Thermal Reactions of Chrysotile Revisited: a 2Esi and 2Smg MAS NMR Study

American Mineralogist, Volume 79, pages 43-50, 1994

Thermal reactions of chrysotile revisited: A 2esi and 2sMg MAS NMR study

K.J.D. MlcKruzrn. R. H. Mrrxuor-o New Zealand Institute for Industrial Researchand Development, P.O. Box 3l-310, Lower Hutt, New Zealand

Ansrntcr The thermal decomposition sequenceof chrysotile samples from New Zealand. ar'd Canadahas been found by'?eSiand'z5Mg MAS NMR to be more complexthan previously indicated by X-ray diffraction and electron optical studies.The greaterpart of the mineral dehydroxylates,producing a Mg-rich X-ray amorphousphase containing essentially Mg in octahedral coordination and characterizedby a broad 2eSiresonance centered at about -73 ppm with respectto tetramethylsilane.This transformsto forsterite,MgrSiOo, at 670-700'C, at which temperaturethe unreactedchrysotile (possibly the deepestlayers of the fibrils) forms a different, Si-rich X-ray amorphous dehydroxylate. This phase,called dehydroxylateII, has a '7eSichemical shift (-97 ppm) not unlike that of talc, but with Mg sites that may be distorted beyond NMR detection, becausemost of the '5Mg spectral intensity is lost over this temperaturerange. Dehydroxylate II forms enstatite(MgSiOr) and free silica at 770-800 "C, further enstatite being formed by reaction of the silica with forsterite at I 150 "C. The stoichiometry of these reactions is consistent with NMR mea- surementsof the partitioning of Si betweenthe various phases.

INrnooucrtoN temperature range, with the first small crystallites of forsterite(MgrSiOo) appearing at about 620 "C and in some Chrysotile (white asbestos)is a predominantly fibrous samplesoverlapping with the remnantsof the chrysotile form of serpentine,Mg.SirOr(OH)o, the trioctahedral X-ray pattern (Brindley and Zussman, 19571-Martin, magnesiananalogue of the l:l layer-structurealumino- 1971).The dehydroxylatedproduct has been variously silicatekaolinite. Becauseof its historical importanceas describedas largely disordered(Ball and Taylor, 1963), an insulating material with refractory applications, the partly disordered (Brindley and Hayami, 1965), and propertiesand thermal reactionsof chrysotilehave been semiamorphousbut preservingsome of the original lat- extensivelyinvestigated. Although the industrial impor- tice order (Martin, 1977).ltscomposition is alsodebated; tanceofasbestiform mineralshas progressivelydeclined Ball and Taylor (1963) consideredthat it contains the as their health hazardshave becomeincreasingly recog- sameatomic ratiosas the parentphase, with the observed nized,their thermal decompositionreactions continue to topotaxy betweenthe reactant and product suggestingin- be of practical interest as a means of converting the ma- herited Mg-O order, rather than an inherited Si-O struc- terial to lesshazardous products (MacKenzie and Mein- ture, as implicit in a schemesubsequently suggested by hold, 1994a).Furthermore, there remain someinteresting Brindley and Hayami (1965). The thermal reaction questionsconcerning the mechanismof this thermal de- schemesso far advanced are summarized in Figure l composition reaction that also bear on the thermal be- Support for the structuralcontinuity of the silicateunits havior of related layer-structure silicates. rather than the Mg-O structure has recently been provid- When chrysotile is heated in air, it undergoesan en- ed by radial distribution function measurements,which dothermic weight loss of about l4olobetween about 600 indicate that the Si-O bond lengths remain constant and 720'C (Khorami et al., 1984),representing the re- throughout the reaction (Datta et al., 1987). The hetero- moval of structuralHrO. The resultingphase, chrysotile geneousreaction mechanismof Ball and Taylor (1963) dehydroxylate,is amorphousto X-rays,but in somesam- demandsa subsequentseparation into a Mg-rich phase ples contains a broad, low-anglediffraction band centered that becomes forsterite and a Si-rich phase that subse- at about l0-15 A @rindley and Zussman,1957). There quently becomesensratite (MgSior) (Fig. 1A). Brindley has been much speculationconcerning the possiblede- and Hayami (1965)accepted the Ball and Taylor concept greeofresidual structure in the dehydroxylatedphase and of the dehydroxylate, but, on the basis of quantitative the origin of the low-angle reflection, which may merely X-ray determination of the amount of forsterite subse- representdiffraction from long cavities observedby elec- quently formed, considered that all the available Mg is tron microscopy to open up in the fiber walls at these accounted for by that phase, leaving a residual material temperatures(Martin, | 97 7 ). of almost pure amorphoussilica, which further reactswith By comparison with kaolinite dehydroxylate, the the forsterite at higher temperaturesto form enstatite(Fig. amorphous phasefrom chrysotile exists over a very short lB). On the basisof electronmicroscopy, Martin (1977) 0003404x/94l0102-o043$02.00 43 MIcKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE

A. Ball - Taylor (f 9ffi1 B. Brindley - Hayami (19651 srgfiPJSItn 6rg"$fJgIln lchrysotilel lchrysotile I (/\ 9Mg"SirO"(OHlo 2llg"SirO"(OHln 4MgrSirO"lOHlo 3Ms"Si"O"(OH Fccepdrl g fd"r.Tl | rcg|on I I l€gpn I ll a_ 3Mg"SirO, I eug 4Mg.Si2O7 og 4Si I eoo t rc,tt]ua-" I F,-.-.-" I t8H20t I ettase i I ehase I 12H20 {+pores} + + 8H.O + 6HrO 35i02 tI +ttgl l*,N <- l-Mg''ichl '- l-silChl t Egktn I t rcglan t si 35i02 6Mg lGti"" I 35i

t TMgrSiOo TllgSiO" MsrSiOn M9SiO. 6llgSiO3 t f"*t"tit.l E;"tit;l |l-"toiG-I E""t"riGl

Fig. 1. Schematicrepresentation ofpreviously proposedchrysotile reaction sequences.(A) According to Ball and Taylor (1963), (B) accordingto Brindley and Hayami (1965),(C) accordingto Martin (1977).

has modified this scheme,suggesting that little separation diffraction observations, which show no sudden change of Mg and Si occursbelow 800 "C, below this tempera- in forsterite crystallinity at this temperature. Datta et al. ture, forsterite forms in regions of the disordered dehy- (1986) attributed the exothermto the partial converslon droxylate by consuming whatever Mg is locally available of chrysotile to forsterite and enstatite, the previously (Fig. lC). Martin suggestedthat the main countermigra- formed forsterite being the transformation product of a tion of Mg and Si occurs rapidly at about 800 "C, the different type of chrysotile particle (type I), which is more resulting spontaneouscrystallization of forsterite and en- reactive. Other workers are more vague, suggestingthat statite giving rise to the exotherm observed in the DTA the exothermis due to "conversionof the chrysotilean- traceat 810 "C (Martin, 1977).This is not consistentwith hydride" within the fibrils (Santos andYada, 1979). the temperature at which enstatite is usually reported to The outstanding problems in the chrysotile reaction se- appear(1000'C: Santosand Yada, 1979),although Mar- quence can thus be summarizedas (l) the nature and Iin (1977) reported the presenceof enstatite crystallites structure(ifany) ofthe dehydroxylate,and their influence in samplesheated at 800 'C for 24 h. on the further course of the reaction sequence,and (2) The origin of this exotherm at 8 10 'C has long been a the causeofthe exotherm at 810'C and its relation to sourceofdebate because,like the exothermat 980'C in the higher temperature products. kaolinite, it is not accompaniedby the suddenappear- As with the thermal reaction sequenceof kaolinite, anceofany new phase.Ball and Taylor (1963)attributed progressin both these areas has been hampered by the the exotherm to the change in O packing that results in lack of crystallinity of the intermediatephases. Solid-state the formation of forsterite, but that is not consistentwith nuclear magnetic resonancewith magic angle spinning the appearanceof forsterite at least 100 'C lower. The (MAS NMR) has provided new information on the be- reactionscheme of Brindley and Hayami (1965)implied havior ofSi and Al during the kaolinitereaction sequence that the exotherm representsthe recrystallization of dis- (MacKenzieet al., 1985),but the application of 'z5MgMAS ordered forsterite,although this is not supportedby X-ray NMR to magnesiosilicateproblems has only recently be- M,c,cKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE 45

come possible, with the increasing availability of high- spinningspeed 19.5-20 kHz, with a 90'(solids) pulseof field spectrometersand high-speedMAS probes.Building 3 ps with a recycle delay of 0. I s. A spectral width of on preliminary work by Dupreeand Smith (1988),which 100000 Hz was usedwith an acquisitiontime of 0.02 s, demonstratedthe feasibility of 2sMgMAS NMR in some giving 4096 points in the free induction decay.After the inorganic systems, the technique has been applied to acquisition of, typically, 400000 transients,the spectra studies of the thermal behavior of several Mg minerals, were processedusing a linear prediction method avail- including magnesite (MacKenzie and Meinhold, 1993), able in the Unity softwareto remove the base-linedis- dolomite (MacKenzie and Meinhold, 1994c), brucite tortions introduced by probe ringdown. For the linear (MacKenzie and Meinhold, 1994d), and hectorite prediction, data acquired after 20 ps with the receiver off (MacKenzie and Meinhold, 1994e). Working at high then had a further ten points (100 ps) dropped,and the magnetic fields and very high spinning speedsdecreases resulting free induction decayused to calculatethe linear some of the problems associatedwith the spin I ,5Mg prediction coefficientsand the missing first l2 points after nucleus(e.g., 100/o natural abundance,low 7, and consid- excitation. The resulting spectrumwas then checkedwith erablesecond-order quadrupolar broadening ofthe cen- those produced by starting the back linear prediction at tral transition). Using high fieldsand high spinning speeds, different points and by applying a polynomial baseline 2sMg the MAS NMR spectraof a number of representa- correction to the spectrum produced with data acquired tive mineral types have been collected, and the effect of immediately after the receiver-off period or after left- the interactionwith the quadrupolar2sMg nucleus of elec- shifting the free induction decayby a variable number of tric field gradients produced by deviations from octahe- points before the Fourier transformation.Gaussian line dral symmetry has been investigated(MacKenzie and broadeningof 200 Hz was applied to all the processed Meinhold, 1994b). spectra, which were referenced to a saturated solution In the presentstudy, MAS NMR is usedto probe the of MgSOo. behavior of the Si and Mg in the chrysotilereaction sequence,with particularattention to the natureofthe de- Rnsur-rs AND DrscussroN hydroxylate and the reactions related to the exotherm at The thermal analysistraces of both chrysotilesamples 810"C. are shownin Figure2. Both chrysotilesamples show similar thermal behavior, the low-temperatureendotherm ExprnrntrNr.q.L being due to the loss of HrO adsorbedon the fiber sur- Two chrysotile sampleswere used in this work. One faces.The endothermicweight loss at about 700'C due was from the Upper Takaka Valley asbestosdeposit, to dehydroxylation appearsfrom the DTG curve to con- northwestNelson, New Zealand(N.2. GeologicalSurvey sistof two overlappingprocesses, most clearlyseen in the sampleno. 4519),the other from the Cassiarmine, Brit- Cassiarsample (Fig. 2E). Similar multistagedehydroxy- ish Columbia, Canada(sample AK 262C91).The X-ray lation has also beenreported by Datta et al. (1986),who diffractionpatterns of both samplesindicated monophase attributed it to the presenceof type I and type II chrys- clinochrysotile.The major impurity (Fe) was presenrin otile particles of different reactivity. X-ray diffraction the New Z.ealandand Canadian samplesat levels of I . I 6 shedsno light on the possiblenature ofthese processes; and 1.12010,respectively. Thermal analysiswas carried samplesheated below 700 oCshow only chrysotilereflec- out in air at a heatingrate of l0'C/min, usinga Stanton tions, which graduallybecome less intense as heating pro- Redcroft TG770 thermobalanceand a Perkin-Elmer DTA gresses,whereas at 700'C, forsteritesuddenly appears, in 1700 differential thermal analyzer.Samples for NMR addition to very weakresidual chrysotile reflections. Both studieswere preparedby heatingfor 15 min in Pt foil- the Nelson and Cassiarsamples heated at 700 "C show lined ceramic boats in a laboratory furnace at tempera- the broad, low-anglereflection commented on by Brin- turesdictated by the thermal analysisresults. After with- dley and Zussman(1957) and Martin (1977).The only drawal from the furnace, the sampleswere examined by changein the X-ray pattern accompanyingthe exotherm X-ray powder diffraction using a Philips PWl700 com- at 810'C is the loss of the broad, low-anglereflection. puter-controlled diffractometer with CoKa radiation and Heating to progressivelyhigher temperaturesresults in a graphitemonochromator. the eventualappearance ofsome enstatite,at 1000"C in The room-temperature,esi NMR spectrawere record- Cassiarchrysotile, and slightlyhigher temperatures in the ed both aI 4.7 T (Varian XL-200 spectrometerwith a Nelson sample. Doty MAS probe, spinning speed4 kHz) and at l l.7 T (Varian Unity 500 spectrometer5-mm Doty MAS probe, NMR spectroscopy spinningspeed 8 kHz). The lI.7 -T spectrawere acquired Typical room-temperature2eSi spectra at ll.7 T of un- usinga 90'pulseof 6 ps with a delayof 0.5 s, all the reSi heatedand heatedchrysotile samplesare shown in Figure spectrabeing referencedto tetramethylsilane(TMS). Af- 3 (Nelsonchrysotile) and Figure4 (Cassiarchrysotile). ter acquisitionofseveral thousand transients, the spectra The chemical shift of the single Si resonancein the were processedwith a Gaussianline broadeningof 50 unheatedminerals (-93 ppm) is in reasonableagreement Hz. The room-temperature2sMg spectra were acquired with the one reportedvalue of -94 ppm for serpentine at lI.7 T using a Doty 3.5-mm high-speedMAS probe, of unspecifiedmorphology (Magi et al., 1984).The spec- 46 MIcKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE

A TG, Cassiar

b< 10 o B o 15 TG, Nelson 0' o o 6

c DTG, Cassiar

D DTG Nelson

E DTA, Ca 29-Si chemical shift 6, (w.r.t- TMS) Fig.3. Room-temperature'?esiMAS NMR spectraof Nelson chrysotileat ll.7 T heatedto variouslemperatures. A : un- hearedand heared to 600,B : 650,c : 700,D : 750,E: 800, F F: 850,G: 1000,H: 1300'C.Units in partsper million. DTA,Nelson

protosilicate gels formed from MgCl, and NaoSiOoin which the ratio of Mg to Si is l: I (Hartman and Millard, 1990).The materialgiving rise to theseresonances is describedas amorphous,with a low degreeof Si polymer- ization (Hartman and Millard, 1990).A similar broad o 2oo 4oo 600 Boo looo peaknear -80 ppm observedin a sampleof MgSiO, with Temperature(oC) perovskite structure was attributed to t4rsi in an amorphous phase with an average of two bridging O atoms Fig.2. Thermalanalysis traces of Nelson(New Zealand) and per tetrahedron,suggesting a composition closeto MgSiO, 'Clmin Cassiar(Canadian) chrysotile samples. Heating rate l0 (Kirkpatrick et al., l99l). Sincethe chemicalshift of the in air. A and B arethermogravimetric curves; C and D aredif- amorphous dehydroxylate peak in the present samples ferentialthermogravimetric curves. E andF aredifferential ther- spansthe orthosilicate region up to the inosilicate mal analysiscurves. Q0 Q'? region, it cannot distinguish between the composition of incipient forsterite(Qo,6: -62 ppm) and incipient en- tra are unchangedby heatingbelow 600 oC,but at about statite (Q2,6 : -83 ppm); the broadnessof the peak this temperature, a slight asymmetry begins to develop suggeststhe presenceofa rangeoftetrahedral configura- in the low-field side of the main peak (Fig. aA). By 650 tions in this phase,which we designatedehydroxylate I. 'C, this feature has developedinto a broad hump at about Heating to 700 "C producesa decreasein the chrysotile -72to -74 ppm (Figs.38, 4C). Sincethe only crystalline resonanceand the appearanceof a sharp new resonance phasedetectable in thesesamples is chrysotile,which is aI -61.5 ppm, consistentwith the chemicalshift reported also evident from the resonanceat -93 ppm in these for forsterite(-61.9 ppm: Magi et a1.,1984; -61.6 ppm: samples,the new broad peak is apparentlyassociated with Hartman and Millard, 1990). The ll.7-T spectrum of the amorphous anhydrous phaseformed in the first of the Cassiarchrysotile heated at 700 'C (Fig. aD) revealsthat two weight-lossprocesses. Similar broad peaksat about the broadnessof the residual chrysotile resonanceat -93 -73 and *80 ppm have beenreported in the spectraof ppm is due to the appearanceof an extra peakat -96.9 M.q.CKENZIEAND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE ppm, which survivesto at least800'C (long after chrysotile has disappearedfrom the X-ray diffraction pattern of the sample), and is more thermally stable than the broad resonanceof the amorphousphase. The chemical shift of this new peak is typical of the Q3 Si atoms oc- curring in layer-structure phyllosilicates such as chrysotile itself, or the related 2: I layer-structure silicate talc, MguSirOro(OH)o,for which the reported ,,Si chemicalshift is -97.2 ppm (Sherriffetal., 1991)to -98.1 ppm (Magi et al., 1984).Although talc is one of the decomposition products of chrysotile under hydrothermal conditions, where it forms in addition to forsterite (Ball and Taylor, 1963),its presencein the presentsamples could not be confirmed by the observation of its characteristic basal spacingin X-ray diffraction patterns. A more likely pos- sibility is that the resonanceat -97 ppm is related to the l: I layer-structureof chrysotile,rendered X-ray amorphous by dehydroxylationbut still containing elements ofthe original silicatestructure, i.e., a Si-rich dehydroxylate phase,which will be called dehydroxylate II. On further heating above 800 'C, the resonancesat -75 and -97 ppm are reduced,and a broad peak ap- pearsat about -108 ppm, correspondingto amorphous silica (Figs. 3F, 4H). The forsteriteresonance at -61.5 ppm, which grows at the expenseof the dehydroxylate I -60 -80 -100 -120 -60 -80 -1( peak at -75 ppm, initially showsa shoulderor tail at about -58 ppm, this disappearswith further heatingas 29-si chemical shift 6 , (w.r.t. TMS) the crystallinity of the forsterite improves. The first ap- Fig.4. Room-temperature'?eSiMAS NMR spectraof Cassiar pearanceof enstatite in the X-ray pattern coincides with : 2eSi -82 chrysotileat | 1.7T heatedto varioustemperatures. A 600,B the appearanceof a resonanceat about ppm : 650,C : 670,D : 700,E : 730,F: 750,G : 800,H : (Fig. 3G), which grows and broadenswith further heating 900"C. Units in partsper million. (Fig. 3H). The characteristicsofthis peakcorrespond bet- ter with those reported for orthoenstatite (single, broad peak centeredaI -82 ppm: Magi et al., 1984)than with havior suggeststhat in chrysotile there are at least two those for clinoenstatite(two peaksat about -81.6 and component lines with very different Zf time constants, -83.8 ppm: Hartman and Millard, 1990;Janes and Old- which have substantially different signal decaycharacter- field, 1985).The initial appearanceof both amorphous istics during ringdown. Processingthe spectrausing a range silicaand enstatiteseems to be at the expenseof the phyl- of left-shifts and fitting the rwo component heights to a losilicate-like dehydroxylate II, although the greatestde- Lorentzian decay gives values for ?"f of 58 and 315 ps, greeof enstatiteformation occursat I130-1150 'C by corresponding to a Lorentzian line width of about 5520 reaction of the excesssilica with forsterite, as suggested and 1010 Hz for the componentsat about -90 and 8 by Brindley and Hayami (1965). ppm, respectively. Although it may be possible that a The 25MgMAS NMR spectraof unheatedNelson and single quadrupolar pattern could have broad and narrow Cassiarchrysotile samples, together with Cassiarchrys- components, we have not hitherto observed the pro- otile heatedto various temperatures,are shown in Figure nounced effectsseen here and propose an alternative ex- 5. In order to processthe 'z5Mgspectra to eliminate the planation. base-linedistortions causedby probe ringdown, the free The ratio (about 9:l to 16:1)of thesetwo typesof site induction decayacquired after a receiver-offperiodof20 with very different electric field gradient (EFG) charac- ps needsto be left-shiftedby eight points (at l0-;rs inter- teristicssuggests that they are not the Ml and M2 octa- vals); this is equivalent to sampling the free induction hedral sites,which, if distinguishable,should be much decay 100 ,.s after excitation.Although this techniqueof more nearly equally populated. Rather, their origin may left-shifting the free induction decay to eliminate probe lie in the fibrousmorphology of the sample;it seemsrea- ringdown effectsappears to be quite satisfactoryfor some sonablethat the octahedralsites in the more tightly coiled other Mg minerals (MacKenzie and Meinhold,, 1994b\, layers located at greatest depth inside the fibrils might the effect in both Nelson and Cassiar chrysotile is to experiencea different EFG from that of the sites in the change the relative areas of the contributing resonances surfacelayers. The Nelson chrysotile spectrum is slightly so that the spectrum processedwith no left-shift has an different from that of Cassiar, showing a third, broad upfield component that is much more intense. This be- component, near -22 ppm. Further evidencethat the 48 M^I.cKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE

changein the shapeof the 2sMgspectra (Fig. 5D, 5E); but at 850 'C the spectrumcould be that of a single EFG- split site (Fig. 5F). However, not all the Mg is accounted for in these spectra; integrated intensity measurements indicate that by 650 "C, only about 360/oof the original Mg intensity is observed, and this figure falls to a mini- mum value of 3oloby 800 "C, rising slightly to 4oloby 850 "C. A similar decreasein the 'z?Alintensity observeddur- ing the dehydroxylation of kaolinite (MacKenzie et al., 1985)was interpretedin terms of the distortion of sites, broadening their NMR resonancebeyond detection. A similar explanation may apply here, with the result that only the Mg in comparativelywell-ordered octahedral sites in chrysotile or its dehydroxylatesare representedin Fig- ure 5. Using the data from the single-crystal study of forsteriteby Derighetti et al. (1978),we have calculated the MAS spectra for the Ml and M2 sites (MacKenzie and Meinhold. 1994b).These calculations show that the maximum peak height is only a few percent of the intensity expectedfor a line of 3 ppm in width, becausemost of the intensity is presentas a broad baseto the peak (the calculatedpeak widths at halfheight and one-tenth height are 72 and 446 ppm for Ml and l19 and 288 ppm for M2, respectively).Thus, the large EFG values in forsterite causethe NMR peaks to be broadenedand the maximum signal height decreasedto such an extent that the signal is largely lost in the background noise. It is this EFG broadening effect that is responsible for the observed loss of the Mg signal. 25 Mg chemical shift 6 (ppm), w r.t. Mg SOa Soln The NMR results suggestthat the thermal behavior of Fig. 5. Room-temperature'z5Mg MAS NMR spectraat I1.7 chrysotile is more complex than postulated in previous T of (A) unheatedNelson chrysotile, and (B-F) Cassiarchryso- reaction sequencesbased on X-ray diffraction or electron tile, unheatedand heatedto various temperatures.B : unheated microscopy, in that two different dehydroxylates are andheated to 650,C: 700,D :750,E: 800,F: 850'C. formed. DehydroxylateI, formed first, containsSi envi- ronments with chemical shifts spanning the range of or- thosilicatesand inosilicates,but with interatomic dis- tances and Mg coordination numbers not very different EFG splittings are morphology-relatedis provided by the from the parent chrysotile(Datta et al., 1987). 25Mg spectrum of the nonfibrous serpentine antigorite, As dehydroxylation proceedsfurther, forsterite crystal- which contains only the site of longer 7"f (lower EFG) lizes from this amorphous phase and dehydroxylate II (MacKenzie and Meinhold, 1994b). The apparent pres- appears,which, althoughX-ray amorphous,has well-de- enceof two and three Mg sites in the Cassiarand Nelson fined Si sites with a fairly narrow chemical-shift range, chrysotile samples,respectively, may also be the origin but shifted upfield by about 5 ppm from the resonance of the differencesin the thermal behavior of the two ma- of the parent chrysotile. Dehydroxylate II also has greater terials, as reflected in the less clearly resolved dehydrox- thermal stability than the amorphous dehydroxylated ylation endotherm of the Nelson sample (Fig. 2F). To try phaseoriginally formed (it is still presentabove 800'C). to compensatefor the effects of T\ decay,all the '?sMg Its formation therefore involves minimal disruption of spectrawere processedusing a linear prediction method. the parent Si structure,but becauseit is virtually anhy- The spectraof samplesheated to the onset of dehy- drous, considerable changesmust have occurred in the droxylation at 650'C suggestonly minor changeshave coordination sphereof the Mg. This is also consistent occurred in the octahedralenvironment of the NMR-vis- with the loss of octahedral25Mg NMR intensity and with ible Mg (Fig. 5B),but possiblywith a slightlydiminished radial distribution function results that indicate a reduc- higher EFG component. Upon the formation of dehy- tion in the coordination number of the Mg, which setsin droxylate I at 700 "C, the 25Mgspectrum narrows, per- ar 700-7 50 'c (Datta et al., 198'7). haps because of the disappearanceof the higher EFG Above 800'C, thereis an exothermicreaction that co- component(Fig. 5C). Heating to 800 "C, in which tem- incideswith the completeloss of OH, the appearanceof perature range further forsterite is formed and dehydrox- further forsterite,and the first indicationoffree silica,the ylate II appears and disappears,produces little further last two at the expenseofdehydroxylate II. This suggests MecKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE 49

chrysotile(f = -92 ppm)

4tg"SirO.(OHln

silice(0 =-1og ppm) 2tgrSiOn 2SiO2 ilgSiO" + tf"r"t".il"-l tsltlca l l;""t"tlt.l ( 6 -.6r.6ppml ( 6 =- lro pF|rl ( 6 =-qrppmt

1000 Tamp6rEture(oC) i/"nrc silgsiO3 Fig.6. Distributionof Si amongthe various phases in heated Nelsonchrysotile, estimated from the 4Si NMR spectraby fit- ,eSi ting peaksto the indicatedresonances. The chemicalshifls (6 =-83pel||l ofthe variousphases are shown in parentheses. Fig.7 . Schematicrepresentation ofthe proposedthermal de- compositionsequence of chrysotile.The "Si chemicalshifts of that the exotherm may be related to the collapse of the the variousphases are shown in parentheses. O layers, with the separation offree silica, as is also the casewith the exothermof kaolinite at 980 "C (Brown et al., 1985). Further heating producesa slow but steady droxylate II, which is just appearing.If, as seemsproba- increasein the amount of enstatite,which acceleratesat ble, the forsterite already present by this stage has also about I 150 "C becauseof the reaction of the amorphous formed at the expenseofdehydroxylate I, then this phase silica with some of the forsterite, as suggestedby Brindley must account for 65-750/oof the total Si in the system. and Hayami (1965). When this is compared with the two alternative hetero- geneousreaction schemes previously proposed, it can be Partioning of the Si among the various phases seenthat the Ball and Taylor scheme(1963) (Fig. lA) To make a semiquantitative estimate of the relative implies the initial transformationof I mol of chrysotile proportion of Si in each phase at each temperature, the into I mol of forsterite * I mol of an amorphous phase 2eSispectra were synthesizedby superposition of a num- with an enstatite-likecomposition, i.e., the initially formed ber ofGaussian peakshapes and the relativeareas under forsterite would contain 500/oof the available Si. The each fitted peak calculated. A similar approach was schemeof Brindley and Hayami (1965)(Fig. lB) propos- adoptedby Brindley and Hayami (1965),who usedquan- es the initial conversionof all the availableMg into for- titative X-ray diffraction to estimatethe amount of Si in sterite ( I .5 mol of forsterite, containing 7 5o/oof the avail- each phase,but the NMR resultsprovide a more com- ableSi) with the excesssilica separating as the amorphous plete Si balance,since they take into account the Si in the phase.Thus the presentresults for the initial stagesofthe amorphousphases that is inaccessibleto XRD. The rel- reaction are more in accord with the schemeof Brindley ative Si contentsin the various phasesformed in Nelson and Hayami (1965), in that about 650/oof the available chrysotile as a function of heating temperatureare shown Si forms forsterite, but at the expenseof dehydroxylate I, in Figure 6. which is Mg-rich, and retains nothing of the original Si If comparable (or fast) relaxation times are assumed configuration. The remaining Si (about 300/o)is associated for all the phasespresent, Figure 7 indicates that at 700 with dehydroxylateII, which, althoughSi-rich, also con- "C, 100/oof the Si is present in unreactedchrysotile, 27olo tains Mg and crystallizesat about 800'C to enstatite,with in forsterite, 500/oin dehydroxylate I, and 140/oin dehy- the separationofthe excessSi as SiO,. Thus, the overall 50 MIcKENZIE AND MEINHOLD: THERMAL REACTIONS OF CHRYSOTILE

reaction sequence,as suggestedby the NMR results,con- Datta, A.K., Mathur, B.K., Samantaray,B.K., and Bhattachedee,S. (1987) tains similaritiesto that of Brindley and Hayami (1965) Dehydration and phasetransformation in chrysotile asbestos:A radial distribution analysisstudy. Bulletin of Materials Science,9, 103- I 10. but is more complex. The suggestedreaction sequence, Derighetti, B., Hafner, S., Marxer, H., and Rager,H. (1978) NMR of 2'Si basedon the NMR results,is shown schematicallyin Fig- and '?5Mgin MgrSiOo with dynamic polarization technique. Physics ure 7. Irtters.66A. 150-152. One outstanding question raised by the reaction se- Dupree, R., and Smith, ME. (1988) Solid-state magnesium-25 NMR Communi- quence spectroscopy.Journal of the Chemical Society, Chemical of Figure 7 is why chrysotile dehydroxylates cations,1483-1485. yielding two different dehydroxylatephases. A clue to the Hartman, J.S., and Millard, R.L. (1990) Gel synthesisof magnesiumsil- answer may lie in the different ways in which the chrys- icates:A 29-Si magic angle spinning NMR study. Physicsand Chem- otile lattice decomposes under dry and hydrothermal istry of Minerals,17, l-8. (1985) nuclear mag- conditions (Ball and Taylor, 1963).Dry decomposition Janes,N, and Oldfield, E. Prediction of silicon-29 netic resonancechemical shifts using a group electronegativity ap- (such as would be expectedin the outer layersofthe fibril proach: Applications to silicate and aluminosilicate structure. Journal bundles,where the HrO formed can readily escape)leads ofthe American Chemical Society, 107, 6769-6775. to decomposition of the silicate structure, with the Mg Khorami, J., Choquette,D., Kimmerle, F.M., and Gallagher,P.K. (1984) coordination remaining largely intact. By contrast, under Interpretation of EGA and DTG analysesof chrysotile asbestos.Ther- mochimica Acta, 76, 87-96 hydrothermal conditions, the Si-O structure remains Kirkpatrick, R.J., Howell, D., Phillips, B.L., Cong, X-D., Ito, 8., and largely intact, with the major changesoccurring in the Mg Navrotsky, A. (1991) MAS NMR spectroscopicstudy of Mg"SiO, with layer(Ball and Taylor, 1963).Although conditionsin the the perovskite structure.American Mineralogist, 76, 673-67 6. innermost layers of the fibril bundles would not neces- MacKenzie,K.J.D., and Meinhold, R.H. (1993)A 25-Mg MAS NMR sarily correspond to the most severehydrothermal con- study ofthe thermal decomposition ofmagnesium carbonate.Journal of Materials ScienceI-etters, 12, 1696-1698 ditions that can be imposed experimentally and that lead -(1994a) A glass-bondedceramic material from chrysotile (white to the formation of the 2: I layer silicate talc (Ball and asbestos).Joumal of Materials Science,in press. Taylor, 1963), nevertheless,the formation of a dehy- - (1994b) r5Mgnuclear magneticresonance spectroscopy ofminerals droxylate retaining phyllosilicate characteristicsmay re- and related inorganics:A survey study. American Mineralogist, 79, in press. quasi-hydrothermal flect the conditions in the deepest -(1994c) Thermal decompositionof dolomite, calcium magnesium layers of the fibril structure. carbonate, studied by 25-Mg solid-state nuclear magnetic resonance. Thermochimica Acta, in press -(1994d) AcxNowr,nncMENTS Thermal decomposition of brucite, Mg(OH)r: A 25-Mg MAS NMR study. Thermochimica Acta, in press. We are indebted to W. waters, N.Z. GeologicalSurvey, and T. Carew, -(1994e) The thermal reactions of synthetic hectorite studied by CassiarMining Corporation, for the chrysotile samples. 29-Si, 25-Mg and 7-Li magic angle spinning nuclear magnetic resonance.Thermochimica Acta, in press. RrrnnnNcns crrED MacKenzie,K.J D., Brown, I.W.M., Meinhold,R.H, and Bowden,M.E. (l 985) Outstandingproblems in the kaolinite-mullite reaction sequence Ball, M.C., and Taylor, H.F W. (1963) The dehydration of chrysotile in investigatedby 29-Si and 27-Al solid statenuclear magnetic resonance. air and under hydrothermal conditions. Mineralogical Magazine, 33, I. Metakaolinite. Journal of the American Ceramic Society, 6E, 293- 467-482. 29't. Brindley, G.W., and Hayami, R. (1965) Mechanism of formation of for- Magr, M., Lippmaa, E., Samosan,A., Engelhardt, G., and Grimmer, steriteand enstatitefrom serpentine.Mineralogical Magazine,35, I 89- A -R. (1984) Solid-state high-resolution silicon-29 chemical shifts in 195. silicates.Journal ofPhysical Chemistry,88, l5l8-1522. Brindley, G.W., and Zussman,J. (1957) A structural study of the thermal Martin, C.J. (1977) The thermal decomposition of chrysotile. Mineral- transformation of serpentineminerals to forsterite. American Miner- ogical Magazine, 41, 453-459 alogist,42,46l-474. Santos,H.D., and Yada, K. (1979) Thermal transformation of chrysotile Brown, I.W.M., MacKenzie,K.J D., Bowden,M.E., and Meinhold,R.H. studied by high resolution electron microscopy. Clays and Clay Min- (l 985) Outstandingproblems in the kaolinite-mullite reaction sequence erals.27 . 16l-17 4. investigatedby 29-Si and 27-Al solid-statenuclear magnetic resonance. Sherriff, B.L., Grundy, H.D., and Hartman, J.S. (1991) The relationship II. High-temperaturetransformations of metakaolinite. Journal of the between29-Si MAS NMR chemicalshift and silicatemineral structure. American Ceramic Society,68, 298-301. EuropeanJournal of Mineralogy, 3, 75 l-7 68. Datta, A.K., Samantaray,8.K., and Bhattacherjee,S. (1986) Thermal transformation of chrysotile asbestos Bulletin of Materials Science,8, Mnnuscn:r'r REcErwDApnn- 26, 1993 497-503. Mem.rscrur'r ACcEFTEDAucusr 26, 1993