American Mineralogist, Volume 76, pages 1279-1290, 1991

Ascent of felsic magmas and formation of rapakivi

HaNN.l, Npxvasrr,Department of Earth and Space Sciences, State University of New York, Stony Brook, New York 11794-2100, U.S.A.

Ansrru.cr

The wide compositional range ofrapakivi-bearing rocks strongly suggests that process,not composition, governs the formation of rapakivi texture. A computational investigationwas undertaken to determine whether decompression can induce the formation of thecommon features of rapakivi, i.e., the presence of mantled ovoid megacrysts of alkalifeldspar (occasionally with inclusions of plagioclase and quartz), embayed early-formedquartz, and the presence of two generations of the phenocrystic phases. Crystallizationpath calculations were conducted for seven rapakivi-bearing magma compositions at pres-sures in the range of l-8 kbar to determine the effect of decompression on the phasestabilities ofplagioclase, quartz, and alkali feldspar. These calculations indicate that de-compression can produce parageneses that can lead to all ofthe characteristic features ofrapakivi if(l) HrO saturation is not attained prior to or during ascent, (2) the cooling rateduring ascent lies in the range of -5-10 "C/kbar, (3) ascent occurs after the magma hasbecome saturated with both alkali feldspar and plagioclase, and (4) the temperature at theonset of decompression is low enough to prevent complete resorption of alkali feldsparduring ascent. These constraints readily explain why rapakivi-trearing rocks are relativelyuncommon in nature, although the compositions that could produce rapakivi are commonand decompression is not an unusual process in the evolution of many felsic magmas.

INrnooucrroN

Many felsic igneous rocks are compositionally similar.The similarities in their bulk compositions and crystal-line assemblages make it difficult to identify specific dif-ferences in their genetic histories. Therefore, any distinc-tive characteristics among felsic rocks (in theirferromagnesian minerals, in feldspar zoning, in their traceelements, etc.) become potential sources of important in-formation. One distinctive characteristic is the presenceof alkali feldspar mantled by plagioclase (rapakivi), whichmay provide a clue to the evolution of certain felsic rocks.Efforts to interpret the origin of rapakivi texture havebeen ongoing since the eady work of Sederholm (1891)on the Fennoscandian rapakivi granites, yet no consensusregarding rapakivi-forming processes has emerged.Nonetheless, such efforts have yielded much petrographic(and more limited analytical and experimental) infor-mation on the characteristics of rapakivi that must beconsidered in evaluating any proposed model of theirgenesis.

Felsic extrusive and intrusive rocks containing rapakivioccur worldwide in spatial and temporal association withrocks as diverse as anorthosites (e.g., Scandinavia; Kranck,I 969), diorites (e.g., New Hampshire; Greenwood, I 95 I ),and silicic rhyodacites (e.g., northern Arizona; Bladh,1980). The general characteristics ofrapakivi granites (andrhyolites) that have emerged from numerous studies ofmany such occurrences are the presence of rounded or

"ovoid" rnegacrysts of perthitic microcline or orthoclase(rarely sanidine), showing incompletely resorbed bound-aries that may or may not be rimmed by oligoclase (wi-borgite-type or porphyritic-type, respectively, using theterminology of Wahl, 1925). (In this article, the term "ra-pakivi" will refer to both types.) Rapakivi-bearing rocksusually contain two generations of quartz and alkali feld-spar (Tuttle and Bowen, 1958). Plagioclase is commonlyfound as phenocrysts and occasionally as crystals poikilit-ically included in alkali feldspar megacrysts (e.g., O'LearyPeak Porphyry, Arizona; Bladh, 1980). Early phenocrys-tic quartz often shows embayment of euhedral forms (e.g.,Gold Butte, Montana; Volborth, 1962). Quartz is alsofound in the cores of megacrystic alkali feldspar (e.g.,Gold Butte, Nevada; Volborth, 1962) and occasionallywithin plagioclase mantles (e.g., Red Beach Granite,Maine; Abbott, 1978). Although biotite (with horn-blende) is the most commonly reported hydrous mineral,muscovite (Maine; Larabee et al., 1965) and riebeckite(e.g., New Hampshire; Greenwood, l95l) have also beenfound. In some cases, hydrous ferromagnesian mineralsare absent and pyroxene is present (e.g., Adirondacks,New York; Buddington and Leonard, 1962). Miaroliticcavities indicating late stage volatile saturation have beencommonly reported in rapakivi granites.

Analysis of feldspar and bulk rock compositions havebeen undertaken by numerous workers and indicate sev-eral general chemical trends. The NarO/KrO ratios of bulk

0003-004x/9 l /0708-l 279$02.00 1279

1280

rocks show wide variation (e.9., 0.42, Enchanted RockBatholith, Texas; Hutchinson, 1956; 1.31, O'Leary Peak,Arizona; Bladh, 1980). However, based on a compilationof compositions of 37 rapakivi granites, Tuttle and Bow-en (1958) concluded that rapakivi granites, in general,have a slightly higher orthoclase content than an averagegranite. Core compositions of alkali feldspar megacrystsare variable, reflecting the variability of the bulk rockNarO/KrO ratios (Anderson, 1980), but most commonlyrange between Oru, and Orro (Bladh, 1980) with low Ancontents. The plagioclase is usually oligoclase (An''-An3s)(Bladh, 1980; Anderson, 1980). Although there is onlyslight variation between the compositions of mantle andphenocryst plagioclase, there is a tendency toward aslightly higher An content in the phenocrysts (Volborth,1962; Ehrreich and Winchell, 1969).

Any process proposed for the formation of rapakivitexture must account for all of the above characteristicfeatures. The compositional diversity (e.9., peraluminousto peralkaline and considerable variation in NarO/KrOratios) and variability in HrO content found among ra-pakivi-bearing felsic rocks (as indicated by the presenceof pyroxene in some and biotite in others) requires a rel-atively composition-independent process. Such a processmust be very different from the one producing incom-pletely resorbed plagioclase rimmed by sodic alkali feld-spar ("anti-rapakivi" texture) that occurs in rocks withina very specific compositional range (including HrO con-tent) (Nekvasil, 1990).

Since the early work of Sederholm (1891), many of thetheories of rapakivi-forming processes that have beenproposed are denied by the current state of our under-standing of phase equilibria of felsic magmas. Tuttle andBowen (1958), for example, attempted to use phase equi-libria in the HrO-saturated haplogranite system to pos-tulate a rapakivi-forming process. They suggested that fora magma to produce rapakivi, the magma must crystal-lize alkali feldspar early, followed by quartz. Further crys-tallization would cause the HrO content of the melt toincrease to the point that the temperature would be lowenough to result in intersection of the alkali feldspar sol-vus. At this point, plagioclase would join alkali feldsparand qtartz, resulting in zonation of the feldspar fromalkali feldspar to plagioclase. Coprecipitation ofall threeminerals would continue until failure of the overlying rockcaused rapid decompression and the formation of a sec-ond generation ofquartz and alkali feldspar. It is gener-ally acknowledged, however, that HrO contents of mag-mas will not increase much beyond the equilibriumsaturation value for any given total pressure, although thekinetics of bubble formation might permit slight super-saturation. Therefore, it is not possible to increase theHrO content of an HrO-saturated melt by 1000/o (or more)by crystallization, as Tuttle and Bowen (1958) suggested.If, instead, crystallization occurred under HrO-undersat-urated conditions at a pressure high enough to result ina ternary eutectic in the haplogranite system upon attain-ment of HrO saturation during unbuffered crystallization,

NEKVASIL: FORMATION OF RAPAKIVI

then plagioclase could eventually be stabilized. However,as will be discussed below, ascent of HrO-saturated mag-ma will not result in the formation of rapakivi texture.

More recent models also recognize the important role

that decompression might play in the formation of ra-pakivi-bearing rocks. Whitney (1975) briefly suggestedthat decompression of certain HrO-undersaturated felsicmagmas could result in crystallization of plagioclase from

a magma that was crystallizing only alkali feldspar. Thiscould be accompanied by partial (incomplete) resorptionof the alkali feldspar. Cherry and Trembath (1978) ex-panded upon this idea by suggesting that decompressioncould produce rapakivi textures but, under either HrO-undersaturated or HrO-saturated conditions, by shiftinga melt from equilibrium with plagioclase, quartz, and al-kali feldspar to just plagioclase and qvartz. However, theprocess requires disequilibrium crystallization at highpressure, large undercoolings during ascent, and rapidcrystallization in a pressure quench process. As will beseen below, however, such a pressure quench process canonly occur if HrO-saturated conditions have been at-tained by the time of emplacement, and it will not lead

to the characteristic features ofrapakivi.To date, none of the proposed rapakivi-forming pro-

cesses has been adequately supported by either experi-mental or computational evidence. Although the role ofdecompression is now generally considered an importantone, no constraints have been placed on this process withrespect to the sequence of appearance of minerals, HrOcontents, requisite magnitude of the decompressionevents, bulk compositions of magmas, or the effects ofdecompression on the compositions of all phases in-volved. The experimental efort that would have to beexpended to evaluate the effects ofthese variables is pro-

hibitive. Therefore, a predictive computational model wasused in the following analysis. This analysis evaluates thedifferences between the equilibrium state of each magmabefore and after decompression. These changes of stateprovide the driving potential for the time-dependent mi-croscopic processes involved in the actual formation ofthe rapakivi textures. Discussion of these time-dependentprocesses themselves, however, is beyond the scope ofthis paper.

The revised quasi-crystalline model of Burnham andNekvasil (1986) and Nekvasil (1986), the ternary feldsparsolution model of Lindsley and Nekvasil (1989), and thecrystallization path calculation methodology of Nekvasil(1988a) were used to investigate the possibility that de-compression could drive a system to produce the char-acteristics of rapakivi. Extensive testing of the model(Nekvasil and Burnham, 1987; Nekvasil, 1988a) indi-cates that it can reliably predict phase relations in thesystem Ab-An-Or-Qz-HrO. Most rapakivi-bearing rockscontain 85-990/o of these components; therefore, the phaserelations calculated for simple analogues should provide

useful information regarding the natural systems.Detailed equilibrium crystallization paths were calcu-

lated for seven rapakivi-bearing granites [Enchanted Rock

Batholith (ERB), Texas, Hutchinson, 1956; O'Leary peakPluton (OLP), Aruona, Bladh, 1980; Wolf River Batho-lith (WA-WRB), Wisconsin, Anderson and Cullers. l97g:Gold Butte (GB), Nevada, Volborth, 1962: Butler Hill(BH) granite, Missouri, Bickford et al., l98l: Deer Island(DI) and Crotch Island (CI), Maine, Stewart, 19591 in thepressrue range of l-8 kbar and several HrO contents. Thevariations in proportions and compositions of all phasesat these pressures, when compared with petrographic andchemical analysis of natural rapakivi-bearing rocks, per-mit the evaluation of the role of decompression in theformation of rapakivi texture.

Prr,lsn R-ELATToNS oF RApAKrvr-BEARTNG MAGMAsEquilibrium crystallization paths for seven rapakivi-

bearing rocks from the localities listed above were cal-culated at several pressures and HrO contents. These pathsassume that the bulk composition of each rock is similarto the magma from which it crystallized. The validity ofthis assumption is difrcult to assess; however, if thesebulk compositions can produce all the characteristics ofrapakivi by the same process, the process will have beenshown generally to be viable (although not necessarilyspecific to a given rock). If the rocks represent fraction-ates from a batholithic suite, then the unit investigatedcould represent a liquid. If some of the crystalline mate-rial present in a unit represents restite, then, as long as itwas in equilibrium with a melt, the calculated equilibri-um crystallization path of the bulk composition is stillvalid. If the rocks represent cumulates, then it should bevery difficult to correlate the calculated crystallizationpaths (which are based on the bulk compositions) withthe petrography. As will be shown below, this was not aproblem for any of the compositions analyzed,. The as-sumption is of most concern if there is evidence for as-simi.lation of xenoliths after the megacrystic (early-formed)phases had crystallized. The compositions used in thisstudy were selected to minimize the chances of late con-tamination playing a role (although this may have oc-curred in OLP; Bladh, 1980).

The bulk compositions for the seven samples were re-cast into normative Ab, Or, An, ez for the modeling.Among the samples, these components summed between86 and 99 wto/o, with the majority lying near 950/0. Asseen in Table I (second column), the samples indicate asignificant compositional range with renormalized nor-mative Ab ranging from 28 to 44 wto/o, Or from 24 to 39wto/0, Qz from 22 to 33 wto/0, and An from 6 to 11 wto/0.Such compositional variability should result in consid-erable differences in saturation temperatures ofthe threecrystalline phases considered, their sequence of appear-ance, and even their compositions.

Sequence of appearance of crystalline phasesTable I shows the results of calculations conducted for

the seven compositions at 2 ard 5 kbar and 2.8 wto/o g0molo/o) bulk HrO (in the system). This HrO conrent (be-tween 2 and 3 wtVd was chosen as a demonstration be-

NEKVASIL: FORMATION OF RAPAKIVI l 2 8 l

cause it lies approximately midway between the dry andHrO-saturated values at low pressure and is that expectedfor equilibrium with biotite or amphibole (Burnham,1979). The temperature of appearance of the liquidusmineral, the saturation temperatures of the second andthird crystalline phases, and the HrO-saturated solidustemperature are given for each composition. Addition-ally, the compositions and proportions of crystallinephases are given in order to indicate the effects ofdiver-sity in bulk compositions on the phase relations.

As anticipated, the compositional variability among therocks selected for modeling results in considerable differ-ences in saturation temperatures and sequence of ap-pearance of the minerals plagioclase, quarlz, and alkalifeldspar at both 2 and 5 kbar. An important invariantcharacteristic, however, does emerge for all these com-positions. Alkali feldspar is not the liquidus phase for anyof these compositions at any pressure (between I and 8kbar) or HrO content (from dry to HrO saturated). Sev-eral additional compositions (showing rapakivi texture,but not indicated in Table l) were also investigated andyielded the same result. Plagioclase is the liquidus phasefor all compositions in Table I at both 2 and 5 kbar.Additional investigation of another BHG composition (no.I 77 in Bickford et al., I 98 l) shows quartz and plagioclaseon the liquidus at 2 kbar and quartz on the liquidus at 5kbar, indicating that plagioclase need not be the liquidusmineral. In fact, just as plagioclase or quartz could be theliquidus mineral for certain bulk compositions, there isno reason that alkali feldspar could not be the liquidusmineral for compositions with sufficiently low bulk Ancontents. Additionally, differences in pressure for a givenbulk composition can change the identity of the liquidusphase for rocks in which the temperature of saturation ofquartz and a feldspar are in close proximity.

The differences in bulk compositions among thosestudied also result in differences in the sequences of ap-pearance of the nonliquidus phases, quartz and alkalifeldspar (see Table l). The sequence of appearance ofthese two minerals is a function of not only the bulkcomposition but also of the pressure and HrO content.As indicated by the experiments of Whitney (1975) andthe calculations of Nekvasil and Burnham (1987), theeffect ofisobarically increasing HrO content is to increasethe thermal stability of alkali feldspar relative to quartz,and the effect ofincreasing pressure at constant HrO con-tent is to decrease the stability of feldspar. Changes inthese variables can induce a change in the sequence ofappearance of quartz and alkali feldspar (see ERB in Ta-ble l). As inferred from additional calculations using thecompositions of Table l, bulk HrO contents of 0-3 wto/owill generally result in quartz appearing before alkali feld-spar; however, for higher bulk HrO contents, alkali feld-spar will commonly be the second crystalline phase toappear.

Several workers have previously suggested that alkalifeldspar is not necessarily the liquidus mineral in rapa-kivi granites. Based on phase equilibrium constraints,

1282 NEKVASIL: FORMATION OF RAPAKIV]

TABLE 1 . Calculated phase relations at 2 and 5 kbar for seven rapakivi-bearing rocks with 2.8 wl"/" HrO

Bulk compositionAb, Or, An, Qz

Sample (wt%)

Liquidus Liquidus phase Second mineralPressure temperature composition saturation I Phase compositions

(kbaO fc) (wt%) fc) (wt%)

ERB Absoor3oAn,,Qzru 5

2

OLP AbnoorroAn"Oz26

WA-WRB Ab3,OrseAn8Qz,,

GB Absror34AnloQz2s 5

BHG Ab3ror26An6ozs 5

Dl Ab34orrsAn,,Qz'

Cl Ab35Or27An7Oz3, 5

2

1 062 An6oAb,.Or,

1038 An8rAb,,Or,

5 1001 AnToAb2eorl

2 978 An66AbsOrl

882 QZ'ooAnu.Ab*Or,

809 Or'3Abr4An3Anl5Absoors

887 QZ'ooAneAb$O13

803 QZ'ooAnr5AbsorT

890 Or,sAb,6AnnAn6rAbsO13

873 Or,rAbroAnoAn5eAboTOrn

905 QZ'ooAnr6Ab4Or,

817 Qz,*An5sAb3eO13

944 QZ,ooAnruAbroOr,

844 Qz,mAn6sAbsoOr,

900 QZ,ooAn."Ab.'Or,

813 QZ,*An4eAb46Org

930 QZ'ooAnr.AbruOr,

841 QZ'ooAnuoAbo"Or,

5 1038

2 1012

1 050

AneAbj6Orj

An.rAb,rOr,

An*Ab,,Or,

2 1026 An.uAb,.Or,

981 AnsoAb'rorr

2 944 An,5Abroor,

5 1069 Ans6Abisorl

2 1045 AneAb,"Or,

999 An61Ab1BOr,

977 An,,Ab22Or'

Stewart (1959) concluded that, for DI and CI, precipita-tion ofplagioclase and quartz preceeded precipitation ofalkali feldspar. He indicated further that this conclusionwas supported by the bulk compositions and petro$aphyof the DI and CI rapakiyi granites. Bladh (1980) con-cluded that plagioclase was the liquidus phase for theO'kary Peak pluton, basing her conclusions on an ex-perimental investigation of the OLP. The calculated re-sults support these conclusions but indicate further thatrapakivi-bearing magmas, in general, may rarely crystal-lize alkali feldspar as the liquidus mineral. Petrographicobservations of inclusions of plagioclase + qtartz withinmany mantled alkali feldspar megacrysts lend support tothe calculated results (e.g., plagioclase poikilitically in-cluded in microcline, ERB, Hutchinson, 1956; oligoclaseinclusions in megacrystic sanidine, OLP, Bladh, 1980;inclusions of both plagioclase and quartz in alkali feld-spar, WA-WRP, Anderson, 1980; quartz inclusions with-in mantled alkali feldspar, GB, Volborth, 1962).

The early phenocryst-forming stage of crystallization

To evaluate decompression as a mechanism for pro-

ducing all the features of rapakivi, the assemblage beforeand after decompression must be considered. For rocksthat contain two generations of crystalline phases, it isreasonable to presume that the early phenocrystic (or

megacrystic) phases indicate the assemblage during thehigher pressure stage of crystallization, and the matrixphases are those produced after emplacement, in the laterstage of evolution. For all the compositions investigated(with the exception of ERB, which lacks early quartz),

two generations ofquartz, plagioclase, and alkali feldsparare reported, indicating an early assemblage containingall three crystalline phases. Comparison of the reportedassemblages with those calculated in the pressure rangel-8 kbar can be used to obtain a reasonable pressure atwhich to model the formation of the early assemblages.In view of the wide compositional ranges of the feldspars

NEKVASIL: FORMATION OF RAPAKIVI r283

Taate 1-Continued

Phase Third mineral phase Solidus phaseproportions saturation f Phase compositions proportions temperature Phase compositions proportions

(wto/") (.C) (wt%) (wt./.) fC) (vur.i.) (wt%)

' t2

01 0

09

01 9

842

797

Or75Ab22An3Qz,ooAn..AborOrnOr7sAb2.An3QZ,ooAnooAbu,Or"O16rAb35An3Qzr*An2sAb67O11oOr.rAb*An.Qz,ooAn,.AbroOr,,OrTBAbleAn3Qz,ooAnruAbo,Or"Or75Abr2An3Qzr*AnJb41Or4Or7eAb16An3QZ,ooAn*Ab"uOr.OrrrAbroAn.QzrooAnszAb@O13OrTrAbr6AngQzroAnsAbso16Ors2Ab€An,QZ,ooAn*AbuoOr,OrT5AbzAn3QzrmAn4sAbnTO14OrTlAbmAn3Qzr.Ani4Ab5lOrsOrr"Ab."An.Qz,*AnorAb*OruOr6eAb2oAn3QZ,ooAn.rAbu"Oru

OrTTAb2iAnlQzr*An6AbsorsOrZAbrrAn,QzrooAn27Ab66An7OrTrAb26AnlQzroAni2AbTsOrroOru"Ab.uAn.Qz,*Anr2Ab7sor,3OrzAbzAnrQzrooAn.uAb"rOr.Or7rAb26An2QzrooAn."Ab*Or,OrTeAbroAnlQz,*An3lAbqorsOf ,oAb"oAn"QZrooAnsAb6rOr5OrTiAb25AnlQzr.An,uAbr"Or,Oru"Ab.rAn"Qz,*Anr7Ab720rr1Orz6AbzlAnrQz,.An"rAburOr"Or72Ab6An2QZrooAnr6Ab65O17Or74AbrsAn1QZ,*An,"AbroOr.OrsAb3oAn,QZrooAnr"AbrrOr,o

04

1 820

2107

230e

271 00

1 1210

1 50b

1 40

1 50

1 31 30I

1 50

2102

230

1 01 40

1 6

680

b / c

372538382537262648272647482230522226402832432829313336343333322741352738333136363133

8 1 6

724

715

77'l

817

0I

01 0

0o

01 4

02

0c

01 2

020

0J

01 0

in these rocks and the limited effect of pressure on feld-spar composition, such comparison does not provide aprecise means of pressure determination. However, a dis-tinct range does emerge from such comparisons.

For the OLP, Bladh (1980) reports an early phenocrys-tic stage consisting of plagioclase [Anru_,,Abro_roOro_n(molo/o) phenocrysts and An2 r_ r 4Ab r r_r rOr' _ro inclusions inalkali feldsparl, qvartz, and alkali feldspar (Or.r_urAb.o_ro-An, megacrysts). The best agreement between these com-positions and the calculated feldspar compositions (in thetemperature range of multiple saturation with plagioclaseand alkali feldspar) is found in the pressure range 4-5kbar. For example, at 5 kbar, the calculated plagioclaseand alkali feldspar compositions (from the onset of alkalifeldspar crystallization to the solidus temperature) areAnrr_,,Abur-rnOrn_,0 and 0161_70Abru_rrAnr_, (in molo/0,converted from Table l), respectively. It may appear un-usual that the calculated Or content of the alkali feldsparincreases with decreasing temperature. It must be re-

membered that the HrO content of the melt is continu-ously increasing and the lowered temperature results in arotation of two-feldspar tie lines to more Or-rich alkalifeldspar; additionally, the rapid decrease in Qz and Ab(through the precipitation of very Ab-rich plagioclase) willconcentrate Or in the melt.

Similar comparisons of calculated and reported feld-spar compositional ranges for compositions GB, BHG,and CI indicate that pressures between 4 and6 kbargivethe best agreement. ERB and DI, however, vary from thisgeneral trend. Hutchinson (1956) reports an early assem-blage consisting of andesine (and oligoclase) and micro-cline without quartz for ERB. For this composition at 5kbar and above, quartz is calculated to be the secondcrystalline phase to form for all HrO contents other thanthat close to saturation. The absence ofquartz, therefore,indicates a pressure of formation of the phenocrystic stageof significantly less than 5 kbar; the best agreement be-tween the reported and calculated compositions is found

t284

a

NEKVASIL: FORMATION OF RAPAKIVI

,aQzzs

Fig. l. Calculated equilibrium crystallization path at 2kbar (dashed) and 5 kbar (solid) for a rapakivi granite fromCrotch Island, Maine, with renormalized compositionAbrorOrrroAnueQz3,, (wto/o) given by the stars in a and b andbulk H,O content of 2.8 wt0/0, projected (a) into the haplogranitesystem and (b) into the feldspar system. The crystallization pathsare for HrO-unbuffered conditions where the HrO content in themelt increases during crystallization of anhydrous minerals. Thesymbols P (plagioclase), Q (quartz), A (alkali feldspar), and V(vapor) indicate the onset of appearance of each phase. S indi-cates the melt composition at the solidus temperature. The seg-ment A-S indicates the variation of the melt when plagioclase,quartz, and alkali feldspar are coprecipitating (that is, as the meltmoves across the four-phase surface in the ganite system). Thecompositions ofthe feldspars at the onset ofprecipitation ofthethird phase and both feldspars at the solidus are shown for 5kbar (triangles 5 and 5') and 2 kbar (triangles 2 and 2', respec-tively) in b.

aI -2kbar.Inasmuch as this rock is similar in bulk com-position to the other rocks (most notably, Dl-see Table

l), the absence of quartz further indicates that the per-

turbation event that disrupted the formation of the earlyphenocrysts occurred at an eailier stage of the differenti-

ation of the magma than for the other granites modeled.

Although no analytical data are available on the exact

compositions of the phenocrystic feldspars, this assertion

is supported by the more calcic nature of andesine phe-

nocrysts and mantles relative to the other granites (in

which oligoclase is the most An-rich plagioclase). DI

shows similar results, with the best agreement between

the reported and calculated compositions at -2 kbar, al-

though in this case, the perturbation event occurred after

the magma became multiply saturated with all three crys-

talline phases.From these comparisons it appears that the pressrue of

formation of the early phases is not of major importance

r050

1Ofi)

20 30 40 50 60 70 80 90 100Phase proportions (w t.Vo )

1050

0 10 20 30 40 s0 60 70 80 90 100Phase proportions (wt.Vo)

Fig. 2. Variations in the proportions of all phases duringequilibrium crystallization are shown in (a) for 5 kbar and in (b)for 2 kbar for 2.8 v"1o/o HrO in the system for the paths in Figuresla and lb. The proportions ofthe crystalline phases (+ hydrousmelt) have been renormalized to 1000/0 after HrO saturation isattained and vapor begins to exsolve. (Symbols: Pl, plagioclase;Af, alkali feldspar; Qz, quartz; L, hydrous melt.)

to the formation of rapakivi textures. However, since thephenocryst phase compositions for all bulk compositionsexcept DI and ERB are consistent with an initial pressure

of 5 kbar, this pressure was chosen as the pressure at

which to model the formation of the early higher pressurephases. In order to determine the changes that would oc-

cur if the phenocrystic assemblages were perturbed by

ascent of the magma, we must first be able to understandhow the magma would have evolved isobarically. The CI

composition is intermediate among the compositionschosen for study and will be used for demonstration pur-poses; the general trends discussed below, however, holdfor all the compositions investigated.

9s0

Oe- 900o

; 8s0LI

E r mo - - -F

750

1000

950

e- 900

; 850!O

F 8ooF

750

700

650

10

An4oAbssors a

An36Ab58Oro

An36Ab58O16

An33Abs1016An34Ab60O16

An32Ab62Or6An31Ab62Or7

An29Ab64Or7

Anl 9Ab71 On 0 An27A667016

lP=sk t . lAnl 9Ab74Or7

An16Ab76OrB )

850

NEKVASIL: FORMATION OF RAPAKIVI l 285

Figures la and lb show the evolution of the melt andcrystals of a magma with composition CI, projected intothe haplogranite system (Fig. la) and into the feldsparsystem (Fig. lb) at 5 kbar (solid curve). The melt evolu-tion path consists of four main segmental trends; the in-flections indicate the appearance of a new phase. Movingfrom higher to lower temperature (i.e., from the bulkcomposition given by the star in Figs. la and lb), thecrystallization of plagioclase Oeginning at P) results indepletion of An and Ab components and enrichment inOr and Qz. Once qtrartz appears (at Q), only Or will besignificantly enriched in the melt. Finally, when alkalifeldspar begins to precipitate (at A), the Qz content of themelt will increase until HrO saturation is reached (at V)and the melt moved along the HrO-saturated five-phasecurve in the simple granite system (Nekvasil, 1988b) tothe solidus composition (S). Figure 2a shows the varia-tions in phase proportions during equilibrium crystalli-zalion at 5 kbar. The labeled inflections seen in Figuresla and lb correspond to the appearance ofeach new phase(including vapor at the lowest temperatures) in Figure 2a.The calculated compositions of the solid solutions at theonset ofprecipitation ofeach new phase and the solidustemperature are shown in Figure lb and Table l.

Figures 3a, 3b, and 3c show expanded views of thecalculated variations in plagioclase, alkali feldspar, andquartz proportions, respectively, once all three crystallinephases have begun to coprecipitate. Atkali feldspar is thelast phase to begin precipitating. Its abundance will,therefore, vary from zero to its solidus proportion. Somequartz and plagioclase are already present when alkalifeldspar appears; their proportions, therefore, will not bezero at the highest temperature terminations of the phaseproportion curves. The inflection in each curve ofFigures3a-3c marks the onset of vapor exsolution (HrO satura-tion). Once the magma is HrO saturated, cotecticlikecrystallization occurs, resulting in large changes in phaseproportions over a small temperature interval. Thesetrends in variation of melt composition and phase pro-portions during isobaric crystallization are similar for allthe compositions studied and generally characterize equi-librium crystallization paths of felsic magmas that are notHrO buffered (Nekvasil, 1988b).

During isobaric crystallization along the four-phasecurve in the granite system at 5 kbar, the phase propor-tions of plagioclase, quarlz, and alkali feldspar all in-crease, albeit to different extents. For the composition CI

(-Fig. 3. Expanded view ofthe variation in phase proportions

during equilibrium crystallization at 2 and 5 kbar for (a) plagio-clase, (b) alkali feldspar, and (c) quartz. The variation in phaseproportions in the system is shown only for crystallization acrossthe four-phase surface (segrnents A-S in Fig. la) when the meltis multiply saturated with all three crystalline phases. Compo-sitions indicated are given in weight percent. The arrow indicatesthe variation in phase proportions after decompression from 5to 2 kbar starting at 780 € with a 7 'Clkbar gradient.

U

L

E zso

F

Q

6 / \ t l

0

F

0 5 1 0 1 5 2 0 2 5 3 0 3 5Plagioclase abundance in the system (wtZo)

0 5 1 0 1 5 2 0 2 5 3 0 3 5

Alkali feldspar abundance in the system (wt%)

s 1 0 1 5 2 0 2 5 3 0 3 5

Quartz abundance in the system (wt%)

850

c)

E 7so

F

P=5kba r

3 bOr73Ab24An3

O169Ab28An3. \Ot71Ab26An2

Ot74Ab24Anz

O172Ab26An2

Ot76A622 An2

- t \Or73Ab25An1

c

r286

at the onset of alkali feldspar saturation (at 5 kbar), thesystem consists of 7 4 wto/o melt and I 30/o each of plagio-clase and quartz (see Table l). However, 50 { below thealkali-feldspar saturation temperature, the proportion ofalkali feldspar rises to l0 wto/o of the system, whereas thatof plagioclase rises to 18 wt0/0 and of qvartz lo 15 wto/0.The greater increase in proportion of alkali feldspar perdegree of cooling (also shown by the shallow slope of thealkali feldspar curve in Fig. 3b) may result in the earlyformation of large crystals of alkali feldspar.

If ascent of a magma is required for the formation ofrapakivi texture, then the viscosity must remain lowenough during the early stages of crystallization to permitthe magma to rise through the crust. The increase in crys-tallinity associated with cooling of a magma is the singlemost important factor in increasing the viscosity of themagma. It is, therefore, important that the magma be-come multiply saturated with the phenocrystic phasesearly in its differentiation history. Table 1 indicates thecalculated variation in abundance of the crystalline phasesas each magma cools from the liquidus to the solidus at5 kbar. Importantly for all these compositions, multiplesaturation with plagioclase, alkali feldspar, and quartz oc-curs when only 20-30 wto/o of the system is crystalline.

The matrix-forming stage of crystallization

If the formation of a second generation of crystallinephases in rapakivi-bearing rocks is a result of perturba-tion caused by decompression, then the compositions ofthe matrix phases and the relative proportions of all phases(matrix and phenocrystic) should reflect low-pressurecrystallization. However, unlike for the megacrysticphases, analytical data on the matrix feldspars are scarce,and this scarcity precluded use of the matrix composi-tions to estimate emplacement pressure. Therefore, forthe analysis below, an emplacement pressure of 2 kbar isused as an example.

To identify differences between high- and low-pressurecharacteristics in the natural rocks, we must first look atvariations in crystallization paths at different pressures.Figures la and lb show the evolution of the CI bulkcomposition at 2 kbar juxtaposed against lhat at 5 kbar.Although the general trends are similar, a major differ-ence in the Qz content of the evolving melts is apparent.As discussed in Nekvasil (1988b), the position of the HrO-undersaturated four-phase surface in the granite systemshifts to higher Qz contents with decreasing pressure. Ifthe matrix phases represent melt after reequilibration atlower pressure, then the composition of the matrix shouldreflect a higher Qz content than that of the melt that wasin equilibrium with the early formed phenocrystic phases.

In addition to the differences in the Qz content of meltscrystallizing at 2 and 5 kbar in equilibrium with all threecrystalline phases, there will also be slight diferences inthe An:Ab:Or ratios in the crystalline solid solutions atthe two pressures. Figure lb shows the calculated com-positions of plagioclase and alkali feldspar when the meltreaches the four-phase surface (where an H2o-undersat-

NEKVASIL: FORMATION OF RAPAKIVI

urated magma is in equilibrium with plagioclase, quartz,

and alkali feldspar) at 5 and 2 kbat (triangles 5 and 2)and at the solidus temperatures (triangles 5' and 2'), re-spectively. The differences in solidus feldspar composi-tions (i.e., higher Or content of the alkali feldspar butlower An content of the plagioclase at 5 kbar comparedwith those at 2 kbar) reflect the effect of pressure andtemperature on the ternary feldspar solvus. Importantly,differences in the compositions of the solid solutions (pla-gioclase and alkali feldspar) are slight between the twopressures (as shown in Figs. 3a and 3b)' These variationsresult in only slight differences in the solidus proportions

ofplagioclase and alkali feldspar (shown by the propor-

tions at the terminations of the variation curves in Figs.3a and 3b). (These small diferences in the high- and low-pressure solidus phase proportions make it difficult to usebulk modal analyses to determine emplacement pressures

because these variations in proportions lie within the un-certainty of most modal analyses.)

The calculated variations in phase proportions for CIat 2 kbar are shown in Figure 2b. The magnitude of thedestabilizing effect of decreasing pressure on the crystal-line phases varies considerably among the three crystal-line phases. For quartz, lowering the pressure results inthe greatest destabilization, which can be readily seen bycomparing the quartz saturation temperatures aI 5 and 2kbar in Figure 2a (930 and 841 "C, respectively). Foralkali feldspar, this destabilization results in a saturationtemperature decrease from 835 "C at 5 kbar to 798"C atZkbar. Plagioclase shows the least effect with only a 22

"C decrease in saturation temperature. Table I indicatesthe effect ofdecreasing pressure on the mineral saturationtemperatures for all the composilions studied. For eachcrystalline phase, the extent of destabilization with de-creasing pressure is generally independent of the bulkcomposition. For plagioclase, the extent of the destabili-zation is 8 t 1 'Clkbar (higher for BHG), for alkali feld-spar, 12 + I "C/kbar (lower for WA-WRP), and for quartz,29 + 4'Clkbar. It is these differences (in the extent ofdestabilization) among the crystalline phases that permit

the formation of rapakivi textures by decompression.

DncovrpnnssroN oF 2 rBr-pspLn + QUARTZ+ MELT (+ v.rpon) ASSEMBLAGES

From petrographic evidence for all the samples, a per-

turbation event resulted in partial resorption of pheno-

crystic alkali feldspar (as evidenced by ovoid alkali feld-spar in all the examples of Table l), partial resorption ofphenocrystic quartz (as evidenced by rounding and em-bayment in all but ERB), and continued precipitation ofplagioclase (based on the absence ofresorption textures,the persistence of euhedral forms on the phenocrysts, andthe formation of mantles on the alkali feldspar). A secondgeneration of each phase that crystallized earlier formedat the later stage.

For all compositions except ERB, the perturbation event(here modeled as decompression) must have occurred af-ter the magma was multiply saturated with quartz, pla-

gioclase, and alkali feldspar. The effects ofdecompressionon these early formed phases in the rapakivi-bearing rockswere determined by investigating the changes induced inthe phase proportions and compositions for several de-compressional end points. These effects indicate the con-ditions under which decompression can produce rapaki-vi.

Thermal history

The simplest decompression thermal process of a mag-ma, although unlikely to occur in nature, is isothermaldecompression. If decompression from 5 to 2 kbar occursisothermally, then the variations in phase compositionsfor CI can be obtained simply from Figures 3a-3c. If, forexample, the system is at 780 "C at 5 kbar (with a phaseassemblage of l2o/o alkali feldspar , 160/o quartz, and 180/oplagioclase) at the onset of decompression, the new as-semblage at 2 kbar and 780 "C would consist of 5 wto/oalkali feldspar, 8o/o quarlz, and l7o/o plagioclase, if equi-librium were eventually reestablished at the lower pres-sure. [Although this temperature was chosen for ease ofdemonstration, it is interesting that feldspar geother-mometry (of Fuhrman and Lindsley, 1988) at 5 kbar forthe phenocrystic alkali feldspar and plagioclase pairOruoAbroAn, and An,rAbrrOrn, which lies within the rangepresented by Bladh, 1980 for the OLP, yields a temper-ature of780 "C.l In this case, partial resorption ofall threephases would take place, with plagioclase undergoing asmall but definite amount of resorption. The decrease inphase abundances during decompression is nearly inde-pendent of the temperature at which ascent begins (asseen by the subparallel nature ofthe 2 and 5 kbar curvesin Figs. 3a-3c) until the HrO content at 5 kbar has reachedthe 2 kbar saturation value (5.7 wto/o HrO in the melt forCI). If this were the case, further crystallization at 5 kbarbefore ascent would result in quenching of the magmaduring ascent because of the lower HrO solubility andhence higher solidus temperature at lower pressures (ascan be seen in Figs. 3a-3c).

Ascent of natural magmas is most likely to take placewith some finite loss of temperature. This can be adia-batic or nonadiabatic (and dependent upon the thermalregime, ascent rate, mode of ascent, and conductivity ofthe crust through which the magma is rising). The com-bination of low magmatic temperature at the onset ofascent and prior heating ofthe wall rocks could result inonly a small average temperature loss during ascent, forexample, 3-5 "C/kbar. As for isothermal decompressionin this case, ascent must not commence before the tem-perature is low enough to retain all phases after decom-pression. However, because ofthe cooling during ascent,decompression could commence at a slightly higher tem-perature than for isothermal decompression without re-sulting in complete resorption of alkali feldspar. In Fig-ures 3a-3c, for example, decompression could commenceat or below 800 'C. If, for demonstration, we consider atemperature of 780 'C at the onset of decompression at5 kbar and a temperature drop of3 "C/kbar during ascent,

NEKVASIL: FORMATION OF RAPAKIVI t287

then the phase assemblage changes from 12 wto/o alkalifeldspar, 160/o quartz, and 180/o plagioclase to 7o/o alkalifeldspar, 9o/o quartz, and l8o/o plagioclase at 2 kbar. Incontrast to isothermal decompression, the proportions ofquartz and alkali feldspar decrease and that ofplagioclaseremains the same.

If the wall rocks were warm and ascent were sufrcientlyrapid, little heat would be lost from the magma duringascent and the temperature decrease of the magma wouldbe close to adiabatic. Rumble (1976) and Sykes and Hol-loway (1987) calculated isentropic adiabats for hydrousalbite melt in the presence of crystals (+ vapor) and ob-tained gradients of -7 'Clkbar. Using this value as anapproximation for the multiphase, multicomponent adi-abatic gradient, decompression from 5 to 2 kbar, begin-ning at 780'C, would result in a lower pressure assem-blage consisting of 9 wto/o alkali feldspar, l0o/o qtartz, andl9olo plagioclase. This change in phase assemblage is con-sistent with what is observed in rapakivi-bearing rocks,that is, little perturbation of plagioclase, some resorption(rounding) of alkali feldspar, and more major embaymentof quartz. As can be readily seen in Figures 3a-3c, av-erage cooling rates greater than -10 "C/kbar will resultin precipitation of alkali feldspar instead of resorption.

From the examples above, it is readily noted that cool-ing rate during ascent plays a critical role in determiningwhether decompression will result in the characteristicsof rapakivi. The maximum extent of resorption of alkalifeldspar and quartz upon decompression occurs with onlysmall amounts of cooling during ascent. However, verysmall amounts of cooling would also result in some re-sorption ofplagioclase in addition to alkali feldspar andquartz. The precipitation of plagioclase can be maxi-mized with larger cooling rates, but large cooling ratesresult in precipitation rather than resorption of alkalifeldspar. The optimum cooling rate for the production ofthe characteristics of rapakivi is, therefore, only slightlyhigher than adiabatic, implying little heat loss to the wallrocks during ascent.

Although this discussion focuses on a closed systemprocess, Figures 3a-3c can also illustrate the effects ofanopen system process such as a magma chamber rechargeevent. In this case, the temperature of the magma wouldincrease isobarically, perhaps because ofthe influx ofhot-ter, less differentiated magma. As can be readily seen inFigures 3a-3c, such an event would result in partial re-sorption of all crystalline phases, including plagioclase,and hence prohibit the formation ofrapakivi texture.

The importance of HrO content

The above discussion focused on the ascent of HrO-undersaturated magma in which the magma remainedHrO undersaturated at the time of emplacement at thelower pressure. As was discussed for isothermal decom-pression, because of the lower HrO solubilities at lowpressure, it is possible that an HrO-undersaturated mag-ma could attain HrO saturation during ascent. This be-comes increasingly likely with large degrees of cooling

I 288

during decompression. If the magma begins decompres-sion at 7 40 "C and 5 kbar and cools at an average rate of7 "C/kbar, the magma would be HrO saturated once itreached 2 kbar (at719 'C). In reflection ofthis, the phaseassemblage would change from l7o/o alkali feldspar, 190/oqtrarlz, and 2lo/o plagioclase to 25o/o alkali feldspar, 27o/oqvartz, and,28o/o plagioclase (as can be seen in Figs. 3a-3c). No partial resorption of any crystalline phase wouldtake place during ascent; instead, a considerable amountof crystallization (230lo) would occur. Slightly larger de-creases in temperature during decompression or a slightlylower temperature at the beginning of decompressionwould result in the magma cooling to the solidus tem-perature before 2 kbar were reached.

IfHrO saturation is attained prior to or during ascentor upon emplacement, a pressure quench phenomenonmarked by rapid crystallization may take place that couldresult in a matrix-forming stage of evolution of the mag-ma. However, as discussed above, this would not be ac-companied by partial resorption of either quartz or alkalifeldspar. Therefore, unlike Cherry and Trembath (1978),this author concludes that such a quench process wouldnot produce rapakivi textures.

Magnitude of the decompression event

Decompression from 5 to 2 kbar has been discussedabove. However, the main trends hold for any amountof decompression. Small changes in temperature are mostconducive to the production ofrapakivi textures, yet smallchanges in temperature also increase the likelihood of theloss of a phase during decompression, particularly for largechanges in pressure. The lower the temperature at theonset of decompression, the less likely the loss of a phasebecomes. However, the lower the initial temperature and,hence, the higher the degree of crystallinity, the morelikely it becomes that the magma will become HrO sat-urated before emplacement or that the viscosity becomestoo high to permit much ascent. Thus, the conditionsunder which the formation of rapakivi is possible becomeincreasingly restricted with larger pressure drops.

Compositional changes during ascent

One of the characteritics of rapakivi-bearing magmasis that, although texturally there is evidence of a majorperturbation event, the compositions of the feldspars donot show marked discontinuities. In fact, for the com-positions investigated, the large alkali feldspar ovoids aresimilar in composition to the matrix alkali feldspar, andplagioclase inclusions (within the alkali feldspar), earlyphenocrysts, and mantles are often indistinguishable incomposition. Stewart (1959), for example, presents com-positions of alkali feldspar in the groundmass for CI,which, when compared with his megacryst compositions,yield a maximum change in the feldspar compositionsresulting from the perturbation event. He reports thepresence of ovoid Orro alkali feldspar megacrysts and Or.nalkali feldspar in the groundmass. (The outer zones ofunmantled phenocrysts lie between Or, and Oru'.) Bladh

NEKVASIL: FORMATION OF RAPAKIVI

(1980) reports compositions of early plagioclase in OLPwithin the range Anr,-,oAbrr-rrOrr-ro, whereas clear pla-gioclase mantles on the alkali feldspar &re of compositionAn,oAbruOrr, also indicating very little change. Althoughsubsolidus reequilibration of alkali feldspar may have oc-curred to smooth compositional diferences, ths sltrggishkinetics of diftrsion in plagioclase (Johannes, 1978) in-dicate that such reequilibration ofplagioclase is unlikely.

The calculated results indicate that these small changesin feldspar compositions are fully consistent with decom-pression as the perturbation event. Figures 3a and 3bindicate the compositions of plagioclase and alkali feld-spar at selected points along the phase assemblage curvesfor the CI bulk composition. Taking as an example adecompression event from 5 to 2 kbar starting at 780'Cwith a temperature drop of 7 "Clkbar, the plagioclase

composition changes from AnroAbuoOru to An'AburOrt(as shown in Figs. 3a and 3b). The calculated alkali feld-spar composition changes from Or- AbroAn, toOr,AbrrAnr. These changes are very similar to those re-ported by Stewart (1959). All of the compositions inves-tigated similarly show only small differences in the cal-culated (and reported) compositions of the phenocryst

and matrix feldspars.The effect of decompression from 5 to 2 kbar on the

melt composition of CI can be readily seen in Figures laand lb. As discussed above, the increase in silica contentof the melt at lower pressures reflects the difference inposition of the four-phase surface (plagioclase, quattz, al-kali feldspar, and melt) at 5 kbar and 2 kbar. If the matrixofeach ofthe rapakivi-bearing rocks represents the liquidafter decompression, then its composition should lie onthe low-pressure four-phase surface in the granite system.The increase in silica content of the melt that would berequired to establish a new equilibrium could be achievedby partial resorption of phenocrystic quartz. Embaymentof quartz is a common characteristic of rapakivi-bearingrocks and is reported for all of the compositions studiedexcept ERB.

CoNcr.usroNs

Investigation ofthe phase stabilities ofseveral rapaki-vi-bearing magmas indicates that decompression canproduce all of the features of rapakivi texture. This in-cludes partial resorption of alkali feldspar and quartz,continu€d crystallization of plagioclase, small changes inalkali feldspar and plagioclase compositions, and the for-mation of two generations of the early formed pheno-

crystic phases. Many felsic magmas are considered to havecrystallized at a higher level than that at which they wereformed; yet, rapakivi-bearing rocks are rare enough to benoteworthy. Ascent alone, therefore, will not necessarilyproduce rapakivi. Instead several constraints on the de-compression event must be met.

l. The bulk composition must be sufficiently Or richto permit saturation with alkali feldspar and plagioclaseduring the early stages ofdifferentiation. This ensures that

the HrO content of the melt will not be very high at theonset ofdecompression. It also ensures that the degree ofcrystallinity will be low enough to permit ascent of themagma after saturation with alkali feldspar. Composi-tions classified as granites, granodiorites, and alkali gran-ites (and their volcanic equivalents) readily fall into thiscategory. Either diorites do not crystallize alkali feldspar,or they do so too late in their diferentiation history topermit the formation of rapakivi. Syenites and trachytesform from higher temperature magmas and may gothrough a region of instability of plagioclase (Nekvasil,1990), making the formation of rapakivi more difrcult.

2. The HrO content of the melt at the onset of decom-pression must be low enough that the magma will notbecome HrO saturated during ascent. If the final pressureis 2 kbar, this requires melt HrO contents at the onset ofdecompression of 5 wto/o or less. (If emplacement occursat even lower pressure, this HrO content must be corre-spondingly less.) Although low HrO content at the onsetof decompression could be most readily assured if thebulk HrO content of the system were low, an HrO-richmagma might also produce rapakivi if the bulk compo-sition permits saturation with plagioclase and alkali feld-spar at a very early stage and decompression takes placesoon thereafter.

3. The percentag€ of alkali feldspar that crystallizes atthe higher pressure must be high enough so that ascentwill not result in complete resorption of alkali feldspar.This constraint, along with the requirements for low HrOcontent and crystallinity, places severe restrictions ontemperature at the onset of ascent.

4. The change in temperature during decompressionmust be close to the adiabatic gradient. Large degrees ofcooling (= 10 "C/kbar) will result in precipitation, ratherthan partial resorption, of alkali feldspar during ascent.Very small degrees of cooling (=3'Clkbar) will result inpartial resorption of plagioclase along with quartz andalkali feldspar and a greater possibility of complete re-sorption of alkali feldspar during decompression. Thisrestriction could explain why rapakivi granites are socommonly found with nonrapakivi-bearing granites andother nonfelsic rocks. If the formation of rapakivi re-quires gradients close to adiabatic, then the country rockthrough which the magma ascends must have been pre-warmed to prevent considerable heat loss during ascent.

This investigation has outlined the changes in equilib-rium assemblages with decreasing pressure that couldprovide the driving force for the production ofall ofthetextural characteristics ofrapakivi. The results ofthis in-vestigation indicate that a simple process such as decom-pression of a closed system and the reestablishment ofequilibrium can produce parageneses that can lead to theformation of rapakivi. However, this does not imply thatmore complicated disequilibrium, open system processessuch as magma mixing (e.g., Hibbard, l98l) cannot playa role. Therefore, the interpretation of the origin of ra-pakivi in a specific body must still be made on an indi-vidual basis after consideration of not only phase equi-

r289

libria but also information about open system processes,provided for example, by textural and isotopic analysis.

AcxNowr-nncMENTs

I thank J.R. Holloway, S.E. Swanson, J.A. Whitney, and D.H. Lindsley

for critical reviews that significantly improved the manuscript and C.R.Bacon for editorial handling. This work was supported in part by NSFgrant EAR-8916050 which is greatly appreciated. SUNY at Stony Brookprovided generous computational support.

RprnnnNcns crrED

Abbott, R.N. (1978) Peritectic reactions in the system An-Ab-Or-Qz-

H,O. Canadian Mineralogist, 16, 245-256.Anderson, J.L. (1980) Mineral equilibria and crystallization conditions in

the late Precambrian Wolf River Massif, Wisconsin. American Journalof Science. 28O. 289-332.

Anderson, J.L., and Cullers, R.L. (1978) Geochemistry and evolution of

the Wolf River Batholith, a late PreCambrian rapakivi massif in north-

ern Wisconsin, U.S.A. PreCambrian Research, 7,287-324.Bickford, M.E., Sides, J.R., and Cullers, R L. (1981) Chemical evolution

of magmas in the Proterozoic terrane of the St. Francois mountalns,

southeastern Missouri: l. Field, petrogSaphic and major element data.

Journal of Geophysical Research, 86, 1 03 6 5- I 0 3 86.Bladh, K.L. (1980) Rapakivi texture from the O'Leary Porphyry, Arizona.

Bulletin Yolcanologique, 43, 1 55-l 7 LBuddington, A.F., and konard, B.V. (1962) Regional geology of the St.

Lawrence County magletite district Northwest Adirondacks, NY U.S.

Geological Survey Professional Paper, 37 6, 1 -259.

Burnham, C.w. (1979) The importance of voltatile constituents. In H.S.

Yoder, Ed., The evolution of the igneous rocks, p. 439-482. PrincetonUniversity Press, Princeton, New Jersey.

Burnham, C.W., and Nekvasil, H. (1986) Equilibrium properties of gran-

ite pegmatite magmas. American Mineralogist, 7 l, 239 -263.

Cherry, M.E., and Trembath, L.T. (1978) The pressure quench formationof rapakivi texture. Contributions to Mineralogy and Petrology, 68,l -6.

Ehrreich, A.L., and Winchell, R.E. (1969) Rapakivi texture in rhyolite.Nature, 224, 905-906.

Fuhrman, M.L., and Lindsley, D.H. (1988) Ternary-feldspar modelingand thermometry. American Mineralogist, 7 3, 2Ol-21 5.

Greenwood, R. (1951) Younger intrusive rocks ofPlateau Province, Ni-geria, compared with the alkaline rocks of New England. GeologicalSociety ofAmerica Bulletin, 62, 115l-1178.

Hibbard, M.J. (1981) The magma mixing origin of mantled feldspars.

Contributions to Mineralogy and Petrology, 76, 158-170.Hutchinson, R.M. (1956) Structure and petrology ofthe Enchanted Rock

batholith, Llano and Gillespie Counties, Texas. Geological Society ofAmerica Bulletin, 67, 763-806.

Johannes, W. (1978) The melting of plagioclase in the system Ab-An-HrOand Qz-Ab-An-H'O at P',o : 5 kbar, an equilibrium problem. Con-

tributions to Mineralogy and Petrology, 66,295-303.Kranck, E.H. (1969) Anorthosites and rapakivi magmas from the lower

crust. New York State Museum and Science Service Memoir, 18, 93-97 .

Larabee, D.M., Spencer, C.W., and Swift, D.J.P. (1965) Bedrock geology

of the Grand Lake area Aroostook, Hancock, Penobscot and Wash

counties, Maine. U.S. Geological Survey Bulletin, 1201-E, El-E38'Lindsley, D.H., and Nekvasil, H. (1989) A temary model for all reasons.

Eos. 70. 506.Nekvasil, H. (1986) A theoretical thermodynamic investigation of the

system Ab-An-Or-QzCHzO) with implications for melt speciation. Ph.D.

thesis, The Pennsylvania State University, University Park, Pennsyl-vanla.

- 1t988a) The calculation ofcrystallization paths ofcompositionallysimple felsic melts. American Mineralogist, 73,956-965.

- (1988b) Calculated effect of anorthite component on the crystal-lization paths of HrO-undersaturated haplogranitic melts. AmericanMineralogist, 7 3, 9 66-982.

NEKVASIL: FORMATION OF RAPAKIVI

t290 NEKVASIL: FORMATION OF RAPAKIVI

- (1990) reaction relations in the granite system: Implications fortrachytic and syenitic magmas. American Mineralogist, 75,560-571.

Nekvasil, H., and Burnham, C.W. (1987) The calculated individual effectsof pressure and HrO content on phase equilibria in the ganite system.The Geochemical Society Special Publication, 1, 433-446.

Rumble, D. (1976) The adiabatic gradient and adiabatic compressibility.Carnegie Institution of Washington Year Book, 7 5, 65 l-65 5.

Sederholm, J.J. (1 89 l) Uber die finnlandischen Rapakiwigesteine. Tscher-maks Mineralogische und Petrologische Mitteilungen, 12, l-30.

Stewan, D.B. (1959) Rapakivi granite from eastern Penobscot Bay, Maine.International Geological Congress Session 20, Sec. I 1, 293-320.

Sykes, ML., and Holloway, J.R. (1987) Evolution of granitic magmaduring ascent: A phase equilibrium model. The Geochemical SocietySpecial Publication, l, 447 -461.

Tuttle, O.F., and Bowen, N.L. (1958) The origin of granite in light ofexperimental studies in the system NaAlSi3O8-KAlSi3O8-SiOr-HrO.Geological Society of America Memoir, 74, 153 p.

Volborth, A. (1962) Rapakivi-type granites in the Precambrian Complexof Gold Butte, Clark County, Nevada. Geological Society of AmericaBulletin, 73, 813-832.

Wahl, W. (1925) Die Gesteine des Wiborger Rapakiwigebietes. Fennia,45, 105-126.

Whitney, J.A. (1975) The efect of pressure, temperature, and X",o onphase assemblage in four synthetic rock compositions. Journal ofGe-ology,83, 1-31.

Maxuscnrrr R.EcErvED Fennunnv 13, 1990MeNuscnrrr AccEPTED Aprul 17. 1991