www.elsevier.com/locate/jnoncrysol

Journal of Non-Crystalline Solids 353 (2007) 1650–1656

1H, 29Si and 27Al NMR study of the destabilization processof a paracrystalline opal from Mexico

Michael Paris *, Emmanuel Fritsch, Bertha O. Aguilar Reyes

Institut des Materiaux Jean Rouxel, UMR CNRS 6502, 2 Rue de la Houssiniere, BP 32229, 44322 NANTES cedex 03, France

Received 20 February 2006; received in revised form 5 October 2006Available online 21 February 2007

Abstract

This is the first NMR study of the destabilization through whitening of paracrystalline opal-CT (SiO2, nH2O). The aim of this study isto understand the process leading to the degradation, particularly any changes in the structure and bonding of this material. 29Si and 27AlNMR signals are not significantly modified. In contrast, using for the first time 1H as a probe in opal, we show that a comprehensiveunderstanding can be obtained on both the local structure of the opal-CT and the destabilization process. Thus, we propose the latter isrelated to some changes in 1H speciation (loss of strongly ‘bound’ water and appearance of new species) which induce structural reor-ganizations on the surface and in the rim of the silica nanograins constituting opal-CT.� 2007 Elsevier B.V. All rights reserved.

PACS: 61.66.Fn; 61.43.Er; 82.56.Dj; 82.56.Hg; 61.72.Hh

Keywords: Silica; NMR, MAS–NMR and NQR; Short-range order; Hydration; Water

1. Introduction

Opal is an amorphous (opal A) or poorly crystalline(opal-CT) variety of hydrous silica, SiO2, nH2O. The watercontent ranges commonly from 6% to 13% weight. Opal isgenerally known as a gem exhibiting ‘play-of-color’ due tothe diffraction of light by a regular network of silica spheresapproximately 200 nm in diameter. Lesser known are vari-eties without play-of-color, called common opal, which arevaluable for their body-color. The better known such opalis the transparent orange ‘fire opal’ found mostly in severalstates of the central Mexican high plateau (Queretaro,Jalisco, Nayarit), but also in Brazil, Kazhakstan, andEthiopia.

Many opals, regardless of variety, are plagued by thefact that they may degrade over long periods of time.For fire opal, this destabilization is well-known to inducewhite, translucent to opaque layers, often found in the cen-

0022-3093/$ - see front matter � 2007 Elsevier B.V. All rights reserved.

doi:10.1016/j.jnoncrysol.2006.12.111

* Corresponding author.E-mail address: michael.paris@cnrs-imn.fr (M. Paris).

ter of the stone in the shape of an egg-shell [1,2]. Dependingon the deposit, this phenomenon may affect as little as 10%to more than 90% of the production. It is the major imped-iment to the economic development of opal mining in Mex-ico. Although it has often been stated that this was due toloss of water, this was never demonstrated.

In general, only limited study of opal destabilization hasbeen pursued. This is in part due to sampling difficulties:the destabilized, white layers are typically very thin, inter-spersed with unaffected opal, making them difficult tolocate precisely by optical microscopy. Only few samplesshow an homogeneous, destabilized white area clearly sep-arated from the original, unaffected material. To be certainthat the white appearance is due to destabilization, andthat the opal was not originally white, one has to workon specimens known to have been of gem quality, whichrequires that an individual has kept a sample of known his-tory after it has deteriorated. Finally, for some techniquessuch as NMR, the quantity of material necessary requireshomogeneous samples of unusually large size 100 mg(0.5 ct).

Fig. 1. {1H}–29Si CP-MAS spectra of opal-CT before (a) and afterdestabilization (b).

M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656 1651

Recent investigations demonstrate that in general, thisdegenerescence is related to a loss of water, ranging from10% to 90%, coupled with an increase in specific surfaceby a factor ranging from only slightly over 1–20 [2,3]. Inaddition, by Raman scattering, it is clear that the broadand asymmetric water-related peak at about 3200 cm�1

decreases, while a new band develops at approximately2950 cm�1, which has been attributed to a new speciationof water, called ‘cristobalitic water’ [2,3].

The present study was prompted by the discovery of aspecimen that did not present all the difficulties mentionedabove, in the stock of an opal dealer in Magdalena, Jalisco,Mexico. The purpose of this study is manifold. It is the firststudy of opal destabilization using NMR. Secondly, NMRstudies of opal have concentrated in the past on 29Si, even27Al (the major impurity), but no study involved so far 1H,although water is a component of opal. A major part of thepresent work is dedicated to the 1H as it provides informa-tion on the opal. Thirdly, NMR results strengthen the pos-tulate that destabilization is accompanied by a change inspeciation of some of the hydrogen, hence presumably ofsome of the water.

2. Experimental procedures

2.1. Materials

The specimen studied originates from the Magdalenaarea of the state of Jalisco, Mexico. The center of this typ-ical irregular opal nodule, approximately 3–4 cm in diame-ter, is an homogeneous and white (destabilized opal),whereas the rim is gemquality, transparent orange (stableopal). It is made of opal-CT, as proven by Raman scatter-ing [4]. This means that it is made of poorly crystallizedcristobalite (C) with significant amounts of tridymite (T)stacking. Its structure is typical of that of many fire opalswith a random aggregation of silica nanograins about20–40 nm in diameter, observed by scanning electronmicroscopy [5]. Its index of refraction (1.45) and specificgravity (about 1.98) are well within the range for fire opal.

2.2. NMR measurements

NMR spectra were acquired at room temperature usinga Bruker Avance 500 MHz spectrometer operating at99.36 MHz, 130.32 MHz and 500.13 MHz for 29Si, 27Aland 1H, respectively, and using a 4-mm double-bearingBruker probehead. Spectra were referenced to TMS for29Si and 1H and a 1 mol L�1 Al(NO3)3 solution for 27Al.

{1H}–29Si CP-MAS (cross-polarization magic-angle-spinning) NMR spectra were acquired using a ramp-ampli-tude sequence [6], a 8 ms contact time and a repetition timeof 5 s. 29Si MAS spectra were acquired with a repetitiontime of 300 s. For all 29Si spectra, MAS spinning ratewas set to 5 kHz and 1H decoupling during acquisitionwas achieved using the TPPM method [7] with a RF fieldof approximately 60 kHz.

27Al MAS spectra were acquired with a p/20 pulse of1.2 ls. No 1H decoupling was used. Repetition time wasset to 10 s and MAS spinning rate to 12 kHz.

For all 1H spectra, MAS spinning rate was set to 5 kHz.1H MAS spectra was acquired with a repetition time of 10 sfor quantitative purposes since 1HT1 was measured to be ofthe order of 1 s. Due to the weak 1H homonuclear dipolarcouplings (see text), dipolar dephasing spectra [8] wasacquired without the need of homonuclear decoupling dur-ing acquisition and thus correspond to rotor synchronizedHahn-echo spectra. Repetition time was set to 1 s anddephasing delays were varied from 0.4 ms to 80 ms. 1Hdouble quanta filtered MAS spectra were acquired usingthe POST-C7 recoupling sequence [9]. MAS spinning ratewas set to 5 KHz and the radiofrequency field to 37 kHz.Excitation time was 400 ls corresponding to two rotorperiods. Repetition time was set to 1 s with prior presatura-tion. 1H MAS rotor synchronized exchange spectra wereacquired without any 1H homonuclear decoupling. Statesmethod was used to give quadrature in the indirect dimen-sion. Repetition time was set to 1 s with prior presatura-tion. Mixing times were varied from 1 ms to 300 ms andspectra were symmetrized.

3. Results

3.1. 29Si single pulse MAS and {1H}–29Si CP-MAS

experiments

Fig. 1 shows the {1H}–29Si CP-MAS spectra of bothnon-destabilized and destabilized opals. These spectra exhi-bit three lines at �112.0, �102.4 and �93 ppm. Fromisotropic chemical shifts, they are, respectively, attributedto Q4, Q3 and Q2 SiO4 species with Qn � Si(OSi)n(OH)4�n

[10]. It should be noted that the –102.4 line could alsobe attributed to the Q4(1Al) species. But, as alreadypointed out by Brown et al. [11], this possibility can beruled out since this line is enhanced by the CP technique.

Fig. 2. 29Si MAS spectra of opal-CT before (a) and after destabilization(b).

1652 M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656

The 29Si widths of the Q4 lines of opals spectra are gener-ally due to an isotropic chemical shift distribution associ-ated to a Si–O–Si bonds angles distribution. For Q4 andQ3 species, both isotropic chemical shifts and full widthat half maximum (7.2 ppm for Q4 lines) agree very wellwith previous results on opals-CT [11–13]. Nevertheless,our results confirm the existence of Q2 species in opals-CT in agreement with Brown et al. [11], and not with thepreviously published data [12–14]. The only differencebetween the two spectra of Fig. 1 is the lower Q3/Q4 inten-sity ratio in the destabilized opal spectrum. This may comefrom a modification of the CP transfer dynamics due to theloss of water during the destabilization process.

Fig. 2 shows the 29Si quantitative MAS spectra of bothnon-destabilized and destabilized opals. The Q2 line is tooweak to be detected. The Q3 and Q4 lines are at the sameposition than in the CP-MAS spectra and their intensityare about 25% and 75%, respectively. The fact that the iso-tropic chemical shifts, the with of the Q4 lines and the Q3/Q4 intensity ratio in the 29Si MAS spectra are the same

Fig. 3. 27Al MAS spectra of opal-CT before (a) and after destabilization(b).

before and after the destabilization process suggests thatthe internal structure of the nanograins is essentiallypreserved.

3.2. 27Al single pulse MAS experiments

Fig. 3 shows the 27Al central transition MAS spectrumof both non-destabilized and destabilized opals. The twospectra exhibit a single line at 53 ppm without any charac-teristic second-order quadrupolar lineshape under MAScondition. Since the line is quite symmetric, this suggestsa chemical shift distribution as for 29Si, rather than an elec-tric field gradient distribution. Moreover, the isotropicchemical shift at 53 ppm agrees with a tetrahedral environ-ment of Al surrounded by four SiO4 units [15]. Finally, nooctahedral environment of Al was detected since there is nosignal in the – 5–15 ppm range. All these results are inagreement with Brown et al. [11].

3.3. 1H single pulse MAS experiments

Fig. 4 shows the 1H MAS spectrum acquired with aMAS spinning rate of 5 kHz of non destabilized opal. Thisspectrum is strongly dominated by a line at 4.8 ppm whichcan be attributed at least in part to free water. Weaker linesat 0.9 ppm, 1.3 ppm and 7.6 ppm can also be seen onFig. 5(a). The two lines at high field are sharp in contrastto the 7.6 ppm one. It is worth noting that the good 1Hspectrum resolution obtained with a MAS spinning rateof 5 kHz is not enhanced by spinning at a 15 kHz MASrate nor by using Frequency Switch Lee Goldburg (FSLG)homonuclear dipolar decoupling sequence [16]. This meansthat the homonuclear couplings between the different 1Hspecies are quite weak.

As previously established for silica materials [8,17,18], itis possible to distinguish three zones in the 1H spectrum.Indeed, the chemical shifts of hydroxyl groups are very sen-sitive to hydrogen bonds formed with water or surroundingOH. The stronger the hydrogen bonds, the higher the

Fig. 4. 1H MAS spectra of non-destabilized opal-CT.

Fig. 5. Detail of 1H MAS spectra of opal-CT before (a) and afterdestabilization (b). Fig. 6. 1H dipolar dephasing spectra of non-destabilized opal-CT at

various dephasing delays.

Fig. 7. 1H dipolar dephasing spectra of destabilized opal-CT at variousdephasing delays.

M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656 1653

chemical shifts of the associated 1H. Thus, the 7.6 ppm linecorresponds to 1H involved in strong hydrogen bonds incontrast to the ones associated to the lines at 0.9 and1.3 ppm which are not hydrogen-bound or isolated hydrox-yls. Hydroxyl groups involved in weak hydrogen bondsand water (free or not) appear typically in the 3–7 ppmrange. Finally, it is interesting to note that in pyrogenic sil-ica, the 1H line at 1.8 ppm was attributed to isolated inter-nal silanols [19].

In both 1H MAS spectra of gem and destabilized opals(Fig. 5), the 1H MAS spectrum is dominated by the freewater line at 4.8 ppm, and lines at 0.9 ppm, 1.3 ppm and7.6 ppm are also visible. After destabilization, we canobserve an increase of the intensity ratio between the1.3 ppm and the 0.9 ppm lines, as well between the1.3 ppm – 0.9 ppm doublet and the 7.6 ppm line. This latterfact agrees well with the above 1H lines assignments andthe loss of water upon whitening. Finally, a new line at3.5 ppm appears for destabilized opal (Fig. 5(b)). Fromspectral decomposition of the 1H quantitative spectra, itis possible to determine the proportion of 1H of water incomparison to the whole 1H proportion in the samplebefore and after the destabilization of the gem. Consideringonly the 4.8 ppm line, we found a change from 65% to 48%.By considering also the 7.6 ppm line (see lines assignmentbelow), we found a change from 67% to 51%. Thus, in bothcases, we observed a decrease of the relative water contentduring the destabilization process as expected.

3.4. 1H dipolar dephasing

Dipolar dephasing experiment [8] allows to preferen-tially select 1H species experiencing the weakest dipolarcouplings (i.e., longest T2 spin–spin relaxation times),which correspond to more isolated 1H. Fig. 6 shows the1H dipolar dephasing spectra of non destabilized opal atdifferent dephasing delays. Two new lines were detected,a narrow one at 4.7 ppm and a broad one around 3 ppm.T2 relaxation times for the 7.6 and 4.8 ppm lines were mea-sured at 2 ms and 7 ms, respectively. In contrast to these

lines, the 1.3 and 0.9 ppm lines both exhibit biexponentialdecays with a long T2 (�14 ms) and a short one (�2 ms)with a 2/1 intensity ratio. Thus, two thirds of these hydrox-yls can be considered as isolated.

Similar measurements were done on destabilized opal asshown in Fig. 7. The 4.7 and 3 ppm lines still exist, but the3 ppm line is less distinct. T2 relaxation time of the 7.6 ppmis similar to that in the non-destabilized opal showing nomodification of the 1H environment for this species. Withdestabilization, the T2 of the 4.8 ppm line increases toabout 14 ms. This may be due to a decrease of the 1H dipo-lar interaction between 1H of the remaining free water mol-ecules. The T2 relaxation time of the 3.5 ppm line is of theorder of 2 ms. This short T2 shows that the associated 1Hare not isolated ones. Finally, the T2 of the biexponentialdecays for the 0.9 and 1.3 ppm lines remain similar. How-ever, the intensity ratio of the two components becomes 1/1. This means that one part of the previously isolatedhydroxyls is not isolated anymore after the destabilizationprocess.

Fig. 8. 1H MAS double quantum filtered spectra of opal-CT before (a)and after destabilization (b).

Fig. 9. 1H MAS exchange spectra of opal-CT before (a) and afterdestabilization at various mixing times (b).

1654 M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656

3.5. 1H double-quantum filtered experiments

This class of experiments provides the opportunity toselect only the resonance of a pair of dipolar-coupled pro-tons through selecting double-quantum (DQ) coherence.Thanks to homonuclear dipolar recoupling sequence underMAS conditions, it is possible to design a 2D DQ-filteredexperiment which correlates the spectrum of pairs of likenuclei coupled by dipolar interaction to the resolved MASspectrum of individuals. Thus, relative spatial informationbetween the different 1H species can be obtained throughthe appearance of off-diagonal peaks in the 2D spectrum.

For both non destabilized and destabilized opals, the 2DDQ-filtered correlation spectra (data not shown) obtainedusing the POST-C7 recoupling sequence [9] do not exhibitoff-diagonal peak. This means that there is no correlationbetween any of the peaks found in the 0–2 ppm range, at4.8 ppm or at 7.6 ppm. The only observed signals areon-diagonal ones. This agrees well with the very goodresolution of the single pulse 1H MAS spectrum even atlow spinning rates. Thus, the strongest homonuclear dipo-lar couplings are between even 1H species since relativelysmall DQ excitation time was used (400 ls correspondingto two rotor periods).

Fig. 8 shows the 1D 1H DQ filtered spectrum of non-destabilized and destabilized opals obtained in the sameexperimental conditions that the 2D spectra. It can beclearly seen that DQ coherence is more efficiently generatedin the 0–2 ppm range for destabilized rather than non-desta-bilized opal. This is in accordance with the conclusion of theT2 relaxation times measurements showing a less isolatedcharacter of theses proton species. Furthermore, since theDQ signal arises from identical 1H species, this can be inter-preted as an increase of the density of hydroxyls protonsassociated to the 0–2 ppm chemical shift range.

3.6. 1H exchange experiments

2D 1H exchange experiments may provide similar spa-tial information to the 2D DQ filtered correlation spectra.

This experiment correlates the MAS spectrum to itself aftera given mixing time. In our case, exchange may arise fromspin diffusion through residual dipolar coupling or fromchemical exchange (water molecules). In presence ofexchange between two sites, off-diagonal peaks appeardemonstrating spatial proximity between the involved 1Hvarieties, closer sites appearing at smaller mixing times.

Fig. 9 shows the 2D 1H exchange spectra for gem anddestabilized opals acquired with mixing times of 1 ms,70 ms and 300 ms. For both samples, exchange progres-sively appears between the 4.8 ppm and the 7.6 ppm linesand between the 0.9 and 1.3 ppm ones. This shows thatthe protons associated to the 7.6 ppm line are not isolatedfrom water and the protons associated to the 0.9 and1.3 ppm lines are spatially close.

For stable opal, there is no exchange between signal inthe 0–2 ppm range, and the 4.8 or 7.6 ppm lines. Thisshows that the 1H species associated to the 0.9 and1.3 ppm lines can be attributed to internal silanol groups,paralleling the proposed assignment of the 1.8 ppm linein pyrogenic silica [19]. For destabilized opal, off-diagonalcorrelation peaks appear between the 0.9/1.3 ppm andthe 3.5 ppm lines as well as between the 4.8 ppm andthe 3.5 ppm lines, starting for mixing time of 30 ms. Atlonger mixing times, the 0.9 and the 1.3 ppm lines alsocorrelate with higher field signals and finally with freewater. This means that when the destabilization processtakes place, one part of the previously isolated internalhydroxyls groups are not isolated anymore from themedium outside the spheres. Hence, these groups find

M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656 1655

themselves particularly close to the newly detected 1H spe-cie at 3.5 ppm.

4. Discussion

Our sample contains about 1% aluminium for whichcharges have to be compensated by others cations, it wasconceivable that Al sites could be implicated in the destabi-lization process. But this hypothesis can be ruled out sincethe 27Al MAS spectrum remains identical after thedestabilization.

Even if the 29Si MAS spectra demonstrates the presenceof Q4, Q3 and Q2 species, this probe fails to give deeperinsights into opal structure. For example, it is very difficultor impossible to discriminate Q3 units on the surface frominternal ones, or even to show the presence of the latter in asilica sample [12]. Moreover, the 29Si MAS spectra areunfortunately very similar before and after the destabiliza-tion. This suggests that the internal and local structure ofnanograins is essentially preserved. Nevertheless, due tothe limited sensitivity and the relatively poor resolutionof the 29Si spectra, we cannot exclude the possibility of amodification of the structure in small proportions.

A very interesting probe of both the local structure ofopals and the destabilization process is offered by 1HNMR spectroscopy. The 0.9 ppm and 1.3 ppm lines areassigned to internal hydroxyls groups, thanks to 1Hexchange spectroscopy. To our knowledge, it is the firsttime that the presence of internal hydroxyls groups isproved in an opal sample. Furthermore, T2 relaxation timemeasurements showed that some of them can be consideredas isolated but the others are not from each other. Whenthe destabilization process takes place, the proportion ofnon-isolated internal silanols increases, as detected by T2

measurements. This can be interpreted as an increase ofthe density of internal silanols inside nanograins, consistentwith the 1H DQ filtered experiments.

The Raman stretching vibration frequency (ms) and the1H chemical shift (dH) of OH species are correlated bythe relation dH(ppm) = D1–D2*ms (cm�1) with D1 =37.9 ppm and D2 = 0.0092 ppm/cm�1 [20,21]. Thus, thebroad and asymmetric water-related Raman band at3200 cm�1 corresponds to a 1H chemical shift of 8.4 ppmclose to the experimentally observed 7.6 ppm line. Keepingtherefore in mind the loss of water during the destabiliza-tion process and the decrease of the 7.6 ppm line, we attri-bute this line to ‘bound’ water (strongly adsorbed)exhibiting hydrogen bonding. Thus, the cross-peaks in 1Hexchange spectra are explained by chemical exchange ofwater molecules. The ‘cristobalitic water’ detected byRaman scattering at 2950 cm�1 would give to a line around11 ppm not detected in the 1H spectrum of destabilizedopal. This could be due to a Raman resonance effect whichcould emphasize this signal compared to the 3200 cm�1 onein contrast to NMR spectroscopy for which the 1H spec-trum (Fig. 5(a)) is quantitative. The 1H exchange spectraof the destabilized opal show that this 1H specie and one

part of the internal silanols are spatially close (typically lessthan 1 nm since the exchange appears from 30 ms on).Moreover, the 3.5 ppm signal exchanges with the 4.8 ppmone assigned to free water molecules. Thus, the 3.5 ppmline likely originates from 1H on the surface of the nano-grains. This line can appear in two ways. First, becauseof the reorganisation of both the surface and the rim ofthe grain upon whitening. Second, because of the loss of‘bound’ water during destabilization, many hydrogenbonds disappear, bringing this line to present a shift fromhigher to lower frequency.

Thus, the destabilization process is correlated to the lossof ‘bound’ water (7.6 ppm line) which also induces a struc-tural reorganization on and fairly close the surface of thenanograins. This process is a slight change as opal-CT isalready poorly crystallized cristobalite. Also, this mustresult in a slight change in index of refraction, inducinglight scattering and the whitish appearance. Nevertheless,at the present stage, it is unclear if there is a direct linkbetween these small structural reorganization induced bychanges in hydrogen speciation and the onset of light scat-tering. Nevertheless, gem opal-CT containing cristobaliteinclusions have been documented, and they have also awhitish appearance due to light scattering [22].

5. Conclusion

We have demonstrated that 1H NMR is a useful probeof the opal structure, one of the few methods sensitive tothe destabilization of opal-CT. The various NMR basedtechniques give consistent results, which can be interpretedin terms of a slight change of structure/texture. The onlydifferences detectable thus far are a change of hydrogenor water incorporation (loss of ‘bound’ water) and the cre-ation of new 1H defects (internal hydroxyls groups) andspecies on the surface of the nanograins. What can beobserved by 1H NMR is located on and near the surfaceof the elementary nanograin.

Acknowledgment

We thank one of the anonymous reviewers for his con-structive criticisms and useful suggestions.

References

[1] E. Fritsch, B. Rondeau, M. Ostrooumov, B. Lasnier, A.M. Marie, A.Barrault, J. Wery, J. Connoue, S. Lefrant, Rev. Gemmol. AFG138&139 (1999) 34.

[2] B.O. Aguilar Reyes, M. Ostrooumov, E. Fritsch, Rev. Mex. Cienc.Geol. 22 (2005) 391.

[3] B.O. Aguilar Reyes, PhD thesis, University of Nantes, France, 2005(in French).

[4] M. Ostrooumov, E. Fritsch, B. Lasnier, S. Lefrant, Eur. J. Mineral.11 (1999) 899.

[5] E. Fritsch, M. Ostrooumov, B. Rondeau, A. Barreau, D. Albertini,A.M. Marie, B. Lasnier, J. Wery, Aust. Gemmol. 21 (2002) 230.

[6] G. Metz, X. Wu, S.O. Smith, J. Magn. Reson. A110 (1994) 219.

1656 M. Paris et al. / Journal of Non-Crystalline Solids 353 (2007) 1650–1656

[7] A.E. Bennett, C.M. Rienstra, M. Auger, K.V. Lakshmi, R.G. Griffin,J. Chem. Phys. 103 (1995) 6951.

[8] C.E. Bronnimann, R.C. Zeigler, G.E. Maciel, J. Am. Chem. Soc. 110(1988) 2023.

[9] M. Hohwy, H.J. Jakobsen, M. Eden, M.H. Levitt, N.C. Nielsen, J.Chem. Phys. 108 (1998) 2686.

[10] R. Dupree, D. Holland, P.W. McMillan, R.F. Pettifer, J. Non-Cryst.Solids 68 (1984) 399.

[11] L.D. Brown, A.S. Ray, P.S. Thomas, J. Non-Cryst. Solids 332 (2003)242.

[12] S.J. Adams, G.E. Hawkes, E.H. Curzon, Am. Mineral. 76 (1991) 1863.[13] H. Graetsch, H. Gies, I. Topalovic, Phys. Chem. Miner. 21 (1994) 166.[14] B.H.W.S. De Jong, J. Van Hoek, W.S. Veeman, D.V. Manson, Am.

Mineral. 72 (1987) 1195.

[15] K.J.D. McKenzie, M.E. Smith, Multinuclear Solid-State NMR ofInorganic Materials, Pergamon Materials series 6 (2002) 273.

[16] M. Lee, W.I. Goldburg, Phys. Rev. A 140 (1965) 1261.[17] D.R. Kinney, I.S. Chuang, G.E. Maciel, J. Am. Chem. Soc. 115

(1993) 6786.[18] I.S. Chuang, D.R. Kinney, G.E. Maciel, J. Am. Chem. Soc. 115

(1993) 8695.[19] J.B. d’Espinose de la Caillerie, M.R. Aimeur, Y. El Kortobi, A.P.

Legrand, J. Colloid Interface Sci. 194 (1997) 434.[20] E. Brunner, H.G. Karge, H. Pfeifer, Z. Phys. Chem. 20 (1953) 1487.[21] E. Brunner, U. Sternberg, Prog. Nucl. Magn. Reson. Spectrosc. 32

(1998) 21.[22] E. Fritsch, E. Gaillou, B. Mocquet, Gems Gemol. 40 (2005) 339.