J. Phys. Chem. B 2001, 105, 1971-1977
1971
Detection, Quantification, and Magnetic Field Dependence of Solid-State 17O NMR of X-O-Y (X,Y ) Si,Ti) Linkages: Implications for Characterizing Amorphous Titania-Silica-Based Materials C. Gervais,† F. Babonneau,‡ and M. E. Smith*,† Department of Physics, UniVersity of Warwick, CoVentry, U.K. CV4 7AL, Chimie de la Matie` re Condense´ e, UniVersite´ Pierre et Marie Curie, CNRS, 4 Place Jussieu, 75252, Paris Cedex 05, France ReceiVed: September 27, 2000; In Final Form: December 11, 2000
solid-state NMR spectra from hybrid TiO2-SiO2-based materials can clearly distinguish Si-O-Si, SiO-Ti, and Ti-O-Ti linkages. Spectra are reported from these groups for applied magnetic fields from 5.6 to 14.1 T, which allows their NMR parameters to be deduced. The quantitative accuracy of these spectra is investigated and the implications for detecting nanoscale phase separation discussed. Changes in the local structure with functionalization of the silicon alkoxide (R′xSi(OR)4-x) and modification of the titanium precursor by acetylacetone are examined. 17O
TABLE 1: Some δ 17O Values Observed for Si-O-Ti Environments
Introduction Solid SiO2/TiO2 materials are of significant technological interest because of their low thermal expansion coefficients1 and catalytic,2 optical,3 photocatalytic, and other chemical properties.4 This has led to the widespread study of these materials by a range of techniques, often with the aim of understanding the dispersion of titanium within the SiO2 matrix. Increasing the concentration of titanium within the SiO2 structure would enhance many of the advantageous properties of these materials. The sol-gel process offers the possibility of introducing more TiO2 into SiO2 materials compared with the usual melt-quench process. Moreover, this kind of synthesis allows the formation of novel silica-titania hybrid organic-inorganic materials, which are a promising system for photonic applications.5-10 Low-loss waveguides have been fabricated by solgel processing of organically modified silicon alkoxides,8-11 and these materials have also been used as hosts for optically active organic molecules.9,11 In the sol-gel process, the major difficulty is to prevent phase separation when silicon alkoxides (R′xSi(OR)4-x) are crosslinked by Ti alkoxides [Ti(OR)4]. The degree of homogeneity of the final materials depends on the ability to favor cocondensation reactions between the precursors, despite their intrinsically different reactivities. A key to understanding this process is the ability to characterize and quantify accurately the mixed oxo-bridges (Si-O-Ti) in these amorphous systems. The importance of quantifying such links is pointed out in many recent articles that have examined this problem with a variety of techniques such as Ti K-edge X-ray absorption near-edge spectroscopy (XANES).12 Oxygen-17 (nuclear spin, I ) 5/2) has been used to study the local structure of a variety of amorphous inorganic solids.13-15 17O NMR parameters have been very sensitive to structure. Hence, the coordination and bonding of oxygen can be distinguished through the 17O isotropic chemical shift (δiso), which * Correspondence to: M. E. Smith, Department of Physics, University of Warwick, Coventry,U.K. CV4 7AL. E-mail:
[email protected] † University of Warwick. ‡ Universite ´ Pierre et Marie Curie.
sample
δ17O (ppm)
fresnoite Li2TiOSiO4 [Ti(acac)2O]2[OSi(C6H5)2]2 TiO2[O4Si4(C6H5)8]2 sol-gel prepared TiO2-SiO2 sol-gel prepared TiO2-SiO2
190 157 333 < δ < 363 295 250 < δ < 314 332
a
environment
ref
20 21 O2CSi-O-TiO5 22 O2CSi-O-TiO3 22 O3Si-O-TiO3 20, 23, a OC2Si-O-TiO5 a
This work.
spans a large range ≈1500 ppm, and quadrupolar coupling parameters CQ (CQ ) e2qQ/h, where eQ is the electric quadrupole moment and eq is Vzz, with Vxx, Vyy, and Vzz the principal elements of the tensor describing the electric field gradient) and ηQ the asymmetry parameter () |Vyy - Vxx|/Vzz). The main drawback of 17O NMR is its poor sensitivity caused by its low natural abundance (0.037%). This can be overcome, in sol-gel chemistry, by hydrolyzing the precursors with 17Oenriched water, because the relative stability of C-O bonds to M-O bonds ensures that 17O is incorporated efficiently into the growing M-O-M oxide network. Studies of connectivity between metallic atoms in transition metal-oxo systems using 17O NMR have been reported recently.16-19 17O NMR data have been reported for Ti-O-Si in fresnoite20 and Li2TiOSiO4.21 In fresnoite a resonance at 190 ppm was observed from the Ti-O-Si unit, but no other parameters could be deduced from the featureless line shape. In Li2TiOSiO4 an isotropic chemical shift of 157 ppm was observed, and CQ ) 3.05 MHz and η ) 0.35 were deduced from the second-order quadrupolar structure. Recently, this was extended to crystalline titanodiphenylsiloxanes containing SiO-TiO3 and Si-O-TiO5 environments with isotropic chemical shifts in the range 295-370 ppm and CQ values in the range 3.0-3.7 MHz.22 In sol-gel-produced TiO2-SiO2, peaks are observed in the range 110-250 ppm20,23 that are assigned to Ti-O-Si groups on the basis of the observations in the crystalline compounds and also their intermediate shift between Si-O-Si and Ti-O-Ti. These different 17O shifts observed for Si-O-Ti environments are summarized in Table 1.
10.1021/jp003519q CCC: $20.00 © 2001 American Chemical Society Published on Web 02/15/2001
1972 J. Phys. Chem. B, Vol. 105, No. 10, 2001
Gervais et al.
TABLE 2: Conditions of Preparation of the Different Samples sample
alkoxide/prehydrolysis time
Ti modification by AcacH
QTi QTiAc TTi TTiAc DTi DTiAc
TEOS/40 min TEOS/40 min MTES/10 min MTES/10 min DMDES/3 min DMDES/3 min
no yes no yes no yes
More detailed interpretation of 17O NMR spectra from such solids, however, requires better understanding of the structural influences on the 17O NMR line shape and position. Also, the ability of such NMR spectra to quantify such linkages needs to be investigated fully. This study has two aims. The first is to characterize further the sensitivity of 17O solid-state NMR parameters for the different X-O-Y (X, Y ) Si, Ti) oxobridges and to evaluate the quantitative integrity of these signals. The second is to use this information to investigate both the influence of the functionalization of the silicon alkoxides and the modification by acetylacetone of the titanium alkoxide on the proportions of the different bonds present. Three functionalized silicon alkoxides were investigated and reacted with titanium alkoxide according to the method introduced by Yoldas.24 The MexSi(OEt)4-x alkoxides (with 0 e x e 2) were prehydrolyzed under acidic conditions before addition of titanium isopropoxide [Ti(OPri)4]. The effect of modification of the titanium precursor by acetylacetone was studied. The prehydrolysis times of the silicon alkoxides were chosen, depending on the silicon alkoxide, to maximize the number of Si-OH groups to encourage the maximum formation of SiO-Ti bonds. Solution-state NMR was used previously25 to observe in detail the hydrolysis of these silicon alkoxides to determine when maximum Si-OH formation occurs. Experimental Section Sample Preparation. Tetraethoxysilane (TEOS, Fluka), methyltriethoxysilane (MTES, Fluka), dimethyldiethoxysilane (DMDES, Fluka), titanium isopropoxide [Ti(OPri)4, Fluka], and acetylacetone (AcacH, Fluka) were used as received. The silicon alkoxides were mixed with ethanol (EtOH/Si ) 4) and then hydrolyzed with 17O-enriched water (H2O/OEt ) 0.5). This rate of hydrolysis was chosen to avoid precipitation of TiO2 when adding titanium isopropoxide. Twenty atom percent enriched water (Isotec, France) was used, and the pH was adjusted to 2 through dilution with HCl-acidified solution of unenriched water at pH ) 1. Acetylacetone modified and unmodified Ti isopropoxides were used (Ti/Si ) 0.5) in separate experiments. Modification by acetylacetone was performed by mixing the two reactants in 2-propanol [Ti(OPri)4/acacH/PriOH ) 1/1/3]. Titanium alkoxide was added to the prehydrolyzed solution of silicon alkoxide (Table 2) when solution-state NMR showed maximum Si-OH formation. All six samples were then airdried for several days until gelation was complete. Samples are designated as XTi, X being Q, T, and D for SiO4, SiCO3, and SiC2O2 environments of the silicon alkoxides, respectively. 17O NMR Experiments. 17O liquid-state NMR experiments were recorded on a Bruker MSL 400 spectrometer operating at 54.22 MHz, using 17O-enriched H2O (10 atom %, Euriso-Top, France) to hydrolyze the silicon alkoxides. The spectra were recorded with a θ ) 90° pulse width (corresponding to a pulse length of 15 µs), and recycle delays of 300 ms and 500-6000 free induction decays (FIDs) were accumulated. 17O magic-angle spinning (MAS) NMR spectra were acquired on Chemagnetics CMX 240, 360, and 600 spectrometers
Figure 1. Evolution of the 17O NMR spectra of a hydrolyzed solution of TEOS (a) before and after addition of Ti(OBun)4 after (b) 15 min, (c) 1 h, and (d) 2.75 h.
operating at 32.29, 48.80, and 81.37 MHz, respectively, using Bruker 4-mm and Chemagnetics 3.2-mm probes. The spectra were recorded using a spin-echo θ - τ - 2θ pulse sequence with extended phase cycling26 to overcome problems of probe ringing and to avoid baseline distortion (θ ) 90° corresponding to a pulse length of 2 µs). The τ delay value was determined by the spinning frequency of the rotor (RO ) 11-20 kHz) and was therefore 50-90 µs (equivalent to 1/RO). A recycle delay between 0.5 and 2 s was used to ensure that fully relaxed spectra were produced. Between 25 000 and 200 000 FIDs were typically accumulated to obtain a reasonable signal-to-noise ratio. Chemical shift values were referenced to tap water (δ ) 0 ppm), and the resulting spectra were simulated using a modified version of the Bruker-Winfit program.27 Results Liquid-State Study of the Reaction between Ti Alkoxides and Prehydrolyzed Solutions of Si(OEt)4 and Me2Si(OEt)2. The 17O liquid-state NMR spectrum of a prehydrolyzed solution of TEOS immediately after addition of titanium butoxide is presented in Figure 1b. It shows Si-OH and Si-O-Si bonds of about 30 ppm, as already observed before addition of Ti alkoxide (Figure 1a), a broad signal in the 200-300 ppm range assigned to Si-O-Ti bridges,28 and two sharper ones at 345 and 530 ppm corresponding to OTi4 and OTi3 environments, respectively.17,19,28,29 At least two components can be distinguished in the spectrum from Si-O-Ti, and their assignment will be discussed below. Aging of the system was followed by recording the 17O NMR spectrum for 17 h. Characteristic spectra are presented in Figure 1. Progressive changes are observed during the first 3 h, and then the system seems to reach a state of equilibrium. The intensity of the signal assigned to Si-OTi bridges decreases, whereas the intensity of the OTi3 and OTi4 signals increases and a new peak caused by OTi2 is visible at about 800 ppm. The decrease in Si-O-Ti bonds corresponds mainly to an increase of the number of Ti-O-Ti bonds, whereas the number of Si-O-Si bonds remains fairly constant.
Solid-State
17O
NMR of Si-O-Ti Linkages
J. Phys. Chem. B, Vol. 105, No. 10, 2001 1973
Figure 2. Evolution of the 17O NMR spectra of a hydrolyzed solution of DMDES/Ti(OiPr)4 after (a) 7 min, (b) 2 h, (c) 6 h, (d) 18 h, and (e) 32 h.
This seems to indicate that redistribution reactions occur with cleavage of the Si-O bonds. Similar experiments were recorded with a prehydrolyzed solution of DMDES reacted with Ti isopropoxide, and the corresponding 17O NMR spectra are presented in Figure 2. Signals are observed at about 67, 295, 360, 540, and 730 ppm and can be assigned to Si-O-Si, Si-O-Ti, OTi4, OTi3, and OTi2 environments, respectively. The evolution of the 17O NMR spectra versus time shows that the intensities of the components due to Si-O-Si and Ti-O-Ti oxo-bridges clearly increase, whereas the signal due to Si-O-Ti decreases and almost disappears after 32 h (Figure 2e). This indicates that redistribution reactions leading to a cleavage of the Si-O bonds in the Si-O-Ti bridges occur in higher proportions than in the TEOS system. It suggests that phase separation will be very likely in this system in the gel state. This preliminary liquid-state study shows that formation of Si-O-Ti bonds is favored kinetically when Ti(OR)4 is added to a Si-OH-rich solution, but the system evolves to a more thermodynamically stable state leading to phase separation between Ti- and Si-rich species. Nonetheless, it is assumed here that this evolution is stopped by the gelation process. Gels Obtained by Reaction of Ti(OPri)4 with Prehydrolyzed Solutions of MexSi(OEt)4-x. The 17O MAS NMR spectra recorded at 5.6 T from samples obtained with the different functionalized silicon alkoxides (0 e x e 2) are shown in Figure 3. They show at high field, between -200 and 100 ppm, a broad signal with a well- defined second-order quadrupolar shape that can be assigned unequivocally to Si-O-Si bonds.16 Two sharper signals are present at about 375 and 540 ppm on all three spectra and correspond most probably to OTi4 and OTi3 environments, respectively, as already observed in solution (Figure 1) and in the solid state.18,30 Some OTi2 environments may also be present at about 760 ppm but at relatively low abundance. An additional signal is also evident with a maximum intensity between 200 and 300 ppm. It is assigned to Si-O-Ti environments as previously identified by 17O liquid-state NMR (Figure 1). Moreover, Si-O-Ti signals observed in the solid
Figure 3. 17O MAS NMR spectra recorded at 5.6 T of (a) QTi, (b) Tti, and (c) DTi samples (*spinning sidebands).
state in sol-gel materials20,23,31 and crystalline titanodiphenylsiloxanes22 have also been reported between 250 and 350 ppm (Table 1). The proportions of these (X-O-Y) (X, Y ) Si, Ti) oxobridges are obviously different in each sample indicating the dependence on the silicon alkoxide used. The DTi spectrum shows higher phase separation than that of QTi. This is manifest as a higher proportion of Ti-O-X (X ) Si, Ti) bonds than Si-O-Si, which will be discussed later. Moreover, the proportion of Si-O-Si compared with Ti-O-Ti is much smaller in DTi because of the presence of two methyl groups linked to each silicon. In all these samples a significant signal corresponding to SiO-Ti occurs. To confirm these assignments and to extract the relative amounts of the different X-O-Y bonds, more precise simulations and consequently a better knowledge of the NMR parameters of these oxo-bridges is required. The shapes of these signals are complex because of overlapping second-order quadrupolar line shapes which can only be partially averaged by MAS.32,33 Second-order broadening was reduced by recording spectra at higher applied magnetic fields, which improved confidence in the simulations of these resonances by providing additional constraints. QTi Sample. Figure 4 shows the 17O MAS NMR spectra recorded at 5.6, 8.45, and 14.1 T for the QTi sample. The two relatively sharp lines occurring at 375 and 540 ppm are assigned to OTi4 and OTi3 environments, respectively. The line widths in Hertz of these resonances scale directly with the applied magnetic field, indicating that they are determined almost completely by chemical shift dispersion. These Ti-O-Ti signals were therefore simulated successfully using simple Gaussians
1974 J. Phys. Chem. B, Vol. 105, No. 10, 2001
Gervais et al. TABLE 4: 17O Solid-state NMR Parameters Extracted from the Simulations of the Spectra Recorded at All Three Applied Fields sample QTi
TTi
DTi
QTiAc
TTiAc
DTiAc
a
Figure 4. Experimental and simulated 17O MAS NMR spectra of sample QTi recorded at (a) 5.6 T, (b) 8.45 T, (c) 14.1 T (*spinning sidebands).
TABLE 3: Line Widths Observed at 5.6 and 14.1 T and Deduced Quadrupolar Coupling Constants for the Si-O-Ti Signals in the Different Samples Studied ∆ν (kHz) ((0.1) δiso (ppm) CQ (MHz) sample ((2) ∆ν(5.6 T) ∆ν(14.1 T) ∆νQ at 5.6 T ((0.1) QTi TTi DTi QTiAc TTiAc DTiAc
290 174 337 332 314 174 360 332
4.2 3.6 3.7 2.5 4.1 3.0 3.6 2.4
4.6 3.9 4.4 4.0 4.9 3.6 3.8 3.9
2.9 2.4 2.3 1.0 2.6 2.0 2.6 0.9
3.1 2.8 3.0 2.1 3.0 2.7 3.0 2.0
with the same line width (in ppm) at each field. The ability to simulate such 17O resonances by Gaussians neglecting secondorder quadrupole effects has already been reported for TiO2 nanoparticles34 and in sol-gel-prepared TiO2-SiO2 materials.20 The broad signal spanning from -200 to 50 ppm is attributed to Si-O-Si bonds and can be simulated at all fields with a second-order quadrupolar line shape characteristic of such oxobridges (Table 3).35 Considering the signal in the region 200-300 ppm assigned to Si-O-Ti environments, the presence of shoulders on this resonance at 5.6 and 8.45 T (indicated with arrows in Figure 4) suggests that there are probably several overlapping components as observed in the liquid state (Figure 1). The spectrum at 14.1 T clearly resolves two signals at about 290 and 170 ppm. The second less intense feature may also arise from an Si-O-Ti environment,31 and its assignment is discussed below. With
δiso (ppm) ((1)
CQ (MHz) (( 0.1)
ηQ ((0.1)
assignment
%a ((2)
42 174 290 375 542 70 337 365 546 86 332 360 542 42 174 314 375 542 70 360 365 546 86 332 370 542
5.1 2.8 3.1 4.8 3.0 4.9 2.1 5.1 2.7 3.0 4.8 3.0 4.9 2.0 -
0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 -
Si-O-Si Si-O-Ti Si-O-Ti OTi4 OTi3 Si-O-Si Si-O-Ti OTi4 OTi3 Si-O-Si Si-O-Ti OTi4 OTi3 Si-O-Si Si-O-Ti Si-O-Ti OTi4 OTi3 Si-O-Si Si-O-Ti OTi4 OTi3 Si-O-Si Si-O-Ti OTi4 OTi3
45 9 31 4 11 56 23 9 12 24 30 14 32 42 6 44 1 7 60 30 5 5 25 23 22 30
Extracted from the simulations of the spectra recorded at 5.6 T.
increasing magnetic field, although these Si-O-Ti signals narrow, they are broader than would be expected for a pure quadrupolar shape (Figure 4). This suggests that the broadening has both quadrupolar and chemical shift distribution contributions. To separate these contributions, if at 5.6 T the contributions are quadrupolar (∆νQ) and chemical shift dispersion (∆νCSD) then the total line width in Hertz ν(5.6 T) can be derived from summing the squared contributions ∆ν2(5.6 T) ) (∆νQ)2 + (∆νCSD)2. Then from the differing field dependences of these two contributions the total line width at 14.1 T, ν(14.1 T), is ∆ν2(14.1 T) ) (∆νQ)2/R2 + R2(∆νCSD)2, where R ) (14.1/5.6). Using the line width ∆νQ obtained from the two field line width data allows CQ to be estimated by spectral simulation (Table 3). The Si-O-Ti signals observed on the QTi sample could be well fitted with parameters close to those found for crystalline titanodiphenylsiloxanes (Table 1). The NMR parameters and the intensity distribution between the different sites extracted from the simulations are given in Table 4. TTi and DTi Samples. OTi3 and OTi4 signals observed in the spectra obtained with TTi and DTi (Figure 3) can be simulated with parameters very close to those used for the QTi sample (Table 4). The spectra recorded at different fields for the DTi sample (Figure 5), which shows a large amount of these two environments, confirm that the line widths (in ppm) of these resonances are not changed on increasing the magnetic field. It appears that the chemical shift of the Si-O-Si bridges increases to higher ppm values (from 44 ppm in QTi to 70 and 86 ppm in TTi and DTi, respectively) when the functionalization of the silicon alkoxide increases (Table 4). This phenomenon has already been observed in the liquid state36 and is very likely caused by the presence of CH3 groups increasing the negative charge on the oxygen. Only one distinct signal corresponding to Si-O-Ti environments is observed in these two samples with higher chemical
Solid-State
17O
NMR of Si-O-Ti Linkages
J. Phys. Chem. B, Vol. 105, No. 10, 2001 1975
Figure 5. Experimental and simulated 17O MAS NMR spectra of sample DTi recorded at (a) 5.6 T, (b) 8.45 T, (c) 14.1 T.
shift values than those observed in QTi. This can also be partly related to the functionalization of the silicon alkoxide, and also may arise from higher coordination of the attached titanium. Indeed, Ti K-edge XANES has suggested that titanium was both four- and six-coordinated in QTi37 whereas it was mainly sixcoordinated in DTi.38 This is consistent with results obtained for titanodiphenylsiloxanes showing shifts of Ti-O-Si signals to higher ppm values when the coordination of Ti increases (Table 1). The CQ and η values for these sites in TTi remain very similar to those in QTi, whereas the corresponding CQ value for DTi appears smaller. In addition, the OTi3 line of the DTi is obviously narrower than QTi and TTi (Figure 3). This may be related to better definition of the TiO2 domains, which indicates more significant phase separation. The CQ values were extracted from the low-field spectra as described previously for the QTi sample, taking into account the broadening due to the chemical shift dispersion, evaluated by comparison with the spectra recorded at 14.1 T (Table 3). In the synthesis presented here, the hydrolysis ratio H2O/ OEt of the silicon alkoxides is only 0.5 to avoid a significant TiO2 precipitation. To complete gelation further reaction with atmospheric unenriched H2O occurs. However the redistribution of 17O between the different (X-O-Y) (X, Y ) Si, Ti) bonds is probably random because solution-state NMR shows that almost complete redistribution occurs before gelation (Figure 2). As expected from the liquid-state NMR study, the QTi sample shows the highest Si-O-Ti/Si-O-Si ratio compared with Ti-O-Ti/Si-O-Si, so it is the best candidate for incorporation of Ti into a Si-O-Si network. TTi and DTi samples are less suitable because larger relative Ti-O-Ti
Figure 6. 17O MAS NMR spectra recorded at 5.6 T for the (a) QTiAc, (b) TtiAc, and (c) DTiAc samples (*spinning sidebands).
signals are observed, which indicates greater phase separation. This is particularly striking in the DTi sample and possibly can be explained by the much smaller number of Si-OH bonds available to react with Ti-OPri to form Si-O-Ti bridges. Moreover, the liquid-state NMR study has shown redistribution reactions leading to a decrease of the Si-O-Ti bonds in favor of the formation of mainly Ti-O-Ti and some Si-O-Si. These reactions occur more particularly in the DMDES system and lead to more significant phase separation (Figure 2). Indeed in this system, the presence of CH3 groups on the silicon increases its hydrophobicity, resulting in easier breaking of the Si-OTi bond. Nevertheless, the significant level of Ti-O-Si bonding observed, which is probably at the interface between TiO2 and SiO2 domains, indicates that the TiO2 domains are rather small. The DTi system may be described as a nanocomposite with chains of poly(dimethylsiloxane)s and TiO2-based particles, where the chains would act as bridges between the inorganic particles.38,39 Gels Obtained by Reaction of Ti(OPri)4 Modified with Acetylacetone and Prehydrolyzed Solutions of MexSi(OEt)4-x. The 17O MAS NMR spectra recorded at 5.6 T for the three samples obtained with the different silicon alkoxides and titanium isopropoxide modified by acetylacetone are presented in Figure 6. The QTiAc and TTiAc samples show much smaller Ti-O-Ti signals than the samples prepared with unmodified Ti alkoxides (Figure 3). This can be explained by the lower reactivity of the Ti(OR)3-(acac) and Ti(OR)2-(acac)2 complexes formed with acetylacetone.19
1976 J. Phys. Chem. B, Vol. 105, No. 10, 2001
Figure 7. Experimental and simulated 17O MAS NMR spectra of sample QTiAc recorded at (a) 5.6 T, (b) 8.45 T, (c) 14.1 T (*spinning sidebands).
QTiAc (Figure 7) and TTiAc (Figure 8) spectra could be simulated with parameters quite similar to those used for QTi and TTi (Table 4), except a small shift to higher values is observed for the chemical shift of Ti-O-Si, possibly related to the modification by acetylacetone. This behavior has already been observed for a TEOS + Ti(OnBu)4 + AcacH system in solution-state NMR work.28 The behavior of DTiAc is slightly different, but this sample showed a very long gelation time (>2 weeks) and was still not a completely rigid solid. The consequent higher mobility in the sample led to a narrower Ti-O-Si signal, which permitted complete resolution of the OTi4 and Si-O-Ti environments (Figure 6). Discussion Observation of Ti-O-Si Linkages. Si-O-Si, Ti-O-Ti, and Si-O-Ti signals are readily separated and characterized in these mixed sol-gel samples, illustrating the ability of 17O NMR to provide detailed atomic scale information about TiO2-SiO2 systems showing structural and/or atomic disorder. Moreover, two types of Si-O-Ti linkages seem to be present in the QTi and QTiAc samples as already observed by 17O liquid-state (Figure 1) and solid-state NMR.31 It was suggested that the distinct Si-O-Ti resonances observed may be due to different coordinations of the attached titanium (i.e., TiO4, TiO6).31 At low titanium concentrations (x < 0.18) (TiO2)x(SiO2)1-x forms a glassy silica framework with titanium substitution in tetrahedral sites after heating to ∼750 °C.37 By correlating the structural changes with the observed 17O NMR resonances, there is no doubt that the 250-ppm resonance
Gervais et al.
Figure 8. Experimental and simulated 17O MAS NMR spectra of sample TTiAc recorded at (a) 5.6 T, (b) 8.45 T, (c) 14.1 T (*spinning sidebands).
Figure 9. Schematic representation of the two kinds of proposed SiO-Ti bonds observed in the QTi and QTiAc samples.
observed in TiO2-SiO2 gels, which remains after heating, must correspond to the Si-O-TiO3 framework sites (Table 1). This assignment is also consistent with the shift range of Si-OTiO3 observed in titanodiphenylsiloxanes. However, the inconsistency then is that the peak that appeared at 150-110 ppm in TiO2-SiO2 gels was previously assigned to a higher coordination of the titanium site.31 The more extensive knowledge of the shift range now available suggests that this assignment is unlikely, so that the resonances observed here at 290 and 174 ppm in QTi could both be assigned to Si-O-TiO3 sites. The peak at 174 ppm is less stable, being removed on heat treatment to leave the other peak as a framework site. We tentatively suggest that this signal could be assigned to a tri-coordinated Si-O-Ti2 environment (Figure 9). The less positive shift compared with the other oxygen between silicon and titanium is consistent with an increase in the coordination number of the oxygen. During the early stages of development of the samples, acetylacetone remains in the structure which is probably the source of the increase in the shift of peaks from 290 to 314 ppm (QTi) and 337 to 360 ppm (TTi). In this study in the initial gel made from the TEOS precursor the two peaks
Solid-State
17O
NMR of Si-O-Ti Linkages
J. Phys. Chem. B, Vol. 105, No. 10, 2001 1977
TABLE 5: Proportions of the X-O-Y (X, Y ) Ti, Si) Bonds Extracted from the Simulations of the 17O MAS NMR Spectra Recorded at 5.6 and 14.1 T for Sample QTi bonds
% at 5.6 T ((2%)
% at 14.1 T ((2%)
Si-O-Si Si-O-Ti OTi4 OTi3
44 40 4 11
50 31 5 14
at 174 and 290 ppm probably correspond to types of structural units similar to those observed at shifts of ∼110 and 250 ppm in the previous study.31 Quantitative Aspect. A primary concern of this investigation was to examine the ability of 17O NMR data to characterize quantitatively the distribution of oxygen between different oxobridges. The relative amount of the X-O-Y (X, Y ) Si, Ti) bonds obtained from the simulations of the spectra recorded for sample QTi at 5.6 and 14.1 T are reported in Table 5. From direct integration of the simulation there are variations between the 5.6 and 14.1 T data, which exceeds the estimated error. Here echoes were used to prevent baseline problems especially at lower applied magnetic fields. Using spectra produced from echoes quantitatively must be approached cautiously if there are significantly different T2 values for different sites, and different nutation rates during the radio frequency pulses for quadrupolar nuclei. Although wide variation of CQ occurred between the sites, all here are effectively “large” CQ limit with the central transition observed.33 Some spectra at 14.1 T were run in single, small tip-angle mode for comparison which produced no noticeable change in the relative peak intensity. In addition, for the Si-O-Si signal the size of CQ means that there will be a distribution of the intensity for the central transition between the centerband and the spinning sidebands, especially at 5.6 T. For the (1/2, -1/2) transition, this depends on the parameter νQ2/νLνR,40 where νQ is the quadrupolar frequency [νQ ) 3CQ/2I(2I - 1)] and νR is the spinning speed. At 14.1 T νQ2/νLνR ) 0.36 (νQ ) 0.765 MHz, νL ) 81.4 MHz, and νR ) 0.02 MHz), which means that most of the intensity is in the centerband; whereas at 5.6 T νQ2/νLνR ) 2.58 (νQ ) 0.765 MHz, ν0 ) 32.3 MHz, and νR ) 0.011 MHz), which implies that the fraction of the magnetization in the centerband is close to 90%. When this factor is taken into account, there is much better agreement of the amount of Si-O-Si between the two experiments. The ratio between the proportion of OTi3 and OTi4 sites remains fairly constant and close to 2.8 at both fields. The slight decrease observed in the amount of Si-O-Ti signal when the field increases is difficult to explain. However, this indicates that using single magnetic field can be misleading and care must be taken in quantitative interpretation of these data. This suggests that absolute quantification may not be straightforward. Nevertheless for comparable samples observations at a constant field will allow relative changes in the bonding and hence the structural variation to be deduced. The QTi and QTiAc show some interesting differences in that the oxygen distributions Si-O-Si:Si-O-Ti:Ti-O-Ti are 45:40:15 and 42:50:8, respectively. Hence modification of the titanium precursor by acetylacetone has clearly promoted Si-O-Ti formation. This article has demonstrated that 17O NMR plays an extremely important role in determining the structure of titaniasilica-based materials.
Acknowledgment. M.E.S. thanks the EPSRC for funding NMR equipment at Warwick and work on 17O characterization of materials through GR/N64267. The collaboration between Warwick and Paris was funded through a CNRS/Royal Society grant. C.G. thanks the EU for a Framework V Marie Curie fellowship. References and Notes (1) Schultz, P. C. J. Am. Ceram. Soc. 1976, 59, 214. (2) Baiker, A.; Dollenmeier, P.; Glinski, M.; Reller, A. Appl. Catal. 1987, 35, 365. (3) Schroeder, H. Phys. Thin Films 1969, 5, 87. (4) Itoh, M.; Hattori, H.; Tanabe, K. J. Catal. 1974, 35, 225. (5) Brusation, G.; Guglielmi, M.; Innocenzi, P.; Martucci, A.; Battaglin, G.; Pelli, S.; Righini, G. C. J. Non-Cryst. Solids 1997, 220, 202. (6) Yoshida, M.; Prasad, P. N. Chem. Mater. 1996, 8, 353. (7) Innocenzi, P.; Martucci, A.; Guglielmi, M.; Armelao, L.; Pelli, S.; Righini, G. C.; Battaglin, G. J. Non-Cryst. Solids 1999, 259, 182. (8) Motakef, S.; Suratwala, T.; Roncone, R. L.; Boulton J. M.; Teowee, G.; Uhlmann D. R. J. Non-Cryst. Solids 1994, 178, 37. (9) Sorek Y.; Reisfeld, R.; Tenne, R. Chem. Phys. Lett. 1994, 227, 235. (10) Yang, L.; Saavedra, S. S.; Armstrong, N. R.; Hayes, J. Anal. Chem. 1994, 66, 1254. (11) Sorek, Y.; Reisfeld, R.; Weiss, A. M. Chem. Phys. Lett. 1995, 244, 371. (12) Kim W. B.; Choi, S. H.; Lee, J. S. J. Phys. Chem. B 2000, 104, 4091. (13) Florian, P.; Vermillion, K. E.; Grandinetti, P. J.; Farnan, I.; Stebbins, J. F. J. Am. Chem. Soc. 1996, 118, 3493. (14) Dirken, P. J.; Kohn, S. C.; Smith, M. E.; van Eck, E. R. H. Chem. Phys. Lett. 1997, 266, 568. (15) Lee, S. K.; Stebbins, J. F. J. Phys. Chem. B 2000, 104, 4091. (16) Walter, T. M.; Turner, G. L.; Oldfield, E. J. Magn. Reson. 1988, 76, 106. (17) Day, V. W.; Eberspacher, T. A.; Klemperer, W. G.; Park, C. W.; Rosenberg, F. S. J. Am. Chem. Soc. 1991, 113, 8190. (18) Bastow, T. J.; Moodie, A. F.; Smith, M. E.; Whitfield, H. J. J. Mater. Chem. 1993, 3, 697. (19) Blanchard, J.; Barboux-Doeuff, S.; Maquet, J.; Sanchez, C. New J. Chem. 1995, 19, 343. (20) Dirken, P. J.; Smith, M. E.; Whitfield, H. J. J. Phys. Chem. 1995, 99, 395. (21) Bastow, T. J.; Botton, G. A.; Etheridge, J.; Smith, M. E.; Whitfield, H. J. Acta Crystallogr. A 1999, 55, 127. (22) Gervais, C.; Babonneau, F.; Hoebbel, D.; Smith, M. E. Solid State NMR, in press. (23) Smith, M. E.; Whitfield, H. J. J. Chem. Soc., Chem. Commun. 1994, 723. (24) Yoldas, B. E. J. Non-Cryst. Solids 1980, 38, 81. (25) Babonneau, F.; Gualandris, V.; Pauthe, M. Mater. Res. Soc. Symp. Proc. 1996, 435, 119. (26) Kunwar, A. C.; Turner, G. L.; Oldfield, E. J. Magn. Reson. 1986, 69, 124. (27) Massiot, D.; Thiele, H.; Germanus, A. Bruker Rep. 1994, 140, 43. (28) Delattre, L.; Babonneau, F. Chem. Mater. 1997, 9, 2365. (29) Scolan, E.; Sanchez, C. Chem. Mater. 1998, 10, 3217. (30) Blanchard, J.; Bonhomme, C.; Maquet, J.; Sanchez, C. J. Mater. Chem 1998, 8, 975. (31) Pickup, D. M.; Mountjoy, G.; Wallidge, G. W.; Anderson, R.; Cole, J. M.; Newport, R. J.; Smith, M. E. J. Mater. Chem. 1999, 9, 1299. (32) Kundla, E.; Samoson, A.; Lippmaa, E. Chem. Phys. Lett. 1981, 83, 229. (33) Smith, M. E.; van Eck, E. R. H. Progr. Nucl. Magn. Reson. Spectrosc. 1999, 34, 159. (34) Scolan, E.; Magnenet, C.; Massiot, D.; Sanchez, C. J. Mater. Chem. 1999, 9, 2467. (35) Babonneau, F.; Maquet, J. Polyhedron 2000, 19, 315. (36) Babonneau, F.; Maquet, J.; Livage, J. Chem. Mater. 1995, 7, 1050. (37) Mountjoy, G.; Pickup, D. M.; Wallidge, G. W.; Anderson, R.; Cole, J. M.; Newport, R. J.; Smith, M. E. Chem. Mater. 1999, 11, 1253. (38) Dire´, S.; Babonneau, F.; Sanchez, C.; Livage, J. J. Mater. Chem. 1992, 2, 239. (39) Dire´, S.; Babonneau, F.; Carturan, G.; Livage, J. J. Non-Cryst. Solids 1990, 121, 428. (40) Massiot, D.; Bessada, C.; Coutures, J.-P.; Taulelle, F. J. Magn. Reson. 1990, 90, 231.