ternary feldspar crystallization in high-temperature felsic magmas › ammin › am77 ›...

American Mineralogist, Volume 77, pages 592404, 1992

Ternary feldspar crystallization in high-temperature felsic magmas

fLc.NNa Nnrv.lsrr,Department of Earth and Space Sciences, State University of New York, Stony Brook, New York 11794-2100, U.S.A.

Ansrn-q.cr

Ternary feldspars crystallizing at equilibrium under conditions of high temperature andlow HrO content in trachytic and dry granitic magmas show distinctly different evolution-ary paths once two feldspars are stabilized if crystallization occurs under conditions offixed HrO activity (HrO buffered) or under conditions in which the HrO content increasesin the melt as a result of precipitation of anhydrous minerals (HrO unbuffered). The lattercondition leads to feldspar paths characterized predominantly by decreasing Or contentin plagioclase (with only slight changes in An content) and significant increases in Orcontent in alkali feldspar with decreasing temperature. Unlike the case for HrO-bufferedconditions, HrO-unbuffered conditions can result in a path that includes a stage of partialresorption of plagioclase by reaction with melt, producing alkali feldspar, followed by astage ofcoprecipitation ofboth feldspars as the HrO content builds up and the odd (peritec-ticlike) region of the 2 feldspar * L or 2 feldspar + Qz + L surface shifts farther towardalbite. The effect of increasing bulk HrO or Qz content is to decrease the extent of Orenrichment in alkali feldspar, the amount of Or depletion in plagioclase, and the amountof resorption of plagioclase. The effect of decompression is complex and depends on thestage of evolution of the magma when decompression commences. Decompression canlead to crystallization or resorption of either feldspar, as well as significant changes in thecompositions of the feldspars.

IrvrnooucrroN ternary alkali feldspar (generally anorthoclase sodic san-idine) reported in numerous trachytes and syenites, e.g.,

Compositionally, quartz-normative trachytes and sy- Sierra el Virulento (Moll, 198 l), Kane Springs Wash cal-enites differ only slightly from rhyolites and granites in dera, Nevada (Novak and Mahood, 1986), Turritablehaving higher normative feldspar and lower quartz con- Falls, southeast Queensland (Ewart, 198 1).tents' Yet the presence of ternary feldspar and anhydrous Details of the positional relations between the solvusferromagnesian minerals indicate high temperatures and and the 2 feldspar + L curve (and the 2 feldspar + Qzpotentially significantly different crystallization path- + L curve) at low HrO contents were calculated by Nek-ways. The differences in crystallization pathways of the vasil (1990) and Nekvasil and Lindsley (1990). These cal-feldspars are controlled by the changes in topology of the culations permitted the determination of the composi-equilibrium phase relations with decreasing H2O content tional regions in which magmas must lie if partial oror H'O activity in the melt. The most significant topo- complete resorption of plagioclase were to occur duringlogic change in the synthetic granite system, i.e., the sys- equilibrium crystallization. This resorption is a conse-tem NaAlSirOr-CaAlrSirO8-KAlSi3O8-SiOr-HrO, is the quence of the high proportion of alkali feldspar in thetermination of the 2 feldspar * L curve, which occurs trachytic compositions and the more An-rich nature ofinside the feldspar subsystem instead of at the Ab-Or this feldspar relative to the melt, which requires resorp-boundary, and the termination of the 2 feldspar + Qz + tion of some plagioclase to release An for further crystal-L curve, which occurs within the An-bearing granite sys- lization of alkali feldspar. Nekvasil (1990), however, cal-tem instead of in the haplogranite system, i.e., the system culated crystallization paths only for the case in whichAb-Or-Qz at low HrO contents. This topologic change the activity of HrO remained unchanged during equilib-results in peritecticlike (or odd) behavior in which the rium crystallization (HrO-buffered crystallization). In thisreaction between a feldspar and melt produces a second case, the relative positions of the solvus and the 2 feld-feldspar at the expense of the reacting feldspar at some spar + L curve remain fixed during crystallization, andpoint along each ofthese curves. As recognized by Car- the resulting resorptional behavior and feldspar compo-michael (1963) and Nekvasil (1990), the most common sitional variations can be readily shown in a ternary sec-reaction would be the production of alkali feldspar through tion of the feldspar-HrO system.the reaction of plagioclase with melt. Such behavior can From the discussions of Tuttle and Bowen (1958), Car-readily explain the presence of partially resorbed cores of michael (1963), Abbott (1978), and Nekvasil (1990), itternary plagioclase (generally andesine oligoclase) within can be concluded that feldspar crystallization in high-

oo03-oo4x/92l0506-0592$02.00 592

temperature trachytic magmas will involve at least partialresorption of plagioclase if the HrO activity of the meltis buffered. Although there are some natural processesthat could buffer the HrO activity of the magma, such asthe crystallization of a hydrous mineral or perhaps thepresence of a large quantity of a mixed volatile fluid,there is no reason to expect that syenites will crystallizeunder buffered conditions. In granitic rocks there is abun-dant petrographic and textural evidence indicating an in-crease in HrO activity during crystallization. Crystalli-zation under HrO-unbuffered conditions is probably alsothe more relevant process for trachytic magmas.

During normal HrO-unbuffered crystallization of an-hydrous minerals, the HrO content of the magma increas-es, and the relative positions ofthe ternary feldspar sol-vus and the relevant HrO-isoactivity three-phase (Pl +Af + L) curve in the feldspar system or the four-phase(Pl + Af + Qz + L) curve in the granite system changecontinually as the temperature is depressed. Additionally,as shown schematically in Figure l, the three- and four-phase HrO-isoactivity curves continually lengthen, andthe location ofthe neutral point on each ofthese curves(where the curve begins to exhibit peritecticlike behavior)lies progressively at a higher Ab content at higher activityof HrO (see Nekvasil, 1990, for further discussion). Im-portantly, however, although evolution along the 2 feld-spar + L or 2 feldspar * L * Qz surfaces on the Ab sideof the neutral curve is a necessary condition for resorp-tional behavior, it is not a sufficient condition, as shownby Nekvasil (1990). The bulk composition dictateswhether or not resorption will take place during equilib-rium crystallization in the odd region of each surface.Should the bulk composition permit resorptional behav-ior, the melt composition at which resorptional behavioris first manifested is not necessarily at the neutral curveand is also dependent upon the bulk composition. Asdiscussed by Nekvasil (1990), the higher the bulk An con-tent, the farther the melt must evolve along the surfacepast the neutral curve before resorptional behavior ismanifested.

Whether or not resorptional behavior will be mani-fested during HrO-unbuffered equilibrium crystallizationis dependent on the relative rates ofincrease ofHrO con-tent in the melt during crystallization of anhydrous phasesand on the Ab enrichment of the melt during crystalli-zation of two feldspars. The increase in HrO content ofthe melt during crystallization of anhydrous phases con-tinually shrinks the resorptional field by shifting the neu-tral point toward Ab and quite likely away from the meltcomposition. The Ab enrichment of the melt during crys-tallization of two feldspars, in contrast, drives the meltdeeper into the resorptional field with decreasing tem-perature. These two competing factors lead to severalquestions. Would the HrO content of unbuffered trachyticmagmas increase so quickly during equilibrium crystal-lization as to inhibit partial resorption ofplagioclase, ormight partial resorption occur during the early stages ofcrystallization and be followed by reprecipitation of both

593

aHzo=o'1auzo=o'5

auzo=1'o

Fig. l. Schematic diagram of the 2 feldspar * L surface inthe system Ab-Or-An-HrO shown by three HrO-isoactivitycurves. The effect ofdecreasing HrO content is to shorten eachcurve, resulting in termination ofthe surface farther and fartherwithin the feldspar system (in projection) and away from theAb-Or sideline. The neutral point on each HrO-isoactivity 2feldspar + L curve, which bounds the region ofeven (or cotec-ticlike) behavior and odd (or peritecticlike) character of eachcurve, also shifts farther from the termination with decreasingHrO content. The dashed curve, which connects the neutralpoints, divides the even portion of the surface (to the right ofthe curve) and the odd portion of the surface.

feldspars as a cotectic region ofthe 2 feldspar + L (or 2feldspar + Qz + L) surface is reached? What effect mightsuch a path have on the evolution of the feldspar com-positions? Is the evolutionary path of the feldspar com-positions significantly different from the buffered case topermit any preserved feldspar zoning in trachytes andsyenites to be used to indicate the behavior of HrO in thenatural magma? What is the effect of the continually in-creasing Qz concentration of the melt? Is the evolutionof a trachytic magma distinguishable from that of a gra-nitic magma during the late stages?

In order to answer these questions, the controls on feld-spar composition must be understood. The composition-al path taken by any feldspar during crystallization is acombination of two limiting path vectors. One vector in-volves a decrease in An content and an increase in Orcontent of plagioclase. For alkali feldspar it involves adecrease in Or and increase (or slight decrease) in An inalkali feldspar. These paths are very similar to those fol-lowed by feldspars crystallizing along an almost isother-mal solvus. This compositional vector is the result of theincreasing Ab content of the melt and hence of the bulkfeldspar compositions with continued crystallization. Theother vector involves a decrease in Or content ofplagio-clase and an increase in Or content ofthe alkali feldsparwith decreasing temperature as the isothermal solvus sec-tions widen because of increasing nonideality of the solidsolutions with decreasing temperature. This can be seenin Figure 2 by tie lines connecting equilibrium feldsparcompositions on the ternary feldspar solvus of Lindsleyand Nekvasil (1989) for a constant bulk feldspar com-

NEKVASIL: TERNARY FELDSPAR CRYSTALLTZATION

594 NEKVASIL: TERNARY FELDSPAR CRYSTALLZATION

Weight%

Fig.2. The ternary feldspar solvus, calculated using the mod-el of Lindsley and Nekvasil (1989), as a series of isothermalsolvus sections (temperature in degrees Celsius). Three tie linesare shown for the same bulk feldspar composition (solid circle)at 900, 800, and 700'C. They indicate that the equilibriumplagioclase compositions become more Or poor and the coex-isting alkali feldspars more Or rich at lower temperatures. Thearrows indicate the possible compositional evolutionary pathsof feldspars during crystallization. If the effect of temperature onwidening the solvus is dominant over the effect of the increasein Ab content ofthe bulk feldspar composition (because oftheincrease in Ab content of the melt during crystallization), pla-gioclase compositions may evolve along a path such as that la-beled a, while alkali feldspar evolves along a path such as thatlabeled d. In contrast, if the changing melt composition is dom-inant, path b may be followed by plagioclase and path c by alkalifeldspar.

position. If the dominant effect on feldspar compositionsis the increase in Ab content of the melt and change inbulk feldspar composition toward Ab, then plagioclasecompositions will follow a path with the characteristicsof arrow b in Figure 2, and alkali feldspar a path with thecharacteristics of arrow c. Because of the shallower T-Xslope around the nose of the solvus, however, it may bepossible that for high temperature magmas, expansion ofthe solvus with decreasing temperature is dominant overthe changing melt composition, and paths such as path afor plagioclase and path d for alkali feldspar might befollowed. In this case, should feldspar crystallize in a step-wise fractional crystallization process, plagioclase will be-come normally zoned with respect to Or content insteadof An content, while alkali feldspar develops reverse zon-ing in Or content.

In order to evaluate the relative magnitude of eachcompositional vector and to answer the questions posedabove, crystallization paths were calculated using a mod-ification of the method of Nekvasil (1988a), the ternaryfeldspar solution model of Lindsley and Nekvasil (1989),and the melt solution model of Nekvasil (1986) andBurnham and Nekvasil (1986). A modification to the

b 100

80

Temperanre (oC)

Fig. 3. Calculated HrO-unbuffered equilibrium crystalliza-tion path for a model Sierra el Virulento rhyolite (trachyte +

Qz) (28.5 rto/o Or, 37.7 wto/o Ab,6.27 wto/o An, 27.6 wtVo Qz) at5 kbar and 3.5 wt0/o bulk HrO content. (a) Melt (L), plagioclase(Pl), and alkali feldspar (A0 compositions projected into thefeldspar system. The symbol Qz indicates the onset of crystal-lization ofquartz. Arrowheads point to the direction ofdecreas-ing temperature. (b) Variations in proportions (mol9o) of melt(L), plagioclase (Pl), alkali feldspar (Af), and quartz (Qz) vs. tem-perature during equilibrium crystallization. The phase propor-tions sum to (l - X",o) x 100.

method of Nekvasil (1988a) was required because itera-tion convergence is very sensitive to the chosen startingcompositions in the albitic region of the ternary feldspar

system. A less efficient, but more readily convergent,method for the calculation of crystallization paths wasused that involved the location ofthe three- or four-phaseisoactivity curve for a given HrO content of the melt and

bulk HrO content, and hence a given fraction of liquid.Then the composition was shifted along the curve untilthe solution of the mass balance equations yielded the

s

" 6 06. i

6

6. 40o

650700800 7508509500 -1000

Siena el Virulento rhyoliteP=5 kbar Xw=0.35

NEKVASIL: TERNARY FELDSPAR CRYSTALLIZATION 595

chosen fraction of liquid. As concluded by Nekvasil andLindsley (1990), the use of several other available solidsolution models does affect the location and temperatureof the three- and four-phase curves for a given HrO ac-tivity; however, the crystallization trends are unaffected.

For modeling purposes a trachytic composition fromSierra el Virulento was used. This composition falls with-in the range of bulk compositions of trachytes and sy-enites in which phenocrysts of ternary anorthoclase sodicsanidine have been reported to contain irregular cores ofandesine oligoclase (see Nekvasil ,1990, for more details).The trends reported here are general and should hold formost syenites and trachytes, although the specific feldsparand melt compositions attained at any stage of differen-tiation will differ for each bulk composition.

Ftr,nsp.ln cRysTALLIzATIoN rN HrO-nrcrrGRANITIC MAGMAS

Feldspar crystallization trends of quartz-normative tra-chytic magmas should contain elements of the crystalli-zation trends of both Qz-saturated granitic magmas andcompositions in the pure feldspar system. Much experi-mental effort has been expended in the determination ofthe feldspar crystallization trends of granitic magmas (e.g.,Tuttle and Bowen, 1958; Johannes, 1978). Although thedetermination of precise trends is problematic because ofkinetic difficulties (Johannes, 1978), the available exper-imental data and complementary computational resultsinvolving use of thermodynamic solution models andmass balance constraints (Nekvasil, 1988a, 1988b) sup-port the common opinion that feldspar crystallization inHrO-unbuffered granitic magmas follows general trendssimilar to those in the binary feldspar systems, i.e., Abenrichment with decreasing temperature.

Figure 3 shows the calculated HrO-unbuffered path, inwhich the HrO content of the melt is free to change uponthe crystallization of anhydrous phases, of a bulk com-position that was obtained by artificially increasing thesilica content of a Sierra el Virulento trachyte composi-tion (Moll, l98l) through projection toward Qz until thecomposition lies in the main compositional field of stan-dard granites, according to Tuttle and Bowen (1958). Forthis composition at 5 kbar and 3.5 wto/o bulk HrO content(Xnro:0.35, where X"ro is the mole fraction of HrO inthe system), plagioclase is the liquidus phase but is joinedby quartz before alkali feldspar becomes stable. Until themagma becomes saturated in alkali feldspar, the plagio-clase compositions describe a smooth curve with decreas-ing An and slightly increasing Or contents. Eventually,the solvus is intersected, and two feldspars are stabilized.Upon further crystallization, both feldspars change com-position along a set of curves across the solvus dome that,if joined, describe a polythermal solvus section. Thepolythermal solvus section described by the feldsparcompositions involves changes in both temperature andbulk feldspar composition as crystallization proceeds. Thepath of compositional evolution of the feldspars de-scribed by such a section differs from an isothermal sol-

vus section because the path crosses innumerable iso-thermal sections that are progressively widening as thetemperature drops. Importantly, the section of the solvustraversed by the feldspars in wet, low-temperature gra-nitic magmas is not very sensitive to temperature. Thisis illustrated in Figure 3a, where the path of feldspar com-positional change still retains the general shape ofan iso-thermal section but shows little change in Or content overthe considerable temperature range of crystallization un-der HrO-unbuffered conditions. For this HrO-rich com-position, the plagioclase compositions will evolve pre-

dominantly by a decrease in An content. Alkali feldsparcompositions, in contrast, will exhibit little change incomposition. They do, however, indicate an importantfeature. While the magma is still HrO undersaturated, thewidening of the solvus with decreasing temperature is thedominant effect, driving the alkali feldspar compositiontoward Or (cf. Fig. 2). Once HrO saturation has beenattained, however, the changes in temperature with com-position are no longer as great because HrO can no longercontinue to build up in the melt and alkali feldspar beginsto evolve along a path similar to an isothermal solvussection. Both of these path segments are very small andwould likely lead to a very homogeneous alkali feldspar,as is commonly found in low-temperature granites.

Figure 3b indicates the variations in phase proportionsfor this composition during HrO-unbuffered equilibriumcrystallization. Importantly, under HrO-unbuffered con-ditions, once the magma is saturated with two feldspars,coprecipitation takes place throughout the crystallizationinterval, and there is a continuous increase in the pro-portions ofeach feldspar from the onset ofprecipitationto the solidus temperature.

The equilibrium crystallization path shown in Figure3a characterizes the feldspar evolution of wet graniticmagmas. This trend should give rise to the normally zonedplagioclase and relatively homogeneous alkali feldsparcommonly found in many granitic rocks. These charac-teristics may be quite different, however, for high-tem-perature, relatively dry, granitic magmas. The discussionbelow presents the results of calculations that were de-signed to assess the differences between such low-tem-perature paths and those for high-temperature trachytes,syenites, and granites (e.9., fayalite-bearing granites).

Fnr-osp.ln cRysrALLIzATroN rN TRACHYTIC ANDSYENITIC MAGMAS

HrO-buffered paths in the feldspar system

Crystallization trends for trachytes under HrO-bufferedconditions were discussed in detail in Nekvasil (1990).These trends provide an important basis for comparisonwith the HrO-unbuffered paths. For example, Figure 4ashows the evolution of a Sierra el Virulento compositionof Moll (1981) modified by removing all of the normative

Qz component so that the composition lies within thefeldspar system, at 5 kbar at constant HrO activity, foractivity of HrO, 4"ro, equal to 0.1 (- 1.6 wto/o HrO). Un-

P = 5 k b a r a w = 0 . 1

596 NEKVASIL: TERNARY FELDSPAR CRYSTALLIZATION

Ba

E- 6 0

E . 4 0od

I roo

I 140 r100 r080

Temperature ("C)

Fig. 4. Portion of the calculated HrO-buffered equilibriumcrystallization path for a model Qz-free Sierra el Virulento tra-chyte (39.3 t40/o Or, 52.0 wto/o Ab, 8.67 wtolo An) at 5 kbar andactivity of HrO (a",o) equal to 0. I . (a) Melt (L), plagioclase (Pl),and alkali feldspar (Af) compositions projected into the feldsparsystem. Arrowheads point to the direction of decreasing tem-perature. Once all of the plagioclase is resorbed, ternary alkalifeldspar precipitates alone, and its composition follows a slightlycurved path, not shown, toward the bulk composition. At theonset of alkali feldspar precipitation, there is a slighr curvatureof the liquid path toward Or (which is difrcult to discern in thediagram) that results in coprecipitational behavior during theearly stages of precipitation of two feldspars. (b) Variations inproportions (molo/o) of melt (L), plagioclase (Pl), alkali feldspar(Af), and quartz (Qz) vs. temperature during equilibrium crys-tallization shown until the plagioclase is completely resorbed.Plagioclase resorption takes place when its curve has a negativeslope. All phase proportions sum to (l - X",") x 100.

der HrO-buffered conditions, the crystallization temper-ature interval is small, and the feldspars follow a pathsimilar to an isothermal solvus section. Before the solidustemperature is reached, however, all of the plagioclase isresorbed, and the feldspar path leaves the solvus.

Figure 4b shows the variations in phase proportionsduring HrO-buffered equilibrium crystallization of the Qz-free trachyte at 5 kbar. Plagioclase is the liquidus phaseand increases in abundance steadily with falling temper-ature until alkali feldspar appears. At this point, furtherprecipitation of plagioclase is slowed down until a reac-tion between melt and plagioclase begins, which resultsin a decrease in abundance of plagioclase and a concom-itant increase in the abundance of alkali feldspar withdecreasing temperature. This reaction continues until allof the plagioclase is consumed. Crystallization of a singleternary alkali feldspar begins at this point (not shown inFigs. 4a or 4b) and continues until its composition is thatof the bulk rock composition.

The resorption ofplagioclase in Figure 4b arises fromthe termination of the 2 feldspar * L surface within thefeldspar system at low HrO activities and the peritectic-like (odd) nature of this surface as the termination is ap-proached. See Nekvasil (1990) and Nekvasil and Lindsley(1990) for further discussion. As discussed by Nekvasil(1990), crystallization of trachytes need not lead to com-plete resorption of plagioclase; rather, the extent of re-sorption of plagioclase will depend on the pressure, thebulk HrO content, and the bulk composition. Partial tocomplete resorption of plagioclase can also occur for Qz-normative trachytes and syenites and high-temperaturegranites under HrO-buffered conditions, particularly if Qzsaturation is not attained until the late stages of crystal-lization.

The feldspars of many natural trachytes show this typeof compositional trend (Fig. 4), e.g., southeast Queens-land (Ewart, l98l), Sierra el Virulento (Moll, l98l).However, before conclusions are drawn that these tra-chytes crystallized under HrO-buffered conditions, the dif-ferences between HrO-buffered and HrO-unbuffered pathsmust be assessed.

HrO-unbuflered paths in the feldspar systern

Equilibrium crystallization paths of the projected Si-erra el Virulento composition were calculated for HrO-unbuffered conditions for several bulk HrO contents, inorder to evaluate the differences between the composi-tional evolution of the feldspars crystallizing under HrO-unbuffered conditions vs. HrO-buffered conditions andto compare such paths with those of granite composi-tions. Figure 5 shows the calculated equilibrium crystal-lization path of the Qz-free projected Sierra el Virulentocomposition at 5 kbar and 25 molo/o (2.2 wto/o), 20 molo/o(1.7 wto/o), and 15 molo/o (1.2 wto/o) bulk HrO contents.For these calculations, the HrO content increases contin-ually to saturation, at which stage a separate fluid phaseappears, which, for modeling purposes, consists of 1000/oHrO.


b r00

80

For all of the three bulk H'O contents, plagioclase isthe liquidus phase. Its early compositional evolution, be-fore the magma becomes saturated with a second feld-spar, is similar to the HrO-buffered case and would bedifficult to distinguish from it. Once a second feldspar isstabilized, however, the similarities end; plagioslsss t.\ilno longer changes its composition by following a pathsimilar to an isothermal solvus section, as is generally thecase for HrO-rich magmas or HrO-buffered magmas, thatis, a path dominated by decreasing An content, but in-stead follows a path dominated by decreasing Or content.Plagioclase follows this path until the liquid becomes sat-urated in HrO. From this point on, as the temperaturedecreases, the HrO content is buffered by the presence ofa fluid, and plagioclase changes its composition along anHrO-buffered trend similar to that in Figure 3a. However,because of the low bulk HrO content, HrO saturation isattained only very close to the solidus temperature, andthis path is followed for only a very short distance.

The alkali feldspar compositions also show a very dis-tinctive evolutionary trend. The alkali feldspar compo-sitions describe a loop reflecting the competing effects ofincreasing Ab content of the melt, which drives the com-positions along a path subparallel to an isothermal sec-tion in the direction of decreasing Or content, and theeffect of the continual temperature depression, which re-sults in a widening of the solvus and hence an increasein the Or content of the alkali feldspar, as shown in Figure2. For this HrO-unbuffered Qz-free trachyte composition,the effect ofdecreasing temperature on expanding the sol-vus, induced by the increasing HrO content of the melt,is much stronger than the effect of the change in meltcomposition. Therefore, the evolving alkali feldspar com-positions show a strong enrichment in Or content with

ts

Fig. 5. Calculated HrO-unbuffered equilibrium crystalliza-tion path for the model Sierra el Virulento Qz-free trachyte at 5kbar and bulk HrO contents of 2.2 wf/o (X'ro : 0.25; dottedcurves), 1.7 wto/o (Xrro : 0.20; dashed curves) and I.2 utto/o (XHzo:0.15, solid curves). (a) Melt (L), plagioclase (Pl), and alkalifeldspar (Af) compositions projected into the feldspar system foreach bulk HrO content. Arrowheads point to the direction ofdecreasing temperature. V indicates the onset offluid saturation.For all of these curves, resorption begins when the solvus isintersected. Solid stars (X"r" : 0.20) and open stars (X",o : 0. I 5)indicate the compositions ofall phases when partial resorptionofplagioclase is replaced by coprecipitation ofboth feldspars for20 molo/o bulk HrO content. (b) Variations in proportions (molo/o)of melt (L), plagioclase (Pl), alkali feldspar (Af) vs. temperaturefor 2.2 wo/o bulk HrO (X^ro : 0.25) content. (c) Variations inproportions (molVo) of melt (L), plagioclase (Pl), alkali feldspar(Af) vs. temperature for 1.7 wt0/0 bulk H rO (Xnro: 0.20) content.The solid star indicates the onset ofcoprecipitational behavior.(d) Variations in proportions (mol7d of melt (L), plagioclase (Pl),alkali feldspar (Af) vs. temperature for 1.2 wto/o bulk HrO (Xrro: 0.20) content. The open star indicates the onset ofcoprecipi-tational behavior. All phase proportions for each bulk HrO con-tent sum to (l - X"ro) x 100.

s

5 e 0o

s

5 6 0o

! + oF

d ' *

80

s

5 6 0o

? + o

0


falling temperature. This enrichment continues until HrOsaturation is reached, at which point the alkali feldsparchanges composition once again, as in the buffered case,that is, subparallel to an HrO-saturated isothermal sec-tlon.

For lower bulk HrO contents, the plagioclase compo-sition at the onset of alkali feldspar crystallization is moretemary than for higher bulk HrO contents (and hencelower temperatures), and the first feldspar pair lies closerto the consolute curve, which connects the consolutepoints of isothermal sections of the two-feldspar solvus.Since the bulk composition is the same for all the pathsin Figure 3a, and the HrO content in the melt builds upto saturation for all the paths, the final assemblage andthe solidus temperature will be the same for all the paths.Therefore, the amount of Or enrichment in the alkalifeldspar that will occur during crystallization will increasewith decreasing bulk HrO content. Furthermore, the low-er the bulk HrO content, the smaller the change in Ancontent of the plagioclase during crystallization once al-kali feldspar is stable (Fig. 5a). In fact, for X",s : 0.15,plagioclase compositions evolve almost along an An iso-pleth.

The changes in phase proportions for the paths in Fig-ure 5a are shown in Figures 5b-5d. As can be seen bycomparing Figures 5b and 5c, the extent of decrease inplagioclase abundance during crystallization increases withlower bulk HrO content. Importantly, unlike the bufferedcase (Fig. 4b, and Nekvasil, 1990), resorption of plagio-clase ceases and coprecipitation commences during thelater stages of crystallization. This will not occur for anybuffered path, whether or not plagioclase undergoes par-tial or complete resorption. The temperature of onset ofcoprecipitational behavior is very similar for Xrro: 0. l5and 0.20 (cf. Figs. 5c and 5d). At the lower bulk HrOcontent (Fig. 5d), this temperature is attained at a laterstage of evolution than at the higher bulk HrO content ofFigure 5c. This can be readily seen by comparing thefraction of liquid remaining at the onset of coprecipita-tional behavior for the two bulk HrO contents (Figs. 5cand 5d). For X",o: 0.25 (Fig. 5b), this temperature isreached just as both feldspars become stable, and no re-sorption of plagioclase takes place during crystallization.Until further calculations are conducted with other bulkcompositions, it is unclear whether the relatively invari-ant temperature of the onset of coprecipitation of twofeldspars is fortuitous for this bulk composition or a re-quirement of the system.

At even lower HrO contents, a significant change in thefeldspar crystallization path occurs. As for the other HrOcontents, at l0 molo/o (0.75 \Mtolo) bulk HrO content (Fig.6), plagioclase is the liquidus phase. As temperature dropsand crystallization continues, the composition of the pla-gioclase becomes more ternary and approaches the sol-vus. Now, however, the temperatures are higher and therelevant region of the solvus lies farther from the Ab-Ansideline. Because of this, the solvus is not intersected atan early stage of crystallization, and the plagioclase com-

positional path now traverses on the solidus around thenose of (but above) the solvus and toward the alkali feld-spar side of the consolute curve. Such a path was alsodeduced geometrically by W. L. Brown (personal com-munication, 1990). Once the feldspar composition haspassed around the nose ofthe solvus, crossed the exten-sion ofthe consolute curve (i.e., the curve connecting theconsolute points of the isothermal solvus sections), andentered the alkali feldspar field, the melt compositionsbegin to evolve along the break in slope that extends pastthe terminations of the 2 feldspar * L H,O-isoplethalcurves describing the 2 feldspar * L surface. As crystal-lization continues, the single feldspar evolves along acurved path toward lower An contents around the solvus,as the melt continues to follow this break in slope of theliquidus surface. The melt HrO content continues to in-crease and the 2 feldspar * L HrO-isoactivity curves (andhence, the 2 feldspar + L surface) extend farther andfarther toward Ab until they finally catch up with the meltat point A (Fig. 6a), at which point the solvus is inter-sected by the feldspar path. At this point, an albitic pla-gioclase begins to crystallize with the earlier formed feld-spar, which now has the compositional characteristics ofan alkali feldspar. Continued crystallization results in aslight increase in An content with a more major decreasein Or content of the plagioclase, while the alkali feldsparcomposition undergoes a major increase in Or content.At the very end stages of crystallization, the melt be-comes HrO saturated, and the plagioclase undergoes aslight decrease in An content and the alkali feldspar aslight decrease in Or as the HrO-saturated polythermalsolvus section is followed.

Figure 6b shows the variations in phase proportionsduring crystallization. This differs dramatically from Fig-ures 5b-5d because of the prolonged crystallization of thefirst feldspar, which changes in compositional character-istics from plagioclase to alkali feldspar. Additionally, oncethe second plagioclase feldspar begins to crystallize, theproportions of the first feldspar decrease, that is, alkalifeldspar undergoes partial resorption during the later stagesof crystallization.

Effect of pressure. The effect of pressure on the HrO-unbuffered equilibrium crystallization paths is shown inFigure 7a for l5 molo/o bulk HrO content by juxtapositionofthe calculated 5-kbar (dashed curves) and 2-kbar (solidcurves) crystallization paths. The higher temperatures ofthe 5-kbar path during the early stages of crystallizationresult in feldspar compositions that move closer to theconsolute curve of the solvus than at2kbar. The higherHrO content at saturation at 5 kbar (and hence lowersolidus temperature), however, results in more extensiveOr enrichment of the alkali feldspar, Or depletion of pla'gioclase, and Or depletion of the liquid during the latestages of crystallization. Figure 7b shows the variationsin phase proportions at2 and 5 kbar. Because the 5-kbarfeldspar path reaches compositions closer to the conso-lute curve, and because melt compositions move furtherinto the odd region ofthe 2 feldspar + L surface during

NEKVASIL: TERNARY FELDSPAR CRYSTALLIZATION

Sierra el Virulento trachyle

{Oz-free)= 5 kbars Xw--O.1

599

bt *

80

1200 1100 ,1000

900 800 700Temperature ('C)

Fig. 6. Calculated HrO-unbufered equilibrium crystalliza-tion path for the model Qz-free Sierra el Virulento trachyte at 5kbar and 0.75 wf/o (X,,o: 0.1) bulk HrO content. (a) Melt (L),the first feldspar (Fl) and the second feldspar (F2) compositionsprojected into the feldspar system. Arrowheads point to the di-rection ofdecreasing temperature. Point A refers to the point onthe liquid path where the 2 feldspar + L surface reaches the meltcomposition and two feldspars are stabilized. (b) Variations inproportions (molo/o) of melt (L), first feldspar (Fl), and secondfeldspar (F2) vs. temperature during equilibrium crystallization.All phase proportions sum to (l - X",o) x 100.

the early stages of crystallization, more resorption of pla-gioclase takes place during crystallization at higher pres-

sures.The effect of decompression can also be seen in Figure

7. It is much more complex for trachytic and syeniticmagmas than for wetter granitic magmas (see Nekvasil,l99l), and the behavior exhibited depends greatly uponthe stage of crystallization (that is, the assemblage) just

b 6 0

)U

1200 1100 1000 900 800 700Tenrperanne ('C)

Fig. 7. Calculated HrO-unbuffered equilibrium crystalliza-tion path for the model Sierra el Virulento Qz-free trachyte at 5kbar and 2 kbar and 1.2 wto/o (X^,o:0.15) bulk HtO content.(a) Melt (L), plagioclase (Pl), and alkali feldspar (Af) composi-tions at 5 kbar (dashed curves) and 2 kbar (solid curves) pro-jected into the feldspar system. Arrows indicate three de-compressional stages: Path A, where only plagioclase and meltare present and decompression commences at 1125 "C with agradient of7 qclkbar. The plagioclase composition changes fromAp5 to Ap2. In path B, plagioclase, alkali feldspar, and melt arepresent at higher pressure, and decompression commences at1020 qC. The plagioclase composition changes from B"r to B"r;the alkali feldspar composition changes from B^, to B^r. In path

C, plagioclase, alkali feldspar, and melt are present at higherpressure, and decompression commences at 880 "C. The plagio-

clase composition changes from C*, to C"r; the alkali feldsparcomposition changes from Cn, to C^r. (b) Variations in propor-

tions of plagioclase and alkali feldspar (molo/o) at 5 kbar (dashed

curves) and 2 kbar (solid curves) with arrows indicating de-compressional pathways as described in a.

Ea

E 4 0

E9 3 0

o

8 2 0

s

c 6 0or

t+od

Sierra el Virulento lrachyle

<-s^--\---t)> ' \ \ ,

. - - \ - I . \

P=5 kbar Xw=0.10 P=5.2kbar Xw=0,15


prior to ascent. As a first case, if decompression occurswhen plagioclase is the only crystalline phase present at5 kbar (path A in Fig. 7), and decompression follows anaverage gradient of 7 'Clkbar (Sykes and Holloway, I 987),then the system must readjust itselfin order to reestablish

Or1 gOzl g Or79Oz1 9

C 100

80

01300 1200 1100 1000 900

equilibrium by undergoing some melting of the higherpressure plagioclase. This can be seen in the phase pro-portion diagram ofFigure 7b by the sliehtly negative slopeof the arrow Aps to Ap2. The effects on plagioclase com-position are indicated in Figure 7a, and they show thatthe lower pressure plagioclase would be more anorthiticthan the previously formed higher pressure plagioclase.Texturally, this may be indicated by an irregular bound-ary around cores of plagioclase (induced by partial melt-ing), mantled by a zone of more anorthitic plagioclase(reverse zone), followed by normal zoning (if a stepwisefractional crystallization process takes place).

If decompression does not begin until alkali feldsparhas appeared, and plagioclase is undergoing reaction witha liquid (path B in Fig. 7), then plagioclase will precipitateupon decompression, as shown by the positive slope ofthe phase proportion arrow B", to B", in Figure 7b. Thehigher pressure alkali feldspar on the other hand will un-dergo a significant amount of melting, as its lower pres-sure equilibrium abundance is about 350/o lower than at5 kbar (as shown by the arrow B^, to Bor). The plagioclasewill become slightly more anorthitic (but not show asmuch An enrichment as for path A), but Or poor. Thealkali feldspar may become reversely zoned with a slightincrease in Or and decrease in An content at the lowerpressure. As the result of decompression, the melt willbecome enriched in Or.

Path C indicates the efects of decompression on thehigh-pressure assemblage if decompression occurs onceplagioclase and alkali feldspar are coprecipitating (towardthe later stages of crystallization). In this case, it is pos-sible to get essentially no change in the abundances ofthephases and to affect only their compositions. For exam-ple, for path C in Figure 7b, the arrows (C", to C", forplagioclase and Co, to Co, for alkali feldspar) show a near-ly zero slope. However, as shown in Figure 7a, the slightAn enrichment and Or depletion of plagioclase and theOr enrichment of the melt, as was seen for path B, arestill manifested, but the magnitude of the enrichment ofeach component declines. The change in composition be-

Fig. 8. Calculated HrO-unbuffered equilibrium cryslalliza-tion path for the Sierra el Virulento trachyte (33.9 uno/o Or, 44.7wto/o Ab, 7.45 wto/o An, 14.0 wto/o Qz) at 5 kbar and 0.75 wto/o(X,,o:0.10) bulk HrO content. (a) Melt (L), plagioclase (Pl),and alkali feldspar (Af) compositions projected into the feldsparsystem. The symbol Qz indicates the onset of crystallization ofquartz. Arrowheads point to the direction ofdecreasing temper-ature. The three outlined regions refer to the plagioclase (P),mesoperthite (M), and cryptoperthite (C) of Parsons and Brown(1988). (b) Melt compositions projected into the system Ab-Or-Qz. Each inflection marks the onset of precipitation of a newcrystalline phase, plagioclase (Pl), alkali feldspar (Af), and quartz(Qz). The last inflection marks the beginning of stability of afluid phase. (c) Variations in proportions (molo/o) of melt (L),plagioclase (Pl), alkali feldspar (Af), and quartz (Qz) vs. temper-ature during equilibrium crystallization.

so

; 6 0

E . 4 0o

o

Sierra el Virulenlo trachyteP = 5 k b a r

Xw=0.10 (0-75 wt%)

O145Oz35

Weight%

Terryeratue fC)

800 700 600


comes smaller and smaller as decompression brings thetemperature closer and closer to the low-pressure solidustemperature. The change in alkali feldspar compositionupon decompression is more pronounced than in path B.Instead of the Or enrichment shown in path B, there willbe a slight Ab enrichment. Just as for plagioclase, thistrend will be enhanced closer to the low-pressure solidustemperature.

HrO-unbuffered paths of Qz-norrnative trachltesand syenites

Natural trachytes differ from compositions within thefeldspar system by containing ferromagnesian compo-nents, and they are commonly quartz or nepheline nor-mative. The Sierra el Virulento trachyte studied here isquartz normative, and the effect of a buildup of the Qzcomponent in the liquid was investigated computation-ally for this composition to determine the differences be-tween crystallization paths that are closer to natural com-positions and those within the feldspar system.

Figure 8a shows the projected path of the Qz-norma-tive Sierra el Virulento composition in the feldspar sys-tem for l0 molo/o bulk HrO content at 5 kbar. The effectof dilution of the feldspar components by the Qz com-ponent can be readily seen by comparing Figure 8a withFigures 5a and 6a. The crystallization path for the Qz-normative composition at l0 molo/o HrO is similar to thatfor higher HrO contents because of the temperature de-pression induced by dilution. Beyond the temperature dif-ferences, the only differences between this path and pathsin the feldspar system are the inflections in the feldsparand liquid paths caused by the precipitation of quartz,which diminishes the continual dilution of the feldsparcomponents (which had been occurring before Qz satu-ration through the buildup of a Qz component and HrOin the melt and through precipitation of the feldsparsthemselves). Figure 8b shows the melt path projected intothe haplogranite system showing the inflections in themelt evolutionary path more clearly.

Figure 8c shows the variations in phase proportionsduring crystallization of this composition. The odd regionof the 2 feldspar * L surface in the granite system isintersected by the melt path, and hence partial resorptionof plagioclase does occur for this composition. Impor-tantly, however, the even (cotecticlike) portion of the 2feldspar + Qz + L surface is reached upon saturationwith quartz, and coprecipitation offeldspars occurs oncequartz appears. The amount of resorption that will occuris very dependent upon the bulk HrO content. With suf-ficiently low bulk HrO contents (<0.2 wto/o), it may bepossible for the melt to intersect the very small odd re-gion of the 2 feldspar + Qz + L surface. At this point itmust again be pointed out that the predicted highest bulkHrO content for which an HrO-buffered path will enterthe odd region of the 2 feldspar * Qz + L surface atsome stage of the crystallization interval depends uponthe ternary solid solution model chosen. The model ofLindsley and Nekvasil (1989) overestimates the amount

ofternary solution in both feldspars at a given pressure,while the model of Nekvasil and Burnham (1987) quitelikely underestimates the amount of ternary solution. Useof the former model, therefore, results in an overesti-mation of the highest bulk HrO content (for which a crys-tallization path of a trachytic magma would enter the oddportion of the 2 feldspar + Qz + L surface), whereas thelatter model probably underestimates this value.

HrO-uunurFERED IATHS oF HIGH-TEMIERATUREGRANITES

Figure 9a shows the evolution of feldspars and meltduring equilibrium crystallization for the same modelrhyolite as in Figure 3 but under conditions oflow bulkHrO content. The crystallization path is very similar tothat for the Qz-free composition at a higher bulk H'Ocontent. Importantly, the evolutionary path of the feld-spars is very similar to that for the trachyte, in that, uponstabilization of two feldspars, plagioclase compositionschange mainly by decreasing Or content, whereas alkalifeldspar changes mainly by increasing Or content. Thispath is very different from that for wet, low-temperaturegranites (cf. Figs. 9a and 3a). However, this diference isonly manifested for HrO-unbuffered conditions. For H'O-buffered conditions, both the wet and dry model rhyoliteshow the same general path as seen in Figure 4a, with themain difference being in the extent of ternary solution. Infact, the wet HrO-unbuffered path does not even differvery much from the HrO-buffered path.

When Figure 9a is compared with Figure 8a, it is readi-ly apparent that toward the end stages of HrO-unbufferedcrystallization of a trachyte, the feldspar and melt com-positions are very similar to those of a granite or rhyolite.This implies that from the standpoint of the behavior ofthe feldspars, it is possible to derive low-temperature, wetgranitic melts from trachytic magmas simply by HrO-unbuffered crystallization. Trachytes crystallizing underHrO-buffered conditions could readily produce high-tem-perature granites during late stage crystallization.

Inrpr-rcarroNs FoR NATURAL syENrrEs, TRACHYTES,AND HIGH-TEMPERATURE GRANITES

A natural magma may undergo various stages of HrOevolution. During the early high-temperature stage, inwhich perhaps fayalitic olivine and pyroxenes precipitate,the HrO content would build up continuously, and crys-tallization would proceed under H'O-unbuffered condi-tions. However, once hydrous minerals (e.g., biotite) be-came stable, the buildup of HrO in the melt (because ofcrystallization of anhydrous phases such as feldspars)would result in the production of more biotite. As longas the melt contains sufficient ferromagnesian compo-nents to continue producing biotite, the magma wouldcrystallize under HrO-buffered conditions. This shouldcontinue until the abundance of the ferromagnesian com-ponents in the melt becomes too small to permit a suffi-cient increase in the abundance of biotite to buffer theHrO content. From this stage on, the HrO content of the

Sietra el Virulento rhyolite


bt*

80

1100 1000 900 800 700 600

Temperature (t)

Fig. 9. Calculated HrO-unbuffered equilibrium crystalliza-tion path for the model Sierra el Virulento rhyolite (28.5 *o/oOr,37.5 wto/o Ab, 5.88 wto/o An, 30.0 wto/o Qz) at 5 kbar and0.75 wto/o (X^,o:0.10) bulk H,O content. (a) Melt (L), plago-clase (Pl), and alkali feldspar (Af) compositions projected intothe feldspar system. Quartz appears soon after plagioclase, sothe inflections in the feldspar paths seen in Figure 7a do notoccur. Arrowheads point to the direction ofdecreasing temper-ature. (b) Variations in proportions (mol7Q of melt (L), plago-clase (Pl), alkali feldspar (Af), and quartz (Qz) vs. temperatureduring equilibrium crystallization. All phase proportions sum to( l - X " , " ) " 1 0 0 .

melt would once again increase and build up to saturationwith an aqueous fluid. Under such conditions, however,the same basic trends would be seen as for the HrO-un-buffered conditions, with, however, a chance of more ter-nary feldspar compositions and the possibility of moreresorptional behavior as well as greater Or depletion of

plagioclase and Or enrichment of alkali feldspar duringthe late stages of crystallization.

The effect of additional melt components and mineralphases must be considered when applying the above pathsto the interpretation ofnatural rocks. The ferromagnesianmelt components and the crystalline phases olivine, py-roxene, and biotite are predominantly those of interest.The olivine and pyroxene melt components will have adefinite effect on plagioclase composition. The anorthitecontent ofplagioclase is very sensitive to the activity ofthe An melt component. This sensitivity to even slightvariations in anorthite activity, when coupled with thedecrease in An activity because of the strong negativedeviations from ideality characterizing the mixing behav-ior of the anorthite and olivine (and to a lesser extentpyroxene) melt components, results in a significant de-crease in the An content of the plagioclase. These devi-ations from ideality are exemplified in the system An-Fo(Andersen, I 9 I 5; Osborn and Tait, 1952), where negativedeviations from ideality are so strong as to result in com-pound formation, as evidenced by the crystallization ofspinel. The effect of ignoring the ferromagnesian com-ponents can be seen from the work ofHousch and Luhr(1991). They compared calculated plagioclase composi-tions using the Nekvasil (1986) and Burnham and Nek-vasil (1986) models with experimentally grown plagio-clase presumed in equilibium with late stage melts frombasaltic magmas. As could be expected from the strongnegative deviations from ideality of the An melt com-ponent with ferromagnesian melt components, even smallamounts of these latter components in the melt resultedin calculated An contents in plagioclase that are too high.This is not a major problem for this analysis because onlythe trends are discussed. Additionally, the uncertainty oflocation of any compositional point in P-T space inducedby ignoring the ferromagnesian components is out-weighed by the differences induced by using various solidsolution models.

In addition to the effect of buffering the HrO activityof the melt, precipitation of biotite may have a significanteffect on the crystallization paths of the feldspars becausebiotite preferentially incorporates Or component relativeto the other feldspar components and changes the evo-lutionary path of the melt. It is difficult to predict themagnitude of the effect of biotite because it depends onthe stage of feldspar evolution at which biotite first ap-pears. The stability relations of high-temperature (e.g.,Ti-rich) biotite are only poorly known, and much exper-imental work remains to be done before the importanceof biotite in dictating the evolutionary paths of the feld-spars can be evaluated.

Because of the marked differences between the feldsparcompositional trends of HrO-buffered and HrO-unbuf-fered trachytes, it should be possible to use natural feld-spar zoning patterns to get an indication ofthe behaviorof HrO during crystallization of the magma. However,ternary feldspars in natural syenites commonly have un-dergone considerable subsolidus exsolution (e.g., Fuhr-

6ao

; 6 0'tro

E 4 0o

Sien'a el Virulento rhvolite


man et al., 1988), which has modified any previouslyexisting zonation. In rare cases, however, it is possible toglean some information on the nature of zoning withinternary feldspars. Careful work by Parsons and Brown(1988), for example, on the perthitic feldspars of theKlokken intrusion in South Greenland has yielded im-portant information regarding feldspar zoning in syenites.Their syenodiorite KB50 and syenite KB56 show threeclusters of feldspar compositions (shown in Fig. 8) thatseem to agree with the compositional types of feldsparsproduced by crystallization under HrO-unbuffered con-ditions. These clusters are, in contrast, quite different fromthose that might be expected for crystallization under HrO-buffered conditions (exemplified in Fig. 4a). Although thefeldspar zoning pattern is not truly definitive, Parsonsand Brown's (1988) analyses of zoned ternary feldsparsof the Klokken intrusion show plagiocase zoned in theusual manner, that is, by decreasing An content, sodicsanidine (their cryptoperthite) zoned by increasing Orcontent, and ternary oligoclase (their "mesoperthite") inwhich there is some evidence of zoning by decreasing Orcontent.

In volcanic rocks, rapid cooling makes late stage zo-nation and the compositions of groundmass phases moredifficult to obtain. Efforts by Ferguson (1978) to evaluatethe compositions of zoned feldspars in trachytic lavasfrom central Victoria indicate that most of the feldsparsfrom the Turritable Falls trachyte flow are zoned suchthat the feldspar compositions move subparallel to anisothermal solvus section as if the magma were bufferedwith respect to HrO content. However, he notes the pres-ence of pegmatoidal clots and veins in which the alkalifeldspar shows reverse zoning (in some cases from Oro,to Oru). This could be a natural example of a path thatinvolves early buffering of the HrO content and laterbuildup of HrO during crystallization.

Natural examples of crystallization of a single feldspararound the nose of the solvus are understandably rarebecause ofthe very low HrO activities required. A naturalexample may be found in the Mboutou layered gabbro-syenite-granite complex of northern Cameroon. Parsonset al. (1986) report feldspars forming a compositionalcontinuum with a compositional range from AnrrOr, toOrrrAn, in the Mboutou. The rock types within the com-plex, however, are quite diverse, and comparison of thefeldspars among these rock types requires that they berelated through a differentiation process.

Reverse zoning of alkali feldspar in Qz-normative tra-chytes, pantellerites, comendites, and phonolites havebeen widely reported, e.g., Carmichael (1965), Nash et al.(1969), Nicholls and Carmichael (1969), and the originofsuch zoning has remained a source ofdebate. Becausemany ofthese rock types show peralkaline characteristics,it was natural for early workers to link this zoning be-havior to peralkalinity through the postulation of com-plex fractionation curves for feldspars in alkali-rich mag-mas (e.g., Nash et al., 1969). The results of the calculationssummarized above indicate that reversely zoned alkali

feldspar may result from as simple a process as stepwiseequilibrium crystallization of a dry felsic magma and doesnot require excess alkalis. Such HrO-unbuffered behaviormay also explain the reverse zones in alkali feldspars fromrocks such as pantellerites, in which only a single feldsparcrystallizes (alkali feldspar) because the Or enrichmentcan be readily linked to the expansion of the An-freealkali feldspar solvus.

From the above analysis of feldspar crystallization be-havior, it is reasonable to conclude that differentiation ofhigh-temperature syenitic magmas can produce magmaswith the compositional characteristics of wet graniticmelts, as long as the syenitic magma crystallizes underHrO-unbuffered conditions (at least during the late stagesof its evolution). In contrast, high-temperature syeniticmagmas crystallizing under H'O-buffered conditions canonly give rise to high-temperature granitic magmas. Forany particular association of syenites and granites, how-ever, a fractionation relationship between the two rocktypes can be accepted only after detailed isotopic andtrace element analysis.

The above discussion centered on an investigation ofequilibrium crystallization pathways. Natural crystalli-zation processes are probably a combination of equilib-rium and fractional crystallization paths. For high-tem-perature felsic magmas, ideal fractional crystallizationpaths can differ significantly from equilibrium crystalli-zation paths. Melts reaching the 2 feldspar * L surfaceon its even side during fractional crystallization will evolvealong the surface, precipitating two feldspars as the bulkcomposition changes and the HrO content increases. Oncethe neutral curve is reached, only a single feldspar be-comes stable, and the melt moves off of the surface andonto the ternary alkali feldspar liquidus surface. Withcontinual buildup of HrO, however, the melt will quitelikely reach the surface again because of the shift of thesurface toward lower An contents and its extension to-ward Ab with increasing HrO content. The details of step-wise fractional crystallization and of equilibrium meltingand stepwise fractional melting paths will be the subjectsoffuture articles.

The calculated paths discussed above serve to indicatethe importance of feldspar zoning in providing importantinformation regarding conditions and processes relevantto the crystallization of felsic magmas. They further il-lustrate that before we can evaluate the efects of opensystem and kinetic processes at work in magmatic sys-tems, we must further our understanding of the equilib-rium state toward which systems strive. Particularlyamong felsic rocks, where bulk compositions are so sim-ilar and the mineral relations fairly indistinct, feldsparzoning may provide an important key to unraveling theevolutionary history of a magma.

AcrNowr,nocMENTs

The careful reviews of S.A. Morse and J.S. Beard contributed greatly

to this manuscript. Support for compulational services from the Depart-ment of Earth and Space Sciences, SUNY-Stony Brook, and partial sup-


port from NSF grant EAR-8916050 and the Center for High PressureResearch are gratefully acknowledged

RrrnnnNcrs crrEDAbbott, R.N., Jr. (1978) Peritectic relations in the system An-Ab-Or-Qz-

HrO. Canadian Mineralogist, 16, 245-256.Andersen, O. (1915) The system anorthite-forsterite-silica. American

Journal of Science, 39, 407-454.Burnham, C.W., and Nekvasil, H. (1986) Equilibrium properties of gran-

ite pegmatite magmas. American Mineralogist, 7 1, 239-261.Carmichael, I.S.E. (1963) The crystallization of feldspar in volcanic acid

liquids. Quarterly Journal of the Royal Society of London, I 19, 95-I 3 1 .

-(1965) Trach).les and their phenocrysts. Mineralogical Magazine,34, rO7-r25.

Ewart, A. (198 l) The mineralogy and chemistry ofthe anorogenic Tertia-ry volcanics of S.E Queensland and N.E. New South Wales. JournalofGeophysical Research, 86, Bl l, 10242-10256.

Ferguson, A.K. (1978) A mineralogical investigation of some trachytelavas and associated pegmatoids from Camel's Hump and TurritableFalls, central Victoria. Journal of the Geological Society of Australia,25, t85-t97.

Fuhrman, M., Frost, R., and Lindsley, D.H. (1988) Crystallization con-ditions of the Sybille Monzosyenite, Iaramie Anorthosite Complex,Wyoming. Joumal of Petrology, 29, 699-729.

Housch, T.B., and Luhr, J.F. (1991) Plagioclase-melt equilibria in hy-drous systems. American Mineralogist, 7 6, 477 -492.

Johannes, W. (1978) The melting of plagioclase in the system Ab-An-HrOand Qz-Ab-An-H2O at PH2o : 5 kbar, an equilibrium problem. Con-tributions to Mineralogy and Petrology, 66,295-303.

Lindsley, D.H., and Nekvasil, H. (1989) A ternary feldspar model for allreasons. Eos. 70. 506.

Moll, E.J. (1981) Geochemistry and petrology of mid-Tertiary ash flowtuffs from the Sierra el Virulento area, eastern Chihuahua, Mexico.Journal ofGeophysical Research, 86, Bl I, 10321-10334.

Nash, W.P., Carmichael, I.S.E., and Johnson, R.W. (1969) The mineral-ogy and petrology of Mount Suswa, Kenya. Journal of Petrology, 10,409-439.

Nekvasil, H. (1986) A theoretical thermodynamic investigation of thesystem Ab-An-Or-Qz(HrO) with implications for melt speciation, 26E

p. Ph.D thesis, The Pennsylvania State University, University Park,Pennsylvania.

-(1988a) Calculation of equilibrium crystallization paths of com-positionally simple hydrous 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, 966-982.

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

-(1991) Ascent of felsic magmas and formation of rapakivi. Amer-ican Mineralogist, 7 6, 1 27 9- 1290.

Nekvasil, H., and Bumham, C.W. (1987) The calculated individual effectsofpressure and water content on phase equilibria in the granite system.Geochemical Society Special Publication, l, 433-44 5.

Nekvasil, H., and Lindsley, D.H. (1990) Termination of the 2 feldspar +liquid curve in the system Ab-Or-An-HrO at low HrO contents. Amer-ican Mineralogist, 7 5, 107 l-1079.

Nicholls, J., and Carmichael, I.S.E. (1969) Peralkaline liquids: A petro-

logical study. Contributions to Mineralogy and Petrology, 20,268-294.Novak, S.W., and Mahood, G.A. (1986) Rise and fall of a basalt-trachlte-

rhyolite magma system at Kane Springs Wash caldera, Nevada. Con-tributions to Mineralogy and Petrology, 94,352-373.

Osborn, E.F., and Tait, D B (1952) The system diopside-forsterite-an-orthite. American Journal ofScience, Bowen Volume, 413-433.

Parsons, I, and Brown, W.L. (1988) Sidewall crystallization in the Klok-ken intrusion: Znned tenary feldspars and coexisting minerals. Con-tributions to Mineralogy and Petrology, 98,431-443.

Parsons, I., Brown, W.L., and Jacquemin, H. (1986) Mineral chemistryand crystallization conditions of the Mboutou layered gabbro-syenite-granite complex, Nonh Cameroon. Journal of Petrology, 27, 1305-1329 .

Sykes, M.L., and Holloway, J.R. (1987) Evolution of granitic magimaduring ascent: A phase equilibriurn model. Geochemical Society Spe-cial Publication, l, 447 -461.

Tuttle, O.F., and Bowen, N.L. (1958) Origin of granite in light of exper-imental studies in the system NaAlSirO'-KAlSiOsSiOyH'O. Geolog-ical Society of America Memoir, 74, 153 p.

MeNuscnrsr RECETVED Jurv 19, l99lMrNuscnrp.r ACcEFTED Jmruenv 9, 1992

ternary feldspar crystallization in high-temperature felsic magmas › ammin › am77 ›...

Documents