The effect of Fe and Mg on crystallization in granitic systems

American Mineralogist, Volume 65, pages 639-653, 1980

The effect of Fe and Mg on crystallization in granitic systems

Mrcuepr. T. NeNev

Chemistry Division, Oak Ridge National LaboratorylOak Ridge, Tennessee 37830

NNO SEUUEL E. SWANSON

Geophysical Institute, University of AlaskaFairb anks, Alaska 99 7 0 I

Abstract

Single-step and multistep undercooling experiments using both Fe,Mg-free and Fe,Mg-bearing model granitic compositions were conducted to investigate the influence of maficcomponents on the crystallization of granitic melts. Crystallization of granite and gran-odiorite compositions in the system NaAlSi3O6-KAlSirOr-CaAlzSi2Os-SiOr-H2O producesassemblages containing one or more of the following phases: plagioclase, alkali feldspar,qloartz, silicate liquid, and vapor. The observed phase assemblages are generally in goodagreement with equilibrium data reported in the literature on the same bulk compositions.

With the addition of Fe and Mg to these bulk compositions six new phases participate inthe equilibria (orthopyroxene, clinopyroxene, biotite, hornblende, epidote, and magnetite).However, crystalline assemblages produced in phase equilibrium and crystal grofih experi-ments brought to the same fnal P-Z-X6,. conditions are in general not equivalent. In crys-tal-growth experiments, nucleation of the feldspars and quartz is greatly inhibited in the pres-ence of Fe and Mg. Indeed, plagioclase is the only tectosilicate to nucleate in the granodioritecomposition. Mafic phases nucleate and grow outside of their thermal stability fields as de-fined by the equilibrium phase diagrams. This contrast in mineral assemblages between theequilibrium and crystal growth experiments is in marked contrast to the results obtained forFe- and Mg-free compositions. Perhaps the addition of Fe and Mg has caused a breakdownof the Si-O framework in the melt, thereby promoting the more rapid nucleation of the ino-and phyllosilicates rather than the framework silicates.

Border zones of granitic plutons, com-only rich in mafic minerals, may result from themore rapid nucleation of mafic phases from the silicate liquid. These zones are thought to de-velop by early crystallization along the walls of the pluton. Our results suggest the maficphases should nucleate more quickly than the feldspars and quartz and thus should enrichthe early crystallization products in ferromagnesian minerals.

Introduction

Crystal-growth experiments were conducted tostudy the effects of iron and magnesium on crystalli-zation processes in granitic magmas. Experimentswere made by undercooling homogenized silicate liq-uids in 5ingle or multiple steps using both Fe,Mg-freeand Fe,Mg-bearing model granitic compositions.

The phase equilibria relationships of the modelgranitic compositions have been previously studied.

I Operated by Union Carbide Corporation for the U.S. Depart-ment of Energy under Contract W-7205-eng-26.

Whitney (1972, 1975) studied the Fe,Mg-free hap-logranitic compositions, and Naney (1977) investi-gated the closely related Fe,Mg-bearing graniticcompositions. Swanson (1977) studied the relation ofcrystal nucleation and growth rates to the develop-ment of granitic textures from silicate liquids gener-ated from the Fe,Mg-free haplogranitic composi-tions. The success of Swanson's study and theavailability of phase equilibria data to guide experi-mentation prompted this investigation.

At the outset it was expected that nucleation andgrowth of the felsic phases from the Fe,Mg-bearing

0003-004x/80/0708-0639$02.00 639

NANEY AND SWANSON: CRYSTALLIZATION IN GRANITIC STSTEMS

compositions would be similar to the behavior ofthese phases in the Fe,Mg-free haplogranitic compo-sitions. Our objectives were to: (l) investigate the de-velopment of textures in granitic rocks by comparinggrowth from Fe,Mg-free compositions with growthfrom Fe,Mg-bearing com.positions, and (2) evaluatekinetic factors irnportant to crystallization of mag-mas (crystal nucleation and growth rates). The delayor complete inhibition of felsic phase nucleation andgrowth from the Fe,Mg-bearing compositions wasunexpected.

We compare the crystallization of the mafic-freewith mafic-bearing granitic systems and offer an ex-planation for the contrasting behavior. One implica-tion of the inhibition of felsic phase growth for natu-ral granitic melts is discussed.

Previous work

In an attempt to understand pegmatite genesis,Jahns and Burnham (1958) grew crystals from a hy-drous silicate liquid. Natural samples of pegmatiteand HrO sealed in noble-metal capsules were heatedto produce silicate liquids and were subsequentlycooled to various subliquidus temperatures to studycrystallization. Formal results of these experimentshave not been reported. However, a photograph pub-lished by Wyllie (1963) illustrates the apparent suc-cess of these experiments.

Mustart (1969) found that crystals of albite up to 3mm in length could be grown from a peralkalinemelt in the system NaAlSi3Or-NaoSirOr-HrO. Mus-tzrt (1972) used the technique of homogenization ofthe silicate gel plus HrO charge above the liquidus,followed by crystalli"ation at subliquidus temper-atures. Substitution of KAlSirO, or SiO, forNaAlSi.O, allowed potassium feldspar and quartz re-spectively to grow in place of albite.

Nucleation and growth kinslis5 of alkali feldsparsfrom hydrous si l icate l iquids in the systemKAlSi.Or-NaAlSirO*-HrO have been studied byFenn (1972, 1977) following the method of Mustart(1972). Experiments were made with both HrO-un-dersaturated and -saturated silicate liquids at 2.5kbar. Alkali feldspars grew elongate parallel to the acrystallographic direction, with relative growth ratesin the unit cell such that a > c > 6. Crystal growthrates up to 6.8 x 10-6 cm/sec were measured parallelto a. Some experiments whic_h were undercooled withrespect to the liquidus for periods of 72 hours ormore showed no signs of crystal nucleation. Becauseof the variability in this nucleation lag time (rime be-tween ? drop to subliquidus value and crystal nucle-

ation) it was not possible to determine the exact timeover which growth took place, and therefore themeasured crystal growth rates are minimum values.The morphology of the crystalline products werespherulitic except in experiments containing an aque-ous vapor phase, which produced single isolatedcrystals at temperatures near the liquidus. Within thehypersolidus region the crystal growth experinentsyielded products predicted by the equilibrium phasediagram.

Growth of plagioclase from hydrous silicate liq-uids has been studied in the system NaAlSirO.-CaAl'Si,O,-H,O by Lofgren (1974r,b). Lofgrenmade both isothermal and step-cooling (periodiclowering of temperature by a prescribed amount)crystal growth experiments at 5 kbar. Morphology ofplagioclase crystals was observed to change from agenerally equant, tabular form at small undercoglingvalues (LT : fiiq"ia"" - Z"*p..i-"o,) to skeletal, den-dritic, and finally spherulitic crystal forms at thehighest AT. Association of spherulitic growth textureswith rapid cooling rates is well illustrated in photo-graphs (Lofgren, 1974a). Spherulitic overgrowthsproduced during the temperature decrease at the ter-mination of the experiment mantle the surfaces oftabular crystals. Lofgren (l97aa) observed that nu-cleation incubation periods increased with the Nacontent of the system at small values of ATin isother-mal crystal growth experiments. One experimentfailed to produce crystals after six days at AZ of40oC. Experirnents made with stepwise cooling pro-files produced normally zoned plagioclase crystalsanalogous to those found in many natural occur-rences (Lofgren, l914a,b).

Lofgren and Gooley (1977) have grown feldsparcrystals containing non-equilibrium intergrowthsthat texturally resemble perthite from melts in thesystem NaAlSirOr-KAlSirOr-CaAlrSirOr-HrO. Ex-perirnents using linear cooling rates or stepwise cool-ing paths both produced plagioclase crystals inter-grown with thin lamellae or patches of aftali feldsparnear their outer margins. They attribute the plagio-clase-alkali feldspar intergrowths to the formation ofa boundary layer of melt enriched in alkali-feldspar-forming components rejected from the advancingplagioclase-liquid interface. The boundary layer de-velops in response to slow diffusion of alkali-feld-spar-forming components in the melt relative torapid growth of the adjacent plagioclase crystal andproduces two changes advantageous to developmentof plagioclase-sanidine intergrowths. The presenceofthe boundary layer causes a breakdown ofplanar

plagioclase-liquid interfaces, resulting in a changefrom a tabular to dendritic growth habit for plagio-clase. Further, the boundary layer eventually be-comes sufficiently enriched in alkali-feldspar com-ponents for local crystallization of alkali feldspar tooccur around the plagioclase dendrites.

Swanson et al. (1972) and Swanson(1977) studiedcrystal nucleation and growth from synthetic graniteand granodiorite melts in the system NaAlSirO*-KAlSi,O.-CaAlrSirO,-SiO,-H,O. Crystal growthrates of plagioclase, alkali feldspar, and quartz werefound to vary as a function of AZ of the experimentfrom lO-u to l0-' cm,/sec. The highest growth rateswere observed in experiments with HrO-under-saturated ssmpositions undercooled approximately200'C. Maximum growth rates observed in the pres-ence of an aqueous vapor phase are two orders ofmagnitude lower than those measured in HrO-under-saturated experiments, and occur at larger under-cooling (Af - 400'C). With the exception of the me-tastable persistence of glass (quenched silicate liquid)at subsolidus temperatures, crystal growth experi-ments produced phase assemblages consistent withthe equilibrium phase diagram. Presumably this me-tastable glass would crystallize if experiments werecontinued for longer periods of time.

Several recent studies have dealt with crystalgrowth from mafic silicate liquids in the basalt-ultra-mafic compositional range. Textural simulation ex-periments have been made with model lunar basaltsby a number of workers, including Lofgren et al.(1974) and Walker et al. (1976). Donaldson (1976,1979) studied the development of crystal morpholo-gies and delay in nucleation of olivine as functions ofcomposition, superheating, undercooling, and cool-ing rate in mafic systems. Gibb (1974) studied crys-tallization of plagioclase from two silicate liquids ofbasaltic composition. Plagioclase nucleation fromthese melts was found to be a function of both themagnitude of AZand duration of the experiment. Heshowed that with slow constant cooling rates(0.05'C/hr) plagioclase and clinopyroxene will nu-cleate from melts of the compositions investigated atsmaller AT than for faster cooling rates (10'C/hr).

Experimental techniques

Bulk compoitions

Prior to making crystal growth experiments, equi-librium phase relations for the compositions ofinterest must be well known. The equilibrium datahelp to guide and interpret crystallization experiments.

G I

Equilibrium phase relations of two anhydroussimplified granitic compositions in each of thesystems NaAlSirO'-KAlSirO'-CaAlrSirO'-SiOzand NaAlSi,O*-KAlSi3O,-CaAlrSirOt-SiO,-KMg,rrFe,orAl,,rSi3O,, with varying amounts ofH,O have been investigated at 8 kbar by Whitney(1975) and Naney (1977') respectively. The same fouranhydrous compositions were used in this study.

Our synthetic granite and granodiorite composi-tions are shown in Table l, together with Nockolds'(1954) average biotite-hornblende granite and horn-blende-biotite granodiorite after which the syntheticcompositions were modeled. The Fe- and Mg-freegranite and granodiorite compositions are designatedRl and R5 respectively. The Fe- and Mg-bearingsynthetic granite and granodiorite were prepared bymechanical addition of l0 weight perc€nt of an 'an-

hydrous biotite' composition (Ml) to both the Fe-and Mg-free granitic compositions, hence they are re-ferred to as Rl*l0Ml and Rs+lOMl respectively.The three basic anhydrous bulk compositions-Rl,R5, and Ml-were synthesized as coprecipitated gels

Table l. Chemical compositions studied

c

E ^

WEIGHT PERCENT OXTDES

70 .56 65 . s0 74 .14 70 .52 40 .4314 .00 15 .55 15 .04 17 .95 12 ,930 . 9 1 1 . 5 3 2 4 . 4 42 . 4 7 2 . 7 90 . 0 6 0 , 0 90 . 4 8 1 . 8 6 1 1 . 9 01 . 6 3 4 . 1 0 1 . 4 9 3 . 9 13 . 5 6 3 . 8 4 3 . 7 0 4 . 3 35 . 3 9 3 . 0 1 5 . 6 3 3 . 2 9 L O . 2 90 . 4 0 0 . 6 10 . r 0 0 . 2 30 . 5 0 0 . 6 9

NANEY AND SWANSON: CRYSTALLIZATION IN GRANITIC STSTEMS

o o

o o

! 5 ! Eo o o oE i F d! + p +E i c h

6 ! d ei i 4 . 4

o vo . o : i o

€ o ! o o H

i o o v G E! l . d d od o € o g t si i o t ! o Io z o . i

d v t t s o o| t r o io o o r! u d ! o oi . i ! o ^ E Fu d C d s ! ! ^o d L d 6 E E 6

; n o ! 6 > > dF O : d o d o o v

SiO2A12O3F e 2 O 3FeOMn0MgoCaoNaz0KzoTiOzP z o sHzOl

70 .77 67 .5Lt4 .83 L7 .452 . 4 4 2 . 4 4

1 . 19 1 . 19L . 3 4 3 . 5 23 . 33 3 , 906 . 1 0 3 . 9 9

NORMATIVE EQUIVALENT

Q u a r t z 2 4 , 5 2 0 . O 2 7 . 9Or thoc lase 31 . 7 17 . 8 33 . 3A- lb i te 29 .9 32 .5 31 .3A n o r t h l t e 6 . 4 1 6 . 4 7 . 4C a S i O 3 0 . 3 0 . 9M g S i o 3 L . 2 4 . 6F e S i O r 3 , 0 2 . 9M a g n e t t t e I . 4 2 . 3r l ren l re 0 . 8 L .2A p a t i t e 0 . 3 0 . 5Hemt i teCorudum O.2

2 4 . 4t9 .43 6 . 6L 9 . 4

23.4 20.236 .0 23 .628 .2 33 .06 . 6 L 7 . 5

3 . 0 3 . 0

2 . 4 2 . 40 . 3 0 . 3o . 2

NANEY AND SWANSON: CRYSTALLIZATION IN GRANITIC SYSTEMS

by a technieue modified after the procedure de-scribed by Luth and Ingamells (1965). Temperaturey.s. Xr,o equi l ibr ium relat ions for Rl, R5,R l + l 0 M l , a n d R 5 + l O M l a t 8 k b a r a f t e rWhitney (1975) and Naney (1977) are shown inFigures I and,2.

Experimental method

Internally-heated pressure vessels modified afterthe design of Yoder (1950) were used for the crystalgrowth experiments (Holloway, l97l). Temperaturemeasurements were made with Inconel-sheathed, Pt/Pt-l0Rh thermocouples which were calibrated peri-odically in their normal experinental confgurationagainst the melting points of gold (1062.5'C) and so-dium chloride (800.5'C). The melting point datawere corrected for pressure effects from the data ofAkella and Kennedy (1971) and Clark (1959). Tem-perature measurements are estinated to be accurateto +l0oC. Pressure was transmitted to the samplesby an argon medium, and was measured with maga-nin cells in conjunction with a Carey,Foster bridge.Pressure measurements are believed to be accurate to1100 bars.

Experi-ments with the Fe- and Mg-bearing compo-sitions were not buffered to maintain a predeter-mined oxygen fugacity. However, the experimentalapparatus yielded oxygen fugacity values inter-mediate to the Ni-NiO and Mt-Hm oxygen bufferreactions. This was determined using capsules con-taining solid buffer reactants. Holloway (1973) alsoobserved that the oxygen fugacity in an internally-heated pressure vessel with argon as a pressure me-dium is in this range. Chou (1978) has used the hy-drogen fugacity sensor technique to determine theoxygen fugacity of similar vessels between the Ni-NiO and MnO-Mn On buffers.

Samples were prepared by encapsulating carefullyweighed amounts of freshly boiled, deionized, dis-tilled water and dried silicate gel in platinum orAg-Pd- tubes which were sealed by welding. TheFe-bearing compositions weri loaded only inAg"oPd* tubes to minimizs loss of Fe to the capsulematerial. The water content of capsules is thought tobe accurate to +0.3 weight percent of the reportedvalue.

Capsules were heated to 1200, 1150, 1100, or1000'C at 8 kbar for l8 to 96 hours. At these temper-atures most of the bulk compositions are above theirrespective liquidus and produce optically-homoge-neous silicate liquids. Liquidus temperatures werenot attained in experirnents with the Rl+loMl and

R5+l0Ml systems containing less than 2 weight per-cent HrO because the Ag"oPd* alloy used for cap-sules began melting at or below the liquidus temper-atures for these granitic compositions (Naney, 1977).

Following production of a silicate liquid, the tem-perature of the experiment was rapidly lowered tothe desired subliquidus value. Samples were allowedto re-equilibrate at this new temperature for 2 to l0days and were then either subjected to another tem-perature decrease (stepwise cooling in 50o or l00oCsteps), or the experiment was terminated by turningoff the furnace power and quenching the samples toroom temperature. The time required to drop thetemperature of an experiment from one isotherm toanother ranged from 3 to l0 minutes, depending onthe magnitude of the temperature decrease. At thetermination of the experiment, the temperaturedropped 300"C within l0 minutes and room temper-ature was reached within 20 minutes. Limitations ofthe gas intensifier system used to pressurize experi-ments did not permit temperature drops to be madeisobarically. Consequently, the pressure comnonlyfell 100 to 500 bars during a temperature drop of 50-100"C. Immediately following temperature re-equili-bration the pressure was readjusted to its previousvalue (8 kbar).

Multistep experiments with the Rl+lOMl andR5+l0Ml compositions were cooled from I150'C inincrements of 50" or l00oC. Experiments with theFe- and Mg-free compositions were cooled in in-crements of 50oC starting from I100"C. Single-stepexperiments were made at 8 kbar with one growthstep of 50oC to 350oC. Note that the actual under-cooling attained is a function of the HrO content ofthe system (Figs. I and 2). Undercooling experiencedby two samples of different HrO contents but cooledto the same temperature is a function of the slope ofthe vapor-undersaturated liquidus curve.

Analysis of experimental products

After completion of an experiment the capsuleswere removed from the pressure vessel and examinedfor leaks. Capsules that showed no indication of leak-age were opened and the contents examined underbinocular and petrographic microscopes to study sur-face textures and determine the presence of crystal-line phases and glass.

X-ray diffraction patterns were obtained fromsome powdered samples to con-firm mineral identifi-cation and for unit-cell refinement of feldspars. ANorelco X-ray diffractometer was used for routineidentification. Powder photographs obtained for

Gro n i teP=8Kbor

Pl+Af+eQ +L

Pl +Af+ aQ+ L

NANEY AND SWANSON: CRYSTALLIZATION IN GRANITIC SYST'MS 643

Gronodror i teP=8Kbo r

Pl+Af +q.O+V

( o ) ( b lFig. l. Phase diagrams for Fe- and Mg-free bulk compositions at 8,000 bars, aft€r Whitney (1975). Phase assemblages are shown as a

function of temperature (7) and water content (weight percent H2O) for the granite (a) and granodiorite (b). Pl is plagioclase, Af is alkali

feldspar, Q is quartz, and L is quenched silicate liquid.

6 8 t O t 2 t 4wt. % H2O

unit-cell refinements were made with a Nonius Guin-ier-de Wolff focusing camera used in conjunctionwith a fine-focus copper-anode X-ray tube. A spinel(o : 8.0833A) internal standard was used in samplesprepared for unit-cell refinement. Positions of dif-fraction lines on frlms were measured on an opticalcomparator with a precision of 0.001 mm (0.0005'20). Measured peaks are thought to be accurate to+0.010o 20 or better. Unit-cell parameters were ob-tained using the least-squares computer program de-veloped by Appleman and Evans (1973).

Electron microprobe analyses were made of crys-tals and glasses in selected samples with an AppliedResearch Laboratories model EMx-sM microprobe.Instrument conditions were as follows: beam-acceler-ating potential-I5 kV (crystals and Fe,Mg-freeglasses), 20 kV (Fe-,Mg-bearing glasses); sample cur-rent-0.02 microamp (crystals), 0.01 microamp(glasses); beam spot size-l to l0 microns. These an-alytical conditions did not produce any noticable ad-verse effects (1.e., alkali boil-otr) in the glasses ana-lyzed. An anhydrous glass of the mafic-free syntheticgranite composition and synthetic plagioclase pre-pared by P. M. Fenn were used as standards for the

analysis of the Fe- and Mg-free experiments. Ana-lyzed natural minerals, glasses, and a synthetic glasswere used as standards for analysis of the Fe- andMg-bearing experiments. Data reduction was pet-formed ssing a modified version of Colby's (1968)MAGIC IV computer program.

Experimental results

Nucleation

A number of crystal growth experiments failed tonucleate any crystals even after periods of up to 144hours at subliquidus temperatures (Figs. 3 and 4).This nucleation lag time has also been observed inother granitic systems (Lofgren, 1974a; Fenn, 1977;Swanson, 1977) and is probably related to the slowapproach to an equilibrium distribution of theatomic clusters in the sitcate liquid prior to the be-ginning of nucleation (Turnbull and Fisher, 1949).

Compositions most commonly failed to nucleatecrystals at low ATand in the presence of an HrO-richvapor phase. In general the Fe- and Mg-bearing sys-tems more successfully nucleated crystals than didthe mafic-free systems. Experiments with the Fe- and

Groni te

r 2 00

I ooo

rec)

8 0 0

o 2 4 6 I t O t 2 t 4 t 6 t 8 2 0wt. % H20

Mg-free compositions indicate that liquids of thesynthetic granite composition (Rl) are more reluc-tant to nucleate crystals than are those of the syn-thetic granodiorite composition (R5).

Crystal morphology

Morphologies of all crystalline phases were foundto be strongly dependent on the magnitude of the un-delss6ling imposed on a sample during a growthstep, as expected from crystal growth theory (Kirk-patrick, 1975). The response of feldspar crystal mor-

NANEY AND SWANSON: CRYSTALLIZATION IN GRANITIC SYST'MS

P= SKbo r Gronod io r i te P=8Kbo r

| 2 0 0

I ooo

T ("C)

8 0 0

6 0 0

( o ) ( b )Fig. 2. Phase diagrams for Fe- and Mg-bearing bulk compositions, after Naney (1977). The symbols and abbreviations used on these

phase diagrams are explained below. Dashed lines indicate uncertainty in position of stability limits for that phase. Opx : O :Onhopyroxene; Bt: B : Biotite; Cpx: C: Clinopyroxene; Hb: H : Hornblende; Ep: E: Epidote; pl: p: plagioclase; Af : A :Alkal iFeldspar; aQ:Alpha-Quartz; BQ:Beta-Quarrz; L:Si l icateLiquid; V:AqueousVapor; ?=eueriedfieldindicatesthatthebeginning ofdry melting is unknown and that the reaction relationships for the breakdown ofbiotite at low H2O content are unknown.

Numbered fields shown in the granite diagram (a) represent the phase assemblages indicated: (l) Opx + Pl + Af + L; (2) Opx + Pl +A f + B Q + L ; ( 3 ) O p x + C p x + P l + L ; ( 4 ) O p x + B t + L ; ( 5 ) O p x + C p x + B t + A f + L ; ( 6 ) O p x + C p x + B t + P l + A f + L ; ( ? ) E p +B t + P l + A f + d Q + L ; ( 8 ) E p + C p x + B t + A f + a Q + L ; ( 9 ) E p + B r + A f + a e + L ; ( 1 0 ) E p + B r + A f + a e + L + V ; ( l l ) C p x +B t + A f + a Q + L + v .

Numbered fields shown in the granodiorite diagram (b) represent the phase assemblages indicated: (l) Opx + Pl + Af + BQ + L; (2)o p x + B t + P l + A f + F Q + L ; ( 3 ) B t + P l + B Q + L ; ( 4 ) H b + B t + p l + A f + a e + L ; ( 5 ) E p + B r + p l + o e + L ; ( 6 ) E p + B t + p l +aQ + L + V; (7) Hb + Ep + Bt + Pl + L; (8) Hb + Bt + L.

o 2 4 6 8 t O t ? t 4 t 6 t 8 2 0

wt "/" H2o

phology to changes in undercooling in the Rl and R5compositions (Fig. 3) follows the same general pat-tern described by Lofgren (1974a). Briefly, experi-ments made at low undercooling (Af : 50 to 100'C)produced large tabular crystals up to 8 mm in lengthwhereas crystals grown at larger A? have skeletal ordendritic morphology. Feldspar crystals have spheru-litic morphology when growth took place at under-coolings ranging from AZ: 100 to 350"C. Withinthis range of AZ there is a gradual change in the ap-pearance of spherulites. Undercooling values at thelow end of this range produce rosettes composed of

r . C p r' 8 1 ' L

B l . L

C a r . B l . P l

C p r . B t . L . V

8 r . P r . A t . r o E p . 8 r . P l . A f . . 0 . V

H b . B i . P I . L

H b . 8 r . P t . r Q . L H b . B r . P t . L . v

H b . E o . B l . P l . L . V

,t E p ' B t. P l . A l . r Q . L

E p . B t . P l . A l . . O . L . V

E p + B f o P l . A f . o Q . V

O G L A S S. TABULARA D E N D R I T I C^ S P H E R U L I T I Co MtcRoLrTtc

A

P t l a f + d Q + L-

A

large, faceted (up to 4 mm long) individual crystalsradiating from a conmon center. The individual ra-diating crystals become smaller and more flbrous asthe magnitude of undercooling increases, finally re-sulting in classical spherulites (Swanson, 197'l).Quartz norphology in the mafic-free systems hasbeen discussed in detail by Swanson (l977r.In sum-mary, quartz forms thin hexagonal plates, whosehabit changes from euhedral, to skeletal, to dendriticcrystals with increased AZ.

The morphology of the mafic minerals, like thoseofthe feldspars and qvartz, is strongly dependent onthe cooling history of the experiment. Ferromagne-sian minerals in general exhibit the same growth pat-tem changes as those observed for the feldspars inthis and other studies (Lofgren, 1974a; Fenn, 1977;Swanson, 1977). Euhedral crystals are observed atsmall ATvalues, hollow or skeletal crystals at shghtlygreater A7, and spherulitic or dendritic forms at thelargest AZ investigated.

In some experiments with the Fe- and Mg-bearingcompositions there are prominent epitaxial over-growths among ferromagnesian phases. Epitaxially-

g5

6 8

wt 7" H2o

related growth pairs consisting of clinopyroxene crys-tals rimmed by orthopyroxene or orthopyroxenerimmed by clinopyroxene have been produced in dif-ferent experiments with composition Rl + l0Ml byusing the same bulk composition and changing thecooling path of the system (single-step vs. multistepcrystallization). In addition, hornblende forms epi-taxial overgrowths on clinopyroxene in experimentswith the granodiorite composition. These over-growths may have been promoted by chemical gradi-ents near the crystal-liquid interface. Vapor bubblesoccur as inclusions in a few clinopyroxene crystals oralong crystal-glass interfaces, even though the bulksilicate liquid was undersaturated with respect toHrO at the conditions of the experiment. Similar va-por bubbles were observed by Fenn and Luth (1973)in aftali feldspars crystallized from vapor-under-saturated silicate liquids. The heterogeneous nucle-ation of crystalline or vapor phases on crystal mar-gins indicates local saturation of the silicate liquidwith respect to a third phase at the crystal-liquid in-terface of a growing crystal. Local saturation is theresult of rapid growth of the substrate crystal relative


r 200

fioo

rooo

T ("C)

900

800

700

600

6 8

wr% Hzo

Fig. 3. Results of isothermal crystal growth experiments with the Fe- and Mg-free bulk compositions. Mineral assemblages are as

indicated by the phase diagram except for those experiments that did not nucleate crystals. Experimental data used in this figure are

available upon request from the authors.

o G L A S S. TABULAR

a D E N D R I T I C^ S P H E R U L I T I Co M I C R O L I T I C

Gron i te


P= 8 Kbo r

L t e u t D l 2 O OO P X } LOPX.CPX.LOPX. CPX.8T .LC P X . B T . L8 T . L

Gronod io r i fe P=8Kbo r

!

3

A

a

| 2 0 0

tooo

T( "C)

8 0 0

6 0 0

t ra

\Orthopyroxcna

HOtOt . t tz / f to t r r t r r ) t r " r , :

Sol idus

to the mobility of components rejected by the grow-ing crystal.

Mineral assemblages in Fe- and Mg-free compositions

Single-step crystal growth experiments. Results of5ingle-step crystal growth experinents with the Fe-and Mg-free compositions that nucleated crystals areconsistent with the equilibrium phase diagram shownin Figure l. Therefore, the phase diagram can be suc-cessfirlly used to predict the crystalline products of acrystal growth experiment. The results shown in Fig-ure 3 duplicate the crystalline phase assemblage pre-dicted by the phase diagram, when crystals do nucle-ate.

.Although results of fractional crystallizallon exper-iments cannot in theory be predicted from the equi-librium diagrams, in practice our predictions havebeen quite successful. This success can be attributedto the degree of crystallization taking place in the ex-periments. In general a small volume ratio of crystals

rooo

T("C)

8 0 0

L IQU I DOPX. PL. LH8. LHB .OPX. CPX.LH8 .CPX.8T . LB T . L

6 0 0

to silicate liquid is observed. Under these conditionsthe silicate liquid composition does not change srg-nifisanlly from the equilibrium composition.

Another test of equivalence between the results ofphase equilibria and crystal growth experimentswould be to determine the composition of the crystalsgrown from a single bulk composition at the sameisotherm. Unfortunately, crystals grown in equilib-riun experiments are very small and are not amen-able to microprobe analysis. In the absence of mi-croprobe data, X-ray powder di-ffraction patternswere obtained from crystal growth samples and theirphase equilibria counterparts (Whitney, 1972, 197 5).Results of the X-ray analysis indicate that corre-sponding feldspars in products of the two experimen-tal methods have nearly identical unit-cell parame-ters. Table 2 contains an example of the unit-celldata obtained for feldspars from experiment Rl-830(Whitney, 1972) and the corresponding crystalgrowth experiment (426). Both experiments were

l a

L lOU tD . VAPOR

a a

\Clinoprroren e

o 2 4 6 I r O t 2 t 4 t 6 t 8 2 0 0 2 4 6 I t O t 2 1 4 t 6 t 8 2 0wr. % Hzo wt . % H2o

( o ) ( b )Fig. 4. Results of isothermal crystal growth experiments in the Fe- and Mg-bearing bulk compositions. Symbols indicate the phase

assemblage observed in the experiments. Phase assemblage boundaries shown are those determined in phase equilibria experiments (Fig.2) and are included for comparative purposes Experimental data used in this figure are available upon request from the authors.

rU PE

aooooA

Aiot irc

Hornblcnd

Table 2. Unit-cell refinements of feldspars

u7

lustrate the type of results obtained frorr experimentswith the granite and granodiorite compositions.

At 8 kbar the granodiorite (R5) phase diagramcontains a large plagioclase + sitcate liquid field be-low the liquidus (Fig. lb). Crystal growth in this sys-tem should produce plagioclase crystals followed atlower temperatures by quartz and alkali feldspar. Amultistep experiment (504) was made using the R5composition plus 4.0 weight percent HrO. A portionof the experimental product is shown in Figure 5a. Asingle L-shaped, twimed crystal of zoned plagioclaseis surrounded by two granophyric zones consisting ofqrrurtz, alkali feldspar, and plagioclase. The fivezones within the plagioclase crystal can be correlatedwith crystallization at 1000, 950,900, 850, and 800"Cwhile the outer two granophyric zones represent crys-ixllization at 75O and 700oC, as shown in Figure 5.The inner granophyric zone (750oC) composed ofplagioclase, alkali feldspar, and qtJartz would be pre-dicted to consist of plagioclase and quartz on thebasis of the phase diagram (Fig. 1b). Coprecipitationof alkali feldspar with plagioclase at this temperatureis probably a reflection of the extraction of plagio-clase from the silicate liquid at higher temperatureand consequent change in the liquid composition im-mediately adjacent to the plagioclase crystal face.The composition of the interface liquid is thereforeeffectively enriched with respect to quartz and alkalifeldspar. Lofgren and Gooley (19'77) reported similarinterface enrichment and resulting intergrowths inthe ternary feldspar system.

Nucleation problems encountered in the single-step crystal growth experiments can be circumventedin multistep growth because each growth zone in thecrystal acts as a substrate for growth at the next lowertemperature. Good minimum values of growth ratemay be determined for the growth zones surroundingthe core as contrasted with the more questionablegrowth rates measured in single-step experiments(Fenn, 1977; Swanson, 1977). Mean growth rates ofthe four zones which surround the core of the L-shaped plagioclase crystal range from 2.3 x lO-e to6.8 x l0-" cmlsec (Fig. 5).

Results of quantitative electron microprobe spotanalyses on the plagioclase and alkali feldspar in theL-shaped crystal of experiment 504 are shown in Fig-ure 5. Normal zoning of the plagioclase is clearly in-dicated and each zone is homogeneous, at the scaleof the electron microprobe analyses. Normally-zonedplagioclase crystals have also been produced by Lof'gren (1974b) using a similar experimental method.He obtained qualitative electron microprobe analyses


A1ka1 i Ie ldspar P l a g i o c I a s eSamp Ie

Nunber Rt-830 4 2 6 4 2 6

a ( A )

b ( A )

" < i l

o ( " ' )

g ( ' ' )

r ( ' ' )

v ( A 3 )

1 . L 7 71 . 0 0 1

9 0

7 . 7 7 71 . 0 0 1

9 0

8 . 5 1 7 8 . 5 1 61 . 0 0 3 1 . 0 0 2

1 3 . 0 1 7 1 3 . 0 1 51 . 0 0 3 1 . 0 0 2

8 . 1 8 4 8 . 1 8 41 . 0 0 5 1 . 0 0 L

1 2 . 8 7 9 1 2 . 8 8 4r . 0 0 5 1 . 0 0 1

7 . 1 2 1 7 . I 2 5r . 0 0 5 1 . 0 0 1

9 3 1 9 . 3 9 3 7 9 . 7! 4 . 7 1 0 . 6

1 1 6 2 8 . 0 7 1 6 2 9 . 4! 2 . 9 I 0 . 6

6 7 1 . 1 8 6 7 ! . 7 4! . 4 5 1 . 0 9

. 0 3 6 . 0 1 9

1 8 / 1 8 6 5 / 6 9

1 1 5 5 8 . 7 1 1 5 5 7 . 3

9 0 9 0

7 7 5 2 1 7 l s . l 7! . 2 5 r . 1 6

8 9 3 s . 7 8 9 3 4 . s! 4 . 7 i 0 . 6

. 0 1 9

r e f . 2 4 / 2 5

. 0 2 5

s , e . i s t h e e s t l n a t e d s t a n d a r d e r r o r o n t h e o b s e r v a t i o n .

te f . i s the n@ber o f peaks used in the re f lnemen! /number o f peaks

subml t ted .

Nunbers be low each o f the un i ! ceL l re f inements a re the ca lcu la teds tandard er ro rs .

Exper inent R1-830 is a phase equ i l ib r ia exper iment f rom Lr t l iEney(1972) aad 426 is a c rys ta l g rowch exper inent Both exper jnenEs were

made v i th the same bu lk conpos i t ion (R1 p lus 3 we igh t percent H20) dnd

. r y s L d l l i z e d r r t h e o a r e c o n d i c i o n s ( 8 0 0 0 o a r s , 7 0 0 " c , 9 6 h r s J .

crystallized at 700oC from the synthetic granite com-position containing 3 weight percent HrO. The phaseassemblage produced in both experiments consistedof plagioclase, alkali feldspar, qtuartz, and silicate liq-uid. Results of the unit-cell refinements are. withinthe error of the refinement, identical.

Multistep crystal growth experiments. Multistepcrystal growth experiments were made in an attemptto reduce the nucleation lag time and to study tex-tures developed by crystallization at progressivelylower temperatures. In the single-stage experiments,spherulites composed of some combination of plagio-clase, alkali feldspar, and quartz were the dominanttextural feature. Identification of individual phaseswithin the spherulites was very difficult, and it washoped that multistage experiments might producecrystals large enough to permit examination of theindividual phases.

Results of the multistep experiments indicate thatsome of the difficulties encountered in the single-stepexperiments were overcome. Multistep experimentsmay allow the nuclei within the silicate melt time toreadjust gradually to decreasing temperature,whereas the rapid large temperature drop used in5ingle-step experiments may not allow sufficient timefor nuclei to re-equilibrate with the melt. Two of themultistep experiments will be discussed in detail to il-

u8

, o. lo 0Dis tonce (mm)

from his samples that indicate the presence of re-versed chemical zoning within some of the growthzones. Extinction patterns observed with polarizedlight in the plagioclase shown in Figure 5 suggest asimilar reversed zoning pattern. However, this re-verse zoning was not detected in the microprobeanalyses of selected spots with a 5 micron beam di-ameter.

Equilibrium phase relations for Rl show a numberof two- and three-phase assemblage fields betweenthe liquidus and the four-phase (plagioclase + alkalifeldspar * quartz + liquid) region (Fig. la). Each of

NANEY AND SWANSON: CRYSTALLIZATION IN GMNITIC SYSTEMS

l x l O - 8

l x lO -9

G

'o1 i t t tq - to

(,ott,

Eo

oo(r

Fig. 5. Results of multistep crystal growth experiment 504 with composition R5 plus 4.0 weight percent H2O. (A) Photomicrograph ofL-shaped plagioclase crystal with five zones of growth corresponding to 1000o, 950o, 900o, 850", and 800oC. Two zones of granophyricalkali feldspar, plagioclase, and quartz surround the plagioclase and represent growth at 7500 and 700'C. Length ofthe small arm on theL-shaped plagioclase crystal is 0.29 mm. (B) Thermal history of experiment 5M. (C) Electron microprobe analyses of L-shaped crystalperpendicular to the long arm of the L, showing the seven zones of growth. (D) Gro'wth rate as a function of crystallographic direction (aaxis and c axis) in the L-shaped plagioclase crystal.

t o o o 9 0 0 8 0 0T ( . C )

these phase assemblages is stable over a relativelysmall temperature interval for a given HrO content.Crystal growth experiments with the Rl compositionshould produce crystals of three different phases atrather small undercaelings (Af : 50"C). This is insharp contrast with results of crystal growth experi-ments w'ith the synthetic granodiorite composition(R5), in which plagioclase was the only crystallinephase for over 200oC below the liquidus.

Crystal growth experiments with the Rl composi-tion were not entirely successful, even when the mul-tistep growth technique was used. Two experiments

Thermol Hls lory

Tims (hru)


with the granite composition (Rl), containing 8 andl0 weight percent HrO, were made concurrently withexperiment 50a fig. 5) and neither produced anycrystals. Though the ratio of crystals to silicate liquidwithin the alkali feldspar + liquid field (Fig. la) wasexpected to be small under these conditions, it washoped that multistep cooling would generate nucleiof sufficient size to allow alkali feldspar growth at700'C. Experiment 500 was prepared by adding 3weight percent HrO to the Rl composition. This ex-periment was also made concurrently with experi-ment 5M. The rosettes shown in Figure 6 illustratethe dominant spherulitic growth form in experimentswith the Rl composition, even when the multistepcooling technique is used. The equilibrium phasediagram for the granite composition indicates experi-ment 500 should have been in the subliquidus regionat each growth step below 950"C. Thus six growthstages are expected in the rosettes shown in Figure 6.

Petrographic examination of the rosettes revealsthree clearly defined zones and one poorly definedzone progressing inward from the spherulite margin.However, no identifiable zones corresponding togrowth at 900o and 950"C were observed. Appar-ently nucleation was inhibited at 950o and 900oC,and the bulk of the spherulite formed at 850'C. Al-ternatively, crystallization of the central portion ofthe spherulite did take place at 950" and 900"C, butthe boundaries between the high-temperature zonescould have been eliminated by diffusion during an-nealing or recrystallization. Experiments with the Rlcomposition have demonstrated a tendency forsluggish nucleation, and because high-temperature

growth zones within the L-shaped plagioclasecrystal of experiment 504 are unaffected by an-nealing, nucleation was probably inhibited at950o and 900"C in experiment 500.

Cores of the rosette-like spherulitic masses (experi-ment 500) appear on first inspection to be composedof subhedral plagioclase crystals radiating from acommon center (Fig. 6). However, detailed examina-tion reveals that these subhedral crystals are not ho-mogeneous, and are made up of parallel filamentaryneedles. The bulk of the three outermost zones of therosettes (800o, 750o, and 700oC growth zones) arecomposed of alkali feldspar (Ab,rOr.,An,). Withinthe alkali feldspar are intergrowths of plagioclaseand quartz (Fig. 6). These intergrowths are very simi-lar to perthitic feldspars grown in experiments byLofgren and Gooley (1977).

Summary

Crystal growth experiments with the Fe- and Mg-free (Rl, R5) compositions yield the crystalline prod-ucts predicted by the phase diagrams, except for ex-periments that failed to nucleate crystals or some in-tergrowths produced along a crystal-liquid interface.Nucleation lag times increased as AZ decreased. Nonucleation was observed in some experiments heldfor perio{s up to 144 hours at P-Z conditions in thesubliquidus-hypersolidus region. Many compositionsfailed to nucleate crystals (within this time frame)when their liquids were subjected to very small un-dercoolings.

The system failed to reach equilibrium for somemultistep crystal growth experiments. In these exper-

Fig. 6. Photomicrographs of experiment 500, Rl plus 3.0 weight percent H2O. Thermal history is the same as experiment 504 (Fig.5B). (A) Coarse-grained spherulites of alkali feldspar surrounded by quenched silicate liquid. White bar is 1.0 mm. (B) Close-up ofspherulites in (A) showing intergrowths of plagioclase and quartz within the alkali feldspar in three growth zones. White bar is 0.1 mm.


iments, phases nucleate at higher temperatures thanpredicted from the equilibriurr phase diagrams. Thenonequilibrium phases always nucleate at the marginof a previously nucleated crystalline phase. The localchange in liquid composition can be great enough tocause precipitation of a second, nonequilibriumphase at the boundary of a growing crystal.

Phase assemblages in Fe- and Mg-bearing composi-tions (RI + IqM 1, R5+ I 0M I )

Phase assemblages in the products of crystalgrowth experiments with the Fe- and Mg-bearingcompositions do not duplicate the assemblages of thephase diagrams in Figure 2. Most notably, alkalifeldspar and quartz were not observed in the prod-ucts of any crystal growth experiment with composi-tions Rl+lOMl or R5+lOMl. Plagioclase was ob-served in only one experiment with the granodiorite(R5+lOMl) composition. However, in this experi-ment plagioclase appeared 250"C below its deter-mined stability limit (plagioclase is the equilibriumliquidus phase at 8 kbar for the R5+l0Ml composi-tion).

Phases observed in crystal growth experimentswith the synthetic granite (Rl + l0Ml) include: glass(quenched silicate liquid), orthopyroxene, clinopy-roxene, biotite, and magnetite. Phase-assemblagedata resulting from crystal growth experiments withthe Rl + l0Ml composition are shown in Figure 4a.

Results of crystal growth experiments with theRs+lOMl composition are shown on the T-Xr,ophase assemblage diagram in Figure 4b. Phases ob-served in crystal growth experiments made with thegranodiorite composition include: glass (quenchedsilicate liquid), orthopyroxene, clinopyroxene, horn-blende, biotite, plagioclase, and magnetite. The pres-ence of clinopyroxene is particularly interesting be-cause equilibrium experiments (Fig. 2) predict itscrystallization only from the granite composition. Al-kali feldspar and quartz present in products of equi-librium experiments were not observed in crystalgrowth experiments.

In crystal growth experiments with either theRl+lOMl or the R5+l0Ml composition, every fer-romagnesian silicate phase was observed outside itsstability field as defined by the phase diagram. Bio-tite is present in crystal growth experiments made attemperatures up to 100'C higher than its upper sta-bility limit as shown on the equilibrium phase dia-gram. This biotite, apparently crystallizing above itsstability limits, is interpreted to have grown duringthe quench. Quench biotite is sparse and nucleates

only on the capsule wall. Biotite which is stable dur-ing the experiment is homogeneously distributedthroughout the experimental product. Hornblendeoccurs in experiments containing lower HrO contentsand persists to lower temperatures than predicted bythe equilibrium phase diagram. Clinopyroxene wasobserved at higher and lower temperatures relative toits equilibrium stability field. The occurrence of or-thopyroxene is restricted to a smaller T-Xr,o fieldthan predicted by phase equilibria experiments.However, orthopyroxene was observed at a temper-ature well below its equilibrium stability field in amultistep experirnent at 8 kbar with the Rl+l0Mlcomposition. However, in this case, orthopyroxene ismantled by clinopyroxene and thus cannot react withthe silicate liquid. Epidote, a stable phase in equilib-rium experiments made with both the Rl+l0Ml andR5+l0Ml compositions, was not observed in theproducts of any crystal growth experiment.

Discussion of experimental results

Absence offeldspars and quartzfrom crystal growthexperiments

Absence of or delayed nucleation of feldspars andqvartz in crystal growth experiments produced phaseassemblages which are strikingly different frorr thoseobserved in equilibrium experiments made at thesame P-T-Xr,o conditions in the Fe- and Mg-bear-ing systems. In contrast, phase equilibria and crystalgrowth experiments with Fe- and Mg-free composi-tions show a high degree of correlation betweenphase assemblages produced in crystal growth andphase equilibria experiments made at the same P-7-&,o conditions. Results of our crystal growth experi-ments plus phase equilibria results reported byMaaloe and Wyllie (1975) indicate that nucleation ofalkali feldspar and qvartz from a granitic liquid isvery sluggish under laboratory conditions. Gibb(1974) reports that plagioclase nucleation is difficultin basaltic systems. Similar results are reported byDonaldson (1979) for the crystallization of olivinefrom basaltic melts. However, ferromagnesian miner-als such as olivine generally nucleate faster than thetectosilicates (Gibb, 1974; Donaldson, 1979). [n ourcrystal growth study alkali feldspar and quartz werenot observed in the Fe- and Mg-bearing systems,even in experiments crystallized at low temperature(650"C). In addition, plagioclase was not present inany of the samples prepared from the synthetic gran-ite composition that contains Fe and Mg. Dependingon the HrO content of the sample, these experiments

NANEY AND SWANSON: CRYSTALLIZATION IN GRANIT:IC SYST:EMS 651

were undercooled with respect to the equilibrium liq-uidus for periods of 6 to l0 days. Some samples withlow HrO contents (less than 3 weight percent HrO)were undercooled up to 250oC below the equilibriumupper stability limits of alkali feldspar and quartz asdefined in Figure 2. Compared to nucleation in-cubation periods for these phases observed in the Fe-and Mg-free granitic compositions (Swanson, 1977),these periods should have been adequate for nucle-ation.

Effect of Fe and Mg on the structure of the silicate meltand cry st alliz atio n kinetic s

Addition of Fe and Mg to the simplified graniticcompositions studied by Whitney (1975) and Swan-son (1977) may modify the structure of the silicateliquid. Kushiro (1975) found that addition of MgOand FeO to simple silicate systems (containing 40 to60 mole percent SiOr) shifts eutectic liquid composi-tions to higher SiO, contents. He believes this com-position shift is a reflection of structural changes oc-curring in the silicate melt. Addition of these cationspresumably breaks Si-O bonds in the silicate liquidstructure and makes the structure more like that ofthe silica-poor crystalline phase. In response to thechange in silicate liquid structure, the stability fieldof the silica-poor phase is enlarged.

Results of phase equilibria experiments (Fig. 2)show that (when mafic phases are disregarded) thestability fields of assemblages consisting of combina-tions of plagioclase, alkali feldspar, qtartz, and liq-uid are little changed from those found by Whitney(1975) in the Fe- and Mg-free granitic compositions(Fig. l). Just as the addition of small amounts of Feand Mg have little effect on the phase assemblagerelat ions, these components (added in smal lamounts) probably have only minor effects on thestructure of the silicate melt under equilibriumconditions. The presence of Fe and Mg in gra-nitic l iquids could restrict the size of regionswithin the melt which have framework-like struc-ture, by breaking the Si-O bonds in the melt. If thesize of these framework structures does not equal orexceed that necessary for formation of critical nuclei(Kirkpatrick , 197 5\, feldspar and quartz are inhibitedfrom growing.

Preferential crystallization of mafic phases ratherthan feldspar and quartz may result from faster crys-tallization kinetics for the Fe- and Mg-bearingphases. Indeed, some authors relate the rapid crystal-lization of olivine and pyroxene relative to plagio-clase in basic intrusives to the structural sfunplicity of

the neso- and inosilicate relative to the tectosilicate(Wager, 1959; Hawkes, 1967). Since mafic phasescrystallize before the tectosilicates in the systems westudied, it follows that after a temperature decreaseto subliquidus values the formation of critical nucleiis faster for mafic phases than it is for tectosilicates.

The difference in nucleation rates may be relatedto size contrasts between critical nuclei of mafic andtectosilicate phases. If the critical nuclei of the tecto-silicates are larger than those of the mafic phasesthey should take longer to form and thus allow ear-lier crystallization of the mafic phases. Formation ofmafic nuclei may also be kinetically favored if theaddition of Fe and Mg causes a breakdown of the Si-O framework structure in the melt. In either of theseviews rapid or preferential crystallization of maficphases compared to tectosilicates is not related to therelative simplicity of mafic phase crystal structures,but rather to the rate of formation of critical nuclei inthe melt.

Conclusions

Natural examples of rapid crystallization kinslis5

are well documented in the literature for maficphases. Examples include the growth of olivine andpyroxene to the exclusion ofplagioclase in the outerzones of pillow basalts (Bryan, 1972), development ofmafic comb-layered zones where mafic phases nucle-ate faster than feldspars or quartz (Moore and Lock-wood, 1973; Lofgren and Donaldson, 1975), and therapid growth of olivine and pyroxene relative toplagioclase in the komatiites (Drever and Johnston,1957; Arndt et al., 1977). Indeed, although Bowen'sreaction series is based on a particular basaltic bulkcomposition, it does illustrate the dominance ofearly-formed mafic phases, a crystalTuation sequencecommon for a variety of igneous rocks.

Our crystal growth experiments with the Fe- andMg-bearing R5 + l0Ml composition illustrate thesluggish crystallization of tectosilicates relative to themafic phases in granitic systems. Clinopyroxene ispresent in crystal growth experiments but is absent inequilibrium experiments made with this composition.This difference can be explained in terms of the su-percooling of plagioclase in crystal growth experi-ments (250'C lower than in equilibrium experimentswhere plagioclase is the liquidus phase), and there-fore lack of competition from plagioclase for calciumin the melt during growth of clinopyroxene. Earlynucleation of plagioclase in crystal growth experi-ments would inhibit the growth of the other Ca-bear-ing phases, as illustrated by the absence of clinopy-

652

roxene from experiments that produce plagioclase. Asimilar argument may be made involving com-petition for potassium between alkali feldspar andbiotite, resulting in crystallization of biotite ratherthan the tectosilicate. However, the absence of quartzfrom the crystal grofih experiments cannot be ex-plained by competition for components within themelt. Indeed, the eady crystallizalisn of maflc phasesshould promote the growth of quartz from the resid-ual liquid, but quartz was not observed. Clearly,some factor in addition to competition for com-ponents in the silicate melt must inhibit the nucle-ation of tectosilicates. The observed nucleation offeldspars and quartz is best explained in terms ofcontrasting rates of critical nuclei formation in thesilicate melt. This efect may be enhanced by com-petition between mafic and tectosilicate phases forcomponents of the silicate liquid common to bothphases.

Petrologic applications

Granitic plutons often show some sort of mineral-ogic zoning related to fractional crystallization of themass of magma (Bateman et al., 1963; Bateman andChappell, 1979). The pattern of the zoning is oftenconcentric about a late-crystallizing interior zone.Later intrusion of magma from the core into the pre-viously crystallized magma often alters the con-centric pattern. Mineralogic zonation in granitic plu-tons may also be asymmetric (Swanson, 1978),forming a continuous variation pattern from one sideofthe pluton to the other. In either case zoning pat-terns typically parallel the walls of the pluton.Boundaries between individual zones may be eithergradational, reflecting subtle changes in the crystal-lizing magma, or sharp, due to intrusion of magmainto previously crystallized zones.

A common feature of mineralogic zoning in gra-nitic plutons is the presence of a border zone en-riched in mafic minerals. Such a mafic border zonecan be identified in a plot of the variation of maficmineral content across the Tuolumne intrusive series(Fig. 7), a zoned granodioritic pluton in the easternpart of Yosemite National Park, California. Onecause of mafic mineral enrichment near pluton con-tacts is contamination of magma by wall-rock mate-rial. However, field evidence indicates that no signifi-cant contamination by wall-rock material hasoccurred. In addition, the wall rocks are not of theproper composition to account for the variation ob-served in the intrusive sequence (Bateman and Chap-pell, 1979). Zonitg in this pluton is thought to have


o

a

Troverso L€ngth ( km)

Fig. 7. Variation in mafic mineral (hornblende + biotite +

magnetite * sphene) content across the Tuolumne Intrusive

Series, a quartz diorite-granodiorite complex in YosemiteNational Park (Traverse A-B of Bateman and Chappell, 1979).

Total distance across the complex is slightly over 2l km.

developed by accretion of material crystallized fromthe magma at the margins of the magma chamber.This model requires the margins of the pluton tocrystallize before the core, and therefore the mineralsat the margin to nucleate sooner than minerals in thecore. Heat losses from a magma to the surroundingwall rocks accelerate coolhg at the margins of amagma chamber relative to interior zones. If inhibi-tion of tectosilicate crystallization is significant atthese "accelerated" but low sssling rates, then devel-opment of mafic-rich margins of this granitic plutonand others may be in part related to the more rapidnucleation of mafic phases relative to the feldspars orqtartz observed at high undercooling in this study.

AcknowledgmentsWe thank W. C. Luth and P. M. Fenn for much assistance and

advice during the course of this study. P. R. Gordon skillfullymaintained the experimental equipment. Among the individualswho have contributed to the ideas presented here through manystimulating discussions, the following deserve special mention: J.A. Whitney, J. G. Blencoe, J. C. Eichelberger, M. P. Taylor, P.Schiffman, G. E. Brown, J. G. Liou, J. D. O'Brient, and B. H. W.S. de Jong. The manuscript was improved by the careful reviewsof E. Dowty and G. E. Lofgren. Financial support for this workwas provided by NSF grants GA 1684 (R. H. Jahns and W. C.Luth) and GA 41731 (W. C. Luth, G. E. Brown, and W. A. Tiller).These studies were done while we were graduate students in theTuttle-Jahns Laboratory for Experimental Petrology, StanfordUniversity the use of these facilities is gratefully acknowledged.

References

Akella, J. and G. C. Kennedy (1971) Melting of gold, silver, andcopper-proposal for a new high-pressure calibration scale. ./.Geophys. Res., 76, 49694977.

Appleman, D. E. and H. T. Evans, Jr. (1973) Job 9214: Indexingand least squares refinement of powder diffraction data. Nat.Tech. Inf. Serv., U.S. Dept. Commerce, Springfield, Virginia,Document PB-216 188.

Arndt, N. T., A J. Naldrett and D. R. Pyke (1977) Komatiitic andiron-rich.tholeiitic lavas of Munro Township, northeast On-taio. J. Petrol., 18,319-369.

I

NANEY AND SWANSON: CRYSTALLIZATION IN GRANIT:IC SYSTEMS 653

Bateman, P. C. and B. W. Chappell (1979) Crystallization, frac-tionation, and solidification of the Tuolumne intrusive series,Yosemite National Patk. Geol. Soc. Am. 8u11.,90,4654g2.

--, L. D. Clark, N. K. Huber, J. G. Moore and C. D. Rinehart(1963) The Sierra Nevada batholith-a synthesis of recent workacross the central part. U. S. Geol. Surv. prof. pap. 414_D.

Bryan, W. B. (19'72) Morphology of crystals in submarine basalts.J. Geophys. Rel, 77, 5812-5819.

Chou, I-M. ( 1978) Calibration of oxygen buffers at elevaled p and? using the hydrogen fugacity sensor. Aa. Mineral., 63, 690_'703.

Clark, S. P., Jr. (1959) Effect ofpressure on the melting points ofeight alkali halides. J. Chem. phys., 31,1526_1531.

Colby, J. W. (1968) Quantitative microprobe analysis of thin in_sulating films. In J. Newkirk, G. Mallett and H. pfeiffer, Eds.,Advances in X-ray Analysis, vol. 2, p. 287. plenum press, NewYork.

Donaldson, C. H. (1976) An experimental investigation of olivinemorphology. Contrib. Mineral. petrol., 57, lg7_213.

- (1979) An experimental investigation ofthe delay in nucle_ation of olivine in mafic magmas. Contrib. Mineral. petrol.,69,2r-32.

Drever, H. L and R.. Johnslon (1957) Crystal growth of forsteriticolivine in magmas and melts. Trans. R. Soc. Edinburgh, 63, pt.i l ,289-317.

Fenn, P. M. (1972) Nucleation and growth of alkali feldspars fromsynthetic melts (abstr.). Trans. Am. Geophys. Ilnion, 53, 1127.

- (1977) The nucleation and growth ofalkali feldspars fromhydrous melts Can Mineral.,lt 135-161.

-- and W C. Luth (1973) Hazards in the interpretation of pri_mary fluid inclusions in magmadc minerals (abstr.). Geol. Soc.Am. Abstracts with Programs. 5,617.

Gibb, F. c. F. (1974) Supercooling and the crystallizarion ofplagioclase from a basaltic magma. Mineral. Mag., 39,64t_65j.

Hawkes, D. D. (1967) Order of abundant crystal nucleation in anatural magma. Geol. Mag., 104, 473486.

Holloway, J. R. (1971) Internally heated pressure vessels. In G. C.Ulmer, Ed., Research Techniques for High pressure and HighTemperature, p. 217-258. Springer-Verlag, New york.

- (1973) The system pargasite-H2O_COr: a model for melt_ing of a hydrous mineral with a mixed-volatile fluid-L Erperi-mental results to 8 kbar. Geochim. Cosmochim. Acta., 37, 651_666.

Jahns, R. H. and C. W Burnham (1958) Experimental studies ofpegmatrte genesis: melting and crystallization of granite andp€gmatite (abstr.) Geol. Soc. Am. Bull., 69, 1592_1593.

Kirkpatrick, R. J. (1975) Crystal growth from the melt: a review.Am. Mineral., 60, 798-814.

Kushiro, I. (1975) On the nature of silicate rnelt and its signifi_cance in magma genesis: regularities in the shift of the liquidusboundaries involving olivine, pyroxene, and silicate minerals.Am. J. Sci . . 275,4l l43l .

Lofgren, G. E. (1974a) An experimental study of plagioclase crys_tal morphology: isothermal crystallization. Am. J. Sci., 274,243-273.

- (1974b) Temperature induced zoning in synthetic plagio_clase feldspar. In W. S. MacKenzie and J. Zussman, Eds., TheFeldspars, p. 362-376. Manchgster University press, Manches-ter, England.

- and C H. Donaldson (1975) Curved branching crystals

and diflerentiation in comb-layered rocks Contrib. Mineral.Petrol.. 49. 309-3 19.

- and R. Gooley ( 1977) Simultaneous crystallization of feld-spar intergrowths from the melt. Am. Mineral.,62,217-22g.

C H. Donaldson, J. J. Williams, O. Mullins, Jr. and T. M.Usselman (1974) Experimentally reproduced textures and min-eral chemistry of Apollo 15 quartz normative basalts. proc. FifthLunar Sci. Conf., 549-567.

Luth, W. C. and C. O. Ingamells (1965) Gel preparation for hy-drothermal experimentation. Am. Mineral., 50, 255-258.

Maaloe, S and P. J. Wyllie (1975) Water content of a granitemagma deduced from the sequence of crystallization deter-mined experimentally with water-undersaturated conditions.Contrib. Mineral. Petrol., 52, 175-191.

Moore, J. G. and J. P. Lockwood (1973) Origin of comb layeringand orbicular structure, Sierra Nevada batholith, California.Geol. Soc. Am. Bull..84. l-20.

Mustart, D. A. (1969) Hydrothermal synthesis of large single crys-tals of albite and potassium feldspar (abstr.). Trans. Am.Geophys. Union, 50, 67 5.

- (1972) Phase Relations in the Peralkaline Portion of the Sys-tem NarO-Al2Oj-SiO2-H20. Ph.D. Thesis, Stanford Univer-sity, Stanford, California.

Naney, M. T. (19'17) Phase Equilibria and Crystallization in lron-and Magnesium-bearing Granitic Systems. Ph.D. Thesis, Stan-ford University, Stanford, California.

Nockolds, S. R. (1954) Average chemical composition of some ig-neous rocks Geol. Soc Am. Bull.,65, 1007-1032.

Swanson, S. E. (1977) Relation of nucleation and crystal-growthrate to the development of granitic textures. Am. Mineral.,62,966-978.

- (1978) Petrology of the Rocklin pluton and associatedrocks, western Sierra Nevada, California. Geol. Soc. Am. &uil.,89.679-686.

J. A. Whitney, and W. C. Luth (1972) Growth of largequartz and feldspar crystals from synthetic granitic liquids(abstr.). Trans. Am. Geophys. Union, 53, ll2'1.

Turnbull, D. and J. C. Fisher (1949) Rate of nucleation in con-densed systems. J. Chem. Phys., 17,71-71.

Wager, L. R. (1959) Differing powers of crystal nucleation as afactor producing diversity in layered igneous intrusions. Geal.Mag., 96,75-80.

Walker, D., R. J. Kirkpatrick, J. Longhi and J. F. Hays (1976)Crystallization history of lunar picritic basalt sample 12002:phase equilibria and cooling rate studies. Geol. Soc. Am. Bull.,87,646-656.

Whitney, J. A. (1972) History of Granodioritic and Related MagmaSystems: an Experimentdl Stuily. Ph.D. Thesis, Stanford Univer-sity, Stanford, California.

- (1975) The effects of pressure, temperature, and Xrro onphase assemblage in four synthetic rock compositions. J. Geol.,8-r, l-31.

Wyllie, P. J. (1963) Applications of high pressure studies to theearth sciences. In R. S. Bradley, F,d., High Pressure Physics andChemistry, vol. 2, p. l-89. Academic Press, New York.

Yoder, H. S, Jr. (1950) High-low quartz inversion up to 10,000barc. Trans. Am. Geophys. Union, 31,827-835.

Manuscript received, June 18, 1979:acceptedfor publication, December 21, 1979.

Documents

The effect of Fe and Mg on crystallization in granitic systems