J. Phys. Chem. C 2007, 111, 18279-18287
18279
Study of the Structure of OH Groups on MgO by 1D and 2D 1H MAS NMR Combined with DFT Cluster Calculations Ce´ line Chizallet,*,† Guyle` ne Costentin,† He´ le` ne Lauron-Pernot,† Michel Che,†,‡ Christian Bonhomme,§ Jocelyne Maquet,§ Franc¸ oise Delbecq,| and Philippe Sautet| Laboratoire de Re´ actiVite´ de Surface, UMR 7609-CNRS, UniVersite´ Pierre et Marie Curie-Paris 6, 4 Place Jussieu, 75252 Paris Cedex 05, France, Institut UniVersitaire de France, 103 BouleVard Saint Michel, 75005 Paris, France, Laboratoire de Chimie de la Matie` re Condense´ e, UMR 7574-CNRS, UniVersite´ Pierre et Marie Curie - Paris 6, 4 Place Jussieu, 75252 Paris Cedex 05, France, and UniVersite´ de Lyon, Institut de Chimie de Lyon, Ecole Normale Supe´ rieure de Lyon and CNRS, 46 Alle´ e d’Italie, 69364 Lyon Cedex 07, France ReceiVed: September 4, 2007
Complex 1H MAS NMR spectra of hydroxylated MgO powders have been assigned by combining DFT embedded cluster calculations and experiments using single pulse, Hahn-echo, and 2D NOESY like sequences. Chemical shifts calculations suggest the qualitative classification of protons into three main categories, characterized by different chemical shifts ranges. The highest chemical shifts (δH > -0.7 ppm) are proposed to be characteristic of hydrogen-bond donor OH groups (threefold O3C-H, fourfold O4C-H, and fivefold O5C-H localized on corners, edges, and in valleys respectively). The lowest chemical shifts (δH < -0.7 ppm) are associated to isolated and hydrogen-bond acceptor twofold O2C-H and onefold O1C-H, whereas the central signal at δH ) -0.7 ppm would correspond to isolated O3C-H and O4C-H on kinks and divacancies. These assignments can be refined by considering dipolar interactions between vicinal protons observed thanks to the NOESY like sequence. It is thus shown that some hydrogen bond donor OH groups are characterized by a lower chemical shift than expected from calculations and also contribute to the central signal. Calculated thermal stabilities and chemical shifts suggest that these protons correspond to O4C-H on monatomic steps. The final assignment is fully consistent with previous experimental results on CD3OH adsorption and quantitative analysis of the evolution of spectra with temperature. This study illustrates the synergism between experiments and theory, by comparison with the results obtained by either one.
1. Introduction The control of the interaction of gaseous water with basic oxides is of prevailing importance in surface science, corrosion and heterogeneous catalysis, because hydroxyls obtained by water dissociation exhibit a specific reactivity in several basecatalyzed reactions.1-8 Due to its simple rocksalt structure, MgO can be considered as a model basic oxide, which the study of is useful to rationalize the role of ion coordination on basicity. As a matter of fact, catalysts used as high surface area powders exhibit a variety of surface irregularities (corners, steps, kinks, vacancies, etc.). For each kind of defect, two kinds of OH groups are generally considered upon water dissociation: MgLC-OH, resulting from hydroxylation of Mg2+LC (LC for low coordination, L being the coordination number of magnesium) and O2-L′C-H produced by protonation of oxide ions O2-L′C (L′: coordination number of oxygen). The variety of OH groups obtained on such hydroxylated MgO powders explains the complexity of their spectroscopic features. Infrared has been widely used in the past, and several models have been proposed to assign the corresponding spectra,9-14 but the dependence of O-H stretching frequency * Corresponding author. Present address: IFP-Lyon, Direction Catalyse et Se´paration, BP3, 69390 Solaize, France. Phone: +334 78 02 55 42. Fax: +331 78 02 20 66. E-mail:
[email protected]. † Laboratoire de Re ´ activite´ de Surface. ‡ Institut Universitaire de France. § Laboratoire de Chimie de la Matie ` re Condense´e. | Ecole Normale Supe ´ rieure de Lyon and CNRS.
on coordination is far from simple, as evidenced by our recent theoretical study of vibrational properties of OH groups on MgO.15 Moreover, there is a need to quantitatively characterize these hydroxyls, which could give a complementary picture of the hydroxylated surface. 1H MAS NMR appears to be a good tool but has been scarcely applied to MgO so far, probably because of the need for high-temperature pretreatment to remove carbonates from the surface. We recently proposed16 a specific procedure to obtain well-resolved 1D 1H NMR spectra of hydroxylated MgO powders free from carbonates, with improved signal-to-noise ratios compared to previous studies.17-19 The 1H NMR spectrum of MgO powder is composed of at least 6 lines, at +1.2, +0.7, 0.0, -0.4, -0.7, and -1.8 ppm.16 Increase of the dehydroxylation temperature leads to a sharpening of the spectrum, concomitantly with a decrease of the intensity of the lines at δH > -0.7 ppm and δH ) -1.8 ppm. A first attempt to assign the lines observed was performed on the basis of deuterated methanol (CD3OH) adsorption to avoid generating MgLC-OH groups. It was concluded that the latter give the lowest chemical shifts, whereas the higher ones are related to OL’C-H, obtained upon protonation of O2-L′C ions. As shown previously, the vibrational properties of OH groups on MgO can be accurately reproduced by using a set of theoretical systems,15 modeling several kinds of irregularities representative of the real MgO surface. The irregularities modeled are summarized in Figure 1. The nature and thermal stability of hydroxyl groups produced by water dissociation on these irregularities have been reported earlier.20 The role of
10.1021/jp077089g CCC: $37.00 © 2007 American Chemical Society Published on Web 11/09/2007
18280 J. Phys. Chem. C, Vol. 111, No. 49, 2007
Chizallet et al. TABLE 1: Specific Surface Areas and [O2-3C]/[O2-4C] Ratios of MgO Samplesa sample
BET surface area (m2.g-1)
R‚[O2-3C]/[O2-4C]
MgO-precipitation MgO-hydration MgO-sol-gel
167 150 150
0.06 0.06 0.15
a
Figure 1. Schematic representation of irregularities on the MgO surface (O2-: red spheres, Mg2+: green spheres), adapted from Figure 11 of ref 21. The terminologies used for the periodic and cluster systems are given in brackets, see section 3.3. and ref 20.
hydrogen bonding, oxygen coordination, and local topology on anharmonic frequencies was stressed, leading to a new assignment of IR spectra consistent with experimental observations. The extension of such a method to solid state 1H NMR is considered in the present work, which has never been proposed for MgO in the literature so far, to the best of our knowledge. One difficulty is the narrow range of chemical shifts observed (3 ppm): our goal is thus to obtain a qualitative analysis of the trends, rather than an absolute quantification of the nuclear properties of each kind of proton. Spectral resolution and information on dipolar coupling in complex 1H MAS NMR spectra of inorganic materials can be obtained thanks to fast MAS techniques combined with double quantum (DQ) excitations.22 However, for sensitivity concerns, the very low proton content of our samples (hydroxyls exist essentially on the surface) prevents the use of 2.5 mm diameter rotors and consequently of the high rotation speed required for the BABA (back to back) sequence23 for example. On the other hand, the NOESY (nuclear Overhauser effect spectroscopy) like sequence is routinely used at lower rotation speed to map direct dipolar coupling or chemical exchange in complex organic molecules. Due to extended use of DQ techniques, NOESY is quite scarcely used to probe spatial proximity of hydroxyl groups on inorganic materials, except when motional averaging processes prevent the observation of off-diagonal correlation in 2D DQ experiments.24 In the absence of chemical exchange, the main interest of such a sequence consists in the identification of close protons (involved in dipolar interaction or spin diffusion), which is of first importance in our study in which hydroxyls obtained by dissociation of the same water molecule are thus likely to be localized in the same region of the MgO surface. The present study thus aims at precisely assigning 1H NMR spectra of hydroxylated MgO powders, by a combined theoretical and experimental approach. Chemical shifts are calculated for each proton on irregularities modeled by a cluster approach. Experimentally, several samples of various morphologies are synthesized so as to vary the relative proportions of ions in low coordination and thus the distribution of OH groups on hydroxylated surfaces. Carbonate-free surfaces with variable degrees of hydroxylation are studied. A deconvolution of the NMR spectrum is proposed by using a set of experimental NMR techniques (single-pulse, Hahn-echo, and NOESY like) in line with the theoretical suggestions. 2. Experimental Section 2.1. MgO Samples Preparations. As reported earlier,25 MgO samples were prepared by thermal decomposition of Mg(OH)2 precursors, obtained as follows:
R is an undetermined constant.
(i) by adding NH3 solution (Carlo Erba, 30%) to a solution (0.1 M) of Mg(NO3)2 (Aldrich, 99.995+%). (ii) by hydration of commercial MgO (Aldrich, 99.99+%). (iii) by the sol-gel method from magnesium methanolate (Mg: Aldrich, 99.5%; methanol: Acros Organics, H2O < 0.005%). The hydroxide samples were then decomposed in vacuum (0.133 Pa) up to 1273 K (ramp 1 K.min-1) and kept at this temperature for 2 h. For the sol-gel synthesis (iii), removal of organic surface residues was performed by subsequent treatment under static oxygen (13.3 × 102 Pa, 673 K, 30 min). The complete transformation into the MgO periclase structure was checked by X-ray diffraction, and the specific surface areas were evaluated by the Brunauer-Emmett-Teller method. The resulting MgO samples are labeled as follows: (i) MgO-precipitation; (ii) MgO-hydration; (iii) MgO-sol-gel. Table 1 gathers specific surface areas and [O2-3C]/[O2-4C] ratios obtained by photoluminescence spectroscopy. In spite of similar specific surface areas, various morphologies were evidenced by TEM and photoluminescence spectroscopy.8,25 2.2. Samples Pretreatment. As reported earlier,16 carbonatefree and hydroxylated surfaces were prepared as follows: (i) removal of carbonates to obtain a clean surface by treating 125 mg of MgO powder in flowing nitrogen (20 cm3‚min-1) in a “U shape” quartz cell equipped with two valves, at 1023 K (ramp 1.5 K‚min-1) maintained for 1 h. (ii) water adsorption on the clean surface cooled at 373 K, by flushing the powder 10 min with water vapor (840 Pa) diluted in nitrogen (20 cm3‚min-1). (iii) control of the hydroxyl content by heating of the sample at the desired temperature Tt (473-873 K, ramp 5 K.min-1) maintained for 2 h, under flowing nitrogen (20 cm3‚min-1). 2.3. 1H NMR Spectra Recording. The MgO pretreated sample was transferred into a glove box where 4 mm zirconia rotors were filled and closed with Kel-F caps, to avoid exposure of the sample to ambient air. Solid state 1H MAS NMR spectra were recorded at room temperature with a Brucker Avance 400 spectrometer at 9.4 T, equipped with a 4 mm probe and with a spinning rate of 12.5 kHz. Three types of sequences were then used: (i) A 90° single pulse of 3.3 µs. (ii) A Hahn-echo (90°-τ-180°-τ) sequence with 16 phasecycling procedure, to suppress the probe signal. τ was fixed to 80 µs (1/(12.5 kHz)) for the “standard procedure”. It was also varied between 80 and 800 µs (values of n/(12.5 kHz), n being an integer) in an attempt to look for a quantitative deconvolution of the spectra, for what will be called the “quantification procedure”. For sequences (i) and (ii), 128 free induction decays per spectrum were accumulated, except for the Hahn-echo procedure with variable τ for which this number was reduced to 64 to avoid long recording duration. A spectral window of 250 kHz was used. Deconvolution of the signal was performed thanks to the DMFIT program.26
Structure of OH Groups on MgO
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18281
(iii) The 2D NOESY experiment (90°-t1-90°-tm-90°) was performed to visualize the spin transfer induced by the proximity between protons, for two pretreatment temperatures Tt ) 473 and 873 K. 128 free induction decays were accumulated for each slice (given t1 value). 210 slices were acquired, with an increment time of 80 µs, which allows a good compromise between accuracy and time. Two values for the mixing time tm were investigated: 1 and 100 ms. The states phase recycling procedure was used. For each sample and for each pretreatment temperature, T1 was estimated using a saturation recovery sequence, and found in the range 2 s < T1 < 10 s. A recycling delay of 5T1 was then used for each sequence and each sample, to allow complete spin relaxation before signal recording. The receiver gain remained constant throughout the experiments. Chemical shifts were determined by reference to an external tetramethylsilane (TMS) sample. The signal of the empty rotor was recorded in the same conditions for each experiment. Some trials were performed at 700 MHz and did not lead to significant improvement of the resolution of 1D spectra: the lines remain quite broad, and no new well-defined peak appears. For practical reasons, the 400 MHz approach was thus preferred: such a combined 1D and 2D study would have not been performed in a reasonable time within the poor access to very high field spectrometers. Moreover, conventional single pulse NMR sequence (i) was preferred to the CRAMPS methods due to the loss of quantitative aspects with the latter. 3. Computational Details 3.1. Methods. Embedded cluster calculations were performed at the DFT level, with the B3LYP hybrid exchange correlation functional,27,28 within the Gaussian 03 code.29 It was indeed previously shown on the clusters used that B3LYP leads to a better description of hydrogen bonds than with GGA,15 which is crucial for the prediction of vibrational and NMR properties of OH groups. As explained earlier,15 the structure of the embedded clusters used was deduced from periodic calculations and further partial geometry optimization, using the 6-311+G** basis set. The Mg LANL2 effective core potential was used for third-order neighbors of hydroxyl groups (magnesium atoms at the border of the clusters noted Mg*) so as to prevent their artificial polarization by the embedding species,30-34 composed of an array of point charges (from 913 to 2100, see ref 15). 3.2. Chemical Shifts Calculations. Chemical shifts were evaluated within the GIAO method,35 with the IGLO-II basis set for oxygen and hydrogen.36 Such a specific basis set is not provided for magnesium, for which the 6-311+G** basis set was thus maintained to perform the evaluation of nuclear properties. The reader is referred to Supporting Information S1 for the evaluation of the effect of exchange-correlation functionals and basis sets on 1H chemical shift values. Chemical shifts are given by reference to TMS (tetramethylsilane), modeled as an isolated molecule. The calculated value of the 1H screening constant (σTMS ) 31.9 ppm) is in very good agreement (within 1%) with other reported computational results37,38 and with the experimental value for gas-phase TMS.39 However, TMS is used experimentally in the liquid state. To the best of our knowledge, no theoretical work has been performed on liquid TMS, and the experimental value of σTMS(liq.) is so far unknown.37 Some authors, using the GIPAW method40-44 bypass this problem by plotting the calculated screening constants as a function of the corresponding experimental chemical shifts, constraining the slope to unity, which enables the determination of σref. This approach is however
Figure 2. Hydroxylated clusters depicting (a) S1 (monatomic step), (b) K-Mg4C-O3C (O2-3C terminated kinks). The coordination and isolated (i), hydrogen-bond acceptor (a) or donor (d) nature of the OH group is indicated. Green spheres: Mg, blue spheres: Mg* (described by LANL2 core potential during geometry optimization), red spheres: O from MgO, yellow sphere: O from the water molecule, white spheres: H. Outermost species only are depicted and embedding charges are omitted for the sake of clarity.
limited to well-defined experimental chemical shifts, which is not our case. Because results reported in the literature and using σTMS ∼ 31.9 ppm are in satisfactory agreement with experiments,39 this value will be used in what follows. 3.3. Clusters. Clusters used to simulate irregularities of the surface of MgO have been described in a previous study15 and are depicted in Figure 2 and Supporting Information S2. Their size has been carefully optimized to allow a converged description of geometric and vibrational properties. Convergence of 1H NMR properties has also been tested for the smallest cluster Mg7O5H2Mg*3 (modeling Mg2+3C terminated corners, see Supporting Information S1 and next paragraph), which allows a fully converged description of chemical shifts for the hydrogen-bond acceptor OH groups, and an uncertainty of only 0.3 ppm for hydrogen-bond donor OH groups, when compared with the bigger cluster selected to model Mg2+3C terminated corners (Mg7O10H2Mg*10, see Figure S2-1d). Monatomic steps are modeled by the S1 cluster (Mg20O15H2Mg*10) and edges and valleys by the S2-ON (Mg10O11H2Mg*10) and S2-IN (Mg22O10H2) systems. Corners are described by the C-O3C (O2-3C terminated corners, Mg13O14H2Mg*12) and the C-Mg3C (Mg2+3C terminated corners, Mg7O10H2Mg*10) clusters and 3C ions in concave environments20 by a divacancy performed in the edge of the monatomic step, called D-Mg3CO3C (Mg9O11H2Mg*13), and by kinks: K-Mg4C-O3C (O2-3C terminated kinks, Mg19O14H2Mg*9) and K-Mg3C-O4C (Mg2+3C terminated kinks, Mg21O15H2Mg*10). All clusters contain a single dissociated water molecule, so that the “-1w-” terminology used in our previous work15 will
18282 J. Phys. Chem. C, Vol. 111, No. 49, 2007
Chizallet et al.
TABLE 2: Calculated 1H Chemical Shifts (GIAO, B3LYP) for OH Groups Obtained upon Hydroxylation of Mg2+LC Ions (Denoted MgLC-O-H) and upon Protonation of O2-L′C Ions (OL′C-H)a MgLC-O-H system S1 S2-ON S2-IN C-O3C C-Mg3C D-Mg3C-O3C K-Mg3C-O4C K-Mg4C-O3C
type
δ (ppm)
-0.1 O2C-H (a) O1C-H (a) -0.5 O2C-H (a) 0.0 O1C-H (a) -0.7 O1C-H (a) -0.9 O3C-H (i) O3C-H (i) 0.4 O2C-H (i) -0.9
OL′C-H type O4C-H (d) O4C-H (d) O5C-H (d) O3C-H (d) O4C-H (d) 0.3 O4C-H (i) O3C-H (i)
δ (ppm) 5.2 9.0 7.4 6.1 5.4 0.5 0.7
a (a): Hydrogen-bond acceptor, (d): hydrogen-bond donor, (i): isolated OH group.
not be added in what follows. The coordination and geometry of OH groups after water dissociation depend on the nature of the dissociating surface site. Most of the OH groups modeled are hydrogen-bonded. For example, hydrogen-bond donor O4C-H (with oxygen bonded to 4 Mg2+) is obtained upon protonation of O2-4C on edges, the corresponding hydrogenbond acceptor hydroxyls being O1C-H. Isolated OH groups are observed in the case of divacancies and kinks only. In concave surface areas20 (valleys, monatomic steps, kinks, and divacancies), contrary to expectation, O1C-H is not always formed, but hydroxylation of Mg2+LC also provides O2C-H (valleys and monatomic steps, O2-3C terminated kinks) or O3C-H (divacancies and Mg2+3C terminated kinks), when bridging between several Mg2+LC is possible. In particular, water dissociation generates two equivalent O3C-H’s in divacancies (D-Mg3CO3C system). 4. Results 4.1. Chemical Shifts Calculations. Calculated chemical shifts are reported in Table 2. Figure 3a-d illustrates how chemical shifts are linked to O-H bond lengths, anharmonic vibration frequencies (values reported in ref 15), hydrogen-bond lengths, and electrostatic charge on the protons, calculated in the same conditions (same clusters optimized with the 6-311+G** basis set, B3LYP, natural population analysis). Such correlations have been established for several compounds in the literature.45-50 In the case of MgO, some nonmonotonous evolutions can be observed (Figure 3-b): IR bands can gather several types of OH groups, as well as NMR lines, but not with the same distribution. This is due to the fact that vibrational properties of OH groups and nuclear shielding of their protons are not influenced by the same factors or not to the same extent: the vibration essentially depends on the mass of the vibrating atoms while the nucleus properties are influenced by the magnetic moments of the surrounding atoms. This opens very positive perspectives of complementary insight of OH groups on MgO: indeed, IR bands were shown to gather significant number of vibrations of OH groups.15 Different gathering afforded by 1H NMR could allow a better discrimination, what will appear in further sections. 4.2. Experimental 2D NOESY Spectra. Figure 4 shows the 2D NOESY spectrum, with a mixing time tm ) 1 ms, obtained for MgO-sol-gel treated at Tt ) 473 K, and some selected corresponding slices. The Hahn-echo spectrum (“standard procedure”, see section 2.3.) is also shown for reference. On the 2D NOESY spectrum, wiggles observed on the vertical line at -0.7 ppm are due to partial truncation of the signal, which could be solved by acquiring a higher number of
slices but would require a very long experiment time. The 2D map remains poorly informative: no explicit correlation appears due to the quite broad resonances. However, maxima on slices enable dipolar interactions to be visualized clearly and give detailed information on the proximity between different kinds of protons. The most representative ones are depicted in Figure 4. Maxima can be observed in a very broad chemical shift domain, from -4 to +5 ppm, which evidences the high heterogeneity of sites involved. The signal however exhibits a significant intensity only in the range -2 to +3 ppm, in agreement with the mono-dimensional spectrum. Protons characterized by the lowest chemical shifts (δH e -1.2 ppm, Figure 4c,g,h) are in dipolar interaction with those relative to the highest chemical shifts (large and complex signal leading to the smooth decrease of intensity above 0 ppm), and with some of the protons represented by the central part of the signal (-0.8 e δH e -0.4 ppm). The comparison of the shapes of the Hahn-echo spectrum (Figure 4a) and of the slices in Figure 4c,e-h shows that the central signal at -0.6 ppm has only a low intensity, and thus those corresponding OH groups are not in dipolar interaction with any other OH group characterized by a different chemical shift. This is confirmed by the slice in panel d, which exhibits a very sharp central signal at -0.6 ppm without any other maximum. Qualitatively, an enlarged mixing time tm ) 100 ms leads to the same observations, with more intense lines on slices. This evidences enhanced dipolar interactions of the same nature than for tm ) 1 ms. Similar experiments have been performed for the sample pretreated at Tt ) 873 K (spectra not shown). In agreement with what was observed with the Hahn-echo sequence,16 an increase of the pretreatment temperature results in a sharpening of the 2D NOESY spectrum around the central signal at -0.7 ppm and an attenuation of the coupling between protons. The major trends observed at Tt ) 473 K are nevertheless confirmed at Tt ) 873 K, insofar as residual protons at δH e -1.2 ppm still interact with the residual protons at δH g -0.7 ppm, and with a small fraction of the protons (the most numerous ones at Tt ) 873 K) at δH ) -0.7 ppm. Most protons at -0.7 ppm are isolated or interact with other protons at -0.7 ppm. The signal of the rotor at 2.8 ppm16 is also more clearly visible as Tt increases (as the amount of OH groups on the samples vanishes), and 2D NOESY unambiguously evidences that the corresponding protons do not magnetically interact with protons on the MgO surface. 4.3. Quantification of the Various OH Groups. For quantification purposes, the deconvolution of the signal obtained with the single pulse sequence has been compared to the “quantification procedure” consisting of the extrapolation to τ ) 0 of the intensities recorded by the Hahn-echo sequence. The time-consuming Hahn-echo “quantification procedure” leads to enhanced uncertainty on the determination of the extrapolated integrated intensities, in spite of the removal of the probe signal. The deconvolution of the spectrum relative to the single pulse sequence was thus preferred. Details will be given in a forthcoming paper. On the basis of the maxima observed on the 2D NOESY spectra (-1.7, -0.7, -0.3, +0.1 and +1.3 ppm mainly), deconvolution of the 1D spectra can be performed, as shown in Figure 5. The rotor and probe contributions are modeled by Gaussian line shapes and the others by Lorentzian ones. At this stage, it is illusory to try to assign each line to a given type of OH group. Nevertheless, from our previous study,16 confirmed by the 2D NOESY approach, the distinction of three chemical shift domains should be considered : δH > -0.7 ppm,
Structure of OH Groups on MgO
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18283
Figure 3. Calculated chemical shifts, reported as a function of other calculated parameters: (a) O-H bond length, (b) anharmonic vibration frequency, (c) hydrogen-bond length for hydrogen-bond donor OH groups only, (d) electrostatic charge on the proton (natural population analysis).
δH ) -0.7 ppm, and δH < -0.7 ppm, excluding the signals induced by the empty rotor and the probe. The corresponding integrated areas for these three domains are reported in Figure 6 as a function of Tt, in the case of MgO-sol-gel. This diagram confirms that the central signal (δH ) -0.7 ppm) corresponds to the most stable OH groups. It also evidences that the relative contributions of the signals at lowest and highest chemical shifts are not equal, even if they decay rather in parallel. The highest chemical shift component is indeed less intense than the lowest chemical shift component. 5. Discussion 5.1. Parameters Influencing Chemical Shifts. Figure 7 summarizes the parameters influencing the relative calculated values of chemical shifts, deduced from values reported in Table 2. All hydrogen-bond donor OH groups are characterized by larger calculated δH than all other hydroxyls, which shows that hydrogen-bonding plays a major role in the chemical shifts values. In agreement with literature data for other systems,41,45,48,49,51 δH is roughly linearly correlated to hydrogenbond length for donor OH groups (Figure 3c), which can be interpreted as deshielding induced by partial electron transfer from the proton to the hydrogen-bond acceptor OH group. On the contrary, protons belonging to hydrogen-bond acceptor and isolated OH groups are not clearly differentiated, as in the case of vibrational properties.15 The coordination of oxygen, which is often called on to rationalize spectroscopic properties of MgO, does not play a monotonous role in chemical shift values, as shown by the sequence δH(O4C-H, S2-ON) > δH(O5C-H, S2-IN) > δH(O3C-H, C-O3C) > δH(O4C-H, C-Mg3C). Nevertheless, it can be observed that, among isolated and hydrogen-bond acceptor OH groups, all O1C-H and O2C-H exhibit lower chemical shifts than all of the other ones, namely O3C-H and O4C-H. Local topology, which plays a leading role
with regard to the thermal stability of OH groups20 and had to be invoked to rationalize their vibrational properties,15 does not monotonically influence the chemical shifts. As usually observed in the literature for acidic OH groups,48,50,52 the chemical shift tends to increase with the electrostatic charge localized on the proton (Figure 3d). The dispersion observed however shows that other parameters (hydrogen-bonding and coordination) influence δH and qH to a different extent. 5.2. Comparison of Experimental and Calculated Chemical Shifts Ranges. Calculated chemical shifts belong to the -0.9 to +9 ppm range, which is found to be larger than the experimental one. The noncompletely accurate description of hydrogen-bonding by DFT can explain the too high values of chemical shifts calculated for hydrogen-bond donor OH groups. DFT indeed overestimates the electronic delocalization in the hydrogen bond, leading to an excessive positive charge and deshielding for the proton of the hydrogen-bond donor hydroxyls. This can be linked to their too low calculated vibrational frequency than expected from experiments, as extensively discussed in our previous work.15 Hybrid functionals (B3LYP in the present case) lead to better results compared to GGA (see Supporting Information S1), but this is still not enough to fit experiments perfectly. Another reason which could explain the overestimation of δH for hydrogen-bond donor hydroxyls is the description of surface defects by clusters: the partial relaxation (only neighboring ions of OH groups are relaxed) could lead to some uncertainty in δH calculations. Calculations thus lead to the following first proposal for the assignment of the three domains distinguished in the 1H NMR spectra: (i) hydrogen-bond donor OH groups are characterized by higher values for chemical shifts. (ii) hydrogen-bond acceptor and isolated O1C-H and O2C-H groups give lower ones (negative chemical shifts). (iii) the central part of the spectrum can be associated to
18284 J. Phys. Chem. C, Vol. 111, No. 49, 2007
Chizallet et al.
Figure 4. 1H NMR experimental spectra of MgO-sol-gel treated at Tt ) 473 K: (a) Hahn-echo spectrum; (b) 2D NOESY with 1 ms mixing time; (c-h) horizontal slices extracted from (b), the position of the diagonal being depicted by a bold arrow.
Figure 5. 1H MAS NMR mono-dimensional spectrum (single-pulse sequence) of MgO-sol-gel treated at Tt ) 673 K, and corresponding deconvolution.
Figure 6. Integrated intensities deduced from the deconvolution of single-pulse 1D 1H MAS NMR spectra of MgO-sol-gel, as a function of pretreatment temperature Tt.
isolated O3C-H and O4C-H in the vicinity of kinks and divacancies. Calculated chemical shifts in this case (∼ +0.5 ppm) should be related with the central peak in the experimental spectrum (-0.7 ppm). 5.3. Adsorption of Protic Molecules Other than Water. As mentioned in ref 16, adsorption of CD3OH instead of water enables us to distinguish between the contributions of MgLCOH (O1C-H, O2C-H, and isolated O3C-H produced by hydroxylation of Mg2+LC), not expected with CD3OH, and of
OL’C-H (isolated and hydrogen-bond donor O3C-H, O4C-H, and O5C-H obtained by protonation of O2-L’C), resulting from water and CD3OH adsorptions. Expected spectra are depicted in Figure 8, assuming that changing the nature of the anionic counterpart of the proton (HO- for water versus CD3O- for deuterated methanol) does not affect the position of the lines. The disappearance of the lowest components of the spectrum (δH < -0.7 ppm) and of part of the central signal (δH ) -0.7 ppm) is predicted, which perfectly fits the experimental results
Structure of OH Groups on MgO
Figure 7. Diagram summarizing relevant parameters influencing calculated 1H NMR chemical shifts. The axis is not scaled for practical reason.
Figure 8. (a) 1H NMR calculated lines positions on MgO upon adsorption of water; (b) estimated lines for adsorption of CD3OH.
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18285 shifts. Diminution of dipolar interactions may also explain the sharpening. This hypothesis however cannot be checked by the present work since the theoretical study of the influence of water coverage would require excessively large clusters. The central signal is experimentally the most stable, as especially evidenced by the corresponding curve on Figure 6. Calculations suggest that this effect can be assigned to the remaining OH groups on kinks and divacancies. These facts confirm the first proposal of assignment of 1H NMR spectra given in section 5.2. However, this first proposal does not explain the deficit of protons characterized by the highest chemical shifts compared to those characterized by the lowest ones (Figure 6). This deficit cannot be explained by the O2C-H of O2-3C terminated kinks only (characterized by δH < -0.7 ppm but not corresponding to any donor OH group), insofar as it is higher on MgOprecipitation (60%) than on MgO-sol-gel (30%) which contains much more irregularities, such as kinks (Table 1). On the other hand, the contribution of some hydrogen-bond donor OH groups to the central signal at -0.7 ppm, which calculated chemical shifts are overestimated by DFT (see section 5.2.), can be called on to explain the deficit. As the integrated intensity of the central line remains nearly constant in the 673-873 K range, the hydrogen-bond donor hydroxyls contributing to it may be the most stable of them, and also the one characterized by the lowest chemical shifts among donors: O4C-H belonging to monatomic steps follows these two trends. If our assignment is correct, it can be noted a posteriori that AK,D, the integrated area defined as
AK,D ) Aδ) -0.7 ppm - (Aδ< -0.7 ppm - Aδ> -0.7 ppm) (1)
Figure 9. Evolution of 1H NMR lines positions of hydroxylated MgO hydroxylated irregularities as a function of temperature, deduced from first principles calculations. Intensities of the lines are arbitrarily normalized.
published previously.16 The comparison of those experimental and theoretical results thus evidence that O1C-H and O2C-H (obtained in the case of water only) are responsible for the signal at δH < -0.7 ppm, whereas other OH groups are responsible for the highest chemical shifts (δH g -0.7 ppm), in line with our first proposal (section 5.2.). 5.4. Thermal Stability of OH Groups, Influence of Morphology. The first principles data on thermal stability of OH groups20 can be combined to the calculated values of chemical shifts (Table 2) to deduce the theoretical evolution of the NMR spectra with temperature. Since thermodynamics was calculated in different pressure conditions (P ) 10-2 Pa) than experiments (“dry” nitrogen flow, water pressure low but unknown), the calculated results reported in Figure 9 can be used only to qualitatively compare our theoretical to real experimental evolution of the system versus temperature. Most of the experimental observations are taken into account by the theoretical model. Indeed, the sharpening of the signal by increasing temperature16 is well reproduced by the recombination and desorption of hydrogen-bond donor and acceptor OH groups responsible for the highest and the lowest chemical
quantifies the amount of Mg2+3C kinks and divacancies, where Aδ) -0.7 ppm, Aδ< -0.7 ppm, and Aδ> -0.7 ppm stand for the integrated areas of the central resonance and the lateral ones, respectively. O2-3C terminated kinks are not included in AK,D as the intensity relative to their O2C-H (contributing to Aδ< -0.7 ppm) is expected to compensate the one of their O3C-H (contributing to Aδ) -0.7 ppm). Values of AK,D for the different samples treated at Tt ) 673 K increase with the irregularity of the sample (43 a.u. for MgO-precipitation, 84 for MgO-hydration, 115 for MgO-sol-gel), confirming that part of the central signal is due to kinks and divacancies. 5.5. Proximity between Protons. NOESY experiments (section 4.2.) confirm that protons responsible for the highest and the lowest chemical shifts are in dipolar interaction, which can be explained by the proximity induced by hydrogenbonding. Moreover, the interaction visualized between protons at lower chemical shifts and those in the central signal (-0.7 ppm) can be rationalized by the following: (i) the O2-3C terminated kink system (O2C-H: δH < -0.7 ppm, O3C-H: central signal around -0.7 ppm). (ii) the proximity of some hydrogen-bond donors and acceptors with isolated OH groups that can be expected, as in the case of the sample pretreated at Tt ) 473 K, as hydroxylation of kinks and divacancies is associated to that of nearby monatomic steps. This could explain the interaction of a few hydroxyls at δH ) -0.7 ppm with other kinds of OH groups. (iii) the contribution of hydrogen-bond donor O4C-H of monatomic steps to the central signal at -0.7 ppm, which counterpart (hydrogen-bond acceptor O2C-H) is characterized by δH < -0.7 ppm, as explained in section 5.4. NOESY experiments thus enable to refine the assignment proposed on the basis of first principles calculations in line with the quantification of experimental 1D spectra (section 5.4.). This
18286 J. Phys. Chem. C, Vol. 111, No. 49, 2007 2D technique applied to MgO enables a detailed mapping of hydrogen-bonding on the surface. 5.6. Synopsis: Assignment of 1H NMR Spectra of Hydroxylated MgO Powder. Our combined experimental and theoretical approach finally leads to the final assignment of 1H NMR spectra of hydroxyls on MgO: (i) δH > -0.7 ppm: most hydrogen-bond donor OH groups. (ii) δH ) -0.7 ppm: isolated O3C-H and O4C-H on kinks and divacancies, and some hydrogen-bond donor hydroxyls, likely O4C-H on monatomic steps. (iii) δH < -0.7 ppm: hydrogen-bond acceptor and isolated O1C-H and O2C-H groups. As expected from the analysis of correlations between calculated chemical shifts and anharmonic frequencies (section 4.1.), the assignments of IR and 1H NMR spectra are not strictly parallel. Hydrogen-bonding (donor or acceptor nature) is the main parameter governing frequencies and chemical shifts, but the latter are not so strongly dependent upon local topology (concave or convex nature of the surface20) as the former. This can be explained by the fact that NMR probes the environment of protons, whereas IR is sensitive to the properties of the whole O-H vibrator. This reduced influence of local topology reveals a stronger dependence on oxygen coordination for NMR, as the lowest coordinated OH groups are responsible for the lowest chemical shifts.
Chizallet et al. and chemical shifts suggest that these protons correspond to O4C-H on monatomic steps. This combined experimental and theoretical approach finally leads to the following assignment of 1H NMR spectra: δH > -0.7 ppm is due to most of the hydrogen-bond donor OH groups, δH ) -0.7 ppm characterizes isolated O3C-H and O4C-H on kinks and some stable hydrogen-bond donor, likely O4C-H on monatomic steps, whereas δH < -0.7 ppm involves isolated and hydrogen-bond acceptor O1C-H and O2C-H. This assignment is fully consistent with experimental results on CD3OH adsorption, insofar as it enables us to explain the loss of the lowest part of the signal, due to O1C-H and O2C-H not obtained from CD3O-H dissociation. The evolution of the shape of the spectra with temperature, and the morphology dependence on protons quantification are also rationalized. This study illustrates that combining experiments and theory leads to a beneficial synergism between both approaches, by comparison with the results obtained by either one. Indeed, cluster modeling has provided a first molecular insight in the relationship between structure and chemical shift, which could have never been reached by the single experimental approach. In return, the theoretical assignment has been refined thanks to complementary experimental aspects suggested by calculations. The final assignment provides a new quantitative tool to characterize the various types of hydroxyl groups on MgO, which will be most useful to study their respective contribution in some specific catalytic processes.
6. Conclusion 1H
MAS NMR spectra of hydroxylated MgO powders free from carbonates are composed of several lines, which behavior with dehydroxylation temperature differs. The central line at -0.7 ppm is the most stable whereas lateral lines decay in intensity with increasing temperature. These lines have been assigned by combining DFT cluster calculations with the B3LYP hybrid functional and experiments using various sequences: Hahn-echo (qualitative aspect), single pulse (quantitative aspect), and 2D NOESY like (visualization of proximity between protons). Clusters representative of hydroxylated edges, monatomic steps, corners, kinks, and divacancies have been used to model irregularities the MgO surface. Chemical shifts evaluated with the GIAO and the B3LYP functional lead to three main domains, suggesting that the highest chemical shifts (δH > -0.7 ppm) would be characteristic of hydrogen-bond donor hydroxyls, the lowest chemical shifts (δH < -0.7 ppm) to isolated and hydrogen-bond acceptor O1C-H and O2C-H and the central signal at -0.7 ppm to isolated O3C-H and O4C-H on kinks and divacancies. The comparison of experimental and calculated chemical shift ranges indicates that DFT overestimates δH of hydrogen-bond donor OH groups because of excessive deshielding. This artifact inherent to the simulation methods used has been unraveled by our combined 1D-2D NMR experimental approach. Indeed, NOESY experiments not only confirm the dipolar interaction between protons characterized by the highest and the lowest chemical shifts, thus associating hydrogen-bond donor and acceptor OH groups, but also reveal that some protons relative to the central line (-0.7 ppm) are close to protons at lower chemical shifts. This coupling, together with the quantification of protons belonging to each chemical shift range (δH > -0.7 ppm, δH ) -0.7 ppm, δH < -0.7 ppm) at increasing pretreatment temperature, suggest that some of the most stable donor OH groups are characterized by chemical shift lower than that expected from calculations. Calculated thermal stabilities
Acknowledgment. Thierry Azaı¨s and Guillaume Laurent, from the Laboratoire de Chimie de la Matie`re Condense´e, and Bruno Alonso and Domininique Massiot, from Centre de Recherche sur les Mate´riaux a` Haute Tempe´rature (CNRS, Orle´ans, France) are gratefully acknowledged for very fruitful discussions on experimental aspects. Most of the calculations were carried out at the IBM-SP at the CNRS-IDRIS computational centre, under project No. 051847. Supporting Information Available: Evaluation of the effect of exchange-correlation functionals and basis sets on 1H chemical shift values, and structures of hydroxylated clusters. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Zhang, G.; Hattori, H.; Tanabe, K. Appl. Catal. 1988, 36, 189197. (2) Prinetto, F.; Tichit, D.; Teissier, R.; Coq, B. Catal. Today 2000, 55, 103-116. (3) Rao, K. K.; Gravelle, M.; Valente, J. S.; Figueras, F. J. Catal. 1998, 173, 115-121. (4) Climent, M. J.; Corma, A.; Iborra, S.; Velty, A. J. Mol. Catal. A 2002, 182/183, 327-342. (5) Choudary, B. M.; Kantam, M. L.; Reddy, C. R. V.; Rao, K. K.; Figueras, F. J. Mol. Catal. A 1999, 146, 279-284. (6) Wang, J. A.; Bokhimi, X.; Novaro, O.; Lopez, T.; Gomez, R. J. Mol. Catal. A 1999, 145, 291-300. (7) Kus, S.; Otremba, M.; Torz, A.; Taniewski, M. Fuel 2002, 81, 1755-1760. (8) Bailly, M. L.; Chizallet, C.; Costentin, G.; Krafft, J. M.; LauronPernot, H.; Che, M. J. Catal. 2005, 235, 413-422. (9) Anderson, P. J.; Horlock, R. F.; Olivier, J. F. Trans. Farad. Soc. 1965, 61, 2754-2762. (10) Shido, T.; Asakura, K.; Iwasawa, Y. J. Chem. Soc., Faraday Trans. 1 1989, 85, 441-453. (11) Coluccia, S.; Marchese, L.; Lavagnino, S.; Anpo, M. Spectrochim. Acta 1987, 43A, 1573-1576. (12) Coluccia, S.; Lavagnino, S.; Marchese, L. Matter. Chem. Phys. 1988, 18, 445-464. (13) Morrow, B. A. Stud. Surf. Sci. Catal. 1990, 57, 161-222. (14) Kno¨zinger, E.; Jacob, K. H.; Singh, S.; Hofmann, P. Surf. Sci. 1993, 290, 388-402.
Structure of OH Groups on MgO (15) Chizallet, C.; Costentin, G.; Che, M.; Delbecq, F.; Sautet, P. J. Am. Chem. Soc. 2007, 129, 6442-6452. (16) Chizallet, C.; Costentin, G.; Lauron-Pernot, H.; Macquet, J.; Che, M. Appl. Catal. A 2006, 307, 239-244. (17) Aramendia, M. A.; Benitez, J. A.; Borau, V.; Jimenez, C.; Marinas, J. M.; Ruiz, J. R.; Urbano, F. Langmuir 1999, 15, 1192-1197. (18) Aramendia, M. A.; Benitez, J. A.; Borau, V.; Jimenez, C.; Marinas, J. M.; Ruiz, J. R.; Urbano, F. J. Solid State Chem. 1999, 144, 25-29. (19) Aramendia, M. A.; Benitez, J. A.; Borau, V.; Jimenez, C.; Marinas, J. M.; Ruiz, J. R.; Urbano, F. Colloids Surf. A 2000, 168, 27-33. (20) Chizallet, C.; Costentin, G.; Che, M.; Delbecq, F.; Sautet, P. J. Phys. Chem. B 2006, 110, 15878-15886. (21) (a) Che, M.; Tench, A. J. AERE Rep. 1980, R-9971, 1-55. (b) Che, M.; Tench, A. J. AdV. Catal. 1982, 31, 77-133. (22) Schnell, I.; Spiess, H. W. J. Magn. Reson. 2001, 151, 153-227. (23) Feike, M.; Demco, D. E.; Graf, R.; Gottwald, J.; Hafner, S.; Spiess, H. W. J. Magn. Reson. 1996, 122, 214-221. (24) Alam, T. M.; Tischendorf, B. C.; Brow, R. K. Solid State Nucl. Magn. Reson. 2005, 27, 99-111. (25) Bailly, M. L.; Costentin, G.; Lauron-Pernot, H.; Krafft, J. M.; Che, M. J. Phys. Chem. B 2005, 109, 2404-2413. (26) Massiot, D.; Fayon, F.; Capron, M.; King, I.; Le Calve´, S.; Alonso, B.; Durand, J. O.; Bujoli, B.; Gan, Z.; Hoatson, G. Magn. Res. Chem. 2002, 40, 70-76. (27) Becke, A. D. J. Chem. Phys. 1993, 98, 5648-5652. (28) Lee, C.; Yang, W.; Parr, R. G. Phys. ReV. B 1988, 37, 785-789. (29) Frisch et al., Gaussian 03, revision C.02; Gaussian, Inc.: Wallingford CT, 2004. (30) Sushko, P. V.; Shluger, A. L.; Catlow, C. R. A. Surf. Sci. 2000, 450, 153-170. (31) Nygren, M. A.; Pettersson, L. G. M. J. Chem. Phys. 1994, 100, 2010-2018. (32) Yudanov, I. V.; Pacchioni, G.; Neyman, K.; Ro¨sch, N. J. Phys. Chem. B 1997, 101, 2786-2792.
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18287 (33) Neyman, K. M.; Ro¨sch, N.; Pacchioni, G. Appl. Catal. A 2000, 191, 3-13. (34) Nasluzov, V. A.; Rivanenkov, V. V.; Gordienko, A. B.; Neyman, K. M.; Birkenheuer, U.; Ro¨sch, N. J. Chem. Phys. 2001, 115, 8157-8171. (35) Ditchfield, R. Mol. Phys. 1974, 27, 789-807. (36) Schindler, M.; Kutzelnigg, W. J. Chem. Phys. 1982, 76, 19191933. (37) Jameson, C. J. Nucl. Magn. Res. 1998, 27, 44-82. (38) Ochsenfeld, C. Phys. Chem. Chem. Phys. 2000, 2, 2153-2159. (39) Alkorta, I.; Elguero, J. Struct. Chem. 1998, 9, 187-202. (40) Pickard, C. J.; Mauri, F. Phys. ReV. B 2001, 63, 245101-245113. (41) Gervais, C.; Profeta, M.; Lafond, V.; Bonhomme, C.; Azaı¨s, T.; Mutin, H.; Pickard, C. J.; Mauri, F.; Babonneau, F. Magn. Res. Chem. 2004, 42, 445-452. (42) Charpentier, T.; Ispas, S.; Profeta, M.; Mauri, F.; Pickard, C. J. J. Phys. Chem. B 2004, 108, 4147-4161. (43) Gervais, C.; Dupree, R.; Pike, K.; Bonhomme, C.; Profeta, M.; Pickard, C. J.; Mauri, F. J. Phys. Chem. A 2005, 109, 6960-6969. (44) Harris, R. K.; Joyce, S. A.; Pickard, C. J.; Cadars, S.; Emsley, L. Phys. Chem. Chem. Phys. 2006, 8, 137-143. (45) Hunger, M. Catal. ReV.-Sci. Eng. 1997, 39, 345-393. (46) Jacobsen, C. J. H.; Topsoe, N. Y.; Topsoe, H.; Kellberg, L.; Jakobsen, H. J. J. Catal. 1995, 154, 65-68. (47) Brunner, E.; Karge, H. G.; Pfeifer, H. Z. Phys. Chem. 1992, 176, 173-183. (48) Gun’ko, V. M.; Turov, V. V. Langmuir 1999, 15, 6405-6415. (49) Harris, R. K.; Jackson, P.; Merwin, L. H.; Say, B. J.; Ha¨rgele, G. J. Chem. Soc., Faraday Trans. 1 1988, 84, 3649-3672. (50) Feng, M. H.; Chao, K. J. J. Mol. Struct. (THEOCHEM) 1996, 364, 51-57. (51) Yesinowski, J. P.; Eckert, H. J. Am. Chem. Soc. 1987, 109, 62746282. (52) Fleischer, U.; Kutzelnigg, W.; Bleiber, A.; Sauer, J. J. Am. Chem. Soc. 1993, 115, 7833-7838.