American Mineralogist, Volume 71, pages 460-47 I, 1986

Trace-elementmodeling of the petrogenesisof granophyresand aplites in the Notch Peak granitic stockoUtah

Pnrnn I. Nlrelr. Department of Geology, University of Missouri-Columbia, Columbia, Missouri 6521I

AssrRAcr The petrogenesesof a granophyre and an aplite in the Notch Peak stock, lJtah, were modeled using published experimentaldata for mineral/melt, mineraVfluid, and melVfluid partitioning of alkali, alkaline-earth and rare-earth elements.The stock rangesin compo- sition from granite to quartz monzonite. The granophyreoccurs along a horizontal contact of the granite with overlying impermeable limestones and contains graphic intergrowths ofalkali feldsparand quartz, displayingunidirectional solidification structures.Calculations suggestthat the granophyre formed by 2-l3o/o crystallization from an aqueousfluid that exsolved from the magma during emplacementand was trapped beneath the limestones. The observed strong enrichment of Ba and Sr in the granophyre relative to the ganite is due to largealkali feldspar/fluid partition coefficientsfor theseelements, while the depletion of large-ionJithophile elementsis due to their small solubility in aqueousfluids. On the other hand, the aplite, one of many crosscuttingthe stock, is depleted in Ba, Sr, and the REEs, while the concentration of highly chargedelements, Zr,Ta, Hf, Th, and IJ, is about the same as in the major lithologic units in the stock. It is not possible to explain the alkaline-earth element and REE concentrationsin the aplite by any model that employs mineral/melt partition coefrcients for anhydrousmelts. However, the "minimum granite" composition and the large concentration of the highly charged elements are difficult to explain by direct crystallization from aqueous fluids. It is apparent that the partition coefficientsof Ba, Sr, and the REEs betweenminerals and melt significantly increasewhen the system becomessaturated with chlorine-rich aqueousfluids. The results of this study have implications for the petrogenesisof pegmatitesand other aplites becausemany have depletion and enrichment characteristicssimilar to the aplite and granophyre at Notch Peak. It is shown that rocks that are the result offractionation offanhydrous melts can be distinguished on the basis of their major- and trace-element chemistries from those that crystallized in the presenceof aqueous fluids or those that crystallized directly from the fluids.

INrnonucrroN to modeling crystaVmelt processesmainly in mafic and The common associationof pegmatites,aplites, and intermediaterocks(Arth,1976; Hanson, 1980).Recently granophyreswith granitic rocks has been recognizedfor published experimental data on partitioning of several a long time, but it was the work of Richard Jahns and trace elementsbetween minerals and fluid and melt and coworkers that demonstratedthe importance of aqueous fluid (e.g., Carron and Lagache, 1980; Volfinger, 1976) fluids in the petrogenesisof these rock types. Jahns and also permit the modeling of crystallization processesthat Burnham (1957, 1969) suggestedon the basis of experi- involve fluids. mental work that pegmatitesand aplites crystallize when The purpose of this paper is to model the petrogenesis a separatesupercritical aqueous-fluid phase forms during of aplitesand granophyresassociated with the Notch Peak the late stagesof crystallization of a granite. The time of granitic stock, House Range,western Utah, and to dem- separationdepends on the solubility of fluids (e.g.,water) onstratethe feasibility of usingtrace elements in modeling in the magma at a given confining pressureon the system. the petrogenesisof pegmatitic rocks. Pegmatites in the A review of pegmatite petrogenesis(eerni, 1982) shows Notch Peak stock occur only as occasionalpods; thus, it that pegmatitesand associatedrocks have been described is difficult to obtain a representativewhole-rock sample. primarily in terms of mineralogy and major-element It is, however, easierto model the petrogenesisof aplites chemistry;however, the abundancesoftrace elementshave and granophyresbecause they can be sampled as whole not been generally used to their full potential to further rocks. Furthernore, the clearly defined mutual relation- elucidate the petrogenesisof theserocks. ships among the aplites, granophyres,and the host intru- The use of trace elementsto model the petrogenesisof sion can be usedto constrainpossible petrogenetic models. igneousrocks has been extensive but generally confined In addition, the petrogenesisofthe stock and the nature 0003{04x/86/03044460$02.00 460 NABELEK:TRACE-ELEMENT MODELING 461

A,

granoPhYre

NW Fig. 2. Cross-sectionthrough the granite body. Note the 1o- cation ofthe granophyrebelow limestone beds.

El camb,isns€drmontary n- of approximately ,l-6 km. "**" l. I limestones and argillites at a depth UTAH The stock is a composite body consisting of a granite and two ditrerent intrusions ofquartz monzonite. The crosscuttingrela- Fig. L Geologic map of the Notch Peak stock with locations tions indicate rhat the granite intruded first, followed by quartz granophyre of the aplite and samplesdiscussed in this article. monzonite I and then by quartz monzonite II (Fig. l). AIl three lithologic units are texturally and mineralogically similar. The granite tends to be porphyritic (Fig. 3a), while quartz of fluid flow in the Notch Peak complex are well under- monzonite I has a seriatetexture and is mineralogicallyand tex- stood from previous studies (Nabelek et al., 1983, 1984, turally the most homogeneousof the three intrusions. Quartz II is porphyritic near the contact with quartz mon- ms.). monzonite zonite I but becomesseriate toward the core.Euhedral oligoclase, AN.cl,YTrcAl, METHoDS equant qrurtz, and minor biotite crystallized early in the para- genetic sequencein all three lithologic units. Although perthite The analysesof Si, Ti, Fe, Mn, Ca, K, Nb, Cu, Ga, Y, and Pb grains are the largest, they apparently crystallized late in the were conducted at Battelle-Northwest, Richland, Washington, parageneticsequence. Titanite, apatite, and zircon are accesso- using an energy-dispersiveX-ray fluorescenceKevex Model 8 l0 ries. The core ofquartz monzonite II containsextensively chlori- systemwith zirconium and silver secondarysources. Al, Mg, and tized biotite and sericitizedplagioclase. Nabelek et al. ( I 983) have Na were analyzed on samplesdissolved in lithium metaborate shown that this alteration is primarily due to interaction with flux by atomic absorption.U.S. GeologicalSurvey standards were fluids that exsolved during crystallization from deeper parts of usedas working standards.Replicate standardand sample anal- the magma chamber. The presenceof a free fluid phase is also ysesyield uncertainties of less than 100/ofor Mn, Nb, Cu, and suggestedby miarolitic cavities, containing quartz and micro- Ga; 590for Si and Ti; 30lofor K, Ca, Al, Na, Mg, Pb, and Y; and cline, that are especiallycommon in the granite. In addition, the lolofor Fe. granite is cut by quartz veins, one about 3 m across.The petro- Other elementswere analyzed by instrumental neutron acti- geneticmodeling of major- and trace-elementdata from the stock vation analyses(INAA) of 0.5-1.0-g irradiated aliquots of about (Nabelek et al., ms.) suggeststhat the three intrusions are coge- 5-kg homogenizedwhole-rock powders. The gamma-raysemit- netic and representvarious stagesof crystallizationand emplace- ted from the sampleswere counted by Ge(Li) detectorsusing the ment of a chemically fractionated diapir. sequentialcounting technique describedby Laul (1979). Zn, Sb, The granophyresare located at horizontal contactswith sedi- Rb, Sr, and Zr data wereobtained by coincidence-noncoincidence mentary rocks. Particularly good examplesare the samples166- counting, which resulted in reduction of the Compton back- 1A and 166-18. These sample were collected from an approxi- ground. Corrections were made for Ia, Ce, Nd, Sm, and Zr formed mately l-m-wide granophyre that occurs along the horizontal by fission of235U and as activation products of23EIJas a result contact betweenthe granite and the limestone layers of the wall of irradiation. U.S. GeologicalSurvey standardsGSP-I, BHVO- rock which form the roof of the magma chamber (Fig. 2). The I , and International Atomic EnergyAgency standardSoil-5 were location of the granophyrenear the top of the granite intrusion used to monitor the accuracyand precision of the results. The is similar to that in other plutons, including gabbros (e.g.,Car- one-standard-deviationuncertainties on replicate standardanal- michael et al., 1974).The granophyreconsists solely ofgraphic ysesare l-2o/o for Na, Ce, Co, Sc, and Zn; 3o/ofor Ba, Eu, Tb, intergrowths of alkali feldspar and quartz (Fig. 3b). The contact and Rb; 4-5o/ofor Sr, La, Sm, Yb, Lu, Hf, Ta, Th, U, and Cs; between the granophyre and sedimentary layers is sharp. The and less than l0Yo for Nd and Zr. K, Na, Rb, Ba, Sr, and Zn contact between the granophyre and the granite is about l0 cm wereanalyzed by more than one method. The replicatesare well thick and is characterizedby rhythmic, about l-cm-thick grano- within the stated analytical uncertainties. For consistency, only phyric and aplitic layers (Fig. 3c). The fine-grained aplitic layers the INAA data are reported. (not to be confused with the aplite dikes discussedbelow) appear Gnor,ocy to have acted as nucleation sites for fanning-out alkali feldspars with intergrown quartz. The $owth direction of the feldspars is The geologyofthe Notch Peak stock is describedin detail in away from the wall-rock contact. Nabelek et al. (ms.) and therefore only a short summary is pre- Aplites occur as dikes cutting all three lithologic units and often sented here. The Jurassic stock intruded interbedded Cambrian cross contacts between the units. The dikes generally strike north 462 NABELEK: TRACE-ELEMENT MODELING

Fig. 3. (A) Typical texture ofthe gxanite.(B) Texture ofthe granophyre. (C) Rhythmic zone between the granophyre and the granite. The direction from the wall-rock contact is shown by the arrow. (D) Texture of the aplite. The scalebars denote I mm. with vertical dips and vary in width from about l0 cm to 2 m. centrations ten times smaller (Fig. 5). The granophyre Thesefeatures suggest that they wereemplaced in fracturesduring pattern has, however, a strong positive Eu anomaly, the last stagesof solidification of the stock. The aplites have an whereasthe granite pattern has a negative Eu anomaly. equigranularsugary texture with grains no more than I mm in Other high-valencelarge-ionJithophile trace elements and maximum The larger have dimension. cores of several dikes a transition metals (Zr,Hf, Nb, Co, Ta, Sc, U, Th) are also coarsepegmatitictexture. The mineralogyincludes albite, quartz, depletedin the granophyrerelative to the granite. Zr val- and graphic intergrowhs of quartz and alkali feldspar (Fig. 3d). uesin parts per million (ppm) are 2D-24inthegranophyre CoprposrrroNs vs. I 15-220 in the granite (see Table 2). On the other Granophl'res hand, the alkaline-earth trace elements are significantly enriched in the granophyre relative to the granite. The The granophyre data for the two analyzedsamples of the granite contains 180-400 ppm Ba, the granophyre about (166-lA 166-lB) The and are shown in Tables I and 2. 3000 ppm, and Sr is enriched approximately three times normative values plotted in Figure 4, Q-Ab-Or are along in the granophyrerelative to the granite. Rb has similar with fields for granite, qvartz morlzonite I, quartz and concentrationsin both (Table 2). monzonite II. The granophyredata lie of the granite field and away from the 200-MPa P"ro minimum on the Q-Or cotectic. The granophyre contains very little FerO, and Aplites CaO relative to the granite and almost no MgO; MnO One aplite sample was analyzed for major and trace was undetected. elements (Tables I and 2). Although the aplites cut all The trace-elementabundances in the granophyre are three main lithologic units, it is best to compare the aplite also significantly different from those in the granite or the chemistry to that of quartz monzonite I as this unit com- other two lithologic units. The rare-earth-element(REE) posesthe bulk of the stock. Furthefinore, the quartz mon- patternsare subparallelto those ofthe granite but at con- zonites were the last units emplaced(Nabelek et al., ms.); NABELEK: TRACE-ELEMENT MODELING 463

Table l. Major-element concentrationsand normative values Table 2. Trace-elementconcentrations in parts per million (ppm)

granophyres apl i te average average r66-',r A 166-1 B 56-2 granite QMI granophyres apl i te 9ranrte QTlI 166--rA 166-18 56-2 rangea range_ (wr. Z) 11.5 78.0 7q q 13,9 Y 1.3 1,3 N.A, 6.4-',13,9 8.1-14.6 Si02 11.3 (300) 12,6 13.9 Sr 450 410 26 100-190(r50) 260-410 A1203 10.9 11.8 12,6 (260) (200) 0.19 0,21 Rb 210 r80 290 195-345 160-230 Ti02 0,16 0.13 0.09 '1.19 (270) 260-7i0(s00) 0.92 2,09 Ba 3320 2760 50 180-400 Fe2b3 0,38 0.36 20_0q ?.8-9,4 0.53 Cs 3.5 I .3 l'4s0 0.03 0,01 Ob 0.21 '1.68 22 22-40 21-49 1.10 Th54 Ca0 a.?1 0,28 0.37 u 1.0 1.8 22 3.8-14,0 3.4-8.6 Na20 2,07 2.48 3.48 3.56 3.93 Zr 24 20 220 11 5-?20 91-170 Iz9 1,78 8.r5 4.90 5,21 3.s',r Hf 0.82 0.63 3.1 3.4-4.8 2.6-4,2 Y2U5 N.A.A N.A.A 0.02 0.07 0.1l Nb 3.4 2.6 N.A. 20-31 16-32 Ta 0.56 0.43 1.8 1.1-2.7 1. 3-8.6 Tota I 98.8 r00. 7 100.4 99.3 99.9 Ga 14 13 N.A. l5-17 14-',18 Sc 0.23 0.I 7 2.4 2.1-3.1 2.8-4.0 Co 0.31 0.26 0.67 1.4-2.9 2.5-4,2 (CIPl,,lnorm) sb 0.08 0.09 0.97 0.07-0.74 0.07-0.28 20-35 37.4 34.4 38.4 32.6 33.4 Pb 28 34 N,A. 22-36 a N,A. 6-45 6-r0 0r 46.0 48.1 29.0 31.1 20.8 Cu79 Zn 4 0.4 t0 10-22 I t-Jo Ab 12.9 15.2 29.4 30.1 33.3 2a 't4.2 1.1 7.6 La 3.38 3.35 18.5-39.1 27.2-48.5 1,1 1.0 Ce 5.93 5.26 24 35-84(62) 48-r30(84) NS 0.78 t.t Nd 3,i 2.8 8,4 1r-30(18) 18-34(26) Di 0,16 0.05 1.7 Sm 0.36 0.36 1.58 1.13-5.19(2.12) 2.45-5.51(3.98) 0.12 0.37 0.26 Eu 0,40 0.38 0,r3 0.32-0.85(0,45) 0.67-0.98(0.76) ny 0.36 11 Tb 0.03 0.02 0.04 0.16-0.40b 0.25-0.45b 0,92 2.0 Yb 0.15 0.'r4 0.64 0.83-1.75(1.14) 0.89-1.52(1.24) FC (0.24) Mt - 0.67 Lu 0.04 0.04 0.?4 0.r6-0.33(0.24) 0.11-0,29 II 0.01 0.01 0.02 0.37 0.09 a Parenthesesindicate average values used in calculations as co's. Ru 0.08 0.23 b Extrapolated values of Gd were used instead of Tb in calculations. The Tn 0. 0, 38 3't values were 1.67 for granite and 2,49 for quartz monzoniteI. Ap 0,05 0.17 0.26 N.A. signifies "not anaiyzed". c 0.94 0.81

Total 98.8 100.5 100.8 100.0 99.8 involve fluids. Unfortunately, these authors did not give a Phosphorus was not analyzed ic the granophyre. o Below detection limrt. the analytical data but presented only exchange distri- bution coemcientsthat they defined as + (l) therefore,the aplites most likely formed in the last stages K"G/O: K"G/0tNa+ K)"2(Na K)J, of crystallization of the quartz monzonites.For theserea- where Ko(s/fl is C"/C'; C, is the concentration of a trace sons, the aplite that is modeled was collected in quartz element in the fluid, and C. is the concentration in the monzonite I (Fig. l). The aplite contains very little Fe2o3 mineral in weight proportions. Carron and Lagachede- and CaO and containsno measurableMgO and only trace fined an analogousequation for the partitioning ofa trace amounts ofMnO and PrOr. Its normative valuesplot near element betweenmelt and fluid: the piercing point for Ab/An : 5.2 on the Q-Ab-Or dia- $am (Fig.4). The trace-elementsystematics of the aplite are quite differentfrom thosein the granophyre.The concentrations of Zr,U, Th, Sc, Ta, and Hf are similar to the ones in quartz monzonite I but are significantly higher than in the granophyre. Ba and Sr have very low concentrations in the aplite at 50 and 26 ppm, respectively.The 290 ppm of Rb are within the range of the concentrations in the quartz monzonite I and are only slightly larger than in the granophyre.The light REEs are depleted approximately five times relative to the quartz monzonite I, and the REE pattern has a slightly negativeEu anomaly (Fig. 5). There is a pronounced relative enrichment of the heavy REEs. PlnrrrroN coEFFrcrENTS Severalmodels for the petrogenesisof the granophyre and the aplite that include mineraVmelt, fluid/melt, and fluid/mineral equilibria are explored in this paper. Min- eral/melt trace-elementpartition coeftcients for the phas- Ab or esinvolved (Table 3) are fairly well known and have been Fig. 4. Diagram of the normative Q-Ab-Or values. Shown petrogenetic used with successin many studies.The data are frelds for the major units in the stock and values for the aplite partitioning ofCarron and Lagache(1980) for the ofRb, sample56-2 and the $anophyre samples166-14 and 166-18. Sr, and Ba betweenhydrothermal solutions and feldspars The boundary for Ab/An : 5.2 is after von Platen (1965); Ab/ and hydrothermal solutions and melts at 700-800'C and An = oo is after Tuttle and Bowen (1958); both are at 200-MPa 200 MPa permit numerical modeling of processesthat pressure. 464 NABELEK: TRACE-ELEMENT MODELING

o = tt c o -c () o o. E cl (t,

Fig. 5. (A) REE patterns for the granophyre samples.The field for granite samplesis shown for comparison. (B) REE pattern for the aplite sample 56-2. The field for the host quartz monzonite I is shown for comparison.

K"(m/f): K"(m/|tNa + K),/(Na + K)-1. (2) valuesremain largeand the Rb K.(s/| for K-feldspar does not differ much from one. For the purposesof this study, the Nernst partition coef- The partitioning of the alkali and alkaline-earth trace ficients,&(s/0 and Ko(m/f), are more useful. By inverting Equations I ard 2, elementsbetween feldsparsand fluid and melt and fluid dependson the chlorinity of the solution to satisfy stoi- &G/0 : K"(s/f)/[(Na + K),/(Na + K)"] (3) chiometry (e.g., Holland, 1972). The exact chlorinity of the solutions at Notch Peak is not known, but fluid-in- K"(m/|: K"(m/f)/tNa + K),/(Na + K)-1. (4) clusion data from the contact-metamorphosedargillite It is therefore necessaryto determine the Na + K in the country rocks suggestthat salinities ofthe infiltrating fluids, feldspars,melt, and fluids at Notch Peak. which were derived from the stock (Nabeleket al., 1984), The partition coefrcientsbetween melt and fluid (Table were as high as 20 wto/ototal dissolved salts or an ap- 4) were calculatedusing the granite value of 70 000 ppm proximately 3M solution (Feldman and Papike, l98l). of Na + K and Ko(m/f) values for QooAbroOrromelt com- This suggeststhat the calculatedKu values may be some- positions as determined by Carron and Lagache(1980). what too large. 59 000 ppm of Na + K were assumedto be in the fluid The amount of anorthite component in the system is since Carron and Iagache (1980) utilized,2M chloide small (Table l). Therefore, the partition coefrcients for solutions in their experimentsand a 2M chloide solution all elements,except perhaps Sr, probably can be approx- in equilibrium with a melt with the composition of the imated by those for albite/fluid. Quartz is assumed to Notch Peakgranitic rocks has molar W(K + Na) of about removeinsignificant quantities of alkali and alkaline-earth 0.4. elementsfrom the melt or fluid. The relative efect of the The most usefrrlmineraVfluidKu(s/f) values are forAb,-, other phasesin the rocks is also considered to be insig- as an analogueto oligoclase,and AbrrOr^ compositions. nificant, except for Rb where the biotite/fluid partition There are approximately 88 000 ppm of Na in albite and coefficientis about four times larger than for orthoclase/ 128000 ppm of Na + K in AbrrOrr,. 59000 ppm of fluid and for Ba wherethe biotite/fluid partition coefficient Na + K in the fluid werealso used to calculatethe mineraV is about the sameas for orthoclase/fluid(Volfinger,1976). fluid K. values, since it was assumedthat the fluid was also in equilibrium with a near-minimum melt. possible PnrnocnNrsrs oF ApLrrEs errors resulting from this assumption are not significant The occurrenceofaplites throughout the stock as dikes, becauseit can be shown that if Na/K in the fluid varies all striking essentiallynorth, suggeststhat the bulk ofthem anywhere between 0 and infinity, the Ba and Sr K"(s/fl crystallizedfrom either a residual melt or an exolved fluid NABELEK:TRACE-ELEMENT MODELING 465

Table 3. MineraVmelt partition coefficientsused in Table 4. MineraVfluid and melVfluid partition coefrcients calculations* used in calculations*

melt/fluid PLAG KFLD BIOT CPX TITN ZIRC APAT 0r75Ab25/f1uidAb169/fluid biot/fluid 0 Rb 0.04 4.1 3,3 0.03 0 0 6.8 I .4 0.1 0,52 0 0 0 Rb 1,1 0.06 5r 4.4 3.9 35 to Ba 0.3 6.'1 6.4 0.13 0 0 0 Sr 370 Ba 270 210 4.5 Ce 0.27 0.044 0.32 0.5 53 35 - q? Ce _ Nd 0.21 0.025 0.29 1.1 88 2.2 57 - Eu 8.3 Sm 0.13 0.018 0.26 1.7 102 11 63 - Gd 5.6 Eu 2.15 l.'13 0.24 1.6 101 3.1 31 _ 12.5 Gd 0,097 0.0i1 0,28 r ,9 102 12 56 Yb Yb 0.006 0.44 r.6 37 270 24 0.049 '1.5 * of the pdrtition coefficients Lu 0.046 0.006 0.33 21 323 20 See text for derivation

Fron compilations bv Arth (1976) and Hanson (1980). Phelpsreported sanidine/melt Ba partition coefficientsof about 22, and Michael (1984) reported calculatedortho- phase within fractures in the nearly totally crystallized clase/melt Ba and Sr Ko(s/m) values of 20 and 6, respec- magma. Therefore, the petrogenesisof the representative tively. The values for other elementsare not significantly aplite sample56-2 is best modeled in terms ofequilibrium different from those listed in Table 3. When Michael's betweenthe melt or fluid and the solidified quartz mon- partition coefficientsare used to model the trace-element zonite. The composition of the quartz monzonite I sample concentrationsusing Equation 5, the Ba and Sr concen- usedfor modeling is very closeto the averagequartz mon- trations become86 and 96, respectively,for .F : 0.01. zonite composition. The ganite cannot be treated as a These values are closer but still 1.7 and 3.7 times larger melt formed alonga fractionation path ofthe stock toward than the observedconcentrations. The problem with the crystallization of the aplite becausethe granite was the high model Rb and REE concentrations, however, still first of the major lithologic units to have solidified. remams. Crystal/melt fractionation The model concentrationof Ba, calculatedwith the use ofthe Rayleighfractionation equation and IQ valuesfrom The aplitescompose only a minute portion of the stock, Table 3, matches closely the concentration in the aplite. certainly no more than l0l0.It seemstherefore appropriate The model concentration of Sr is, however, an order of to model the trace-elementconcentrations in the aplite magnitude smaller than the observed value, whereasthe by the batch-equilibrium equation ofShilling (1966), model Rb concentration is unreasonably htgh and the c^/Co:t/lD(l- n+rf, (5) model chondrite-normalized REE plot is not shown in Figure 6 becauseit falls offthe scale. if the processof melt separationwas sudden, as it may The resultsfor Rb and the REEs,except for the direction have been during fracturing ofthe stock. C- in Equation of the Eu anomaly, are somewhat better if it is assumed 5 is the concentration of a trace element in the melt, Co that the analyzedaplite sample is the product of 5olofrac- is the concentration of the trace element in the system,D tional crystallization from the residual melt. This result is the weightedsum of mineral/melt partition coemcients was obtained by calculating the bulk trace-elementcon- according to the fraction of each phase in the solid, and centration in the crystallizing solid using the Rayleigh Fis the fraction of melt. In the lesslikely caseof complete fractionation law (Hoefs, 1980), disequilibrium betweenthe interior of the solid phasesin the quartz monzonite and the residual melt, the trace- e": co(l - FD)/Q- n. (7) element concentration in the melt can be modeled using The normative proportion of minerals used in the cal- the fractionation equation of Rayleigh ( I 886); culation is listed in Table 5. The Sr and Ba values are, C^/Co: I'ro-rt. (6) however, much larger than the observed concentrations (Table 6). The resultsdo not improve with larger amounts The representativequartz monzonite I trace-elementcon- of crystallization becausethe Sr and Ba concentrations centrations (Table 2) were used for Co, while normative cannot decreasebelow the values in the starting melt. mineral proportions (Table 5) in conjunction with the The model REE results have to be interpreted with mineraVmelt partition coefrcients (Table 3) were used to some caution. A large proportion of the REE in granitic calculateD. A comparison of the model trace-elementconcentra- tions in the aplite, which were calculatedusing Equation Table5. Mineralproportions used in calculations* 5 (Table 6; Fig. 6) with the observedconcentrations (Table 2;Fig. 5), showsthat the model concentrationsof Rb, Sr, granophyre Ql'4I Apl i te Ba, and the REEs, for both F values of 0.01 and 0.3, are a 0.40 0.33 0.38 0r 0,60 0.28. 0.40 significantly higher than those in the aplite. Ab 0 0.33 0.20 Apat 0.005 0.0025 0.0005 keman and Phelps (1981) and Mahood and Hildreth Titn 0 0.0025 0 Zirc 0 0.00026 0.00044 (1983) have shown that in very high silica systems(more Biot 0 0.03 0 * than about 74 wto/oSiOr), trace-elementpartition coeffi- A11 pyrox. componentsfrom Table 1 were assigned to biotite and cients may becomevery large.For example,Leeman and al6ite componentwas proportioned to makeK-feldspar = 0175. 466 NABELEK:TRACE-ELEMENT MODELING

Table 6. Calculatedtrace-element concentrations in aplite and granophyre modeling O aplite apl ite nodels granophyre 'tB 'lc mode'ls rA tD 2 34384 tr 5% crystlliz. ',160 trom resid. melt Rb 116 420 6319 330 78 78 186 Sr 11 7 143 0.21 280 3 341 540 52 8a 313 351 31 572 6 755 3Z4O 60 Q residualmelt 1A: in resid. melt formedby 99 per cent equil. crystalliz, of qtz. monz,. 1B: in resid. per 'lC: f,elt formed by 70 cent equil. crystalliz. of qtz. monz.. in resid, melt formed by 99 per cent fract. crystalliz. of qtz. nonz.. lD; in rock formed by 5 per cent fractional crystalliz. from melt formed ds in 14, 2: in fluid which is jn equr'libriun with solidified quartz monzonjte, 3A: in rock formed by nearly 0 per cent crystallrz. from qranite melt - assumingBa and Sr alk. feld./melt K6's are 6.12 and 3.g7, resp.. o 38: in rock formed as in 34 but with Ba ind Sr K;'s of 20 and 6, resp., .= 4! in fluid " in equillbrium with qranite melt. E c o -c o rocks is often confined to minor phases, o such as allanite CL or monazite. However, the error introduced into the cal- E ct culations by this possibility is probably small for the Notch o Peak rocks, becauseit can be easily shown from mineral- separatesdata (Nabelek et al., ms.) that essentiallyall the REEs are confined to titanite and apatite, which were incorporated into the present calculations. Michael (1984) modeled the petrogenesisof aplite dikes around the margin of the Cordillera Paine granite, Chile. Those aplites have very similar chemistry and chemical relation to the host granite as the aplites at Notch Peak. Michael concluded,primarily on the basis of low Ba and Sr concentrations in the aplite, that the aplites are the crystallizaton product from a residual melt derived by Fig.6. ModelREE patterns for residualmelts formed by 70 fractional crystallization of the host granite. Although he and 990/ocrystallization of quartz monzoniteI and a calculated ascribed low REE concentrations in the aplite to their pattern for 5o/ocrystzlllzation from the 1% residualmelt. The removal granite, by allanite in the he did not model the observedaplite pattern is shownfor comparison. REEs rigorously and the Rb concentrationsnot at all. As shown above, it is especially the concentration of these incompatible elementsthat suggests,at leastfor the Notch fluid. Note the very small calculatedconcentrations of Ba, Peak situation, that the model of crystallization of the Sr, and REEs in the fluid. aplites from a polymerized, anhydrous residual melt is The secondstage of aplite petrogenesisthat was mod- inappropriate. eled was crystallization from the fluid. Owing to the lim- ited solubility of aluminosilicatesin fluids, the aplite clear- Crystallization from fluid ly could not have formed by l00o/ocrystallization of the At the other extreme is the possibility that the aplite fluid. The concentrationsof the trace elementsin the bulk crystallized from an aqueousfluid, becauseit is thought aplite were calculated assumingRayleigh fractionation @q. that the fracturing of a solidifying pluton is often accom- 7). panied by flow of a fluid into the fractures (e.g.,Norton The calculated fractionation curves for Ba and Sr (Fig. and Taylor, 1979).Such a fluid probably can be assumed 8) demonstratethat the concentrationsof theseelements to be in equilibrium with the nearly solidified host rock. rapidly decreasein the crystallizing solid to levels found The concentrationsof trace elementsin a fluid phase in in the aplite. In fact, their concentrations in the aplite equilibrium with the quartz monzonite I (Table 6;Fig.7) might suggest,ifthis model is correct, that the aplite formed werecalculated using Equation 5 and the sameparameters by l2-l4o/o crystallization from the fluid. The concentra- as in the calculation of concentrationsin a residual melt tion of Rb cannot be used to estimate the amount of (previoussection), except that mineral/fluid partition coef- crystallization, becausethe Rb fractionation curve is es- ficients were used (Table 4). There are no direct deter- sentially flat. Nevertheless,the calculated concentration minations of mineral/fluid REE K.G/| values for alkali of about 290 ppm is in agreementwith the observedvalue. feldspars.Therefore, these were approximated by dividing The calculatedREE concentrationsand the shapeofthe mineraVmelt partition coefrcients (Table 3) by fluid.zmelt pattern for 130/ofractional crystallization are also in good partition coefficients(Flynn and Burnham, 1978; Table agreementvrith those observed(Fig. 7). The relative en- 4). The melt/fluid partition coefficientsused are from a richment of the heavy REEs is due to crystallization of run conducted at 125 MPa and 800.C with 0.914M Cl zircon. Although zircon occursin only a small proportion NABELEK:TRACE-ELEMENT MODELING 467

GRANOPHYRE APLITE aPlite n I \. Q model aplite

\ tr vapor in equil. tt with stock o o- .= o- !t c10 o E o o5 ct E 6 10 o l.o o.g 0.8 0.7 0 6 r.o 0.9 0.8 0.7 0.6 FRACTIONOF VAPOR FRACTIONOF VAPOR Fig. 8. Plotsofcalculated concentrations ofBa, Sr, and Rb in modelgranophyre and aplite during crystallization from fluids. The blackcircles denote the concentrationsofBa and Sr in the samples.Thc observedvalues may bracketthe amount of crys- tallization.The Rb concentrations,also denoted by black circles, cannotbe used to estimatethe percentage ofcrystallization; there- Fig. 7. Model REE patternsfor a fluid in equilibrium with fore,their positionin termsofF hasno significance.See text for solidified quartz monzoniteand for an aplite formed by 130/o furtherdiscussion. fractionalcrystallization from the fluid. The observedpattern is shownfor comparison. Furthermore, the relatively large quantities of Zr, Hf, Sc' Sb, Ta, U, and Th do not appear to be consistent with in the aplite (the amount was calculatedfrom Zr concen- crystallization from a fluid becausethese elements are tration assumingstoichiometric zircon), its preferencefor thought to have small solubilities in aqueousfluids. the heavy REEs has an especially significant effect on the petrogenesis shapeof the pattern in a systemwhere all other minerals Discussionof aplite have REE Ku(s/| values much less than l. Thus, zircon It is apparent that neither the residual anhydrous melt essentiallybehaves as a cumulate.The largestdiscrepency model nor the model of crystallization from hydrothermal is in the calculatedand observedEu values.However, the fluids is satisfactory to explain the petrogenesisof the partitioning ofEu betweenphases is very sensitive to/o, aplites. The correct model must explain the "minimum and is thus more difrcult to model accurately.A similar granite" composition of the aplite (Fig. a); the relatively enrichment of the heavy REEs was observedin quartzo- large concentration of the highly chargedZx, Hf, Th, U, feldspathicveins in the Cortlandt Complex of New York Sc, Sb, and Ta; and the small concentration of the REEs, that include garnet as the only phasewith a strong pref- Ba, and Sr. To explain thesecharacteristics, it is suggested erencefor the heavy REEs (Bender, 1980). that the aplites crystallized from a melt saturatedwith a These calculations point to a model of crystallization Cl-rich aqueous fluid that intruded fractures in the last of the Notch Peak aplites from an aqueous-fluid phase stagesof crystallization of the stock. that exsolved during the waning stagesof crystallization The discussionby Lagacheand Carron (1982) oftheir of the stock. Owing to very large Ba and Sr crystaVfluid earlier experimental results (Carron and Lagache, 1980) partition coefficients and moderately large coefficients for implies that the crystaVmelt partition coefficients(Table the REEs, the presumed fluid was depleted in these ele- 3) used here thus far to model the petrogenesisof the ments, which resulted in their small observedconcentra- aplite may be appropriate to model crystallization of a tions in the aplite. fluid-undersaturatedmelt but not the formation of a re- There are difrculties with this fluid model, too. The sidual melt saturatedwith a Cl-rich aqueousfluid. Lagache required I 30losolubility of aluminosilicatesin the hydro- and Carron obtained their mineraVmelt partition coefr- thermal fluid is two to three times larger than that pre- cients indirectly by assuming that Tuttle (1969).In addition, the position dicted by Luth and C"/C^: (C"/C) x (C'/C^) (8) of the normative Q-Ab-Or values of the aplite near the ternary 200-MPa minimum for Ab/An : 5.2 (Fig. 4) is becausethey did not obtain direct measurementsof min- difrcult to explain by crystallization from a fluid in equi- erallmelt partition coefrcients.Ifthe assumptionin Equa- librium with the quartz monzonite I whose Ab/An is ap- tion 8 is correct, then the Ba, Sr, and Rb quartz mon- proximately 5. Luth and Tuttle (1969) suggestedthat the zonitelaplite melt bulk distribution coefficients (D) are composition of the fluid phasein equilibrium with a solid approximately17 ,39, and 0.46,respectively. The Ba and granitejust below the solidus lies closeto the quartz apex. Sr D values are little more than an order of magnitude 468 NABELEK:TRACE-ELEMENT MODELING larger than the D values derived using partition coem- largepegmatites also have generallysmall concentrations cients from Table 3, while the Rb D is about twice as of alkaline-earthelements and lower concentrationsof the large.If theseD valuesand an,F value of 0.01 are used REEs than the associatedparental rocks and tend to be in Equation 5, the model melt has 30, 9, and 430 ppm of enriched in highly chargedelements (e.g., Walker, 1984). Ba, Sr, and Rb, respectively.These values are reasonably However, the petrogenesisof suchpegmatites may be dif- closeto the measuredconcentrations. Iflagache and Car- ficult to model numerically to a large precision because, ron's partition coeftcients are correct, then these results as shown here, the partition coefficientsof trace elements suggestthat the aplite did indeed crystallize from an betweenminerals and anhydrous granitic melts may not aqueous-fluid-saturatedmelt. be applicable.Furthermore, direct measurementsof par- The reasonfor the apparentdifference between mineral/ tition coefficientsbetween minerals and melts saturated anhydrous melt and mineral/Cl-rich-fl uid-saturatedmelt with Cl-rich fluids are still lacking. Mineral/fluid and melV partition coefficientsis not clear.Chlorine doesnot greatly fluid coefficientsmay neverthelessbe used to bracket the affectthe melt structurebecause it partitions preferentially most plausible models. into the aqueous-fluid phase (e.9., Kilinc and Burnham, 1972). However, a large concentration of chlorine in the Prrnocnutsrs oF GRANopHyRE fluid phase may affect the partitioning of alkaline-earth The very low concentrationsoflarge-ion-lithophile ele- elementsin the solid-melt-fluid system. ments in the granophyre preclude that the granophyre The partitioning between minerals and melt may be representsa residual melt formed on crystallization of the also affected by changesin the melt structure when as- stock, because crystal/melt fractionation processesin sociatedwith a free fluid phase.Stolper (1982) has shown granitic rocks generally lead to enrichment of such ele- that at high water contents,a substantialfraction ofwater ments (see,for example,the previous section on aplites). dissolvedin a silicatemelt may be in the form of molecular However, the granophyre may be an early-crystallized HrO. Perhapsunder such conditions, alkaline-earth ele- rock that formed by fractionation from the granite melt, ments are substantially rejected from the melt structure or it may be a rock that crystallizedfrom an aqueousfluid while the effect on Rb and Cs is smaller, as evidencedby that exsolved sometime during solidification of the stock their similar concentrationsin the quartz monzonite and and subsequentlypooled under a limestone bed (Fig. 2) the aplite. that wasimpermeable to fluids (Nabeleket al., 1984).Both The structure of the silicate melt may also be affected possibilities are numerically modeled here. by fluorine, which preferentially dissolvesin melts (e.g., Bailey, 1977). However, there is no strong evidence for Crystallization from melt substantial amounts of fluorine in the Notch Peak com- The position of the normative composition of the plex. Thus, its efect remains speculative. granophyre on the 200-MPa Q-Or cotectic boundary on It appears,therefore, that the retention characteristics the granite ternary diagram (Fig. a) and the positive Eu of Ba, Sr, and Rb in a melt saturatedwith a Cl-rich fluid anomaly in the REE pattern (Fig. 5) could be consistent are similar to those in an aqueousfluid. This is substan- with crystal accumulation from the granite magma. Crys- tiated by melVfluid partition coefficientsthat are no more tal accumulation is best modeled by Rayleigh fractiona- than 5 (Table 4). The REE melt/fluid partition coefrcients tion where the averageconcentration of a trace element are not much larger; therefore, the REE bulk partition in the solid, C., is given by Equation 7. The maximum coefficients(D) betweenthe quartz monzonite and a melt and minimum possibleconcentrations of compatible and saturated with a Cl-rich fluid may also be greater than incompatible trace elements, respectively, in the grano- unity. The melt should have, therefore, a REE pattern phyre (Fig. 9, Table 6) were obtained by multiplying D similar to the one calculated for a fluid in equilibrium by Co using the averagetrace-element concentrations in with the solidified quartz monzonite (Fig. 7). The relative the granite (Table 2) for Co,the partition coefficientslisted enrichment of the heavy REEs, which is observedin the in Table 3, and the normative mineral proportions in the aplite pattern, neverthelessneeds to be explained by ac- granophyre(Table 5) for D. The proportion of apatite was cumulation of zircon. obtained by point counting becausephosphorus analyses The partitioning behavior of highly charged elements were not obtained for the granophyre samples. suchas Zr,Th, Hf, Ta, and U betweenminerals and melt The calculatedBa and Sr concentrationsare not as large may not be affectedmuch by saturation of the systemwith as in the granophyre,whereas the calculatedRb concen- Cl-rich fluids becausethese elements may behave as net- tration is about three times less than the observedvalue work-formers. In a melt with a high HrO content, their (compare with Table 2). The calculated REE pattern is, solubility may in fact increaseto a limited extent, as it however, very close to the observed pattern. When Mi- does, for example, for Zr (Watson and Harrison, 1983). chael's (1984) Ba and Sr K"(s/f) values of 20 and 6, re- The model for the petrogenesisof the Notch Peakaplite spectively, are used, the calculated concentrations arc 3240 may be applicableto the petrogenesisofmany pegmatites. ppm for Ba and 540 ppm for Sr. These values are close Aplites throughout the world have elemental enrichment enoughto the measuredconcentrations to suggestthat the and depletion characteristics,relative to the host granites, granophyremay have formed by fractional crystallization similar to thoseof the Notch Peakaplites. Moreover many of the granite magma. However, other features of the NABELEK:TRACE-ELEMENT MODELING 469 granophyre,which are discussedbelow, precludethis pos- sibility. fgranophyre 166-1A Q model granophyre Crystallization from fluid D vapor in equil. The granophyremay have crystallizedfrom a fluid that exsolvedfrom the granite magma on emplacement.This involves a two-stageprocess: (1) formation of the fluid (2) granophyre. possible o and crystallization of the It is to = calculatethe concentrationof Ba, Sr, Rb, Ce, Eu, Gd, and L !,c Yb in the fluid using Equation 5, with the substitution of o Crfor C^, and melt/fluid partition coefficientsfrom Table E(, 4. The calculatedconcentrations, assuming an f' value of o 0.01, are listed in Table 6, and the REE pattern is shown CL E in Figure 9. o The concentrationsof trace elementsin the granophyre o on crystallization from the fluid are best modeled using Equation 7. The Rb, Sr, and Ba K.(s/| values used are given in Table 4. The D(s/f) values used were calculated by multiplying the Kd(s/0 values by the mineral propor- tions shown in Table 5. The mineral/fluid REE K"(s/f) valueswere again approximated by dividing mineral/melt partition coefficients(Table 3) by fluid/melt partition coef- ficients(Flynn and Burnham,19781,Table 4). As discussed Fig. 9. Model REE patternsfor a fluid in equilibrium with in the sectionon the petrogenesisof the aplites, this meth- the granitemelt and for granophyrecrystallized from the fluid. od may lead to some error. The calculated REE pattern Themodel granophyre pattern is for 590fractional crystallization. is shown in Figure 9, and the fractionation curves for Rb, Theobserved pattern is shownfor comparison.The REE pattern Sr, and Ba in Figure 8. for fractionalcrystallization from the granitemelt (not shown) The fractionation curves indicate that the large con- coincideswith the pattern for the model granophyrecrystalli- centrations of Ba and Sr are easily obtained on crystalli- zationfrom the fluid. zation from a fluid phase, even if the concentration of theseelements in the fluid phaseis relatively small (Table 6). This is due to the extremely large alkali feldspar/fluid It is not necessarythat the fluid from which the gran- partition coefficients.There is, however, some discrep- ophyre crystallized exsolved while the granite was still ancy in the amount of crystallization indicated by the Ba mostly molten. Satisfactoryresults are also obtained if the and Sr data: the concentrationofBa indicates 20loand the fluid was in equilibrium with the already-solidifiedstock. concentrationof Sr 130/ocrystallization. These values may This is the same situation as in the fluid model for the bracket the solubility of aluminosilicates in the fluid, al- petrogenesisofthe aplite, which was previously discussed. though errors in the partition coefficientsused may lead If it is assumedthat the value of -Fapproaches1 and that to somewhatincorrect resultsand explain the discrepancy. the concentrationsof Rb, Sr, and Ba in the fluid are 330, The calculated Rb concentration is very similar to the 3, and 6 ppm, respectively,concentrations of 337, 666, observedvalues, suggestingthat this model may be cor- and 972 ppm, respectively, are obtained for the grano- rect. The observed Rb concentrations again cannot be phyre. The Rb and Sr concentrationsmatch the observed used to constrain the amount of crystallization, because concentrationsrelatively well. The calculatedBa concen- they remain essentiallyconstant with fraction of fluid re- tration would match the measuredvalues well if about 20 maining. ppm of Ba were dissolvedin the fluid. Given the probable The 2-13 wto/ocrystallization necessaryto reproduce error in the partition coefficientsused, such variation in the concentrationsof Ba and Sr encompassesthe 5 wto/o the Ba concentration in the fluid is reasonable.It, there- solubility of aluminosilicatesin pure water vapor at mag- fore, does not matter when in the crystallization history matic temperatures at 200 MPa obtained by Luth and of the stock the fluid from which the granophyrecrystal- Tuttle (1969).The solubility of aluminosilicatesin Cl-rich lized exsolved and pooled below the roof of the magma fluids is not known, but it is probably larger. chamber. The calculated REE pattern (Fig. 9) is for 5olocrystal- lization, and it fits well the observedpattern in both the Discussionof granophyre petrogenesis shape(as much as the shapecan be determined from the The preceding calculations suggestthat the granophyre four calculated REEs) and the concentrations.The simi- formed by probably less than 5o/ocrystallization from a Iarity in the REE pattern to the one obtained by modeling fluid phasethat exsolved from the granite magma while crystallization from melt is in part due to the way the REE the magma was still largely molten or during later crys- mineral/fluid partition coefficientswere approximated. tallization of the stock. However, if large Sr and Ba min- 470 NABELEK: TRACE-ELEMENT MODELING

centrationsof their essentialstructural constituentsin the fluid may be too small for saturation. Apatite may be an o cl exception,because the solubility of phosphorusin a fluid 6 E' may relatively large. Phosphorusforms similar struc- o be

I tural units in melts to silicon (Hess,1980) and, by analogy, o probably in fluids where silica solubility at magmatic con- o ditions is about 5 g per I kg of pure HrO (Holland and E for this is the fact that o Malinin, 1979). Some evidence z found in pegmatites,including pegmatite + apatite is often :< fracturefrllings that apparentlycrystallize from fluids (e.g., )< Walker, 1984). ot* The relatively large concentrations of the chalcophile r,l*o*o t?itporlo't elements,especially of Pb, in the granophyreis also con- Fig. 10. Plotofequilibrium molar compositions of coexisting sistentwith the fluid model. Urabe (1985) has shown that fluid-meltand fluid-alkali feldspar. The stippled part ofthe feld- at relatively high chlorinities, chalcophile elements, es- spar-fluidcurve denotespossible compositions of feldsparin pecially Pb, partition preferentially into the fluid phase equilibrium with fluid that exsolvedfrom the granitemagma. over the melt. The Pb in the granophyre may be stored in the K-feldspars and some in the apatites. Thus, the chemistry and texture of the granophyre at Notch Peak by crystallization from a fluid phase erallmelt partition coemcientsare used,the model results are best explained exsolved from the melt some time during crystalli- are also consistentwith a very small amount of fractional that zation of the stock and that subsequentlypooled below crystallization from the granite melt. The textures,major- roof of the magma chamber. element, and other trace-elementdata suggest,however, the that the granophyre crystallized from a fluid phase. The position of the normative compositions of the CoNcr,usroNs granophyreon the Q-Or cotecticto the right of the ternary Granophyresand aplites composeonly a small portion minimum (Fig. a) is inconsistentwith a small amount of of the Notch Peak stock. Trace-elementmodeling of their fractional crystallization becausethe granite contained petrogenesesusing available mineral/melt, mineral/fluid, quartz and feldsparphenocrysts when it was emplaced.If and melVfluid partition coefficientsprovides significant fractional crystallization occurred,it is expectedthat pla- constraints on the nature of the media from which they gioclasewould be a cumulate phase, not K-feldspar. In crystallized. The extreme enrichment of Ba and Sr, the addition, the texture ofthe granophyre(Fig. 3b) is clearly low concentrationsof large-ion-lithophile elements, and not a cumulate texture. the high K/Na ratio of the granophyreare best explained The high K./Na composition of the granophyreis better by crystallizationfrom a fluid that exsolvedfrom the crys- explained by crystallization from a fluid phase.An alkali tallizing stock and pooled under the roof of the magma feldspar that crystallizesfrom an aqueousfluid will tend chamber. to be more potassicthan the fluid (e.g.,Orville, 1963; The aplite dikes at Notch Peakare apparentlythe result Carron and Lagache,1980; Fig. l0). The K,rNa of a fluid of crystallization from a melt saturatedwith Cl-rich fluids in equilibrium with a melt is approximately 0.75 of the that intruded fractures in the waning stagesof crystalli- molar ratio in the melt (Holland, 19721'Carron and La- zation of the stock. The large concentration of highly gache,1980). The granitemagma at Notch Peakhas molar chargedlarge-ionJithophile elementsin the aplite and the K/(K + Na) between 0.4 and 0.5; therefore, a fluid in aplite's "minimum granite" composition strongly suggest equilibrium with the granite magma should have molar that it crystallized from a melt. This melt either had re- K/(K + Na) between about 0.3 and 0.4. A feldspar that tentive characteristicsfor the alkaline-earthelements and crystallizesfrom such a fluid should have K/(K + Na) REEs approachingthose ofa halide-rich fluid, or the pres- between 0.4 and 0.9 (Fig. l0). The K/(K + Na) of the ence of the fluid affected the mineral/melt partitioning granophyreis about 0.7 and is within the predicted range. suchthat the concentrationsof thesetrace elements in the The fact that the granophyre lies on the Q-Or cotectic aplite can be modeled only in terms ofcrystallization from suggeststhat the fluid was also saturatedwith quartz. a fluid phase.The aplite texture may be merely a reflection The small concentrationsof Mg, Ca, and the large-ion- of rapid nucleation and crystallization on "pressure" lithophile trace elementsin the granophyreare also better quenchingduring fracturing ofthe stock. explainedby crystallization from a fluid than from a melt. This study showsthat the petrogenesisofgranophyres, The Mg and Ca fluid/melt partition coefficientsare much aplites, and pegmatitescan be constrained by modeling less than one, even at high Cl concentrations (Holland, their trace-elementcontents. It is shown that pegrnatitic 1972), indrcating that the solubility of these elementsin rocks that crystallizefrom fluids can be distinguishedfrom a fluid is very small. Thus, a rock that crystallizesfrom a thosethat crystallizefrom residual pegmatitic melts when fluid is not expectedto contain phasessuch as plagioclase, their trace-elementcharacteristics are compared to their biotite, and probably zircon and titanite becausethe con- parental magmas. Owing to the large Ba and Sr alkali NABELEK:TRACE-ELEMENT MODELING 471 feldspar/fluid paftition coefrcients and the apparently large experimental melting and crystallization of Harding, New feldspar/pegmatitic melt partition Mexico, pegmatite.Geological Society ofAmerica Bulletin, 68, alkali coefficients, frac- r75r-17 52. pegmatitic tional crystallization from melts and fluids can - (1969) Experimental studies of pegmatite genesis:l. A lead to large gradients of Ba and Sr in large pegmatites. model for the derivation and crystallization of granite peg- It is also shown that irrespective of which medium the matites. Economic Geology, 64, 843-864. pegmatites crystallize from, their REE concentrations Kilinc, I.A., and Burnham, C.W. (1972) Partitioning of chloride betweena silicate melt and coexisting aqueousphase from 2 probably will be lower than those in the parental magma. to 8 kbars.Economic Geology, 67,231-235. Direct measurements of trace-element partition coeffi- Lagache,M., and Carron, J.-P. (1982) Zonation des dl€mentsen cients between minerals and fluids saturated with Cl-rich trace au cours de la croissancedes cristaux dans les bains fluids and the determination of the concentration of all silicatEs:L'example de Rb, Cs, Sr et Ba dans le systdmeQz- Acta, 46, 2I5l- the elements in such fluids are needed for more precise Ab-Or-HrO. Geochimica et Cosmochimica 2158. modeling. Laul, J.C. (1979) Neutron activation analysis of geologicalma- terials. Atomic Energy Review, I7, 603-695. AcrNowr,nocMENTS Leeman,W.P., and Phelps,D.W. (1981)Partitioning of rareearths This research was funded by a fellowship at Battelle-Pacific and other trace elementsbetween sanidine and coexistingvol- glass. Research,86, 10193-10199. Northwest Laboratories with the Northwest College and Uni- canic Journal ofGeophysical Luth, W.C., and Tuttle, O.F. (1969) The hydrous vapor phase versity Associationfor Science(University ofWashington) under in equilibrium with granite and granite magmas. Geological Contract EY-76-S-06-2225with the Department of Energy.Ad- Societyof America Memoir 115, 513-548. ditional support was provided by the University of Missouri Mahood, Gail, and Hildreth, W. (1983) Large partition coefr- GeologyDevelopment Fund. Many thanks go to Jim Papike and cientsfor elementsin high-silicarhyolites. Geochimica et Cos- Gil Hanson for discussionson the subject,Neal White for assis- mochimica Aclz, 47, l l-30. tance in the field, and to Rich Walker and Carol Nabelek for Michael, P.J. (1984) Chemical differentiation of the Cordillera reviews of an earlier version of the manuscript. Two anonymous Paine granite (southern Chile) by in situ crystallization. Con- reviewersand Gordon E. Brown, Jr., provided many useful sug- tributions to Mineralogy and Petrology, 87, 179-195. (1983) phase gestions. Nabelek,P.I., O'Neil, J.R., and Papike,J.J. Vapor exsolutionas a controlling factor in hydrogenisotope variation RnrnnnNcns in granitic rocks: The Notch Peak granitic stock, Utah. Earth futh, J.G. (1976) Behavior of trace elementsduring magmatic and PlanetaryScience Letters, 66, 137-150. (1984) processes-A summary of theoretical models and their appli- Nabelek,P.I., tabotka, T.C.,O'Neil, J.R.,and Papike,J.J. cations. U.S. Geological Survey Journal of Research,4, 4l- Contrasting fluid/rock interaction between the Notch Peak 47. granitic intrusion and argillitesand limestonesin westernUtah: Bailey,J.C. ( I 977) Fluorine in granitic rocks and melts: A review. Evidence from stable isotopes and phase assemblages.Con- Chemical Geology, 19, l-42. tributions to Mineralogy and Petrology, 86,25-34. Bender,J.F. (1980)Petrogenesis of the Cortlandt Complex. Ph.D. Norton, D., and Taylor, H.P., Jr. (1979) Quantitative simulation thesis,State University of New York at Stony Brook. of the hydrothermal systemsof crystallizing magmas on the basisoftransport theory and oxygenisotope data: An analysis Carmichael, I.S.E., Turner, F.J., and Verhoogen, J. 097q h- neouspetrology. McGraw-Hill, New York. ofthe Skaergaardintrusion. Journal ofPetrology, 20, 421-486. Carron, J-P., and lagache, M. (1980) Etude exp6rimentale du Orville, P.M. (1963) Alkali ion exchangebetween vapor and fractionnementdes 6l6mentsRb, Cs, Sr, et Ba entre feldspaths feldspar phases.American Journal of Science,26I, 201-237. alkalins, solutions hydrothermaleset liquides silicat6sdans le Rayleigh,J.W.S. (1886) Theoretical considerations respecting the separationofgases by difrrsion and similar processes.Philo- systemeQ.Ab.Or.HrO i 2 kbar entre 700 et 800'C. Bulletin sophical Magazine,42, 77 -lO7 . - de Min€ralogie,103, 571-578. Cernf, Petr, Ed. (1982) Granitic pegmatitesin scienceand in- Shilling, J.-G. (1966) Rare earth fractionation in Hawaiian vol- dustry. Mineralogical Association of Canada, Short Course canic rocks. Ph.D. thesis,Massachusetts Institute of Technol- Handbook 8. ogy, Cambridge,Massachusetts. Feldman, M.D., and Papike,J.J. (1981) Metamorphic fluid com- Stolper, Edward (1982) The speciationof water in silicate melts. positions from the Notch Peak aureole, Utah. (abs.) EOS Geochimica et Cosmochimica Acta, 46, 2609-2620. granite (American GeophysicalUnion Transactions),62, 435. Tuttle, O.F., and Bowen,N.L. (1958) The origin of in Flynn, R.T., and Burnham, C.W. (1978) An experimental de- light of experimental studies in the system NaAlSirOr-KAl- termination of rare earth partition coefrcientsbetween a chlo- Si3Os-SiOr-HrO.Geological Society of America Memoir 74. ride containing vapor phaseand silicate melts. Geochimica et Urabe, Tetsuro (1985) Aluminous granite as a sourcemagma of CosmochimicaActa, 42, 685-701. hydrothermal ore deposits:An experimental study. Economic Hanson,G.H. (1980) Rare earth elementsin petrogeneticstudies Geology,80,148-157. of igneous systems.Annual Reviews of Earth and Planetary Volfinger, M. (1976) Etret de la temp6raturesur les distributions Sciences,8,371-406. de Na, Rb et Cs entre la sanidine,la muscovite, la phlogopite pression Hess, P.C. (1980) Polymerization model for silicate melts. In et une solution hydrothermale sous une de I kbar. R.B. Hargraves, Ed. Physics of magrnatic processes,3-48. Geochimica et Cosmochimica Acta, 40,267-282. granitischer Princeton University Press,Princeton, New Jersey. von Platen, H. (1965) Kristallisation Schmelzen. Hoefs, Jochen(1980) Stableisotope geochemistry. Springer-Ver- Contributions to Mineralogy and Petrology, 11, 334-381. lag, New York. Walker, R.J. (1984) The origin of the Tin Mountain pegmatite, Holland, H.D. (1972) Granites, solutions, and base metal de- Black Hills, South Dakota. Ph.D. thesis. State University of posits. Economic Geology, 67, 281-301. New York at Stony Brook. (1 Holland, H.D., and Malinin, S.D. (1979) The solubility and oc- Watson,8.B., and Harrison, T.M. 983)Zircon saturation:Tem- currenceof non-ore minerals. In H.E. Barnes,Ed. Geochem- peratureand composition efects in a variety of crustal magma istry ofhydrothermal ore deposits,2nd edition, 46 1-508. John types. Earth and Planetary ScienceL€tters, 64,295-304. Wiley, New York. MlNuscnrpr REcETvEDFpnnuenv 28, 1985 Jahns,R.H., and Burnham,C.W. (1957)Preliminaryresultsfrom MeNuscnrpr AccEp'rEDOcrosrn I, 1985