Chemical Shift Prediction in the 29Si MAS NMR of ... - ACS Publications

compounds and especially for predicting 29Si chemical shifts in titanium-containing silicates. The purpose of this work is to see whether we could est...
0 downloads 0 Views 123KB Size
J. Phys. Chem. B 1998, 102, 2897-2904

Chemical Shift Prediction in the

29Si

2897

MAS NMR of Titanosilicates

Andrea Labouriau, T. J. Higley, and William L. Earl* Chemical Science and Technology DiVision, Mail Stop J514, Los Alamos National Laboratory, Los Alamos, New Mexico 87545 ReceiVed: May 20, 1997; In Final Form: February 18, 1998

We have related the 29Si MAS NMR chemical shifts of titanosilicates and highly siliceous zeolites linearly with structural factors. Good correlations in titanosilicates were obtained using the model proposed by Sherriff et al.,1,2 where the chemical shift is described in terms of the summation of valence, dipole, and hybridization terms. This empirical correlation provides a basis for semiquantitative interpretation of spectra for titaniumbearing silicates. Our original goal was to understand the lack of a 29Si NMR peak assignable to silicon that is next-nearest neighbor to titanium in titanium-containing zeolites. Although the equations and correlations presented work very well for most titanium-containing silicates, we still do not understand the 29Si MAS NMR spectrum of these zeolites because our results suggest that there should be a distinguishable peak for TS-1 and TS-2. No such peaks are detected in the actual spectra.

Introduction We are interested in characterizing the catalytically active site in titanium-substituted zeolites. These materials, TS-1, TS2, Ti-beta, etc., are selective oxidation catalysts, isostructural with ZSM-5 (MFI), ZSM-11 (MEL), and zeolite beta (BEA). There are extensive publications using various spectroscopies to characterize the titanium site. However, as suggestive as the results may be, none of them clearly prove that titanium is at tetrahedral sites in the framework. We give a short summary of spectroscopic measurements of TS-1 and TS-2 below. Early 29Si NMR data claimed that the high field shoulder on the main peak of TS-1 is due to silicon sites that are next-nearest neighbor to titanium. More recent work by Tuel et al. makes a strong case for this shoulder simply being due to phase transitions or structural distortions in the MFI structure with increasing titanium content in the material.3 Figure 1 is a plot of the 29Si magic angle sample spinning NMR spectrum of highly siliceous MFI (silicalite) taken at about 325 K (trace A) and a digitally broadened plot of the same spectrum (trace B). This can be compared to trace C, which is the spectrum of TS-1. This supports Tuel and co-worker’s claim that the upfield shoulder on the TS-1 spectrum is due to the MFI structure rather than silicon nuclei with titanium as next-nearest neighbors. Solid-state 29Si NMR spectra of zeolites substituted with heteroatoms, e.g., aluminum and boron, have easily identified peaks for those Si nuclei with the heteroatom as next-nearest neighbor. We were perplexed by the lack of a clear 29Si NMR peak due to silicon associated in some fashion with the titanium in the structure and decided to pursue the structural implications of 29Si chemical shifts in these materials. Our goal is to use 29Si NMR of titanozeolites in analogy to the way that it is used in alumino-zeolites to probe structure, substitution, and structural changes in the zeolite as a function of synthesis and perhaps of reaction conditions. It is possible to calculate chemical shifts from quantum mechanical first principles, and this is often done in 13C NMR to relate the spectrum to chemical structures. Recent work by Moravetski et al.4 extends these types of calculations to 29Si NMR. However, ab initio calculations for such complex

Figure 1. 29Si MAS NMR spectra of TS-1 and siliceous ZSM-5. The lower spectrum (A) is from a very crystalline sample of ZSM-5. Spectrum B is the same as (A) but with a large Gaussian line broadening applied prior to Fourier transformation. The upper spectrum (C) is from a very good sample of TS-1, but it is representative of many other samples we have synthesized.

structures can be time-consuming and are often not convenient for the experimentalist. Additionally, we are interested in understanding chemical shift differences of only a few parts per million which is a difficult task for quantum mechanical methods. However, there are a significant number of papers correlating the detailed structure around a given silicon nucleus with the 29Si NMR chemical shift in different silicate materials.1,2,5-14 These publications correlate geometrical parameters such as bond angles and interatomic distances to 29Si NMR chemical shifts in silicates and aluminosilicates. These correlations can be very good when a small number of very similar samples are used but often break down when more than a few materials are included in the regression. Our experience is that the empirical correlation proposed by Sherriff and Grundy1 gives the best results over a widest variety of compounds and especially for predicting 29Si chemical shifts in titanium-containing silicates. The purpose of this work is to see whether we could establish similar correlations for titanium silicates and then to use that correlation in a predictive fashion to better understand why there is no distinctive 29Si NMR peak

S1089-5647(97)01670-2 CCC: $15.00 © 1998 American Chemical Society Published on Web 03/31/1998

2898 J. Phys. Chem. B, Vol. 102, No. 16, 1998 due to silicons that are next-nearest neighbor to titanium, if such really exist in these titanozeolites. Correlations of chemical shift with local bond angles and bond lengths recognize the fact that the local structure is intimately related to the detailed electronic environment which is, in turn, responsible for the measured NMR chemical shift. Structure-shift correlations can be accurate within 1 or 2 ppm and so could effectively be used to aid in structural studies of unknown materials. The hope is to establish chemical shift limits that are consistent with zeolite and titanosilicate structures and to see whether these aid in understanding the lack of a distinctive peak for silicon in titanozeolites. The goal is to use high-resolution solid-state 29Si NMR to obtain information about the nature of Ti in titanozeolites in analogy to methods used to study aluminum substitution in aluminum-containing zeolites and aluminosilicate minerals.15 These structural studies rely on the fact that the 29Si isotropic chemical shift is dependent on both the substitution and geometry of Si-O-X bonds. Trace C of Figure 1 is a representative spectrum of TS-1. There is a barely discernible peak at -101 to -103 ppm. This downfield peak is due to Si-O-H defects as was aptly shown by Maciel16 and Dessau.17,18 This peak is larger in spectra from samples with more defects in the structure. We present a short digression into methods that have been used to characterize titanium-containing zeolites followed by a discussion of the derivation of Sherriff and Grundy’s equations. We then demonstrate that these equations can be effectively applied to a few purely siliceous zeolites because zeolites are porous materials with somewhat flexible frameworks, and it is possible that this will affect the structure-shift correlations in an unknown way. The three samples chosen are ones that have excellent crystal structures published in the literature; thus, the deviations in the regression are not due to uncertainties in the structure. Finally, we apply the correlational formalism to a suite of titanosilicates of differing structures. Most of the samples are condensed, crystalline minerals, but a few are porous materials. Although the results leave the questions of the structure of titanozeolites still unanswered, we believe that these equations, in conjunction with powder diffraction data, are useful aids in determining zeolite structures. Characterization of Titanozeolites As mentioned above, there has been extensive spectroscopic work to characterize titanium-containing zeolites. It is not possible to give a complete review of this literature here, but we would like to comment on the uniqueness of the various characterizations. The primary methods are infrared and Raman spectroscopy, powder X-ray diffraction (XRD), diffuse reflectance ultraviolet spectroscopy (DRUV), X-ray adsorption near edge spectroscopy (XANES), extended X-ray absorption fine structure (EXAFS), NMR, elemental analysis, and interpretation of catalytic reactions. Powder X-ray diffraction techniques are invaluable for determining that samples have the correct framework topology, e.g., that a particular TS-1 sample has the MFI structure and that there is little amorphous background. Careful studies of the powder XRD as a function of titanium in the synthesis mixture indicate that the unit cell of TS-1 undergoes an expansion with isomorphous substitution of Si by Ti.19,20 However, the MFI framework is notoriously flexible,21,22 and it is likely that occluded TiO2 in the zeolite would also cause expansion in the unit cell. Vibrational methods (infrared and Raman) show a band at 960-970 cm-1, whose

Labouriau et al. intensity increases with titanium content.23,24 Samples without this absorption band invariably have no catalytic activity so it has been assigned to a titanyl group or to a symmetric stretching vibration mode of the SiO4 unit attached to a Ti4+ ion.23 However, silicates with a large number of defects or cations also have a vibrational band at about 960 cm-1 25 so it is difficult to ensure that this band is due to titanium in the framework or just defects created in the synthesis because of titanium in the reaction mixture. According to Deo et al., all solid compounds with Ti-O bonds possess Raman bands below 900 cm-1 which are not seen in the titanosilicalites.26 They conclude that vibrational spectroscopies do not provide direct structural information about the titanium species in the Ti silicalites since the vibrations observed are essentially those arising due to silicon-oxygen vibrations. Diffuse reflectance UV spectra of titanozeolites have an adsorption at about 330 nm when anataselike phases are present. Good samples of titanozeolites (those that are catalytically active) do not have this characteristic anatase-like absorption. Conversely, those that do have an anatase-like adsorption simply convert H2O2 to water.23,24,27,28 However, it is possible that small, nearly molecular species of titanium oxide are site isolated in the zeolite pores. Such species would not have anatase structure and thus would not have the characteristic DRUV spectrum. X-ray absorption studies, XANES and EXAFS, are potentially very useful in determining the details of titanium coordination and bond lengths. Early structural work on titanium-containing materials29-31 strongly suggests that the preedge features of XANES spectra can be used to deduce the coordination geometry of titanium. Despite this, Lopez32 and Behrens33 propose that titanium occupies octahedral and square-pyramidal sites while Davis34 and Bordiga et al.35 conclude that Ti is in tetrahedral coordination. Trong et al. investigated TS-2 using EXAFS and XANES at the Ti K-edge and conclude that titanium is located on a framework defect and not on a regular T-site.36 Since TS-1 and TS-2 have similar structures and catalytic activity, Trong et al. suggest that titanium sites in both materials are similar, i.e., framework defects. Since interpretation of X-ray adsorption is somewhat subjective, it is likely that the different interpretations are due to varying sample synthesis and preparation techniques. Some of the X-ray absorption studies include analyses of the fine structure (EXAFS), but the low concentration of Ti in the catalytic samples makes it difficult to get very good signal-to-noise. The titanium oxygen bond lengths are between 1.80 and 1.85 Å with the bond lengths being longer in hydrated or solvated samples than in very dry samples. The coordination numbers reported go from 4 to 6, but most authors indicate that the errors in calculating coordination numbers are very large. NMR spectroscopy has not proven to be very useful as an analytical tool to understand titanozeolites. A cursory examination of the magnetic properties of titanium would indicate that 47Ti and 49Ti would be slightly more difficult than 27Al NMR in other zeolites. There is one report of titanium NMR in TS-1 materials.37 However, in our laboratory we have been successful in obtaining 47,49Ti spectra of a number of titanium-containing materials, including some of the minerals in this report, but have been unsuccessful with TS-1. We believe that the reasons for this lack of success are the low density of Ti in titanozeolites and the probability that the Ti site is somewhat distorted, producing very wide NMR lines. As mentioned above and the fundamental reason for this study, 29Si solid-state NMR of titanozeolites does not have a clearly distinguishable peak for silicon adjacent to titanium such as seen in alumino-zeolites.

29Si

MAS NMR of Titanosilicates

J. Phys. Chem. B, Vol. 102, No. 16, 1998 2899

TABLE 1: Source and Properties of the Samples Studied; 29Si Chemical Shifts Were Calculated Using Eq 5 sample

origin

ideal formula

titanite benitoite fresnoite43 lorenzenite narsarsukite vinogradovite

Gilgit, Pakistan Ontario, Canada San Benito, CA synthetic Kola Peninsula, Russia Sage Creek, MT Kola Peninsula, Russia

CaTiOSiO4 BaTiSi3O9 Ba2TiOSi2O7 Na2Ti2Si2O9 Na2TiOSi4O10 Na2Ti2O2Si2O6(Si,Al)2O5(H2O,K2O)

zorite

Kola Peninsula, Russia

Na6[Ti (Ti0.9Nb0.1) 4(Si6O17)2(O,OH) 5H2O

penkvilksite

Kola Peninsula, Russia

Na4Ti2Si8O22‚4H2O

Finally, an interpretation of the catalysis chemistry is one of the most important of the parameters that can be used to characterize titanozeolites. These catalysts are interesting selective oxidation catalysts, converting phenol to catechol and hydroquinone and alkanes to simple alcohols and ketones using H2O2 as the oxidant.19,38-40 It is also reported that large organic peroxides are ineffective in this catalytic reaction. Several groups have incorporated Ti in microporous sieves isomorphous with MCM-41 to overcome the pore size limitations of MFI and MEL zeolites.41,42 We also find that poor syntheses of titanozeolites (those with anatase-like phases) simply convert H2O2 to water and O2 without oxidizing the organic substrate. All of this chemical information indicates that the Ti sites in the titanozeolites must be inside the pores and must be site isolated. As suggestive as the evidence is, however, it is not direct evidence for structural titanium. Experimental Section Materials. The titanium silicates used in this study are representative of the Q4 silicon environment, the superscript indicating the number of silicon or other cations attached to the SiO4 tetrahedron. The titanium-silicate mineral samples studied were titanite, benitoite, fresnoite, narsarsukite, lorentzenite, vinogradovite, zorite, and penkvilksite. We did not obtain a sample of natural fresnoite so the NMR data are from a publication by Dirken et al.43 Both zorite and penkvilksite are relatively rare minerals. The samples used in this work were loaned to us by researchers at Engelhard Corp. The narsarsukite, lorentzenite, and vinogradovite were obtained from the Excalibur-Cureton Co. in Peekskill, NY. The titanite and benitoite were purchased from David New, Anacortes, WA. We also studied two synthetic titanosilicate materials: the porous ETS10 and the new layered titanosilicate material, JDF-L1. These two materials were generously provided by the Engelhard Corp. All samples were used without further characterization. We have obtained spectra of numerous different TS-1 samples synthesized in our laboratories using slightly different synthetic routes. The TS-1 spectrum in Figure 1 was obtained from a sample synthesized using a synthetic procedure by Thangaraj and Sivasanker.44 However, the spectra are all nearly the same in TS-1 samples with good X-ray powder diffraction patterns and which have good catalytic activity. In some samples the peak at -101 to -103 ppm, attributed to defect silanols, is larger. We also obtained the spectrum of a single sample of TS-2 which was prepared using the synthesis reported by Reddy et al.45 For comparison, we include calculations and NMR data for siliceous samples of the zeolites ZSM-5, ZSM-12, and ferrierite. The sources and properties of the samples are given

SiO4 environment

av SiO bond length (Å)

av TiO bond length (Å)

δ exp (ppm)

δ calc (ppm)

0Si,4Ti 2Si,2Ti 1Si,3Tisemioct 2Si,2Ti 3Si,1Ti 2Si,2Ti

1.64 1.62 1.615 1.632 1.62 1.63

1.96 1.96 1.93 1.99 1.97

-78.5 -93.4 -82.0 -90.0 -94.2 -90.6

-77.3 -90.3 -83.4 -87.6 -95.8 -89.5

3Si,1Ti 2Si,2Ti

1.66 1.62

1.97 1.96

N/A -90.2

-96.1 -87.5

3Si,1Tisemioct 2Si,2Ti 3Si,1Ti

1.64 1.63

1.89 2.00

N/A -95.6 -100.0

-92.7 -93.7 -98.2

in Table 1. The analytical formulas were obtained from typical literature values for these minerals. NMR Spectroscopy. The 29Si MAS spectra were obtained at room temperature on a Varian Unity-400 spectrometer at a frequency of 79.452 MHz. A Varian MAS probe with 7 mm (outside diameter) rotors was used. The typical π/2 pulse width was 6 µs, and a 7 s recycle delay was used. The spinning rate was typically 6 kHz. Chemical shifts are reported relative to TMS but were actually measured by referencing to internal poly(dimethylsiloxane), which has a chemical shift of -22.3 ppm relative to TMS. The number of scans varied from 208 to 3400 depending on the particular sample. We used 8264 scans for the zorite sample because only a very small quantity of this material was available. The linear regressions for correlating chemical shifts with structures were performed using IGOR, a general mathematical package, on a Macintosh computer. Equations for Predicting 29Si Chemical Shifts. The chemical shift in NMR is the result of the detailed electronic structure surrounding the nucleus of interest. With modern quantum chemical methods, it is possible to calculate such shifts to reasonable accuracy, especially 13C shifts where the majority of this research has concentrated. However, this is timeconsuming and often not an option to the experimentalist.4 Recognizing that the detailed structure of the molecule also reflects the electronic structure, it is likely that a simple empirical formula might be obtained to predict shifts from structural parameters, and vice versa. As mentioned above, numerous groups have attempted such correlations with varying degrees of success. If one constrains the correlation to a limited number of very similar compounds, it is possible to obtain an excellent correlation. However, with varied compounds and with different substituent atoms the correlation often breaks down. Sherriff and co-workers published two papers correlating measured 29Si chemical shifts with the structure of silicate minerals.1,2 In our hands, their equations work reasonably well, but since these papers are not generally recognized, we will review the equations and their rationale. The original concepts go back to early work by McConnell46 and have been used to predict proton NMR shifts in multiply bonded organic molecules and organometallic complexes.47 When a molecule is placed in an external magnetic field, the electrons in bonds adjacent to the Si of interest will circulate, producing internal dipolar fields that result in chemical shifts. Figure 2 is a reproduction of the diagram used by Sherriff et al. and is used to define the terms used in the equations below. It is a schematic representation of one of the four substituents in a Si(OX)4 tetrahedron and will be used to explain the ideas of the structure-chemical shift correlations.

2900 J. Phys. Chem. B, Vol. 102, No. 16, 1998

Labouriau et al.

Figure 2. Schematic drawing of the angles θ and R and the lengths R and D for the Si-O-X unit, where X is any neighbor cation. The notation for this figure is the same as used by Sherriff et al.1,2

The largest effect on the chemical shift comes from the electrons in the O-X bond and is affected by the valence of that bond. Altermatt and Brown48 correlated an empirical valency, si, in terms of structural parameters for 1000 structures in the “Inorganic Crystal Structure Database” and arrive at the following operational definition:

si ) exp[(r 0 - ri )/0.37]

(1)

where ri is each O-X bond length and r0 is an average bond length for a given O-X pair. The values 0.37 and r0 are obtained from their fit, and r0 can be found for a large number of inorganic pairs in their paper. These concepts can be used to loosely define the O-X bond as a magnetic dipole with the following influence at the Si nucleus:

(dipole term)i ) si(1 - 3 cos2 θi)/(3Ri3)

(2)

where Ri is the distance to the center of the dipole defined as the distance between the silicon atom and the midpoint of the O-X bond. The angle θi is calculated from Ri and the SiO bond length (see Figure 2). This simple dipole equation is insufficient to predict chemical shifts from structure; it also requires a term to account for changes in hybridization when the Si-O-X angle is far from the average of 143° found in many silicates.49

(hybridization term)i ) (cos R i/(cos Ri - 1)) The final geometrical term is

Ω′ )

∑i

{[ si

][

]}

1 - 3 cos2 θ

cos Ri

3Ri3

cos Ri - 1

(3)

(4)

This describes the structure around the Si atom that affects the 29Si chemical shift. Finally, the structural factor can be correlated with measured chemical shifts in a linear regression. Sherriff and co-workers have done this for over 100 silicon sites in minerals and synthetic compounds using chemical shifts that they measured, chemical shifts from the literature, and structures available from several databases.1 They obtain an excellent correlation with

δ ) 650.08Ω′ - 56.06

(5)

In a second paper, Sherriff et al. used a slightly different equation, inserting a different function for the hybridization term:2

(hybridization term)i ) log Di

(6)

where Di is each Si-X distance (see Figure 2). With this hybridization term, their equation correlating the chemical shift with the structure is

δ ) 701.6Ω′′ - 45.7

(7)

Figure 3. Experimental 29Si chemical shifts versus calculated chemical shifts for (9) ZSM-5, (2) ZSM-12, and (b) ferrierite.

where

Ω′′ )

∑i

{[

] }

1 - 3 cos2 θi

si

3Ri3

log(Di)

(8)

It is our experience that the hybridization term described by eq 3 provides slightly better correlations than the one using the log(Di) function. Results and Discussion The quality of eq 4 will be demonstrated using the purely siliceous zeolites ZSM-5, ZSM-12, and ferrierite because the equation was derived by Sherriff et al.,1 primarily for dense minerals. Since zeolites and other porous materials have somewhat flexible frameworks, we want to demonstrate that the structure-shift correlations work well in these materials. Additionally, they had to make corrections to their structural parameters to account for silicon-aluminum disorder in some of the samples. The reason that we choose ZSM-5, ZSM-12, and ferrierite is that these three zeolites have excellent diffraction structures in the literature,50-55 and their highly resolved 29Si NMR spectra are also published.10,54-56 We believe that the final quality of the chemical shift-structure correlation is dependent upon the quality of the diffraction-derived structure. More recent structures tend to be better because of improvements in instrumentation and software. Single-crystal structures are normally more accurate than those obtained by Rietveld analysis of fine powders. Figure 3 is a plot of the observed vs calculated 29Si chemical shift for the three zeolites using eq 5 with r0 ) 1.60 Å. This value is smaller than the value of r0 used for titanosilicates (1.63 Å) because the mean Si-O bond length in highly siliceous zeolites is about 1.60 Å. The quality of the fit of the experimental data to eq 5 is not as good as the correlation obtained by Fyfe and co-workers for ZSM-5 and ZSM-12.10 Fyfe et al. fit the measured 29Si chemical shifts to several different equations, but the two with the best fits contained either (1) the mean Si-Si distance or (2) the average of cos R/(cos R - 1), where R is the Si-O-Si bond angle. They performed linear regressions for each zeolite separately and obtained different fitted parameters for each zeolite. Thus, the set of chemical shifts (and corresponding geometries) is much smaller for each of their regressions. Although the quality

29Si

MAS NMR of Titanosilicates

Figure 4. 29Si MAS NMR spectra of titanosilicates samples: (a) titanite, (b) lorenzenite, (c) benitoite, (d) zorite (the peak at -75 ppm is a sideband from poly(dimethylsiloxane) which is used as an internal chemical shift reference), and (e) penkvilksite.

of the fit is good, this method lacks generality because their results only apply to the particular zeolite for which they were derived. In other words, the equations cannot be accurately applied to some unknown structure to assign 29Si chemical shifts or conversely to derive information about the structure from measured chemical shifts. This same problem exists with most of the other shift-structure correlations in the literature; the set of samples is very limited. On the other hand, Sherriff et al.’s work may be slightly less precise, but it has more general utility in predicting structures from chemical shifts, or vice versa. The average standard deviation between the measured and calculated chemical shifts for the three zeolites is 1.5 ppm. This is reasonable and useful for predictive purposes. However, eq 5 cannot be applied indiscriminately. The quantity r0, reported by Altermatt and Brown48 and earlier by Brown and Shanon,57 covers a range of values. It is clear that these bond lengths have certain trends with different types of inorganic materials. In other words, the mean Si-O bond length in siliceous zeolites is different from that in the polymorphs of silica. The experimentalist, using these equations, must use reasonable judgment in selecting the input parameters. Figure 4 shows 29Si MAS NMR spectra of the titanosilicate samples investigated. We studied two different samples of natural titanite (CaTiOSiO4), one from Pakistan and one from Canada. In titanite there is one tetrahedral Si in the asymmetric unit. It shares oxygen atoms with four separate TiO6 octahedra in three different chains. The average Si-O bond is 1.64 Å, and the six Ti-O bond distances are 1.77, 1.97, 1.98, 1.99, 2.01, and 2.03 Å.58 The 29Si MAS NMR spectra of both samples show one peak at -78.5 ppm. So we assign peaks in this region to SiO4 tetrahedra attached to four Ti atoms. The spectrum from sample from Pakistan had somewhat better signal-to-noise, presumably because of paramagnetic (iron) impurities in the Canadian sample. Hawthorne et al.59 reported 29Si MAS spectra for several titanite samples with different amounts of iron and other impurities. They measured chemical shifts between -74 and -79 ppm with the best sample having a shift of -78.9 ppm. This is very close to our measured value. Fresnoite (Ba2TiOSi2O7) consists of tetrahedral corner-shared silica dimers, [Si2O7] 6-, and titanium-centered square pyramids linked to form flat sheets. The Ti-O bond distances are 2.00 ( 0.04 Å for the four oxygen atoms that form the base of the pyramid and 1.66 ( 0.08 Å for the fifth oxygen atom. The tetrahedral SiO4 is attached to three titanium atoms and one silicon atom. The Si-O distance for those oxygen atoms associated with the Ti-centered square pyramid is 1.61 Å, and

J. Phys. Chem. B, Vol. 102, No. 16, 1998 2901 the “apical” oxygen is 1.59 Å. The Si-O distance for the oxygen atom shared by the Si tetrahedra is 1.65 Å. The average Si-O distance is 1.615 Å.60 Unfortunately, we were not able to obtain a natural fresnoite sample for this study. However, Dirken et al. published the 29Si MAS NMR spectrum for a synthetic sample, which shows a single resonance at -82 ppm.43 The 29Si MAS NMR spectrum for lorenzenite (Na2Ti2Si2O9) has a single resonance at -90 ppm. The SiO4 tetrahedra in this mineral form pyroxene-type chains in the direction of the b axis by sharing two oxygens with neighboring silica tetrahedra. The two remaining oxygens are shared by two Ti atoms. The average Si-O bond length is 1.632 Å. The resonance observed at -90 ppm is thus associated with an SiO4 tetrahedron attached to two Ti atoms.61 In benitoite, BaTiSi3O9, there is one Si site that is surrounded by four oxygens in a pseudotetrahedral arrangement. The space group symmetry repeats the tetrahedron to form a threemembered cyclosilicate ring (Si3O9). These rings are linked by an almost regular (TiO6) octahedron. The average bond lengths are 1.96 Å for the Ti-O bond and 1.62 Å for the Si-O bond.62 The 29Si MAS NMR spectrum for this sample shows a single resonance at -93.4 ppm. This peak is assigned to an SiO4 tetrahedron attached to two Ti atoms and two Si atoms. The 29Si MAS NMR spectrum for narsarsukite (Na2TiOSi4O10) shows a single resonance at -94.2 ppm. The structure of narsarsukite can be described as made up of tubes having composition Si4O10 parallel to the c axis. The parallel Si4O10 tubes are linked together by chains of titanium octahedra which also run parallel to the c axis. The silicon tetrahedron is regular, with average Si-O distance of 1.62 Å. The octahedron of oxygen atoms surrounding a titanium atom has Ti-O bond lengths of 1.904 and 2.07 Å and four bonds of 1.966 Å.63 The peak observed at -94.2 ppm is associated with the SiO4 tetrahedron attached to one Ti atom and three Si atoms. In vinogradovite, Na2Ti2O2Si2O6(Si,Al)2O5‚(H2O,K2O), there are two different silicon-oxygen chains: the dimetasilicate band [Si4O10] 8 and the single pyroxene metachain [Si2O6]. The two crystallographically inequivalent Si atoms form tetrahedra of different sizes. The mean Si-O distance is 1.626 Å for the Si2O6 chains and 1.660 Å for the Si4O10 chains. The Si1 tetrahedra (pyroxene chains) are linked to two Ti atoms and two Si atoms, whereas the Si2 tetrahedra (dimetasilicate chains) are linked to one Ti atom and three Si atoms. There is one type of Ti atom in the usual octahedral coordination to oxygen with Ti-O bond lengths of 1.815, 1.868, 1.938, 1.953, 2.139, and 2.128 Å.64 Vinogradovite is a porous mineral with two types of channels: one small pore, occupied by Na+ cations, and a larger one containing H2O and K2O.64 The 29Si MAS NMR spectrum for this sample shows a single broad resonance at -90.6 ppm. We associate this resonance with Si1 tetrahedra that are condensed into pyroxene chains. We were unable to detect the signal associated with the Si2 tetrahedra. It is known that there is extensive aluminum substitution in the Si2 site of this mineral. The lower concentration of Si and disorder in bond lengths and angles due to this substitution is a probable cause for a loss in signal intensity and broadening which explains our inability to observe this expected resonance. Zorite (Na6[Ti(Ti0.9Nb0.1)4(Si6O17)2(O,OH)5]‚11H2O) is another mineral where we only observe one of two expected 29Si NMR resonances. The porous framework of zorite is made up of xonotlite-like bands [Si6O17] joined by Ti semioctahedra and octahedral [(Ti,Nb)2O10] chains. There is a significant amount

2902 J. Phys. Chem. B, Vol. 102, No. 16, 1998 of niobium substituted for the Ti in these octahedral chains. Zorite contains two types of channels with minimum cavity diameters of 4.3 and 4.6 Å. The framework is filled with Na+ cations and water molecules which support the structure.65 There are two crystallographically different Si sites with average Si-O bond lengths of 1.62 and 1.64 Å. The structure contains a Ti-O square pyramid and a Ti-O octahedron. The bond lengths for the Ti pyramid are four bonds of 1.94 Å and one bond of 1.67 Å, and the Ti octahedron has four bonds of 1.98 Å and two of 1.93 Å. The 29Si MAS NMR spectrum for zorite has only one resonance at -90.8 ppm. From the chemical shift, we assign this peak to the SiO4 tetrahedron attached to two Ti atoms and two Si atoms. We believe that the peak which should be obtained for the second Si site is not detectable because of high aluminum substitution on this site. Anderson also only observed a single resonance in zorite. Their explanation was that there is paramagnetic substitution (Fe) in the mineral that destroys the second peak. Our experience is that paramagnetic metals in the structure and pores of many materials can have profound effects on the 29Si NMR spectra. However, we have never found spatial selectivity that would result in the loss of the NMR peak associated with one crystallographic site and not another one in close proximity. We note that zorite is a relatively rare mineral; both Anderson and ourselves obtained spectra from very small samples, and the resultant signal-to-noise is not sufficient to draw very strong conclusions about the details of a weak or nonexistant peak. Finally, the 29Si MAS NMR spectrum for penkvilksite (Na4Ti2Si8O22‚4H2O) has two resonances: one at -95.6 ppm assigned to SiO4 tetrahedra attached to two Ti atoms and two Si atoms and another resonance at -100.0 ppm assigned to the SiO4 tetrahedron attached to one Ti atom and three Si atoms.66 All measured chemical shifts are listed in Table 1. We also included chemical shift and structural parameters for the synthetic materials ETS-10 and JDF-L1 in the correlations. A crystal structure model of ETS-10 was proposed by Anderson et al., based on structural modeling, chemical analysis, electron diffraction, HREM, and 29Si NMR chemical shifts.67,68 This structure comprises corner-sharing SiO4 tetrahedra and TiO6 octahedra linked through bridging oxygen atoms. The 29Si MAS NMR spectrum shows four lines with chemical shifts at -94.1, -95.8, -96.5, and -103.3 ppm. Anderson assigned the first three resonances to the SiO4 tetrahedra connected to three silicon atoms and one titanium atom and the resonance at -103.3 ppm to the SiO4 tetrahedra connected to four silicon atoms. Samples of ETS-10 are invariably a mixture of two polymorphs: polymorph A which contains 20 T-sites for silicon and polymorph B which has 11 T-sites. In our calculations using eq 5, we considered both polymorphs giving a total of 31 chemical shifts. Generally, we find chemical shifts for Si connected to one titanium, i.e., 3Si,1Ti, in the range between -102 and -105 ppm and chemical shifts associated with Si(4Si,0Ti) environments at about -116 ppm. These are significantly different from the shift/assignments proposed by Anderson and coworkers. There are several potential reasons for this discrepancy between experimental and calculated chemical shifts. The first reason may be that the structure is not quantitatively correct. There are two polymorphs of ETS-10 with considerable disorder. Disorder makes it difficult to measure structural parameters to the accuracy required to get good agreement with the NMR shifts. We use an average bond length, r0, of 1.60 Å for X ) Si in eq 5. The Si-O bond lengths in Anderson’s structure of ETS-10 ranges from 1.58 to 1.59 Å. If real, this difference can account for some of the discrepancy in the calculated shifts.

Labouriau et al.

Figure 5. Experimental 29Si chemical shifts of the titanosilicates plotted as a function of the structural parameter, Ω. The symbol 0 and associated straight line is the data for eq 5, using cos R/(cos R - 1), and O is the data for eq 7 using log Di for the hybridization term. The minerals are symbolized by (a) titanite, (b) fresnoite, (c) lorenzenite, (d) zorite, (e) vinogradovite, (f) benitoite, (g) narsarsukite, and (h) penkvilksite.

The most likely source of the difference between the measured and calculated shifts is the fact that ETS-10 is microporous and requires a large number of Na+ cations and water molecules for charge balance. Anderson did not find the cations in his structural determination, either because he did not solve for them or because they are mobile. As mentioned above, in many cases we must include cations in our calculations to get good correlations of chemical shifts with eq 5. Preliminary work at Los Alamos and Engelhard on the neutron powder diffraction structure of ETS-10 shows excess scattering in the pores which is attributed to the cations. These preliminary results indicate that the proposed topology is correct. However, the detailed structure probably contains enough error to explain the observed differences between computed and measured chemical shifts. The new material, JDF-L1 (Na4Ti2Si8O22‚4H2O), is a layered titanosilicate.69 For our purposes, it is interesting because the Ti atoms are five-coordinate, and we would like to evaluate the generality of the regression equations when the “X atom” has different coordination. Each titanium is linked to SiO4 tetrahedra forming the continuous sheets. The 29Si MAS NMR spectrum contains a single peak at -107.2 ppm. This is in the correct region for a SiO4 tetrahedron attached to one Ti atom and three Si atoms. The calculated chemical shift is -105.4 ppm. The difference of 1.8 ppm between computed and measured shifts is very satisfying. Another interesting result in this sample is that we see a second, small 29Si NMR resonance at -90.1 ppm from our sample. In a later paper, Du et al. claim that zorite is a frequent impurity in syntheses of JDF-L1.70 This resonance is at exactly the correct position for the major peak of zorite. Figure 5 is a plot of the measured chemical shifts for the titanosilicate samples versus the structural parameter Ω′. The structural quantities necessary to calculate Ω′ were obtained by entering the crystal structures in the molecular modeling program, Cerius2, and extracting bond lengths and bond angles. The measured chemical shifts are correlated to the structural parameters using both of the equations proposed by Sherriff et al. (eqs 4 and 6). We used the values of r0 ) 1.63 Å when the X atom is Si and r0 ) 1.86 Å when X ) Ti (Figure 2). The straight lines in Figure 5 are plots of eqs 5 and 7. The numerical constants in these equations are those proposed by Sherriff and

29Si

MAS NMR of Titanosilicates

Grundy, not new regression equations. This demonstrates the generality of these equations for computing 29Si chemical shifts. The structural parameters Ω′ and Ω′′ are the sum of contributions from all of the Si-O-X next-nearest neighbors. Normally this would be the sum over four contributions because silicon is usually four-coordinate (nominally tetrahedral). However, in many cases there are other metal cations in the mineral structure, e.g., Ba in benitoite, and contributions from these must be included for the correlation to be very good. These additional terms in the summation may include atoms 3-4 Å away from the silicon In eqs 4 and 8, it can be seen that the contributions get smaller as one gets more distant from the silicon so there is a natural convergence. The chemical shift contribution from these, usually distant, cations is only a few parts per million. We now return to the original question of understanding 29Si NMR results for TS-1 and other titanozeolites. The above resultsscorrelations of the chemical shifts with well-determined structures of both siliceous zeolites and titanosilicate minerals with varying composition and structuresare very satisfying. Equation 5 represents a good correlation for computing silicon shifts from the known structure of titanosilicates of varying composition and structure. The correlations between structure and chemical shift are within 1.5 ppm. There are a few cases that are outside these limits. They are due to inaccuracies in the diffraction determined structures or due to cations that are not included in the correlation. We would like to use eq 5 to compute the 29Si NMR spectrum of TS-1; however, there are several different ways to do this. It is clear that we cannot simply use the established structure of siliceous ZSM-5 to obtain atomic positions because the unit cell dimensions of TS-1 increase with increasing titanium content, and we expect the Ti-O bond to be longer than an Si-O bond. To get a qualitative idea of the chemical shift, we have taken the SiO-Ti geometry from narsarsukite because this mineral contains a silicon with a single titanium as next-nearest neighbor. Then we replace one silicon next-nearest neighbor in each of the 12 T-sites with the narsarsukite geometry. Since there are four next-nearest neighbors, this results in 48 different structures. The chemical shift for each of these structures can be calculated using eq 5. This results in chemical shifts of 5-10 ppm downfield of the shift in the purely siliceous material with the exact shift being different for each T-site. Now considering the lower spectrum in Figure 1, the most upfield peak (at about -116.3 ppm) would fall under the main peaks if the Ti shift were 5 ppm but would actually be resolved at the other extreme of 10 ppm. Peaks in the center or toward the left edge of the spectrum, if they contained titanium, would be separated from the rest of the spectrum even with a 5 ppm shift due to titanium as next-nearest neighbor. We also measured the 29Si MAS spectrum of TS-2. The sample has an X-ray powder pattern consistent with the MEL structure with no peaks for anatase or silica. The 29Si NMR spectrum consists of two relatively broad, featureless peaks centered at -103 and -114.3 ppm. The lowfield peak is assigned to silanol groups in defects or on the surface of the crystallites. The up-field one is assigned to Q4 units in the zeolite structure. If we make assumptions similar to those above (for TS-1), we expect a peak between the two observed for silicon atoms that are next-nearest neighbor to titanium. No such peak is seen. The results on titanium silicalite samples present a dilemma: With so much evidence that titanium is part of the structure of titanium zeolites, why do we see no 29Si NMR peak for those silicon atoms attached (through oxygen) to titanium? An unlikely possibility is that titanium occupies a nonframework

J. Phys. Chem. B, Vol. 102, No. 16, 1998 2903 position, and site isolation as well as association with the zeolite gives the titanium its unusual catalytic properties. A second possibility is that there is an error in our correlations that makes it impossible for us to calculate shifts for titanosilicalites. A third possibility is that the titanium has a dynamic equilibrium of its coordination number, e.g., between 4- and 5-coordinate, or there is some water dynamics that result in an average shift for silicon that is very close to the shift in pure silicalite. It is clear that titanium silicalites elude an understanding of the titanium structure by using normal 29Si MAS NMR. A recent publication by Camblor et al.71 describes the preparation of low defect, aluminum-free titanium zeolite β. This sample will make an interesting comparison with the TS-1 and TS-2 spectra reported here and alluded to in many publications in the literature. Conclusions Our original goal in this work was to understand the fact that there is no peak in the MAS 29Si NMR spectrum of TS-1 that corresponds to silicon atoms that are next-nearest neighbor to titanium in the structure. With all of this work, an explanation still eludes us. There are a number of other spectroscopic and analytical techniques that infer that titanium is in the zeolite structure. At this time we believe that if titanium is tetrahedrally bound in TS-1, there must be something about the detailed structure that we do not understand. We continue to seek NMR spectroscopic and other means of resolving this issue. There is a long history of using NMR chemical shifts to obtain structural information. In a simplistic sense, this is the utility of routine NMR as an analytical tool. In the solid state, the molecules are “locked” into a single conformation by steric constraints from their neighbors. This is ideal for correlating detailed structural information with the NMR shift measurement. After some consideration of the quality of the correlations obtained in this paper and others in the literature, we believe that the quality of diffraction derived structures is one of the limiting factors in these structure-shift correlations. Of all of the papers on this subject, we find that the work of Sherriff and Grundy is the most general. They derived equations using pure silicates and aluminosilicates. In our work, we have demonstrated that it is possible to take their equations and apply them to situations where silicon is next-nearest neighbor to titanium. It appears that these equations work well for titanium in both six- and five-coordinate environments. However, there are still a very few samples in our repertoire that seem not to behave as expected. One of these is TS-1 which shows no NMR resonance where it might be expected. We also find that other cations in the structure can influence the silicon shifts, even when they seem relatively distant from the Si nucleus. For this reason, these shift correlations must be used judiciously in porous or dynamic systems where extraframework cations can cause 29Si shifts and where their location in the structure is often not known. Acknowledgment. We are grateful for collaborations with Engelhard Corp. In particular, we thank G. Koermer for interesting and useful discussions and for the samples of zorite and penkvilksite and S. Kusnicki for providing the sample of JDF-L1. This work was performed as part of the Los Alamos Catalysis Initiative supported by the Department of Energy under Contract W-7405-ENG-36. References and Notes (1) Sherriff, B. L.; Grundy, H. D. Nature 1988, 332, 819.

2904 J. Phys. Chem. B, Vol. 102, No. 16, 1998 (2) Sherriff, B. L.; Grundy, H. D.; Hartman, J. S. Eur. J. Mineral. 1991, 3, 751. (3) Tuel, A.; Ben Taarit, Y. J. Chem. Soc., Chem. Commun. 1992, 1578. (4) Moravetski, V.; Hill, J.-R.; Eichler, U.; Cheetham, A. K.; Sauer, J. J. Am. Chem. Soc. 1996, 118, 13015. (5) Lippmaa, E.; Magi, M.; Samoson, A.; Engelhardt, G.; Grimmer, A. R. J. Am. Chem. Soc. 1980, 102, 4889. (6) Lippmaa, E.; Magi, M.; Samoson, A.; Tarmak, M.; Engelhardt, G. J. Am. Chem. Soc. 1981, 103, 4992. (7) Magi, M.; Lippmaa, E.; Samoson, A.; Engelhardt, G.; Grimmer, A. R. J. Phys. Chem. 1984, 88, 1518. (8) Smith, J. V.; Blackwell, C. S. Nature 1983, 303, 223. (9) Thomas, J. M.; Kennedy, J.; Ramdas, S.; Hunter, B. K.; Tennakoon, D. T. B. Chem. Phys. Lett. 1983, 102, 158. (10) Fyfe, C. A.; Feng, Y.; Grondey, H. Microporous Mater. 1993, 1, 393. (11) Engelhardt, G.; Radeglia, R. Chem. Phys. Lett. 1984, 108, 271. (12) Radeglia, R.; Engelhardt, G. Chem. Phys. Lett. 1985, 114, 28. (13) Newsam, J. M. J. Phys. Chem. 1987, 91, 1259. (14) Sherriff, B. L.; Singh, V.; Liang, J.; Grundy, H. D. Comput. Geosci. 1991, 17, 967. (15) Engelhardt, G.; Michel, D. High-Resolution Solid-State NMR of Silicates and Zeolites; John Wiley & Sons: New York, 1989. (16) Maciel, G. E.; Sindorf, D. W. J. Am. Chem. Soc. 1980, 102, 7606. (17) Dessau, R. M.; Schmitt, K. D.; Kerr, G. T.; Woolery, G. L.; Alemany, L. B. J. Catal. 1988, 109, 472. (18) Dessau, R. M.; Schmitt, K. D.; Kerr, G. T.; Woolery, G. L.; Alemany, L. B. J. Catal. 1987, 104, 484. (19) Notari, B. Stud. Surf. Sci. Catal. 1988, 37, 413. (20) Kaliaguine, S.; Nagy, J. B.; Gabelica, Z. Stud. Surf. Sci. Catal. 1988, 35, 381. (21) Fyfe, C. A.; Strobl, H.; Kokotailo, G. T.; Kennedy, G. J.; Barlow, G. E. J. Am. Chem. Soc. 1988, 110, 3373. (22) Conner, W. C.; Vincent, R.; Man, P.; Fraissard, J. Catal. Lett. 1990, 4, 75. (23) Zecchina, A.; Spoto, G.; Bordiga, S.; Ferrero, A.; Petrini, G.; Leofanti, G.; Padovan, M. Stud. Surf. Sci. Catal. 1991, 69, 251. (24) Huybrechts, D. R. C.; Buskens, P. L.; Jacobs, P. A. J. Mol. Catal. 1992, 71, 129. (25) Camblor, M. A.; Corma, A.; Perez-Pariente, J. J. Chem. Soc., Chem. Commun. 1993, 557. (26) Deo, G.; Turek, A. M.; Wachs, I. E.; Huybrechts, D. R. C.; Jacobs, P. A. Zeolites 1993, 13, 365. (27) Klaas, J.; Kulawik, K.; Schulz-Ekloff, G.; Jaeger, N. Stud. Surf. Sci. Catal. 1994, 84, 2261. (28) Klaas, J.; Schulz-Ekloff, G.; Jaeger, N. I. J. Phys. Chem. B 1997, 101, 1305. (29) Waychunas, G. A. Am. Mineral. 1987, 72, 89. (30) Dumas, T.; Petiau, J. J. Non-Cryst. Solids 1986, 81, 201. (31) Babonneau, F.; Doeuff, S.; Leaustic, A.; Sanchez, C.; Cartier, C.; Verdaguer, M. Inorg. Chem. 1988, 27, 3166-3172. (32) Lopez, A.; Kessler, H.; Guth, J. L.; Tuilier, M. H.; Popa, J. M. Proc. 6th Int. Conf. X-ray Abs. Fine Struct. 1990, 549. (33) Behrens, P.; Felsche, J.; Vetter, S.; Schulz-Ekloff, G.; Jaeger, N. I.; Niemann, W. J. Chem. Soc., Chem. Commun. 1991, 678. (34) Davis, R. J.; Liu, Z.; Tabora, J. E.; Wieland, W. S. Catal. Lett. 1995, 34, 101. (35) Bordiga, S.; Coluccia, S.; Lamberti, C.; Marchese, L.; Zecchina, A.; Boscherini, F.; Buffa, F.; Genoni, F.; Leofanti, G.; Petrini, G.; Vlaic, G. J. Phys. Chem. 1994, 98, 4125. (36) Trong, O. D.; Bittar, A.; Sayari, A.; Kaliaguine, S.; Bonneviot, L. Catal. Lett. 1992, 16, 85. (37) Lopez, A.; Tuilier, M. H.; Guth, J. L.; Delmotte, L.; Popa, J. M. J. Solid State Chem. 1993, 102, 480.

Labouriau et al. (38) Tatsumi, T.; Nakamura, M.; Negishi, S.; Tominaga, H. J. Chem. Soc., Chem. Commun. 1990, 476. (39) Tuel, A.; Ben Taarit, Y. Appl. Catal. A 1993, 102, 69. (40) Kraushaar-Czarnetzki, B.; van Hooff, J. H. C. Catal. Lett. 1989, 2, 43. (41) Corma, A.; Navarro, M. T.; Perez-Pariente, J. J. Chem. Soc., Chem. Commun. 1994, 147. (42) Tanev, P. T.; Chibwe, M.; Pinnavaia, T. J. Nature 1994, 368, 321. (43) Dirken, P. J.; Smith, M. E.; Whitfield, H. J. J. Phys. Chem. 1995, 99, 395. (44) Thangaraj, A.; Sivasanker, S. J. Chem. Soc., Chem. Commun. 1992, 123. (45) Reddy, J. S.; Kumar, R.; Ratnasamy, P. Appl. Catal. 1990, 58, L1. (46) McConnell, H. M. J. Chem. Phys. 1957, 27, 226. (47) McGlinchey, M. J.; Burns, R. C.; Hofer, R.; Top, S.; Jaouen, G. Organometallics 1986, 5, 104. (48) Altermatt, D.; Brown, I. D. Acta Crystallogr. 1985, B41, 244. (49) Gibbs, G. V. Am. Mineral. 1982, 67, 421. (50) van Konigsveld, H.; Tuinstra, F.; van Bekkum, H.; Jansen, J. C. Acta Crystallogr., Sect. B 1989, 45, 423. (51) van Konigsveld, H. Acta Crystallogr., Sect. B 1990, 46, 731. (52) van Konigsveld, H.; Jansen, J. C.; van Bekkum, H. Zeolites 1990, 10, 235. (53) Fyfe, C. A.; Gies, H.; Kokotailo, G. T.; Marler, B.; Cox, D. E. J. Phys. Chem. 1990, 94, 3718. (54) Lewis, J. E., Jr; Freyhardt, C. C.; Davis, M. E. J. Phys. Chem. 1996, 100, 5039. (55) Morris, R. E.; Weigel, S. J.; Henson, N. J.; Bull, L. M.; Janicke, M. T.; Chmelka, B. F.; Cheetham, A. K. J. Am. Chem. Soc. 1994, 116, 11849. (56) Fyfe, C. A.; O’Brien, J. H.; Strobl, H. Nature (London) 1987, 326, 281-283. (57) Brown, I. D.; Shannon, R. D. Acta Crystallogr. 1973, A29, 266. (58) Mongiorgi, R.; Sanseverino, L. R. Mineral. Petrogr. Acta 1968, 14, 123. (59) Hawthorne, F. C.; Groat, L.; Raudsepp, M.; Ball, N. A.; Kimata, M.; Spike, F. D.; Gaba, R.; Halden, N. M.; Lumpkin, G. R.; Ewing, R. C.; Greegor, R. B.; Lytle, F. W.; Ercit, T. S.; Rossman, G. R.; Wicks, F. J.; Ramik, R. A.; Sherriff, B. L.; Fleet, M. E.; McCammon, C. Am. Mineral. 1991, 76, 370. (60) Moore, P. B.; Louisnathan, J. Science 1967, 1361. (61) Sundberg, M. R.; Lehtinen, M.; Kivekas, R. Am. Mineral. 1987, 72, 173. (62) Zachariasen, W. H. Z. Kristallogr. 1930, 74, 139. (63) Peacor, D. R.; Buerger, M. J. Am. Mineral. 1962, 47, 539. (64) Rastsvetaeva, R. K.; Andrianov, V. I. SoV. Phys. Crystallogr. 1984, 29, 403. (65) Sandomirskii, P. A.; Belov, N. V. SoV. Phys. Crystallogr. 1979, 24, 686. (66) Merlino, S.; Pasero, M.; Artioli, G.; Khomyakov, A. P. Am. Mineral. 1994, 79, 1185. (67) Anderson, M. W.; Terasaki, O.; Ohsuna, T.; Phillippou, A.; MacKay, S. P.; Ferreira, A.; Rocha, J.; Lidin, S. Nature 1994, 367, 347. (68) Anderson, M. W.; Terasaki, O.; Ohsuna, T.; O’Malley, P. J.; Phillippou, A.; MacKay, S. P.; Ferreira, A.; Rocha, J.; Lidin, S. Philos. Mag. B 1995, 71, 813. (69) Roberts, M. A.; Sankar, G.; Thomas, J. M.; Jones, R. H.; Du, H.; Chen, J.; Pang, W.; Xu, R. Nature 1996, 381, 401. (70) Du, H.; Fang, M.; Chen, J.; Pang, W. J. Mater. Chem. 1996, 6, 1827. (71) Camblor, M. A.; Costantini, M.; Corma, A.; Gilbert, L.; Esteve, P.; Martinez, A.; Valencia, S. Chem. Commun. 1996, 1339.