J. CHEM. SOC. FARADAY TRANS., 1994, 90(4), 675-682 675 Al-Pillared Saponites Part 2.7-NMR Studies Jean-Franqois Lambed,* Sophie Chevalier, Raymonde Franck, Helene Suquet and Denise Barthomeuf Laboratoire de Reactivite de Surface, URA 1106 CNRS Tour 54-55,2eme etage, Universite Pierre et Marie Curie 4, Place Jussieu, 75252 Paris Cedex 05,France A saponite from Ballarat was intercalated with Al polycations by four different procedures involving different sources of polycations, and stabilised by calcination (pillared). 29Si and 27AI solid-state MNR of both the inter- calated and the pillared samples have been measured and are discussed. 29Si NMR leads us to propose a pillaring mechanism, involving H+ attack of the clay tetrahedral sheet followed by AIO, tetrahedra inversion, similar to that previously proposed for beidellites. 27AI NMR reveals that the pillars are not fundamentally modified upon calcination at 500°C although they undergo reversible dehydration reactions; at 750 "C and above, a strong pillar reorganisation occurs prior to collapse of the global structure.In a companion paper,' we described the study of Al-pillared saponites by several techniques, including XRD, N, physi- sorption, thermoanalytical methods, and, mainly, IR spec- troscopy of both the solid matrix and adsorbed pyridine. We analyse here the results of solid-state NMR studies under- taken on the same materials. The importance of NMR for the characterisation of inter- calated and pillared clays stems from the lack of periodicity in the organisation of the interlayer, which precludes the use of diffraction methods.The literature on this topic up to 1988 has been reviewed by Fripiat;, since then, however, little fun- damental progress has been made while researchers concen- trated their efforts on the development of new pillared materials and intercalation procedures rather than on funda- mental understanding of the mechanisms involved. Ballarat saponite is a good material for NMR studies owing to its low structural iron content' which allows one to obtain well resolved NMR lines, not broadened by paramagnetic inter- action. Experimenta1 Starting Materials and Catalyst Preparation The mode of preparation of the samples used in this study is described in full detail in ref.1 and 3. The parent clay was a saponite from Ballarat, which was submitted to pillaring by A1 polycations from four different source solutions: A, a com- mercial Chlorhydrol solution, diluted to [Altotall= 0.1 mol 1-' and aged at 60 "C for 2 h (natural pH = 4.8); B, solution A adjusted to pH = 6 with concentrated NaOH immediately prior to use; C, solution A mixed with a 2 mol 1-' ammon-ium acetate solution (NH, :A1 = 15 : 1, final pH = 7.0); D, a solution obtained by addition of aqueous NaOH to AlCl, (final A1 concentration = 0.1 mol 1-', final pH = 5.0) and containing mainly the Keggin, Al13, polycation: this solution is denoted AHY. The resulting materials, after washing and drying, are termed intercalated samples and designated A, B, C and D, respectively.Portions of these samples were then heated to 500 or 750"C with an optimised temperature ramp. The resulting materials are called pillared samples and designated A500- D500 or A750-D750. In addition to the main experiments on Ballarat saponite, a sample of Otay montmorillonite (a dioctahedral clay with octahedral substitutions) was submitted to intercalation fol- t Part 1 :J. Chem. SOC., Faruduy Trans., 1994,90,667. lowing procedure D: the resulting sample is called MD, and the same sample calcined at 500 "C is MD500. "Si and "A1 Solid-state NMR 29Si MAS NMR spectra were recorded on an XC300 Bruker spectrometer operating at a B, field of 7 T (corresponding to a Larmor frequency of 59.62 MHz for 29Si) and in some cases on an MSL400 Bruker spectrometer (field 9.3 T, ,'Si Larmor frequency 79.5 MHz).We used a single pulse sequence with quadrature detection, a pulse length of 5 ps corresponding to a 42 flip angle, and a recycle time of 1 s (it was verified that longer recycle times did not modify the observed spectrum). The magic-angle sample rotation frequency was 3 to 4 kHz. Chemical shifts are referenced with respect to TMS at 0 ppm, with Q8M8 as a secondary standard. 29Si Spectra were decomposed into a sum of Lorentzian components for quan- tification. Contrary to initial expectations, a decomposition into Gaussian components gave poorer fits with the experi- mental data; the overall observed trends were, however, the same in both cases.27Al MAS NMR spectra were recorded on an MSL500 Bruker spectrometer operating at a B, field of 11.7 T, corre- sponding to a Larmor frequency of 130.29 MHz for 27Al, with a single pulse sequence. A short pulse-length of 0.6 ps was chosen so that the results could be quantitatively exploited [flip angle < .n/(21 + l)]." The recycle time was 0.5 s and the spinning frequency was 10 kHz or 5.5 kHz (some samples could not easily be spun at high speeds). Signal posi- tions are referenced with respect to a 0.1 mol 1-' solution of aluminium nitrate at 0 ppm, with yttrium aluminium garnet as a secondary standard. Several double rotation (DOR) spectra were also obtained on an experimental probehead for 27Al, with spinning fre- quencies of 1 kHz for the outer rotor and 4 kHz for the inner rotor.Results 29SiNMR The 29Si NMR spectrum of the starting saponite is shown in Fig. l(a). It consists of two peaks in the Q3 range, at -95.8 and -90.5 ppm. These can be attributed to silicon atoms in the tetrahedral layer: the peak at -95.8 ppm corresponds to Si with no aluminium neighbour (designated Q3-OAl) and the peak at -90.5 ppm to Si with one aluminium neighbour (designated Q3-lAl), in conformity with the results of Sanz and Serratosa.' The probability of having a silicon with two -9 5.7 I ~~ -8 0 -9 0 -100 6 Fig. 1 29Si Solid-state MAS NMR spectrum of the initial saponite SNa: (a) after drying at 60°C; (b)after calcining at 500 "C;(c)after calcining at 750 "C or three aluminium neighbours is too low to give rise to an observable peak.A decomposition of the observed signal into two Lorentz- ian lines gave intensity contributions of 71 & 1% for Q3-OAl and 29 1% for Q3-lAl. We may then calculate the Si/A1 ratio by using the well known formula? where n is the number of A1 neighbours and In denotes the intensity of the peak corresponding to Q3-nAl. This formula simplifies here into Si/Al = 3(Z0 + 11)/11, yielding Si/ A1 = 10.4, and a formula of (Si7,30A10.70)for the tetrahedral sheet, in good agreement with the chemical analysis which provides (Si7.26 A10.74). In Fig. l(b) and (c), it is shown that the spectrum of a saponite heated at 500°C does not undergo any noticeable modification, while in a saponite heated at 750"C both peaks are somewhat broadened, resulting in some loss of resolution but no change in the relative contributions of Q3-OAl and ~3-1~1.By comparison with the results on pillared saponites (cf: infra), we think that a small peak is present at -85 ppm, in the Q2region. J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 Fig. 2-4 represent the 29Si NMR spectra of saponites sub- mitted to A1 intercalation through procedures A, C and D, respectively, (a) immediately after intercalation and after heating at (b) 500 and (c)750 "C. The intercalated samples (i.e. those not submitted to a calcination step) show no per-ceivable modification in peak positions or relative intensities with respect to the spectrum of the initial Na-saponite.In contrast, for samples heated at 500°C [Fig. 2(b)], the peaks are only poorly resolved: the peak at -90 to -91 ppm now appears as a shoulder of the main peak at -95 to -96 ppm. A mathematical decomposition shows that the relative contribution to the signal of the -90 ppm peak has noticeably decreased, falling to 21% for sample D, 14% for sample A and 7% for sample C. Although the precision of these decompositions is estimated between f1.5 and f4%, we are confident that the decrease of the peak at -90 ppm is directly correlated to the amount of aluminium fixed in the interlayer space (see Table 1). This correlation is rather obvious on the spectra even without decomposition. The spectra of samples heated at 750°C show a more con- siderable loss of organisation, together with the appearance of a signal at -85 to -83 ppm, in the region of Q2 Si.This signal is especially prominent for sample D, which has the least fixed Al. 27~1NMR As expected, the starting saponite SNa shows only one 27A1 peak in the region corresponding to tetrahedral A1 [S,,, = +64.5 ppm, Fig. 5(a)],confirming the absence of aluminium 6 Fig. 2 29Si Solid-state MAS NMR spectrum of sample A (a) after A1 intercalation and drying at 60°C; (b)after calcining at 500°C; (c) after calcining at 750 "C J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 I I I -80 -90 -100 6 Fig. 3 29Si Solid-state MAS NMR spectrum of sample C: (a)-(c) as Fig. 2.0,spinning side bands (SSBs). in the octahedral sheet. Recording the spectrum under DOR conditions [Fig. 5(b)]resulted in a considerable narrowing of this line, indicating that its residual width under MAS is mainly due to second-order quadrupolar interactions. The spectra of samples B and D are shown in Fig. 6 and 7, respectively, showing in each case the intercalated sample, the pillared sample calcined at 500°C and the pillared sample calcined at 750°C. In addition, Tables 2 and 3 summarise Table 1 Correlation between the amount of fixed A1 and the rela- tive importance of the Q3-lAl 29Si NMR line integrated intensity fixed Al/ of Q3-lAl 29Si NMR line sample mmol (g clay)-' (Yo) SNa 0 29 (unpillared saponite) D 2.1 21 A 2.7 14 C 4.5 7 Likely error on integrated intensity (estimated from independent attempts at deconvolution): between f1.5% (unpillared saponite) and f4% (sample C, pillared).-9f.5 I 1 I I I 1-60 -80 -1 00 -120 6 Fig. 4 29Si Solid-state MAS NMR spectrum of sample D: (a)-(c) as Fig. 2. 0,SSBs. 6 4.5 I I I I I I I 1 I ' lb0 90 80 70 60 50 40 30 20 pp n 6 Fig. 5 27AI Solid-state NMR spectrum of the initial saponite SNa dried at 60°C: (a)MAS spectrum; (b) DOR spectrum. *, SSBs. Table 2 Evolution with the calcination temperature of the "A1 NMR lines intensity ratio R, R, for R, for pillared sample intercalated procedure sample 500 "C 750 "C A 1.37 1.25 0.73 B 2.23 2.18 0.61 C 3.03 2.58 0.73 D 1.13 1.44 0.21 J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 63.9 I 6.5n I I 1 I I 1 I I 120 80 40 0 4 6 Fig. 6 27Al Solid-state MAS NMR spectrum of sample B: (a)-(c) as Fig. 2. 0,SSBs. relevant features of the spectra for all four samples (A, B, C and D) submitted to the same treatments. All intercalated samples show two peaks: a rather broad, asymmetrical peak with d,,, = +5.7 to +7.4 ppm, i.e. in the region of six-coordinated Al, and a narrower, more symmetri- cal one at +63.9 to 65.5 ppm, in the region of four-coordinated Al. Quantification of the A1 NMR peaks should be possible since we used pulse lengths <.n/(21 + 1). The ratio of the integrated intensities of the two peaks, taking into account the main spinning side bands, was given as R,= Z6-coord AJ14-coordThe interpretation of R, is addressed in the Discussion section. Table 2 shows the R, evolution with the calcination temperature.After heating at 50O0C, the R, values are unaltered, although the peak widths vary somewhat in samples B and C. Peak positions do not change significantly except for a small upfield shift for both six-coordinated and four-coordinated A1 in sample D. After heating at 750°C,in contrast, R, falls to values < 1, the most dramatic decrease in six-coordinated A1 intensity occurring for sample D (prepared with 'pure' AlI3). In addi- I I I I I I I I 120 80 40 0 -4 6 Fig, 7 27Al Solid-state MAS NMR spectrum of sample D: (a)-@) as Fig. 2. 0,SSBs. tion, the position of the four-coordinated A1 peak undergoes a significant negative shift for the samples with the lower fixed amounts of Al: -4.5 ppm for sample A and -6.3 ppm for sample D.It would be very interesting to discriminate between tetra- hedral A1 in the clay layers and tetrahedral A1 in the pillar, as is possible with intercalated and pillared beidellite~.~In general, we did not clearly observe more than one signal in the region of tetrahedral A1 for MAS spectra. This result is not very surprising since the resonance of the central A1 in All, is expected at +62.5 ppm (from our own liquid-state NMR results and literature data3v8), too close to the substi- tuting A1 of saponite sheets (+ 64.5 ppm) to be discriminated. We tried to record some DOR spectra to remove second- order quadrupolar broadening and improve the resolution.Fig. 8 shows the DOR 27Al NMR spectra of samples A, A500, B500 and 0.For A, B500 and C500, no improve- ment in resolution is observed owing to severe overlap with DOR SSBs. In the case of A500, however, we can distinguish between a sharp component at +63.4 ppm and a broader one at +56 ppm. This observation is interesting if we keep in mind that Plee et d7observed a similar shoulder at +56.5 Table 3 Evolution with calcination temperature of the positions of the observed 27Al NMR peak maximum for four-coordinated A1 and six-coordinated A1 peak maximum (ppm) six-coordinated A1 four-coordinated A1 calcined calcined procedure intercalated 500"C 750 "C intercalated 500 "C 750 "C A B C D +5.8 +6.8 +5.7 +7.4 +6.2 +6.4 +6.0 +4.5 +5.4 +6.5 +4.1 +4.2 +65.5 +63.9 +63.9 +65.4 +65.0 +64.1 +64.1 +63.4 +60.5+63.9 +64.6 +57.1 J.CHEM.SOC. FARADAY TRANS., 1994, VOL. 90 4.1 I 63.5I 100 80 60 40 20 0 -20 -40 Fig. 8 "A1 DOR NMR spectra: (a)-(c) as Fig. 2 ppm in pillared beidellite and tentatively attributed it to A1 in inverted A10, tetrahedra. In view of this, we are led to think that the asymmetry observed in DOR at the low ppm side of the tetrahedral A1 peak for sample A500 may be attributable indeed to a second signal. However, we did not try to decom- pose the observed signal into its components, a notoriously difficult task for a spin 5/2 nucleus. Another result from DOR NMR, albeit a negative one, is that the broadening of the six-coordinated A1 signal is heter- ogeneous rather than homogeneous since suppression of the quadrupolar interaction does not result in significant line narrowing.In other words, the local environment of octa- hedral A1 is somewhat variable, in both intercalated and pil- lared saponites; indeed, the shape of six-coordinated A1 peaks is reminiscent of that found in glas~es.~ Discussion 29SiNMR The 29Si NMR spectra of the parent saponite confirm the stability of the layers up to at least 750°C in the starting material (the peak at -85 ppm may indicate a very slight disorganisation after 4 h at 750 "C). In contrast, the 29Sispectra of samples A500, B500 and C500 indicate that the pillaring step induces a degree of local reordering in the tetrahedral sheet.We shall be mostly con- cerned here with the decrease of the intensity ratio of the -90 ppm peak with respect to the -96 ppm peak upon heating at 500 "C. If we insist on attributing these peaks to Q3-lAl and Q3-OA1 as in the parent saponite, we would have to admit that some of the aluminium ions in the tetrahedral sheet are replaced by silicon ions, a conclusion difficult to substantiate. It seems easier to rationalise this observation in terms of a mechanism put forward by Plee et al.,' in which chemical (covalent) bonds between the pillars and layers are formed by the inversion of some TO, tetrahedra (T = A1 or Si) of the clay tetrahedral sheet to yield T-0-A1,~,,,,, linkages.There is good reason to believe that this inversion is initiated by the Brernsted acidity provided by the decomposition of the pillars, and that it occurs in the vicinity of Al-substituted tetrahedra, which are the 'weak point' of the clay layer. The mechanism favoured by Plee et al. consisted of opening of an Si-0-A1 linkage by H+ (the appearance of Si-OH-A1 groups upon H+ attack of the Si-0-A1 links during the calcination step was indeed observed in the IR study of our samples') followed by inversion of either the AIO, or the Si04 tetrahedron to give A14-coord-O-A16-coord or ~~~~~~~~coor~ ~ ~ ~ c o o r linkages. These mechanisms are sum- ~ marised in Fig. 9A(ii) and B(ii), respectively. Mechanism A($ is the one favoured by Plee.Fig. 9B(i) illustrates the inversion of an SiO, tetrahedron with no A1 neighbours. We have included this mechanism because it cannot be completely excluded in view of prelimi- nary data obtained by us on pillared montmorillonites. Fig. 10 indicates that while the 29Si spectrum of intercalated mont- morillonite, MD, looks like that of the starting Na-montmorillonite, in the pillared sample, MD500, a large upfield shift and broadening of the resonance is obtained, indicative of strong local modifications in the tetrahedral layer upon pillaring. A similar shift was obtained by Butruille et a!. for pillared fluorhectorite,lO*" at some variance with previous results from the same tearn.l2 We may try to discriminate between these mechanisms by assessing their effects on the 29Si NMR signal and comparing with the experimental data.First, let us notice that in each mechanism, all Si remain Q3, and their number of A1 neigh- bours does not change. However, some of the Si-0-T tetrahedral angles do change, and they have a relevance for the 29Si chemical shift.I3 If we consider a given tetrahedron and assume it is inverted in the pillaring process, it is easy to see from Fig. 9 A s --..._ Interlayer spa-Fig. 9 Illustration of possible pillaring mechanisms (see discussion in the text): A, Preferential inversion of A10, tetrahedra. (i) Effect on Q3-OAl SiO, tetrahedron: unchanged. (ii) Effect on Q3-1A1 SiO, tetrahedron. B, Preferential inversion of SiO, tetrahedra. (i) Inversion of a Q3-OAl tetrahedron.(ii) Inversion of a Q3-lAI tetrahedron. that all three T-0-T angles formed with its neighbours will increase in the process. In contrast, if we suppose that it is one of its neighbouring tetrahedra that is inverted in the process, only one out of three T-0-T angles will increase while the others remain constant (as a first approximation, neglecting relaxation effects). The most precise and complete data on the correlation between average T-0-T angle and 29Si chemical shift published in the literature for tetrahedrally substituted phyl- losilicates were obtained by Sanz and Robert' working on the Na+-saturated form of synthetic saponites (also relevant is Weiss et all4). By extrapolating the data of Sanz and Robert, inversion of a given SiO, tetrahedron with no modi- fication of the first neighbours should yield a downfield shift of ca.-15 ppm on the resonance position of its central atom. In fact, in our case, the purely geometrical modification would have to be accompanied by a change in first neigh- bours (appearance of a Si-0-A1 linkage with the inter- calated polycation) having an opposite effect on the chemical shift (generally ca. +5 to +7 ppm in zeolites and clays): we thus estimate that the net chemical shift modification would be in the range -8 to -10 ppm, i.e. it should easily be obser- vable. On the other hand, inversion of a TO, tetrahedron neigh- bouring the concerned SiO, should only yield a -5 ppm shift since only one out of its three T-0-T angles is modi- fied in first approximation.Now if mechanism B(i) was operating, some isolated Q3-OAl Si would undergo a -8 to -10 ppm chemical shift, while their neighbours (also Q3-OAl but approximately three times more numerous) would undergo a -5 ppm shift. Since no signal is observed in the foreseen region, mechanism B(i) can be discarded here: i.e. in saponite, no SiO, tetrahedra without A1 neighbours are inverted on pillaring. Mechanism B(ii), preferential inversion of SiO, tetrahedra with one A1 neighbour, Q3-lAl Si, would result in a -8 to -10 ppm downfield shift of some Q3-lAl, but also in a -5 ppm shift of twice the amount of Q3-OAl, the Si labelled with a star in Fig. 9B(ii). This does not seem to correspond with experimental results either.Mechanism A($ would only result in a -5 ppm upfield shift of some of the Q3-lAl, those neighbouring the AlO, tetrahedra that effectively form bonds with the pillars. They would be shifted to ca. -95.5 ppm, too close to the Q3-OAl to be resolved from it. All Q3-OAl Si, and those Q3-lAl Si neighbouring non-linking AlO, tetrahedra, would be unaf- fected: the net result would be an apparent decrease of the initial Q3-lAl peak at -91 ppm compensated by an increase of the peak at -96 ppm. This decrease would be more pro- nounced for samples with high amounts of fixed A1 because of an elevated number of pillar-layer links. It can be seen that mechanism A(ii) is the only one compat- ible with the observed evolution of the 29Si NMR spectra.In particular, it is interesting that there is a positive correlation between the amount of A1 fixed in the interlayer space and the decrease of the -91 ppm contribution (Table 1). Samples with high amounts of fixed A1 were obtained from Chlorhydrol and may contain polycations more condensed than However, we do not expect the bonding between these higher polymers and the tetrahedral sheet to be very different from the All3 case. Therefore, we conclude that the formation of bonds between the pillars and the layers in pillared saponites occurs through the specific inversion of AlO, tetrahedra rather than of SiO, tetrahedra, in confor- mity with the hypothesis of Plee et al.' for beidellite. This conclusion is of course limited to pillared clays containing A1 in their tetrahedral sheet.For octahedrally substituted clays such as montmorillonite, the commonly held opinion is that J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 the pillar-layer interaction remains rather weak and of non- covalent nature even after calcining at high temperatures ; however, our data indicate that heating an All ,-intercalated montmorillonite at 500 "C involves considerable modifi-cations of the tetrahedral sheet with changes in the short- range order (Fig. 10). A more precise study is underway on the effect of pillaring on octahedrally substituted clays. The 29Si signals of pillared clays calcined at 750°C give information on the stability of the layers under severe thermal treatments.For the low pillar-density D750, it can be seen that the layers structure has been partly destroyed, with the appearance of a strong signal at -83.4 ppm probably corresponding to a new phase: at the same time, the BET 'surface area' collapses. ' In contrast, in both the unpillared saponite and pillared samples with higher pillar-density, the 29Si NMR signals are hardly modified with respect to the 500°C calcined samples. There seems to exist a critical pillars concentration that causes instability of the pillar-clay layer system: the reason for this critical concentration cannot be inferred from the data presented here although it is possible that variations in layer flexibility are involved. 27~1NMR As stated earlier, interpretation of the chemical shifts of tetra- hedral A1 is hampered by the accidental coincidence of the 6,,, for A1 in the saponite sheet and in the pillar.However, the DOR spectrum of pillared sample A500 yields support to the supposition that mechanism A(ii) is operative here since it shows the signal at +56 ppm attributed by Plee to inverted AlO, tetrahedra. Some remarks can be made on the evolution of the inten- sity ratio, R,, defined in the Results section. Octahedral A1 belongs only to the intercalated polycation; tetrahedral A1 I-80 -9b -A0 -1lo 6 Fig. 10 29Si Solid-state MAS NMR spectrum of: (a)Otay montmo- rillonite; (b)intercalated Otay montmorillonite dried at 60 "C (MD); (c)pillared Otay montmorillonite (MD500) J.CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 has two components, one corresponding to the (constant) amount of A1 in the saponite tetrahedral sheet, while the other contribution stems from tetrahedral A1 in the poly- cation and will of course depend on the number of inter- calated polycations. We can attempt to predict the NMR peak intensity ratio from the amount of fixed A1 (from chemi- cal analysis) if we make a hypothesis on the stoichiometry of the intercalated polycation. If we are dealing exclusively with Al13, each polycation will provide 12 octahedral Als, and one tetrahedral Al. It is then easy to see that the predicted inten- sity ratio, which we will call Ran, is: Ran = 12X/(1 + X) where X is the number of intercalated polycations normalised to the number of substituting A1 in the sheet, or X = x/y = (number of pillars per unit cell)/(number of substituted A1 per unit cell); y is known to be 0.70, while x can be imme- diately obtained from chemical analysis.In fact, it is consistently observed that Ran> R,. A similar observation was made by Schoonheydt16 who raised the possibility that the assumed stoichiometry for the inter-calated polycation (A14-coord)l(A16-coord)lcould be incorrect : the polycation would then have to contain more tetrahedral A1 related to octahedral A1 than was supposed. R, is not expected to be exactly equal to the molar ratio between the two types of A1 since some intensity may be lost to the satellite transitions.The question has been addressed in detail by Massiot et a1.,17 who designed tables of correc- tion factors applicable to the quantification of NMR data on quadrupolar nuclei for various values of the MAS rotational frequency (arf)and the quadrupolar frequency of the observed nucleus (aQ).Taking as a basis the apparent line- width for six-coordinated Al, we would favour a rather low value for the quadrupolar coupling constant (QCC) and thus for oQ:in this case, the corrections to R, due to intensity losses to the satellites would not be over 10% and the cor- rected R, would still be significantly too low with respect to Ran. However, basic aluminium sulfate which contains free [All,] ions shows a clear quadrupolar lineshape for the six- coordinated Als, and allows one to obtain for the QCC a rather high value of 9.8 to 10.5 kHz." One does not expect that the environment of the six-coordinated Als should gain in symmetry upon intercalation; therefore, we suspect that the QCC of six-coordinated A1 might be much more impor- tant than a cursory examination of the NMR signal seems to indicate.The singularities of the quadrupolar lineshape would then be masked by extreme broadening (homogeneous and/or heterogeneous) and the discrepancy between R, and Ran would be much reduced. Furthermore, the values obtained from DOR spectra for R, are higher than those from MAS spectra, and they are rather close to the Ran. This leads us to believe that the NMR intensity ratios do not demonstrate conclusively that aluminium species different from All, are present in the interlayer space.Thus, even though the quantitative data on 27Al NMR of the pillars are not conclusive at the moment, they have the merit of raising the question of the precise stoichiometry of the intercalated polycation in the various preparation condi- tions. Future studies will have to focus on the precise deter- mination of NMR parameters such as the quadrupolar shift 6,, the quadrupolar frequency aQand the asymmetry parameter qQ. Potentially, the evolution of these parameters in going from the isolated pillar (in such model compounds as All, sulfate) to the intercalated pillar might reveal a lot about the layer-pillar interaction. If we compare samples A, B, C and D and plot R, as a function of the amount of A1 fixed in the intercalation step, it seems to vary linearly up to a maximum of 4 mmol A1 (g saponite)-(samples D, A and B).For the highest amount of fixed aluminium (sample C), there is a deviation from linear- 68 1 it^.^ This means that in a first approximation, the species intercalated is the same in D, A and B: the effect of the higher pH value for sample B (6.0) would then essentially be to lower the positive charge on the polycation and thus allow intercalation of a higher amount of polycations to compen- sate for the constant cation-exchange capacity (CEC). The equilibrium form for aluminium is not expected to be the All, ion at this pH value, but the kinetics of further polym- erisation is rather slow" compared to the timescale of the intercalation procedure.In contrast, for sample C, a different phenomenon occurs, which may be cation polycondensation or surface precipi- tation of amorphous alumina on the clay tactoids (in rather low amounts since no alumina phase was observed by elec- tron microscopy). When considering the spectra of pillared samples (A500- D500), no significant variation is observed either in the R, value (Table 2) or in the position of the octahedral peak (Table 3, the position of the tetrahedral peak is not very informative as stated earlier). Apparently, the aluminium pillars have very much the same structure as the initially intercalated polycation. This is puzzling since All should lose all of its H20 ligands, and even condense its OH ligands, well before 500"C,8 a fact which seems confirmed by the TGA-DTA of our intercalated sample^.^ Both phenomena should result in changes in A1 coordination, easily detectable by NMR, which are, in fact, not observed.One possibility is that the pillars' Als do indeed lose some of their ligands upon heating at high temperature, but quickly regain them on exposure to room humidity before the spectrum is recorded. However, the spectra clearly rule out the transformation of the pillars to alumina which was sometimes proposed in the first studies on pillared clays, as already pointed out by Fripiat.2 In other words, an equation of the type: 13/x[A1,0,]'3x-2y)+ + zH' + (24 -~/2)H20 (1) where z = 7 -(13/x)(3x -2y), can still probably be written for the polycation calcination process, but one cannot assume .x = 2, y = 3 as written before.Most probably, the pillar nuclearity does not change and thus x = 13. The stoichiom- etry of eqn. (1) has received surprisingly little attention so far, and we are currently conducting a series of experiments in which the calcined pillared samples will be transferred to the NMR rotors under controlled atmosphere, in the hope that A1 coordination changes will then be noticeable. After calcination at 750 "C, all samples show significant modifications in their 27Al NMR spectrum. These are espe- cially marked for sample D750, where the global structure has already collapsed (see above), but also very significant for B750 and C750 where it has not: thus, an important trans- formation occurs in the structure of the pillars prior to the collapse of the clay layers, and possibly accelerates it.The pillars transformation seems to depend sensitively on their density in the interlayer space. In no case did we observe 27Al NMR signals reminiscent of a spinel structure. Conclusion 29Si NMR sheds light on the pillaring mechanism of All, intercalated saponites, which involves proton attack of Si-0-A1 linkages in the tetrahedral sheet followed by spe- cific inversion of AlO, tetrahedra and formation of a chemi- cal bond to the pillars, in a manner similar to other tetrahedrally substituted pillared clays such as beidellite.27Al NMR of the intercalated samples does not provide any definite evidence of the presence of species different from J. CHEM. SOC. FARADAY TRANS., 1994, VOL. 90 All, in samples A, B and D, while some alumina precipi- tation may occur in sample C. 27Al NMR of the pillared samples reveals that the pillars retain their integrity after calcining at 50O0C,and suggests a reversibility of their dehydration/dehydroxylation reactions. In contrast, after calcining at 750 "C, important structural rearrangements of the pillars occurs even before the global structure collapses, but no spinel-like alumina is formed. The authors are indebted to Drs. H. Zanni and G. Herrman (Bruker Spectrospin) for facilitating access to the NMR spec-trometers, and to Dr.A. Samoson for running the DOR 4 5 6 7 8 9 10 11 12 V. H. Schmidt, in Lecture Notes, Amp2re Summer School II, Yugoslavia 1971, ed. R. Blinc, Institute R. Stefan, Ljubljana, 1971, p. 75. J. Sanz and J. M. Serratosa, J. Am. Chem. SOC., 1984,106,4790. G. Engelhardt, and D. Michel, in High-resolution Solid State NMR of Silicates and Zeolites, Wiley, Chichester, 1987, p. 150. D. Plee, F. Borg, L. Gatineau and J. J. Fripiat, J. Am. Chem. SOC.,1985, 107, 2362. J. T. Kloprogge, Ph.D Thesis, Utrecht, 1992. F. Taulelle, personal communication. J. R. Butruille, L. Michot, 0. Barres and T. J. Pinnavaia, CEAPLS Meeting on Pillared Clays, Athens, November, 1992. J. R. Butruille, Ph.D. Thesis, Michigan State University, 1992. T. J. Pinnavaia, S. D. Landau, M-S. Tzou and I. D. Johnson, J. spectra. C. Davesne conducted the experimental work on montmorillonite pillaring. 13 14 Am. Chem. SOC.,1985,107,7222. J. Sanz and J-L. Robert, Phys. Chem. Miner., 1992, 19, 39. C. A. Weiss, S. P. Altaner and R. J. Kirkpatrick, Am. Mineral., 1987,72,935. 15 G. Fu, L. F. Nazar and A. D. Bain, Chem. Muter., 1991,3,602. References 16 R. Schoonheydt, in CEAPLS Symposium on Adsorption, Separa- tion and Environmental Applications of Pillared Layered Struc- 1 S. Chevalier, R. Franck, H. Suquet, J-F. Lambert and D. Bar- tures, Antwerp, June 1993. thomeuf, J. Chem. SOC., Faraday Trans., 1994,90,667. 2 J. J. Fripiat, Catal. Today, 1988, 2, 281. 17 D. Massiot, C. Bessada, J. P. Couture and F. Taulelle, J. Magn. Reson., 1990,90,23 1. 3 S. Chevalier, Ph.D. Thesis, Paris, 1992. 18 J-F. Lambert and L. Bergaoui, unpublished results. Paper 3/04611G; Received 2nd August, 1993