Travel & Places

A multinuclear magnetic resonance study of the structure of hydrous albite glasses

A multinuclear magnetic resonance study of the structure of hydrous albite glasses
of 11
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  Geochimica et Carmorhimica Ada Vol. 3. pp. 2925-2935 Copyright 8 1989 Fkrgamon Ras plc.F’rintcd n U.S.A. 0016-7037/89/$3.00 .oO A multinuclear magnetic resonance study of the structure of hydrous albite glasses S. C. KOHN, ' R. DUPREE,* and M. E. SMITH*‘Department of Geology, University of Manchester, Manchester, M I3 9PL. U.K.*Department of Physics, University of Warwick, Coventry. CV4 7AL, U.K. (Received March 28, 1989; accepted in revised orm September 6, 1989) Abstract-The structures of a series of hydrous albite glasses quenched from melts at high pressures andtemperatures have been studied using 2gSi, 23Na, 27A1,and ‘H nuclear magnetic resonance. Changes inthe isotropic chemical shift, the chemical shift dispersion, and the mean nuclear quadrupole couplingconstant for 23Na as a function of dissolved water concentration were deduced from spectra obtained attwo different magnetic fields. Major changes in the sodium environment occur, but the spectra for *‘Siand 27Al, and hence their structural environments, are similar throughout the range of water concentrationsstudied (O-67 mol%). No previous model is consistent with the results of this study. The data suggestthe existence of the following structural features: i) exchange of H + for Na+ as a charge-balancing cation;ii) formation of Na( OH) complexes; iii) incorporation of molecular water; iv) no octahedrally coordinatedaluminium; v) no Al-OH or Si-OH. These features can be summarised in terms of the equilibriumNaAlS&Og + Hz0 = HAlSi308 + Na( OH).In contrast to all previous interpretations, we see no evidence for depolymerisation of the aluminosilicateframework, although an increase in the symmetry of the aluminium environments and decrease in thechemical shift dispersion of the sodium environments suggests a more ‘ordered’ structure than in the dryglass. If the structures of hydrous albite melts are the same as those of the glasses studied here the currentunderstanding of the effect of dissolved water on the physical properties of felsic melts must be reassessed.INTRODUCTION IN STUDIES OF VOLATILEdissolution in melts albite (Na-AISiJOs) has been the most frequently used composition. Itis a major component of felsic rocks so it can reasonably betreated as a simple model for such systems, and it has evenbeen used as a model for all natural magmas (BURNHAM,1974, 1975). Furthermore, it melts at experimentally acces-sible temperatures ( 1118°C under anhydrous conditions atone bar pressure), and its melts can be quenched to a clear,bubble-free glass even with several weight percent of waterin solution. Despite having been the subject of many studies,the water solubility mechanisms are still poorly understood.High-resolution solid state nuclear magnetic resonance(NMR) has already made valuable contributions to the studyof local order in anhydrous glasses (KIRKPATRICKet al., 1986;DUPREE and HOLLAND, 1988). In this paper we have usedNMR to study glasses quenched from hydrous melts to pro-vide a new insight into the long-standing problem of waterdissolution mechanisms.X-ray diffraction (TAYLOR and BROWN, 1979) has beenused to show that anhydrous albite glass has a frameworkstructure based on linked [ SiOd] and [ AlO tetrahedra. Ac-cording to a model based on these results the tetrahedra arejoined to form six membered rings such as those found intridymite, with Na+ cations in interstices between the rings.This “stuffed tridymite” model contrasts with the structureof crystalline albite which contains four membered rings oftetrahedra. A wealth of evidence exists to show that all thealuminium is tetrahedrally coordinated as part of the glassnetwork (e.g., SEIFERT et al., 1982; MCI&OWN et al., 1985;OESTRIKE et al., 1987).Early studies of volatile dissolution in silicate melts werelimited to studies of solubility ( HAMILTONet al., 1964) andthe effects on phase relations (e.g., TUTTLE and BOWEN,1958). BURNHAM 1974, 1975, 1979) formulated a generalmodel for the mechanism of dissolution of water based largelyon solubility data for albite. In Bumham’s model water dis-solves in albite melt according to the following reactions.i) Below 50 mol%NaAISi,OB + H20 + HAISi,O,( OH)(ONa); ( 1)i.e., tetrahedrally coordinated Al is now charge balanced byH+ instead of Na+. One Si-0-Si bond has been rupturedforming one Si-OH and one Si-O- locally charge balancedby Na+. Using NMR terminology one QJ-OH silicon andone Q3-O- silicon have been generated.ii) Above 50 mol%HA1Si307( OH)( ONa) + Hz0 -HAlSi,O,(OH),(ONa), etc.; (2)i.e., all further dissolved water breaks Si-0-Si bonds to formadditional Qt4_x)-( OH), units. The relationship betweenmolar water solubility in albite melt and water fugacity led OXTOBY and HAMILTON ( 1978) to suggest that water dis-solves by one mechanism below 50 mol% and by a differentmechanism above. They further suggested that one mole ofwater is associated with one mole of NazO or six moles ofSiOz. On the basis of the effect of water on phase relationsin the granite system, it has been suggested (MANNING et al.,1980: PICHAVANT,1987) that dissolution of water in feldsparmelts results in a change of aluminium coordination from4-fold in the dry melt to 6-fold in the hydrous melt. Changes2925  2926S. C. Kohn, R. Dupree, and M. E. Smithin the physical properties of melts can also provide evidencefor changes in structure accompanying water dissolution. Foralbite, as for other melts with framework structures, viscosityis dramatically reduced by the introduction of water (DING- WELL and MYSEN, 1985; DINGWELL, 1987). For melts ofrelated compositions, it has been shown that water dissolutionincreases electrical conductivity ( LEBEDEV nd KHITAROV, 1964) and cation diffusion rates ( JAMBONet al., 1978). Allthese observations have been interpreted in terms of reducedpolymerisation in the melts.In the last ten years spectroscopic measurements on hy-drous glasses have been widely used to deduce structural fea-tures of hydrous melts, the assumption being made that thestructures of hydrous glasses are similar to the melts fromwhich they are quenched. The available evidence ( AINES etal., 1983; STOLPERet al., 1983; MYSEN and VIRGO, 1986a)suggests that this is a reasonable assumption and one whichwill also be made here.MYSEN et al. ( 1980) used Raman spectroscopy to studyhydrous albite glasses. They proposed a complex reactionmechanism involving the following species: i) S&OH units(Q:,-OH) attached to the framework and produced by therupturing of Si-0-Si linkages; ii) chain units without attached-OH but containing Si and Al; these would therefore containQz -( 0-h silicons; iii) residual framework units unaffectedby water dissolution, i.e., Q4 silicons; iv) OH- anions locallycharge balanced by Na+. They also stated that no molecularwater was present and that A13+ ions are expelled from tet-rahedral coordination, changing their structural role from anetwork former to a network modifier. However, these resultswere controversial. FREUND ( 1982) disputed some of theirband assignments and suggested that their results did notrule out the possible existence of Hz or Al-OH units.MCMILLAN et al. ( 1983) repeated the Raman measurementson another hydrous albite glass and observed a weak peakcorresponding to molecular water. STOLPER ( 1982a,b)changed many of the accepted ideas on water dissolutionmechanisms in this system by his careful infrared measure-ments and interpretations. His results strongly suggested theexistence of both B-OH and molecular water. Most recentdiscussions of water solubility mechanisms have concentratedon the relative concentrations of the two species. MYSEN andVIRGO ( 1986b) published a further study of hydrous alu-minosilicates, including albite. In contrast to their earlier workthey observed the presence of molecular water and furthersuggested that for hydrous albite glass Si-OH and Al [ 41 -OHbonds are absent and that hydroxyl is associated with networkmodifying Al 3+or Na+. They again observed vibrations whichthey interpreted as due to the presence of anhydrous depo-lymerised silicate units containing Qz( -0-h silicons.Aspects of the structure of hydrous glasses have also beeninferred from calorimetric studies. CLEMENSand NAVROT-SKY ( 1987) measured heats of mixing in the system Na-AlSi30s-Hz0 and observed a minimum at a mole fractionof Hz0 of around 0.25. This was interpreted ( NAVROTSKY,1987 ) in terms of a solubility mechanism involving both mo-lecular water and hydroxyl species. It was suggested that theminimum arose from competition between an endothermicheat of dissolution of molecular water and an exothermicheat of dissociation. OKUNO et al. ( 1987) have performedX-ray radial distribution analysis of hydrous albite glass. Theyfound that the hydrous glass has a similar structure to thedry glass in that it still consists of TO4 tetrahedra and thatall the aluminium is in four-fold coordination. Their radialdistribution functions show that hydrous albite glass has amore “ordered” structure than the dry glass and that thechange between dry and hydrous glasses is greater for albitethan for the other compositions studied (Abj3w7 andAn,,Qzs5). They proposed that the change is due to modi-fications of the linkages between tetrahedra and suggestedthat clustering takes place.No previous multinuclear NMR study of glasses quenchedfrom melts in the system NaAlSisOs-Hz0 exists. YANG andKUUCPATRICK1989) have obtained NMR spectra of hydrousglasses in their study of the dissolution of albite glass inaqueous solutions. However, their results are not applicableto hydrous albite melts since their reaction products arequenched from well below the liquidus and have composi-tions which are considerably different from their starting ma-terial in that they contain potassium and are generally defi-cient in total alkalis. EXPERIMENTAL METHODS The dry albite glass is the same as that used by OXTOBY nd HAM- ILTON 1978) and was prepared by melting at - 15OO’C n a largevolume furnace. This glass (in the form of a crushed powder) wasused as a starting material for all the hydrous glasses except samplesAWB and AWA. These were synthesized from a gel of albite com-position and also contain 0.14 wt% MnO which was added to reduce29Si elaxation times.The synthesis conditions and water concentrations of all samplesare listed in Table 1. The water concentrations quoted for the water-saturated samples (AB7, AB4B, AB4A, ABSD, and ABSG) are basedon previous measurements of water solubility ( OXTOBYand HAM- ILTON, 1978; BLAMARTt al., 1986,unpubl. ahstr.) and were checkedby comparing the amount of water added to the capsule with theamount remaining in a fluid phase after the run; for the water-un-dersaturated samples (AB6, ABSB, ABSF), the concentration of waterquoted is simply derived from the amount of water added to thecapsule. Measurements of weight losses at 1000°C were performedon fragments of each of the glasses as a further check, the waterconcentrations are believed to be correct to within 10% of the valuesquoted. All NMR experiments were performed on clear, bubble-freeglasses with the exception of ABSG which was a slightly translucentglass. When glasses ABSD and ABSF were recovered from their plat- Table 1.Water concentrations and synthesis condftions of allsamples studiedSamDle ABS AMA67ABOBAB4AA656AB5DABSGAB5F mater concentrations Apparatus(wt X) (mol I)P/kb Run T/Y Durationhrs0 0 LVF 0.001 -15000.85 IO IHPV0.51270 42.6 29 IHPV0.51270 44.3 40 IHPV1.01300 76.3 50 IHPV2.01300 68.2 57 PC10.0 1100 3 9.5 60 IHPV5.0 900 3 9.5 60 IHPV 5.0900 312.2 67 PC 15.0 1100 3 LVF = Large-volunh? ne-atnosphere furnaceIHPV = Internally heated pressure vessel with argon as pressuremediumPC = Solid media piston cylinder apparatus  NMR study of hydrous albite glass2921inum capsules, they were found to be mainly clear and bubble-freebut with part of each sample being translucent. These translucentparts were removed and NMR experiments were only performed onthe clear parts of these samples. Sample ABSB was examined usingan SEM and found to be free of bubbles larger than - 100 nm (theinstrumental resolution). The slightly translucent glass, ABSG, con-tained a few bubbles ranging in size from 300 to 10000 nm. Noneof the samples studied possess water~on~ning fluid inclusions asthe static ‘H NMR spectra (see later) do not contain a narrow com-ponent. Infrared spectra between 500 and 5000 cm-’ were recordedfor ABSB and are very similar to those reported by STOLPER ( 982a)for an albite glass containing 6.85 wt% HzO.NMR experiments were performed using a Bruker MSL-360 spec-trometer (BO = 8.45 T) operating at 71.535 MHz (z9Si), 93.83 MHz(“Al). 95.22 MHz (“Na), and 360.13 MHz (‘H). Some ad~tional23Na and “Al spectra were collected using a Bruker CXP-200 {I&= 4.7 T) spectrometer operating at 52.90 MHz and 52. I 1 MHz, re-spectively. All one-pulse spectra were acquired using short (<r/6)rf-pulses with sufficient recycle delays to prevent saturation. Spectrawere externally referenced to the accepted standards for these nuclei( TMS for ‘H and %i, 1 M aqueous NaCl for 23Na,and 1 M aqueousAlC13 for *‘Al). Magic-angle spinning between 3.5 and 9 kHz wasused with the highest spinning speeds generally producing the clearestspectra. The -Si spectra were smoothed with 50- 100 Hz exponentialbroadening to improve the signal-to-noise ratio. Some 23Na nd *‘Alspectra were accumulated using high-power proton decoupling. ‘H-‘?Sicross-polansation (CP) experiments were carried out with contacttimes between 0.5 and 20 ms, and cross-polarisation experiments toz3Na and *‘Al were also attempted. Some ‘H spectra were obtainedunder static conditions using a probe with a negligible proton back-ground signal and a dead-time of 1 ~1s. his enabled ‘H spectra tolx obtained by single-pulsemethods rather than a solid-echo echnique(e.g., ECKERT et al., 1988). Delays of I-60 s were employed.RESULTS29SiThe single pulse mSi spectrum for dry albite glass is shownin Fig. 1a. The width of the 29Si resonance in dry albite glassis greater than that in SiOz glass ( PETTIFER et al., 1988) dueto the presence of a range of Q4( n Al) next-nearest neighbourenvironments. We observe an approximately Gaussian peakwith a maximum at -97.8 ppm and a width (FWHM) of16.2 ppm. This position is similar to previous measurementsof -98.7 ppm (OESTRIKE et al., 1987) and -97.9 ppm (MURDOCH et al., 1985 ). Single pulse 29Si spectra of the hy-drous albite glasses (Fig. 1b) are very similar to that of dryalbite glass. In some spectra there appeared to be a slightincrease in intensity at the deshielded side of the peak in thehydrous glasses, but this probably does not exceed the ex-perimental error.Use of cross-polarisation MASNMR (PINES et al., 1973)was less informative than for hydrous silica glass ( FARNAN et al., 1987), though small changes in the appearance of thespectra as a function of contact time are apparent. Spectraobtained using long contact times (5 and 20 ms) are almostindistinguishable from single pulse spectra. At shorter contacttimes (1 ms and 500 ps) the peak maximum moves to- -92.5 ppm with the full width at half maximum (FWHM)remaining approximately constant at 16.5 ppm. The CPspectra for AB6 ( 10 mol% HzO) and ABSB (57 mol% HzO),shown in Fig. Ic and d for a contact time of 500 gs, areidentical (the feature at -40 ppm in the spectrum for AB6is an artefact), but acquisition of the spectrum for the highwater concentration sample was much more rapid ( 1467 0 -2M 1000 -1m-EmFe.3 PRI FIG. I. “Si MASNMR spectra of afbite glasses. (a) and (b) aresingle-pulse spectra obtained using a pulselength of 2 ps (=x/6) anda recycle delay of 75 s, (a) AB5, dry and (b) ABSD, hydrous. (c)and (d) are cross-polarisation spectra obtained with contact times of 500 ps, 50 ms of proton decoupling following the pulse and a recycledeiay of 5 s, (c) AB6, 10 molW H20 and (d) ABSB, 57 mol% HzO.Spinning speed for all spectra was -3.5 kHz, and the number ofscans was- 1000 (a and b).-12,~(c),and-15~(d). pulses compared with 12057 pulses for the spectra shown inFig. lc and d.).We have also performed single-pulse experiments understatic (non-spinning) conditions for AB5 (dry) and ABSB( 57 mol% H20). This technique is sensitive to the differencesin chemical shift anisotropy between Qs and Q4 silicons whichgives different peak widths and shapes for the two species( STEBBINS,1987 1. However, the resonances for the dry andhydrous albite giasses (whose peaks are at the same isotropicchemical shift as the MAS re~nances) have the same shapesand widths (47 t 5 ppm ). again suggesting that the concen-tration of QI silicons is low. Finally, we have performed astatic cross-polarisation experiment on the hydrous glass,ABSB. If any QJ-OH units were present, their larger chemicalshift anisotropies would result in the observation of a broaderresonance. in fact, the resonance (which is at -90 k 3 ppm)is narrower ( 33 + 5 ppm) than the single pulse spectrum andshows no signs of a shape characteristic of chemical shiftanisotropy, providing further evidence for the absence of Si-OH units.The quadrupolar nature of the sodium nucleus (nuclearspin, I = %) results in the linewidth and peak position beingdependent on a number of factors. Measurements carriedout at two magnetic field strengths (here 8.45 and 4.7 T)allow deconvolution of these effects.The 23Na spectra of three representative samples at 8.45T are shown in Fig. 2a, b, and c and at 4.7 T in Fig. 2d, e,and f. At a magnetic field strength of 8.45 T the peak max-imum for dry albite glass is at -2 1.5 ppm (Fig. 2a) similar  2928 S. C. Kohn, ,....l,...I,,..I....ILI 200No 0-100 -zoo sy) zoo 0 -mm -4oQRy Pm FIG. 2. 23NaMASNMR spectra of dry and hydrous albite glasses. (a, b, c) obtained at a magnetic field strength of8.45 T and with MAS at 7-8 kHz. (d, e, f) obtained at a magnetic field strength of 4.7 T with MAS at -4 kHz. (a)and (d) AB5, dry. (b) and (e) AB7,29 mol% HzO. (c) and ff) ABSB, 57 mol% H20. A pulselength of 1~ (<n/6 on( % ++ -% ) transition in a site where cq is large) and recycle delay of 1 were used with the number of scans rangingbetween 500 and that of -20.7 ppm reported by GEISINGER et al. ( 1988)at the same field and -20.3 ppm reported by OESTRIKE etal. ( 1987) at 11.7 1. At both fields the peak maximum ( bpak)is found to he a function of dissolved water concentration(Fig. 3); however, the difference between the trends at thetwo fields is considerable. At 8.45 T, 13,~ becomes lessshielded with increasing water concentration throughout therange of con~n~tio~ studied, whereas at 4.7 T, 6, firstbecomes less shielded then mom shielded with increasingwater concentration. There are also differences in the trendsin peak width (FWHM). At 8.45 T, FWHM generally de-creases with increasing water concentration, whereas at 4.7 -.v -15 1 . . ... m -20 ” ‘. l q El0‘peak _26 - n(pwd.LI-30 --35 -D-40’ .,.I J0 102030 40 50 60 70 Mole % dissolved waterFIG. 3. Position of the “Na peak maximum (8,) as a functionof dissolved water concentration at magnetic field strengths of 8.45T (filled symbols) and 4.7 T (open symbols). T it remains approximately constant for the samptes withthe lower water con~nt~tions but increases for the highestwater concentrations (Table 2). ‘H-‘jNa CP experiments us-ing NaB& and NazB407 - lOH20 (borax) to set the matchconditions (HARRIS and NESBIT~, 1988) produced no cross-polarisation for the hydrous glasses. Experiments involvinghigh-power proton decoupling have also been performed. Nochanges between sit&e pulse and decoupled spectra were ob-served either for static or MAS conditions, indicating thateffective dipolar coupling to the sodium is weak.The 23Na resonance in dry sodium disilicate glass was alsoexamined at both fields in order to provide a comparisonwith the albite glasses. The peak position changes from -2ppm at 8.45 T to -5 ppm at 4.7 T, with the width changingfrom 3700 Hz to 1900 Hz. Z7AI The peak maximum of the “Al resonance in dry aibitegiass is 50.5 ppm at a field of 8.45 T, This compares with avatue of 54.8 ppm at 11.7 T (OESTR~WS t al., 1987). Withincreasing water concentration the *‘Al peak maximum shiftsslightly, reaching 53 ppm for ABSG. A much more markedchange is that the peak becomes narrower and more sym-metric (Fig. 4) with the FWHM changing from 23 ppm forAB5 to 20.5 ppm for AB6 and 14 ppm for ABSG. No peaknear 0 ppm corresponding to octahedrally coordinated ahr-minium is present. The normahsed aluminium signal inten-sity from all the albite glasses remains constant and, comparedto our a-Alz03 standard, 15-208 more signal was observed.  NMR study of hydrous albite glass2929 sample ExperimentalSimulated SimulatedSimulatedualdcal shiftLinewidthsLinewidthscontributionscontributionsdispersionU'W4)/Hr(FWHll)/Hrat 4.7TiHZar 8.45Tl"ZJdi/ppm4.71 8.45T4.1I 8.45T"9 "b"9 %A85 2380 2730 2390 2730 ,260 1127 700 2030 21.3A86 2495 2930 2‘80 2920 1224 L250 680 2156 23.7ABI 2410 2450 2400 2450 1500 900 830 ,620 17.0ABhB 2381 2360 2360 2360 1530 820 850 I~180 15.5An58 2625 ,790 2645 1790 2160 220 1200 400 4.2ABSG 2770 1590 2770 1570 ,920 0 1070 0 0 Hence all the 27Al nuclei are observed in these glasses (seeDUPREE et al., 1988, for a discussion of the quantificationtechniques). High-power proton decoupling experimentswere again performed with no detectable changes in the res-onance, and *‘Al CP experiments, using the alum, NH* Al( S04h - 12H20, to set up the match conditions gaveno signal. “Al MASNMR spectra for sample AB6 were runat both 8.45 T and 4.7 T. The peak position shifts from 5 1.5ppm to 48.8 ppm while the FWHM increases slightly from2070 Hz to 2180 Hz at the lower field. 150 100 50 PPM J 0 -50 FIG. 4. 27A1MASNMR spectra of albite glasses. (a) ABS, dry (b)AB6. 10 mol% Hz0 (c) ABSB, 57 mol% HrO (d) ABSG, 60 mol%HrO. 200 1 ps (<k/6 on (r/r ++ -r/r) transition in a site where c, islarge) pulses, a recycle delay of 1 s and MAS at -7 kHz were used. ‘H The single-pulse static spectra consist of two components:a broad one due to Hz0 and a narrower one due to OH. Theexistence of two components is more difficult to deduce fromthe MAS spectra, shown in Fig. 5 for sample ABSB. All MASspectra consist of one broad resonance centered on 3.5-3.8ppm with several pairs of spinning sidebands. Resolvable res-onances such as those observed in hydrous silica glasses ( KOHN et al., 1989) are not observed, presumably becausethe OH and Hz0 have similar shifts and an increased residualwidth. As the total dissolved water concentration of the glassis increased, changes in the MAS spectra broadly in agreementwith ECKERT et al. ( 1988) are observed, although the changeswere easier to detect in the static spectra.There are several problems associated with quantifying thewater speciation from ‘H NMR spectra in these glasses.Firstly, it was noted that the relaxation time for protons inthe Si-OH group were very dependent on water concentration.At high total dissolved water concentrations no differencewas observed in the spectrum as a result of varying the re-laxation delay between Is and 60 s. However, for AB6 ( 10mol% HzO) at least 60 s recycle delay was required to allowcomplete relaxation. Secondly, the residual width of the MASlines increases with increasing water concentration (due toincreased homogeneous coupling); quantification based onpeak heights using MAS may therefore be in error. Lastly,there were problems in completely eliminating backgroundsignals due to the probe and water adsorbed onto the surface ‘HFIG. 5. ‘H NMR spectra for ABSB (a) Static (b) MAS. Around1000 scans were obtained using a 2 ps pulse ( T/ lo), a 2 s recycledelay. and (for (b)) MAS at -4 kHz.
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks