Hydration of Mineral Surfaces Probed at the Molecular Level

Nov 4, 2008 - multisite complexation (MUSIC) model.4,5 In this approach, the ..... 1994,. 94, 2095. (50) Wittbrodt, J. M.; Hase, W. L.; Schlegel, H. B...
0 downloads 0 Views 986KB Size
13434

Langmuir 2008, 24, 13434-13439

Hydration of Mineral Surfaces Probed at the Molecular Level Mathias Flo¨rsheimer,*,†,‡,§ Klaus Kruse,†,§,⊥ Robert Polly,†,§ Ahmed Abdelmonem,†,§ Bernd Schimmelpfennig,†,§ Reinhardt Klenze,†,§ and Thomas Fangha¨nel‡,§,| Institute for Nuclear Waste Disposal, Research Centre Karlsruhe, Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany, Institute of Physical Chemistry, UniVersity of Heidelberg, Im Neuenheimer Feld 252, D-69120 Heidelberg, Germany, Virtual Institute Functional Properties of Aquatic Interfaces, Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany, and European Commission, Joint Research Centre, Institute for Transuranium Elements, Helmholtz-Platz 1, D-76344 Eggenstein-Leopoldshafen, Germany ReceiVed June 3, 2008. ReVised Manuscript ReceiVed September 25, 2008 By employing the nonlinear optical, interface selective experiment of sum frequency spectroscopy together with independent ab initio and density functional theory calculations, we determine the functional species of a corundum (001) surface: doubly coordinated OH groups which differ in their bond tilt angles. The interaction of the functional species with the adjacent water molecules is also observed. In a large pH range around the point of zero charge, the interaction is not controlled electrostatically but by hydrogen bonding. The functional species’ tilt angles are crucial parameters, determining whether the species act as hydrogen bond donors or acceptors.

Introduction The interaction of a mineral with water is controlled by the mineral’s surface functional groups.1-3 These species determine, for example, the chemical equilibrium constants for the sorption of ions and molecules at the mineral/water interfaces in the Earth’s near subsurface. Knowledge of these constants is highly important for modeling the migration of elements and molecules in the aquifer.1-3 Our particular objective is to reliably predict radioactive ion migration in the frame of long-term safety assessment of nuclear waste repositories. A difficulty, however, is that the speciation of the functional groups of most mineral surfaces in water is not definitely known so far. The most basic reaction the species are expected to undergo is their protonation and deprotonation upon a pH change. A widely applied approach for understanding proton affinity of mineral/water interfaces at the level of functional species is the multisite complexation (MUSIC) model.4,5 In this approach, the chemical composition of the functional groups (sites) at a mineral’s surface is postulated based on the structure of the ideal (nonrelaxed) termination of the bulk crystal. The charge distribution within the species or within an adsorbed surface complex is then estimated, applying Pauling’s bond valence concept with corrections for hydrogen bonding6 and for the * To whom correspondence should be addressed. E-mail: [email protected]. † Institute for Nuclear Waste Disposal, Research Centre Karlsruhe. ‡ University of Heidelberg. § Virtual Institute Functional Properties of Aquatic Interfaces. | European Commission, Joint Research Centre, Institute for Transuranium Elements. ⊥ Present Address: euro engineering AG, Friedrichshafen, Germany. (1) Brown, G. E., Jr.; Henrich, V. E.; Casey, W. H.; Clark, D. L.; Eggleston, C.; Felmy, A.; Goodman, D. W.; Gra¨tzel, M. M.; Maciel, G.; McCarthy, M. I.; Nealson, K. H.; Sverjensky, D. A.; Toney, M. F.; Zachara, J. M. Chem. ReV 1999, 99, 77. (2) Sposito, G. The Surface Chemistry of Natural Particles; Oxford University Press: Oxford, 2005. (3) Lu¨tzenkirchen, J., Ed. Surface Complexation Modelling; Elsevier: London, 2006. (4) Hiemstra, T.; Van Riemsdijk, W. H.; Bolt, G. H. J. Colloid Interface Sci. 1989, 133, 91. (5) Hiemstra, T.; De Wit, J. C. M.; Van Riemsdijk, W. H. J. Colloid Interface Sci. 1989, 133, 105. (6) Hiemstra, T.; Venema, P.; Van Riemsdijk, W. H. J. Colloid Interface Sci. 1996, 184, 680.

potential gradient in the electric double layer7 near the surface. This semiempirical model successfully explains major differences in the properties of different surfaces and species.8 It has also been refined in various studies by quantum chemical means.8-10 In an alternative approach, molecular dynamics (MD) simulations were applied to small aqueous clusters11,12 and mineral/ water interfaces12,13 in order to derive site specific reaction rates, bond lengths, and binding energies. Interaction potentials for the MD simulations could be partially obtained from ab initio quantum chemical computations.14 For the simulation of the mineral surfaces, plausible functional species were assumed whereas the structures of the small clusters could be obtained from X-ray diffraction studies.12 In additional site specific 17O NMR experiments at the small clusters, oxygen-isotope-exchange rates were determined in order to derive reaction rates for water detachment. The rates were shown to be reasonably in agreement with the rates for the same cluster sites obtained from the MD simulations.12,15 For analogous sites of the clusters and the mineral surfaces, a strong correlation was found between the calculated water detachment rates and the calculated bond lengths between the water molecule and the active site, indicating the feasibility to determine structure/reactivity relations which are valid for clusters as well as for mineral surfaces.12,15 In spite of all this progress in understanding mineral surfaces at the molecular level, direct experimental information on the chemical composition and geometric structure of the functional species at real mineral/water interfaces is still lacking. X-ray reflectivity experiments are a means for studying mineral surfaces (7) Hiemstra, T.; Van Riemsdijk, W. H. J. Colloid Interface Sci. 1996, 179, 488. (8) Hiemstra, T.; Van Riemsdijk, W. H. J. Colloid Interface Sci. 2006, 301, 1. (9) Bickmore, B. R.; Tadanier, C. J.; Rosso, K. M.; Monn, W. D.; Eggett, D. L. Geochim. Cosmochim. Acta 2004, 68, 2025. (10) Bickmore, B. R.; Rosso, K. M.; Tadanier, C. J.; Bylaska, E. J.; Doud, D. Geochim. Cosmochim. Acta 2006, 70, 4057. (11) Rustad, J. R. Geochim. Cosmochim. Acta 2005, 69, 4397. (12) Wang, J.; Rustad, J. R.; Casey, W. H. Inorg. Chem. 2007, 46, 2962. (13) Felmy, A. R.; Rustad, J. R. Geochim. Cosmochim. Acta 1998, 62, 25. (14) Wasserman, E.; Rustad, J. R.; Xantheas, S. S. J. Chem. Phys. 1997, 106, 9769. (15) Casey, W. H.; Rustad, J. R. Annu. ReV. Earth Planet Sci. 2007, 35, 21.

10.1021/la801677y CCC: $40.75  2008 American Chemical Society Published on Web 11/04/2008

Hydration of Mineral Surfaces

in situ in contact with water.16-18 The experiments provide the atomic layer sequence near the interface, and the layer distance can be determined with high precision. From the interpretation, rich information on the surface speciation can be obtained. This technique, however, is not sensitive to hydrogen. Hence, small functional species such as OH groups cannot be directly probed. Direct and selective access to such species is the benefit of sum frequency (SF) spectroscopy.19 We take advantage of this nonlinear optical method which allows us additionally to probe the hydrogen bonding network of the water molecules near the interface. This network exhibits a preferential polar order due to the presence of the surface. We determine and interpret the sign20 of polar water orientation as a function of the surface speciation. Polar water ordering near hydrophobic surfaces could be well understood at the molecular level during the recent years.21 Hydrophobic hydration is enormously important because it is the driving force for key biological phenomena such as protein folding and self-assembly of membranes. Hydration of hydrophilic entities such as mineral surfaces, however, is still far from being understood. We apply our spectroscopic method to the corundum (001)/water interface and compare the results with independent quantum chemical ab initio and density functional theory (DFT) calculations. Corundum (R-Al2O3) is selected because it can be considered as a simple model for the naturally widespread clay minerals and analogous iron phases. In our study, it turns out additionally that the speciation of the functional groups of the corundum (001) surface is particularly well suited for examining predictions on the hydrogen bonding interaction of water with the functional species of a surface, obtained from recent statistical simulations.22-24 In these studies, it was postulated that the concentration ratio22,23 of hydrogen bond donor and acceptor species of a surface is crucial for the preferential order of the adjacent water molecules. This hypothesis was verified in statistical simulations at artificial model walls.22,23 In very recent computer simulations,24 a species’ geometry was systematically varied. It was then observed that its ability to act as donor or acceptor depends on the geometry. From such theoretical investigations, we expect that hydration of a surface is controlled not only by the chemical composition of the functional species but also by their geometric structure. In order to verify the expected effect of the geometry in an experiment, we need a real interface in which we can observe an alteration of the species’ geometry without variation of their chemical composition. Our investigations below show that the corundum (001)/water interface exhibits various doubly coordinated OH species, differing in a single geometric parameter, the bond tilt angle. An OH species can interact with adjacent water molecules as hydrogen bond donor or acceptor. We expect that the tilt angle determines whether binding as acceptor or as donor is preferred. At the corundum (001)/water interface, we observe that the relative concentrations of the differently tilted species depend on the pH. By altering the pH, we are thus able to investigate the effect of (16) Eng, P. J.; Trainor, T. P., Jr.; Waychunas, G. A.; Newville, M.; Sutton, S. R.; Rivers, M. L. Science 2000, 288, 1029. (17) Zhang, Z.; Fenter, P.; Cheng, L.; Sturchio, N. C.; Bedzyk, M. J.; Pøredota, M.; Bandura, A.; Kubicki, J. D.; Lvov, S. N.; Cummings, P. T.; Chialvo, A. A.; Ridley, M. K.; Be´ne´zeth, P.; Anovitz, L.; Palmer, D. A.; Machesky, M. L.; Wesolowski, D. J. Langmuir 2004, 20, 4954. (18) Fenter, P.; Sturchio, N. C. Prog. Surf. Sci. 2004, 77, 171. (19) Shen, Y. R. Nature 1989, 337, 519. (20) Shen, Y. R.; Ostroverkhov, V. Chem. ReV. 2006, 106, 1140. (21) Chandler, D. Nature 2005, 437, 640. (22) Hayashi, T.; Pertsin, A. J.; Grunze, M. J. Chem. Phys. 2002, 117, 6271. (23) Besseling, N. A. M. Langmuir 1997, 13, 2113. (24) Janecˇek, J.; Netz, R. R. Langmuir 2007, 23, 8417.

Langmuir, Vol. 24, No. 23, 2008 13435

different tilt angles and different donor/acceptor concentration ratios on the preferential order of the hydrogen bonded water molecules near the interface.

Experiment and Spectra The SF generation experiment19 provides a signal selectively from the mineral surface and from the adjacent water film whose molecules exhibit a net polar orientation due to their interaction with the mineral. In the bulk of water, the molecular dipoles cancel each other. Thus, no bulk water signal is generated. The same is valid for the bulk of many crystals including corundum. The experimental geometry is shown schematically in the insets of Figure 1a and b with two incident laser beams (indicated as red and green arrows) and a radiated signal (blue) at the sum of the fundamental frequencies. Chemical analytical information is obtained by tuning the frequency of one of the two lasers over the infrared (IR) spectrum, thus probing the vibrational resonances of the interface. The light of the other (auxiliary) laser impinges on the interface exactly at the critical angle of total internal reflection (TIR). The TIR geometry provides superior SF generation efficiencies as compared to external reflection experiments and thus facilitates recording of low-noise spectra. The corundum crystals for the experiments were obtained from Kyburz, Safnern, Switzerland. They were epi polished to serve as substrates for epitaxial film growth. We cleaned the surfaces by subsequently submerging them into chloroform, methanol, and water at pH 12. The pH of the electrolytes was obtained by applying HCl and NaOH. Solutions with ionic strength I of less than 50 mM were adjusted to I ) 50 mM using NaCl. In order to exclude CO2, the electrolytes were purged with Ar prior to the experiment. The spectrochemical cell was kept in Ar atmosphere during the measurements. After having taken a series of spectra at different pH values, we always repeated the first spectrum in order to make sure that no surface alteration or degradation occurred. The OH stretch region of a series of spectra is given in Figure 1. The spectra of Figure 1a and b have been taken at pH 12 applying different combinations of polarizer and analyzer orientation (see insets). The two experiments are thus differently sensitive to the dipole components of the species at the interface. With the auxiliary laser beam at the critical angle, the experiment of Figure 1a is selectiVely sensitive to the normal dipole contributions. (For more details, see the Supporting Information.) In-plane and normal dipole components contribute to the spectrum of Figure 1b. The quantitative comparison of the spectra provides the axial orientation of the species’ OH bonds. The effect of a pH change is then shown in the series from Figure 1b-d which have all been taken in the geometry of Figure 1b.

Experimental Results and Discussion Identification of the Functional Species. The relatively low noise level of the spectra facilitates their deconvolution into several bands whose magnitudes are drawn in Figure 1. The two broad resonances at smaller wavenumbers (blue lines in Figure 1) are due to the water molecules near the interface which exhibit a preferential polar order. These bands are well-known from other aquatic interfaces.20,25-28 At higher frequencies, we, however, observe a surprisingly large number of additional bands (red lines in Figure 1) which were not resolved previously. They are due to different OH surface species. The bands given in Figure 1 are the result of the simultaneous fit of all four spectra. We compared the result of the deconvolution with fits of several additional series of spectra, in which we altered the pH, the ionic strength, and the polarizer orientations. (25) Richmond, G. L. Chem. ReV. 2002, 102, 2693. (26) Miranda, P. B.; Shen, Y. R. J. Phys. Chem. B 1999, 103, 3292. (27) Raymond, E. A.; Tarbuck, T. L.; Brown, M. G.; Richmond, G. L. J. Phys. Chem. B 2003, 107, 546. (28) Gopalakrishnan, S.; Liu, D.; Allen, H. C.; Kuo, M.; Shultz, M. J. Chem. ReV. 2006, 106, 1155.

13436 Langmuir, Vol. 24, No. 23, 2008

Flo¨rsheimer et al.

resonant background (green lines in Figure 1 with maxima near 4050 cm-1). We quantified this background, which is not due to the OH vibrations of the interface, in independent experiments (see the Supporting Information). The peak in the spectra with its maximum near 3700 cm-1 exhibits a relatively sharp apex and is asymmetrically broadened. Broadening at the higher frequency side of the apex is mainly due to the resonant background. The apex originates from a relatively sharp aluminol band at 3690 cm-1. Broadening at the lower frequency side is due to (at least) two additional aluminol bands whose widths are similar to the width of the lowest frequency aluminol band at 3460 cm-1. This latter band is clearly visible in the spectra of Figure 1b-d which are sensitive to in-plane and normal dipole contributions. Comparing the amplitudes of this band in the two experiments at pH 12 (Figure 1a and b) shows us that the OH bond of the corresponding species is oriented relatively flat. The band positions of the different aluminol species are given in Figure 1b together with the quantitative tilt angles relative to the surface normal. Decreasing resonance frequency is observed to correlate with increasing tilt. From the ideal termination29 of the bulk crystal structure, one single type of aluminol species is expected to exist at the (001) surface: an OH species in the bridge position to two Al atoms. Our spectra show directly that the structure of the real interface is more complicated. From the extensive literature on IR and Raman spectroscopy of the different phases of aluminum (hydr)oxide colloids and powders,30 we know that a decreasing resonance frequency could also be due to an increase in the coordination number. In principle, the aluminol band of the lowest frequency in our spectra could originate from an OH group, tribridged to three Al atoms. The OH bond of such a species, however, should exhibit a small tilt, but the opposite is true as we have measured. Most likely, all the OH species of the (001) surface are doubly coordinated. They differ in their OH bond tilt angle. The geometries of the species with the largest and the smallest tilt are drawn schematically in Figure 1a. The flat oriented species are expected to form intrasurface hydrogen bonds with neighboring oxygen atoms.31 A recent interpretation of the band at 3460 cm-1 to be due to a water species32 can be ruled out.33 A few mineral/water interfaces20,25,26,32,34-37 including the corundum (001)/electrolyte interface32,34-37 have recently been investigated by SF spectroscopy. The spectra from samples which have been cleaned in a similar way as in our experiments with solvents at room temperature resemble our spectra, although the substructure of the spectral peak near 3700 cm-1 has not been resolved previously. The different tilt angles of the species have also not been observed previously. Cleaning the sample surfaces by plasma treatment34 or annealing the crystals at elevated temperature32,36 leads to significant alterations of the spectra. Investigating the temperature dependence of the spectra is of considerable geochemical interest, but it is not in the scope of Figure 1. Series of sum frequency spectra taken from a corundum (001) surface in situ in contact with water at different pH values and different experimental geometries. The spectra (b-d) have been taken in the geometry shown in (b). Four doubly coordinated surface OH species are observed (bands marked as red curves) whose resonance frequencies and bond tilt angles relative to the surface normal are given in (b) (positions in cm-1/angles in deg). The polar water orientation (phases of water bands) at pH 6 and 2.7 cannot be explained by electrostatic forces (inadequate interpretation struck out in (c)) but by hydrogen bonding. Flipping of the net polar water orientation from pH 6 to 2.7 is due to the concentration increase of the flat oriented OH species with their larger probability to act as hydrogen bond acceptors as compared to the steeper oriented OH species.

In all cases, we obtained the same set of bands. A key feature for the correct fitting and understanding of the spectra is the

(29) Barro´n, V.; Torrent, J. J. Colloid Interface Sci. 1996, 177, 407. (30) Morterra, C.; Magnacca, G. Catal. Today 1996, 27, 497. (31) Hass, K. C.; Schneider, W. F.; Curioni, A.; Andreoni, W. J. Phys. Chem. B 2000, 104, 5527. (32) Braunschweig, G.; Eissner, S.; Daum, W. J. Phys. Chem. C 2008, 112, 1751. (33) At high pH, the surface is charged negatively due to partial deprotonation. The thickness of the electric double layer at that surface and the number of water molecules in the layer depend on the ionic strength of the electrolyte. We verified that a change of the ionic strength affects the amplitudes of the two water bands in our spectra but not the amplitude of the band at 3460 cm-1. (34) Yeganeh, M. S.; Dougal, S. M.; Pink, H. S. Phys. ReV. Lett. 1999, 83, 1179. (35) Ma, G.; Liu, D.; Allen, H. C. Langmuir 2004, 20, 11620. (36) Chandraskharan, R.; Zhang, L.; Ostroverkhov, V.; Prakash, S.; Wu, Y.; Shen, Y. R.; Shannon, M. A. Surf. Sci. 2008, 602, 1466. (37) Zhang, L.; Tian, C.; Waychunas, G. A.; Shen, Y. R. J. Am. Chem. Soc. 2008, 130, 7686.

Hydration of Mineral Surfaces

the present Article in which we study the room temperature properties. An argument against using corundum as a model mineral for surface speciation studies is its non-negligible solubility, particularly at high and at low pH. We observe, however, that our results are reproducible. After having taken a series of spectra, we always repeated the first spectrum in order to make sure that no surface alterations occurred, as mentioned above. Spectral changes at the (001) surface can be observed after several days of contact with water. The alterations might be due to the creation of defects, to anisotropic etching and excavation of differently indexed surfaces, or to the precipitation of different aluminum oxide phases. For the quantitative studies described here, we thus used our crystals for ∼30 h in contact with electrolytes at maximum. They were then freshly polished and cleaned. In this way, we avoided measuring artifacts which would originate from the nonvanishing solubility of corundum. In the literature,38,16,39 there are various indications that corundum crystals, exposed to humid air or water, exhibit surface structures similar to the corresponding surfaces of the mineral gibbsite [γ-Al(OH)3] or bayerite [R-Al(OH)3]. In recent quantum chemical40,41 studies, it was found that the gibbsite (001) surface is composed of various doubly coordinated OH species which differ in their bond tilt angle. This result is very similar to our observation at the corundum (001)/water interface. The agreement supports our interpretation of the corundum (001) surface in the aquatic environment to be far from the ideal bulk terminated structure. Hydration of the Mineral Surface. Now, we apply the corundum (001) surface as model for studying hydration. The observation of various functional species which differ only in their OH bond tilt opens the possibility to investigate the preferential water order as a function of this geometric parameter. For this purpose, we take advantage from the phase20 information in the coherent nonlinear signal. So far, we have only considered the magnitudes of the bands in Figure 1a and b. The spectral deconvolution, however, provides also the relative phases which are given in Figure 1 at the resonance maxima (+90° or -90°) for all the OH bands. The aluminol species exhibit the same phase, shifted by 180° with respect to the phase of the water bands in Figure 1a and b. The shift of 180° means that the absolute orientation of the corresponding dipoles is opposite. We assume that the net dipole resulting from the OH bonds of the aluminol species is oriented normal to the surface and points into the electrolyte. (The in-plane dipole components cancel each other upon integrating over the surface.) The net dipole moment of the water molecules then points to the mineral surface (see inset in Figure 1a). This is exactly the result we expect at high pH due to partial deprotonation. Indications of protolysis were also observed in a recent SF spectroscopic adsorption study of the base piperidine to corundum (001)/humid air interfaces.35 The deprotonated aluminol species can of course not be observed directly in the spectra of the OH stretch vibrational region. They could, however, be detected in SF spectra of the surface lattice vibrations42 which have, however, not been measured for corundum so far. (38) Liu, P.; Kendelewicz, T., Jr.; Nelson, E. J.; Chambers, S. A. Surf. Sci. 1998, 417, 53. (39) Rabung, T.; Geckeis, H.; Wang, X. K.; Rothe, J.; Denecke, M. A.; Klenze, R.; Fangha¨nel, T. Radiochim. Acta 2006, 94, 609. (40) Dubot, P. Ecole Nationale Superieure de Chimie Paris (ENSCP), Paris, France. Personal communication. (41) Bickmore, B. R. Brigham Young University, Provo, UT. Personal communication. (42) Liu, W.-T.; Shen, Y. R. Phys. ReV. Lett. 2008, 101, 016101.

Langmuir, Vol. 24, No. 23, 2008 13437

Next, we consider the polar water orientation upon reduction of the pH within our series of spectra (Figure 1b-d). We may expect to find a crossover point (point of zero charge, pzc) without a net water dipole moment at which the negatively charged regime passes over into a positively charged (protonated) one.43-45 Such a crossover point was observed in recent second-harmonic generation (SHG) experiments at different mineral/water interfaces.43-45 SHG is an interface selective experiment similar to SF generation, but it is usually carried out with a laser at fixed frequency. Thus, no spectroscopic information is obtained. The measured signal is an integral over the nonresonant contributions from all functional species and from the polar ordered water film. For the interpretation46 of the data, it was assumed that the signal change upon the alteration of the pH originates mainly from the reorientation of the water molecules and changes in the speciation of the functional groups can be neglected. For corundum (001)/electrolyte interfaces, a pzc in the pH region between 4 and 6 was then derived.43,44 Our spectroscopic experiment, however, provides additional surprising details. At pH 6.0 (Figure 1c), the phase of the water bands is shifted47 by 180° with respect to the phase at pH 12 (Figure 1b), indicating that the net water dipole moment points into the direction of the bulk liquid. If this phenomenon originated from electrostatic forces, the surface would be positively charged at pH 6 due to protonation. If this was true, however, the surface would still be positively charged at lower pH. However, at pH 2.7 (Figure 1d), we observe that the water dipoles flipped again. Polar water ordering can hence not be explained by electrostatic forces in Figure 1c and d. In a broad pH range around the pzc, spanning at least from pH 6 to 2.7, mineral/water interaction must be controlled predominantly by hydrogen bonds. This observation is well in agreement with results obtained from the MUSIC model for the gibbsite (001) face.48 In this model, the gibbsite (001) surface is considered to be composed of doubly coordinated OH species. They are calculated to be partially deprotonated at pH 10 and higher. For pH values below 10, little surface charging is expected. For a precise determination of the pzc from the SF data, we had to consider water band amplitudes only from pH regions in which the mineral/water interaction is dominated by electrostatic forces. This, however, is not in the scope of the present Article. The band amplitudes from hydrogen bonded water in Figure 1c and d are small. Hence, we expect that contributions from such water to a SHG signal are small, too. Possibly, the extrapolation of the SHG amplitudes from pH regions in which the mineral/ water interaction is controlled electrostatically onto the pH regions of vanishing surface charge leads to reasonable estimations of the pzc. This hypothesis could be verified by calculating the expected SHG signals as a function of the pH from corresponding series of SF spectra. Here, we are, however, mainly interested in the molecular interpretation of the polar water ordering phenomena, observed in the pH region of dominating hydrogen bonding interaction. In order to explain the inversion of the net polar water orientation (43) Stack, A.; Higgins, S. R.; Eggleston, C. Geochim. Cosmochim. Acta 2000, 65, 3055. (44) Fitts, J. P.; Shang, X.; Flynn, G. W.; Heinz, T. F.; Eisenthal, K. B. J. Phys. Chem. B 2005, 109, 7981. (45) Kataoka, S.; Gurau, M. C.; Albertorio, F.; Holden, M. A.; Lim, S.-M.; Yang, R. D.; Cremer, P. S. Langmuir 2004, 20, 1662. (46) Ong, S.; Zhao, X.; Eisenthal, K. B. Chem. Phys. Lett. 1992, 191, 327. (47) The minimum signal in the spectrum of Figure 1c near 3250 cm-1 is due to net destructive interference of the fields from all the contributing species at that frequency. The net constructive interference in Figure 1b at the same frequency is a clear indication that the sign of the water dipoles flips between pH 12 and 6. (48) Hiemstra, T.; Yong, H.; Van Riemsdijk, W. H. Langmuir 1999, 15, 5942.

13438 Langmuir, Vol. 24, No. 23, 2008

from Figure 1c to d, we have to assume an alteration of hydrogen bonding between the water molecules and the aluminol groups. This alteration must be induced by a change of the surface structure. We determined that the species’ tilt does not vary with pH. For a constant tilt, the band area of a functional group is a measure of its surface concentration. In the series of spectra from Figure 1b to d, we observe that the concentrations of the three lower-frequency species increase with decreasing pH, particularly the concentration of the lowest-frequency species with the largest tilt angle. So far, we do not know the reason and the mechanism of this alteration. The observed phenomenon, however, allows us to explain the change of polar water orientation between pH 6 and 2.7 qualitatively: The oxygens of the aluminol groups with large tilt (see inset of Figure 1d) are well exposed for hydrogen bonding with adjacent water molecules. The oxygens as hydrogen bond acceptors lead to a net dipole moment of the bonded water molecules (black arrow in the inset of Figure 1d), in agreement with the measured signal phase. At medium pH (Figure 1c), we expect that bonds between the hydrogens of aluminol species and neighboring water molecules dominate. The resulting water dipole moment (see scheme in Figure 1c) is again in agreement with the measured phase. The formation of such bonds is less probable at low pH due to the increased concentration of flat oriented OH species with their intrasurface31 hydrogen bonds. The breaking24 of these bonds in order to form mineral/water bonds is energetically unfavorable. It is more favorable for a water molecule at pH 2.7 to bind with the oxygen of a flat oriented OH species (Figure 1d). Theoretical Methods. From the experimental investigations, we have thus obtained a consistent picture of surface speciation and polar water ordering at the molecular level with outstanding richness of details. In order to confirm this result, we performed independent quantum chemical calculations. There are several possible theoretical approaches, by means of either molecular dynamics simulations12,13,31 or ab initio and DFT calculations.49 Here, we use a combination of ab initio and DFT methods. Ab initio methods such as coupled-cluster singles doubles with triplets corrections theory (CCSD(T)), Møller-Plesset perturbation theory of second order (MP2), as well as DFT developed into valuable tools for the understanding of many-body systems. The recent development of these techniques facilitates the efficient treatment of a quite large chemical system. DFT has already been applied to the corundum (001) surface in contact with water31,50,51 but not with emphasis on the determination of the vibrational resonance frequencies of the OH species and their bond orientations. In our theoretical approach, we focused on these experimentally observable quantities. The ab initio and DFT techniques provide common tools for calculating the second derivative matrix of the potential energy and in turn for computing of the vibrational frequencies.52 For the CCSD(T) calculations, we used the MOLPRO53 program package. For the other computations, the TURBOMOLE54 program package was applied. We considered aluminum (hydr)oxide clusters as model systems and fully optimized the structure of these neutral clusters without fixing any degree of freedom using all the different applied theoretical methods. A thorough account of this theoretical study will be published elsewhere.55 (49) Sauer, J.; Ugliengo, P.; Garrone, E.; Saunders, V. R. Chem. ReV. 1994, 94, 2095. (50) Wittbrodt, J. M.; Hase, W. L.; Schlegel, H. B. J. Phys. Chem. B 1998, 102, 6539. (51) Yong, C. W.; Warren, M. C.; Hillier, I. H.; Vaughan, D. J. Phys. Chem. Miner. 2003, 30, 76. (52) Kubicki, J. D. In ReViews in Mineralogy & Geochemistry; Cygan, R. T., Kubicki, J. D., Eds.; The Mineralogical Society of America: Washington, DC, 2001; Vol. 42, p 459.

Flo¨rsheimer et al.

Figure 2. Calculated equilibrium geometry of the Al20O38H12 cluster. The 12 OH species are all doubly coordinated. They differ in their OH bond tilt angle.

In a first step, we did benchmark calculations on these small clusters to assess the accuracy of the DFT method compared with the ab initio techniques for this class of systems. We studied isolated Al7O12H331,51 clusters as well as the associative and dissociative adsorption of water molecules on Al4O650 clusters. All calculations were carried out using the cc-pVDZ basis set. For the DFT computations, we used the Becke-Perdew 86 (BP86) functional throughout. We compared the OH bond lengths, obtained from the different techniques, as well as the OH vibrational frequencies. The CCSD(T) calculations were only possible for the determination of the OH bond lengths of adsorbed water on the Al4O6 cluster. For this system, the MP2 and CCSD(T) results agree very well. The largest deviation for the OH bond lengths is ∆rOH ) 0.1 pm. The discrepancy between the MP2 and DFT results is slightly larger but still small with ∆rOH ) 1.3 pm. For the larger Al7O12H3 cluster, the agreement for the OH bond lengths is excellent as well, with the largest deviation of ∆rOH ) 1.8 pm between the MP2 and DFT results. For the frequencies, we get an outstanding agreement between the MP2 and DFT data with a rather small maximal deviation of ∆VOH ) 18 cm-1 for all the three test systems. This part of our investigation confirms that DFT with BP86 is an appropriate tool to describe the regarded system with sufficient accuracy. Another important result is that using the rather small cc-pVDZ basis set in combination with the BP86 functional allows for an accurate description of the systems under consideration. Not having to use larger basis sets, such as aug-cc-pVDZ or ccpVTZ, reduces the computing time for large systems considerably. In a second step, the sizes of the clusters were increased successively by joining several units together in order to approximate the mineral surface. The largest cluster calculated (Al20O38H12, Figure 2) is treated at the DFT level only. For more (53) Werner, H.-J.; Knowles, P. J.; Lindh, R.; Manby, F. R.; Schu¨tz, M. MOLPRO, Version 2006.1, a package of ab initio programs, 2006; University College Cardiff Consultants Limited: Cardiff, Wales, U.K. (http://www. molpro.net). (54) Ahlrichs, R.; Ba¨r, M.; Ha¨ser, M.; Horn, H.; Ko¨lmel, C. Chem. Phys. Lett. 1989, 162, 165–169. (55) Polly, R.; Schimmelpfennig, B.; Flo¨rsheimer, M.; Kruse, K.; Abdelmonem, A.; Klenze, R.; Rauhut, G.; Fangha¨nel, T. Submitted to J. Chem. Phys.

Hydration of Mineral Surfaces

Langmuir, Vol. 24, No. 23, 2008 13439

to the surface. The results indicate that two water populations occur, one with its net dipole pointing to the surface and the other one with its net dipole pointing oppositely. The different polar signs are due to different bonding with flat and steeper oriented aluminol species which act as hydrogen bond acceptors and donors, respectively, confirming our experimental results.

Comparison of the Experimental and Theoretical Results with Results of Monte Carlo Simulations

Figure 3. Comparison of measured and theoretical spectra. The calculated spectrum (red line) was obtained from the theoretical resonance frequencies and tilt angles for the OH species of the Al20O38H12 cluster. Since the theory considers a neutral cluster without water, the experimental spectrum (×) of Figure 1c (pH 6.0, small water contribution) is used for comparison. The experimental spectrum is corrected for the background contributions. The agreement of the two spectra is excellent, confirming that the cluster is an appropriate model for the corundum (001) surface, and DFT provides reliable information. The agreement corroborates also that our interpretation of the experimental spectrum (doubly coordinated OH species with different tilt) is correct.

details of the calculations, the interested reader is referred to the Supporting Information. Theoretical Results and Comparison with the Results of the Experiments. The Al20O38H12 cluster in Figure 2 exhibits 12 doubly coordinated OH groups with different tilt angles. The results of the calculations show that increasing tilt angles are correlated with decreasing resonance frequencies just as observed experimentally. A simulated spectrum using the theoretical resonance frequencies and tilt angles from a DFT(BP86)/ccpVDZ calculation is given in Figure 3. The characteristic shape of the theoretical spectrum with two maxima of different magnitudes is due to the vibrational distribution of the calculated resonance frequencies and due to the fact that the efficiency of signal generation depends on the species’ tilt relative to the polarizer orientations of the simulated experiment. The details of the calculations are given in the Supporting Information. Since the computation considers a neutral cluster without water, we compare the theoretical spectrum with the spectrum measured at pH 6 (Figure 1c) which exhibits only a small water contribution. The agreement of the two spectra in Figure 3 is excellent if one considers that two totally different methods have been applied. The theory supplies independent, strong evidence for the experimental results. We additionally investigated Al20O38H12 clusters with up to seven water molecules on top. These calculations aim to determine the different possible orientations of the water molecules close

Very recently, the effect of varying tilt angles of OH species at model surface lattices was systematically studied in Monte Carlo simulations24 as mentioned above. At medium tilt angles, the species were observed to act as acceptors and donors. Flat species, however, exhibited a strong preference for serving as acceptors, leading to a preferential polar water orientation in agreement with our studies. Steep OH groups were observed to form no bonds as acceptors, which is also consistent with our results. For the steepest species, the simulations additionally revealed a small probability for forming hydrogen bonds as donors although the hydrogens are well exposed to the water half space. This result seems contradictory to the widely accustomed expectation that a polar species should be hydrated. The simulation, however, shows that it is energetically most favorable for a water molecule near the interface to contribute to the hydrogen bonding network with other water molecules in this case instead of forming a bond with a steep oriented functional OH group. In our spectra, we can directly observe this predicted phenomenon. The small width of the band from the steepest aluminol species at 3690 cm-1 is a clear sign25 for the deficiency of hydrogen bonds.

Conclusion We determined the coordination structure and geometry of the functional groups at a mineral surface. In a large pH range around the pzc of the corundum (001)/water interface, mineral/ water interaction is controlled by hydrogen bonding. We give the first direct experimental evidence that the geometry of the functional species is crucial for the preferential water order at the interface. The experimental results agree with results from quantum theory and from recent computer simulations. A strong impact of this study for geochemical transport modeling and for the understanding of hydrophilic hydration is expected. Acknowledgment. Financial support from the German Research Foundation is gratefully acknowledged. We thank M. Plaschke for investigating the surface quality of the sapphire samples by atomic force microscopy. Supporting Information Available: Details of the experimental and theoretical methods. This material is available free of charge via the Internet at http://pubs.acs.org. LA801677Y