Polarization- and Azimuth-Resolved Infrared Spectroscopy of Water

---

Letter pubs.acs.org/JPCL

Polarization- and Azimuth-Resolved Infrared Spectroscopy of Water on TiO2(110): Anisotropy and the Hydrogen-Bonding Network Greg A. Kimmel,*,† Marcel Baer,† Nikolay G. Petrik,† Joost VandeVondele,‡ Roger Rousseau,† and Christopher J. Mundy*,† †

Chemical and Materials Sciences Division, Pacific Northwest National Laboratory, P.O. Box 999, Richland, Washington 99352, United States ‡ Department of Materials, ETH Zurich, Wolfgang-Pauli-Strasse 27, 8093 Zurich, Switzerland ABSTRACT: We have investigated the structure and dynamics of thin water films adsorbed on TiO2(110) using infrared reflection−absorption spectroscopy (IRAS) and ab initio molecular dynamics. Infrared spectra were obtained for sand p-polarized light with the plane of incidence parallel to the [001] and [11̅0] azimuths for water coverages ≤ 4 monolayers. The spectra indicate strong anisotropy in the water films. The vibrational densities of states predicted by the ab initio simulations for 1 and 2 monolayer coverages agree well with the observations. The results provide new insight into the structure of water on TiO2(110) and resolve a long-standing puzzle regarding the hydrogen bonding between molecules in the first and second monolayers on this surface. The results also demonstrate the capabilities of polarization- and azimuth-resolved IRAS for investigating the structure and dynamics of adsorbates on dielectric substrates. SECTION: Surfaces, Interfaces, Catalysis

T

he interactions of water with surfaces are pervasive in nature and technology and are important for numerous scientific disciplines.1−3 Because TiO2 has many uses, including as a photocatalyst to remove organic pollutants from water4 and as thin-film coatings for self-cleaning surfaces,1 the interactions of water with TiO2 are particularly important. Surface science techniques have proved particularly useful for developing a molecular-level understanding of water’s interactions with wellcharacterized surfaces. Because rutile TiO2 single crystals are readily available and the (110) face is the most stable, many fundamental investigations have considered water adsorption on TiO2(110).5−18 The most common defects on the surface of reduced TiO2(110) are vacancies, VO, in the bridge-bonded oxygen (BBO) rows (see Figure 1). Water dissociatively adsorbs in the vacancies, creating two bridging hydroxyls.8,15,19,20 Water subsequently adsorbs above five-fold-coordinated titanium (Ti5c) and then above BBO sites.8 We will refer to these as H2OTi and H2OBBO, respectively. Early experiments suggested that water molecularly adsorbs on the Ti5c sites.8,9,21 However, a recent XPS result suggested a “mixed” adsorption mode with both molecular and dissociative adsorption.22 Theoretically, the adsorption state of water on TiO2(110) is also an area of active debate,18 with dissociative, molecular, and mixed adsorption modes all predicted in various simulations.6,13,14,23−28 Less information is available regarding the bonding within water films on TiO2(110). High-resolution electron energy loss spectroscopy (HREELS) suggested that H2OBBO does not form hydrogen bonds (H-bonds) with H2OTi.8 In contrast, work © 2012 American Chemical Society

Figure 1. IRAS geometry for s-polarized and p-polarized light (E(s) and E(p), respectively) incident along the [001] azimuth of TiO2(110). In this case, s-polarized spectra are sensitive to vibrations perpendicular to the BBO rows, while p-polarized spectra are sensitive to vibrations both normal to the surface and parallel to the surface along the BBO rows. Experiments were also performed with the plane of incidence perpendicular to the BBO rows (i.e., in the [110̅ ] azimuth).

function measurements indicated that H2OTi adsorbs with its C2v axis “tilted strongly off the surface normal,” and that Received: January 27, 2012 Accepted: February 28, 2012 Published: February 28, 2012 778

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

H2OBBO is H-bonded to H2OTi with its dipole parallel to the surface.9 X-ray scattering measurements for liquid water on TiO2(110),29,30 H−D exchange in 2 ML films,31 and ab initio molecular dynamics (AIMD) simulations14,23,27 also suggest Hbonding between H2OBBO and H2OTi. Infrared (IR) spectroscopy is a powerful tool for investigating the chemical identity, structure, and dynamics of adsorbates on surfaces.2,3,32 Because the OH stretch frequency of H-bonds depends on characteristics of the bond, IR spectroscopy is particularly useful for investigating H-bonded systems, including thin water films.2,3,33,34 However, because the signals in Fourier-transform infrared reflection−absorption spectroscopy (IRAS) on dielectric substrates are typically an order of magnitude smaller than those on metals, relatively few experiments have been done on well-characterized, dielectric single crystals.35−38 Here, we use IRAS with s- and p-polarized light incident along both the [001] and [11̅0 ] azimuths and AIMD simulations to investigate D 2O films (coverages, θ, ≤4 ML) on rutile TiO 2(110). The IRAS spectra provide information on the vibrations perpendicular and parallel to the surface and reveal strong asymmetries in the bonding of water on TiO2(110). For θ = 1 ML, comparison of the experimental and calculated vibrational spectra indicate that the films consist primarily of lines of fully coordinated molecules that adsorb nearly parallel to the surface on Ti5c sites. Two monolayer D2O films also form fully coordinated “strands” of H-bonded molecules that run parallel to the BBO rows. In contrast to the 1 ML films, these strands also have vibrations normal to the surface. Our results provide new insights into the structure and dynamics of water on TiO2(110). Figure 1 shows the IRAS geometry for s- and p-polarized light with the plane of incidence, defined by the light’s wave vector, k, and the surface normal, in the [001] azimuth. The electric field vector for s-polarized light (Figure 1, E(s)) is parallel to the surface and perpendicular to the BBO rows. Thus, the IRAS spectra are sensitive to vibrations in that direction. For p-polarized light, the E field has components both perpendicular and parallel to the surface, (Figure 1, E(p)), and those spectra reflect vibrations in the [001] direction and normal to the surface. For the IR beam in the [11̅0] azimuth (not shown), the s-polarized spectra are due to vibrations parallel to the BBO rows, and the p-polarized spectra reflect vibrations perpendicular to the BBO rows and normal to the surface. Figure 2 shows the s-polarized spectra for 0.8 ≤ θ ≤ 4.0 ML in [110̅ ] and [001] azimuths. D2O was deposited at 130 K, and the spectra were obtained at 30 K. For θ = 0.8 and 1.0 ML, the spectra in the [11̅0] azimuth consist primarily of a single, relatively narrow peak at 2605 cm−1 (Figure 2, top). Note that this peak is not from non-H-bonded D (i.e., “dangling” ODs), which are seen at 2725 cm−1 (see Figure 3). For coverages greater than 1 ML along [11̅0], the spectra begin to show absorption in a broader band centered at ∼2410 cm−1 that dominates the spectra for larger coverages. Along the [001] azimuth for θ ≤ 1.5 ML, the largest feature in the spectra is a band centered at ∼2335 cm−1, and a small peak is also observed at ∼2605 cm−1 (Figure 2, bottom). For θ = 2.0 and 3.0 ML, the largest absorbance shifts to ∼2250 cm−1. For s-polarized light (and unpolarized light), no reliable signal was detected for θ ≤ 0.5 ML. (As a next step in this research effort, we will try investigating water films for θ < 1

Figure 2. S-polarized IRAS spectra for D2O on TiO2(110) with the IR beam in the [11̅0] and [001] azimuths (top and bottom, respectively). The (scaled) s-polarized spectrum for 100 ML amorphous D2O is shown for comparison.

Figure 3. P-polarized IRAS spectra for D2O on TiO2(110) along the [11̅0] and [001] azimuths (top and bottom, respectively). The (scaled) p-polarized spectrum for 100 ML amorphous D2O is shown for comparison.

ML by increasing the amount of signal averaging at these lower coverages.) A key observation in Figure 2 is that for the two different azimuths, the s-polarized spectra are quite different for θ ≤ 4.0 ML due to the anisotropic interactions of water with TiO2(110). Figure 3 shows the p-polarized IRAS spectra for 0.8 ≤ θ ≤ 4.0 ML for the two azimuths. D2O was deposited at 130 K, and the spectra were obtained at 30 K. Due to optical effects, these signals are approximately 10 times larger than the corresponding spolarized spectra (Figure 2). For θ = 0.8 and 1.0 ML along both 779

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

Figure 4. Structure and dynamics for θ = 1 ML D2O on TiO2(110). (a) Top view and (b) side view of the calculated structure. (c) Calculated VDOS (solid lines) and d(ω) parameters (dashed lines) versus frequency. The same colors are used for each direction for both the VDOS and d parameters. (d) IRAS spectra for s- and p-polarized light along both azimuths (solid lines) and simulated spectra (dashed lines) using the d(ω) shown in (c).

incidence relative to the surface normal.41 For p-polarized light, the reflected amplitude, rp(x), is given by

azimuths, the absorbances are positive and small. However, for θ ≥ 2.0 ML, the spectra are primarily negative with stronger absorbances. For all of the coverages shown, there are significant differences between the spectra for the two azimuths. This anisotropy arises primarily from the small component of the E field that is parallel with the surface and thus is sensitive to the anisotropy observed in the s-polarized spectra in Figure 2. The D2O bending mode (1225 cm−1) is seen for 100 ML amorphous D2O films (Figures 2 and 3) but is not reliably observed for θ ≤ 4 ML, presumably due to the small signals. Figure 4 compares the calculated structure and dynamics for a defect-free surface with experiment for θ = 1 ML. In the simulations, water molecularly adsorbs on the defect-free surface above the Ti5c sites, nearly parallel to the surface, forming a line. Each D2OTi forms two H-bonds with its neighbors in the [001] direction. The other D forms a relatively strong bond with an adjacent BBO atom. To better accommodate a tetrahedral bonding geometry, the bonds between D2OTi and BBO atoms alternate between the BBO rows on either side (Figure 4a). The H-bonds within the calculated structure are indicated by the dotted white lines (Figure 4a and b). The solid lines in Figure 4d show the s- and p-polarized IRAS spectra of 1 ML D2O for both azimuths. Because optical effects on dielectric substrates complicate the interpretation of the spectra, particularly for p-polarization,38,39 we use d parameter theory to simulate the reflections from the adsorbed water films.40,41 For x and y parallel to the surface and z normal to the surface, the molecular absorptions in those directions are related to the diagonal elements of the d matrix, dxx(λ−1), dyy(λ−1), and dzz(λ−1), respectively. For s-polarized light with k∥x, the complex reflected amplitude, rs(x), is given by rs(x) rs0(x)

=1+i

2π cos θin d yy(λ−1) λ

rp(x) rp0(x)

=1+i

2π cos θin λ

⎡ d (λ−1)(ε − sin 2 θ ) − d (λ−1)ε sin 2 θ ⎤ 1 in zz 1 in ⎥ × ⎢ xx ⎢⎣ ⎥⎦ ε1 cos2 θin − sin 2 θin (2)

where ε1 is the permittivity of the substrate. (For k∥y, simply swap the x and y subscripts in eqs 1 and 2.) The measured absorbance is 41

A i (ω) = log10[(ri /ri0)(ri /ri0)*]

(3)

where the i = s or p for s- or p-polarized light and (...)* denotes the complex conjugate. The spectra simulated with d parameter theory (Figure 4d, dashed lines) faithfully reproduce the experiments. The imaginary components of the d parameters, which are proportional to the polarizabilities,41 are shown in Figure 4c (dashed lines). The solid lines in Figure 4c show the calculated vibrational density of states (VDOS) along the [001], [11̅0], and [110] directions for 1 ML D2O. The agreement between the experimentally derived Im[d(ω)] and the calculated VDOS is excellent. Note also that the IRAS spectra can be described by a single set of d parameters, which indicates that the water films had similar structures on the two different crystals used for the experiments on the [001] and [110̅ ] azimuths. The AIMD simulations provide a simple interpretation of the IRAS spectra. For θ = 1 ML on a defect-free surface, every molecule is nearly equivalent. Thus, there are two types of Hbonds and one IR band associated with each. Once the projections of each bond along the different directions are included, the bands associated with the two bonds can be clearly identified. The first band, due to H-bonds between adjacent D2OTi’s (see Figure 4), are primarily parallel to the BBO rows and are responsible for the narrow peak at 2605 cm−1 in the experiments and at 2572 cm−1 in the VDOS (Figure 4c, red lines). Becuase these bonds are not completely

(1)

where rs0(x) is the reflected amplitude from the bare substrate, λ is the wavelength, and θin is the angle of 780

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

Figure 5. Structure and dynamics for θ = 2 ML D2O on TiO2(110). (a) Top view and (b) side view of the calculated structure. (c) Calculated VDOS (solid lines) and d(ω) parameters (dashed lines) versus frequency. The same colors are used for each direction for both the VDOS and d parameters. (d) IRAS spectra for s- and p-polarized light along both azimuths (solid lines) and simulated spectra (dashed lines) using the d(ω) shown in (c).

though the p-polarized spectra are nominally dominated by the vibrations normal to the surface due to the grazing angle of incidence, the spectra obtained along the two azimuths are quite different. For the [001] azimuth, the absorbance is negative with a single broad peak at ∼2380 cm−1 (Figure 5d, dark blue). On the other azimuth, there are additional peaks at 2605 and 2725 cm−1 (Figure 5d, light blue). Finally, there is a positive “peak” at 2235 cm−1 that is absent for the [001] azimuth. The solid lines in Figure 5c show the projections of the calculated VDOS in the [001], [11̅0], and [110] directions for the 2 ML D2O film. Compared to the experimentally determined absorptions (Figure 5c, dashed lines), the calculated VDOS covers a broader range of frequencies. Furthermore, the VDOS has three resolved peaks, while the experiments do not. However, there is good qualitative agreement between the experiments and the calculations. First, theory and experiment both show a relatively narrow peak due to vibrations in the [001] direction, which also has smaller components in the other directions. Second, as indicated in Figure 5c, the experimental absorption band centered at 2275 cm−1 for the [11̅0] azimuth appears to be composed of two unresolved peaks. This hypothesis is consistent with the observation that for vibrations in the [11̅0] direction, the 2 ML spectra have additional intensity at lower frequencies in both the experiments and the theory when compared to the 1 ML spectra. In the calculated structure for θ = 2 ML (Figure 5a and b), all of the molecules in the first layer are nearly equivalent, as are all of the molecules in the second layer. Therefore, there should be four different H-bonds and four associated vibrational bands, while the calculated VDOS shows only three major peaks. However, theory shows that the H-bonds connecting neighboring molecules along the [001] direction in the first and second layers both contribute to the peak at 2580 cm−1 in the VDOS (and 2605 cm−1 in IRAS). This peak is very similar to the one observed for θ = 1 ML. Because these bonds are not completely parallel with the [001] direction, they also make contributions to the spectra along the other two directions. The VDOS for the [11̅0] and [110] directions has two more peaks

parallel to the BBO rows, the same peak is also evident in the other directions. The second H-bond, between D2OTi and BBO, is primarily along [11̅0] and leads to the peak at 2320 cm−1 in IRAS and that at 2360 cm−1 in theory (Figure 4c, black lines). The D2OTi's adsorb nearly parallel to the surface for θ = 1 ML (Figure 4b), leading to almost no absorption normal to the surface. According to eq 2, if dzz(ω) = 0, then the p-polarized spectra are proportional to the s-polarized spectra, but the absorbance will be positive, as seen experimentally. Thus, the theory explains two of the key observations for the 1 ML D2O films. However, as vibrations normal to the surface develop for the thicker films (e.g., 2 ML), the p-polarized spectra begin to display negative absorbances and no longer bear such a simple relationship to the s-polarized spectra (see Figure 2). Figure 5 compares the calculated structure and dynamics for θ = 2 ML with the experiments. Figure 5a and b shows the structure. No dissociation is observed in the simulations on a defect-free surface. The calculated structure is comprised of “strands” of water that run parallel to the BBO rows that do not form bonds with adjacent strands. Within each strand, D2OTi bonds to the Ti5c sites through an electron lone pair, and D2OBBO bonds to BBO via a D atom. The D2OTi and D2OBBO both form H-bonds with their neighbors in the [001] direction (Figure 5a). In contrast to the 1 ML film, D2OTi no longer bonds to the BBO. Instead, they are hydrogen donors to D2OBBO. D2OBBO adsorbs in a bridging position between the D2OTi and the BBO (see Figure 5b), consistent with previous measurements of the H−D exchange in 2 ML water films on TiO2(110).31 Note that for the calculated structure, all of the water molecules form four H-bonds, and there are no dangling ODs or lone pair electrons. The solid lines in Figure 5d show the s- and p-polarized IRAS spectra for θ = 2 ML. The fits of the spectra obtained using d parameter theory (dashed lines) are in excellent agreement with the results. The corresponding d parameters are shown in Figure 5c (dashed lines). Unlike the 1 ML D2O films, the 2 ML films have significant absorption from vibrations perpendicular to the surface. As a result, the p-polarized spectra now look quite different from the s-polarized spectra. Furthermore, even 781

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

centered at 2050 and 2210 cm−1. The H-bonds between the first- and second-layer water molecules give rise to the VDOS peak at 2050 cm−1, which we tentatively assign to the lowfrequency side of the unresolved peak in d([11̅0]) at 2275 cm−1 (Figure 5c, black dashed line). The H-bonds between D2OBBO and BBO are responsible for the peak at 2210 cm−1 that is seen in the VDOS in the [11̅0] and [110] directions and probably give rise to the peak at 2370 cm−1 in d([110]) (Figure 5c, blue dashed line). Because this bond is more normal to the surface (Figure 5b), it contributes more strongly in the absorption normal to the surface, as seen in both the theory and the experiments. While the calculated structure for θ = 2 ML has no dangling bonds (see Figure 5), the IRAS spectra for θ = 2 ML have a weak absorption at 2725 cm−1 that increases for θ = 3 and 4 ML and corresponds to the dangling ODs seen in amorphous D2O (see Figure 3). The origin of this small peak is uncertain, but it might be due to changes in the water bonding arrangement in the vicinity of bridging hydroxyls (which are not included in the AIMD). There are several important issues related to the agreement between theory and experiment seen in Figures 4 and 5. As already noted, the MD simulations were for a defect-free surface. In contrast, the experimental surface had bridging hydroxyls with concentrations of ∼0.1−0.16 ML due to dissociative adsorption of D2O in the vacancies. In the vicinity of the hydroxyls, the water films presumably do not have the same structures as those calculated for the defect-free surface. Thus, although we have good agreement between simulations and experiment, it is unlikely that the calculated structures for θ = 1 and 2 ML correspond to the true minimum-energy configurations realized by experiment. However, the agreement between theory and experiment (Figures 4 and 5) suggests that the observed signals primarily arise from molecularly adsorbed water bound to normal Ti5c and BBO sites with local structures similar to those calculated for θ = 1 and 2 ML. It is well-known that hydroxylated surfaces contain polarons in the sublayer. The precise treatment of these polarons using DFT is still controversial and must utilize a higher level of theory to treat the additional localized electron at significant computational expense. We believe that correct treatment of the dynamics of the excess charge due to the defects is critical for the calculations on defective surfaces. For both theory and experiment, the interaction of bridging hydroxyls with adsorbed water films will need to be addressed in future work. Another important issue concerns the dependence of the VDOS on the initial configuration used for the adsorbed water in the AIMD. For θ = 1 ML, preliminary calculations (not shown) started with a non-H-bonded film with the oxygen atoms above the Ti5c sites and all of the D’s pointing away from the surface. After geometry optimizations followed by lengthy MD, a VDOS was obtained that did not match the IRAS experiments. Although the simulations were of high quality, the amount of simulation time required for the water to completely rearrange into the lower-energy H-bonding pattern that agrees with experiment could not be realized with the standard NVE sampling that was performed here. Given that this naive choice of system preparation fits within any reasonable protocol, the preliminary exercise demonstrates the synergy needed between simulation and experiment to deduce the proper starting configurations and thus provides extra confidence in our results presented here. While the 2 ML films is composed of noninteracting strands running in the [001] direction (Figure 5), thicker films form a

three-dimensional network of H-bonded water, leading to IR spectra more typical of crystalline D2O ice or amorphous D2O. The results for 3 and 4 ML films (Figures 2 and 3) show increasing absorbance in the range expected for these threedimensional ices, but anisotropies due to the TiO2(110) persist. The transition from anisotropic, surface-mediated absorption to isotropic, bulk absorption will be explored in more detail in a future publication. Water adsorption on TiO2 particles and thin films has been investigated many times with IR spectroscopy.42 However, those samples precluded the kind of detailed investigation presented here. For experiments on single-crystal TiO2,8,17 the most detailed results are from HREELS on TiO2(110).8 For θH2O ≈ 0.9 and 1.3 ML, HREELS showed a narrow peak at 3505 cm−1 that corresponds to the peak at 2605 cm−1 for D2O.43 For θH2O ≈ 1.8 ML, this peak persisted with an additional broader peak at 3205 cm−1. Because the similarity of the 3505 cm−1 peak for θ = 1 and 2 ML suggested that the firstand second-layer water molecules were not H-bonding to each other, 8 this result has generated considerable attention.10,14,18,25,27,31,44,45 However, our results clearly resolve this issue by showing that the oscillator responsible for this peak is not normal to the surface, as would be expected for a dangling OD. Because the peak is observed in the s-polarized spectra and is a positive absorbance in the p-polarized spectra, we know that the oscillator is essentially parallel to the surface. Furthermore, the specific bond responsible for the peak at 2605 cm−1, the bond between adjacent molecules in the [001] direction, is similar for water in the first and second layers. The IRAS and AIMD also show that there is H-bonding between the layers that manifests itself at other OD stretch frequencies. The interactions of water with rutile TiO2(110) have been the subject of numerous theoretical investigations. Sun et al. have recently reviewed the work in this area,18 including a discussion of the long-running debate about the extent of dissociative versus molecular adsorption.6,14,23−26,28,46 A recent investigation indicates that molecular adsorption is preferred when sufficiently large simulation cells are used.27 However, the requirement of large cell sizes often limits the scope of AIMD simulations of the dynamics. Despite the difficulties, several AIMD studies have investigated the vibrational dynamics of water on TiO2(110).13,14,23,25−27 One of the most detailed studies predicted strong orientation effects in three-layer water films at ∼300 K due to the anisotropy of the substrate and the strong water−TiO2 interactions.23 In the calculations, water adsorbed at Ti5c sites with second-layer water preferentially adsorbed in a bridging position between the BBO and the H2OTi (i.e., similar to the 2 ML structure in Figure 5). However, the predicted vibrational spectra consisted of a narrow dangling OH peak and a broad OH stretch band that extended to quite low frequencies (e.g., ∼2500 cm−1 for H2O or 1850 cm−1 for D2O) due to the strong H-bonds formed between the surface and the adsorbed water. Compared to the results presented here, the calculations overestimated the strength of water’s interactions with TiO2(110), possibly because the simulation cell consisted of only three Ti layers.27 In summary, we have investigated the structure and dynamics of D2O films on rutile TiO2(110) using polarization- and azimuth-resolved infrared spectroscopy and ab initio molecular dynamics. We find that the bonding of water on TiO2(110) is highly anisotropic. For θ = 1 ML, the D2O adsorbs nearly parallel to the surface. Weak H-bonds are formed between adjacent D2OTi in the [001] direction, leading to a narrow, 782

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

ensemble with a typical time step of 0.5 fs. A plane-wave cutoff of 400 Rybergs was used for the density. All production level runs were of 25 ps at an average temperature of 100 K.

high-frequency OD stretch peak, while stronger bonds between D2OTi and BBO lead to lower-frequency OD stretches perpendicular to the BBO rows. For θ = 2 ML, long, weak bonds in the [001] direction between adjacent molecules in both the first and second layers again lead to a narrow, highfrequency peak. Stronger bonds between first- and second-layer D2O and between second-layer D2O and BBO atoms produce lower-frequency vibrations perpendicular to the BBO rows and normal to the surface. These results provide new insight into the binding of water on TiO2(110) and resolve a long-standing puzzle regarding the H-bonding between water molecules on Ti5c and BBO sites on this important surface. More generally, the results demonstrate the ability of polarization- and azimuthresolved IRAS to investigate the structure and dynamics of adsorbates on dielectric substrates.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: 509-371-6134 (G.A.K.); E-mail: [email protected]. Phone: 509-3752404 (C.J.M.). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported by the U.S. Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences & Biosciences. M.B. is grateful for the support of the Linus Pauling Distinguished Postdoctoral Fellowship Program at PNNL. Pacific Northwest National Laboratory (PNNL) is a multiprogram national laboratory operated for DOE by Battelle. A portion of the research was performed using EMSL, a national scientific user facility sponsored by the Department of Energy’s Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory. All simulations were performed under an INCITE 2011 award, using the resources of the Oak Ridge Leadership Computing Facility at the Oak Ridge National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725.

■

EXPERIMENTAL AND THEORETICAL METHODS The experiments were conducted in an ultrahigh vacuum system that has been described previously.33 The system was equipped with a closed-cycle helium cryostat, a quadrupole mass spectrometer, a molecular beamline for dosing adsorbates, and a Fourier-transform infrared spectrometer (Bruker Vertex 70). The infrared light was incident on the TiO2(110) at ∼84° with respect to normal and detected in the specular direction. A wire-grid polarizer was used to produce polarized light. Different crystals, which were prepared using the same procedures, were used for the experiments with the IR beam along the [001] and [110̅ ] azimuths. The crystals were prepared by multiple sputtering/annealing cycles with 2 keV Ne+ ions and annealing for between 2 and 10 min in vacuum at 950 K. The water (D2O or H2O) coverages, θ, were determined via temperature-programmed desorption (TPD).8 The coverages are defined relative to the water monolayer on TiO2(110), corresponding to 5.2 × 1014 #/cm2 per ML.8,9,44 With this definition, the “second” water monolayer also forms bonds directly to the surface.8,9,12,14,44 The bridging oxygen vacancy concentration on the surfaces, determined from the OD (or OH) recombinative desorption peak in the water TPD,8 were ∼0.05−0.08 ML. The IRAS spectra were obtained using a spectrometer resolution setting of 4 cm−1. The reported spectra are averages of multiple experiments (∼10 typically), each of which was the average of 1000 scans. Due to slow changes in the baseline over time, spectra obtained from (for example) 10 experiments with 1000 scans were found to be more reproducible than a single experiment with 10 000 scans. We subtracted a smoothly varying baseline such that the absorbance of the resulting spectra was zero at frequencies greater than and less than the OD stretch region. All density functional theory (DFT) calculations were performed with the CP2K package.47 DFT calculations are based on the Gaussian and plane wave method in conjunction with the PBE exchange−correlation functional with pseudopotentials.48,49 All D2O were described with a basis set at the triple-ζ level with double polarization.50 The TiO2 slab utilized an optimized basis set.51 A six TiO2 trilayer-thick, defect-free 6 × 2 rutile TiO2(110) surface slab model was employed to minimize long-range electrostatic effects that are known to influence the equilibrium between adsorbed and dissociated water molecules,27,28 with the last row of oxygens in the TiO2 system constrained to yield a single free interface. The 1 and 2 ML water systems were constructed with 12 and 24 water molecules. All simulations for the spectra were run in the NVE

■

REFERENCES

(1) Fujishima, A.; Zhang, X. T.; Tryk, D. A. TiO2 Photocatalysis and Related Surface Phenomena. Surf. Sci. Rep. 2008, 63, 515−582. (2) Henderson, M. A. The Interaction of Water with Solid Surfaces: Fundamental Aspects Revisited. Surf. Sci. Rep. 2002, 46, 1. (3) Hodgson, A.; Haq, S. Water Adsorption and the Wetting of Metal Surfaces. Surf. Sci. Rep. 2009, 64, 381−451. (4) Fujishima, A.; Honda, K. Electrochemical Photolysis of Water at a Semiconductor Electrode. Nature 1972, 238, 37. (5) Allegretti, F.; O’Brien, S.; Polcik, M.; Sayago, D. I.; Woodruff, D. P. Adsorption Bond Length for H2O on TiO2(110): A Key Parameter for Theoretical Understanding. Phys. Rev. Lett. 2005, 95, 226104. (6) Harris, L. A.; Quong, A. A. Molecular Chemisorption As the Theoretically Preferred Pathway for Water Adsorption on Ideal Rutile TiO2(110). Phys. Rev. Lett. 2004, 93, 086105. (7) Brookes, I. M.; Muryn, C. A.; Thornton, G. Imaging Water Dissociation on TiO2(110). Phys. Rev. Lett. 2001, 87, 266103. (8) Henderson, M. A. An HREELS and TPD Study of Water on TiO2(110): The Extent of Molecular versus Dissociative Adsorption. Surf. Sci. 1996, 355, 151−166. (9) Hugenschmidt, M. B.; Gamble, L.; Campbell, C. T. The Interaction of H2O with a TiO2(110) Surface. Surf. Sci. 1994, 302, 329−340. (10) Onda, K.; Li, B.; Petek, H. Two-Photon Photoemission Spectroscopy of TiO2(110) Surfaces Modified by Defects and O2 or H2O Adsorbates. Phys. Rev. B 2004, 70, 045415. (11) Onda, K.; Li, B.; Zhao, J.; Jordan, K. D.; Yang, J. L.; Petek, H. Wet Electrons at the H2O/TiO2(110). Surf. Sci. 2005, 308, 1154− 1158. (12) Predota, M.; Bandura, A. V.; Cummings, P. T.; Kubicki, J. D.; Wesolowski, D. J.; Chialvo, A. A.; Machesky, M. L. Electric Double Layer at the Rutile (110) Surface. 1. Structure of Surfaces and Interfacial Water from Molecular Dynamics by Use of Ab Initio Potentials. J. Phys. Chem. B 2004, 108, 12049−12060. (13) Zhang, C.; Lindan, P. J. D. Towards a First-Principles Picture of the Oxide−Water Interface. J. Chem. Phys. 2003, 119, 9183−9190. 783

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784

The Journal of Physical Chemistry Letters

Letter

(14) Zhang, C. J.; Lindan, P. J. D. Multilayer Water Adsorption on Rutile TiO2(110): A First-Principles Study. J. Chem. Phys. 2003, 118, 4620−4630. (15) Bikondoa, O.; Pang, C. L.; Ithnin, R.; Muryn, C. A.; Onishi, H.; Thornton, G. Direct Visualization of Defect-Mediated Dissociation of Water on TiO2(110). Nat. Mater. 2006, 5, 189−192. (16) Ketteler, G.; Yamamoto, S.; Bluhm, H.; Andersson, K.; Starr, D. E.; Ogletree, D. F.; Ogasawara, H.; Nilsson, A.; Salmeron, M. The Nature of Water Nucleation Sites on TiO2(110) Surfaces Revealed by Ambient Pressure X-ray Photoelectron Spectroscopy. J. Phys. Chem. C 2007, 111, 8278−8282. (17) Takahashi, K.; Yui, H. Analysis of Surface OH Groups on TiO2 Single Crystal with Polarization Modulation Infrared External Reflection Spectroscopy. J. Phys. Chem. C 2009, 113, 20322−20327. (18) Sun, C. H.; Liu, L. M.; Selloni, A.; Lu, G. Q.; Smith, S. C. Titania−Water Interactions: A Review of Theoretical Studies. J. Mater. Chem. 2010, 20, 10319−10334. (19) Wendt, S.; Matthiesen, J.; Schaub, R.; Vestergaard, E. K.; Laegsgaard, E.; Besenbacher, F.; Hammer, B. Formation and Splitting of Paired Hydroxyl Groups on Reduced TiO2(110). Phys. Rev. Lett. 2006, 96, 066107. (20) Zhang, Z.; Bondarchuk, O.; Kay, B. D.; White, J. M.; Dohnalek, Z. Imaging Water Dissociation on TiO2(110): Evidence for Inequivalent Geminate OH Groups. J. Phys. Chem. B 2006, 110, 21840−21845. (21) Krischok, S.; Hofft, O.; Gunster, J.; Stultz, J.; Goodman, D. W.; Kempter, V. H2O Interaction with Bare and Li-Precovered TiO2: Studies with Electron Spectroscopies (MIES and UPS(HeI and II)). Surf. Sci. 2001, 495, 8−18. (22) Walle, L. E.; Borg, A.; Uvdal, P.; Sandell, A. Experimental Evidence for Mixed Dissociative and Molecular Adsorption of Water on a Rutile TiO2(110) Surface without Oxygen Vacancies. Phys. Rev. B 2009, 80, 235436. (23) Kumar, N.; Neogi, S.; Kent, P. R. C.; Bandura, A. V.; Kubicki, J. D.; Wesolowski, D. J.; Cole, D.; Sofo, J. O. Hydrogen Bonds and Vibrations of Water on (110) Rutile. J. Phys. Chem. C 2009, 113, 13732−13740. (24) Langel, W. Car−Parrinello Simulation of H2O Dissociation on Rutile. Surf. Sci. 2002, 496, 141−150. (25) Lindan, P. J. D.; Harrison, N. M.; Gillan, M. J. Mixed Dissociative and Molecular Adsorption of Water on the Rutile (110) Surface. Phys. Rev. Lett. 1998, 80, 762−765. (26) Lindan, P. J. D.; Zhang, C. Exothermic Water Dissociation on the Rutile TiO2(110) Surface. Phys. Rev. B 2005, 72, 075439. (27) Liu, L. M.; Zhang, C. J.; Thornton, G.; Michaelides, A. Structure and Dynamics of Liquid Water on Rutile TiO2(110). Phys. Rev. B 2010, 82, 161415. (28) Stefanovich, E. V.; Truong, T. N. Ab Initio Study of Water Adsorption on TiO2(110): Molecular Adsorption versus Dissociative Chemisorption. Chem. Phys. Lett. 1999, 299, 623−629. (29) Zhang, Z.; et al. Ion Adsorption at the Rutile−Water Interface: Linking Molecular and Macroscopic Properties. Langmuir 2004, 20, 4954−4969. (30) Zhang, Z.; Fenter, P.; Sturchio, N. C.; Bedzyk, M. J.; Machesky, M. L.; Wesolowski, D. J. Structure of Rutile TiO2(110) in Water and 1 molal Rb at pH 12: Inter-relationship among Surface Charge, Interfacial Hydration Structure, and Substrate Structural Displacements. Surf. Sci. 2007, 601, 1129−1143. (31) Petrik, N. G.; Kimmel, G. A. Hydrogen Bonding, H−D Exchange, And Molecular Mobility in Thin Water Films on TiO2(110). Phys. Rev. Lett. 2007, 99, 196103. (32) Chabal, Y. J. Surface Infrared-Spectroscopy. Surf. Sci. Rep. 1988, 8, 211−357. (33) Feibelman, P. J.; Kimmel, G. A.; Smith, R. S.; Petrik, N. G.; Zubkov, T.; Kay, B. D. A Unique Vibrational Signature of Rotated Water Monolayers on Pt(111): Predicted and Observed. J. Chem. Phys. 2011, 134, 204702. (34) Kimmel, G. A.; Matthiesen, J.; Baer, M.; Mundy, C. J.; Petrik, N. G.; Smith, R. S.; Dohnalek, Z.; Kay, B. D. No Confinement Needed:

Observation of a Metastable Hydrophobic Wetting Two-Layer Ice on Graphene. J. Am. Chem. Soc. 2009, 131, 12838−12844. (35) Rohmann, C.; Wang, Y. M.; Muhler, M.; Metson, J.; Idriss, H.; Woll, C. Direct Monitoring of Photo-Induced Reactions on WellDefined Metal Oxide Surfaces Using Vibrational Spectroscopy. Chem. Phys. Lett. 2008, 460, 10−12. (36) Wang, Y. M.; Woll, C. Chemical Reactions on Metal Oxide Surfaces Investigated by Vibrational Spectroscopy. Surf. Sci. 2009, 603, 1589−1599. (37) Xu, M.; Gao, Y.; Moreno, E. M.; Kunst, M.; Muhler, M.; Wang, Y.; Idriss, H.; Woll, C. Photocatalytic Activity of Bulk TiO2 Anatase and Rutile Single Crystals Using Infrared Absorption Spectroscopy. Phys. Rev. Lett. 2011, 106, 138302. (38) Xu, M.; Gao, Y.; Wang, Y.; Woll, C. Monitoring Electronic Structure Changes of TiO2(110) via Sign Reversal of Adsorbate Vibrational Bands. Phys. Chem. Chem. Phys. 2010, 12, 3649−3652. (39) Hansen, W. N. Reflection Spectroscopy of Adsorbed Layers. Symp. Faraday Soc. 1970, 4, 27. (40) Chen, W.; Schaich, W. L. The Optical Effects of an Adsorbate Layer  D-Parameter Description of Simple-Models. Surf. Sci. 1989, 218, 580−602. (41) Schaich, W. L.; Chen, W. Nonlocal Corrections to Fresnel Optics  How to Extend D-Parameter Theory Beyond Jellium Models. Phys. Rev. B 1989, 39, 10714−10724. (42) See: Takahashi, K.; Yui, H. J. Phys. Chem. C 2009, 113, 20322 for references. (43) We have a limited set of IRAS spectra for H2O on TiO2(110). For 2 ML H2O, the s- and p-polarized spectra with k∥[11̅0] look very similar to those presented in Figure 4d, with the frequency scaled appropriately. We observe a peak at 3515 cm−1 corresponding to the peak at 2605 cm−1 for D2O. (44) Diebold, U. The Surface Science of Titanium Dioxide. Surf. Sci. Rep. 2003, 48, 53−229. (45) Panayotov, D. A.; Yates, J. T. Depletion of Conduction Band Electrons in TiO2 by Water, Chemisorption  IR Spectroscopic Studies of the Independence of Ti−OH Frequencies on Electron Concentration. Chem. Phys. Lett. 2005, 410, 11−17. (46) Goniakowski, J.; Gillan, M. J. The Adsorption of H2O on TiO2 and SnO2(110) Studied by First-Principles Calculations. Surf. Sci. 1996, 350, 145−158. (47) The CP2K developers group. http://cp2k.org/, (2009). (48) Perdew, J. P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77, 3865−3868. (49) Hartwigsen, C.; Goedecker, S.; Hutter, J. Relativistic Separable Dual-Space Gaussian Pseudopotentials from H to Rn. Phys. Rev. B 1998, 58, 3641−3662. (50) VandeVondele, J.; Krack, M.; Mohamed, F.; Parrinello, M.; Chassaing, T.; Hutter, J. QUICKSTEP: Fast and Accurate Density Functional Calculations Using a Mixed Gaussian and Plane Waves Approach. Comput. Phys. Commun. 2005, 167, 103−128. (51) Schiffmann, F.; VandeVondele, J.; Hutter, J.; Wirz, R.; Urakawa, A.; Baiker, A. Protonation-Dependent Binding of Ruthenium Bipyridyl Complexes to the Anatase(101) Surface. J. Phys. Chem. C 2010, 114, 8398−8404.

784

dx.doi.org/10.1021/jz3001079 | J. Phys. Chem. Lett. 2012, 3, 778−784