Hydroxyl Identification on ZnO by Infrared Spectroscopies: Theory and Experiments

Here we present a thorough density functional study combining experiments on ZnO nanostructures aimed at the identification, by means of Infrared (IR) spectroscopies, of hydroxyl and hydride species formed on most stable low-index Miller surfaces of würtzite ZnO; namely, Znand O-terminated (0001) and (0001) polar surfaces, and nonpolar (1010) and (1120) surfaces. The Perdew-Burke-Ernzerhof functional was employed on the periodic slab calculations, and all possible H and OH adsorption modes were studied at medium and full coverage, while IR spectra were simulated for most favourable situations. This information was used to model the most likely surface arrangements when exposed to either H2 or H2O. IR experiments on ZnO surfaces and nanoparticles are discussed based on the calculated adsorption energy values and simulated IR spectra. The study emphasizes the detailed assignment of OH moieties with the help of IR and their interpretation as fingerprints of surface morphology, allowing for a consistent interpretation of water adlayers stability and their corresponding vibrational fingerprints as a function of coverage, low-index Miller surface, and hydrogen source.


Introduction
Würtzite Zinc Oxide (or zinczite) is a material that has driven very much research interest in the last decades, mostly due to its optical and electronic properties. It displays high electron mobility and thermal conductivity, and a direct bandgap (~3.4 eV) with large exciton binding energies (~60 meV). 1,2 These properties have spurred its usage in semiconductors, 3 field-effect transistors, 4 photodetectors, 5 blue-and ultraviolet-light (UV-light) emitting and laser diodes, 6,7 gas sensors, 8 piezoelectric generators, 9 transparent electrodes, 10 and cells for solar light harvesting. 11,12 In chemistry, ZnO has recently become the focus of many researches addressing its usage as a catalyst for a variety of reactions; from being an active phase in methanol synthesis using the ternary Copper-based Cu/ZnO/Al 2 O 3 catalyts, [13][14][15] to water 16 and sulphur hydride 17 dissociations, desulfurization processes, 18,19 the water gas shift reaction, 20 the activation of CO 2 , 21 and finally, the conversion of maleic anhydride into 1,4-butanediol. 22 Furthermore, the above-commented UV-light absorption capability has unfolded its use in light-triggered catalysis, such as for dye decomposition, 23,24 the treatment of volatile organic compounds, 25 the peroxide synthesis, 26 water splitting, 27 and alcohol photodegradation. 28,29 Nevertheless, these catalytic or photocatalytic processes are far from being well understood and efficiently mastered, and the aspects on which the catalytic activity and selectivity hang are still a matter of debate. Many experimental works have addressed the point by relating the catalytic activity with structural factors, this is sampling a variety of ZnO morphologies -including single crystal surfaces, 30 thin films, 24 nanostructures, 31,32 and wellfaceted nanoparticles 23,33 -with distinct polar/nonpolar facet ratios. This proportion is often argued to be a key aspect on the photochemical catalytic activity: Some working groups assign to polar surfaces -the Zn-and O-terminated (0001) and (0001) surfaces-the highest catalytic activity, 23,31 whereas other point to nonpolar surfaces -the (1010) and (1120) surfaces. 30 Along this line there is also an open discussion on the role of surface hydroxyl and hydride species on most of the aforementioned catalysed reactions. This is because in many of them one or more reaction steps involve hydrogenation/dehydrogenation of reaction species, likely to be carried out by these surface moieties. 34 For instance, previous works pinpoint that the hydroxyl scavenger character of isopropanol is decisive in its degradation. 35,36 In another related study, the pivotal role of acid/basic sites on the catalytic activity and specificity is highlighted. 37 Indeed, in many of the commented processes either heterolytic dissociation of H 2 or water splitting is actually considered the rate-determining step.
Clearly, the unambiguous identification of the reacting hydroxyl or hydride species under working conditions is vital for the real-time observation of the catalytic process and the outline of the reaction mechanism, eventually allowing for a posterior improvement of the reaction setup. This is usually experimentally tackled by a combination of spectroscopic and microscopy techniques. Vibrational spectroscopies can be highlighted as prototypical techniques among them due to their (experimental and theoretical) simplicity and potential use in all conditions (e.g. ex and/or in situ conditions), providing additionally information of the conformation and bond strengths of surface species with direct relation to surface/catalytic phenomena. 38,39 However the assignment of the vibrational fingerprints is by no means easy, and often assumptions based on common knowledge -e.g. vibrational fingerprints of transition metal complexes-drive to misassignments and, therefore, to erroneous interpretations. 40 40,42,43 In the case of würtzite ZnO there is quite a number of recent experimental studies aimed at this issue. It is worth to highlight the paramount works of Wang et al. using High-Resolution Electron Energy Loss Spectroscopy (HREELS) on (1010) and (0001) single crystal surfaces, [44][45][46] 34,47 In both research lines assignments of different features to specific surface hydroxyl or hydride groups are posed, yet some surfaces remain unexplored, and even the studied assignment seems to contradict earlier and recent experiments. 34,[48][49][50] From the theoretical point of view the matter is far more shaky. Despite there is also a number of DF studies addressing the structure and stabilization mechanisms of the bespoken surfaces saturated with either water or hydrogen, 51-55 these computational studies did not address their vibrational identification. To the best of our knowledge, there is no previous DF theoretical study consistently addressing the assignment of vibrational spectroscopy features to surface hydride or hydroxyl moieties when exposing ZnO systems to either H 2 or H 2 O.
In the present work we study, by means of state-of-the-art DF calculations on proper slab models, the stability of polar -(0001) and (0001)-and nonpolar -(1010) and (1120)-surfaces of würtzite ZnO exposed to H 2 or H 2 O. Simulated IR vibrational spectra are gained for the most stable surface adsorbate arrangements as a function of coverage and hydrogen source. A full analysis of hydroxyl-related species at surface model systems as well as on ZnO nanoparticles is presented. the atomic cores, 57 allowing for obtaining converged results -variations in energy below 0.01 kJ mol -1 -with a cut-off kinetic energy of 415 eV for the plane-wave basis set.
Geometry optimizations were performed using a conjugated gradient algorithm and applying a tetrahedron smearing method with Blöchl corrections with a 0.2 eV width, although final energy values were corrected to 0 K (no smearing). The structural optimization was finalized when forces acting on atoms were below 0.01 kJ mol -1 pm -1 . Unless stated otherwise, all calculations were carried out in a non spin-polarization fashion. All DF calculations have been carried out using the Perdew-Burke-Erzenhof (PBE) exchange correlation functional, 58 proven to deliver a realistic description of bulk ZnO and low-index Miller surfaces, 59,60 as well as to properly capture the interactions of atomic hydrogen, hydroxyl moieties, and water molecules upon. 16,61 From previous X-Ray Diffraction (XRD) studies 11,23,25,29,32 it is clear that würtzite ZnO preferentially displays the nonpolar (1010) and (1120) surfaces, and the Zn-and O-terminated (0001) and (0001) polar surfaces. The latter ones are simultaneously created when cutting the crystal along a basal plane, see Figure 1. Nonpolar surfaces exhibit perfectly stable unreconstructed terminations. Polar surfaces feature a surface energy instability issue originated from their different charged terminations, which exert a net dipole moment that, de facto, increases with the separation between them. This introduces an electrostatic component to the surface energy which diverges with the surface separation. 62,63 The stabilization mechanisms to nullify this dipole moment are nowadays a hot topic, with some experiments suggesting the existence of unreconstructed polar surfaces, 64 implying a charge transfer among the two polar surfaces, 59,64,65 whereas other studies suggesting the formation of vacancies regularly found on the surface, 66,67 or concentrated at step edges. 51,53 Regardless of the previous, when fully hydrogenated/hydrated both polar surfaces show a perfect (1×1) arrangement with apparently no surface vacancies. 67,68 Moreover, unreconstructed polar surfaces can be perfectly stable in oxide nanoparticles. Thus, we decided to use unreconstructed and vacant-free surface supercell slab models to simulate ZnO single crystals and the facets of nanoparticles. Last but not least, note that when fully hydrogenated/hydrated, nonpolar surfaces also display a (1×1) periodicity -with the caveat of H 2 O on (1010) which, according to previous simulations and experiments, may display a (2×1) pattern. 16,69 Towards this end, 8-layer slab -constructed from the orthogonal unit cell instead of the primitive hexagonal unit cell, see Figure 1-(1×1) supercell models have been used for each of the four surfaces, in which the upmost four layers are completely allowed to relax, and the bottommost four layers are kept fixed as in the bulk-optimized positions. Note that cell contains a single surface ZnO pair for the (1010) surface, but two units for polar and (1120) surfaces. To correctly study the (1010) surface a (2×1) unit cell has been used when needed. Present tests and several past studies reveal that the employed slab width and relaxation approximation is enough to ensure converged results of surface properties, 52,59-61,64,65,70-72 except for cleavage energies of polar surfaces, which should be derived with an extrapolation to infinite slab thickness. 52,59,60,64 A minimum vacuum of 1 nm was applied in the surface direction to avoid interaction between repeated slabs. Optimized slabs with larger vacuum gap showed deviations in the total energy below 0.02 kJ mol -1 . To compensate the long-range dipole-dipole interactions among translationally repeated slabs a counterdipole is placed in the middle of the vacuum gap. An optimal Monkhorst-Pack k-points grid of 17×17×1 has been used for the ZnO surface calculations. This guarantees a convergence of the energy to values below 0.01 kJ mol -1 as tested using denser grids. The adsorption energy, !"# , of a species A -H 2 O, OH, or H-on a substrate Beither the polar (0001) or (0001) surfaces, or the nonpolar (1010) or (1120) surfaces-is defined according to the following equation; where EA/B is the energy of the complete system where A is adsorbed on the surface B, and EA Harmonic frequencies were obtained through numerical calculation and diagonalization of the Hessian matrix taking into consideration only the adsorbate molecules and, in the case of atomic H, the surface species it is bonded to. The Hessian matrix is constructed from finite differences of analytical gradients by calculating energy changes due to independent displacements of 3 pm of every atom in each direction of the unit cell vectors.
Shorter displacements lead to frequency shifts below 4 cm -1 , i.e. below standard experimental resolution. 47 Negligible coupling with substrate phonons is expected since the latter exhibit distinctive lower frequencies, typically ranging 100-600 cm -1 , according to Raman spectroscopies. [75][76][77] Test calculations of the adsorbate harmonic frequencies accounting for variables of influence, such as phonon coupling, different k-points grids, and the absence of a counterdipole resulted in frequency variations below 2 cm -1 .
Simulated IR spectra have been obtained estimating the intensity of a band through the change of the dipole moment component normal to surface accompanying a given vibration.
The spectra have been drawn by smoothing the peaks with a Gaussian function of 100 cm -1 half width. This procedure has been widely used in the past for calculations of adsorbate spectra on solid metal surfaces. 40

Experimental Details
Materials were prepared using a microemulsion method using n-heptane (Scharlau) as organic media, Triton X-100 (C 14  Analysis of the spectra was carried out considering the derivative spectra to extract the number of components. Fitting was performed with the peakfit program using Gaussian shapes for the components.

H Adsorption
First we examined the adequacy of the computing level by analyzing ZnO bulk and surface properties (see Supporting Information). In summary, a very good agreement is found between present PBE results and most recent experimental bulk structural parameters, 83 with a slight overestimation of the bond strength per ZnO unit of ~20 kJ mol -1 compared to the experimental value, 84 in the order of standard DF methods accuracy. The relaxed surface structure of (1010) is in excellent agreement with very recent aberration-corrected HRTEM images, 85 and earlier experiments. [86][87][88] Additionally, the relaxation of polar (0001) surface is also in excellent agreement with structural data derived from Surface XRD (SXRD) and Grazing Incidence X-Ray Diffraction (GIXD) experiments. 64,89 Thus, we conclude that the employed methodology appears to be suited for the description of the systems under scope.
Concerning the adsorption of species, we tackled firstly the simplest atomic H adsorption at two different coverages (θ), this is, at half and full coverage, hereafter also referred as 0.5 and 1 monolayers (ML). All possible adsorption sites on the modeled surfaces have been sampled, yet only the most stable case on each surface is further discussed. For  Table 1, while views of the adsorbed conformations together with their simulated IR spectra are illustrated in Figure 2. Note that when taking H 2 molecule in vacuum as the energy reference the adsorption energies shown in Table 1  Photoelectron Spectroscopy (XPS) measurements. 55 Note that the desorption energy of a H 2 molecule at θ = 0.5 ML, 388 kJ mol -1 -calculated as twice the atomic H desorption energy reported in Table 1 but accounting for half H 2 dissociation energy, as above-commented-, is far larger than that for H 2 at θ = 1 ML (152 kJ mol -1 ). When we compare the adsorption energy per unit cell -proportional as if done per surface area-, the desorption energy at half coverage is 194 kJ mol -1 -note that at this coverage two unit cells with an adsorbed H atom are needed to desorb a H 2 molecule-reflecting that more energy is released -about 40 kJ mol -1 -when covering half of the (0001) surface with atomic H instead of a full coverage, and so, such a (1×2) arrangement is thermodynamically clearly preferred.
The 1 ML situation is still exothermic, and so, a (1×1) periodicity is actually feasible.
Indeed, previous experiments combining LEED and He-Atom Scattering (HAS) experiments detected such a (1×1) arrangement. 67 In this latter study the activation energy for H 2 -desorption was found to be 141 kJ mol -1 , in perfect line with the present estimate for the desorption energy of 152 kJ mol -1 at θ = 1 ML, corroborating the possible existence of a (1×1)

arrangement. This is also back-supported by previous XPS simulations and calculations
showing that core level O 1s signals assigned to surface hydroxyl groups can only be explained when having full hydrogen coverage. 67,90 When addressing the arrangement issue from the point of view of IR techniques one encounters a few discrepancies in the literature. Previous IR studies on the H(1×1)-ZnO(0001) pattern formed after exposing the ZnO single crystal surface to water vapor at room temperature found a peak at 3572 cm -1 , assigned to the hydroxyl stretching vibration, 91 whereas a more recent HREELS study assigned a peak at 3621 cm -1 to the aforementioned hydroxyl vibration after exposing 2 Langmuir (L) of H 2 O to the single crystal surface, also at room temperature. 44 Indeed, a simpler explanation for this discrepancy is that values were obtained at moderate 91 and low 44 water partial pressure. Previous theoretical studies show that in H-rich conditions (high pressure) and high temperature (800 K) coverage above 0.5 ML and even the full coverage situation approach in stability to the (1×2) arrangement. 52 However at such elevated temperature values surface H is known to recombine and desorb as H 2 , as found already to happen at 547 K for a (1×1) situation. 67 Nevertheless the full coverage phase at H-rich conditions could be plausible at lower room temperature.
Present IR simulations on Figure 2 show that at 1 ML coverage only the OH symmetric stretching vibration is active, centered at 3486 cm -1 , sensibly below the peak found when exposing the single crystal surface to water. An important point here is that there is a very large difference between this IR fingerprint and that of coverage 0.5 ML. When decreasing the coverage to the (1×2) pattern, the hydroxyl structure is only very weakly modified, with OH bond length decreased by only 2 pm, and the hydroxyls slightly less normal to surface. However, the bonding energy is increased by more than 100 kJ mol -1 , which has a critical impact on the hydroxyl stretching frequency, which blue-shifts more than the 3621 cm -1 peak observed at low pressures, 44 where the formation of a (1×2) pattern seems to be clear. The additional experimental peak 91 at 3572 cm -1 appears at a somewhat larger value than the one calculated at saturation coverage, and so probably belongs to an intermediate situation, e.g. 2 / 3 or 3 / 4 ML, as previously found to be also energetically competitive situations. 52 (0001) surface: Previous HAS studies showed that when this surface is exposed to hydrogen, a weakly bound (1×1) hydride pattern is formed. Reports combining Scanning Tunneling Microscopy (STM) images and DF calculations showed that such a single crystal surface displays triangular pits, 51,53 and recent theoretical surface phase diagrams assign the (1×1) pattern to a situation with full hydride coverage on triangular pits and islands. 92 Current computed adsorption sites and energies are in perfect agreement with those of Kresse and coworkers, 53 showing an essentially isothermic adsorption at θ = 0.5 ML and a disfavored adsorption process at 1 ML, also in line with a previous experimental report showing a H 2 desorption energy of a few kJ mol -1 . Probably the instability at full coverage is at the origin of the observed reconstruction and formation of a rough surface. 93 In any case, the predicted arrangements are similar to those of (0001) surface, but with a sensibly larger ZnH bond length of 155-160 pm, in very good agreement with recent DF calculations at Local Density Approximation (LDA) level. 94 The IR simulated spectra signals are attenuated by ~30 times compared to the OH stretching vibrations found on (0001) surface, here the peaks situated around 1833 and 1499 cm -1 at half and full coverage, respectively, allowing also for a clear distinction of the degree of coverage based on IR measurements,.
(1010) surface: The interaction of hydrogen with the nonpolar (1010) surface has driven lately very much attention due to the so-called surface metallization effect appearing when exposing this surface to hydrogen at room temperature. 45 95 and it has also been found not to induce a metallization of the (1010) surface. 45,98 According to present calculations, the adsorption of H to form surface hydroxyls would be, similarly to (0001) surface, an isoenergetic process when taking H 2 in vacuum as the energy reference. The formation of sole ZnH hydride species was found to be thermodynamically prohibitive. Concerning the IR simulated spectra in Figure 2, the surface hydroxyl stretching is active and observable, although the intensity is clearly inferior to those spectra of (0001) polar surface. The attenuation of the signal is more acute in the θ = 0.5 ML case, due in principle to the tightening of the O s -H bond, as found also for the polar surfaces.
During the geometry optimization the system undergoes a peculiar relaxation, which involves an inversion of the ZnO angle -7.1º with respect the surface plane, see Supplementary Information-having Zn as the most exposed surface atom, similarly to earlier results obtained at B3LYP hybrid functional level. 98 We found an elongation of 29.5 pm of the surface ZnO bond length, and a small contraction of ZnH bond of 2 pm (Table 1). The adsorption energy accounting for H 2 gas dissociation results in an exothermic process of 43 kJ mol -1 , at variance with previous theoretical studies using pair-potentials, which erroneously found such dissociated system to be endothermic by around 500 kJ mol -1 , a feature arising from the limitations of such empirical approach. 99 When having both species, the overall adsorption strength is enhanced, due to the bulk near-tetrahedral coordination geometry recovery. 100 The simulated IR spectra shown in Figure 3 displays two peaks at 1797 and 3601 having an angle of 6.8º but exposing surface Zn atoms (Table 1 and Supplementary   Information). The process is estimated to be exothermic by 47 kJ mol -1 . Figure 3  Note that when studying ZnO nanoparticles one has to consider the ratio between polar/nonpolar surfaces. Previous HRTEM and FTIR experiments have shown that, typically, on ZnO nanopowder particles, 80% of the surface corresponds to nonpolar faces. 95,104 In this sense, the relatively low IR intensity of nonpolar hydrogenated surfaces is partially counterbalanced by their highest exposure, and so, in principle observable. This is in perfect agreement with the FTIR signal at 3672 cm -1 observed when exposing ZnO nanoparticles to atomic hydrogen in UHV conditions. 34 In these experiments, the rapid appearance of a dominant peak at 3618 cm -1 corresponds to the strong hydrogen adsorption on the (0001) facets forming a (1×2) pattern of hydroxyls, and only after more exposure the nonpolar surfaces are hydrogenated, although (1010) and (1120) become indistinguishable. According to present results displayed in Figure 3, a distinction of both surfaces could only be reached by scanning the 1650-1850 cm -1 region with a high-resolution IR technique.

OH Adsorption
Here  Table 2.
(0001) surface: Here only one stable site is found, having the hydroxyl adsorbed on a 3-fold hollow site -connecting three surface Zn species and therefore, a µ 3 adsorption mode-with no ZnO oxygen directly beneath, see Figure 4. The µ 3 mode with an underneath oxygen is destabilized by more than 50 kJ mol -1 , and therefore has not been further considered. In the µ 3 conformation the adsorption strength is rather high, almost 400 kJ mol -1 at θ = 0.5 ML, and the hydroxyl stays essentially perpendicular to the (0001) surface. By reaching full coverage the lateral repulsion reduces the adsorption energy by ~120 kJ mol -1 , with very small structural changes and only a small increase of the molecular perpendicularity towards the surface. The vibrational frequency at half coverage of 3755 cm -1 , gets red shifted by ~11 cm -1 when reaching a full coverage, providing a noticeable IR signal as seen in Figure  4. It is worth to mention that in the earlier study of Wander and Harrison on water adsorption on a ZnO(0001)/(0001) slab 72 the hydroxyls were put directly on top of surface Zn atoms in the (0001) surface, although according to present results such mode is found to be unstable.
Actually (1010) and (1120) nonpolar surfaces: Curiously, for both surfaces, two competitive adsorption minima are found. For the (1010) surface these are named µ 1 and µ 2 conformations. In the latter, the OH Oxygen atom attaches bridging two adjacent Zn atoms, with its H atom pointing towards the vacuum, see Figure 4. This particular coordination mode has been recently found to be a metastable fragment in water dissociation, according to B3LYP and PBE DF calculations. 61  given the shorter distances 191 pm in between surface hydroxyls. As seen in Figure 4, hydroxyls align themselves along the (1120) surface grooves, staying also highly planar to surface. Therefore, its IR peak located at 3366 cm -1 displays low intensity, and would be observed as a broad peak in experiments. As seen below, this particular arrangement is a clear precursor of the fully hydrated (1120) surface. The other mode, named η 1 , simply corresponds to a hydroxyl adsorption on a surface Zn atom, with H pointing towards the vacuum. This conformation is only 2 kJ mol -1 less stable and so considered isoenergetic to the η 2 mode.
This mode is IR-active, with a stretching frequency at 3733 cm -1 at half coverage. At full coverage, the hydroxyls align, maybe making H-bonds (267 pm), with a characteristic peak situated at 3660 cm -1 .

Full Analysis of H/OH/H 2 O Species on Polar and Nonpolar Surfaces
The information of the preferential adsorption of -OH and -H moieties has been used to sample different arrangements of split water molecules on the different surfaces. Let us consider first of all ZnO slabs simultaneously exhibiting (0001) and (0001) surfaces. ML is shown in Figure 5 for both polar surfaces. The small difference of 11 cm -1 between frequencies of hydroxyls of different polar surfaces (Tables 1 and 2) 16 The stability of this peculiar arrangement was compared to the non-dissociated water monolayer, and also to the completely dissociated case. 16,54,61 Notice that H 2 O molecules can be adsorbed on the ZnO dimer stripes, above the surface grooves, or alternatively sampling stripe and groove positions. In addition, the distinction between stripe and groove positions affects the number of possibilities for the mixed dissociated case. Furthermore, as shown in Table 2, the hydroxyl moieties can display µ 1 and µ 2 conformations. All possible moiety combinations meeting the (2×1) surface pattern have been optimized, the most stable one being precisely that previously proposed. 16,54,61 Within this arrangement every second H 2 O molecule is adsorbed over the surface groove, and the other half becomes dissociated into hydroxyl and hydrogen atoms, see The same procedure was carried out for the (1120), accounting for both η 1 and η 2 adsorption modes. Here, in accordance to previous simulations at LDA level, 54 Table 3, and their vibrational fingerprints are shown in Figure   5. On both nonpolar surfaces a distinction is made for hydroxyls arising from water splitting (-OH), and those formed by the adsorption of a H adatom on a surface Oxygen atom (O s -H).
Considering the (1010) surface, the simulated IR spectra are governed by two peaks, located at 3155 and 3765 cm -1 . The experimental HREELS by Wang et al. showed that after exposing the (1010) surface to water, two main features were detected; a broad band at 3195 cm -1 , assigned, in principle, to a water -OH implied in a H-bond, and another at 3670 cm -1 assigned to a partially dissociated H 2 O molecule. 46 However, in light of the present results, this assumption is proven to be only half true. Whereas the peak at 3670 cm -1 would actually belong to an O-H stretching from a partially dissociated water molecule -the underestimation of ~100 cm -1 in our calculations is mainly due to the lack of description of anharmonicity, as shown below-the peak at 3195 cm -1 is, according to present model calculations, due to O s -H hydroxyl stretching. In the HREELS experiments a shoulder of the 3670 cm -1 signal -specified as a peak located at 3700 cm -1 -was assigned to a non-Hbonded OH from the intact water molecules. However, as observed in Figure 5, these modes are actually located below 3000 cm -1 .
Note that water molecule lays rather planar to the surface, and consequently its IRactive vibration modes are rather weak in intensity. Indeed, its stretching modes are found to be slightly coupled with the O s -H mode, and it is because of this that they have an appreciable intensity. All these modes are responsible of the broad band centered at 3195 cm -1 as found by Wang et al., 46 which indeed could be composed of many superimposed contributions. Last but not least, according to present calculations, there is no explanation for the peak at 3700 cm -1 , suggesting that such a peak might be apparent, and in reality not belonging to any particular vibrational mode. This statement is supported by FTIR UHV experiments of Noei and coworkers 47 on ZnO nanoparticles, who did not detect such a signal, despite (1010) surface, due to its stability, is the dominant surface on many ZnO nanoparticles. Although not shown on Figure 5, the (1010) surface displays a water scissor mode calculate to be at 1651 cm -1 displaying a low intensity, in accordance to FTIR experiments by Noei et al who detected such mode at 1617 cm -1 . 47 As far as the (1120) nonpolar surface is considered, there is, regretfully, no IR experiments -to the best of our knowledge-carried out on single crystal surfaces. (0001) hydroxyl stretching modes, 47 in accordance to simulated IR shown in Figure 5 -the experimental peak difference of 50 cm -1 is somewhat underestimated by 30 cm -1 according to present calculations. Curiously, UHV-FTIR experiments revealed an emerging peak at ~3440 cm -1 , which, according to present calculations, would well belong to the OH stretching modes on the nonpolar (1120) surface. The exact assignment of the contribution located around 3400-3450 cm -1 is however complex, and is further analyzed in the following section.

Surface Species at ZnO Nanoparticles Interacting with H 2 O
To analyse the significance of our IR study for water-related species at ZnO nanostructures, we synthesized two materials: The next two regions contain contributions for isolated OH species [47][48][49][50]107 and have variable importance in the two samples. Such issue becomes visually evident by the different colourcode used in Figure 7 for the analysed ZnO surfaces. The fitting here illustrated using 8 contributions is suggested by the derivative of the spectra (shoulders during the slow decay are particularly evident in sample B) as well as theoretical predictions previously discussed.
Furthermore, larger differences between the IR spectra of both samples are noteworthy for the higher dehydration temperature, as the masking contribution from adsorbed water is significantly decreased, highlighting alongside the predominance of the peak at ~3400 cm -1 on sample B. Last but not least, the higher dehydration temperature unveils a peak located at ~3670 cm -1 , which could belong to either (1010) or (0001) hydroxyl stretching modes, in accordance to previous DRIFTS and FTIR experiments. 47 Note that bands in Figure  Information gives details about the statistical significance of using 8 contributions, as well as additional physical insights. More importantly, rather than the existence and/or definitive assignment of each band presents in an experimental spectrum, the dominance of polar or nonpolar surfaces at one specific frequency region (above or below the two cut-off points mentioned) would be the key factor in interpreting the IR-derived morphology information at nanoparticles.
The detailed interpretation of the IR active species experimentally detected is therefore presented in Figure 8. Such interpretation comes from analysing the relative contribution of the IR bands observed in Figure 7 and confronting such result with the theoretical expectation derived from Figure 5. Note that this latter figure displays the intensity as well as the frequency of the corresponding surface moieties, allowing a meaningful description of the vibrational features arising from surface species. Most important differences between our theoretical and experimental results would be related to a frequency shift between them; experimental results presenting always lower frequency values than theoretical ones, as expected due to anharmonic contributions not considered in the calculations and to an  bands between 3100 and 3370 cm -1 ). In agreement with this, we can see that the brick-type A sample presents a significantly larger intensity contribution than needle-type B sample.
The zone around 3370-3450 cm -1 displays an important single contribution. As it presents more significant intensity in sample B -mainly exposing polar (0001)/(0001) facets-than sample A it could be expected to be related to polar surfaces. We may suggest that it would correspond to O s -H species on (0001) polar surfaces. This correlates well with the features presented in Figure 2 and Table 1 considering that the exact peak position for such adspecies depends critically on coverage (as it was discussed above, contributions located between the two frequencies presented in Figure 2 are possible in experimental studies in dependence on the exact coverage of the surface). Note in passing that, as above explained, coverage beyond 0.5 ML is achievable depending on the water exposure.
Furthermore, even coexistence of domains with different local coverage cannot be discarded.
According to Figures 2 and 5, the O s -H(0001) species assignment seems the only possible from our data and it should be noted that contributions in the 3370-3450 cm -1 have been frequently observed at ZnO nanoparticles faceting polar surfaces. 34,[48][49][50] Nevertheless other unexplored situations (such as edges, kinks, etc.) were previously claimed as a possibility to assign such contribution. Thus, the assignment of the 3410 cm -1 presented in Figure 8 to a polar contribution seems highly likely although not conclusive.
Above 3500 cm -1 we can see signals from both polar (0001) and nonpolar (1010) surfaces although the latter seems more important at higher wavenumbers. This fact becomes more evident with the treatment at higher temperature. However the high frequency peak displays relative small intensity and would require further analysis for definitive analysis of this region. Thus the uncertain assignment of peaks at this zone has been visually translated in overlapping colours in Figure 8.

Summary and Outlook
The present work presents an extensive and systematic computational study aimed at the identification, by means of simulated IR spectra, of hydroxyl and hydride species formed When dosing H 2 O, the situation changes; a (1×2) pattern is clearly preferredadsorption energy of 286 kJ mol -1 -having hydroxyl and hydride moieties on (0001) and (0001) surfaces, respectively. This leads to additive IR signals at 3744 and 3755 cm -1 , although the former signal is ~100 times more intense, and the only thought to be detectable.
In any case, when dosing H 2 O to these surfaces, hydroxyl signals on nonpolar surfaces are the most intense: On the (1010) water arranges in a mixed dissociated state, with an adsorption energy of 112 kJ mol-1, in which the hydroxyls formed from water splitting feature a stretching frequency at 3765 cm -1 , this is, an IR signal blue shifted from the polar surfaces.
Furthermore, H adatoms formed from H 2 O partial dissociation produce surface hydroxyls with a characteristic IR stretching at 3159 cm -1 . This mode is coupled with non-dissociated H 2 O stretching modes -2864 and 2994 cm -1 -forming a broad band. Finally, water is completely dissociated on the (1120) surface, revealing two clearly distinguishable peaks at respectively.
The combination of experimental and theoretical methods does not only allow one to discriminate among different water-related species in IR studies concerning model surfaces, but also in photocatalytically relevant ZnO nanoparticles. A very important finding is that experimental frequency regions between 3100-3370 and 3370-3500 cm -1 seems distinctive of nonpolar and polar surfaces, respectively, and would thus allow to identify the dominance of such type of surfaces at in situ conditions where the application of microscopy or other techniques is rather complex. This indicates that IR is a powerful tool in describing shape and surface morphology of ZnO nanostructures and allows to follow active surface sites under reaction conditions.       Table 1.  Table 2.  Colour-code is the same as the used in Figure 1 labelling the surfaces Figure 8: Intensity of the contributions presented in Figure 7 A/B and ascription to different surface planes. Colour-code is the same as those in Figure 1 to label the surfaces.