INFORMATlON TO USERS

This manuscript has beeir reproduced from the microfilm master. UMI films

the text dimdy from the original or wpy submitbd. Thus, some thesis and

dissertation copies are in m e r face, Mi le others may be from any type &

cornputer printer.

The qurlity of thb irproduction k dipendent upon tha qurlity of tim copy submitted. Broken ar indistinct print, ooCored or poor quaMy illustr'afions

and photographs, pn'nt Meedthrough, suôstandard margins, and improper

alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manusaipt

and there are missing pages, aiese will be nated. Also, if unauthorized

copyright material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., m e s , drawings, charb) are repmduced by

sectiming the original, beginning at the upper letthand corner and continuhg

from left to rigM in equal sons uith small werlaps.

Photographs included in the original manuscript have beeri reproduœd

xemgraphically in this -y. Higher quality 6. x W Mac)< and white

photographie prints are availabk for any photographs or illustrations appearirig

in this copy for an additional charpe- Contact UMI d i r e to order.

Bell & Howell Information and Leaming 300 Norlh Zeeô Road, Ann Arbor, MI 481 06-1 346 USA

-521 -0600

AN lNVESTlGATlON O F Z R AND Tl-BEARING

ALKALI ALUMINOSILICATE GLASSES: SOLUBlLlTY EXPERIMENTS, RAMAN

SPECTROSCOPY AND 2 3 ~ ~ NMR ANALYSES

by RoSert A. Man

May, 1998

A thesis submitted to the Faculty of Graduate Studies and Research in partial

fuifillment of the requirements of the degree of Ooctor of Philosophy

Department of Earth and Planetary Sciences

McGill University

Monléal, Québec

Canada

O Robert A. Man - MCMXCVlll

National Library 1+1 ofcanada Bibliothèque nationale du Canada

Acquisitions and Acquisitions et Bibliographie Services services bibliographiques

395 Wellington Street 395. rue Wellington Ottawa ON KlA ON4 OttawaON KlAûN4 Canada Canada

The author has granted a non- L'auteur a accordé une licence non exclusive licence aiiowing the exclusive permettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, loan, distribute or seil reproduire, prêter, distribuer ou copies of this thesis in microform, vendre des copies de cette thèse sous paper or electronic formats. la forme de microfiche/film, de

reproduction sur papier ou sur format électronique.

The author retains ownership of the L'auteur conserve la propriété du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substantial extracts fiom it Ni la thèse ni des extraits substantiels may be prùited or othewise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation,

The solubility of Zr-bearing minerals in peralkaiine, H,O-saturated alkali

aluminosiiicate melts with or without added F reaches a maximum of about 4 and 3.5 W.%

23-0, respectively at approximately 57 wt% Si02 The saturating phase for melu with SiO2

content above this threshold is zircon. In the halogen-free experiments, the saturating phase

for melts with lower silica content is wadeite (K2ZrS &O9), while ZQ crystallizes from melts

with I wt.% added F. Experiments with Cl-bearing melts indicate no maximum solubility

of Zr minerals; the solubility remains a: 2 to 2.2 wt.% Zr02 across a wide compositional

range. The saturating phases in the Cl-bearing melts are the same as those of the F-bearing

melts and the transition from ZrO, to zircon occurs at the same composition. The positive

dope of the wadeite saturation curve on X,-X,, plots for the halogen-- melts suggests

that increasing SiOl activity increases the solubility of wadeite. This behaviour of wadeite

would not be predicted based on a simple dissolution mechanism. Therefore, a more complex

equilibriurn involving melt structure is proposed. Observations that zircon is scarce, whereas

el pidite (N- &O,, -3H2 O) and other alkali and alkalineearth Wconosilicates are abundant

in the Strange Lake peralkaline intrusive complex in northem Quebec/Labrador indicate that

the parent magma did not saturate with a Zr-bearing mineral until it cooled to a low

temperature, probably less than 600 OC.

Peralkaline Ti,Zr-bearing sodium duminosilicate glasses have been analyzed by

Raman spectroscopy to determine the effect of CI on glas stxucture. The spectra of the Ti-

bearing glasses show a significant difference between the CI-free and the Cl-bearing

composition. The Cl-free glass spectrum contains a strong, asymmetrïc peak at 900 cm"

which is associated with Ti in five-fold coordination. This peak is shified to higher

frequency and becomes more symmetric with the addition of 0.3 wt% CI. Deconvolution

of the high-fiequency waveband suggests that differences between spectra result from the

contribution of a peak at 945 cm-'. This peak is believed to be the result of Ti-O vibrations

in fûliy-polymerized titanate tetrahedra. It is proposed that the addition of Cl destabiiizes

[ T i in favour of tevahedral coordination as a result of cornpetition between Cl and titanate

groups for aikalis. The spectra of Zr-bearing glas show a strong peak at 980 cm-' which is

not seen in spectra of the base glass, indicating a vibrational mode associated with a

zirconosilicate stmcturai environment. There is no discemible difEerence between the

Raman spectra of Zr-bearing glasses with and without added Cl, indicating that the local

structure around Zr is insensitive to the presence of Cl.

=Na NMR MAS analyses of a suite of Na-duminosilicate glasses with NdAi = 2 and

v w n g SiO, content has revealed a trend toward more negative chemical shifi ( p a t e r

shielding of the nucleus) as the glas structure becomes more polymerized, Le. the average

number of non-bndging oxygen atoms per tetnhedron (NBOm decreases. This trend is

obsemed only for glasses with NBO/T s 0.3. For more polymerized glasses no change in

chernical shift is measured. Past studies have indicated a strong correlation between

increasing Na-O bond length and increasingly more negative .=Na chemical SM. This

suggests that for Na-aiurninosilicate glasses with NBO/T s 0.3, Na-O bond lengths would

be expec ted to increase as polymerization increases.

La solubilité des minéraux de Wcmium dans les liquides silicates peralcalins, saturés

en H20 et formés d'aiuminosilicates alcalins, aux quels on a ajouté ou non du fluor, atteint

respectivement un maximum à environ 4 et 3.5% poids de ZrO, à environ 57% SiO,. La

phase qui sature les liquides ayant un contenu en SiO, au-dessus de ce seuil est le zircon.

Dans les expériences sans halogène, la phase qui sature les liquides avec un un contenu plus

faible en SiO, est la wadéite (I(2ZrSi3O,), tandis que le ZrO, cristallise à partir de liquides

avec I % de fluor. Les expériences avec des liquides contenant du chiore n'indiquent pas de

maximum de solubilité pour les minéraux de zirconium; la solubilité reste entre 2 et 2.2%

ZrO, à l'intérieur d' une vaste gamme de compositions. Les phases qui saturent les liquides

chlorés sont les mêmes que pour les liquides fluorés, et la transition de ZrO, au zircon se

produit à la même composition. D'après la pente positive de la courbe de saturation de la

wadéite sur un diagramme Xm2-XSa2 pour les liquides sans halogène, une augmentation de

l'activité de la silice augmente la solubilité de la wadéite. Ce comportement de la wadéite

est inattendu si on s'entient à un simple mécanisme de dissolution. Par conséquent, un

équilibre plus complexe impliquant la structure du liquide semble indiqu6. Dans le complexe

intrusif perdcalin de Strange Lake dans le nord du Québec-Labrador, le zircon est peu

abondant, tandis que I'elpidite (Na&3,0,,-3H20) et d'autres zirconosilicates alcalins et

dcaiino-terreux le sont davantage, indiquant que le magma parent n'était pas saturé avec des

minéraux de zirconium jusqu'à ce qu'il évolue vers une basse température, probablement

moins de 600 OC.

Des verres d'aiuminosilicates de sodium avec Na/Ai > I et dopés avec Ti ou Zr ont

été analysés par spectroscopie de Raman afin de déterminer l'effet du chlore sur la structure

du verre. Les spectres du verre contenant du titane montrent une différence importante entre

les compositions possédant ou non du chlore. Le spectre du verre sans chlore contient un

important pic asymétrique à 9 0 cm-' qui est associé avec le [ q i . Ce pic est déplacé à une

plus haute fréquence et devient plus symétrique avec l'addition de 0.3 % de chiore. D'après

la déconvolution des bandes de hautes fréquences, les différences entre les spectres résultent

de la contribution d'un pic à 945 cm-'. Ce pic résulterait de vibrations Ti-O dans les

tétraèdres de titane dans la trame. L'addition de chlore déstabiliserait le IqTi en faveur de la

coordinance tétraédrique à cause de la compétition entre le chlore et les groupes de titane

pour les alcalins. Le spectre du verre contenant du zirconium montre un pic important à 980

cm-' absent dans les spectres du verre de base, indiquant un mode de vibration associé à la

structure du zirconosilicate. Il n'y a pas de différence discemable entre les spectres de Raman

des verres contenant du Zr avec ou sans chlore. indiquant que la structure locale autour du

zirconium est insensible à la présence du chiore.

Des analyses de "Na NMR MAS d'une série de verres d'aluminosilicates de sodium

avec NdAi = 2 et un contenu variable de SiOz ont révélé une tendance vers un dédoublement

chimique plus fortement négatif (plus grand camouflage du noyau) lorsque la structure du

verre devient plus fortement polymérisée, Le. le nombre moyen d'atomes d'oxygène non liés

à deux tétraèdres divisé par le nombre de tétraèdres (NBO/T) diminue. Cette tendance est

observée seulement pour les verres avec NBO/T 5 0.3. Pour les verres plus fortement

polymérisés, aucun changement dans le dédoublement chimique n'est décelé. Des études

précédentes ont montré une forte corrélation entre la longueur de la liaison Na-O et un

dédoublement chimique du %a de plus en plus fortement négatif. Pour les verres

d'aluminosilicates de sodium avec NBOm 5 0.3, les longueurs de la liaison Na-O devraient

donc augmenter avec une augmentation du degré de polymérisation.

Acknowledgments

1 wish to acknowledge the many individuals who contributed much to this thesis.

First of ali, 1 would like to thank my supervisors, Don Baker and AE. "Willy" Williams-

Jones. They provided guidance and fmanciai support, which were absolutely essentid to the

success of this undertaking. 1 owe them much. Numerous people provided technical expertise

and their patience is greatly appreciated. Glenn Poirier was always very helpful as I stniggled

to acquire good electron microprobe analyses of my diffcult samples. Andrew Vreugdenhil

spent many hours with me collecting Raman spectra and 1 am very grateful for his efforrs.

Fred Morin provided the NMR MAS analyses. 1 dso wish to thank Linda Reven, Grant

Henderson, Bob Martin and Michel Bouley who provided valuable insights.

1 have many fnends here at McGiU University whose support and insight were

invaluable in the completion of this endeavor. In particular, 1 would Iike to acknowledge

Pierre Hudon, Claude Daipé, Soûad Guemina, Leyla Houssain, Lawrence Yane, Sandy

Archibald, Jim Davies and Kate Ault. Thank you, my fnends.

Finally, for their love and support, 1 wish to thank my parents and my beloved Vice.

This thesis is as much a result of their efforts as mine.

....................................... ............... 3 -3 -2 Raman spectroscopie analyses .... 48 3.4 Results ............................ .- ....... ................................................... 48

........................................................................... 3 -5 Discussion ........ .... ............... .. 61 3.6 Conclusions ................................................................................................... 70

Chapter 4: "Na NMR MAS Analyses of Na-aluminosilicate Glasses with Varying SiO,

.............................................................................................................................. Content. 72

4.1 Abstract .......................................................................................................... 73 4.2 Introduction ................................................................................... .. .......... 74

.................................................................................. 4.3 Experimental Procedure -74 4.4 Results ............................................................................................................ ..75

....................................................................................... ............ 4.5 Discussion .. 80 ...................................................................................................... 4.6 Conclusions 89

Chapter 5: ALkali Zirconosilicate Phase Equilibria in the Systern: N~~o .K~o .AI~o~&~o~ .

................................................................. ....................................................... H.OG ....... -90

................................................................................................... 5.1 Introduc tion... 91 ................................................... ......................... 5.2 Experimental Procedure ....... 92

5.3 Results ......................................................................................................... 94 ...................................................................................... 5.4 Discussion .......... ... 96

...................................................................................................... 5.5 Conclusions 97

Chapter 6: ..................................................................................................................... Conclusions -98

...................................................................................................... 6.1 Conclusions 99 .................. 6.2 Contributions to Knowledge ... ................................................ 100

............................................................... 6.3 Recommendations for Future Work 101

........................................................................................................................ References 103

....................................................................................................................... Appendix 1 127

........................................................................ Appendix 2 ....................................... 129

....................................................................................................................... Appendix 3 131

vii

Preface This thesis is divided into six chapters: an introductory chapter, three manuscnpts,

a chapter reporting the results of phase equilibriurn experiments which will not be submitted

for publication and a concluding chapter. Each of the three manuscripts is CO-authored by my

thesis supervisors Drs. Don Baker and A.E. Williams-Jones, who provided advice on

research methodology and assisted in interpretation of results. Dr. Andrew Vreugdenhil of

the Department of Mining and Metallurgy at McGill is the third author of Chapter Three.

He contributed his expertise in Raman spectroscopy and perfomed the anaiyses with my

assistance. All interpretation of Raman spectra are, however, the sole responsibility of this

author. Glenn Poirier provided assistance in the collection of the electron microprobe data

reported in this thesis, although the author performed the anaiyses. Dr. Fred Morin of the

Department of Chemistry at McGill performed the nuclear magnetic resonance analyses

reported in Chapter Four.

The foiiowing is an excerpt from the "Guidelines Concerning Thesis Preparation" as

required by the Facuity of Graduate Studies and Research at McGill University:

Candidates have the option of including, as part of the thesis, the text of one or more papers submitted or to be subrnitted for publication, or the clearly- duplicated text of one or more published papers. These texts must be bound as an integral part of the thesis-

If this option is chosen, comecting texts that provide logical bridges between the different papers are rnanda!ory. The thesis must be written in such a way that it is more than a mere collection of manuscnpts; in other words, results of a senes of paper must be integrated.

The thesis must s a conform to al1 other requirements of the "Guidelines for Thesis Preparation". The thesis must include: A Table of Contents, an abstract in English and French, an introduction which clearly States the rationaie and objectives of the study, a review of the litenture, a finai conclusion and summary, and a thorough bibliography or reference list.

Additional matenai must be provided where appropriate (e.g. in appendices) and in sufficient detail to U o w a clear and precise judgement to be made of

the importance and ongiaaiity of the research reported in the thesis.

In the case of manuscripts CO-authored by the candidate and others, the candidate is required to make an explicit statement in the thesis as to who contributed to such work and to what extent. Supervisors must attest to the accuracy of such statements at the doctoral oral defense. Since the task of the examiners is made more difficult in these cases, it is in the candidate's interest to make perfectly clear the responsibilities of di authors of the co- authored papers.

Chapter 1

General Introduction

1.1 Introduction

P e r a h h e , felsic rocks represent a smdl proportion of the total volume of igneous

rock in the crust. However, such rocks have received a disproportionate share of scientific

and economic interest due to their having elevated concentrations of rare metals such as Zr,

Y, Nb and rareearth elements (REE). Adding to the curious nature of these rocks is the fact

that these rare metais are generally found in unusuai minerais such as REE fluorocarbonates

and alkali and aikaline earth Wrconosilicates- Whiie it has long been thought that these

minerais might serve as useful petrogenetic indicators, efforts to employ them in this mannet

have been hampered by the absence of thermodynamic data. Those attempts which have

been made have mostiy been limited to chemographic analysis (Williams-Jones and Wood,

1992; Marr and Wood, 1992) or synthesis experiments (Maurice, 1949; Baussy et al., 1974;

Caruba, 1975; Canrba et al., 1973; Christophe-Michel-Lévy, 196 1 ; Haschke, 1975; Haschke

and Eyring, 1971; Shunhua er al., 1986; Christensen, 1973; Chai and Mroczkowski. 1978;

Kutty et al., 1978; Tareen and KUW, 1980; Wakita and Kinoshita, 1979; Metz. 197 1). One

notable exception is the phase equilibna study of Cume and Zaleski (1985) which

constrained the pressure and temperature (P-T) conditions of the reaction:

elpidite (N@r!3i6O,,-3H20) - viasovite (Na7ZrSi,0,,) + 2 quartz + 3 H20

Given the relative paucity of reliable phase equilibria data, this project was

undertaken to elucidate the physico-chernical conditions of crystaiiization of the alkali

zirconosilicates and other Zr-bearing minerals at magmatic conditions and to describe the

structure of magmas from which these minerals are derived. It was found that two minerals;

wadeite (K2ZrSi,0,) and zircon (ZrS iO,), are predominant at these conditions. There fore,

in Chapter 2, the solubility of these mineds has been measured for several perallcaline melts.

Chapten 3 and 4 are Raman spectroscopy and = ~ a NMR MAS studies focused on the

structure of such melts.

1.2 Mineralogy of the Alkali Zirconosilicates

Aikali zirconosilicates have been reported in most perallcaIine rocks around the world

including reknowned loçalities such as Lovozero, Russia, Ascension Island, Ilimaussaq,

Greenland and Mont St. Hilaire, Canada (Table 1.1). It should be noted that many alkaiine

earth and mixed alkaii-alkaline earth zkconosilicates have also k e n discovered. In ail, there

are over two dozen zirconosilicates which have been given approval by the IMA

Commission on New Minerds and Mineral Names.

For most of the minerals in Table 1.1, there has only k e n one or two reponed

occurrences and none of these exceptionally rare minerais hâs ever been successfÛiiy

s yn thesized. The more common aikali zirconosilicates do warrant some afrention, ho wever,

as attempts to study h e m experïmentally have met with Limited success. These minerais

include: elpidite, vlasovite, catapleiite, parakeldyshite, dalyite and wadeite.

EIpidite has been reported at Lovozero, Russia (Neronova and Belov, 1963) as weU

as Mont S t. Hilaire, Quebec (Chao, 1967) and Strange Lake, QuebecLabrador (Miller,

1986). while vlasovite has k e n reported at Lovozero (Tikhonenkova and Kazakova, 1961),

Ascension Island (Fleet and Cam, 1967), Kipawa, Quebec (Gittins et al., 1973) and Strange

Lake (B irke tt et al., 1992). Several successfui syntheses of elpidite and vlasovite have k e n

done (Christophe-Michel-Lévy, 196 1; Baussy et al., 1974; Camba, 1975, Currie and Zaleski,

1985). The upper thermai stability Mt of elpidite has been established at 1.0 kbar

(595d°C) and 2.0 kbars ( 6 4 4 ~ I " C ) at which point it breaks down to vlasovite + quartz + H,O (Currie and Zaleski, 1985).

Catapleiite is another relatively common Na-zirconosilicate which has k e n reported

at several localities including Strange Lake (Birkett et al., 1992). Catapleiite has been

successfully synthesized at temperatures of 350 to 450°C and at a pressure of 0.7 kbar

(Baussy er al., 1974). although no reversed experiments have been performed with this

Mineral 1 Composition 1 Reference(s)

elpidite

vlasovite

catapleiite

gaidonnayite

hilairi te

parakeldys hi te

petarasite

te rs kite

wadeite

dalyite

umbite

khibinskite

kostylevite

Table 1.1 Naturaliy-occumng Na and K-zirconosiiicates. References are ris follows: 1. Cannillo et al- (1973); 2. Tikhonenkova and Kazakova (1962); 3. Chao et al. (1973); 4. Chao and Watkinson (1974); 5. Chao (1985); 6. Chao et al. (1974); 7. Rade and Mladeck (1977); 8. Chao et al. (1980); 9. Ghose et al. (1980); 10. Khomyakov et al. (1983a); 11. Prider (1939); 12. Henshaw (1955); 13. Van Tassel (1952); 14. Khomyakov et al. (1983b); 15. Khomyakov and Voronkov (1973); 16. Khomyakov et al. (1974); 17. ïiyushin et al. (198 1); I 8. Khomyakov et al. (1983~); 19. Boggs and Ghose (1985); 20. Ghose and Thakur (1985).

phase.

Parakeldyshite is one of two phases which had originaily b e n identified as keldyshite

(Gerasimovskii, 1962). Later workea detemllned that the original materid described by

Gerasmovskü actually contaïned two trichic phases, NaJrSi,O, (Phase II) and another with

indeterminate composition, but probably involving significant substitution of Na by H,O

(Phase 1) (Khomyakov et al. 1975; Khalilov et al. 1975). A second occurrence of Phase II,

which was named parakeldyshite (Khomyakov et al. 1975), was reported at Lagendalen,

Norway (Raade and Mladeck, 1977). Several successful syntheses have k e n accomplished

(Baussy et aï., 1974; Carub;r, 1975, Currie and Zaleski, 1985).

Wadeite and dalyite are the two most cornmon K-zirconosilicates, although the

number of reported occurrences is relatively smaü compared to that of Na-zirconosilicates.

Wadeite was first reported in the West Kimberley area of Australia (Prider, 1939) and was

fully characterized by Henshaw (1955). Wadeite has been successfully synthesized using

severai techniques (Caruba, 1975; C m b a et al. 1973). Dalyite was fmt reported at

Ascension Island (Van Tassel, 1952) and has been successfuily synthesized in one study

(Camba, 1975). No reversed experiments with K-zirconosilicates have ever k e n performed.

1.3 Raman Spectroscopy of Silicate Melts

Fundamental to any study of alkali zirconosilicate crystaliization is understanding the

nature of the silicate melt. Raman spectroscopy is a valuable technique for detemining melt

structures. At the most basic level, components of a silicate melt are thought of as network-

formers and network-modifiers. Network-fonning cations are those which contribute to the

nework of linked teuahedn. which make up a silicate melt Silicon is the most hpo-t of

these eiements, but aluminum as weil as boron and, in some cases, titanium act as network-

fomers &o. Al1 other cations act as network-modifiers which means that they alter the

structure of the meit in their vicinity. This alteration involves the transformation of bridghg

oxygen a t o m (those oxygen atoms that Link tetrahedra) to non-bridging oxygen atoms

(oxygen atoms which are part of a tetrahedron, but are bonded to one or more network-

modifien). This process, known as depolymerization, has a tremendous effect on melt

properties such as viscosity and difisivity.

A signifcant contribution to our kmwledge of the structure of silicate melts has been

provided by Raman scattering spectroscopy. When an incident beam of light, usually visibie

laser Light, is directed toward a sûmple, a s m d fiaction of that light is scattered. Most of the

scattering is elastic, known as Rayleigh scattering, and results in no change in wavelength.

However, a very small fraction of the scattered light results fiom inelastic interaction with

a vibrational mode (McMillan and Hofmeister, 1988). This is known as- aman scattering

and the scattered light in this case is shified to longer wavelengtb (Stokes scattering) or

shorter wavelength (anti-Stokes scattering).

The inelastic interaction of the incident photons with certain molecular vibrational

modes which genemtes Raman scattering arises because the electric field associated with

incident light oscillates at a frequency, v', much higher than the oscillation frequency for any

given molecular vibration, vi, which lies in the range 1012 to 10i4 Hz (McMillan and Wolf,

1995). The electronic response to this osciiiating electric field is a time-dependent, induced

dipole moment. /~,,(t), which is a function of electric field strength, E,, the oscillation

frequency and a proportionality constant, a, known as the molecular polarizability (McMillan

and Wolf, 1995):

The molecular polarizability depends on the relative positions of nucleü w ithin the molecule,

and therefore, the vibrationai frequency. For srnail vibrational displacement, qi, a may be

represented by a Taylor series (Wilson et al. 1955):

Substituting this expression into that for the induced dipole moment:

/lin, ( t ) = îXo Eo C O S 27W*t +

The new t e m which appear in the expression represent scattered light with frequencies

iower (va - vi)-and higher (v' + vi) than that of the incident beam, Stokes and anti-Stokes

Raman scattering respectively (McMillan and Wolf, 1995).

Interpretation of the Raman spectra of glasses and melts involves cornparison of the

spectra to those of crystaiiine matends with well-defined structures. The Raman spectra of

aikali and aikaline-earth silicate and durninosilicate @asses and rnelts have k e n studied

thoroughiy (Brawer and White, 1975; Fukumi et al. l99U; Furukawa et al. 198 1 ; Matson et

al. 1986; McKeown et al. 1984; McMillan, 1984a,b; McMiilan et al. 1982; McMUan et al.

1992; Mysen. 1990sb; Mysen, 1995; Mysen and Frantz, 1992; Mysen and Frana, 1994;

Mysen et al. 1985; Seifert et al. 1982). Therefore, many of the recent Raman snidies bave

concentrated on the structural effects of transition met& such as Y and Zr (Ellison and Hess,

1994) and Ti (Alberto et al. 1995; Henderson a d Fleet, 1995; Mysen and Neuville, 1995)

or halogens such as F (Mysen and Virgo, 1985a,b). The Raman study undertaken as part of

this thesis is intended to expand on these recent efforts and to include CI, a halogen known

to have a si gnificant effect on silicate melt properties (Hirayama and Camp, 1969; Baker and

Vaillancourt, 1995), but which has not, to date, been studied with this technique.

1.4 Sotid State Nuclear Magnetic Resonance

Nuclear magnetic resonance (NMR), like Raman spectroscopy, is a useful tool for the

study of silicate melts, and the two techniques are complirnentary. NMR is used to observe

changes in the local environment around a specific nucleus as a result of changes in the

electronic density function, unlike Raman spectroscopy, which detects the presence of

vibrations associated with certain bonds.

1.4.1 Basic Theory

Al1 sub-atomic particles possess 2 spin quantum number (1) of '/t. The manner in

which the spins pair up determines the net spin. Isotopes with an even number of both

protons and neutrons (i.e. "c, 160) have a net spin, 1 = 0, while those with an odd number

of both protons and neutrons (Le. '"N) have a net spin equal to some integer. It is those

nuclides with an odd number of protons and an even nurnber of neutrons or vice versa which

have a net spin 1 = !4, or some multiple thereof that are useable for nuclear magnetic

resonance (NMR). The list of NMR-active isotopes includes 'H, I3C, "O, =Na, "Al,

' 9 ~ i , 3 1 ~ and 35C1 to name a few.

Nuclear spins possess angular momentum, 1. which has discrete orientations in the

presence of an applied magnetic field. This gives rise to 21 + 1 spin States, which are

degenerate in the absence of a magnetic field. Spinning charges create a magnetic moment,

p. In the case of a spinning nucleus, the expression for this magnetic moment is:

where h is Planck's constant and y is the gyrornagnetic ratio, a quantity specific to the

nucleus and different for each isotope. The magnetic moment and the spin angtdar

momentum behave as parallel vectors, that is, the angles between the vector of the appIied

magnetic field, H, and those of the magnetic moments are fixed and quantized. However,

the moments are not static and they precess amund the extemal magnetic field at a frequency

given by the equation:

This expression, known as the Larmor equation, states that the Larmorfrequency, o,, is

directly proportional to the applied magnetic field.

The energy levels of the nuctear spin states are split by Zeeman interaction. This is

illustrated for h e simplest case, I = Yi, in Fig. 1. la. Two energy -1evels. referred to as the a

and p states appear with magnetic spins of +!A and -!4 respectively. The net magnetization

of al1 spins in the lower energy a state describes a vector parallel to that of the applied

rnagnetic field, while the net magnetization of the higher energy P state is antiparaliel. This

is illustrated in Fig. 1. lb. Note that the number of spins in the a state is higher than in the

p state, resuiting in an overall net magnetization parallel to the externat magnetic field. The

distribution of spins shown in Fig. 1. lb has been exaggerated for clarity. In fact, hE is srnall

relative to thermal motion and (Np - NJN, = IO-'.

A nuctear magnetic resonance experirnent is perfonned by subjecthg a sample in an

Fig. 1.1 A) Nuclear spin energy level diagram for a nuclide with spin, 1 = S. AE is directly

proportional to the intensity of the applied magnetic field, Ho. B) Graphical representation

of spin orientations for a nuclide with 1 = Yz The siightiy larger number of spins in the lower

energy cc state result in a net rnagnetization, w, paraiiel to the appiied field.

Fig. 1.2 The application of a transverse RF field, H,, results in an effective magnetization

vector, H,, precessing about the axis of the applied magnetic field, H,. Within a rotating

fnme and at resonance (Le. = q), H8 = 8, and precesses about the axis of the RF field.

applied magnetic field as descnbed above to a transverse radio fiequency (RF) field. This

generates a second magnetic field vector, Hl, perpendicular to H, and rotating at frequency,

Q, in the x-y plane. Therefore, the nuclear spins see an effective magnetic fieid, E&, as

shown in Fig. 1 . 2 ~ At resonance, when Q equals the Larmor fiequency, the spins wili only

see H~ = H, and the net magnetization now precesses around H, (Fig. 1 -2b). The angle of

rotation around H l is called the nutation angle, 0, and is controlled by the time duration of

the RF pulse. In other words, a 90" pulse is one of suficient duration to rotate the net

magnetization by 8 = 90". Afier the RF pulse, the net mapetkation retums to its

equilibrium position. The RF coi1 used to emit the RF pulse now serves as a receiver as it

detects the sinusoidally decaying transverse magnetization resulting from the relaxation of

the net magnetization. This signai is c d e d the free induction decay 0). The FID is

converted from a time domain signal to a frequency spectmm using a Fourier transfom.

1.4.2 NMR of geological materials ami rnagic angle spinning (MAS)

There are three important interactions which cornplicate interpretation of NMR

results when dealing with minerals and glasses. These are 1) dipole-dipole interaction, 2)

c hemical shift aniso tropy and 3) quadrupolar interaction (Kirkpatnck, 1988). di pole-dipole

interaction is an interaction between the dipole moments of individual nucleii. This is

illustrated in Fig. 1.3. The spins in nucleus j see the applied magnetic field, Ho, resulting in

a net magnetization. pj. However. nucleus j is also influenced by the dipole moment of

nucleus i, resulting in a second magnetization vector Bi. Note that the magnitude of the

interaction is inversely proportional to the cube of the internuclear distance. Chernical shift

anisotropy is non-sphencal electronic shielding at individuai sites. Quadrupolar interaction

affects NMR-active nuclides with 1 > '/z. These include many geologically important

isotopes such as "0. UNa, *'AI, 35C1 and "K. This interaction arises because of a non-

sphericai distribution of spins in the nucleus. As noted earlier, in the presence of an extemal

magnetic field, there SI + 1 spin states. Therefore, for a nuclide with 1 = 3/2, this results in

four spin states.

The diffïculties in interpretation of NMR spectra which result from these interactions

c m be mitigated by a technique known as magic angle spinning (MAS). This methoci

involves spinning the sample at an angle around the axis of the applied magnetic field. By

doing this, the dipole-dipole interaction, the chemicai shift anisotropy and the fmt-order

quadrupolar interaction are reduced to zero (Kirkpatnck, 1988). A superficial understanding

of why this technique works cm be derived from Figure 1.3. Note that the equation for the

interaction (in this case, dipole-dipole interaction) includes the term (3cos28-l), where 0

represents the ange between the extemal magnetic field vector and the p ~ c i p a l axis of the

interaction. At 8 = 54.7" (the magic angle), that term becomes zero. Rotation of the sample

at that angle relative to the applied field causes the average angle of the interaction to equal

54.7 O resulting in greatly improved fine resolution. This effect applies to all solids, although

it is less striking for amorphous materials due to stnictunl disorder (Kirkpatrick, 1988).

Fig. 1.3 Gnphical representation of dipole-dipole interaction. Nuclide 'j' is influenced by

the applied magnetic field, I&, as well as the dipole moment of nuclide 'i'. The influence

of the dipole-dipole interaction results in a second magnetic vector, Bi, which is proportional

to Urij3.

1.5 Notation

There are two types of notation associated with melt structure used in this thesis

which require some explanation. The fmt of these is Q" notation. This notation appiies to

silicate and aluminate tetrahedra in a glasdmelt network. The value of 'n' is an integer

between O and 4 inclusive and refers to the number of bridging oxygen atoms present in the

tetrahedron, that is, the number of oxygen atoms which are bonded to two tetrahedra.

Therefore, a fuily-polyrnensed network in which al1 tetrahedra are linked consists entirely

of Q* stmctures.

The second notation refers to the coordination number of a given ion and is presented

as ["'A in which n is the coordination number and A is the symbol of the element. This

notation is dso used to denote coordination numbers in crystdine materials.

1.6 Thesis Organization

Three of the chapters have been or will be submitted for publication. Chapter 2 is

a study of the solubility of Zr-bearing minerais in Na,K-aluminosilicate melts with

alkali/Al> 1 as an analogue for naturai felsic, peralkaline magmas. Among the findings

reported in this manuscript is the observation that adding s m d amounts of CI to a H,O-

saturated, perallcalïne melt significantly lowers the solubility of Zr miner& in such a

meit, relative to those with no Cl. A stnictural change in peralkaline melts resulting fiom

the addition of Cl is implied. Chapter 3 is a Raman spectroscopie study of Zr and Ti-

bearing aikali aluminosiiicate glasses, with and without added Cl. Ti-bearhg glasses

were included because Ti and Zr are chemically-similar (Le. both are high field-strength

elernents; similar valency) and because a prominent peak associated with Ti-O bonding

has been reported in previous Raman spectra (Mysen and Neuville, 1995; Henderson and

Fleet, 1995).

The results of the solubility experiments in Chapter 2 indicate a maximum

solubility of Zr-bearing minerals at about 57 wt.% SiO, for H20-saturated peraikaline

rnelts with and without added F. This dso represents the compositional threshold

between the stability field of zircon and that of wadeite (KZZrSi,O,) [for F-fke melts] or

ZrO, [for F-bearing melts]. Since alkalis are believed to play an important role in the

dissolution mechanism of Zr in perakaiine melts, an investigation of the structural

environment of Na as a function of SiO, was undertaken. Chapter 4 is a =Na nuclear

magnetic resonance magic-angle spinning (NMR MAS) study of Na-aluminosilicate

giâsses. NMR MAS was chosen as the andyticd technique for this smdy since it directIy

observes the nuciide of interest, whereas Raman spectroscopy analyses changes in overidi

glas stmc ture resulting from cornpositional ciifferences.

Collectively, this thesis represents an effort to establish physical and chernical

constraints on the stability of zirconium-beiuhg minerals in perdkaline, magmatic

environments. It also includes the application of Raman spectroscopy and NMR MAS

techniques to the study of perdkaline giasses in order to place solubiiity and phase

equiiibria experirnents in a context of melt structure.

Chapter 2

Chemical Controls on the Solubility of Zr-bearing Phases in

Peralkaline Magmas and Application to the

Strange Lake Intrusion, Canada

Robert A. Marr, Don R. Baker, A.€. WilliamsJones

Department of Earth and Planetasr Sciences, McGil University, 3450 University Street, Montreal, Québec, H3A 2A7

2.1 Abstract

The solubility of Zr-bearing mùierals was measured in H,O-saturated, peralkaline

haplobasaltic, haplosyenitic and haplogranitic melts CkF at 800 OC and 1 kbar. In

halogen-free and F-bearing melts, the solubiiity of Zr-besiring phases reached a maximum

of 4 and 3.5 W.% ZrO, respectively in melts with at 57 to 60 wt.% silica- No such maximum

was observed for C I - b e h g compositions which became saturated in a Zr-bearing phase at

~ r O f "" concentrations of 2 to 2.2 wt.%. A silica concentration of 57 to 60 wt.8 represents

a threshold abovewhich zircon is the saniraihg Zr-bearing phase for ail compositions. For

compositions with SiO, content below this range, wadeite (K,ZrSi,O,) crystalkes in the

halogen-free experiments while ZrOz is the saturating phase for the F and Cl-bearing run

products. The positive slope of the wadeite saturation curve on XmIrO?-&i, plots suggests that

increasing silica activity increases the solubility of wadeite. Since this behaviour of wadeite

is not expected based on a simple dissolution mechanism, a more complex equilibnum

involving different structural units is proposed. Observations that zircon is scarce, whereas

elpidite (Na&!%@,,-3H20) and other alkali and alkaline-earth zirconosilicates are abundant

in the Strange Lake perakaline intrusive complex in northern QuebecLabrador indicate that

the parent magma did not saturate with a Zr-bearing mineral until it cooled to a low

temperature, probably less than 600°C.

2.2 Introduction

Peralkaline, felsic rocks have k e n a focus of study by petrologists and economic

geologists because of the exotic minerais cornmonly associated with them and their potential

as ore deposits of rare metals such Zr, Y, Nb and the rareeuth elements (RE). The unusual

rnineralogy of these rocks has long held promise of providing a means for constraining the

pressure, temperature and chernical conditions of crystallization. The wide array of alkali

and alkaiineearth zirconosilicates reported in peralkafine intrusions (e-,o., Chao er al,, 1967;

Semenov, 1972; Harris, 1983; Currïe, 1985; Birken et al., 1992; Marr and Wood, 1992; Sdvi

and Williams-Jones, 1995) is particularly promising in this respect.

Cume and Zaleski (1985) pubiished the fmt phase-equilibrium resuits involving

alkali zirconosilicates and consuained the reaction:

Na,ZrS i,O,,BH,O (elpidite) * NaJrSi,O,, (vlasovite) t 2 S i 4 + 3H,O

at 0.5, 1 .O, 1.5 and 2.0 kbars. These experiments were limited to the system Na@-Si0,-ZQ-

H,OiCl. Marr and Wood (L992) later produced P-p,, (as an analogue for P-T reactions

involving hydrated phases) petrogenetic grids for the systems: Si0,-Na&03-H,0 and SiO2-

CaZrO,-H,O, To our knowledge, however, no attempt has been made to utilize phase

equilibria involving alkaii zirconosilicates to constnin petrogenetic conditions in systems

more closely approximating those of natural peralkaline magmas.

The solubility of Zr-bearing rninerals in peralkaline melts is much greater than in

peraluminous or metaluminous melts. Watson (1979) conducted the earliest systematic

study of zircon solubility in such melts. For perdkaline felsic rnelts, he suggested a solution

mechanism involving the formation of alkali zirconosiiicate' complexes of the fom:

A,ZrSi,O,,,, in which A represents an alkali metal. The existence of such complexes was

supported by Raman spectroscopie analyses (EUison and Hess, 1994). which indicate the

presence of SiO, tetrahedra with one non-bridging oxygen atom shared by one Zr and two

K ions in anhydrous, Zr-bearing, K-alutninosilicate glasses. The purpose of our snidy is to

ex tend the work of Watson ( 1979). by including S i02 under-saturited peralkaline

compositions, such as are found in well-known localities as Mt- St- Hilaire, Canada and

Lovozero, Russia. The effects of Cl and F on the solubiiity of Zr also were hvestigated, and

the results applied to the petrogenesis of the Strange Lake intrusion.

2.3 Experimental Procedures

We chose as starting rnaterids, glasses with compositions ranging from 50 to 70

wt.% SiO2, and with Na+K/Ai and Na/K values of about 1.4. These compositions span the

range comrnonly found in peralkaline intrusions. Additionally, results on the more Si0,-nch

samples c m be compared with the results of Watson (1979) who performed s idar

experiments on compositions ranging from 68 to 79 W.% SiO,. Glasses were prepared using

reagent-grade SiO2, GO,, NhCO,, &Cq, NaCl, NaF and Zrq . The carbonates and oxides

were ground together and decarbonated at 750°C for 24 hours, then reground and fused in

air at 1450°C for only one hour to avoid volatilization of Na. The resulting giasses were

then ground a third time, with NaF or NaCl added as appropriate.

For each experiment, 0.01 g of sample was sealed in a 02x1 -2 cm gold capsule with

2 PL of distilled water. The experiments were performed for 2 weeks at 800°C and 1 kbar

in cold-seal pressure vessels, utilizing Ar as a pressure medium. The pressure vessels were

air-quenched, achieving subsolidus temperatures in approximately 20 S. After the

expenments, the capsules were reweighed to ensure that there was no loss of volatiles. Only

those samples that tiad excess water upon opening were retained.

The samples were analyzed using a JEOL 8900 electron microprobe. The results of

these analyses are presented in Table 2.1. Crystalline phases were analyzed with a beam

current of 20 nA and a beam width of 5 prn to minimize damage to standards and samples.

The instrument was calibrated using zircon (Zr), albite (Na,Si,Al), orthoclase (K), fluorite

(F) and scapolite (Cl) standards. The beam width was increased to 20 pm for glasses. Glas

SiO,

A1203 Na,O

&O ZQ aZr0,

F =O

total

K+Na/AI

SiO, AI203

Na20 &O

drû, total

K+Na/AI

SiOz

Na20

60 Zr02 atm2

CI =O

total

K+Na/Al

Table 2.1 : Compositions of melts saturated with a Zr-bearing phase. Errors reported are for one standard deviation. Experiments denoted by "r" were reversed (see text). Sampies RMG6a and RMG6c are run products of the variable duration experiments shown in Fig. 2.2. These experiments were performed for one and three weeks respectively. Totals do not iaclude loss of dissolved H20 during analysis.

standards with NdSi values similar to the samples were used to calibrate the instrument for

alkalis, Al and Si where analyses of glasses were performed. This was done to account for

the effects of Na loss during analysis.

2.4 Results

S yn thetic wadeite (&ZrSi,O,) satunted the me1 t in experiments employing low-silica

glasses with no added halogens. Wadeite appeared as large (up to 50 pm-diameter), equant

crystds with weli-defmed boundaries; it was accompanied by ieucite and nepheline in some

experiments and small clusters of unreacted ZrO, were occasionaiiy detected in the cores of

wadeite crystals. Leucite exhibited Little solid solution, with Na representing about 5 mol.%

of total akalis. Nepheline displayed a very nearly ideal (Na,K)AI,Si,O,, composition. The

addition of F or Cl destabilized wadeite at 800 OC and ZrO, crystallized instead, appearing

as cIusters of smd , rounded grains. Since the source of Zr was ZrO,, the presence of this

phase in the run products did not inspire confidence that equiiibnurn was achieved.

Therefore, a senes of experiments were performed using F-bearing glasses at 650°C and

1 kbar for one week. These experiments yielded synthetic wadeite as observed in the

halogen-free expenments (Figure 2.la). This material was then ground and used in

experiments at 800°C for 2 weeks and the results compared to the initiai results at those

conditions. Zirconia reappeared as the saturating Zr-bearing phase in the reversed

experiments as elongated, euhedrd crystals 10 to 15 prn in length (Figure 2.lb). The Zr

concentrations in melts formed in the reversed experiments (Table 2.1) were similu to those

in the other experiments, but dways slightly Iower. The equilibrium concentration of Zr is

presumed to be between measurements of the forward and reversed experiments.

Disequilibrium concentrations of Zr in the melts are deemed to be much less Likely for chose

experiments in which the starting Zr-bearing phase and the saturating phase were different.

Reagent ZrO, would fmt have to dissolve iato the melt and then crystailize the new phase

Fig. 2.1. Secondary electron images of a) wadeite crystal with nepheline from sample RMGlr

at 650°C and 1 kbar after one week and b) ZrO, cqstal from the sarne sample at 800°C and

1 kbar after two weeks.

To further establish that the Zr content in the rnelts reached equiiibrium, or at least steady

state after two weeks, one sample (RMGo) was selected for a series of variable duration

experiments at 800°C and 1 kbar. The results, presented in Figure 2.2, indicate that there is

no change in ZrO, content in the melt from two to three weeks. Since Watson (1979) also

found that two weeks was a sufficient duration for experiments on zircon solubiiity, we

accept that these results represent good evidence that steady state was achieved.

Zircon was the saturating phase in ai i experiments with SiO2 contents above o r

approximately equd to 55 wt-96. Occasionaily small amounts of unreacted ZQ were

observed as ctusters of rounded grains similar to those described above, each grain measuring

about 1 or 2 p m in diameter. They do not comprise euhedrd crystals iike those observed in

the reversed, F-bearing experiments. Chemicd analyses of run products (Table 2.1; Figure

2.3) indicate a maximum solubility of -4 W.% ZrO, at 55 to 60 wt.% SiOl for the haiogen-

free and -3.5 W.% ZrOS in F-bearing experiments. As mentioned, this SiO, concentration

represents a threshold above which zircon is the saturating phase for al1 experiments (Figure

2.3). The solubility of Zr-bearing phases in the Cl-bearing run products is much lower, near

2 wt.% ~ r 0 ~ " ~ " , and shows no such maximum across the entire range of S i 4 content

-2.5 Discussion

2.5. I Zr minerals coexisting wirh -0-saturated melts

The positive slope of the curve representing the solubility of Zr in the low silica

region of the halogen-free experiments in Figure 2.3 is puzzling. It suggests that as the

activity of silica in the melt increases, the precipitation of wadeite is suppressed. This cannot

be reconciled with the reaction:

K20m" + Z I ~ ~ ~ ~ " + 3s i02mc'i 4 ~ ~ Z r ~ i ~ 0 ~ - (1)

because increased %, would favour wadeite crystallization and lower Zr& concentrations

Fig. 2.2. Concentration of ZrOl in the melt as a function of run duration for sample RMG6.

- - E R O F k-sreprese~t 0~eatandaddeviaUan.- - - - - - - - - - - - - - -

12 16

Time (days)

Fig. 2.3. Solubility of Zr-bearing phases as a function of silica concentration for peraikaline

alkali duminosilicate @asses with no added halogens (round syrnbols), with adcied F (square

syrnbols) and with added Cl (triangular symbols). The holIow circle represents data fiom

Watson (1979). A second-order polynomid was Gtted to the data for zircon-saturated, F-

karing and halogen-free experiments. Linear fits have been applied to aii other data. Dashed

lines represent extrapolations of these regressions. Error bars represent one standard

deviation.

ZrO, or wadeite satüration I zircon saturation

f

-saturated (this study) b -saturated (Watson, 1979)

Only with a strong decrease in the alkali activity would this reaction be compatible with our

results. Given the relarively low enthalpy of mïxïng in the binary alkali-silica systems [Le.

me tas table Iiquid-iiquid immiscibility gap (Haller et al., 1974)], this seems unlikel y.

Therefore, a more complicated explanation must be proposed, One possibility is the

presence of a high-silica, alkali zirconosilicate complex in the melt. Such alkali

zirconosiiicate complexes have been proposed by Watson (1979) and by Eiiïson and Hess

(1994) with (Na,K)/Zr varying from 2 (Ellison and Hess, 1994) to 4 (Watson, 1979).

Ellison and Hess (1 994) attributed the difference in aiktli/Zr ratio to the higher temperature

at which their experiments were performed and possibly a different coordination geometry

for Zr in the two studies as a result. Unfortunately, the approaches taken by Eiiison and Hess

(1994) and Watson (1979) did not allow the number of SiO, tetrahedra associated with these

melt complexes to be detennuied. The Raman spectra of Ellison and Hess (1994) can only

provide indirect information on the local environment around Zr ions in a glas or melt by

detecting changes in the structure of the network of (Si&-tetrahedra resulting from its

presence, whereas solubility studies like those of Watson (1979) provide no information

about melt structure. Thus, the presence of a high-siiica complex is speculative.

An alternative approach is to mode1 a reaction in terrns of structural units. Eiiison

and Hess (1994). concluded, on the bais of Raman spectra, that a Q3 structure (a SiO,

tetmhedron with one non-bridging oxygen atom) is involved in the aikali zirconosilicate melt

complex. Assurning this to be the case, we have used (NaK)Jr[3Si0,]m"t to represent the

aikali zirconosilicate complex, based strictly on charge balance. For simplicity, we have also

assumed ideal niixing among al1 species in the melt. Because the structure of wadeite is

based upon three-membered rings of QZ SiO, tetrahedra (Henshaw, 1955), we may write its

formula as (Na,K)2ZrZr[3Si0,]c~? Using this notation, we can write an equation that

schematicaily accounts for the increase in Zr concentrations in wadeite-saturated melts as

silica concentration increases:

Kz~r[3Si~s]m"' c K J ~ D S ~ O ~ ] ~ ~ + ~ [ S ~ O J ~ ~ ' (2)

Written in this way, the reaction proceeds to the left as the SiO, concentration increases,

consistent with the solubiiity data reported in this study.

2.52 The effect of halogens on solubility of Zr-bearing rninerals

The addition of hdogens significantly rtffects the solubility of Zr-bearing phases in

perdkaline melts. As indicated in Figure 2.3, the saturation curves of Zr-bearing minerals

in fluorinated and halogen-free, H,O-saturated melts are similar, indicating that F and H,O

affect Zr in a simiiar manner in the melt. However, there are ciifferences in the solubilities

of Zr-bearing phases in the F-beuing and F-free melts that must be addressed The solubility

of Zr-bearing phases in F-bearing, low-silica glasses is approximately equal to that in

halogen-free compositions, whereas the solubility of zircon is decreased in the F-bearing

glasses relative to the F-free glasses for the hi@-silica compositions. This observation must

be related to the effect of F on melt structure and alkali activity because direct complexation

of F with Zr is unlikely (Farges, 1996). Previous expenments have demonstrated that the

addition of F to metaluminous, haplogranitic melts increases the solubility of zircon

(Keppler, 1993). This increase results from the creation of additional non-bridging oxygen

atorns as aluminofluoride complexes are formed (Mysen and Virgo, 1985; .Kahn et al., 199 1 ;

Schaller er al., 1992). k is not clear whether such a mechanism is applicable in peralkaline

melts, however, as the presence of alkalis in excess of what is needed to charge-balance Al

may favour direct alkali-F bonding. The observation in this study that the presence of F

appears to decrease zkcon solubility for the more Sioz-rich compositions bears this out,

suggesting that (Na,K)-F bonding may be involved.

As in the halogen-free expenments, a positive slope is also observed for the

saturation curve of 210, in the F-bearing experiments below the threshold of zircon

saturation (Figure 2.3). As ZrO, is the satunting phase and given the reaction:

(~a,K),Zr(Si,0,),""' * + (Na,K),SiO,m'L + 5Sio2me't (3)

it is clear that increasing the activity of SiO, or aikalis in the melt would favour an increase

in the solubility of Zr02' The solubility of Zr-bearing phases increases with the addition of

F in the low-silica melts, but the different saturating phases in the F-free (wadeite) and F-

bearing (ZrO,) ex periments make cornparison less meaningful.

The reason for the difference in saturating phase between the low-silica F-free and

F-bearing compositions is unclear. Wadeite crystallues from the F-bearing melts at 650°C,

but is replaced by a at 800°C. ïndicating that F lowers the upper stabiiity iirnit of wadeite.

Based on the reaction:

K , Z ~ S ~ , O , C ~ ~ * ZrOZm + KzSi03"t + 2 ~ i 0 , ~ " ' (4)

the destabilization of wadeite in favour of ZrO, could be achieved by lowering the activity

of either silica or rtlkalis. Although the results of this study do not indicate which of these

mechanisms is responsible for the observed ciifferences between the F-free and F-bearing

experiments, it should be noted that the (Na,K)-F complex we propose could result in a

decrease in the availability of alkalis to interact with the duminosilicate network.

Chlorine lowers the solubility of zircon and ZQ across most of the compositional

range investigated in this study. Furthemore, the solubiiity maximum observed in the

hdogen-free and F-bearing compositions is not present in the CI-bearing experiments. It has

been known for sorne time that Cl affects melt properties differently than F or H20.

Viscosity studies of (NoCa) and (K,Ba)-silicate melts by Hirayama and Camp (1969)

indicated that the addition of F resulted in a decrease in viscosity, while Cl increased it.

Since viscosity is a reflection of the degree of polymerization, the observed increase in

viscosity mems that Cl polymerizes a silicate melt, at least for alkali and alkaiine-earth

silicate compositions. A m e n t diffusion study lends further support to this proposal.

Activation eaergies for diffusion determined by Baker (1993) were Iowered by the presence

of fluorine, whereas the addition of chiorine increased the activation energy of ciiffision.

Baker (1993) proposed that these results are due to decreases and increases in the degree of

melt polymerization produced by the presence of F and Cl respectively. Solubility and

partitioning experiments (Metrich and Rutherford, 1992; Webster, 1992) provide evidence

of a solution mechanisrn for CI which could account for this behaviour. The results of these

studies suggest that Cl forrns meft complexes with network-modifying alkalis (at l e s t for

compositions with alkalis/Alz 1) rather than with network-forming Al as is the case with F

(Mysen and Virgo, 1 9 8 5 ~ Kohn et al., 1991; Schaller et al., 1992). This solution mechanism

would lower the activity of aikalis and, therefore, be expected to reduce the solubility of Zr-

bearing phases by destabilizing alkali zirconosilicate melt complexes. Given the signifîcant

decrease in solubility of Zr-bearing minerals in Cl-bearing meits, it is clear that more

investigation is needed of the effect of Cl on the concentration of high field-strength

elements (HFSE) in perdkaline magmas.

2-5-3 Applications to the Strange Lake peralkaline granite

The mid-Proterozoic Strange Lake (Lac Brisson) peralkaline granitic pluton on the

QuebeciLabrador border may represent the extension into Labrador of Gardar anorogenic

igneous activity (Cume, 1985). The pluton comprises two main rock types; a hypersolvus

granite and a volatile-saturated, typicdly altered, subsolvus granite (Nassif, 1993). The

hypersolvus granite consists of mesoperthitic alkali feldspar, interstitial quart. and late-

crystallizing arfiedsonite with accessory fiuorite, zircon, thorite, aenigmatite and

as trop hylii te.

The subsolvus granite is either equigranular or porphyritic. It contains idiomorphic

quartz, arfvedsonite (occurrhg as phenocrysts in the porphyritic granite) and two alkali

feldspars. Accessory minerals in the subsolvus granite include late zircon, elpidite

(Na,ZrSi60,,-3H,O), annstrongite (CaZrSi,O,,-2.5H20), gittinsite (CaZrSi,O,), kainosite

(Ca,[Ce,Y,HREE],Si,O ,,[CO3] .HO) and calcite. Pegmatites and miarolitic cavities are

associated with the subsolvus granite, ixidicating that it crystaüized under water-saturated

conditions.

Elpidite has been identified as the primary, magmatic Zr-bearing minerai in the

subsolvus granite (Salvi and Williams-Jones, 1995). It appears as doubly terrninated

orthorhombic prisms up to 3 mm in length that may be flow-siligned (Miller, 1996). It shows

varying degres of rirn and core replacement by gittinsite or ~unstrongite. Anhecirai elpidite

ais0 has been reported in the porphyritic granite either as single crystals interstitial to quartz

and feldspar or as large aggregates of smdi p i n s (4 .1 mm in diameter) (Salvi and

Williams-Jones, 1995).

Althougfi elpidite was the prïmary, magmatic Zr-bearing mineral at S trange Lake, the

experïments performed in this study have yielded only zircon, ZrO, and &ZrSi,O,. This

suggests that the upper stability limit of elpidite is exceeded at the P-T conditions of these

experiments. This is consistent with the results of Currie and Zaieski (1985), which

indicated that the transition from elpidite to vlasovite + quartz + H20 at P(H20) = 1 kbar

occurs at 595 & 4°C. The inferred pressure of crystakation at Strange Lake, based on fluid-

inclusion data (Salvi & Williams-Jones, 1992), is 0.7 kbar, assurning Ph, = Pmm. At this

pressure, the upper thermal stabiliv limit of elpidite is about 57S°C, a low temperature for

a magma, but close to the approximate lower-tempenture iimit suggested by Nassif (1993)

of 590°C for the subsolvus-granite-fonning magma at Strange Lake. The solidus temperature

proposed by Nassif ( 1993) is based on that reported for a topaz-bearing quartz keratophyre

(ongonite) Erom Mongolia (Kovalenko et al., 1971) with a high F content similar to the

Strange Lake intrusion. Thermometric studies of melt inclusions in topaz from this locality

indicated a solidus between 550 and 600°C (Naumov et al., 1971), which Nassîf (1993)

estimated to be 575°C at 1 kbar and 590°C at 0.7 kbar. Thus, the solidus temperature is not

well-constrained. Furthemore, the Mongolian ongonite is not peraikaiîne. Experiments have

indicated that perdkaline rnelts have solidi signifïcantly lower than those of hapiogranitic

melts (McDoweil and WyUie, 1972). Indeed, some unusual granitic and syenitic rnelts have

been reported with solidus tempentures below 500°C (London et cil., 1989; Piotrowski and

EdgarJ970). Therefore, lowenng the proposed solidus temperature for the Strange Lake

subsoIvus magma can be justified.

Since zircon is the saturathg phase at 800°C and 1 kbar, granitic intxusions which

contain elpidite as a magmatic phase must not have been saturated until late in the

crystdlization process or must have crystdized from a Iow-temperature magma. Textural

evidence fiom the subsoIvus granite (Salvi and W'iams-Jones, 1995) and related pegmatites

(Miller, 1996) indicate that in the case of Strange Lake, elpidite crystaiiized early. Based on

the chemopphy of the common aikali zirconosilicate (AZS) minerais, which generally lie

on the SiO2-(Na,K)JrO, pseudobinary ( M m and Wood, 1992), zircon is an indifferent

phase and is, therefore, stable at the same P-T conditions as elpidite, wadeite, vlasovite and

other cornrnon AZS minerais. Consequently, if zircon crystallized from a magma at a

temperature above the upper stability b i t of elpidite, it would remain stable as the magma

cooled to a tempenture at which elpidite would also be stable. indeed, small arnounts of

early zircon with a reaction rim of vlasovite or elpidite have been reported in the

hypersolvus granite at Strange Lake (Birkett et al., 1992). These may represent localized

pockets i t which zircon saturated early or may be xenocrysts. Furthermore, vlasovite at

Strange Lake commonly displays a reacîion rirn of elpidite in contact with quartz (Birkett

et al., 1992). This sequence indicates a reaction senes of zircon - vlasovite - elpidite as

temperature decreased.

Thus, presence of magmatic alkaii ùrconosilicates in peralkaline granites implies

either a lower initial Zr content in the melt or a lower melt tempenture at the tirne of

emplacement than if zircon is present. Three possible paths of melt evolution are indicated

in Fig. 2.4. Here we assume that Zr behaves incompatibly until either zircon or elpidite

saturation in the meit. Path A illustrates the evolution of a relatively high-temperature

magma with a relatively hi@ initial Zr content. As the magma cools, the concentration of Zr

in the me1 t increases until zircon saturates. Path B , indicates how a high-temperature magma

with a low initial Zr content would evolve differently from that illustrated by path A. Once

again, the magma becomes enriched in Zr as it cools. However, saturation in zircon does not

occur and the melt continues to cool until elpidite saturates. An alternative path that could

lead to elpidite saturation is represented by B,. A magma that has a high initial Zr content,

but is emplaced at a lower temperature could avoid saturation in zircon and crystallize

elpidite. It is this third alternative which is the most compatible with the petrographic

textures observed at Strange Lake.

This mode1 must be viewed as highly qualitative, since it does not take into account

the effect of variable agpaitic index (alkWAl); the index will increase in a fkactionating,

peralkaline, felsic melt. Nor does it consider pressure which has a significant effect on the

stability field of elpidite, especiaily for pressures below 1 kbar (Currie and Zaleski, 1985).

However, the mode1 does provide some insight into the evolution of the peralkaline felsic

magmas. High concentrations of Zr can be reached in the late stages of magmatic

crystallization from melts with reiativety low initial Zr content. On the other hand, high

initial Zr concentrations may result in early saturation of zircon, which would be dispersed

throughout the magma, yielding final Zr contents comparable to those of magmas with low

initial concentrations.

2.6 Conclusions

1) For H,O-saturated, peralkaiine, Zr-rich, aikali duminosilicate melts with 55 W.%

SiOl or more, zircon is the saturating phase at 800°C and 1 kbar, whereas wadeite appears

in melts with lower silica contents.

2) The addition of F Lowers the upper thermal stabiiity limit of wadeite in favour of

Fig. 2.4. Schematic representation of cooling paths for three perdkaline magmas assuming

Zr behaves as an incompatible element. Magma A represents a relatively high temperature

melt with high initial ZrOz content which becomes saturated in zircon. Magmas B, and B,

represent melts at high T and low initial ZrO, content, and low T and high initial Zr&

content, respectively, which avoid early saturation in zircon and crystallize elpidite.

ZrO, in low siiica melts and decreases the solubiiity of zircon in hi&-silica melts

3) There is a maximum in the solubility of Zr-bearing minerals at ktween 55 and 6û

wt.% SiO2 for water-satunted perdkaline melts with and without added F.

4) The addition of Cl serves to depress the solubility of Zr-bearing phases,

panicularly in the compositional range where a maximum is observed in Cl-& experiments.

This may be the result of lower alkali activity due to alkali-Cl complexing.

5) Fluorine appears to destabilize aikali zirconosiiicate meIt complexes, possibly as

a result of aikaiï fluoride cornpiexation.

Introduction to Chapter 3

In Chapter 2, it was reported that the addition of CI to perdkaline melts depresses the

solubility of Zr-bearing minerais across a wide range of SiO, content. Ellison and Hess

(1994) indicated the presence of a peak in the Raman spectra of K,Zr-silicate glasses which

is associated with Si-O stretching vibrations in Si-O-Zr bridges. This suggests the possibility

of using Raman spectroscopy to determine how CI affects the solution mechanisrn of Zr.

Ti-bearing alkali silicate glasses have also k e n studied by Raman spectroscopy

(Mysen and Neuville, 1995; Henderson and neet, L995), but the pezk appearing in the

spectra of Ti-karing @asses is associated with Ti-O vibrations. Therefore, Raman analyses

of Ti-bearing glasses with added Cl may serve as a useful cornparison since any changes

observed in the spectra would be the result of some direct effect on Ti-O bonding.

Chapter 3 is a Raman study of Zr and Ti-bearing alkali aluminosilicate glasses with

and without added Cl. The objective is to determine changes in the local structural

environment of these two metais in glasses and, by inference, melts which resuIt from adding

Cl.

Chapter

The Effect of CI on the Structure of Zr and Ti-bearing alkali

aluminosilicate glasses

Robert A. Marr', Don R. Baker', Andrew Vreugdenhi12 and A.€. VftJliamsJones'

'Dept of Earth and PIanetary Sciences, McGill University, Montreal, QC 'Oept of Mining and Metallurgy and Dept of Chemistry, McGill University, Montreal, QC

3.1 Abstract

Na-alurninosilicate glasses wi th Na/Al> 1 and added Ti or Zr have been analyzed

by Raman spectroscopy to determine the effect of Cl on g las structure. The spectra of

the Ti-bearing glasses show a significant difference between the Cl-free and the C1-

bearing composition. The Cl-free glas spectrurn contains a strong, asyrnmetric peak at

900 cm-' which is associated with Ti in five-fold coordination. This peak is shifted to

higher frequency and becornes more symmetrïc with the addition of 0.3 wt.% Cl.

Deconvolution of the high-frequency waveband suggests that ciifferences between spectra

are the result of a contribution from a peak at 945 cm-'. This peak is believed to be the

result of Ti-O vibrations in fully-polymerized titanate tetrahedra It is proposed that the

addition of Cl destabilizes [''Ti in favor of tetrahedral coordination as a result of

cornpetition between Cl and titanate groups for alkalis.

The spectra of Zr-bearing @as show a strong peak at 980 cmaL which is not seen

in spectra of the base glass, indicating a vibntional mode associated with a

zirconosilicate structure. There is no discernible difference between the Raman spectra of

Zr-bearing glasses with and without added Cl, indicating that the local structure around

Zr is insensitive to the presence of CL.

3.2 Introduction

Viscosity studies of CI- and F-bearing alkali-alkiiiine-earth silicate rnelts indicate a

decrease in viscosity with added fluorine, but an increase with added chlorine (Hirayama and

Camp, 1969). These viscosity measurements can be interpreted to indicate that F

depolymerizes silicate melts, whereas CI increases the degree of melt polperization. This

increase in viscosiq with addition of Cl was also observed by Baker and Vaillancourt (1995),

in a study comparing the viscosity of a CI-bearing albitic melt with that of pure albite melt

at 1400°C and 1.5 GPa. A recent diffusion study lends funher support to this proposal;

activation energies for difision detennined by Baker (1993) were lowered by the presence

of fluorine, whereas the addition of chlorine increased the activation energy of diffusion.

Baker proposed that these data reflect an increase in the degree of melt polymerisation due

to the presence of Cl. Recent solubility and partitionhg experiments provide evidence for

a solution mechanism for CI which could account for this behaviour (Metrich and

Rutherford, 1992; Webster, 1992). The results of these studies suggest that Cl forms melt

complexes with network-rnodifjring aikalis (at l e s t for compositions with alkalis/Alz 1)

nther than with network-forming AI as is the case with F (Mysen and Virgo, 1985a; Kohn

et al., 199 1; Schaller et al., 1992).

While the effect of Cl on the physical properties of silicate melts and glasses is well-

established, the changes in network stmcture responsible for these changes have not yet been

detected. Raman and NMR MAS ('?U, '%i and = ~ a ) analyses of albite glass with 0.5 wt.%

Cl (Baker and Vaillancourt, 1995; Baker, unpubl. data) indicate little change in the spectra

as compared with those of pure albite glass, suggesting little discemible change in glas

structure. Yet, the effect of Cl on viscosity must result in some measure from structural

change.

One possible approach to examine the effect of Cl on melt stmcture is to study it

indirectly by analyzing samples with an additional element which may interact with CL.

Phase equilibna in the system N%O-Si0,-Zr02-H2W1 suggests that Zr may be such an

element (Currie and Zaleski. 1985). Efforts to synthesize sodium zirconosilicates revealed

that the source of Na affected which phase would crystdlize. Specificdiy it was reported

that pwtkeldyshite (NaJrSi20,) woulci appex if NaCl was used as opposed to vlasovite

(NaJrSi,O, ,) if N-CO, was used as the source of N a This suggests that Cl may alter the

local environment around Zr. The siniilarity in chernical behavior between Zr and Ti irnplied

by thek proximity in the periodic table indicates that the latter rnay be affected similariy by

CI.

Raman spectra have been reported for Zr-bearing potassium silicate glasses (Ellison

and Hess, 1994) and Ti-bearing sodium silicate glasses and melts (Mysen and Neuviiie,

1995). In both of these studies peaks were reported which were not observed for the base

glasses free of Ti or Zr, but were associated specifically with the transition metal studied.

In light of the fact that these peaks appear even with only a few weight percent of Ti or Zr,

it is possible that the addition of Cl, which has a Iow solubility compared to F or H,O, may

have some noticeable effect on the spectra. The purpose of this study is to examine the

Raman spectra of Ti- and Zr-bearing Na-aluminosiiicate glasses. By this approach, it is

hoped that the effect of added Cl on the glas structure around these two met& may be

de termined.

3.3 Experimental Procedure

3.3. I Sumple Preparation

Samples were prepared from reagent grade Na,CO,, SiO, N203, ZrO,, TiOz and

NaCl. Carbonate and oxide powden were ground together and decarbonated at 750°C for

48 houn. The sample material was ground and fused at 1450°C for one hour in air, then

reground with added NaCl as appropriate, sealed in Pt capsules and fused at 1600°C and 1.5

GPa for two hours in a piston-cylinder apparatus using NaCl-crushable Al,03-fused silica

assemblies (Hudon et al., 1994). The samples were prepared at high pressure to ailow for

high temperature homogenization of the $lasses without CI volatilization. The effect of

pressure on the structure of the samples is expected to be negligïble. Mysen et al. (1983) has

s h o w only srnail changes in the average T-0-T angle and the absence of new peaks in the

Raman spectra of Na-duminosilicate glasses as pressure is increased hm 1 atm to 3.0 GPa

The sample compositions were verified by electron microprobe analyses using a 40 pm beam

diameter and a 2 nA current calibrated with g l a s standards. Using these andytical

conditions, Na migration was eliminated. The results are reported in Table 3.1. It should be

noted that while no water was added to the samples, previous studies in Our laboratory have

indicated that a srnall amount of dissolved water, on the order of 0.5 wt.%, is inevitable in

melts prepared in this manner. Glas samples were examined opticaily and those devoid of

inclusions were retained for Raman analysis.

3.3.2 Raman Spectroscopie Analyses

The 5 14.5 nm line of an Ar' laser was used for excitation at a power of 200 mW. The

spectrometer was a JY Spex 750M single monochrometer with a Liquid N,-cooled CCD

detector. Spectra were acquired by averaging 10 iotegrations of 20 s each. Slit width was set

at 300 Pm. The monochrometer was calibrated ushg MMT [(C,H4CH3)Mn(CO),] which

provides strong Raman peaks across the entire frequency range of interest.

The background was removed manuaily and the spectra were smoothed with a 9-point

smoothing function. Peak intensities were nomdized using the 470 cm-' peak in order to

allow cornparison of the high-frequency region between 850 and 1200 cm".

3.4 Results

nie spectra of the base glasses with no added Zr or Ti (Fig. 3.1) show a band in the

low frequency region in which two peaks are observed, a large peak at 485 cm" and a

I - - -

RML RMLC RMLT RMLTC RMLZ RMLZC

- - - - -

total 99.86 99.67 99.56 9959 99.67 9934

Table 3.1. Electron microprobe analyses of g l a s samples. Al1 compositions are averaged from ten or more analyses. Errors represent one standard deviation.

Fig. 3.1. Raman spectra of base glasses: RML and RMLC, Cl-free and CI-be;uing,

peraikaline Na-duminosilicate glass, respectively.

much smaller one at approximately 560 cm? The large peak is not weH-understood although

it has been noted that it is associated with a large displacement of oxygen (Galeener and

-Mikkelson, 198 1) and it h a , therefore, been interpreted to represent symmetric vibrations

in the plane bisecting Si-O-Si bridges (McMillan et al., 1982). nie srnalier peak at 560 cm-'

is believed to represent the "breathing" mode of three-membered ring structures (Matson et

al., 1986; Kubicki and S ykes, 1993). This peak appears at about MW) cm'' in spectra of fused

silica, but the presence of Al tetrahedra in these ring structures shifts it to Iower fkquencies.

in the medium frequency range, there is an asymmetric peak at 780 cm". ui fused

silica and albite glas, this peak is found at approximately 800 cm", with probable

components at 790 and 830 cm" (Mysen et al., 1982). This peak is beiieved to represent

vibration of silicon within its oxygen cage with Little associated displacement of oxygen

(McMillan et al., 1982). The position of this peak in less polymerized @asses is shified to

lower frequencies due to the decrease in rigidity of the network.

There is a large, asymmetric peak in the high-fiequency region, between 850 and

1200 cm? This band has a peak at 1060 cm-' with a shoulder at a lower frequency. A peak

at about 1100 cm-' in alkali silicate @asses is generaiiy assigneci to vibrations associated with

Q' structures, that is, silicate tetrahedra with one non-bridging oxygen (Brawer and White,

1975; Virgo et al., 1980; Furukawa er al., 198 1 ; Mysen and Frantz, 1994), although it has

been suggested that it may also be associated with alkali-oxygen bonding (Fukumi et al.,

1990; McMillan et al., 1992). This peak shifts to lower frequency in alkali aluminosilicate

$asses as the AU(Al+Si) ratio increases (Mysen, 1990). The shoulder probabiy indicates the

presence of a significant number of Q' structures (SiO, tetrahedra with .two non-bridging

oxygen atoms) which are manifested in Raman specua by a peak at 950 cm" (Brawer and

White, 1975; Furukawa et al., 198 1; McMilIan, 1984a; Mysen and Frantz, 1994). It should

be noted that there is no demonstnble ciifference between the spectra of the Cl-free and CI-

bearing compositions in Fig. 3.1.

The addition of Zr has a sipiîlcant effect on the Raman spectra in the hi&-fkquency

region (Fig. 3.2). A new peak emerges ar 980 cm-' which may correspond to the 1010 cm-'

peak reported by Ellison and Hess (1994). Etlison and Hess proposed that tbis peak

represents the Si-NB0 stretch of a species in which the non-bndging oxygen is shared by

a Zr and two aikali ions. The shift in the position of this peak to lower frequency may be a

result of the effect of Al-containhg tetrahedra reducing the overail rigidity of the glass

network; the compositions studied by Ellison and Hess did not contain Al. Altematively, the

choice of dicati metai used in the glasses studied may account for the difference in peak

positions; in this study, Na was use& while Ellison and Hess used K in the preparation of

their glasses. This latter explmation is deemed less plausible, however, as significant

structural change due to the larger K+ ion seems unlikely in the open structure of an alkali

duminosikate g las .

Very little difference is observed between the CL-bearing and Cl-free gIasses in Fig.

3.2. A slight broadening of the high-frequency band toward higher wavenumbers may

indicate an increase in more highly-polyrnerized species, but the difference is too small to

be definitive. The spectra of the Ti-bearing glasses (Fig. 3.3) show the strong peak at 1060

cm", as well as a prominent peak at 9 1 O to 930 crd. This latter peak is more prominent, less

asymmetrical and shifted to higher frequency in the Ci-bearing glass. A comparative

deconvolution of the high-frequency region of the base glass and the Ti-bearing glasses is

presented in Fig. 3.4. The deconvolution of the base glass spectmm (Fig. 3.4A) was

performed assuming the presence of two peaks, the positions of which were lefi

unconstrained. The peak positions and intensities derived are indicated in Table 3.2.

Examination of the spectra of the Ti-bearing glasses indicates the presence of a third major

peak at approximately 900 cm''. Therefore, the deconvolution of these spectra was

perfomed assuming three peaks, with the positions of the peaks at 950 and 1060 cm-'

Fig. 3.2. Raman spectra for samples RML, RMLZ and W C . These samples are the base

glass, the base composition with added Zr and with added Zr and Cl, respectively. The peak

at 975 cm-' observed in the Zr-bearing compositions is associated with an alkali

zirconosilicate melt complex involving a Q~ structure with the loue non-bridging oxygen

being shared by one Zr and two alkali ions (Ellison and Hess, 1994).

cm-

Fig. 3.3. Raman specua for samples RMli, RMLT and RMLTC. the base glass, with 1.9

wt.% added Ti and with Ti and 0.3 wt.% added CI respectively. A prominent peak appears

at 900 cm-' in the Ti-bearing glasses which is associated with Ti-O bonds.

I 800 cm-'

Fig. 3.4. Deconvolution of the high-frequency region of the Raman specua for (A) the base

glass and the Ti-bearing samples. (B) RMLT and (C) RMLTC. The increase in intensity of

the peak at approximately 950 cm*' may result from an increase in the number of fuiiy-

polymerised, titanate tetrahedn as a result of added Cl. See Table 2 for exact peak positions

and intensi ties.

1m 1 la, cm"

IRML RMS: 0.0 172

1 RMLT RMS: 0.0186

- -

position

1 RMLTC RMS: 0.0288

949 0.320 28.8

intensity

position

position intensity integral

integral

intensity integral

Table 3.2. Positions and intensities of peaks fitted CO spectra of the high-fiequency region of the base glass and Ti-bearing glasses with and without added Cl. The residuai root-mean- square (RMS) error is included for each sarnple. Peak positions were unconsvained in the specûum of the base glass. The positions acquired for the base giass were used to constrain the 950 cm" and 1060 cm'' peaks in the spectra of the Ti-bearing glasses.

899 0.833 53.8

constrained to approximately those of the base glas. The results in Fig. 3.4 indifate that the

addition of Cl serves to increase the 950 cm-' pedc relative to those at 900 and 1060 cm-'.

The peak at 900 cm-' also appears diminished relative to that at 1060 cm", although the

difference is much more subtle and the validity of the observation more uncertain. Once

again, exact peak positions and intensities are indicated in Table 3.2.

In addition to the clifferences between the spectra of Ti-bearing and base glass in the

high-f~quency region, the presence of a low-fkquency shoulder on the 485 cm" peak is also

observed. This shouider is more prominent in the CI-bearing sample, but is present in both

Ti-bearing spectra. Henderson and Fleet (1995) repoaed that sorne peaks of anatase and

rutile (polymorphs of T i 0 3 occur in this region of the spectrum. However, optical analysis

of the Ti-bearing glasses indicated no evidence of phase separation which might account for

the appearance of peaks associated with @ ? ~ i .

3 5 Discussion

It is clear fiom the spectra in Figs. 3.2 and 3.3 that the addition of Cl has a significant

effect on the structure of Ti-bearing akali duminosilicate glasses and little discernible effect

on the structure of Zr-bearing glasses with similar compositions. This results from d i f f e ~ g

solution mechanisms for these metals in silicate melts. In the case of Zr, it has been proposed

that in silicate melts in which (Na+K)/N > 1, Zr forms alkali zirconosilicate meit complexes

of the form (Na,K),ZrSi,02+h,,, (Watson, 1979, Ellison and Hess, 1994). Some

disagreement exists between the two studies regarding the ratio of total akalis to Zr in the

melt cornplex, but Eiiison and Hess (1994) suggest that this is a function of the g las

transition temperature, that is, the temperature range across which a supercooled melt

becomes rigid and, thereby, "freezes" the melt structure in the gassy state. The (1-x)

K,O-xZr0,*4SiO, glasses analyzed by Eliison and Hess were prepared at 1600 to 1800°C,

while the experiments of Watson (1979) were performed at 700 and 800°C. Since the glass

transition temperature is a function of the cooling rate, the high temperature, rapidquench

experiments of Ellison and Hess would have a higher glass transition temperature than those

of Watson and would, therefore, have the melt structure at a higher temperature "fiozen"

into the glas. Eiiïson and Hess proposed that a peak which appears at 1010 cm-' with the

addition of Zr to the base glas represents a Q' structure in which the lone non-bridging

oxygen is shared by one Zr and two K. A peak at approximately this same location is

indicated in both spectra presented in Fig. 3.4.

Raman spectra of Ti-beanng Na-siiicate glasses and meIts showed a peak at 880 cmeL

which increases in prominence with increasing Ticontent (Mysen and Neuviiie, 1995). A

similar peak is reported by Hendenon and Fleet (1995) at 843 cm". The authon of the latter

study compared the spectra of titaniferous Na-silicate glasses to those of several crystalline

Ti-bearing compounds with differing coordination geometries for Ti. Based on this

cornparison, they concluded that the peak at 843 cm-' results ikom vibratious associated with

Ti-O bonding in tetrahednl coordination at low TiO, concentrations. A slight shift to higher

frequency with increasing titanium concentrations was reported by Hendenon and Fleet,

which they proposed resulted from Ti in five-fold coordination. A recent XANES study with

a more polymensed composition confmed the presence of primarily [ q i in titaniferous Na-

silicate giasses (Farges, 1997). The coordination geometry proposed for [''Ti consists of a

square pyramidal structure with a short Ti=O bond and four long Ti-O bonds (Farges et al.,

1996; Farges, 1997; Cormier et al., 1997). Bond valence-bond length correlations of the

titanyl unit (Ti&) favor a non-bridging oxygen bonded to 3 or 4 low field strength ions such

as alkaiis (Farges et al., 1996). The rernaining oxygens are probably bridging oxygens

bonded to s ikate tetrahedra or a mix of silicate and titanate tetrahedra (Fig. 3.5).

In the current study, a large peak appears around 900 cm'' for the Ti-bearing glasses,

which is not seen in the base glasses. It seems likely that this peak results from a vibrational

mode associated with Ti-O bonding, but the different peak positions reported by Henderson

Fip. 3.5. A possible bonding mode1 for Ti in five-foId coordination based upon bond valence-

bond lene& correlations of Farges et al. (1996). The structure iiIustnted is that of NfliSiO,

and is similar to fiesnoite (BaJiS&O,). Titanium in this structure is bonded to both bridging

and non-bndging oxygens (modified after Farges et aL, 1996).

and Fleet (1995), Mysen and Neuville (1995) and the current study îndicate differing bond

strengths in each case. The base composition in Henderson and Fieet was N@i03, while that

of Mysen and Neuville was N@iO,. a composition which would result in a more

polymerized melt structure compared to NqSiO,. As a shift of 37 cm-' exists in peak

positions between the two studies for the same concentration of dissolved Ti, it is concluded

that the iower degree of polymerisation in the glass compositions of the former study is

responsible for the shift to iower frequency. C o r n p a ~ g the resuiü of these previous studies

to those of the present study lends further credence to diis conclusion. The glasses used in

this study are much more polyrnensed than those of either of the previous studies and the

peak associated with Ti-O bonding is shifted to still higher frequency. A comparison of peak

position as a function of the average number of non-bndging oxygens per tetrahedra

(NBOm, assuming T= [Si+Al] is presented in Fig. 3.6. Given the uncertainty in establishing

the exact peak positions when dealing with the broad bands produced by silicate glasses, the

trend shown in Fig. 3.6 appears linear. The NBO/T values presented are for base glasses as

the effect of Ti is diff~cuit to quanuS. However, if the effect is constant for alI three studies,

then the comparison is vdid assuming approxirnately equal concentrations of dissolved Ti.

The relatively low sensitivity of peak position to Ticoncentration reported by Mysen and

Neuville (1995) and Henderson and Fleet (1995) gives a high degree of confidence in

interpolating peak positions for Ti concentrations similar to that in this study.

Henderson and Fleet (1995) also produced Raman spectra for Ti0,-SiO, glasses.

They reported that the addition of Ti to vitreous silica results in the appearance of peaks at

945 and 1 LOO cm" which were interpreted to represent the reaction of the silicate network

with Ti. The peaks were interpreted to represent the stretching vibrations of Si-NB0 bonds

in Q' and Q3 structures respectively. resulting from disruption of the silica tetrahedral

network by Iarger Ti-tetrahedn. A later EXAFSKANES study (Henderson and Fieet, 1997)

indicated that the addition of Ti to fused siiica has little effect on the network. This suggests

Fig. 3.6. Cornparison of the peak position for the peak associated with Ti-O bonding as a

function of the number of non-brid=eg oxygen per tetrahedron. NBOiT values are for base

glass compositions and are, therefore. only approximate for Ti-bearing compositions.

- This study

\ Mysen and Neuville (1 995)

Henderson and Flee

that the Raman peaks at 945 and 1 100 cm" represent vibrations associated with Ti-BO

(bridging oxygen) bonds in fully-polymerised tetrahedra. However, EXAFS/XANES may

not be able to distinguish fully-polymerised silica tetrahedra fkom those with one or two non-

bridgïng oxygens (G. Henderson, perscomm.). Therefore, the interpretation that the peaks

at 945 and 1100 cm-' represent @ and @ structures (Henderson and Fieet, 1995) cannot be

d e d out. In both cases, these peaks are associated with the presence of fully-polymerized

Ti-tetnhedn, either directly as a result of vibrations of Ti-O bonds or due to the appearance

of Q2 and Q3 stnrctures involving Si andlor Al.

The addition of Ti to the base glass composition in this study resutts in an increase

in the peak at 950 cm" in addition to the appearance of a new peak at 900 cm-' (Fig. 3.4).

A further increase in this peak is observed with the addition of Cl. Since no change is

indicated in the spectra of the base glas or the Zr-bearing glas with the addition of Cl, this

increase in intensity must be rdated to changes in the local structure around Ti. The fact that

Hendenon and Reet (1995) reported a peak at 945 cm-' resulting from the presence of [''Ti

in fused silica suggests that the increase in intensity of the peak at 950 cm-' may also indicate

the presence of tetrahedrdy-coordinated Ti.

The solubility of Ti, like that of Zr, is greatly enhanced in natural silicate melts in

which alkaii/Al> 1 (Gwinn and Hess, 1989). Furthemore, Ti often crystallizes unusual aikali

titanosilicate minerals such as narsarsukite [Na2(Ti,Fe)Si,0,,(OH~] and aeni=patite

[~a,Fe;*TiSi,0~J from such rnelts, in much the same way that Zr forms alkaii

zirconosiiicates such as catapleiite ~a&Si,O,-SH,Oj and wadeite [KZZrSi,Og] (Salvi and

Williams-Jones, 1990). It is not unreasonable to suggest, therefore, that Ti forms alkali

titanosiiicate complexes in alkali-rich silicate melts, in much the same way as Zr forms alkali

zirconosilicate melt complexes. Indeed, just such a complex of the form K,Ti,Si,O,, was

proposed for Ti-bearing, hi&-silica, K-aluminosilicate glasses (Dickinson and Hess, 1985).

The I s i~ i structure proposed by F-es et al., (1996) may be thought of in terms of an alkali

titanosîiicate melt complex. One of the three possible bond-valence models they proposed

requires the presence of some Ti-O-Si M a g e s and al1 their models require the titanyl unit

(Ti=O) to bond witb several alkalis in order fulfd its bond-valence requirements- Thus, it is

possible that if Cl bonds preferentiaily with alkalis in the melt structure, a cornpetitive

situation could arise between CI and an a h d i titanosilicate melt complex. The composition

of the Cl-bearing titaniferous glas (RhUTC) from Table 3.1 has only 3.2 mol% Na,O in

excess of that needed to charge-balance Ai. This same glas has 1.9 mol% TiO, and 0.6

mol% Cl. Thus, if Cl cornpetes effectiveiy for Na, Ti may be forced to alter its soiution

mechanism from five-fold coordination in an alkaii titanosilicate melt complex to network-

forming, tetrahedral-coordination.

The finding that adding Cl to the Zr-bearing glass has little effect on the Raman

spectra must be addressed. It is generally accepted that Zr is 6-fold or 8-fold-coordinated in

silicate glasses and melts. Such coordination geometries do not require that one of the

oxygen atoms be bonded only with akalis to satisfy its bond-valence requirements as is the

case in a [']Ti structure. Therefore, it is reasonable to conclude that such a coordination

geometry would be far less sensitive to cornpetition from Cl for alkalis. Furthemore, the

large ionic radius of Zr4 makes it impossible for it to revert to tetrahedral coordination.

Therefore, addition of Cl to the melt would not cause any detectable change in the intensity

or position of any Raman peak associated with Zr.

The phase equiiibria study of Cunie and Zdeski (1985) indicated that Cl exerts some

control over the crystallization of allcali zïrconosiiicate phases. Since the results of this study

indicate that no discernibte change in structure results from adding Cl to a Zr-bearing, alkali

aiuminosiiicate melt, the findings of Cume and Zaleski are possibly the result of Cl affecthg

the activity of alkalis or -0.

3.6 Conclusion

The solution mechanism of Ti"' in alkali aluminosilicate melts is believed to be

primarily that of alkali titanosilicate melt complexes, with Ti in five-fold coordination.

Rarnan spectroscopy provides evidence to suggest that Cl can destabilize these structures to

sorne extent in favor of fully-polymerized Ti-tetrahedra, especidly if the availability of

network-modimng alkalis is limited. The cornpetition for aikalis between CI and non-

bridging oxygen atoms associated with Ti-bearuig structures suggested by the results of this

study would aIso be expected to increase the viscosity of these melts by forming fûlly-

polymerized Ti-teuahedra. These results also support the mode1 that NaCl melt

complexation may be responsible for viscosity increases in other Cl-bearing silicate melts

as well.

Introduction to Chapter 4

Since alkalis play an important role in tbe solution mechanism of Zr and other bigh-

field strength elements in perdkaline magmas, it is useful to understand the structural

environment of alcalis and how it changes, if at aii, as other compositional parameters

change. A good technique for such a study is nuclear magnetic resonance magic-angle

spinning (NMR MAS) as it provides a direct analysis of changes in local environment of the

nuclide of interest. In Chapter 4, =Na NMR MAS analyses are performed on a suite of Na-

alurninosilicate @asses with varying Si02 to determine any change in locai structure such as

coordination number or average Na-O bond length.

Chapter 4

23Na NMR MAS Analyses of Na-Aluminosilicate

Glasses with Varying SiO, Content

Robert A. Man, Don R. Baker, A.E- WilliamsJones

Department of Eadh and Pianetary Sciences, McGili University, 3450 University Street, Montreal, Québec, H3A 2947

4.1 Abstract

=Na NMR MAS analyses of a suite of Na-aluminosiiicate glasses with Na/N = 2 and

varyîng SiO2 content h a revealed a negative shift in peak position as the glass structure

becomes more polymerized, i.e. the average nurnber of non-bridging oxygen atoms per

tetrahedron (NBOm decreases. This trer,d is observed only for giasses with NBOm s 0.27.

For more polyrnerized glasses no change in chemical shifi is measured. Past studies have

indicated a strong correlation between increasing Na-O bond length and increasingiy more

negative isotropie chemical shift in a ~ a NMR spectra. This suggests that for Na-

aiuminosilicate glasses with NBOK s 0.27, Na-O bond lengths would be expected to

increase as polymerïzation increases.

4.2 Iatroduction

Interpretation of UNa NMR MAS spectra in glasses is complicated by the

quadrupolar nature of this nuclide. In general, researchers compare the spectra of giasses

with those of crystalline phases to determine approximations of coordination number and

bond length (Maekawa et al., 1997). It then becomes possible to compare the spectra of

different giasses in order to observe how glass suucture responds to changes in composition.

The most conunonly studied characteristic of NMR spectra of glasses is the isotropic

chemical shift, the shifi in peak position resulting from differences in the degree of magnetic

shielding of the nucleus by local electrons. Changes in local structure distort electronic

orbitals and thus, change the amount of magnetic shielding. More negative chemical shifts

are indicative of p a t e r shielding. Systematic studies of glasses using = ~ a NMR MAS

indicated that the chemical shift trends toward more negative values with increasing

Si/(Si+Al) and Na/(Na+K) for compositions in the system NaAISi,Og-KAISi,O,-SiO,,

indicative of p a t e r shielding of the nucleus (Oestrike et al., 1985; Maekawa et al., 1997).

A sirnilar trend has been reported for melts with v;isring degrees of polymerization (Gee et

al., 1997; Xue and Stebbins, 1993; Maekawa et al,, 1997). The results of these studies

indicate a trend toward more negative chernical shifts as the mean number of non-bridghg

oxygen atorns per tetrahedron (NBO/T) is decreased. This paper reports the results of a

series of single-pulse * ~ a NMR MAS measurernents of a suite of Na-alumhosilicate glasses

with Na/Al=2 and varying SiOt content in order to identiw any shifi in peak position and

relate such shifts to structural changes in the glass network.

4.3 Experimental Procedure

Glasses were prepared using reagent-grade SiO, gel, A1203 and Na2C0,. A staRii1g

composition of 2N%O*Al,03-3Si02 was prepared and decarbonated at 1 atm and 750°C for

24 hours, then fused at 1 atm and 1450°C for 1 hour in a Pt crucible- From this base glass,

6 samples were prepared having the composition 2Na,O-Ai@,-nSiO,, with n = 3,4,5.7.9

and 11. For each composition (including the base composition), the sample was fbsed a

second time at 1450°C for 1 hour. AU samples were examined opticaiiy to ensure that no

crystals were present.

= ~ a NMR MAS spectra were acquired at a magnetic field strength of 7.1 T for a

Larmor frequency (vJ of 79.25 MHz with a spin rate of 5.5 kHz. Utilizing a Chemagnetics

CMX-300 instrument, analyses comprised 1000 acquisitions using a pulse width of 0.5 ps

and a recycle time of 200 m. The s h i f t in peak position measured is reported in parts per

miilion difierence in the nuclear resonance fiequency from that of solid NaCI. Typical error

in peak position is estimated to be about 0.1 ppm.

4.4 Resulr

The =Na NMR MAS spectra for ai i samples are presented in Fig 4.1. Each spectrum

consists of a single broad peak and spinning sidebands which overlap the main peak. The

peak positions. provided for each spectmm, become more negative as the S i 4 content in the

g l a s increases and the average number of non-bndging oxygen atoms per tetrahedron

(NBO/T) decreases. In Fig. 4.2, shifts in peak position are plotted as a hinction of (N'Born.

indicating that the trend is not continuous. Rather. the degee of shielding becomes constant

above NBOlI' - 0.27. indicating that continued depolymerization exerts less effect on the net

magnetization of the nucleus.

Fig. 4.1. =Na NMR MAS spectra of six Na-aluminosilicate glasses with constant Na/N and

vaqing SiO, content. The chernical shift relative to crystalline NaCl is indicated for each

spectrum.

O 400 PP*

Fig. 4.2. Plot of peak position as a hnction of the average number of non-bndging oxygen

atoms per tetrahedron (NBO/T). A negative trend in peak position is observed for glasses

with NBO/T 5 0.27. No shifi occurs for less polymerked glasses.

4.5 Discussion

The trend towards more negative chemical shifts in the NMR MAS spectra of =Na

with decreasing NBO/T results from greater s hielding of the nucleus from the appiied

magnetic field by the local electrons which generate a magnetic field of their own. The

magnitude of this shielding is a function of local structure as different coordination

geometries distoa electronic orbitals in different ways. which, in turn, affects the magnetic

field generated by the electmns in these orbitais as felt by the nucleus (Stebbins, 1995). " ~ a

NMR MAS spectra of Na-silicate and duminosiiicate crystdine phases (Maekawa er al.,

1997) have shown that the average Na-O distance is the most important factor affecting the

chemical shifi. Table 4.1 presents =Na chemical shifrs for several Na-silicate minerais as

weli as coordination numbers and mean Na-O distance. These data cleariy indicate that such

a relationship exists.

It should be noted, however, that the chemical shift does not change in a h e a r

fashion with increasing bond lenad. A plot of mean Na-O distance vs. chernical shift (Fig.

4.3) for several crystaliine Na-silicates is approximately iinear and negative in dope up to

a Na-O bond length of about 2.7 A. Beyond this point, the dope decreases slightly.

indicating that increasing bond length becomes Iess effective as a means of increasing the

degree of magnetic shielding.

The correlation between increasing polymensation and greater shielding of the

nucleus observed in this study for @asses has also been reported for melts. ResuLts from

severai studies of Na-silicate and aluminosilicate melts are piotted in Fig. 4.4. Above NBOn

= 1, the chemicai shifts become more positive very pdualiy, suggesting little change in the

local environment of Na. For more polymerized melts, the trend is still that of greater

shielding with decreasing NBOfT, but the slope is steeper. The implication is that as the melt

approaches full polymerization. the Na-O bond length becomes longer.

The implications of the trend in Fig. 4.4 for melts are somewhat counter-intuitive

phase ai (C0n.r Original C.N. (Na) d (W) NMR Structura mndud rdemna nfnma

l

NqBaSi,O, 1 Na@)

jadeite NaAtSi,O,

nep heline N~~(N&)ddAISiO,

Na site K site

albite NaAISi,O,

microciine

~ ~ ~ 1 7 U A ~ S ~ 3 O 8 sodalite Na.AI,Si,O,,CI anhydrous sodalite NaJAISiO J, natrolite NaJI,Si,O,,-2H,O tourmaline

NaMg&i,(B03,Si,O,,(OH), N a,LlYS &O,, Na(1) -1 7.95r0.5 -1 7.95

Na(2) -5.4=1 -5-4 narsarsukite

1 M NaCl 1 M NaCl

1 M NaCl 1 M NaCI 1M NaCl 1 M NaCl 1 M NaCl 1 M NaCl

1 M NaCl 1 M NaCl

1 M NaCl 1 M NaCl

1 M NaCl 3M NaCl 3M NaCl

1 M NaCl

1M NaCl

solid NaCl

1M NaCI 4+2H20

I ~k~ i s i . ~ . , -4.01 -4.0 1M NaCl 7 2-536 21 'Chernical shifts relative to IM NaCI standard; 0.2 ppm added to resutts using 3M NaCI (Templeman and van Geet, 1972); 7.3 pprn added to results using solid NaCl (Kundla et al, 198 1) bDat.a are simulations o f spectm rneasured at 200°C

Table 4.1. =Na chemicai shi f ts and structural data for c r y s h e Na-silicate phases. Coordination numbers indicated for Na represent bonding with oxygen atoms only. in some instances, Na may ais0 bond with other Ligands as noted. References are as follows: 1. Maekawa et al. (1997); 2. Xue and Stebbins, (1993); 3. Stebbins et al. (1989); 4. Phillips er al. (1988); 5. Kirkpatrick et ai. (1985); 6. Kundla et al. (198 1); 7. Buhl et al. (1988); 8. Pant and Cruickshank, (1968); 9. Pant (1968); 10. McDonald and Cruickshank, (1967); 11. Gunawardane et al. (1973); 12. Prewitt and Burnharn, (1966); 13. Daliase and Thomas, (1978); 14. Harlow and Brown, (1980); 15. Da1 Negro et al. (1978; 1980); 16. Barth (1932); 17. Felsche et al. (1986); 18. Pechar et al. (1983); 19. Buerger et al. (1962); 20. Gunawardane et al. ( 1982); 2 1. Peacor and Buerger, (1962)

Fig. 4.3. Mean Na-O distance vs. = ~ a chemicd shifi for Na-bearing silicate crystabe

compounds. Al1 chemical shifts indicated are relative to 1M NaCl solution, either directly

measured or corrected.

2.20 2-40 2.60 2.80 3-00 3.20

Mean Na-O distance (A)

Fig. 4.4 Plot of =Na chernicd shift vs. number of non-bndging oxygen atoms per

tetrahedron (NBO/T) for several Na-silicate and Na-duminosilicate meIts.

Na-silicate melts (Stebbins et al., 1985) Na-silicate melts (Maekawa et a1.,1997)

O Na-sil icate melts (Stebbins et a1.,1985) A Na&-silicate melts (Stebbins and Faman, 1992) 0 Na,Al-silicate melts (Maekawa et al., 1997)

Since the "Na chemical shifl becomes more negative as the SiO2 content is increased and the

melt structure becomes more polymerized, the mean Na-O distance would appear to be

increasing. Yet, data from Riebling (1966) indicate that for any constant Na/AI ratio. the

molar volume of a melt decreases with increasing mole fraction of SiO,. Unfortunately. Na-

O bond lena& have only been measured for a few glasses and melts. The few results which

have been reported are summarized in Table 4.2. Clearly, these Limited data cannot provide

much insight into the trend observed in Fig. 4.2. Indeed, the only snidy which compared

glasses of vruying composition (McKeown et aL, 1985) indicated virh~dly no change in Na-

O bond lene&.

It should be noted that the shifts in peak position of the glasses andyzed in this study

are si,pificantly more negative than those of the melts in Fig. 4.4, even taking into account

a correction of +7.3 ppm for the soiid NaCl standard (Kundla et al. 1981) used in this study.

This difference results from second order quadnipolar shifts in the specua of the glasses.

The measured shift. 6,,, of a quadrupolar nucleus such as =Na is actuaily the sum of the

isotropic chemical shift, 4,. and the second order quadrupolar coupiing, 6,. The latter can

be determined by using the equation for the isotropic average of the second order

quadmpolar coupling (Youngman and Zwanziger, 1994):

With the quadrupolar coupling constant, QCC = 3 MHz and the asymrnetry parameter, qQ

= O (Xue and Stebbins, 1993; George and Stebbins, 1996), and a Larmor fi-equency (vJ of

79.25 MHz, the contribution of quadmpolar effects is about 35 ppm. When this contribution

is added, gNa chemical shifts measured in this study average about 15 to 20 ppm higher than

those of compositionaiiy-simïiar rnelts (Maekawa et al.. 1997; Xue and Stebbins, 1993;

87

Composition T s Method A N . (Nap r(Na-O) Reïerence

298 298 1 273 298 298

NaAlSiO, 298 Na&4I,,Si,O, 298 Na, ,AI,,Si,,,O, 298

J I 298

X-ray rdf neutron rdf X-ray rdf

Na-EXAFS Na-EXAFS Na-EXAFS Na-EXAFS Na -WFS Na-EXAFS

Na,CaSi,O.- 298 Na-EXAFS 'Coordination numbers rounded off to nearest integer.

Table 4.2. Na coordination numbers and Na-O bond lengths for NaAl-silicate @asses.

References cited are as foiiows: 1. Waseda and Suito, (1977); 2. Greaves et al., (1981); 3. McKeown et al., (1985).

Stebbins et al., 1985). A similar shifi to more positive 6- was noted for Na-silicate glasses

at ambient temperature as compared to melts of the same composition by George and

Stebbins (1996) for both static and MAS NMR. They observed that the chemical shift

became increasingiy more negative as the glasses were heated, which they interpreted to

result from thermal vibration of Na within its coordination sheli, or alternatively, increases

in the size of the Na site. Quadrupolar effects become unimportant for melts as al1

transitions become fully averaged.

4.6 Conclusions

The results of this snidy agree with those of several previous NMR studies of

alkali silicate and duminosilicate glasses and melts confinning that there is a trend toward

greater shïelding of the nucleus with increasing polymenzation. This study, however, has

reveaied that for aikali duminosilicate glasses, that the trend appiies ody to compositions

with NBO/Ts 0.27. For less polymerized glasses, the zNa chemical shîft and, by inference,

the average Na-O bond length remain approximately constant,

Chapter 5

Controls of Alkali Zirconosilicate Crystallization

in the System: Na20-K20-AI,O,-Si0,-Zr02-H20*F

at Temperatures between 575 and .ïOO O C

5.1 Introduction

Although zircon is the most common zirconium-bearing minerai, there are a

bewildering array of alkali and aikaiine-earth Wconosilicates associated with felsic,

perdkaline rocks. Nearly two dozen of these minerals have been reported to date (Marr and

Wood, 1992; Birkett et aL, 1992). Among the most widely reported of these are elpidite

(Na@i,O,,*SH,O), catapleiite (Na&Si,0~2H20), vlasovite (Na2ZrSi JO, ,), wadeite

(KJr!Si,O,), dalyite (&2rS&O,,) and eudialyte (Ng Ca ZrSi, q, (OEQ ). The potential exists

for using these minerals as petrogenetic indicators, but only M t e d experimental data exist

on the physico-chernical conditions of crystaiiization of these minerals.

An important phase equilibrium study in the N@-Si0,-m-H20&1 system by

Cume and Zaleski (1985) constrained temperature of the reaction: elpidite = vlasovite + quartz + H20 for pressures of 0.5 to 2.0 kbars. In the same study, the authors reported that

parakeldyshite (Na,ZrSi,O,) + quartz was produced in synthesis experiments at temperatures

less than 700°C. when NaCl was used instead of N a m , as the source of sodium in the

starting materids. The authors attributed this to an effect of Cl- on the activity of silica or

H20. Whiie these were syntheses, not reversed experiments, and therefore, the assemblage

observed mayhave been metastable, the results suggest that halogens may- play a significant

role in the speciation of zirconosilicate minerals.

The aim of the present snidy is to define the role of halogens in the speciation of

akali zirconosiLicate complexes in feisic, perdkaline melts. was chosen for study

as it is one of the most important halogens in geological systems. Furthemore, the

complexation behaviour of fluorine in sodium aluminosilicate melts has been weU-

characterized (Mysen and Virgo, 1985a,b), thus allowing more meaningful interpretation of

results.

5.2 Experimental Procedure

The starting matenals were @asses with compositions as presented in Table 5.1. The

glasses were prepared from reagent-grade SiO,, Ai,O,, Nago , , &CO, and NaF powders.

The mixtures were decarbonated at 750°C in air for 48 &ours, ground. to a fine powder and

melted at 1550°C in air for one hour. The resulting glass was ground and fused a second

time. After the glasses were prepared, they were inspected optically for the presence of

crystals. Those samples with more than approxirnately one percent crystailine material were

discarded. ï h e samples were analyzed using an electron microprobe to determine the degree

of homogeneity and aikaii loss. Zirconium was added to the glasses as reagent-grade 21-0,.

Initial experiments involved the addition of 1.0 W.% Zr& to each composition. The glasses

were then fused a third time, analyzed by eiectron microprobe, ground and stored in a dryhg

oven until used.

Experhents were performed using cold-sed vessels (Tuttle, 1949). The cold-seal

apparatus in the Department of Earth and Planetary Sciences at McGill utilizes argon as a

pressure medium and can be employed up to pressures of 1.5 kbar. Pressure is monitored

using Astra Bourdon-tube gauges calibrated against a large dia1 Heise gauge which was

calibrated at the factory. Pressure measurements are considered accurate to +cO bars.

An Omega mode1 650 themocouple themorneter with chromel-alumel

thermocouples was used to determine temperature near the sample to within H C . The

thermocouples were calibrated against the melting points of H20, Sb, Bi and NaCl.

The experiments involved sealing 0.010 g of sample and 2.0 to 2.5 pL of distilled

water in a gold capsule. The capsules were welded and the weid was inspected opticaiiy. The

capsules were heated at 1 10°C for 15 minutes and reweighed to detect any fluid loss. The

capsules that underwent no mass loss were placed in the cold-seal vessels dong with a

ceramic filler rod. The experiments were run for one week at 1 .O kbar water pressure and

tempentures indicated in Table 5.2.

SiO, 54.70 66.37 77.67 53. 17 66.26 78.50

40, 22.97 17.17 1 1.88 24.3 1 17.51 11.70

N-O 1 1.38 8.18 5 -40 1 1.32 8.25 5.36

&O 12.46 9 -49 6.57 12.56 9.39 5-36

F - - 0.49 0.23 0.34 - O = F - - - -0.2 1 -0.10 -0.15

Tot al 10 1.5 1 101.21 10 1.52 101.64 10 1.54 101.1 1

Alk/Al I .40 1.38 1.35 1.33 1.36 1 .25

Na/K 1.39 1.3 1 1.25 1.37 1.34 1-52

Table 5.1: EIectron microprobe analyses of starting materials for fluorine-free and fluorine-bearing experiments. Results represent the average of ten andyses. Compositions 1, 2 and 3 correspond to a leucitite, alkali feldspar syenite and a grnite respectively.

After the experiments were compIeted, the reaction vesseIs were airquenched,

attaining subsolidus tempentures in under 30 seconds. The capsules were then removed h m

the vessels and the charges were exarnined. Since these experiments were performed

below the d q granite solidus, the presence of significant amounts of residuai glass in the run

products was taken as good evidence that water vapour did not escape' fiom the capsule

during the experiments. The capsules were also reweighed to detect any loss of volatiles.

Oxygen fugacity was not explicitiy controlled because no redox reactions were

invoIved in the experiments. However, fO, is expected to be within a log unit or two of the

Ni/NiO boundary as a result of the buffering effects of the reaction vessels which were made

of a nickel-bearing alloy (Chou, 1987).

The run products were malyzed using an electron microprobe with an accelerating

voltage of 15 kV and a beam current of 20 nA. Analyses of gisses was achieved by using

a wide beam (10 to 15 pm) to minimize sample damage. Albite (Na, Al, Si), orthoclase (K),

fluorite (F) and zircon (Zr) were used as standards,

5.3 Results

Wadeite (&ZrSi,O,) was obsewed in the run products of the haiogen-fiee

experiments for both the leucitite and the alkali feldspar syenite. The observed assemblages

are indicated in Table 5.2. The granitic composition yielded no zirconium-bearing minerais,

except for smdl aggregates of ZrO, grains, which are believed to be undissolved starting

material. Zircon was not observed in any of the run products.

Microprobe analysis of the feldspars, feIdspathoids and alkali zirconosilicates in the

run products of the fluorine-free experiments indicate very Little solid solution between Na

and K except at the highest temperatures. Even for the experiments at 700°C, the alkali

zirconosilicates and feldspathoids retain compositions close to that of ideal end-members.

Some representative mineral analyses are provided in Appendix 1.

590°C ne + Ksp + wd + glass

640°C ne + Ksp + wd + g l a s

ne + Ic + wd + glass

Ksp + wd + glass

Ksp + wd + glass

Ksp + qz + glass glas -

ne t Ksp + wd + glass

Ic+ Ksp+wd+@;tss

Ic + wd + glas

Ksp + pk* + &us

Ksp + glass

Ksp + glas

Ksp + qz + glas

1 3F 700°C Symbols used as foIlows: ne = nepheline; Ic = Ieucite; Ksp = K-feIdspar. qz = q u m ; wd = wadeite; pk = parakeldyshite; *(my be metastable)

Table 5.2: Observed assemblages in run products of expenments performed in this study. Al1 expenments were run for one week at 1 kbar pressure.

The addition of fluorine to the glasses resulted in the appearance of a Na-

zirconosïiicate phase in the aikali feldspar syenite composition, but only for the lowest

temperature expriment. This phase had the chernical composition of parakeldyshite

(N+ZrSizQ), but only appeared at a temperature si@cantly below that reported by Currie

and Zaleski (1985) for the appearance of parakeldyshite in their experiments. Furthemore,

this phase lacked rationai crystai boundaries and may have k e n metastable. Aside fiom this

one difference, the observed assemblages remain unchanged from the fl uorine-fkee run

products. As was the case with the fluorine-free experiments, most of the 2r0, remained

undissolved in the granitic melt. Zircon was not observed in any of the fluorine-bearing run

products.

Once again, significant solid solution between Na and K-rich end-members was not

evident at lower temperatures and not even in the high temperature experiments for the

zirconosiIicates and feldspathoids.

5.4 Discussion

Silica activity appears to have an effect on the solubiiity of Zr-bearing phases in

peralkdine melts. Analyses of the residud @ass from experimental run products of

composition IA at 640°C and 2A at 600°C (Appendix 2) indicate nearly an order of

magnitude increase in the solubility of wadeite in the more silica-nch melt. A similar trend

was observed in the solubility experiments reported in Chapter 2 for SiO, contents below

approximately 57 wt.%, although the degree of enrichment was more modest. Yet, Watson

(1979) reported that the soIubility of zircon in peralkaline grmitic mefts is insensitive to SiO,

content. Therefore. it appean that the solubility of wadeite and possibly other alkali

zirconosilicates is more sensitive to silica activity in melts than is that of Wcon.

The presence of wadeite as the saturating phase in the run products of experiments

using composition 2A is interesthg, as zircon was the saturating phase for aU experhents

with S i 4 above 55 W.% in Chapter 2. Since these latter experiments were ai l performed

at 800°C, this indicates that increasing the siiica activity depresses the upper thermal stabïiity

Iimit of wadeite in favour of zircon. This Iends further support to the conclusion from

Chapter 2 that peralkaline melts with high initial 210, content would be expected to

crystallize zircon rather than alkali zirconosilicates,

5.5 Conclusions

Severai concIosions may be drawn fiom the results of these experiments. First of aii,

the solubility of wadeite is greatly affected by silica activity in the melt, increasing

dramatically with the SiO, content of the melt. Secondly, wadeite is the saturathg Zr-

bearing phase across a wide range of thermai and chernical conditions and may be the fmt

alkali zirconosilicate to crystallize fiom many peralkaline melts in nature. Findiy, the upper

thermal stability Limit of wadeite is depressed as the silica content in the melt increases.

Therefore, hi$-silica, peralkaiine melts may crystallize zircon rather than an alkali

zirconosilicate, particularly if initial Z a content is hi@.

Chapter 6

Conclusions

6.1 Conclusions

The solubility of Zr-bearing xnherals has been previously reported to be Iargely

unsected by SiO, content in peralkaline grrinitic melts (Watson, 1979). However, the

results of this study indicate that for syenitic compositions, does become an important

control on the solubility of zircon and wadeite (K,ZrSi,O,). For -0-saturated peralkaline

melts at 800°C and 1 kbar, the solubiiity of zircon reaches a maximum of approximately 4

W.% Zr02 at a SiO, content of about 57 wt.%. For compositions with less silica, the

saturating phase becomes wadeite and the saturation Ievet drops sharply with decreasing

SiO,. Similar behaviour is observed in experiments with about 1 wt.% F, with two important

exceptions. The solubility reaches a maximum at about the same SiO, as that of the F-free

experiments, but the satuntion concentration of ZrO, is approximately 3.5 wt.%. Aiso, the

saturating phase for the Iower siiica compositions is ZrO, nther than wadeite. Experiments

with the same compositions at 650°C do crystdlize wadeite, however, indicating that the

addition of F Iowers the upper thermal stability limit of wadeite. The addition of small

amounts of Cl (less than 0.5 wt.%) to perdkaline granitic and syenitic melts lowers the

solubility of Zr-bearing minerais and no peak appears even though the transition from zircon

saturation to ZrOz saturation SUU occurs at about the sarne SiO, content. The saturation

concentration for al1 Cl-bearing melts in this study was about 2 to 2.2 wt-% ZrO,.

Raman spectra of aikali duminosilicate and Zr-bearing akali duminosilicate @asses

show no detectable change as small concentrations of CI are added. However, some

differences are detected in the spectra of Ti-bearing alkali aluminosilicate glasses with and

without added Cl. The spectrum of the Cl-free glas contains a strong, asymmetric peak at

900 cm-' which is associated with Iq~i-O stretching vibrations. This peak is shifted to higher

frequency and becomes more symmetric with the addition of 0.3 wt.% Cl. Deconvolution

of the high-frequency region in each spectra suggests that the observed differences are the

result of a contribution from a peak at 945 cm-'. This peak is believed to be the result of Ti-

O vibrations in fuiiy-polymerized titarate tetrahedra. It is proposed that the addition of Cl

destabilizes five-fold coordinated Ti in favour of tetnhedral coordination as a result of

cornpetition between Cl and titanate groups for alkali ions.

NMR MAS analyses of a suite of Na-aluminosilicate @asses with NalAl= 2 and

varying SiOz content have revealed a trend toward more negative chemicai shifts (greater

shielding of the nucleus) as the average number of non-bndging oxygen atorns per

tetrahedron (NBOm decreases. This trend is observed only for glasses with NBO/T a 0.27.

For more polymerized glasses no change in chernicd shifi is evident. Negative =Na chernical

shifts for crystaiiine Na-silicates are indicative of p a t e r Na-O bond length (Maekawa et al.

1997). This suggests that for Na-aluminosilicate glasses with NBOlT i 0.27, Na-O bond

lengths increase as full polymerization is approached.

6.2 Contributions to Knowledge

1. The solubilities of Zr-bearing rninerals have been detemiined for syenitic,

peraikaline melts. Prior to the completion of this thesis, th is data was only available for

grmitic compositions despite the fact that many Zr-rïch plutons such as Mont S t Hilaire and

Lovozero are syenitic.

2. It has k e n demonstrated that for H20-satunted. perdkaline melts, the addition of

F has either little effect or a detrimental effect on the solubility of Zr-bearing minerais

depending on bulk composition. This indicates that F affects the dissolution mechanism of

Zr in peraikdine melts differently than in metduminous melts since previous work has

indicated that F enhances the solubility of zircon in metaluminous, granitic melts (Keppler,

1993).

3. It has also k e n demonstrated that Cl has a detrimental effect on the solubility of

Zr-bearing minerals regardless of melt composition.

4. Raman spectra of Ti-bearing alkali aluminosilicate glasses have indicated that

smaii amounts of Cl result in a signifcant change in a peak which has k e n associated with

Ti-O "stretching" vibrations. A mode1 has been proposed suggesting that cornpetition

between alkalis and five-fold coordinated Ti results in some Ti becoming tetrahedrally-

coordinated. This has impIications for the stability of &di titanosilicate minerals such as

narsarsukite [N+(Ti,Fe)Si,O,,(OH,F)j and aenigmatite W F ~ ; ~ T ~ S & O ~ ] wwhh may

crystallïze from an alkali titanosilicate melt cornplex based on five-fold coordinated Ti.

5. U ~ a NMR MAS spectra of Na-aiuminosilicate glasses have indicated that for

compositions with NBO/T < 0.27, the chernicd shift becomes increasingly negative with

increasing polymerization, implying that Na-O bond length increases. Although previous

studies of Na-bearing crystalline phases have established the relationship between Na-O

bond length and chemical shift, this study is the first to systematically measure chemical shift

for highly-polymerized alkali aluminosilicate glasses, which more closely approximate

naturai melts.

6.3 Recommendations for Future Woïk

Much more phase equiiibria work must be done before the alkali zirconosilkate

minerais will be usehl as petrogenetic indicaton. Literally dozens of possible reactions

remain to be constrained. Additionally, calorimeuic studies of some of the more common

alkali zirconosilicates could provide valuable thermodynamic data which would allow

modelling of some kineticdly-unfavounble reactions.

The role of CI in granitic and syenitic melts needs to be explored more thoroughly.

Although it is known that CI affects the physical properties of silicate melts very differently

than the other common halogen F, theories on its solution mechanism are purely speculaîive.

Spectroscopie studies usuig X-ray absorption techniques or U ~ a NMR MAS studies of Cl-

bearing Na-silicate basses rnay provide some insight. Furthemore, while the solubility of

Cl in granitic melts has k e n studied (Webster, 1992), it has not been determined for silica-

undersaturated melts. Given that many nepheline syenites such as that at Mont St. Hilaire

have elevated Cl concentrations (i.e. normative sodalite), a systematic study is called for.

References

ALBERTO, H.V., AYRES DE CAMPOS, N. and MYSEN, B.O. (1995) The structural role

of titanium in silicate glasses: a Raman study of the system Cao-Si02-TiO,. Phys.

Chem.Glasses, 36, 1 14- 122.

BAKER, D.R. (1993) The effect of F and Cl on the interdifision of peralkaline

intemediate and siIicic meits. Am. Mineral., 78,3 16-324.

BAKER, D.R. and VAILLANCOURT, J. (1995) Cl in albite melt: A volatile which increases

melt viscosity. (abs.), Eos Tram., 76, AGU Fa11 Meeting, San Francisco, F646.

BARTH, T.F.W. (1932) The structures of the minerals of the sodaiite family. Z Knstaiiogr.,

83,405-4 14.

BAUSSY, G., CARUBA, R., BAUMER, A. and TURCO, G. (1974) Mïnéralogie

expérimentale dans le système ZQ-Si02-Na,O-H,O. Corrélations pétrogénétiques. Bull.

Soc. fr. Minéral. Cristal., 97,433-443.

BIRKEïT, TC., MILLER, R.R., ROBERTS, AC. and MARLANO, A.N. (1992) Zirconium-

bearinp minerais of the Strange Lake intrusive complex, Quebec-Labrador. Cm. Minerai.,

30, 191-205.

BOGGS, R.C. and GHOSE. S. ( 1985) Georgechaoite NaKZrSi,O,-2H,O, a new mineral

species from Wind Mountain, New Mexico. Cm. Mineral., 23, 1-4.

BRAWER, S.A. and WHITE, W.B. (1975) Raman spectroscopic investigation of the

structure of silicate glasses. 1. the binary silicate glasses. J. Chem. Phys., 63,2421-2432.

B UERGER, M.J., B U R m M , C.H. and PEACOR, D.R. (1962) Assessrnent of the several

structures proposed for tourmaline. Acta Crystallogr., 15,583-590.

BUHL, J.-C., ENGLEHARDT. G., WLSCHE, J. and LUGER, S. (1988) "Na MAS-NMR

and 'H MAS-NMR studies in the hydro-sodalite system. Ber. Bunsenges. Phys. Chem., 92,

176-181.

CANMLLO, E., ROSSI, G. and UNGARE?Ti, L. (1973) The crystal structure of elpidite

Am. Mineral., 58, 106- 109.

CARUBA, R. (1975) Étude expérimentale de la cristallochimie, de la motphologie, de la

stabilité et de la génése du zircon et des zirconosilicates en vue d'applications

pétrogénétiques. Ph.D. thesis, University of Nice, 324 p.

CARUBA, R., BAUMER, A. and TURCO, G.H. ( 1973) Synthèse et étude morphologique,

radiocrystallographique et spectromeuique IR de la wadéite K,ZrSi,O,; domaine de

104

formation; remplacements isomorphiques; application à la géochimie du zirconium. Chem.

Geol., 11,49-69.

CHAI, B.H.T. and MROCZKOWS KI, S. ( 1978) Synthesis of rare-earth carbonates under

hydrothermal conditions. J. Crysr. Growth, 44, 84-97.

CHAO, G.Y. (1967) Leucophanite, eIpidite, and narsarsukite from the Desourdy Quarry,

mont St. Hilaire, Quebec- [abs.]. Cm. Mineral., 9,286.

CHAO, G.Y. (1985) The crystal structure of gaidomayite Na,ZrSi,O,-2H,O. Can. Mineral.,

23, 11-15.

CHAO, G.Y. and WATKINSON, D.H- (1 974) Gaidomayite, Na,ZrSi,O,-2H20, a new

minera1 from Mont S t. Hilaire, Quebec. Can. Mineral., 12, 3 16-3 19.

CHAO, G.Y., HARRIS, D.C., HOUNSLOW, A.W., MANDARINO, J.A. .and PERRAULT,

G. (1967): Minerals from the nepheline syenite, Mont St. Hilaire, Quebec. Can. Mineral.,

9, 109-123.

CHAO, G.Y., ROWLAND, J.F. and CHEN, T.T. (1973) The crystal structure of catapleiite

[abs.]. In Abstracts with Programs, GSA ,Qnnuaf Meeting, Dallas, 572.

CHAO, GY., WATKINSON, D.H. and C m , T.T. (1974) Hilairite, Na,ZrSi,O,-3&O, a

new mineral from Mont St, Hilaire, Quebec. Cm. Mineral., 12,237-240.

CHAO, GY., CHEN, T.T. and BAKER, J. (1980) Petarasite, a new hydrated sodium

zirconosilicate hydroxychiorosilicate minerai from Mont St-Kifah, Quebec. C a Mineral,

18,497-502.

CHOU, LM. (1987) Oxygen buffer and hydrogen sensor techniques at eievated pressures and

temperatures., In Hvdrothermal Exuenmental Techniaues. Barnes, H.L. and Ulmer, G.C.,

eds., John Wiiey, New York, 6 1-99.

CHRISTENSEN, AN. (1973) Hydrothermal preparation of rare earth hydroxy-carbonates.

The crystal structure of NdOHCO,. Acta Chem. Scand., 27,2973-2982.

CHRISTOPHE-MICHEL-LÉVY, M. ( 196 1) Reproduction artificielle de quelques minéraux

riches en zirconium (zircon, eudialyte, catapléite, elpidite); comparaison avec leurs

conditions natureIIes de formation. Btïli. Soc. fr. Minéral. Cristal., 84, 265-269.

CORMIER, L., GASKELL, P.H., CALAS, G., ZHAO, J. and SOPER, A.K. (1997) The

titanium environment in a potassium silicate glass measured by neutron scattering with

isotopic substitution. Physica B, 234,393-395.

CURRIE, KL. (1985) An unusual peraikaline granite near Lac Brisson. Quebec-Labrador.

Geologicai Survey of Canada Report 85-l A, p.73-90.

CURRIE. K.L. and ZALESKI. E. ( 1985) The relative stability of elpidite and vlasovite: A

P-T indicator for peralkaline rocks. C a n Minerai., 23, 577-582.

DALLASE, W.A. and THOMAS, W.M. (1978) The crystal chemistry of silica-rich, alkali-

deficient nepheline. Contrib. Mineral. Petrol., 66,3 11-3 18.

DAL NEGRO, A., DE PIERI, R., QUARENI, S. and TAYLOR, W.H. (1978) The crystal

structures of nine feldspars from the Adamello Massif (northern Italy). . Acta Cysfallogr.,

B34,2699-2707.

DAL NEGRO, A., DE PIERI[, R., QUAREM, S. and TAYLOR, W.H. (1980) The crystal

structures of nine feldspars from the Adamello Massif (northem Italy): erranim. Acta

Crystallogr., B36, 32 1 1.

DICKINSON, J.E. and HESS, P.C. (1985) Rutile solubility and titanium coordination in

silicate melts. Geochirn. Cosmochim. Acra, 49, 2289-2296.

ELLISON. A.J.G. and HESS, P.C. (1994) Raman study of potassium silicate glasses

containing Rb+, S P , Y3+, and Zr*: Implications for cation solution rnechanisms in

multicomponent silicate Iiquids. Geochin;. Cosmochim. Acta, 58, 1877- 1887.

FARGES, F. (1996) Does Zr-F "complexatiod' occur in magmas? Chem- GeoL, 127,253-

268.

FARGES, F. (1997) Coordination of Ti* in silicate glasses: A high-resolution XANES

spectroscopy study at the Ti K edge. Am, MineraL, 82, 3643.

FARGES, F., PONADER, C. W. and BROWN, Jr. G.E. ( 199 1) S tructurd environments of

incompatible elements in silicate glass/melt systems, 1. Zr at trace levels. Geochim

Cosmochim. Acta, 55, 1 563- 1 574.

FARGES, F., BROWN Jr., G.E., NAVROTSKY, A., GAN, H. and REHR, J.J. (1996)

Coordination chemistry of Ti(IV) in silicate glasses and melts: II. Glasses at arnbient

temperature and pressure. Geochim. Cosrnochim. Acta, 60, 3039-3053. -

FELSCHE, J., LUGER, S. and BAERLOCHER, C. (1986) Crystal structures of the hydro-

sodalite NaJNSiO,],*8H2O and of the anhydrous sodalite NaJAiSiO,],. Zeolires, 6, 367-

372.

FLEET, S.G. and CANN, J.R. (1967) Vlasovite: a second occurrence and a triclinic to

monocliaic inversion. Mineral. Mag., 36,233-24 1 .

IWKUMT, K., HAYAKAWA, I. and KOMIYAMA, T. (1990) Intensity of Raman bands in

silicate glasses. J. Non-Crystai- Soli&, 119,297-302.

FURUKAWA, T., FOX, K.E. and WHm, W.B. (198 1) Raman spectroscopie investigation

of the structure of silicate glasses. iXI. Raman intensities and structural uni& in sodium

silicate glasses. Chem. Phys., 153, 3226-3237.

GALEENER F L and MIEGELSEN, K. (198 1) Vibrational dynamics of "O-substituted

viueous SiO,. Phys. Rev. B, 23,5527-5530.

GEE, B., JANSSEN, M. and ECKERT, H. (1997) Local cation environments in muted alkali

silicate glasses studied by multinuclear single and double resonance ma@-angle spinning

NMR. J. Non-Crysr. Solids, 215,4 1-50.

GEORGE, A.M. and STEBBINS. J.F. (1996) Dynamics of Na in sodium aluminosilicate

glasses and Liquids. Phys. Chem. Minerais, 23,526-534.

GERASIMOVSKII, V.1. (1962) Keldyshite - a new minerai. Dokl. A M Nauk SSSR, 142.

9 16-9 18 (in Russian).

GHOSE, S. and THAKUR, P. (1985) The crystal structure of georgechaoite

NaKZrSi,0,?H20. Can. Mineral., 23,s-10.

GHOSE, S., WAN, C. and CHAO, G.Y. (1980) Petarasite Na,ZrZSi,O,,(CI,OH)-2w, a

zeolite-type zirconosilicate. Can Mineral., 18, 503-509.

GiTïINS, J., GASPARRINI, E.L. and FLEET, S.G. (1973) The occurrence of vfasovite in

Canada, Can. Mineral,, 12-2 1 1-214.

GREAVES, G.N., FONTADE, A., LAGARDE, P., RAOUX, D. and GURMAN, S.J. (198 1)

Locai structure of silicate @asses. Nature, 293, 6 1 1-6 16.

GUNAWARDANE, R.P., CRADWTCK, MIE. and GLASSER, L.S.D. (1973) Crystal

structure of Na,BaSiz06. J, Chem. Soc. (Dalton), 2397-2400.

GUNAWARDANE, R.F., HOWIE, R.A. and GLASSER, L.S.D. (1982) Structure of lithium

sodium yttrium silicate Na&iYSi60,,. Acru Crystallogr., B38, 1405-1408.

GWINN, R. and HESS, P.C. (1989) iron and titanium solution properties in peraluminous

and perallcaline rhyolitic liquids. Contrib. Mineral. Petrol., 101,326-338.

HALLER, W., BLACKBURN, D.H. and SIMMONS, J.H. (1974) Miscibility gaps in alkali-

silicate binaries - data and thermodynamic interpretation. J. Am. Cer. Soc., 57, 120- 126-

HARLOW, G.E. and BROWN Jr., G.E. (1980) Low albite: an X-ny and neutron diffraction

1 IO

study. Am. Minerai-, 65,986-995.

HARRIS, C. (1983) The petrology of lavas and associated plutonic inclusions of Ascension

Island. J. Petrol., 24,424470.

HASCHKE, J.M. (1975) The Ianthanum hydroxide fluoride carbonate system: The

preparation of synthetic bastnaesite. J. Solid Srate c'hem., 12, 1 15- 1 2 1.

HASCHKE, J.M. and E Y W G , L. (197 1) Hydrothermai equiiibria and crystal growth of rare

earth oxides, hydroxides, hydroxynitrates and hydroxycarbonates. Inorg- Chem., 10,2267-

2274.

FIENDERSON, G.S. and FLEET, ME. (1995) The stmcture of Ti-silicate glasses by micro-

Raman spectroscopy. Con. Mineral., 33,399-408.

HENDERSON, G.S. and FLEET, M.E. (1997) The structure of titaniurn silicate glasses

investigated by Si K-edge X-ny absorption spectroscopy. J. Non-Crysrai. Solids, 211,2 14-

HENSHAW, D.E. (1955) The structure of wadeite. Mineral. Mag., 30,585-595

HIRAYAMA, C. and CAMP, F.E. (1969) Effect of fluorine and chiorine on viscosity and

111

fining of soda-lime and potassium-barïum silicate glas. Glass Tech., 10, 123-127.

HUDON, P., BAKER, D.R. and TOFT, P.B. (1994) A hi&-temperature assembly for the

1 -9 1 -cm (U-in.) pistoncylinder appannis. Am. MineraL, 79, 145- 147.

ILYUSHIN, G.D., KHOMYAKOV, A.P., SHUMYATSKAYA, N.V., VORONKOV, A.A.,

NEVSECiT, N.N., ILYUKHIN, V-V. and BELOV, N.V. (1981) Crystal structure of a new

natural zirconosilicate, K4- i,O ,,-2H20. Dokf. Akad. Nauk SSSR, 256. 860-863 (in

Russian).

KEPPLER, H. (1993) Influence of fluorine on the e ~ c h m e n t of high field strength trace

elements in granitic rocks. Contrib. Mineral. Petrof., 1 1 4 , 4 7 9 4 8 .

KKALILOV, A.D., KHOLMYAKOV, A.P. DZHAFAROV, N. Kh., EYUBOVA, S.M. and

POVARENNYKH, (1975) X-ray investigation of keldyshite. DOW. Akad Nauk Ukrain. SSR,

Ser. B , 12, 1095-1098 (in Russian).

KHOMYAKOV, A.P. and VORONKOV, A.A. (1973) New zirconium silicates in the

Lovozero and Khibina massifs. Trudy Minerai. Muzeya Fersman, A M Nauk SSSR, 22,2 15-

2 17 (in Russian).

KHOMYAKOV, A.P., VORONKOV, A.A., LEBEDEVA, SI., BYKOV, V.P. and

112

YURKIVA, KY. ( i 974) Khibinskite, KJrSi,O,, a new minerai. Z i i s k i Vses. Minerai,

Obshch., 103, 1 10- 1 16 (in Russian).

KHOMYAKOV, A.P., VORONKOV, A.A., KAZAKOVA, M.E., VLASOVA, E.V. and

S MOLYANINOVA, N.N. ( 1975) Investigation on minerals of the keIdys hite group. T r d y

Mineral. Mrizeya Fersman, Akad. Nuuk SSSR, 24, 120-13 1 (in Russian)-

KHOMYAKOV, A.P., SEMENOV, EL, VORONKOV, A.A. and NECHELYUSTOV, G.N.

( l983a) Terskite, Na,ZrSi,O ,,*2H2O, a new mineral. Zàpiski Vses. Mineral. Obshch., 112,

226-232 (in Russian).

KHOMYAKOV, A.P., VORONKOV, A.A., KOBYASHEV, Yu-S. and POLEZHAEVA,

L.I. (1983b) Umbite and paraumbite, new potassium zirconosilicates from the Khibina

dkalic massif. Zapiski Vses. Mineral. Obshch., 112,46 1-469 (in Russiaa).

KHOMYAKOV, A.P., VORONKOV, A.A., POLEZHAEVA, L.L and SMOLYANINOVA,

N.N. (1983~) Kostylevite, K4Zr,S&0,,-2-0, a new mineral. ïizpiski Vses. Minerai. Obshch.,

112,469-474 (in Russian).

KIRKPATRICK, R J . (1988) MAS NMR spectroscopy of minerals and glasses. In

Spectroscopie Methods in Mineralogy and Geology, MSA Reviews in Mineruiogy, 18,341-

387.

KIRKPATRICK, R.J., KINSEY, R.A., SMlTH, KA., HENDERSON, D.M. and

OLDFIELD, E. (1985) High resolution solid-state sodium-23, aluminum-27, and silicon-29

nuciear magnetic resonance spectroscopic reconnaissance of alkali and plagioclase feldspars.

Am. Minerai,, 70, 106-123.

KOHN, S.C., DUPREE, R., MORTUZA, M.G. and HENDERSON, C.M.B. (1991) NMR

evidence for five- and six-coordinated aluminum fluoride complexes in F-bearing

duminosilicate glasses. Am. Minerai., 76,309-3 12.

KOVALENKO, V.L, KUZ'MIN, MA., ANTIPIN, V.S- and PETROV, L.L. (1971) Topaz-

bearing quartz keratophyre (ongonite), a new variety of subvolcanic igneous vein rock Doki.

Akad. Nartk SSSR, 199, 132-135.

KUBICKI, J.D. and SYKES, D. (1993) Molecular orbital calculations of vibrations in three-

membered aluminosiiicate rings. Phys. Chem. Minerais, 19,38 1-39 1 .

KUNDLA, E., SAMOSON, A. and -LlPPMAA, E. (1981) High-resolution NMR of

quadru polar nuclei in ro tating solids. Chem Phys. Lert., 83,229-232.

KUTTY, T.R.N., VISWANATHIAH, M.N. and TAREEN, J.A.K. (1978) Hydrothermal

equilibna in Nd20,-H,O-CO2 system. froc. Ind. Acad. Sci., 87A, 69-74.

LASAGA, A.C. (1982) Optimization of CNDO for molecular orbital caicdation on silicates.

Phys. Chem. Minerals, 8,3646.

LONDON, D., MORGAN, G.B. and HERVIG, R.L. (1989) Vapor-undersaturated

experiments with macusani @ass plus H20 at 200 MPa, and the internai differentiation of

granitic pegmatites. Contrib. Mineral* Perrol., 102, 1 - 17.

MAEKAWA, H., NAKAO, T., SHIMOKAWA, S. and YOKOKAWA, T. (1997)

Coordination of sodium ions in N M 0 , - S i 0 melts: a high tempennire =Na NMR study.

Phys. Chern. Minerals, 24,53-65.

MARR, R.A. and WOOD, S.A. (1992) Prefiminary petrogenetic grids for sodium and

calcium zirconosilicate minerals in felsic peralkaiine rocks: The SiO,-N-ZrO, and S i -

CaZrO, pseudobinq systems. Am. Mineral., 77,8 10-820.

MATSON, D.W., SHARMA, S.K. and PHILPOTTS, J.A. (1986) Raman spectra of some

tectosilicates and glasses dong the orthoclase-anorthite and nepheline-anorthite joins. Am.

Mineral., 71,694-704.

MAURICE, O.D. (1949) Transport and deposition of the non-sulphide vein minerals. V.

Zirconium minerais. Econ. Geol., 44,72 1-73 1.

McDONALD, W.S. and CRIJICKSHAPYK, D.W.J. (1967) A reinvestigation of the structure

of sodium metasilicate, N-SiO,. Acta Crystallogr., 22,37-43.

McDOWELL, S.D. and WYLLIE, P.J. (1972) Experirnental studies of igneous rock series:

the Kûngnât syenite compIex of southwest Greenland. J. Geol., 79, 173- 194.

IMcKEOWN, D.A., GALEENER, F.L. and BROWN Jr., GE- f 1984) Raman studies of Al

coordination in silica-rich sodium duminosilicate glasses and some reiated minenls. J. Non-

Crystal. Solidir, 68,36 1-378.

McKEOWN, D.A., WAYCHUNAS, G.A. and BROWN Jr., GE. (1985) EXAFS and

XANES study of the local coordination environment of sodium in a series of silica-rich

glasses and seIected minerals within the Na,0-A120,-Si02 system. J. Non-Crystal. Soiids,

74, 325-348.

McMILLAN, P.F. (1984a) Structural studies of silicate @asses and melts - applications and

limitations of Raman spectroscopy. Am. Mineral., 69,622-644.

McMILLAN, P.F. (1984b) A Raman spectroscopie study of giasses in the system Cao-MgO-

SiO,. Am. MineraL, 69,645-659.

McMILLAN, P.F. and HOFMEISTER. A.M. (1988) Infrared and Raman spectroscopy. In

116

Spectroscopie Methods in Mineralogy and G e o l o ~ , MSA Reviews in Mineralogy, 18,99-

159.

McMILLAN, P.F. and WOLF, G.H. (1995) Vibrational spectroscopy of silicate liquids. In

Structure, Dynamics and Properties of Silicate Melts. MSA Reviews in Mineralogy, 32, 247-

3 16.

McMILLAN, P.F., PIRIOU, B. and NAVROTSKY, A. (1982) A Raman spectroscopie snidy

of glasses dong the joins silica-calcium aluminate, silica-sodium aluminate and

silica-potassium aluminate. Geochim. Cosmochim. Acra, 46,202 1-2037.

McMILLAN, P.F., WOLF, G.H. and POE, B.T. (1992) Vibrationai spectroscopy of silicate

liquids and glasses. Chem. Geol., 96, 351-366.

iMETRICH, N. and RUTHERFORD, M.J. (1992) Experimentd study of chlorine behavior

in hydrous silicic melts. Geochim. Cosmochim. Acta, 56,607-6 16.

METZ, M.C. (197 1) The geology of the Snowbird Deposit, Mined Couaty, Montana M.Sc.

thesis, Washington State Univ.

LVIILLER. R.R. (1986) Geology of the Svange Lake aikalic complex and the associated Zr-Y-

Nb-Be-REE mineraiization. Nfid. Dep. Mines, Mineral Develop. Div., Rep. 86-1, 11-19.

MILLER, R.R. (1996): S tnictural and t e x t d evohtion of the S trange Lake peraikaline rare-

element (NYF) granitic pepatite, Quebec-Labrador. Cm. Mineral-, 34.349-37 1.

MYSEN, B.O. (1990a) Effect of pressure, temperature and bulk composition on the structure

and species distribution in depolymerised alkali aluminosilicate melts and quenched melts.

J. Geophys. Res., 95, 15733- 15744.

MYSEN, B.O. (1990b) The role of Al in depolymerised, penlkdine duminosilicate melts

in the systems LiO-A1203-SiO2, Na,O-Al,O,-SiO2 and &O-&03-Si02- Am Minerai-, 75.

120- 134.

MYSEN, B.O. (1995) Structural behavior of AIY in silicate melts: in-situ, high-temperature

measurements as a function of bulk cherrical composition. Geochim. Cosmochim. Acrn, 59.

455-474.

MYSEN, B.O. and FRANTZ, J.D. (1992) Raman spectroscopy of silicate melts at magmatic

temperatures: N-O-SiO2. &O-SiO, and Li,O-SiO, binary compositions in the temperature

range 25-1475°C. Chem. Geol., 96,321-332.

MYSEN, B.O. and FRANTZ, I.D. (1994) Structure of haplobasaltic liquids at magmatic

temperatures: In situ, high-temperature study of melts on the join Na2Si20,-NaflaAi)20,.

Geochim. Cosmochim. Acrn, 58, 17 1 1-1733.

MYSEN, B.O. and NEUVILLE, D. (1995) Effect of temperature and TiO, content on the

structure of Na_SiIO,-N-Ti,, melts and glasses. Geochim Cosmochim. Acta, 59,325-342.

MYSEN, B.O. and VIRGO, D. (1985a) Structure and properties of fluorine-bearing

aluminosilicate melts: the sys tem N+O-Al,O3-Siq-F at 1 ami. C o n f ~ b . Minerai. Petrol- , 91,

205-220.

MYSEN, B.O. and VIRGO, D. (1985b) interaction between fluorine and siiica in quenched

meks on the joins Sioz-ALF, and SiO2-NaF determined by Raman spectroscopy. Phys. Chem.

Minerals, 12.77-85.

MYSEN, B.O., VIRGO, D. and SEIFERT, F.A. (1982) The structure of silicate melts:

Implications for chernical and physical properties of naturai magma. Rev. Geophys. Space

Phys., 20, 353-383.

MYSEN, B.O., VRGO, D., DANCKWERTH, P., SEIFERT, F.A. and KUSHiRO, 1. (1983)

Influence of pressure on the structure of melts on the joins NaAI0,-SiO,, CaAl,O,-Si0 and

MgAi,O,-SiO2. Neues Jahr. Miner. Abh., 147,28 1-303.

MYSEN, B.O., VIRGO, D. and SEIFERT, F.A. (1985) Reiationships between properties and

structure of aluminosilicate melts. Am. Minerai., 70, 88- 105.

NASSIF, G.J. (1993): The Strange Lake Peralkaline Cornplex, Québec-Labrador: The

Hypersolvus-Subsolvus Granite Transition and Feldspar Mineralogy. M.Sc. thesis, M c G i

Univ.

NASSIF,G.J. and MARTIN, R.F. (1991) The hypersoIvus granite - subsolvus granite

transition at Strange Lake, Quebec - Labrador [abs.]. In Prognm with abstracts, GAC-MAC-

SEG joint annual meeting, Toronto, A89.

NAUMOV, V.B., KOVALENKO, V.I., KUZ'MIN, M.L, VLADYKIN, N.V. and IVANOV,

G.F. (197 1) Thennometrîc study of incliisions of melt in topaz from topaz-bearing quartz

keratophyre (ongonite). Dokl. Akad. Nauk SSSR, 199, 104- 106.

NERONOVA, N.N. and BELOV, N.V. (1963) The crystal structure of elpidite. Dimorphism

of dimetasilicate radicais. Dokl. Acad Nauk SSSR, 150,3 (In Russian).

OESTRKE, R., YANG, W.-H., KIRKPATRICK, R.J., ERVIG, R.L., NAVROTSKY, A.

and MONTEZ, B. (1987) High-resolution UNa, "~1, and "Si NMR spectroscopy of

framework aluminosilicate glasses. Geochim. Cosmochim. Acta, 51,2 199-2209.

PANT, A.K. (1968) A reconsideration of the crystal structure of b-Na2Si,0,. Acta

Crysrallogr., B24, 1077- 1 083.

PANT, A.K. and CRWCKSHANK, D.W.J. (1968) The crystal structure of a-Na&O,. Acta

Crysrallogr., B24, 13- 19.

PEACOR, D.R. and BUERGER, M.J. (1962) The detennination and refinernent of the

structure of narsarsukite, N+TiOSi,O,,. Am. Minerai., 47,539-556.

PECHAR, F., SC-, W. and WLL, G. (1983) A neutron diffraction refinement of the

crystal structure of naturai natrolite, Na&Si,O,,-2H20. 2. Kristallogr., 164, 19-24.

PHILLIPS, B.L., KIRKPATRICK, R.J. and HOVIS, G.L. (1988) " ~ l , 2 9 ~ i , and " ~ a MAS

NMR study of an Al, Si ordered alkali feldspar solid solution series. Phys. Chem. Minerais,

16,262-275.

P U T , D., BONHOMME, M.G., DUTHOU, J.L. and SHENEVOY, M. (1989)

Chronologie Rb& et K/Ar du granite perdkalin du lac B ~ s s o ~ , Labrador central, Nouveau-

Québec. Can. J. Earth Sci.. 26, 328-332.

PIOTROWSKI, LM. and EDGAR, A-D. (1970) Meking relations of undersaturated aikaline

rocks from South Greenland compared to those of Afiica and Canada. Medd Gr@nl., 181,

i -62.

PREWTï, C.T. and B-, C.W. (1966) The crystal structure of jadeite, NaAiSi,O,.

121

Am. Mineral., 51,956-975.

PRIDER, R.T. (1939) Some minenls from the Ieucite-rich rocks of the West Kimberley area,

Western Australia. Mineral. Mag., 25,373-387,

RAADE, G. and MLADECK, M.H. (1977) Parakeldyshite from Norway. Can. Minerai., 15,

102- 107.

RIEBLING, E.F. (1966) Structure of sodium duminosilicate meIts containing at least 50

mole% SiOl at 1500°C. J. Chern. Phys., 44,2857-2865.

SALVI, S. and WILLIAMS-JONES, A.E. (1990) The roIe of hydrothermal processes in the

granite-hosted Zr, Y, REE deposit at S trange Lake. Quebec/Labrador: Evidence from fluid

inclusions. Geochim. Cosmochim. Acta, 54,2403-241 8.

S A L K S. and WILLIAMS-JONES, A.E. (1992) Reduced orthornagrnatic C-O-H-N-NaCl

fluids in the Strange Lake rare-metal grmitic cornplex, Quebeckabrador, Canada. Eur. I.

MineraL, 4 , 1 155- 1 174.

S ALVI, S. and WILLIAMS-JONES, A.E. (1995) Zirconosilicate phase relations in the

Strange Lake (Lac Brisson) pluton, Quebec-Labrador, Canada Am. Mineral., 80, 103 1-1040.

SCHALLER, T., DINGWELL, D.B., KEPPLER, H., KNOILER, W., MERWJN, L. and

SEBALD, A. ( 1992) Fluorine in siiïcate giasses: A multinuclear nuclear magnetic resonance

study. Geochim. Cosrnochirn, Acta, 56,76) 1-707.

SEMENOV, E.I. (1972): Mineraiogy of the Lovozero Aikaline Massif. Akad. Nauk.,

Moscow, 305 .pp. (In Russian).

SEIFERT, F., MYSEN, B.O. and VIRGO, D. (1982) Three-dimensionai network structure

of quenched melts (glass) in the systems: SiO,NaAiO,, SiO,CaAl,O, and Si0,-MgAI,O,.

Am. Minerai., 67,696-7 17.

SHUNHLJA, K., ZONGGANG, W., ZHONGNEI, Z. and SONGYW, H. (1986) An

experimentai study of the conditions of the formation of bastnaesite. Acra Min. Sinica, 6,

155-160 (in Chinese).

STEBBINS, J.F. (1995) Dynamics and structure of silicate and oxide melts: nuclear magnetic

resonance studies. In Structure, Dynamics and Properties of Silicate Melts, MSA Reviews in

Mineralogy, 32, 19 1-239.

STEBBINS, J.F. and FARNAN, L (1992) Effects of hi& temperature on silicate liquid

structure: a multinuclear NMR study. Science, 255,586-589.

STEBBINS, J.F., MURDOCH, J.B., SCHNEIDER, E., CARMICHAEL, LS. and PINES, A.

( 1985) A hi&-temperature high-resolution NMR study of =Na, and "si in molten

silicates. Nature, 314, 250-252.

STEBBINS, J.F., FARNAN, I., WILLIAMS, E.H. and ROUX, J. (1989) Magic angle

spinning NMR observation of sodium site exchange in nepheline at 500°C. Phys. Chem

Minerais, 16,763-766.

STEBBINS, J.F., FARNAN, I., DAWO, N. and TZENG, S .-Y. ( 1 992) Solids and liquids

in the NaF-AE3--0, sysrem: a high-temperature NMR study. J. Am. Cerm. Soc., 75,

300 1-3006,

TAREEN, J.A.K and KüTIY, T.R.N. (1980) Hydrothermd phase equilibria in Ln,4-H,O-

CO2 systerns. 1. The lighter lanthanides. J. Cryst. Gmvth , 50,527-532.

TEMPLEMAN, G.J. and VAN GEET, A.L. (1972) Sodium magnetic resonance of aqueous

salt solutions. J. Am. Chem. Soc., 94,5578-5582.

TWONENKOVA, R.P. and KAZAKOVA, M.Y. (1962) Vlasovite --a new zirconium

silicate from the Lovozero massif. Proc. Acad. Sci. USSR, 137.45 1-452.

TWTLE, O.F. (1949) Two pressure vessels for silicate-water studies. GeoL Soc. Am. Bull.,

124

VAN TASSEL. R. (1952) Dalyite. a new potassium zirconium silicate, from Ascension

Island, Atlantic. Minerai. Mag., 29,850-857.

VIRGO, D., MYSEN, B.O. and KUSHIRO, L (1980) Anionic constitutio-n of silicate melts

quenched at 1 atm from Raman spectroscopy: Implications for the structure of igneous melts.

Science, 208, 137 1- 1373.

WAKITA, H. and KINOSHITA, S. (1979) A synthetic snidy of the solid solutions in the

s ystems La,(CO,),- 8H20-Ce,(CO,),-8H20 and La(OH)CO,-Ce(OH)CO,. Bull. Chem. Soc.

Japan, 52,428432-

WASEDA, Y. and S W O , H. (1977) The structure of molten alkaii metal silicates. Trans.

lron Steel Insr. Jnpcrn, 17, 82-9 1.

WATSON, E.B. (1979) Zircon saturation in felsic liquids: experimental results and

applications to trace element geochemistry. Contrib. Minerai. PetroL, 70 ,407419 .

WEBSTER, J.D. (1992) Water soIubility and chlorine partitionhg in Cl-nch granitic

system: Effects of mett composition at 2 kbar and 800°C. Geuchim. Cosmuchim. Acta, 56,

679-687.

'WILLIAMS-JONES, A.E. and WOOD, S.A. (1992) A prelimuiary petrogenetic grid for REE

fluorocarbonates and associated rninerals. Geochim Cosmochim- Acta, 56,725-738.

WILSON, E.B., DECIUS, J-C. and CROSS, P.C. (1955) Molecular Vibrations. McGraw-

Hill, New York (reprinted 1980, Dover Publications).

XUE, X. and STEBBINS, J.F- (1993) %a NMR chemical shifts and locd Na coordination

environmen ts in silicate crystals, melts und glasses. Phys. Chem. Minerais, 20,297-307.

YOUNGMAN, RE. and ZWANZIGER, J.W. (1994) Multiple boron sites in borate glass

de tec ted wi th dynamic angle spinnint nuclear magnetic resonance. J. Non-Crystal. Solids,

168,293-297.

Appendix 1

Appendix 1 : Represenrritive elecuon microprobe andyses of crystalline phases appearing in run products of phase equilibria experhents described in Chapter 5. -Mineral abbreviations are 3s follows: ne: nepheline; fp: dkaIi feldsprir, wd: wadeite: Ic: Ieucite: pk: pankeIdyshite. Analyses have been recalculated to the appropriate number of oxygens in the formuia unit of each mineral.

nelA/610 ne lA164O f'p l M 0 wdlM640 w d l M 0 0 f'p3A/600 fp3A/600 fp3N550 Si 1 .O3 1 -04 3.0 t 3 .O2 3.03 3 .O2 3.03 3

wdl F/665 fp 1 FI665 Ic 1 F/665 lc 1 F/665 fptFl575 p WFl575 pk2 F/575 fp2Fl595 fp2Fl595 Si 3.1 3 2.0 1 2-06 3 .O2 2.03 2.07 3 .O2 3 .O3

Appendix 2

Appendix 3: Electron microprobe anaiyses o f some residual glasses from run products of experiments described in Chapter 5. Totds do not indude loss on ignition (LOI) resultinp from dissoived H20.

Appendix 3

Appendix 3: Electron microprobe analyses of residual glass for al1 experiments reported in

Chapter 2. Totals do not include dissolved H,O.

b

SiO, Al203 Na,O K2O m. F =O

total

Si02 A1203 Na20 K2O Zr02 F =O

total

SiO, AI203 Na20 K2O ZrO, F =O

total

SiOz A I 2 0 3

NazO KzO 202 F =O

total

r

SiO, A1203

Na20 K2O ZQ F =O

total

Si02 A1203 Na20 K20 Zr02 F =O

total

Si02 Al203 Na20 KzO 2 0 2 F =O

total

Si02 A1203 Na20 K20 ZrO, F =O

total

RMG4 RMG4 RMG4 RMG4 RMG5 RMG5 RMG5 RMG5 RMGS RMG5 RMG5 RMGB' 54.30 54.14 54.02 54.05 55.01 55.09 54.90 54.88 54.87 55.22 54.90 55.61

95.61 9527 95.25 95.19 95.09 95.06 94.91 94.83 94.96 95.22 94.82 95.83 RMG5 RMG5 RMG5 RMGS RMG5 RMGS RMG6 RMG6 RMG6 RMG6 RMG6 RMG6 55.48 55.24 55.46 55.61 55.65 55.64 56.43 56.26 56.28 56.41 56.52 56.35

total

SiO2 AI203 Na,O K20 ZrO, F =O

total

SiO, A(2O3 Na20 K2O ZQ F =O r =O

total

RMG7 RMG7 RMG7 RMGlr RMGlr RMGlr RMGlr RMGlr RMGlr RMGlr RMG2r RMG2r 69.53 69.52 68.86 50.27 49.90 49.95 50.01 50.05 50,09 49.67 51 -47 51.44

93.69 93.89 93.75 94.34 93.68 93.67 95.61 95.52 94,83 95.42 94.87 95.21 RMG& RMG6c RMG6c RMGGc RMGGc RMG6c RMG6c RMG6c RMGGc RMG6c RMG6c RMG6c

57.01 56.84 57.03 56.79 56.65 56.79 56.74 56.97 56,72 56.64 56.96 56.49

C

SiO, A1203

Na20 K2O Zr02 CI =O

total

Si02 4 0 3

Na20 K*O Zr02 CI =O

total

Si02 A1203 Na,O K20 2 0 2 CI =O

total

Si02 A1203

Na20 K20 Zr02 CI =O

total

RMJ1 RMJl RMJl RMJ1 RMJ1 RMJl RMJ1 RMJ1 RMJ1 RMJ1 RMJ2 RMJ2 52.88 52.85 52.78 52.92 52.70 52.74 52.77 52.36 52.77 52.58 56.13 56.36

94.79 94.79 95.08 94.92 94.43 94.72 94.83 94.52 94,77 94.72 94.91 95.22 RMJ2 RMJP RMJ2 RMJP RMJP RMJ2 RMJP RMJP RMJ2 RMJ2 RMJ2 RMJP' 56.51 56.26 56.40 56.32 56.31 56.45 56.33 56.03 56,07 56.15 55.82 56.06

total

SiOz

95.10 95.09 94.68 94.85 95.02 94.98 95,Ol 94.74 94.97 95.18 94.93 95.20 RMHl RMH1 RMH1 RMH1 RMH1 RMH1 RMH1 RMH1 RMH1 RMHI RMH1 RMH1 51.61 51.80 51.82 51.73 51.66 51.42 51.43 51.13 51.50 51.79 5150 51.69

total

Si02

93.92 93.85 93.78 93.91 93.88 93.65 93.61 93.06 93.43 93.89 93.05 93.72 RMH1 RMH2 RMH2 RMH2 RMH2 RMH2 RMH2 RMH2 RMH2 RMH2 RMHP RMH2 51.49 52.25 52.01 52.19 52,30 52.21 52.15 52.01 52.35 52.42 52.39 52-16

total K

SiO2

ZrO, total

RMH3 RMH3 RMH3 RMH3 RMH3 RMH3 RMM RMH4 RMH4 RMH4 RMH4 RMH~' 55.12 54.44 54.53 54.73 54.67 54.93 54.71 56.87 56.65 56.77 57.00 56.51 17.90 18.07 18.24 18.20 18.06 17,94 18.10 16.50 16.55 16.36 16.56 16.63 8,89 8.83 8.86 8.87 8.90 8.70 8.93 9.04 9.04 8.95 9.07 8.89 9.42 9.43 9.40 9.36 9.36 9.45 9.46 8.45 8.43 8.34 8.33 8.35 3.34 3.72 3.29 3.34 3.33 3.39 3.37 4 .O2 4 ,O6 4.07 3.93 4.03

94.67 94.49 94.32 94.50 94.32 94.41 94.57 94.88 94.73 94.49 94.89 94.41 RMH4 RMH4 RMH4 RMH4 RMH4 RMH4 RMH4 RMH4 RMH4 RMHS RMHS RMHS 56.68 56.56 56,62 56.76 57.02 56.63 56.75 56.77 56,69 63.79 64.31 64 .O3 16.63 16.58 16.78 16.45 16.45 16.73 16.50 16.53 16.58 12.74 13.08 12.72 8.89 9.08 9.07 8.86 9.02 9.15 8.94 8.85 9.1 1 7.46 7.39 7.40 8.33 8.30 8.37 8.32 8.40 8.23 832 8.45 8.35 6.23 6.57 6.43 3.74 3.96 4.00 3.95 3.94 4.01 3.92 3.96 3.97 3.81 3.64 3.69

94.27 94.48 94.84 94.34 94.83 94.75 94.43 94.56 94.70 94.03 94.99 94.27 RMH5 RMH5 RMHS RMH5 RMH5 RMHS RMH5 RMH5 RMH5 RMH5 RMHS 63.96 64.38 64.25 64.11 64.37 64.01 6458 64.48 64.26 64,27 63.63 13.38 12.82 12.87 12.62 12.85 12.87 12.82 12.78 12.81 13.15 12.71 7.13 7.46 7.59 7.72 7.76 7.58 7.66 7.51 7.51 6.93 7.17 7.01 6.28 6.39 6.30 6.17 6.31 6.24 6.35 6,30 6.88 6.33 3.46 3.93 3.89 4.43 4.06 3.91 3.91 3.87 3.77 3.41 3.76

94.94 94.87 94.99 95.18 95.21 94.68 95.21 94.99 94,65 94.64 93.60 I