当前位置:首页 >> 电力/水利 >>

Solid-state NMR characterisation of the thermal transformation of a Hungarian white illite


ARTICLE IN PRESS

Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 www.elsevier.com/locate/ssnmr

Solid-state NMR characterisation of the thermal transformation of a Hungarian white illite
D.L. Carrolla, T.F. Kempa, T.J. Bastowa,b, M.E. Smitha,?
b

Department of Physics, University of Warwick, Coventry CV4 7AL, UK CSIRO, Division of Manufacturing and Infrastructure Technology, Private Bag 33, South Clayton MDC, Clayton, Vic. 3169, Australia Received 27 October 2004; received in revised form 29 March 2005 Available online 17 May 2005

a

Abstract
1 H, 27Al, 29Si and 39K solid-state NMR are reported from a Hungarian illite 2:1 clay for samples heated up 1600 1C. This singlephase sample has a small amount of aluminium substitution in the silica layer and very low iron-content ($0.4 wt%). Thermal analysis shows several events that can be related to features in the NMR spectra, and hence changes in the atomic scale structure. As dehydroxylation occurs there is increasing AlO4 and AlO5-contents. The silica and gibbsite layers become increasingly separated as the dehydroxylation progresses. Between 900 and 1000 1C the silica layer forms a potassium aluminosilicate glass. The gibbsite-layer forms spinel/g-Al2O3 and some aluminium-rich mullite. Then on heating to 1600 1C changes in the 29Si and 27Al MAS NMR spectra are consistent with the aluminosilicate glass increasing its aluminium-content, the amount of mullite increasing probably with its silicon-content also increasing, and some a-Al2O3 forming. r 2005 Elsevier Inc. All rights reserved.

Keywords: Illite; Solid-state NMR; Clay mineral; Thermal processing; Dehydroxylation

1. Introduction Clays form a technologically important group of minerals that have numerous industrial uses being major components of whiteware products and ?reclay bodies such as bricks, roof tiles and cements. Clay ?ring is a very mature part of the ceramics industry and this thermal reaction is amongst the most studied of all ceramic reactions [1–3]. There is usually a clear idea about the equilibrium end point of the thermal breakdown, but the details of the intermediate states, which usually have signi?cant amorphous components have remained largely elusive. To obtain a fundamental understanding of the thermal breakdown mechanism that will allow optimisation of the processing of clay minerals it is necessary to study these materials with

?Corresponding author. Fax: +44 24 7669 2016.

E-mail address: m.e.smith.1@warwick.ac.uk (M.E. Smith). 0926-2040/$ - see front matter r 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.ssnmr.2005.04.001

leading-edge, modern spectroscopic techniques that can provide new unambiguous structural information. Clay minerals are layer structures with the layers lying in the ab plane, which are then stacked up parallel to the c-axis. The layers are of 2 types; silica-like layers made up of SiO4 units and gibbsite-like layers made up of AlO4(OH)2. The two main groups of clays are those which comprise repeat units of one silica and one gibbsite layer; 1:1 clay minerals, and those made up of a sandwich of a single gibbsite layer between two silica layers; 2:1 clay minerals. Between these repeated layers there are interlayer spaces (Fig. 1). Complexity in clays may arise through atomic substitutions that can occur within these layers. Some examples are aluminium substituting for silicon in the silica-like layers, and iron and magnesium substituting for aluminium in the gibbsite-like layer. These substitutions usually produce a negative layer charge, which then needs to be balanced by cations (e.g. Na+, K+), which reside in the interlayer space. The interlayer space can also accommodate

ARTICLE IN PRESS
32 D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43

Potassium

Oxygen Interlayer spacing containing K and H2O. Tetrahedral (silica) layer: Variable Al substitution. Aluminium Silicon Hydrogen

Octahedral (gibbsite) layer.

Fig. 1. Schematic structural picture of a 2:1 clay mineral such as illite.

varying amounts of water to hydrate these cations. The various layer substitutions and interlayer contents can produce signi?cant changes in the properties of the clay including its thermal breakdown mechanism. An important group of 2:1 clay minerals termed illite was ?rst proposed in 1937 [4] to denote mixtures of minerals, including muscovite and feldspar whose physical properties resemble those of the mica family. It has been generally accepted that the main difference between mica and illite is that micas contain more potassium as interlayer cations, whereas illite contains more water and silica. The general formula for illite can be expressed as K1.5Al4(Si6.5Al1.5)O20(OH)4 [5]. Earlier studies on illite-rich clays [6–9] using various techniques such as ESR and Mossbauer show there are four main ¨ reaction processes which occur; viz. dehydration, dehydroxylation, structural breakdown and re-crystallisation. Interlayer water is driven off by 350–400 1C, followed by dehydroxylation between 450 and 700 1C, and irreversible structural breakdown $900 1C. The formation of spinel typically occurs at 1000 1C and continues to increase in amount and particle size with increasing temperature. Other phases can form such as haematite, corundum, mullite and glassy phases at higher temperatures, depending on the composition [9]. The presence of aluminium and silicon makes clays an obvious target for solid-state NMR because both of these nuclei are very amenable to such study [10–13].

There have been various multinuclear NMR reports (especially 29Si and 27Al) of the thermal decomposition of clay minerals. The most comprehensive NMR data set from clays has been for 1:1 clay minerals, such as kaolinite [14–24] and related phases such as halloysite [25]. For 2:1 type clay minerals pyrophyllite is a cationfree clay that has no iron and it shows different thermal breakdown characteristics from the cation-free 1:1 clay minerals, with much more AlO5 in the intermediate dehydroxylate state [26–28]. Studies on muscovite using solid-state 29Si and 27Al NMR [29] showed that the dehydroxylate phase decomposes ($1100 1C) by separation of the silica layers, which then combine with the cations (K+) forming a feldspar-like phase. The remainder of the structure forms a spinel/g-Al2O3, which then transforms into aluminous mullite, and then corundum, and on further heating the mullite becomes more silica-rich by reacting with the feldspar-like phase. There have been more recent attempts to carry out multinuclear solid-state NMR on clays that are more complex, often smectite-illite mixtures, which more closely resemble clays that are used industrially. A calcium montmorillonite (Fuller’s Earth) showed subtle structural changes of the layers up to $900 1C when there is a signi?cant break down into alumina-rich and silica-rich domains [30]. In Lode an illite-rich clay of the mica group with a small amount of interstrati?ed smectite, it was observed that compared to other 2:1

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 33

clays 27Al MAS NMR showed very little AlO5 in the dehydroxylate, and at high temperatures an aluminosilicate glass was the dominant phase [31]. The study of Lode indicated that exactly the same spectra were obtained from powdered, pressed and extruded samples [31]. Another model illitic clay, Silver Hill, again showed no AlO5 in the intermediate dehydroxylate [32]. One of the key dif?culties with these previous NMR studies of illites is that they typically had iron-contents of 4.8–7.1 wt% Fe2O3, which affects the NMR signal intensity, losing signal from those nuclei close to the paramagnetic centres. This means that the spectra can only really be interpreted qualitatively. For example, in Lode for silicon $41% of the expected intensity was observed at all temperatures, whereas for aluminium the signal intensity decreased from 41% in the room temperature sample to only 22% in the sample heated to 1200 1C. As new phases form, the iron partitioning is often not uniform. In Lode and Silver Hill samples Xray diffraction (XRD) showed that mullite forms, but it was absent in the NMR data. Elemental analysis shows that the iron-content of the mullite is signi?cant and is subsequently lost from the NMR spectra. This study reports multinuclear (1H, 27Al, 29Si, 39K) magnetic resonance from a Hungarian white illite, which has a very low iron-content (0.4 wt%). This greatly extends previous NMR studies of illitic clays since the intensity complications that arise from paramagnetic ions are not present and quantitatively accurate results should be obtained. This study includes quantitative 1H measurements, 27Al at two applied magnetic ?elds (14.1 and 18.8 T) and 39K. 39K has a natural abundance of 93.1% and a relative receptivity of its central transition comparable to that of 13C. It has been relatively little studied because it has a small magnetic moment making it a low-g quadrupolar nucleus [11,33]. However, in the few 39K NMR studies of solids that have been carried out spectra are usually readily obtained. For example, static 39K NMR spectra from a range of model inorganic potassium salts showed relatively small dipolar coupling, resulting in sharp powder pattern features of the observed central transition allowing accurate extraction of the quadrupole parameters as chemical shift anisotropy appeared to make only a very small contribution [34]. In potassium-containing aluminosilicates it has been shown that from the position of the centre-of-gravity (cg) of the static 39K resonances, the tectosilicate orthoclase ($205 ppm) could be distinguished from phyllosilicates in which the cg position is below 40 ppm [35]. Potassium NMR could also distinguish structural and exchangeable potassium in the clay mineral montmorillonite as it undergoes wetting and drying [35]. For potassium aluminosilicate geopolymers the cg positions varied from ?47 ppm (more typical of a hydrated potassium phase) in the unheated geopolymer, to ?120 ppm in the

material heated at 1300 1C, consistent with the position of an unheated feldspar [11].

2. Experimental methods Chemical analysis of the white Hungarian illite studied here is shown in Table 1 which is typical for potassium illite (of model composition (K1.5Al4(Si6.5Al1.5)O20(OH)4)). In the illite phase studied here there is some calcium and a very small amount of iron substitution for aluminium in the gibbsite layer. There were no other phases detected. Experiments were carried out on powders, which were heated to and from the desired temperature in the range 200–1600 1C by ramping at 2 1C/min, and were held at temperature for 2 h. Differential scanning calorimetry (DSC) was carried out on a Netzsch DSC 404 instrument using a ramp rate of 10 1C/min under a static air atmosphere up to 1300 1C. Thermogravimetric (TG) analysis was carried out on a Stanton Redcroft TG-750/770 instrument at a ramp rate of 10 1C/min under ?owing air up to 1000 1C. XRD diffraction was carried out on powder samples using a Bruker D5005 powder diffractometer, with the patterns measured over a 2y range 10–901 at a scan rate of 0.151(2y)/min using Cu-Ka radiation of wavelength ? 1.5418 A. 29 Si MAS NMR experiments were carried out on a Varian-Chemagnetics CMX 360 In?nity Plus (8.45 T magnetic ?eld) spectrometer. 29Si magic angle spinning (MAS) NMR spectra were collected using a 6 mm probe spinning at $5 kHz. A 15 s recycle delay was employed with a 1.2 ms (301) pulse, which allowed relaxed spectra to be obtained. The spectra were collected at 71.54 MHz and referenced against TMS at 0 ppm. The 27Al MAS NMR spectra were collected at 18.8 T using a 4 mm probe spinning at 15 kHz with 1 ms ($151) pulses and a 1 s recycle delay on a Varian Chemagnetics In?nity Plus spectrometer operating at a frequency of 208.39 MHz. The spectra were repeated at 14.1 T on a VarianChemagnetics In?nity spectrometer operating at a frequency of 156.37 MHz. The pulse and recycle delay conditions were the same as at 18.8 T, but a 3.2 mm Chemagnetics probe was used spinning at 18 kHz. The
Table 1 Chemical analysis of Hungarian white illite Component SiO2 Al2O3 CaO Fe2O3 K2O Loss on ignition (400 1C) Loss on ignition (1000 1C) Content (wt%) 49.6 33.0 1.6 0.4 6.6 4.3 3.9

ARTICLE IN PRESS
34
27

D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43

Al spectra were referenced against the AlO6 resonance of Y3Al5O12 at 0.7 ppm.1 1H MAS NMR spectra were measured at 360.10 MHz using a Doty 4 mm probe spinning at 15 kHz. A recycle delay of 20 s was used to ensure that the samples were fully relaxed. For the proton spectra the samples were weighed and the integrated signal intensity was compared to that from an adamantane sample, which was also weighed. An 1H spectrum was run on the probehead with an empty rotor under identical conditions and the integral was subtracted from that of the samples. The adamantane was used as a secondary 1H reference at 1.82 ppm. 39K static data was collected on a Varian Chemagnetics CMX 300 In?nity (7.05 T magnetic ?eld) spectrometer at a frequency of 14.01 MHz using a static 10 mm probe. An echo sequence was used with two 5 ms (sub-901) pulses separated by 200 ms with a recycle delay of 0.05 s. Extended phase cycling was used to suppress direct magnetisation and any ringing [36]. 39K spectra were referenced against a static sample of solid KBr at 55.1 ppm.

3. Results DSC results show a good deal of structure (Fig. 2(a)) with a number of endotherms and one exotherm. The sharp exotherm is the a ! b phase transition of quartz that is used as a reference point. TG analysis (Fig. 2(b)) shows that there is initially a sharp drop of mass ($4%) up to $100–150 1C which is the loss of interlayer water, continuing much more slowly up to 400 1C. Then another signi?cant mass drop, commencing at 450 1C is observed, with two distinct steps, corresponding to the dehydroxylation process. There is no mass loss beyond $800 1C stabilising at $91% of the initial value. There is good agreement of this mass loss with the loss on ignition of the clay up to 1000 1C of 8.2% determined during chemical analysis, which corresponds to the loss of water and hydroxyl species (Table 1). XRD patterns from selected samples are shown in Fig. 3. In the initial room temperature sample a set of broad, but nevertheless de?nite re?ections are present. Such a pattern is typical of a clay and the pattern indexed in accordance with the patterns available in the crystallographic databases for illite (i.e. K1.5Al4(Si6.5Al1.5)O20(OH)4). After heating to 1000 1C some very weak re?ections are observed which are dif?cult to assign and there are some very broad features from the glass present. Then heating to 1600 1C the crystalline re?ections have narrowed considerably, and can be assigned as a
1 Note that the secondary shift references used here are quoted relative to the accepted zero shift references for each nucleus i.e.1H TMS, 27Al in [Al(H2O)6]3+ in aqueous 1M Al(NO3)3 and 39K as K+ in aqueous 0.1 M KCl.

Fig. 2. (a) Differential scanning calorimetry and (b) thermogravimetric analysis from a white Hungarian illite heated at 10 1C/min in air.

mixture of mullite and a-Al2O3. There is still some broad underlying intensity from the glass that remains in the sample. 29 Si MAS NMR spectra are shown in Fig. 4 from all the samples, including the initial clay at room temperature and those heated from 200 to 1600 1C, with the peak positions and linewidths summarised in Table 2. At p400 1C, the spectra remain effectively unchanged showing two resolved peaks at ?86.8 and ?92.0 ppm, and a small shoulder at ?82.9 ppm (see Table 2). The parameters for the peaks were deduced by ?tting three Gaussian lines. Above 400 1C, heating up to 900 1C, there are only some minor changes in the spectrum with the clear resolution of the two main peaks being lost and a small shift of the peak (peak position and width, Table 2). Between 900 and 1000 1C the 29Si MAS NMR spectrum undergoes a very signi?cant change, with the peak shifting to ?100 ppm and the resonance getting very much broader. From 1000 to 1600 1C only small, but nevertheless distinct changes occur. The main broad resonance starts at ?101 ppm, but shifts progressively to ?95.9 ppm at 1600 1C. At 1200 1C there is also a small, sharper peak that appears at $?86 ppm. This small peak increases in intensity with increasing heat treatment, although it becomes less resolved as the main broader resonance moves to more positive shift.

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 35

Fig. 4. 29Si MAS NMR spectra of a Hungarian white illite clay heated at various temperatures up to 1600 1C for 2 h.

Fig. 3. XRD patterns from a Hungarian white illite clay at (a) room temperature, and after heating to (b) 1000 1C and (c) 1600 1C.

The corresponding 27Al MAS NMR spectra collected at 18.8 T are shown in Fig. 5 and are described here in terms of their peak positions (Table 2) with the actual isotropic chemical shifts given in Table 3. As with the 29 Si data no signi?cant change is observed in the spectra of samples heated to p400 1C with two resolvable peaks present, a smaller one at 70 ppm (Table 2) corresponding to AlO4 and the major peak at 4.4 ppm assigned to AlO6. Then at 650 1C some noticeable changes start to occur with both the initial resonances broadening. The AlO6 signal in particular develops a distinctive tail to negative shift. There is also some evidence of an additional resonance at intermediate shifts ($25 ppm, Table 2). Above 600 1C, the AlO6 peak continues to

decrease in intensity until 900 1C, when the original sharp peak at 2.5 ppm has very low intensity. The octahedral peak is replaced by AlO4 and AlO5 signals, and also a much broader AlO6 resonance. At 1000 1C there is a marked change in the spectrum with the AlO6 region now showing two AlO6 resonances at 12 and 4 ppm. The AlO5 peak completely disappears and there is a profound change in the AlO4 peak. The AlO4 peak changes both its shape and shows a signi?cant shift of $10 ppm. Then with increasing heat treatment temperature the fraction of AlO4 increases and there is changing intensity between the two AlO6 resonances observed, until at 1400 1C there is very little intensity left in the 12 ppm AlO6 resonance. At 1500 1C a new AlO6 resonance appears at 14.9 ppm, which increases in intensity on heating up to 1600 1C. By carrying out the experiment at 18.8 T, the high magnetic ?eld reduces second-order quadrupole effects [37] so that in some cases better resolution can be observed than at 14.1 T, for example, between the different AlO6 resonances seen in samples heated in the range 1000–1400 1C. When the lineshapes from quadrupole nuclei do not show well-de?ned second-order quadrupolar features, such as singularities, unambiguous simulation of the lineshape is dif?cult, especially if the data is obtained at only one magnetic ?eld. Taking data at a second magnetic ?eld allows more de?nite

ARTICLE IN PRESS
36 D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 Table 2 Summary of 29Si MAS NMR linewidths and peak positions and temperatures up to 1600 1C Sample heat treatment temperature/1C RT 200 400 650 700 750 800 900 1000 1100 1200 1300 1400 1500 1600 Si Peak position/ppm 70.2 ppm ?82.9, ?86.8, ?92.0 ?82.6, ?86.8, ?92.1 ?82.5, ?86.6, ?91.4 ?91.9 ?92.6 ?92.8 ?92.6 ?93.2 ?100.4 ?101.3 ?101.1 ?97.1 ?95.8 ?95.6 ?95.9
29

27

Al peak positions at 18.8 T from a Hungarian white illite heated at various
29 27

Si linewidth Hz

Al Peak position/ppm 70.2 ppm 71.3, 70.8, 71.3, 66.5, 65.9, 65.0, 65.2, 63.6, 55.4, 55.4, 55.6. 55.8, 55.8, 56.0, 54.6, 4.4 4.5 4.3 27.4, 25.6, 24.3, 24.8, 24.3, 12.0, 11.8, 12.0, 11.8, 14.9, 14.7, 14.4,

170720, 310725, 420725 190720, 300725, 450725 210720, 300725, 500750 800750 700750 800750 900750 10007100 14007100 14007100 12007100 14007100 12007100 14007100 13007100

3.9 3.9 3.2 3.7 2.5 4.3 4.2 3.7 3.0 3.1 2.9 2.3

Fig. 5. 27Al MAS NMR spectra of a Hungarian white illite clay heated at various temperatures up to 1600 1C for 2 h at a magnetic ?eld of 18.8 T. (*Indicates spinning sidebands.)

lineshape analysis. The spectra have been simulated using a programme QuadFit assuming second-order quadrupolar lineshapes including a distribution of the interaction parameters. A simple Gaussian distribution of the quadrupolar coupling constant about a mean value is taken along with an additional Gaussian broadening representing a chemical shift distribution (D). The parameters are iterated between the two ?elds until a single set of parameters (that include the isotropic chemical shift (diso ), the quadrupole coupling constant (wQ ) and the width of the distribution in the quadrupole interaction (DwQ ) and the additional Gaussian broadening (D)) are arrived at that gives the best

simulation at both ?elds [38]. Some examples of typical simulations are shown in Fig. 6. The NMR interaction parameters derived from the two ?eld simulation are given in Table 3. In samples from the unheated one to that heated to 400 1C the spectra are essentially identical, with $2873% of the aluminium present as AlO4. The simulations are of the centrebands so that with ?nite spinning speeds the amount of intensity will depend on the factor (w2 =no nr ) where no is the Larmor Q frequency and nr is the spinning speed [11,39]. There is only 4% difference in this factor between the two sets of conditions used here. To get the true intensity distribution between the different sites the simulated intensity needs to be corrected to account for the fraction of intensity that appears in the centreband [11,39]. However, for most sites (wQ o7:5 MHz) this correction is smaller than the 3% error estimated in the simulation and even for the largest value of wQ of 10 MHz because of the high ?elds and quite fast spinning employed here 92% of the intensity still appears in the centreband simulation and so any correction will only be small. At 400 1C there is still not really any change, e.g. in the aluminium distribution, but there are some signs of precursors of change in that there is a slight increase in the quadrupole coupling constant (wQ ). After heating to 650 1C the spectrum begins to change more rapidly as the AlO6-content starts to drop as dehydroxylation occurs. The strong overlap between the two AlO4 signals means that there is probably a larger degree of error in their parameters, but nevertheless it is clear that there are two AlO4 species. The effect of the high ?eld providing extra resolution can be seen from there being a clear shoulder on the AlO4 signal. The AlO5 site that is formed only has a moderate wQ but has a much larger broadening that scales with the magnetic ?eld, indicative

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 37 Table 3 Summary of 27Al NMR interaction parameters deduced by 2 ?eld NMR data simulated using the QuadFit programme from a Hungarian white illite heated at various temperatures up to 1600 1C Heat treatment temperature (oC) RT 200 400 650 Peak diso (ppm) 70.2 1 1 1 1 1 1 1 2 1 1 1 2 1 1 1 2 1 1 1 2 2 1 1 2 2 1 1 2 1 1 2 1 72.4 6.0 71.8 5.6 72.8 5.5 72.3 68.8 25.5 4.9 71.9 67.2 24.8 4.4 65.6 63.5 33.6 7.0 68.5 60.2 18.8 7.2 70.0 59.8 15.9 9.8 59.7 17.6 10.0 61.5 17.1 8.9 wQ (MHz) 70.5 2.9 3.7 2.6 4.0 3.5 4.0 4.5 5.0 4.0 3.5 4.0 4.5 4.0 4.5 3.0 8.0 10.0 7.0 4.0 6.5 9.0 6.0 4.0 5.5 5.0 9.0 6.0 4.0 9.5 6.5 4.2 8.5 DwQ (MHz) 70.5 1.4 2.2 1.7 2.0 1.0 3.0 1.0 1.0 0.1 4.0 1.0 1.5 0.1 4.0 1.0 3.0 2.0 2.0 1.0 2.0 2.5 1.5 1.0 2.0 1.0 2.0 3.0 1.0 2.2 2.0 1.2 2.0 D (Hz) 720 100 70 100 80 100 60 100 120 500* 80 100 130 500* 70 200 250* 250* 50 300 250* 150 90 300 300* 230 100* 300* 200* 50* 300* 90 40 % 73 26 74 28 72 30 70 14 20 14 52 7 34 16 43 23 40 24 13 11 54 12 3 8 58 12 22 71 3 27 71 12 17

750

900

1000

1200

1400

1600

AlO4, AlO6, AlO4, AlO6, AlO4, AlO6, AlO4, AlO4, AlO5, AlO6, AlO4, AlO4, AlO5, AlO6, AlO4, AlO4, AlO5, AlO6, AlO4, AlO4, AlO6, AlO6, AlO4, AlO4, AlO6, AlO6, AlO4, AlO6, AlO6, AlO4, AlO6, AlO6,

diso Is the isotropic chemical shift, wQ is the quadrupole coupling constant [37], DwQ is the width of the Gaussian distribution of quadrupole coupling constants, D is the additional Gauusian broadening employed, *these widths scaled approximately with the ?eld and the value quoted is that used for the simulation at 14.1 T. % Is the integrated intensity of the corresponding centreband. The asymmetry parameter of the quadrupole interaction is taken as zero.

of the chemical shift distribution being dominant. At 900 1C when the structure is beginning to signi?cantly breakdown there is a large distortion of some of the environments. At 1000 1C there is a large change in the spectrum as the AlO5 is completely lost. There are de?nitely two AlO6 sites, with the main AlO6 site, which is presumably from a transitional alumina/spinel component, having a large ?eld gradient with a distribution showing that this site has a large variation in parameters throughout the sample. There are two AlO4 sites, with the site at $60 ppm probably from the aluminosilicate glass. At 1200 1C only a single AlO4 site is distinguishable. The mullite-related AlO6 site remains with similar parameters, but above 1400 1C a new AlO6 peak appears with a shift $17 ppm that increases in intensity with heat treatment temperature. The proton MAS NMR spectra (Fig. 7) show that there are two distinct resonances in the initial clay at 4.9

and 2.0 ppm. Neither of these two resonances has strong sidebands. There is some sign of two peaks persisting even after signi?cant heat treatment, but the higher shift (4.9–4.6 ppm) peak decreases in both peak amplitude and shift, and is probably completely lost at 750 1C. The 2.0 ppm peak persists up to 900 1C, while at higher temperatures still a very much broader, weak peak is observed with a shift of $3.7 ppm. The absolute intensity of the proton signal was also deduced by comparing the integrated signal from a weighed adamantane standard. The normalised proton intensity (in mol per g) are shown in Fig. 8. It can be seen that the proton-content remains very approximately constant up to $400 1C and the spectra do not really change, although the peak at $4.9 ppm does get a bit broader than at lower heat treatment temperatures so that the resolution is worse. There is no increase in the absolute intensity of the 2.0 ppm peak within the accuracy of the

ARTICLE IN PRESS
38 D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43

Fig. 6. Examples of the simulation of the 27Al MAS NMR spectra of a Hungarian white illite clay using the QuadFit programme for samples heated for 2 h to 1600 1C (A,B), 1000 1C (C,D) and 750 1C (E,F). Data was taken at 14.1 T (A,C,E) and 18.8 T (B,D,F) with each showing the separate components (lower), complete simulation (middle) and the experimental data (top).

measurements. From 400 1C there is a decrease in the overall intensity with increased heat treatment, up to $800 1C, after which the proton-content remains approximately steady at only $5% of the original level. 39 K static NMR spectra are shown in Fig. 9. In the initial clay there is a well-de?ned single resonance at ?41 ppm, with good signal to noise obtained even though the potassium-content is only $3.6 at%. The lineshape shows some asymmetry to negative shift. There is no real change up to 900 1C with the peak position remaining approximately constant and a small increase in the linewidth up to this point. After heating at 1000 1C although the peak position does not really

change, remaining at $?50 ppm, the resonance has become much broader. With further heat treatment the width of the spectra becomes slowly greater and increasingly asymmetric.

4. Discussion In heating a clay the main reactions that occur initially are the loss of protons both as water from the interlayer space that is hydrating the cations, and from the hydroxyls of the gibbsite layer. It would appear from the absolute signal intensity that there is very little

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 39

Fig. 7. 1H MAS NMR spectra of a Hungarian white illite clay heated at various temperatures up to 1100 1C for 2 h.

Fig. 9. 39K static NMR spectra of a Hungarian white illite clay heated at various temperatures up to 1600 1C for 2 h.

Table 4 Summary of 1H MAS NMR peak positions from a Hungarian white illite heated at various temperatures up to 1100 1C Fig. 8. The absolute proton-content of a Hungarian white illite clay heated at various temperatures up to 1100 1C for 2 h as deduced by quantitative proton NMR and corrected for probe background. Sample heat treatment temperature/oC RT 200 400 600 650 700 750 800 900 1000 1100
1

H peak position/ppm 70.1 ppm 2.0 2.1 2.2

proton signal loss up to 400 1C, but TG analysis shows a 4% mass loss up to $400 1C. The difference between thermal analysis and NMR experiments is that the samples for the 1H MAS NMR spectra have been cooled back to room temperature and the water can re-enter the system, whereas the thermal analysis is of course carried out at temperature. The high shift signal at $4.9 ppm gets broader and changes shift (Table 4) suggesting that this is associated with the water, and as it is interacts again with the potassium it appears less ordered than in the sample prior to any heat treatment. The other peak at 2.0 ppm corresponds to the hydroxyls, which is in good agreement with the shift determined for hydroxyls in kaolinite at 2.8 ppm [40]. Then at higher temperatures this peak gets much weaker as the process is no longer reversible. The signi?cant loss of proton

4.9, 4.8, 4.6, 2.1 2.2 2.0 2.0 1.8 3.8, 3.5 4.2

2.0

signal intensity, effectively starting at $400 1C and carrying on up to $800 1C is associated with the dehydroxylation of the gibbsite layer. Finally there is a very broad proton peak remaining even at 41000 1C. This is a very weak signal of only $5% of the initial intensity and is probably only the surface interacting

ARTICLE IN PRESS
40 D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43

proton-species that are almost always present on the surfaces of such aluminosilicates. It is clear from both the 29Si and 27Al NMR that below 400 1C the dehydration process which occurs in the interlayer space has no affect on the environments in the silica and gibbsite layers. The 27Al MAS NMR spectra show that as expected most of the aluminium is present as AlO6 from the gibbsite layers. There is some AlO4 that corresponds to substitution in the silica layer. The AlO6:AlO4 ratio determined experimentally of 2.670.1 agrees well with the ratio expected for the ideal illite composition of 2.7. The aluminium substitution in the silica layer conveys a negative charge to these layers and is charge balanced by the interlayer potassium. In the un?red clay the 29Si MAS NMR spectra reveal a silicon shift characteristic of Q3 sites. For layer aluminosilicates such as illites and closely related clay minerals such as muscovite resolution of different sites is often limited as the presence of iron greatly degrades the resolution of such 29Si MAS NMR spectra. In Lode for instance [31] there were hints of at least two distinct silicon sites. For other closely related micaceous minerals resolution of different Q3 silicon sites have sometimes been observed [41]. These different sites are usually attributed to a single crystallographic distinct silicon site in the layer with varying numbers of next nearest neighbour (nnn) aluminiums. The three 29Si resonances observed here can therefore be assigned as Q3(0Al), Q3(1Al) and Q3(2Al) to the peaks at ?92.0, ?86.8 and ?82.9 ppm, respectively, [41]. The relative intensities of these lines obtained from ?tting the spectrum are 6074%, 3574% and 572%. Various models have been advanced for the dispersion of the aluminium within such aluminosilicate sheets. Lowenstein’s rule means that direct Al–O–Al bonding is avoided. It has been found that there is a tendency for the dispersion to be stronger than this, various studies indicating a homogeneous dispersion within the layer usually gives a better description [42,43]. Comparing the relative intensity of the three species observed in the spectrum, Lowenstein’s rule would give 45% Q3(0Al), 41% Q3(1Al) and 12% Q3(2Al), while a homogeneous distribution would give 54% Q3(0Al), 40% Q3(1Al), and 5% Q3(2Al). It is clear from the intensity distribution determined from the illite studied here that the aluminium distribution much more closely matches the homogeneous model than that from Lowenstein’s rule. After heating to 650 1C, small but de?nite changes commence in the layer structure, most noticeably for the aluminium sites where there is a small broadening of the AlO6 site, but the most noticeable difference is the formation of some AlO5 ($14%). The AlO4 resonance broadens and shows a shift. The 650 1C data taken at 18.8 T is quite revealing in that the AlO4 region shows de?nite structure, probably as a result of contributions from both aluminium still residing in the silica layer and

that forming as dehydroxylation occurs. The 29Si MAS NMR spectrum shows only minor changes with a slight decrease of shift and increase in broadening so that the two main sites can no longer be clearly resolved. This probably comes about as the silica layer distorts slightly as the adjacent gibbsite layer starts to dehydroxylate. It is clear from the samples heated up to 800 1C that the peak position changes from $?89 to $?93 ppm. There is no reason to suppose at these temperatures the aluminium-content of this layer is changing. For aluminosilicates there is generally observed to be a decrease of the isotropic 29Si chemical shift with an increase in the average Si–O–Si bond angle [11]. Hence a more negative 29Si chemical shift may indicate that as the adjacent gibbsite layer dehydroxylates and starts to separate the average Si–O–Si bond angle increases. Then, with continued heating up to 900 1C the silicate layer shows little further change in both its peak position and broadening. The 27Al MAS NMR spectra always show some AlO5 in the same temperature range, and increasing AlO4 as the AlO6-content decreases as dehydroxylation proceeds, reducing the coordination of the gibbsite layer. It is interesting to compare the development of AlO5 here with that formed in related clay phases. Calcium montmorillonite shows a small amount of AlO5 [30], with none present in Lode [31], indicating that there is great variation in the stability of AlO5 in these intermediate, disordered phases formed from illite-related clay minerals. In pyrophyllite all aluminium becomes AlO5 in the intermediate state [26]. One of the reasons that has been advanced for this difference is the nature of the cations in the structure; Lode’s main cation is K+ and for Fuller’s Earth (montmorillonite) it is Ca2+, while no cations are found in pyrophyllite. However, potassium is also present in the illite studied here and there is a de?nite AlO5 signal. Hence, the results reported here indicate that the effects of the nature of the interlayer cation are at most weak. It could be that the lack of AlO5 signal in Lode is that the iron present is closely associated with the AlO5 that wipes out this signal from the NMR spectrum. This data suggests that it is now more likely that other possible reasons that have been advanced for differences between the amount of intermediate AlO5 formation are more likely. For example the amount of intermediate AlO5 formation could be determined by whether the hydroxyls in the octahedral layer are either cis–cis or cis– trans oriented, where cis has O–Al–OH and trans OH–Al–OH con?gurations and this varies between different illites [44]. At 1000 1C very signi?cant changes occur in the spectra. For example, the silicon moves from a shift of ?92 ppm, typical of a Q3 peak, to ?101 ppm in the Q4 chemical shift range. This is indicative of the silica layer now becoming a connected three-dimensional network. The 29Si NMR peak has also signi?cantly broadened,

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 41

indicating that the new silica-based phase forming is much more structurally disordered than the initial layer. 27 Al MAS NMR also shows large changes in heating from 900 to 1000 1C. The AlO5-content drops rapidly and the AlO4 changes peak position from 65 to 55 ppm. There are two distinct AlO6 sites, both showing a signi?cant tail, indicating a large distribution of quadrupole interactions, although the spectrum is dominated by the tail from the peak with an isotropic chemical shift of 18.8 ppm. The large changes in both the 29Si and 27Al MAS NMR spectra are indicative of large structural changes that are occurring at this point. Thermal analysis shows a broad endothermic event from $900 to 1100 1C, which is probably due to the layers forming three-dimensional nanodomains and silica-based layer forming a glass. The changes in the 29Si chemical shift and the formation of the AlO4 peak at 52 ppm are both consistent with formation of an aluminosilicate glass. Potassium in the interlayer space experiences some electric ?eld gradient as indicated by the linewidth and asymmetry of the peak. As the aluminosilicate glass forms, the potassium shows signi?cant broadening, probably an indication that potassium is taken up by the aluminosilicate glass, and in the glass potassium experiences a much larger electric ?eld gradient. It is interesting to compare these results with those from other illite-rich clays such as Lode [31] and Silver Hill [32]. After glass formation the 27Al MAS NMR spectra from these samples showed a peak at $52 ppm, indicating that all the aluminium became tetrahedrally coordinated. The dif?culty with these previous NMR studies of these clays was the largely unknown effect of the iron-content, especially with respect to the differential loss of signal from different phases as the iron present does not partition equally between them. In the current illite, where interpretation is not hampered by signal loss through paramagnetic broadening, there is certainly signi?cant AlO6 content. There are two AlO6 peaks, one of which at 4 ppm is probably from mullite. This mullite formation appears to occur before the 1200 1C found for halloysite [25] and kaolinite [16,20,22]. However, there is certainly another AlO6containing phase which could be a g-alumina or spinelbased phase present in this illite at these temperatures corresponding to the peak at $12 ppm. Both the shift and wQ of the peak at $12 ppm are characteristic of a transitional alumina [12,45,46]. Then with more heating, further, but more subtle changes occur with the main peak in the silicon spectrum shifting from ?101 at 1000 1C to ?95 ppm at 1600 1C. The AlO6 peak at $4 ppm gets more intense while that at 12 ppm decreases. A new peak at 15 ppm grows after heating above 1400 1C. These changes can be explained by the potassium aluminosilicate glass phase being present throughout this temperature range, but is gradually changing its composition, dissolving more

aluminium as the heat treatment temperature is increased. The evidence for this is that the average 29Si isotropic chemical shift decreases, consistent with increasing numbers of aluminium next nearest neighbours, and the amount of AlO4 in the 27Al MAS NMR spectra increases. The accepted overall illite formula (K1.5Al4(Si6.5Al1.5)O20(OH)4) is in very good agreement with the chemical composition quoted in Table 1. Hence, as the silica-based layers form the initial potassium aluminosilicate glass, the Si:Al atomic ratio is likely to be Si6.5Al1.5 charge balanced with K1.5. Based on this ratio of metals and simply scaling to re?ect the formulae of the common oxide components gives a chemical composition of K2O.Al2O3.8.6SiO2. The eutectic composition in this region of the phase diagram corresponds to K2O.Al2O3.11.2SiO2 with glass of this composition ?rst appearing [47,48]. Then the temperature dependence of the liquid phase boundary is such that the silica-content of the liquid phase decreasing until at $1100 1C the liquid phase should then match that of the initial layer composition (i.e. K2O.Al2O3.8.6SiO2). At higher temperatures still the liquid should also have this composition, but is in contact with an aluminium-rich phase so it is plausible that a further increase in the aluminium-content might occur. Hence, the glassy phase that initially forms can be expected to be more silica-rich than the average layer composition. As it is heated to higher temperatures the phase diagram tie lines to the liquid phase indicate an increasing aluminium-content. Hence, the observed trends in the 29 Si chemical shift can be rationalised in terms of the expected changing composition of the glass. The 29Si MAS NMR spectrum from the sample heated to 1200 1C shows the small, sharp peak at ?86 ppm that is partially resolved from the main glass resonance. This probably corresponds to crystallisation of mullite, indicated by the small, sharp endothermic event at $1200 1C. The 27Al MAS NMR spectra also show formation of an additional aluminous phase above 1400 1C. XRD is dominated by mullite with some aAl2O3, the latter agreeing with the 15 ppm peak, and the glass component of the sample does not show up strongly in the XRD. There is an interesting comparison of the 27Al NMR interaction parameters that have been determined for the phases here and those that have appeared previously in the literature. Mullite has at least three distinct aluminium sites, one AlO6 and two AlO4, with one of the AlO4 having an unusual isotropic chemical shift at 53 ppm from tetrahedra joined as triclusters [49,50]. Detailed spectral simulations using satellite transition spectroscopy showed 3 different AlO4 sites, with the AlO6:AlO4 ratio $45:55. Considering the mullite in the sample heated to 1600 1C, from the intensity observed in the AlO6 centreband associated with the mullite, this could mean that up to $20% of the AlO4 in these 27Al NMR spectra could be from the

ARTICLE IN PRESS
42 D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43

mullite, although mullites with much lower AlO4content can occur [25]. However, no AlO4 site from the mullite can be distinguished from the main AlO4 from the glass in the 1D MAS NMR spectra. The shifts of the 2 AlO6 sites at 17.1 and 8.9 ppm agree, within the errors, to the known values for a-Al2O3 and mullite, respectively. However, the quadrupole coupling constants a-Al2O3 wQ is $2 larger than well crystallised, single phase samples [45]. This is probably a result of it still being effectively nanocrystalline and re?ected in the local sites being more distorted. For the AlO6 in mullite there is excellent agreement between the wQ from the simulation and the estimate made from the linewidth both of which give 8.5 MHz, and there is reasonable agreement with the previous study of a mullite that gave wQ of 7.3 MHz [50]. For this site the lineshape has a very pronounced tail indicative of a signi?cant spread of interaction parameters at these sites. Mullite is a solid solution Al2(Al2+2xSi2?2x)O10?x and there have been careful studies examining how the NMR parameters and occupancies of the different sites vary. For 27Al the NMR parameters and the AlO4:AlO6 ratio vary with composition (x) [51,52]. Also, the peak intensities of the different 29Si sites vary with the aluminium/vacancy composition of the sample [51,53,54]. Hence, in principle if the 29Si and 27Al MAS NMR spectra can be accurately simulated more details about the nature of the mullite and any changes with heat treatment would be revealed. The problem here is that the mullite is occurring with at least two other phases with signi?cant spectral overlap that precludes suf?cient accuracy to allow this approach to be followed. The multinuclear NMR data presented here reveal some key details of the structural breakdown of this illite. In comparing the breakdown of illite to other clays there are some points in common and clear differences. The silica and gibbsite layers to some extent behave somewhat independently as the sample is heated. As dehydroxylation occurs the average aluminium coordination is reduced, as a mixture of AlO5 and AlO4 form from AlO6. The amount of AlO5 that forms is intermediate to that seen in other clay minerals. Effectively, the layers start to form separate phases. When the silica layer starts to form a separate aluminosilicate glass it dissolves the interlayer potassium ions. From the amorphous dehydroxylated alumina phase a mixture of other aluminous phases, probably spinel/transitional alumina and mullite form. Then at higher temperatures some atomic intermixing occurs, with the glass phase probably increasing its aluminiumcontent. Although not well resolved, which makes accurate deconvolution dif?cult, there is certainly a 29 Si peak at $?86 ppm, that appears to increase with increasing heat treatment For chemically simpler clays such as halloysite the mullite phase initially formed is alumina-rich and its silicon-content increases at higher

temperatures [25]. It may be that the same thing happens here. The resolution observed here in the 29Si MAS NMR spectra, has shown for the ?rst time in an illite that the glass composition probably varies with heat treatment, consistent with increasing AlO4-content of the sample. The assignments and suggestions made here for the thermal breakdown of this very low iron-content illitic clay could be further tested by multiresonance NMR experiments such as 1H-29Si cross-polarisation and correlation, and multiple quantum and satellite transition MAS of 27Al. However, this paper illustrates the ability of multinuclear one-dimensional MAS NMR to give insight into the atomic scale structural rearrangements that occur in the thermal decomposition of a clay.

5. Conclusion This report of multinuclear solid-state NMR from a very low iron-content illitic clay has provided atomic scale insight into the structural rearrangements that occur in this group of clays as they are heated. It improves on previous work on such clay minerals through the study of 4 separate nuclei, and the fact that complications resulting from differential signal loss due to paramagnetic iron are removed. The NMR complements bulk thermal and powder XRD analysis. Loss of interlayer water has little effect on the clay layers and is largely reversible, although the 1H peak assigned to this water does shift on thermal cycling. Once hydroxyls start to come off from the gibbsite layer the associated formation of lower coordination aluminium species (both AlO4 and AlO5) is observed, with continued heating producing slightly more AlO5, but considerably more AlO4. As the gibbsite layer dehydroxylates there are only small corresponding changes in the neighbouring silica layers. After dehydroxylation the major structural event occurs between 900 and 1000 1C where the silica layer forms an aluminosilicate glass that contains the potassium ions. At the same point, alumina-rich mullite and spinel/g-Al2O3 form. A common theme observed in all clays to date studied in detail by solid-state NMR is that at the point where the remnant two-dimensional layer structure is lost the silica and gibbsite layers form essentially separate threedimensional nanodomain structures, although they are intimately mixed. The data collected here also suggests that the glass subsequently changes composition with heat treatment, dissolving more aluminium. At 1600 1C the sample contains a mixture of potassium aluminosilicate glass, mullite and probably a-Al2O3, although for the crystalline samples the quadrupole interactions and the spread of values suggest that the local structure remains quite distorted even after signi?cant heating. Insight is clearly provided by MAS NMR studies, such

ARTICLE IN PRESS
D.L. Carroll et al. / Solid State Nuclear Magnetic Resonance 28 (2005) 31–43 43

that thermal processing of these industrially important materials can be better understood.

Acknowledgments The EPSRC and University of Warwick are thanked for funding NMR equipment, the Nuf?eld Foundation is thanked for providing a summer bursary to D.L. Carroll and the Leverhulme Trust for providing a visiting professorship to T.J. Bastow. S. Certoks, R. Svinka and V. Svinka, Riga Technical University are thanked for provision of the sample and an accurate chemical analysis. References
[1] G.W. Brindley, M. Nakahira, J. Am. Ceram. Soc. 42 (1959) 311. [2] G.W. Brindley, J. Lemaitre, Chemistry of Clays and Clay Minerals, Longman, Harlow, 1990. [3] W.E. Worrall, Clays and Ceramic Raw Materials, Elsevier, Barking, 1986. [4] R.E. Grim, R.H. Bray, W.F. Bradley, Am. Mineral. 22 (1937) 813. [5] W.A. Deer, R.A. Howie, J. Zussman, An Introduction to the Rock-Forming Minerals, Longman Group, Hong Kong, 1992. [6] C.M. Earnest, Lecture Notes in Earth Sciences, Springer, Berlin, 1991, p. 270. [7] R.E. Grim, W.F. Bradley, J. Am. Ceram. Soc. 22 (1940) 242. [8] P.J. Michael, W.R. McWhinnie, Polyhedron 8 (1989) 2709. [9] E. Murad, U. Wagner, Clay Minerals 31 (1996) 45. [10] G. Engelhardt, D. Michel, High Resolution Solid-State NMR of Silicates and Zeolites, Wiley, Chichester, 1987. [11] K.J.D. MacKenzie, M.E. Smith, Multinuclear Solid-state NMR of Inorganic Materials, Pergamon, Oxford, 2002. [12] M.E. Smith, Appl. Magn. Reson. 4 (1993) 1. [13] D.E. Woessner, Am. Mineral. 74 (1989) 203. [14] K.J.D. Mackenzie, I.W.M. Brown, R.H. Meinhold, M.E. Bowden, J. Am. Ceram. Soc. 68 (1985) 293. [15] T. Watanabe, H. Shimizu, K. Nagasawa, A. Masuda, H. Saito, Clay Minerals 22 (1987) 37. [16] J. Sanz, A. Madani, J.M. Serratosa, J.S. Moya, S. Aza, J. Am. Ceram. Soc. 71 (1988) C418. [17] J.F. Lambert, W.S. Millman, J.J. Fripiat, J. Am. Chem. Soc. 111 (1989) 3517. [18] J. Rocha, J. Klinowski, Phys. Chem. Minerals 17 (1990) 179. [19] R.C.T. Slade, T.W. Davis, J. Mater. Chem. 1 (1991) 361. [20] D. Massiot, P. Dion, J. Alcover, F. Bergaya, J. Am. Ceram. Soc. 78 (1995) 2940. [21] K.J.D. Mackenzie, J.S. Hartman, K. Okada, J. Am. Ceram. Soc. 79 (1996) 2980. [22] P.A. Schroeder, R.J. Pruett, Am. Mineral. 81 (1996) 26. [23] J. Rocha, J. Phys. Chem. B 103 (1999) 9801.

[24] H.P. He, J.G. Guo, J.X. Zhu, C. Hu, Clay Minerals 38 (2003) 551. [25] M.E. Smith, G.S. Neal, M.B. Trigg, J. Drennan, Appl. Magn. Reson. 4 (1993) 157. [26] J.J. Fitzgerald, S.F. Dec, A.I. Hamza, Am. Mineral. 74 (1989) 1405. [27] P.J. Sanchez-Soto, I. Sobrados, J. Sanz, J.L. Rodriguez, J. Am. Ceram. Soc. 76 (1993) 3024. [28] J.J. Fitzgerald, A.I. Hamza, S.F. Dec, C.E. Bronnimann, J. Phys. Chem. 100 (1996) 17351. [29] I.W.M. Brown, K.J.D. MacKenzie, R.H. Meinhold, J. Mater. Sci. 22 (1987) 3265. [30] S.R. Drachman, G.E. Roch, M.E. Smith, Solid-state NMR 9 (1997) 257. [31] G.E. Roch, M.E. Smith, S.R. Drachman, Clays Clay Minerals 46 (1998) 694. [32] C.J. McConville, G.E. Roch, W.E. Lee, M.E. Smith, CIMTEC 2002 Science for New Technology of Silicate Ceramics, (2003) 97. [33] M.E. Smith, Ann. Rep. NMR Spectrosc. 43 (2000) 121. [34] T.J. Bastow, J. Chem. Soc., Faraday Trans. 87 (1991) 2453. [35] J.-F. Lambert, R. Prost, M.E. Smith, Clays Clay Minerals 40 (1992) 253. [36] A.C. Kunwar, G.L. Turner, E. Old?eld, J. Magn. Reson. 69 (1986) 124. [37] M.E. Smith, E.R.H. van Eck, Prog. NMR Spectrosc. 34 (1999) 159. [38] T.F. Kemp, M.Sc. Thesis, University of Warwick, 2004. [39] D. Massiot, C. Bessada, J.P. Coutures, F. Taulelle, J. Magn. Reson. 90 (1990) 231. [40] S. Hayashi, T. Ueda, K. Hamamizu, E. Akiba, J. Phys. Chem. 96 (1992) 10922. [41] J. Sanz, J.M. Serratosa, J. Am. Chem. Soc. 106 (1984) 4790. [42] C.P. Herrero, J. Sanz, J.M. Serratosa, J. Phys. Chem. 93 (1989) 4311. [43] S.K. Lausen, H. Lindgreen, H.J. Jakobsen, N.C. Nielsen, Am. Mineral. 84 (1999) 1433. [44] V.A. Drits, G. Besson, F. Muller, Clays Clay Minerals 43 (1995) 718. [45] D. Freude, Encyclopedia of Analytical Chemistry, in: R.A. Meyer (Ed.), Wiley, Chichester, 2000, p. 12188. [46] G. Kunath-Fandrei, T.J. Bastow, J.S. Hall, C. Jager, M.E. Smith, ¨ J. Phys. Chem. 99 (1995) 15138. [47] J.F. Schairer, N.L. Bowen, Bull. Soc. Geol. Finlande 20 (1947) 74. [48] J.S. Brown, in: A.M. Alper (Ed.), Phase Diagrams in Advanced Ceramics, Academci Press, London, 1995, p. 55. [49] L.H. Merwin, A. Sebald, H. Rager, H. Schneider, Phys. Chem. Minerals 18 (1991) 47. [50] G. Kunath-Fandrei, P. Rehak, S. Steuernagel, C. Jager, Solid¨ state NMR 3 (1994) 241. [51] I. Jaymes, A. Douy, D. Massiot, J.P. Coutures, J. Chim. Phys. PCB 92 (1995) 1851. [52] P. Rehak, G. Kunath-Fandrei, P. Losso, B. Hildmann, H. Schneider, C. Jager, Am. Mineral. 83 (1998) 1266. ¨ [53] T. Ban, K. Okada, J. Am. Ceram. Soc. 76 (1993) 2491. [54] I. Jaymes, A. Douy, D. Massiot, J.P. Coutures, J. Non-Cryst. Solids 204 (1996) 125.


赞助商链接
相关文章:
更多相关标签: