American Mineralogist, Volume 79, pages 504-512, 1994

A phase equilibrium study in the system KAlSi3Os-NaAlSi3Os-SiO2-Al2SiOr-HrO andpehogenetic implications

Da.vrn B. JovcnChemistry Division, Oak Ridge National Laboratory, P.O. Box 2008, M.S. 6110, Oak Ridge, Tennessee 37831-6110, U.S.A.

DoNllo E. VorcrDepartment of Geosciences, Deike Building, Pennsylvania State University, University Park, Pennsylvania 16802, U.S.A.

ABSTRACT

The effect of saturating haplogranitic melts with sillimanite at 2 kbar was examined bydetermining near-eutectic or near-minimum melt phase relations in the sillimanite-freeand sillimanite-saturated haplogranite systems. Three series of compositions were studiedwith Ab/(Ab * Sa) ratios of 0. 0.545. and 1.0. The solubilities of sillimanite in these meltswere nearly indistinguishable from each other (3.5 + 0.5, 3.7 + 0.3, and 3.7 + 0.5 wt0/0,respectively). Under sillimanite-saturated conditions, the eutectic or minimum-melt tem-peratures were lowered by 15-29 "C, and the melt compositions were shifted to morequartz-rich compositions by 1.5-2.5 wto/o.

At 2 kbar, the temperatures and compositions (on an anhydrous, sillimanite-free basis)ofthe ficur- and five-phase points are

A b + Q z + L i q + V 7 4 8 + 5 ' C

A b + Q z + S i l l + L i q + V 7 3 6 + 5 " C

S a + Q z + L i q + V 7 6 1 + 7 " C

F s p + Q z + L i q + V 6 8 5 + 7 " C

Fsp + Qz + Sil l + Liq + V 656 + 10'C

Sa+ Qz + Si l l + L iq + V 737 + 5"C 43.5 + l .0wto/oQz

36 + 1.0 wto/o Qz

38 + 1.0 wto/o Qz

42 + l .0wto/oQz

33.5 + 2wto/oQz

36.0 + 2wto/oQz.

The solubility of sillimanite in these melts was found to be insensitive to P and Z overthe range studied, 2.0 and 6.0 kbar, 680-900 'C, but preliminary experiments indicatethat it is increased by the presence of mafic components. This lowering of the minimum-melt temperature may explain the observation of apparently primary magmatic andalusitein some granitic rocks.

Iurnooucrrou the melting point depressions of feldspars and quartz duepartial melting of alumrnous metaser r of sillimanite into the melt, the phase

commonly hypothesized to produce la eutectic point were redetermined in the

nitic magmas. A prominent feature of s sillimanite. The experiments focused

their peraluminous nature (Chappell . phase relations near the eutectic points

Aluminum silicate minerals are found ems, and near the minimum-melt com-peraluminous granites (Clarke et al., l, Lplogranite system at 2.O kbar-que, 1971, 1973; Price, 1983), and silhmanrte rs a com-mon constituent of metapelites at the high grades of PnnvrOuS woRKmetamorphism corresponding to beginning of melting. The Ab-Qz-HrO system was studied by Tuttle andThe present study was undertaken to determine the effect Bowen (1958) up to 4.0 kbar. The best estimate of theof excess alumina in the form of sillimanite (Sill) on the HrO-saturated eutectic point at 2.0 kbar from their dataphase relations in the haplogranite system. is 7 52 + I0 "C, 62 + 2.0 wto/o albite. Shaw ( 1963) deter-

Experiments were performed first to measure the sol- mined the HrO-saturated eutectic point in the Sa-Qz-HrOubility of sillimanite in near-eutectic melts. Subsequently, system. At 2 kbar he found the eutectic point to be at 58the phase relations near the eutectic or minimum melt in wto/o sanidine ,767 + 7 "C. Chatterj ee (1974) determinedsillimanite-absent systems were determined more pre- the solidus temperature as a function of pressure in thecisely than had been done previously. Finally, to evaluate Ab-Qz-aluminum silicate-HrO system. In that study, the0003-004x/94l0506-0504$02.00 504

JOYCE AND VOIGT:PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM 505

solidus temperature at 2.0 kbar l ies at 730 + 15'C. Thecompositions of the melts were not determined.

There are very few experimental data on the solubili-ties of aluminous minerals in granitic melts. Dimitriadis(1978) determined, by microprobe analysis, the Al con-tents of corundum-bearing synthetic glasses in the per-aluminous haplogranite system at 2.0 kbar, 760'C, andat 5.0 kbar, 680'C. The compositions had from 2. I I to3.56 wto/o sillimanite (correspondingto 1.33-2.24 wto/onormative corundum), with no apparent systematic vari-ation with pressure or composition of the glass.

Solubility data can be obtained from studies in naturalsystems. Winkler and von Platen (1961) melted para-gneisses made from graywackes. Two of these composi-tions contained sillimanite at the highest metamorphicgrade before melting. Comparison of the modes of thesegneisses to those of the crystalline residues 85 'C abovethe solidus (-70o/o melt) indicates a solubility of silli-manite between 2.9 and 4.1 wto/o. Also, Winkler (1979)synthesized a gneiss containing 3.0 wto/o sillimanite. At2.0 kbar, 770 "C (HrO-saturated), this material was 800/omolten. All the sillimanite had dissolved, resulting in-3.75 wto/o sillimanite in the melt. The average silliman-ite content of these three natural melts was about 3.6 wto/o(correspondin g to 2.27 wt0/o normative corundum).

Clemens and Wall (1981) seeded two peraluminousgranitic glasses with finely ground sillimanite and quartz,and partially crystallized them at 5 kbar and 800 and 850"C, with a variety of HrO-CO, fluid compositions. Theyanalyzed the residual glasses by electron microprobeanalysis. The values ranged from 4. I to 7.5 wt0/o silli-manite, correspondingto 2.6-4.7 wt0/o normative corun-dum. No estimate of error was given. These compositionsincluded ferromagnesian components, which preliminaryexperiments have shown to increase the solubility of sil-limanite.

Recently, Holtz et al. (1992a, 1992b) investigated theeffects of excess AlrO, on the phase relations in the Qz-Ab-Or system at 2 kbar with ar,o : 0.5. They reporteda solubility of mullite equivalent to 2.38-3.66 wto/o silli-manite (correspondin g to | .5-2.3 wto/o normative corun-dum), and a shift in the composition of the boundarybetween quartz and feldspar to more quartz-rich com-positions (3 + I wto/o more quartz-rich) in the peralu-mrnous system.

ExpnnrurNTAL TEcHNreuE

Starting material

Adularia (OR-l), quarlz, sillimanite, and Amelia al-bite, as well as glasses made from mixtures of these min-erals, were used in the Ab-Qz and haplogranite parts ofthis study (Table l). Sillimanite and glassy and crystallinematerial prepared from gels were also used in all parts ofthis study. In addition, the solubility of sillimanite inhaplogranitic melts was initially determined using a glassdonated by G. Fine and E. Stolper (Table l).

Brazilian quartz was ground to <400 mesh and rinsed

TABLE 1. Analysis of albite, adularia (OR-1), and minimum-meltglass starting materials (wt%), with the compositionsexpressed as mole percent end-member minerals

Oxide Amelia albite- OR-l* Min. meltt

sio,Al,o3CaONaroKrOL.O r .

Total

b / . b

18.90.09

1 1 . 3

86.067.720.002.433 3 3na

100.0

61.0018.200 0 01.03

0.14 14.732.7O na

100 6 94.96Calculated mole percent of end-member minerals

NaAfSi3Os 97.62 9.1 2O.7KAlSi30sCaAl,Si,O€

100.01 100.0

0 80 90.90.43 0.01 . 1 6 0 . 0

19 .80.0

59.5100 0

si4osTotal

'Analyst H. Gong. Sample no. 83-246" By electron microprobet Analysis supplied by G. Fine: includes 0.46% Sb,O3.

in HCI and HNO.. The sillimanite, from Williamstown,Australia, was donated by J. G. Blencoe. For the Sa-Qzpart of this study, it was ground to a size < 5 pm, and thevery fine material was removed by flotation. For the Ab-

Qz and haplogranite parts of this study, the sillimanitewas ground and sieved for the <80 and >200 mesh. X-raypowder diffraction (XRD) analysis revealed only silli-manite. The Amelia albite was prepared by crushing,hand-picking, grinding, and sieving for the fraction <80and > 200 mesh and then performing density separationsin bromoform + acetone liquids. Optical and XRD ex-amination revealed pure albite only. The albite was an-alyzed spectrochemically at the Mineral ConstitutionLaboratory, Pennsylvania State University (H. Gong, an-alyst); this analysis and the calculated composition of thealbite are presented in Table I . Adularia (OR- I ) was usedto make glasses in the haplogranite part of the study andwas analyzed by electron microprobe (Table l). An ap-proximate minimum-melt composition glass, donated byG. Fine and E. Stolper, was used to make a preliminarydetermination of sillimanite solubility. An analysis of thisglass is also presented in Table l.

For the Sa-Qz part of this study, mixtures of crystallinesanidine and quartz were used as the starting material.The sanidine starting material was crystallized from a gelprepared according to the procedure outlined by Luthand Ingamells (1965). The gel had a calculated NarO con-tent of 0. I wto/0. It was crystallized with excess HrO at5.0 kbar, 800 "C, in an IHPV for one week. The resultingmaterial was identified as sanidine by XRD. Microscopicexamination showed no signs of residual gel. Gels ofcompositions AburQz., and AbroSaro were prepared usingthe same technique. The AburQz* gel was fired to a glass,which was used in the initial determination of sillimanitesolubility in the Ab-Qz system. The AbroSaro gel was hy-drothermally crystallized and used as seeds in the hap-logranite part ofthis study.

Glasses of compositions spanning the Ab * Qz eutecticor granite minimum-melt composition were made by

506

810

800

wt%o QzFig. l. The phase relations near the HrO-saturated eutectic

in the KAlSirO8-SiO?-H,O system at 2.0 kbar, projected ontothe anhydrous sanidine-quartzbrnary. Solid circles : liquid sta-ble. Open triangles : sanidine stable. Open diamonds : quartzstable. Open circle indicates indeterminate phase relations thatare interpreted to lie close to a phase boundary.

weighing out appropriate amounts of albite, orthoclase,and quartz, mixing them in an agate mortar, and meltingthem hydrothermally. Microprobe analyses of these hy-drous glasses confirmed their compositional homogenei-ty, but, because of low probe totals, the gravimetricallydetermined compositions were considered more accurate.

In the Ab-Qz part of the study, glasses were made frommixtures of 52-70 wto/o albite in 2 wto/o intervals. Totalamounts of each composition ranged from 400 to 840mg. Each composition was ground in an agate moftar for30 min and then split in half. Between 3.7 and 3.45 wo/osillimanite was added to one of the splits of each com-position. These compositions were melted hydrother-mally at 2.5 kbar,900 "C, for 6 d. The resulting glasseswere finely ground, and the sillimanite-free glasses wereseeded with 0.5-4.0 wto/o crystalline albite and quartz, thebulk composition of the seed material being the same asthat of glass. The sillimanite-containing compositions wereseeded with sillimanite and, in some cases, albite or quartzas well. In experiments with glasses containing only sil-limanite seeds, it was found that the stable phase nucle-ated heterogeneously on the sillimanite crystals.

Haplogranitic glasses were made from mixtures pre-pared by weighing out appropriate amounts of OR-I,Amelia albite, and Brazilian quarlz to produce compo-sitions at 31.5, 34.5,37.5, and 41.0 wto/o quarlz, with thebalance being feldspar. The ratio of albite to sanidine inthese compositions was constant at 1.2, with the albitecontent of the OR- I adularia and the sanidine content ofthe Amelia albite taken into account. These mixtures werehydrothermally melted at2kbar,800 "C, for 4 d. A small

JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

P zsoF\

780

amount of crystalline material was still present in mostcases. These glasses were finely ground, and splits weretaken of each. Between 6.7 and 8.0 wt0/0, sillimanite wasadded to one split of each composition. These composi-tions were then reloaded and hydrothermally remelted at2 kbar, 750-780 "C-approximately 50 oC above the liq-uidus temperature. Seeds of composition AbroSaro wereprepared by hydrothermally crystallizing a gel. Silliman-ite-free compositions were seeded with 0.4-0.5 wto/o feld-spar, quartz, or both.

Equipment

The Sa-Qz and Ab-Qz parts of this study were con-ducted in large-volume internally heated pressure vessels(IHPV) (Holloway, l97l). Experiments with haplogranit-ic compositions were conducted in cold-seal vessels. Phaseboundaries were commonly bracketed by capsules placedin the same vessel and were thus reacted at nearly iden-tical conditions. IHPV experiments were quenched byturning offthe power to the furnace, and cold-seal exper-iments were quenched by holding the vessel under a jetof air. Isobaric conditions usually were maintained untilthe temperature had fallen to at least 400 "C.

Manganin cells (Harwood Engineering, Walpole, Mas-sachusetts) were used to monitor the pressure in the IHPV,and a Bourdon-tube gauge was used to monitor the pres-sure in the cold-seal system. The pressure determinationsare considered accurate to +50 bars.

Temperature was measured by sheathed Chromel-Alu-mel (type K) thermocouples. The maximum fluctuationof the temperature from the set point was always <2.0"C and usually <1.0 "C. The thermocouples were cali-brated against a NBS standard K-type thermocouple, themelting points of HrO and Al, and the boiling point ofHrO. Temperatures are considered accurate to + l0 "C.

Samples were made by loading 30-70 mg of materialinto Pt capsules with a 5-mm o.d. with l0-20 wto/o HrO.This HrO content was sufficient to produce a HrO-satu-rated sample at experiment conditions.

Experiment times varied from 4 to 14 d. In experi-ments with seeded samples, this length of time was suf-ficient for the small amount of seeds to react in the melt(dissolve completely, or show significant growth). Twoexperiments (AQSL- I I and AQSL- 16) were reacted atthe same temperature, but for l0 and l4 d, respectively.There were no detectable differences in the kind. texture.or amounts of the various phases in equivalent samplesbetween these two experiments. When crystalline, mi-cronized starting material was used, the samples wereheated for l0- 14 d to promote the growth of larger crys-tals.

Identification of phases

Crystalline phases were distributed uniformly through-out the quenched glasses after the experiments and wereidentified optically with a petrographic microscope in allsamples. Certain samples, including all haplogranitecompositions, were atalyzed by XRD, verifying the op-

5048464442403836

o

Q z + L

' , , ' L + S i l l

tor' o

O S a + Q z + S i l l

JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

750

740

730

507

780

P 760

F\

P 760F\

750

7303840424, r ' -464850

wt%o QzFig. 2. The phase relations near the H'O- and sillimanite-

saturated eutectic in the KAISi:OE-SiOr-AlrSiO5-HrO system at2.0 kbar, projected onto the anhydrous sanidine-quartz binary.Solid circles : liquid stable. Open triangles : sanidine stable.Open diamonds : quartz stable Open square : sanidine andquartz stable. Open circles indicate indeterminate phase rela-tions that are interpreted to lie close to a phase boundary. Thedashed lines are the phase boundaries from the sillimanite-freesystem. superimposed for comparison.

tical identifications of sillimanite, qtaftz, and feldspars.In several cases, the XRD patterns constrained the Ab/Sa ratios of the alkali feldspars, in accordance with themethod of Tuttle and Bowen (1958). Sillimanite had adistinctive fibrous, needlelike habit, high relief, and par-allel extinction. Sanidine and alkali feldspar formedequant, euhedral crystals, which facilitated identification.Albite usually formed rectangular laths with inclined ex-tinction. The morphology of quartz ranged from anhedralblebs to subhedral doubly terminated bipyramids.

Srr,r,rvrlxrrE soLLJBrLrrY

The solubility of sillimanite in near-minimum or near-eutectic melts was determined by heating samples withvarious amounts of sillimanite and noting its disappear-ance in undersaturated samples and its persistence in over-saturated samples. This technique is considered morerobust for these hydrous alkali glasses than electron mi-croprobe analysis because ofthe problem ofalkali vola-tility and because oversaturated samples generally con-tained extremely fine-grained, dispersed needles ofsillimanite, making it difficult to analyze only the glass.

The solubility of sillimanite in the near-eutectic meltsin the Sa-Qz system was determined for the SarrQzo,composition at a temperature of 790 "C by analyzingsamples with 2.0, 3.0, 4.0, 5.0, 10.0, 15.0, and 20.0 wto/o

28 30 32 34 36 38 40 42wtVo Qz

Fig. 3. The phase relations near the Hro-saturated eutecticin the NaAlSi.O8-SiOr-HrO system at 2.0 kbar, projected onto

the anhydrous albite-quartz binary. Solid circles : liquid stable.

Open triangles : albite stable. Open diamonds : quartz stable.Open square : albite and quartz stable. Open circle indicatesindeterminate phase relations that are interpreted to lie close to

a phase boundary.

sillimanite. Sillimanite was visible optically in the samplethat contained 4.0 wtolo sillimanite, but not in the samplethat contained 3.0 wto/o sillimanite. Therefore, the solu-bility is taken to be 3.5 + 0.5 wto/o sillimanite.

The solubility of sillimanite in Ab-Qz melts was deter-mined first for a glass made from a gel of compositionAbu,Qz., by heating samples with 2.0, 3.0, 4.0, 5.0, 10.0,and 15.0 wto/o added sillimanite, with excess H'O at2.0kbar and 7 40 "C for 13 d. No sillimanite was observedin the 2.0 or 3.0 wto/o sillimanite compositions, but a tracewas present in the 4.0 wto/o sample. The 5, 10, and 15wto/o compositions contained abundant sillimanite. The2.0 wto/o sample also had nucleated albite and quartz. Incontrast, the 3.0 wt0/o sample contained albite, but notquartz-qualitative evidence that the eutectic shifts tomore quartz-rich compositions with the addition of sil-limanite. The 4, 5, 10, and 15 wtolo samples containedonly sillimanite and glass.

To ensure that equilibrium was attained, the sampleswith 3.0 and 4.0 wto/o sillimanite were reground, reloadedinto capsules, and held at 2.0 kbar and 750'C for another14 d. No changes were observed. A glass (from crystals)of AbrrQzo, containing 3.39 wto/o crystalline sillimanitewas also heated to 750 "C at 2.0 kbar. No sillimanite wasobserved in this glass at the end of the experiment. Thus,the solubility is taken to lie at 3.7 ! 0.3 wto/o sillimaniteat the Ab + Qz + Sill + Liq + HrO eutectic.

The solubility of sillimanite seems to be independent

: Qz+L

508 JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

800

780

720

700

680

660

760Q

740Fr

r \ 7 6 0

t\ 750

740

730

72028 30 32 34 36 38 40 42 44 46

wtVo QzFig. 4. The phase relations near the HrO- and sillimanite-

saturated eutectic in the NaAlSisOs-SiOr-AlrSiOr-HrO systemat 2.0 kbar, projected onto the anhydrous albite-quartz binary.Solid circles : liquid stable. Open triangles : albite stable. Opendiamonds : quartz stable. Open square : albite and quartz sta-ble. The dashed lines are the phase boundaries from the silli-manite-free system, superimposed for comparison.

of pressure. Two samples with the AburQzr* compositionheld at 6 kbar, 760 'C, bracketed the solubility of silli-manite between 3.5 and 4.00/0, indicating a solubility of6 kbar of 3 .7 o/o'r 0.3o/o silli manite, indistinguishable fromthe solubility at 2.0 kbar.

The solubility of sillimanite in haplogranitic mini-mum-melt was determined at 2 kbar and 900'C, using aglass donated by E. Stolper and G. Fine (Table 1).Mixtures of glass with 1.5, 2.5, 3.5, and 4.5 wto/o sil l i-manite were held at 900'C for a total of 100 h. Opticalexamination of the products showed that only the 4.5wto/o sillimanite mixture contained sillimanite. This sam-ple contained abundant sillimanite. Accordingly, the sol-ubility of sillimanite is estimated to be 3.7 + 0.5 wto/0.

A reconnaissance study of the Ab-Sill-HrO system wasundertaken wherein crystalline Amelia albite was mixedwith 0.5, 1.0, 3.0, and 4.0 wto/o sillimanite and held at840 "C, 2.0 kbar, for 4 d. The 0.5 and 1.0 wto/o sillimaniteexperiments contained albite and glass. The 3.0 and 4.0wto/o experiments contained albite, sillimanite, and glass.The percentage of glass in the samples increased percep-tibly with increasing sillimanite content. Because the as-semblage liquid + H,O-rich fluid + albite + sillimaniteis an isothermal, isobaric invariant point, the assem-blages in the 3.0 and 4.0 wto/o experiments represent dis-equilibrium. The data indicate that the minimum solu-bility of sillimanite in Ab-Sill melrs is > I wto/o at theseconditions.

28 30 32 34 36 38 40 42wt%o Qz

Fig. 5. The phase relations near the Hro-saturated mini-mum melt in the NaAlSi.O8-KAlSi3OE-SiOr-HrO system at 2.0kbar, projected onto the anhydrous Abr. , Sao, ,-quartz join. Sol-id circles : liquid stable. Open triangles : alkali feldspar(-AbrrSao,) stable. Open diamonds : q\artz stable. Open squares: alkali feldspar and quartz stable. Open circles indicate inde-terminate phase relations that are interpreted to lie close to aphase boundary.

TrrB nrrnct oF srLLrMANrrE oN THE FELDspAR ANDQUARTZ LIQUIDI AND SOLIDI

The Sa-Qz system

In the Sa-Qz system, the effect of sillimanite on theliquidus and solidus temperatures was determined usingcrystalline starting material, owing to problems with nu-cleation and growth from the melt in this very viscoussystem. Mixtures of sanidine + quartz were prepared inlarge quantities ofnearly a gram to decrease the relativeerror associated with a fixed weighing error. Mixtures ofcompositions 52, 54, 56, 58, 60, and 62 wto/o sanidinewere prepared. Each composition was homogenized bymixing it dry in a mortar for 15 min. The error in thesecompositions is estimated to be + 1.0 wt0/0.

The liquidus was located by noting the presence ofcrystalline phases of larger size and more euhedral char-acter than in the starting material, indicating crystalgrowth in a solid + liquid field, or an absence of crystals,indicating melting in the all-liquid field. The eutectic waslocated at the intersection ofthe quartz and sanidine liq-uidi.

Four wto/o sillimanite was added to the mixtures, andthe liquidi were then redetermined. The sillimanite actedas seeds for heterogeneous nucleation, and some deter-minations were made by soaking the samples for 48-72h at 800 "C, where all compositions were in the liquid +

- A b + L - - -+ sill

A b + Q z + S i l l

/oo

+Lsill

o

Af+Qz+S i l l

JOYCE AND VOIGT:PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

TABLE 2. Diagnostic results: Sa-Qz-Sill-HrO

Expt. t(h) ffc) Composition' Products.'

509

U

Fr

760

740

720

700

680

660

640

62030 32 34 36 38 40 42

wt%o QzFig. 6. The phase relations near the HrO- and sillimanite-

saturated minimum melt in the NaAlSirOr-KAlSi3Os-SiOr-HrOsystem at 2.0 kbar, projected onto the anhydrous Ab'oSaro-quartzjoin. Solid circles: liquid stable. Open triangles: alkali feldspar(-Abr.Oro,) stable. Open diamonds : qnarlz stable. Open cir-cles indicate indeterminate phase relations that are interpretedto lie close to a phase boundary. The dashed lines are the phaseboundaries from the sillimanite-free system, superimposed forcompanson.

sillimanite field, and then lowering the temperature to thefinal experimental conditions. The compositions that werein a liquid + solid + sillimanite field then showed growth

of large euhedral crystals of the stable phase, whereascompositions below the solidus contained sanidine,quartz, and sillimanite. The solidus was again located atthe intersection of the liquidus curves and was confirmed

through bracketing, by noting the first presence of melt in

experiments performed without prior soaking at 800 "C'

Ab-Qz and haplogranite

The method used in the Ab-Qz and haplogranite partsof this study to determine the liquidus and the solidusinvolved seeding a glass of known composition, con-ducting the experiment at the desired temperature, andexamining the experiment products to determine whichof the seed phases had grown and which had dissolved.The small amounts of crystalline seeds thus acted as aprobe of the equilibrium phase assemblage. The disap-pearance of seed crystals of both phases indicated thatthe charge was in the all-liquid field. The seeds overcamethe problem of requiring homogeneous nucleation offeldspars or quartz in these viscous melts. Previously madeglasses (both with and without sillimanite) were seededwith crystalline material ranging in amounts from a fewtenths ofa percent to a few percent. In each case, the seedmaterial either was of the same composition as the glass,or was present in such small quantities that the bulk com-position was essentially the same as that of the glass.

s/18-1 2405t18-2 2405/18-5 2405/18-6 2403t4-2 168314-3 1683t4-5 1687131-3 336314-6 1683114-9 1683114-10 1683114-11 1685/19-3 2405119-4 2405/19-5 24010/19-1 33610119-2 336913-4 240913-2 2409/3-3 24010t12-1 24010121-1't 33610121-12 3369t10-4 5529/10-8 5529/10-9 5s210121-8 336't114-15 3361114-13 3361114-2 3362t1-4 16810121-10 1682t1-5 168

O z + LLLS a + LQ z + LLLS a + LS a + LO z + LLS a + LO z + LLS a + LQ z + L + S i l lL + SillL + SillS a + L + S i l lS a + L + S i l lS a + L + S i l lL + SillQ z + L + S i l lSi l l + LLLS a + L + S i l ls i i l + LQ z + L + S i l lS a + L + S i l lS a + L + S i l lS a + O z + S i l l + LS a + Q z + S i l l

800 Sau.800 StLo800 Sd"o800 Sau,790 Sduo790 S€Lu79O S?uo790 Slruo790 Sau,780 S?.u780 Sa."780 Sauo773 Sduu773 S€1,"773 SEL'77O Sas, + 4% Sill77O Sa54 + 4% Sill770 Sass + 57" Sill770 Sa58 + 5% Sill77O Sa6o + 5% Sill760 Sa5s + 4% Sill760 Sa56 + 47" Sill760 Sa54 + 47" Sill759 Sa5s + 47" Sill759 Sass + 3% Sill759 Sa5s + 27" Sill75O Saso + 4% Sill75O Sa56 + 47" Sill740I Sa,u + 4% Sill74OI Sa5o + 47" Sill74O+ Sa6o + 4% Sill735t Sa6o + 4Y" Sill735+ Sa6o + 47" Sill

' Subscript represents wto/o Sa on an anhydrous, sillimanite-free basis.Wt% Sill on an anhydrous basis.

-. L : melt (silicate liquid), Sill : sillimanite, Sa : sanidine, and Oz :

quartz. All experiments are HrO-saturatedt Held at 800 rc for 24 h, then lowered to experiment temperature.

f Brought up temperature to experiment conditions

The determination of the stable phases depended onthe assemblage observed after quenching, as well as onthe textures observed. In a few cases, grain mounts oftheunreacted glasses * seeds were used for comparison. Thetextural criteria for determining that an observed phase

is stable were (l) the presence of euhedral crystals, dis-tinct from the irregular, fractured grains added as seeds,(2) the presence of crystals much larger than the restrictedsize range ofthe seed fragments, (3) overgrowths on op-tically homogeneous but irregularly shaped grains, indi-cating homogeneous growth on seeds, and (4) heteroge-neous nucleation and growth of the phases on sillimaniteseeds.

Several charges yielded ambiguous results. The crys-talline phase observed in these charges after the experi-ment was present in smaller amounts than what was load-ed and appeared as rounded or embayed subhedral toanhedral blebs. These same compositions had been heldat very similar temperatures (within 5'C) and had shownunambiguous results (growth or complete dissolution).Also, in some of those cases, experiments adjacent incomposition at the same temperatures showed consistent,unambiguous results. Therefore, the ambiguous experi-

5 1 0 JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

Tlau 3. Diagnostic results: Ab-Qz-Sill-H,O

Expt f (h) P (kbar) rfc)Composition.

Glass + (seeds) Results"

AOS.2

AO-5AQL-9

AOSL-10AOSL-11

AQSL-15

AQSL.17

AQSL-19

AQSLM-20

AQSL-21

AOSS-23AAQS-5-7AAO5-5-6

312 2 .o312 2.O312 20312 2.0312 2.0142 2.O360 2.O360 2.O312 2.0336 2.0336 2.0336 2.0't92 2.0192 2.O'192 2.O168 2 .01 6 8 2 01 6 8 2 0168 2 .0168 2.0168 2.0240 2.O240 2.O240 2.O240 20240 2096 2.096 2.096 2.O96 2.096 2.O

168 2 .01 6 8 2 0168 2.0168 2.0168 2.O168 2.O48 2.O

316 6.03 1 6 6 0

L + A b + ( Q z )L + A bL + SitlL + S i l lgrowth of Ab + QzL + A b + Q ZL + O z

L + SillL + A b + S i l lL + SitlL + Q z + S i l lL + Q Z + S i l IL : A b + Q ZL + Si t lL + A bL + Ab (trace)LL + O zL + A b + S i l IL + S i l lL + A b

l * R n * s i t tL + SitlL + Q z + S i l lAbL + A bL + A bL + A b + S i l lL + A b + S i l lL + Si t lL + SitlL + Q z + S i l lLL + Qz (trace)LA b + Q z + S i l lL + Si t lL

740740740740740755759759750740740740747747747755755755755755755/bu760760760760840840840840840772772772772772772731765/ o c

Ab6' + (2% SilDAb6' + (3% Siil)Ab6' + (4% Silt)Ab6' + (4% Siil)A b 6 r + ( A b + Q z )Ab""tA b 6 ' + ( A b + o z )A b 6 ' + ( A b + Q z )Ab* + (siil)Ab64 + (sitDAb6' + (sill)Ab6o + (Siil)Ab@ + (sitDA b 6 r + ( A b + Q z )Ab64 + (sitDA b 6 s + ( A b + Q z )A b 6 6 + ( A b + o z )A b 6 1 + ( A b + o z )A b 6 r + ( A b + Q z )A b 6 s + ( A b + Q z + S i l l )A b 6 6 + ( A b + Q z + S i l DAb6s + (Ab)A b 6 6 + ( A b + Q z )A b 6 s + ( A b + Q z + S i l l )A b 6 6 + ( A b + o z + s i i l )A b s s + ( A b + o z + s i i l )Ab'ootAb,oo + 0.5% silltAb,oo + 1% SiiltAb,oo + 3% SilttAb,@ + 4% SiiltA b ? o + ( A b + Q z + S i l l )A b s s + ( Q z + S i l l )A b 5 6 + ( Q z + S i l l )A b 6 ' + ( A b + Q z )A b e + ( A b + Q z )Ab?o + (Ab)Ab64 + 6% SiiltAb6o + 4% SilttAb6' + 3.5% Siltt

. Subscript represents wt% Ab on an anhydrous, sillimanite-free basis. Wt% Sill on an anhydrous basis" L : melt (silicate liquid), Sill : sillimanite, Sa : sanidine, and Oz : quartz. All experiments are Hro-saturatedt Purely crystalline starting materiat.

ments are interpreted to lie virtually on the phase bound-aries and are plotted as open circles in Figures l-6.

Results

Results from experiments in the Sa-Qz subsystem arelisted in Table 2 and shown graphically in Figures I and2. The location ofthe 2-kbar four-phase point Sa + Q +Liq + HrO at Sa58 (anhydrous basis) and at767 + 5 "Ccompares very well with Shaw's (1963) determination atSa$ and 767 + 7 "C.

The effect of saturating the near-eutectic Sa-Qz meltswith sillimanite is shown in Figure 2, where the curvesare projected from AlrSiO, and HrO onto the KAlSi.Or-SiO, binary join. Also shown for comparison are the liq-uidi and solidus in the sillimanite-free system, from Fig-ure l. The eutectic point has shifted to 737 + 5 'C at aprojected composition of Saru, on an anhydrous, silli-manite-free basis. Although the position of the sanidineliquidus is relatively unaffected, the quartz liquidus isdepressed significantly.

Results of all experiments in the Ab-Qz subsystem arepresented in Table 3. The best estimates of the liquidinear the eutectic in the Ab-Qz-HrO system are presentedgraphically in Figure 3 and those for the Ab-Qz-Sill-H,Osystem in Figure 4. The eutectic compositions in the sil-limanite-saturated and the sillimanite-free systems werefound at the intersections of the liquidi for quartz andalbite at 62 and 64 + | wto/o albite, respectively (anhy-drous, sillimanite-free basis). The composition of theTuttle and Bowen (1958) Ab + Qz + HrO eutectic at 2.0kbar is 62 + 2 wto/o albite. The differences between theseeutectic determinations are less than the combined un-certainties of the two studies.

Results from the haplogranite system are presented inTable 4 and plotted in Figures 5 and 6 for the sillimanite-free and sillimanite-saturated systems, respectively. Thebest estimate of the eutectic points are at 687 + 7 "C,33.5 + 2wto/oquartz (sil l imanite-free),andat653 + l0"C, 36.0 + 2 wto/o quartz (sillimanite-saturated). The ef-fect of sillimanite in this subsystem is similar to the effectin the Sa-Qz system, resulting in a depression of the eu-

JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

Trsle 4. Results in the haplogranite system

Expt. f(h) rfc) Products'

5,4a4,2a3,2a1 , 36,3b

3 , 1 b4,5a5,3a6,3a

200384168112156

11216890

200156

200384165112

165384200

200384168

384165112156

112165168384200156

200384165

31.5 wt% Oz720 L709 L7O1 L + FsP694 L + FsP680 Fsp + Qz

34.5 wto/o Qzo / 5683689700680

37.5 wto/o Qz

Qz + Sao6sX I S + LX l s + L

Qz + Fsp

S i l l + Q z + F s p + LS i l l + F s p + LS i l l + F s p + L

L + Si l lS i l l + L + Q zs i i l + L + o z

5,2b4,3b2,2b1 , l

2 , 14,'l5 , 1

5,4b4,2b3,2b

4,4b2,41 , 46,2a

1 , 52,33 ,1a4,4a5,3b6.2b

651674.5660

739 L7 2 9 O z + L7 1 9 O z + L7 0 9 Q z + L

0

41.0 wto/o Qz751.5 Qz + L768 L+Qz ( t r ace )780 L

31.5 wto/o Oz + Sill720 L + Si t l709 L + Sill701 L+S i l l +FsP( t r ace )

34.5 rttlo/o Oz + Sill672 L + Sill

500 600 700 800

T"CFig. 7. The aluminum silicate phase diagram of Robie and

Hemingway (1984), plotted with the granite minimum meltingcurve (Tuttle and Bowen, 1958; Luth eI al., 1964), and a hypo-thetical peraluminous granite minimum melting curve depressed30 "C by the dissolution of an aluminum silicate mineral. Thestippled area represents the P, T region where andalusite couldprecipitate from a peraluminous. granilic magma.

tectic temperature by -29 "C, and shifting it to morequartz-rich compositions by -3 wto/0.

Although not directly comparable, the results of thisstudy are in overall good agreement with the results ofHoltz et al. (1992a, 1992b). The present experiments werecarried out with du.o : 1.0, and the aluminum silicatephase was sillimanite, whereas the experiments of Holtzet al. were carried out at aH,o : 0.5, and the aluminumsilicate phase was mullite. In that study, the haplograniteminimum-melt temperature was depressed 25 + l0 'C

upon the addition of an aluminum silicate phase, vs. 29+ l0'C in this study. This small difference may be in-significant, or it may be due to the lower a",o, indicatinga possible effect of cH.o on the solubility of AlrO. in thesemelts. Holtz eI al. (1992a, 1992b) reported a solubility ofmullite equivalent to 1.9 + 0.4 wIo/o normative corun-dum. The solubility of sillimanite in haplogranite meltsobserved in this study corresponds to 2.3 + 0.3 wto/o nor-mative corundum. With structural equilibrium in the melt,less dissolved alumina produces a concomitantly smallerdepression of the minimum-melt temperature. Therefore,the difference in melting point depression may be due todifferent amounts of alumina being dissolved in the meltsin these two cases. Mullite, being metastable, should havea higher solubility than sillimanite, all other factors beingequal. The factors that were different were aH,o and, to alesser extent, temperature and composition. Thus, the re-sults obtained by Holtz et al. (1992a, 1992b) may indicatean effect of HrO activity, temperature, or composition onthe solubility of mullite.

37.5 wto/o Qz + Sill6 7 5 O z + L + S i l l689 Sill + L683 Sill + L672 Oz + Sill + L700 L + Sill660 L+S i l l +Qz ( t r ace )

41.0 wto/o Qz + Sill739725719

- L : melt (silicate liquid), Sill : sillimanite, Fsp : "1L""

feldspar' Xls :

unidentified crystalline phases, and Oz : quartz. All experiments are H,Osaturated.

Pnrnor,ocrcAr. rMPLrcATroNs

Formation of peraluminous melts

The results of this and other studies (Voigt, 1983; Joyce'1986; Voigt and Joyce, 1991; Holtz etal.,1992a,1992b)imply that compositions in the AlrSiO5-saturated haplo-granite system begin to melt at temperatures 15-30 "Clower than in the AlrSiOr-free system, to produce peralu-

minous, quartz-enriched melts containing -2.3 w|o/o nor-

mative corundum.However, the results of this study probably do not rep-

resent the upper limit of the alumina content of natural

melts. Preliminary experiments on the solubility ofgarnetin Ab-Qz melts indicates that the dissolution of 9 wto/ogarnet will increase the solubility of sillimanite from 3-7

to 4.5 wo/o (resulting in a total of 4.8 wto/o normative

corundum). Clemens and Wall (1981) determined the

normative corundum content of a natural garnet + cor-

dierite + biotite granite melt saturated with sillimaniteat 800 'C. Aside from the mafic components, the com-

5,2a4,3a2,2

5t2

position of their starting material is very similar to theone used in this study. Their results (4.70lo normativecorundum) also suggest that mafic components increasethe peraluminous nature of melts in equilibrium with analuminum silicate mineral.

AlrSiOs in granitic rocks

Although aluminum silicate minerals in granitic rocksare often interpreted to be restite phases, some petrog-raphers have observed andalusite in granitic rocks and,on the basis oftextural or chemical evidence, have inter-preted it as magmatic (Clarke et al., 1976; de Albuquer-que, 1971, 1973; Price, 1983). However, the aluminumsilicate phase diagram (Holdaway, l97 l; Robie andHemingway, 1984) shows no overlap between the anda-lusite stability field and the granite minimum-meltingcurve (Luth et al., 1964; Tuttle and Bowen, 1958), asshown in Figure 7. Price (1983) discussed this discrep-ancy and noted that the addition of B, Li, and F loweredthe granite solidus (Wyllie and Tuttle, 1961, 1964; Chorl-ton and Martin, 1978; Manning, l98l). However, theresults of this study suggest that one large effect in low-ering the haplogranite minimum-melting temperature isdue to andalusite saturation itself. The peraluminousgranite minimum is plotted as the curve 30 .C lower thanthe simple granite minimum in Figure 7, representing ahaplogranite composition saturated in an aluminum sil-icate mineral. This depression of the granite minimumresults in a region ofoverlap (the stippled area in Fig. 7),where andalusite can crystallize directly from a peralu-minous grani l ic magma.

ACKNoWLEDGMENTS

D B.J. $atefully acknowledges the generous support ofa Carnegie In-stitution of Washington Geophysical Laboratory postdoctoral fellowship.John Frantz is thanked for his help in the experiments and Larry Fingerfor his help with XRD analyses. Special thanks go to James G. Blencoeand J.C Clemens for their insightful reviews, which significantly im-proved this article, and to l-orna Perkins for her reviews and editing. Thisstudy was also supported by the Geosciences Program of the Office ofBasic Energy Sciences, U.S. Department of Energy (DOE), under contractnumber DE-AC05-84OR21400 with Martin Marietta Energy Systems,Inc. The senior author participated in the U.S. DOE postgaduate pro-gram at Oak Ridge National Laboratory, administered by Oak RidgeAssociated Universities. Parts ofthis study are from the master's thesesofthe authors C.W Burnham is acknowledged for suggesting the researchtopic and for his support.

RppenrNcns crrED

Chappell, B.W., and White A.J.R. (1974) Two contrasring ganite types.Paci f ic Geology. 8, l l3-174.

Chatterjee, N.J (l 974) Crystal-liquid-vapor equilibria involving parago-nite in the system NaAlSi.Os-Al.O.-H.O. Indian Joumal of Earth Sci-e n c e , 1 , 3 - l l .

Chorlton, L.B , and Martin R.F. (1978) The effect ofBoron on the granitesolidus. Canadian Mineralogrst, 16, 239-244.

Clarke, D.B., McKenzie, C.B, Muecke, G K., and Richardson, S.W (1976)Magmatic andalusite from the South Mountain batholith, Nova Scotia.Contributions to Mineralogy and Pelrology, 56,279-287

Clemens, J.D., and Wall, V.J (1981) Origin and crystallization of some

JOYCE AND VOIGT: PHASE EQUILIBRIA IN A HAPLOGRANITIC SYSTEM

peraluminous (S-tlpe) ganitic magmas Canadian Mineralogist, 19, I I l-l 3 l .

de Albuquerque, C.A.R. (l 97 l) Petrochemistry ofa series ofgranitic rocksfrom northem Portugal. Geological Society of America Bulletin, 82,2783-2798.

-(1973\ Geochemistry of some biotites from granitic rocks, north-ern Portugal. Geochimica et Cosmochimica Acta, 37, 1779-1802

Dimitriadis, S. (1978) Some liquid compositions in the peraluminoushaplo$anite system. Neues Jahrbuch fiir Mineralogie Monatshefte, 8,377-383.

Holdaway, M J (1971) Stability ofandalusite and the aluminum silicatephase diagram. American Journal of Science,27l,97-131.

Holloway, J.R. (1971) Internally heated pressure vessels. In G.C. Ulmer,Ed., Research techniques for high pressure and high temperature, p.217 -258. Springer-Verlag, New York

Holtz, F, Johannes, W., and Pichavant, M. (1992a) Peraluminous gran-ites: The effect of alumina on melt composition and coexisting min-erals. Transactions ofthe Royal Society ofEdinburgh: Earth Sciences,83,409-416.

- (1992b) Effect of excess aluminium on phase relations in the sys-tem Qz-Ab-Or: Experimental investigation at 2 kbar and reduced HrO-activity. European Journal ofMineralogy, 4, 137-152.

Joyce, D.B. (1986) A phase equilibrium study in the system NaAlSirOs-SiO=-AI,SiO,-H,O at 2 kilobars and petrogenetic implications, 60 p.Master's thesis, Pennsylvania State University, University Park, Penn-sylvania.

Luth, W.C, and Ingamells, C.O. (1965) Gel preparation of staning ma-terials for hydrothermal experimentation American Mineralogist, 50,255-258.

Luth, W.C , Jahns, R.H , and Tuttle, O F. (1964) The granite system atpressures of 4 to l0 kilobars Journal of Geophysical Research, 69,7 59-773.

Manning, D.C (1981) The effect ofF on liquidus phase relationships inthe system Qz-Ab-Or with excess water at I kbar. Contributions toMineralogy and Petrology, 7 6, 206-215

Price, R.C. (1983) Geochemistry of a peraluminous granitoid suite fromnorth-eastern Victona, south-eastern Australia. Geochimica et Cos-mochimica Acra, 47, 3 l -42.

Robie, R A , and Hemingway, B.S. (198a) Entropies of kyanite, andalu-site, and sillimanite: Additional constraints on the pressure and tem-perature of the AlrSiO5 triple point. American Mineralogist, 69,298-306.

Shaw, H.S (1963) The four-phase curve sanidine-quartzJiquid-gas be-tween 500 and 4000 bars American Mineralogist, 48, 883-896.

Tuttle, O.F., and Bowen, N.L. (1958) Origin of granite in the light ofexperimental studies in the system NaAlSi.Or-KAlSi.O8-SiOr-H:O.Geological Society of America Memoir, 74, 153 p.

Voigt, D.E. (1983) The solubility of Al.SiO, in the system KAlSir08-SiOr-H.O at 2 kbar, and its implication for melt speciation, 30 p. M.S thesis,Pennsylvania State Universjty, University Park, Pennsylvania.

Voigt, D.E., and Joyce, D B (1991) Depression of the granite minimumby the addition ofsillimanite (abs.). Eos, 72, 304.

Winkler, H.G. (1979) Petrogenesis of metamorphic rocks (5th edition),348 p Springer-Verlag, New York.

Winkler, H.G., and von Platen, H. (1961) Experimentelle Gesteinsme-tamorphose. IV. Bildung anatektischer Schmeltzen aus metamorphis-ierten Grauwacken. Geochimica et Cosmochimica Acta, 24, 48-69.

Wyllie, P.J., and Tuttle, O.F. ( l96l) Experimental investigation of silicatesystems containing two volatile components. The effects of NH. andHF in addition to HrO, on the melting temperatures of albite andgranite. American Journal of Science, 259, 128-1 43.

-(1964) Experimental investigations of silicate systems containingtwo volatile components III. The effects of SO., P,Or, HCl, and Li.Oin addition to HrO on the melting temperatures of albite and granite.American Journal of Science. 262. 93O-939.

Meroscrum RECETYED Aucusr 13, 1993MeNuscrrm AccEmED Jnm;.qnv 7, 1994