 2 /Pt(111) lms, 31 when compared with the z15% TPD tail in our ultrathin lms.For our powder samples, it may come as a surprise that almost no strongly-bound H 2 O is detected.To some degree, this may be an experimental artifact: as the reference spectrum was acquired aer high-temperature annealing; very strongly bound hydroxy groups or H 2 O readsorbed while cooling in vacuum would be included in the reference spectrum and go undetected.The density of these species should be small, however, as indicated by the small negative peak occurring when the preadsorbed H 2 O joins the H-bonded adsorbates.The low density of defect sites in our powder samples may be due to the extensive hightemperature annealing (1 h at 1273 K in pure O 2 ), which is also responsible for the low specic surface area.
So what are these "defects" with high adsorption energy?We start this discussion by analyzing the DFT results for the ultrathin lm.DFT indicates that this lm is strongly distorted.Considering the sharp TPD peak (Fig. 1) and the large variation of the DFT adsorption energies (Fig. 7), we believe that DFT overestimates the heterogeneity of the surface (and, related, the distortions).Nevertheless, DFT tells us that the lm structure is rather unstable, prone to distortions, and breaking of Zr-O bonds that lead to twofold-oxygen sites.We consider it likely that this kind of instability is responsible for the changes induced by multiple adsorption/desorption cycles, which create additional defect sites.DFT reveals a complex potential-energy landscape of the uncovered ultrathin oxide, but also the variety of bulk structures of similar energy 73 can be seen as the root of this instability.The STM results indicate stronger water bonding at one or two of the 12 Zr sites per ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 unit cell, in agreement with the z12% defect sites.At room temperature, water is stable at such sites only above subsurface dislocations of the substrate, again showing that a minor disturbance (slight distortions of the substrate) signicantly inuences the ultrathin oxide.The calculated E ads ¼ 1.05 eV for the most stable site in the supercell indicates that dissociated water should be barely stable for extended times at room temperature, which nicely ts these STM results, so these defect sites are probably similar to Fig. 7(c).
DFT suggests that twofold oxygen is the site of H 2 O dissociation and provides an anchor 65,72 for further H 2 O by H bonding.Given that the most stable surfaces 74 of monoclinic ZrO 2 exhibit at least 1/4 twofold oxygen atoms in the surface, and our FTIR study of well-annealed m-ZrO 2 suggests a low defect concentration, it is unlikely that a twofold oxygen is sufficient to qualify as a defect site.Our DFT results (Fig. 7(a)) show substantially different adsorption energies adjacent to twofold-oxygen sites, suggesting that the details of the geometry at a twofold oxygen and the Zr atom where H 2 O adsorbs (and dissociates) strongly inuences the bonding strength.This implies that not all twofold oxygens provide strong bonding to H 2 O. On the other hand, DFT predicts much stronger H 2 O adsorption on m-ZrO 2 than found in many experiments (including ours), which would indicate that the twofold oxygen atoms on m-ZrO 2 are not present or somehow blocked in reality.
Of course, there might be also other defects.We can probably exclude the most common defect 75 of oxide surface science, oxygen vacancies: for the tetragonal ZrO 2 (101) surface (equivalent to {111}-type surfaces of cubic or monoclinic ZrO 2 ), the formation energy of oxygen vacancies at the surface is higher than in the bulk. 70Also for m-ZrO 2 ( 111), our calculations yield a high oxygen vacancy formation energy of 7.1 eV.This means that the concentration of oxygen vacancies at these surfaces will be negligible.Calculations for the ultrathin ZrO 2 trilayer suggest much lower oxygen vacancy formation energies (2.23 eV). 70Oxygen vacancies, if present, should be easily lled by H 2 O, and the desorption barrier at these sites should be high.Our TPD + XPS experiments exclude a signicant amount of such species, however.Therefore, we believe that the concentration of oxygen vacancies at the surface of the trilayer ZrO 2 lm is also negligible.Concerning minority sites, we should also consider that our ultrathin-lm model system contains 3D ZrO 2 clusters. 36The XPS signal from the clusters (Fig. 2(a)) is about 18% of that from the ultrathin lm.Our STM images indicate that the average cluster height is at least four trilayers.With these values, an XPS simulation 59 indicates that clusters account for less than 7% of the surface area.This area is not enough to have a large impact on our XPS and TPD results; e.g., a picture where all strongly bound or dissociated water is adsorbed on the 3D clusters would be inconsistent with the area fraction of the clusters.
Let us now compare H 2 O adsorption on ZrO 2 to structurally similar oxygen-terminated surfaces.The TPD peak temperature of H 2 O on the ultra-thin ZrO 2 /Pt 3 Zr(0001) at 180 K is higher than on other oxygen-terminated oxides such as ultrathin alumina (164-168 K) 76,77 or a 2 ML-thick FeO lm (H 2 O: 166 K, D 2 O: 169 K). 78 At least for the case of alumina, the top O-Al interlayer spacing (40 pm) 79 is less than in the present case, so the stronger bonding of H 2 O on ZrO 2 must be explained by the large in-plane O-O spacing (z350 pm) on ZrO 2 ; this makes the metal cations more accessible than on most other oxygen-terminated oxides, where O-O distances are around 300 pm.Strong adsorption requires that the metal cations are easily accessible; for O-terminated surfaces this means large O-O distances.Examples of structures with O-O distances similar to ZrO 2 are CeO 2 (111) 80 with the main TPD peak at 200 K and In 2 O 3 (111) with H 2 O adsorption energies of 1.2 eV (TPD peak well above room temperature). 57inally, how good a model system is the ultrathin ZrO 2 lm for "real-world" ZrO 2 surfaces?The ultrathin lm is based on a trilayer of cubic ZrO 2 (111) with additional distortions and some twofold-coordinated O atoms at the surface.Thus, geometrically it is similar to the two energetically most favorable monoclinic ZrO 2 surfaces, ( 111) and (111). 74When comparing with ZrO 2 powder, nanocrystalline material will also exhibit a substantial number of edge and corner sites (more than the step sites in our model system), but for the hightemperature annealed material used for our FTIR study we do not expect enough of these sites to play an important role.As is generally true for ultrathin oxide lms on metal substrates, the main restriction as a model system applies if adsorption involves charge transfer.In such a case, one should expect stronger bonding on the ultra-thin lm than on thick lms or bulk materials, where charge transfer is difficult or impossible. 34,81,82For adsorption of H 2 O, in molecular or dissociated form, there is no net charge transfer, so the delity of the lm as a model system is probably limited by differences in geometry between the lm and bulk structures.Unfortunately, the current DFT models do not agree well enough with experimental data for an exact comparison.Nevertheless, the good agreement of experimental H 2 O adsorption energies is encouraging.

Conclusion
We have investigated the adsorption and desorption of water on a trilayer of ZrO 2 grown by oxidation of a Pt 3 Zr(0001) single crystal.With a combined TPD, XPS, and STM approach, we could show that z88% of the water molecules in the rst monolayer adsorb molecularly, with an adsorption energy of E des ¼ 0.57 AE 0.04 eV.This agrees well with DFT, which nds dissociative adsorption in only a few sites, with the proton accepted by twofold coordinated oxygen.Experiments also showed that the adsorption energies of dissociated water vary over a wide range, depending on local details of the structure (e.g.substrate dislocations below) and pretreatment (increasing number of "defect" sites aer multiple adsorption-desorption cycles).FT-IR measurements on ZrO 2 powder showed similar adsorption energies as our UHV TPD study, indicating that the ultrathin (trilayer) ZrO 2 lm is a valid model system for water adsorption on well-annealed real-world ZrO 2 surfaces.
3.1.2X-ray photoelectron spectroscopy.To determine whether (or to what extent) water adsorbs in molecular or dissociated form on the ZrO 2 /Pt 3 Zr surface, we have analyzed XPS O 1s data.When dosing 2 ML D 2 O at 100 K, four different peaks are distinguished as originating from D 2 O, OD, 3D ZrO 2

Fig. 1
Fig. 1 TPD of D 2 O on ZrO 2 /Pt 3 Zr (heating rate 1 K s À1 ; no fresh sample preparation between spectra with different coverage).(a) The TPD spectra consist of three regions: the multilayer peak (150 K), the monolayer peak (180 K) and a high-temperature tail up to z540 K, see inset.(b) Logarithmic plot of the spectra in (a) vs. 1/T.The linear onset of the multilayer peak gives E des ¼ 0.47 eV.(c) Inversion analysis showing the desorption barrier E des vs. coverage using n ¼ 10 15AE1 s À1 .In the 0.5-0.95ML range, E des is 0.57 AE 0.04 eV.(d) Amount of desorbed D 2 O (calculated from the integrated TPD signal) as a function of gas dose.The TPD intensities were calibrated assuming a sticking coefficient of one, thus 1 Langmuir (L) ¼ 10 À6 torr s corresponds to 0.48 ML.The red line is a linear fit.

Fig. 2
Fig. 2 XPS of D 2 O on ZrO 2 /Pt 3 Zr.(a) O 1s region with 2 ML D 2 O deposited at 100 K, then annealed to 160 K to remove the multilayer.(b) Comparison of spectra before and after D 2 O adsorption, and after additional annealing steps.All XPS spectra were taken at 100 K. (c) Peak areas of the spectra in (b) for the peaks assigned to D 2 O, OD, 3D ZrO 2 clusters, and the ZrO 2 trilayer.(d) TPD measurement (1.54 ML) for comparison with the XPS results in (b, c).

Fig. 3
Fig. 3 STM of small H 2 O coverages on ZrO 2 /Pt 3 Zr.(a) 0.12 AE 0.04 ML H 2 O, dosed at 100 K, measured at 78 K.The surface consists of stable regions with water species arranged in the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 structure (yellow dashes), regions with higher coverages and unstable imaging (fuzzy gray patches), and apparently uncovered regions of the ZrO 2 trilayer.(b) After dosing 15 L of H 2 O at RT. Water-related species, namely O w H, are found preferentially above subsurface dislocation lines (grey areas in the image).They locally show the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 order (yellow).(c) The surface after dosing 30 L of H 2 O at RT and storage in UHV for 8 hours.The image was obtained with a special tip that shows OH as depressions (black).Apart from the few hydroxy groups, the clean oxide surface can be seen with atomic resolution.The red lines in the zoomin (d) indicate that the OH are on top of Zr atoms (protrusions in the image).
of the ZrO 2 trilayer and the dislocation lines appear unchanged.A small number of water-induced features with an apparent height of 120-150 pm survives the 550 K annealing; their number increases with more adsorption/desorption cycles, and amounts to approx.0.01 ML per adsorption cycle, tting the results from the decrease of the ML peak mentioned above.Similar to the OH formed at room-temperature adsorption, these features preferentially bind to sites above the subsurface dislocation lines.Aer 5 cycles, see Fig.5(b), some ordering of these water-induced features can be made out: in the Fourier transform of the positions of these species (inset of Fig.5(b)), the strong outer spots (marked by red circles) indicate equivalent positions with respect to the 0.35 nm ZrO 2 lattice.There are also weaker spots corresponding to the ð ZrO 2 trilayer on the oxide (blue circles),

Fig. 4
Fig. 4 Water-induced modification of the ZrO 2 /Pt 3 Zr trilayer.(a) Repeated D 2 O TPD runs (heating rate 2 K s À1 ) with a starting coverage of 2 ML each.The monolayer peak decreases while the high-temperature tail increases.(b) CO 2 TPD (initial dose z 1 ML, heating rate 1 K s À1 ) before (blue curve) and after (red) 18 D 2 O TPD cycles.The increasing intensity in the 110-125 K range indicates an attractive interaction of CO 2 with the waterinduced defect sites.Annealing to 1205 K restores the original state of the surface (orange, dashed).

Fig. 5
Fig. 5 Effect of H 2 O adsorption/desorption cycles on the ZrO 2 /Pt 3 Zr trilayer seen by STM (T STM ¼ 78 K).(a) After one adsorption/desorption cycle.The water-induced bright species form preferentially on top of dislocation lines (gray bands) as well as at a rotational domain boundary [between the red arrows in the margins; the orientation of the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 unit cell on both sides of the DB is indicated by yellow lines].(b) After five adsorption/desorption cycles, the coverage of the bright spots (O w H) has increased by almost a factor of five.An FFT of the bright features (inset) indicates equivalent positions on the 0.35 nm ZrO 2 lattice and some influence by the ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 superstructure.

Fig. 6
Fig. 6 Infrared absorption of H 2 O on sintered monoclinic ZrO 2 powder.(a) Spectra at 298 K for selected partial pressures of H 2 O. Spectra are shifted vertically for clarity, and a few regions of dense gasphase absorption peaks are indicated.(b) Differential adsorption energies from the isobaric and isothermal measurements.The vertical lines span the results for the two extreme cases of zero (top) and maximum (ice-like, bottom) adsorbate entropy.Coverage values are normalized with respect to the value at p H 2 O ¼ 25 mbar, which corresponds to 2-3 H 2 O per surface Zr atom.

Fig. 7
Fig. 7 H 2 O adsorption on a trilayer ZrO 2 film in DFT.(a) Top view of the fully-relaxed ð ffiffiffiffiffi 19 p Â ffiffiffiffiffi 19 p ÞR23:4 model cell including adsorption energies (in eV) at water adsorption sites.For selected sites, energies of dissociative adsorption are given after the slash.Values in parentheses indicate that the ZrO 2 film had moved and changed upon adsorption, resulting in a site of different character.The yellow rhombus indicates the unit cell, and the dashed triangle shows the group of six Zr atoms higher than the others.Twofold-coordinated oxygen atoms are marked by yellow asterisks.Side view of (b) molecularly bound and (c) dissociated water at the site marked "0.98/1.05",with bond lengths in pm.These two configurations are indistinguishable in the STM simulations (d, e).
65,72ms on Pt(111)31is an exception, probably due to the simplistic assumption of the prefactor being 10 13 s À1 : the TPD peak in ref. 31 is 20 K above the multilayer peak, comparable to ours.Strongly bound water seems to be mostly dissociated; the weakly bound H 2 O is in molecular form at least at low temperatures on our ultrathin lms.Both our DFT results for the ultrathin lms and DFT calculations for H 2 O on monoclinic ZrO 2 from the literature65,72predict dissociative adsorption at (some or all) twofold-oxygen sites, otherwise mostly molecular adsorption, in many cases forming H-bonded structures already at low coverage.This also nicely ts our XPS data as well as FTIR, where the absence of sharp peaks in FTIR spectroscopy (except for the negative peak from the initial coverage) can be explained only by ensembles of H-bonded species on the monoclinic powder sample.With the H 2 O molecules on top of the Zr atoms, and in-plane Zr-Zr