Contrib Mineral Petrol (1995) 118 : 420M32 �9 Springer-Verlag 1995

Yuenyong Panjasawatwong �9 Leonid V. Danyushevsky Anthony J. Crawford �9 Keith L. Harris

An experimental study of the effects of melt composition on plagioclase - melt equilibria at 5 and 10 kbar: implications for the origin of magmatic high-An plagioclase

Received: 31 August 1993/Accepted: 20 May 1994

Abstract An experimental investigation of plagioclase crystallization in broadly basaltic/andesitic melts of variable Ca# (Ca/(Ca+Na)* 100) and AI# (A1/(AI+Si)*I00) values and H20 contents has been carried out at high pressures (5 and 10 kbar) in a solid media piston-cylinder apparatus. The H20 contents of glasses coexisting with liquidus or near-liquidus pla- gioclases in each experiment were determined via an FTIR spectroscopic technique. This study has shown that melt Ca# and AI#, H20 content and crystalliza- tion pressure all control the composition of liquidus plagioclase. Increasing melt Ca# and AI# increase An content of plagioclase, whereas the effect of increasing pressure is the opposite, However, the importance of the role played by each of these factors during crys- tallization of natural magmas varies. Melt Ca# has the strongest control on plagioclase An content, but melt AI# also exerts a significant control. H20 content can notably increase the An content of plagioclase, up to 10 mol % for H20-undersaturated melts, and 20 mol % for H20-saturated melts. Exceptionally calcic plagio- clases (up to An~00) in some primitive subduction- related boninitic and related rocks cannot be attributed to the presence of the demonstrated amounts of H20 (up to 3 wt %). Rather; they must be due to the involve- ment of extremely refractory (CaO/Na20 > 18) magmas in the petrogenesis of these rocks. Despite the refrac- tory nature of some primitive MORB glasses, none are in equilibrium with the most calcic plagioclase (An94) found in MORB. These plagioclases were likely pro-

Y. Panjasawatwong (N~) - L.V. Danyushevsky �9 A.J. Crawford K.L. Harris Geology Department, University of Tasmania, GPO Box 252C, Hobart, Tasmania, Australia 7001 Y. Panjasawatwong Department of Geological Sciences, Chiang Mai University, Chiang Mai, 50002, Thailand

Editorial responsibility: T.L. Grove

duced from more refractory melts with CaO/Na20 = 12-15, or frOlyl melts with exceptionally high A1203(> 18%). Magmas of appropriate compositions to crystallize these most calcic plagioclases are sometimes found as melt inclusions in near liquidus phenocrysts from these rocks, but are not known among wholerock or glass compositions. The fact that such melts are not erupted as discrete magma batches indicates that they are effectively mixed and homogenized with volumet- rically dominant, less refractory magmas. The high H20 contents (~ 6 wt%) in some high-A1 basaltic arc magmas may be responsible for the existence of pla- gioclases up to An95 in arc lavas. However, an alterna- tive possibility is that petrogenesis involving melts with abnormally high CaO/Na~O values (> 8) may account for the presence of highly anorthitic plagioclases in these rocks.

Introduction

Extremely calcic plagioclase compositions (Ansi100) are quite common in high-A1 basalts and gabbroic nodules in island-arcs (Arculus and Wills 1980; Brophy 1986; Crawford et al. 1987; Stolz et al. 1988; Beard and Borgia 1989), in boninitic lavas and gabbros from intraoceanic forearcs (Falloon and Crawford 1991; Thy et al. 1989), and in mid-ocean ridge basalts (MORB) (Donaldson and Brown 1977; Fisk 1984; Stakes et al. 1984; Autio and Rhodes 1984; Price et al. 1986; Koyaguchi 1986; Wilkinson 1982; Sinton et al., 1993). One-atmosphere experiments suggest that such highly anorthitic plagio- clases only crystallize from melts with unusually high CaO/Na20 values (e.g. Duncan and Green 1987). Other controls on plagioclase-melt equilibria include: 1. the presence of H20 in a magma, as proposed by Kudo and Weill (1970) and Arculus and Wills (1980) and demonstrated by Sisson and Grove (1993),

2. high melt A1203 conten t (Beard and Borgia 1989), 3. elevated crystal l ization pressure, and 4. oxidat ion state of the magma.

We have carried out an experimental invest igat ion of plagioclase crystal l ization to evaluate the effects of fac- tors 1 to 3 above. To investigate the role of H20, our experiments were based a r o u n d an island arc tholeiitic basalt from Vanua tu . The effects of melt A1 content have been studied by adding various propor t ions of plagioclase to the chosen start ing composi t ion. To document the effects of pressure, we have run the experiments at 5 and 10 kbar.

Experimental techniques

Starting materials

Eight starting compositions (Table 1) were prepared by mixing AR grade oxides (SiO2, TiO2, A1203, Fe203, MnO and MgO), carbon- ates (CaCO3, Na2CO3 and K2CO3) and Ca2P207 under acetone in an agate mortar and firing at 1000~ for 12 h to drive off all CO2 and react the oxides. Pure fayalite, synthesized ti'om a stoichio- metric mixture of Specpure Fe metal sponge, Fe203 and SiOz heated for 3 days in an evacuated silica tube at 1000~ was added to each mixture in the amount required to yield the desired FeO and SiO2 concentrations in the bulk composition. After mixing under ace- tone, each starting mix was dried by heating at 450~ under argon. To avoid possible Na20 loss by volatilization from the melt, the prepared starting mixes were not melted before use in experiments.

Mix A, with bulk Ca# (100*Ca/(Ca+Na)) of 79.4 and AI# (100*A1/(AI+Si)) of 32.9 (Table 1) is chemically equivalent to an arc tholeiite (sample 71047) from western Epi, Vanuatu Arc (Barsdell and Berry 1990) + 10% plagioclase Ans0. Compositions for Mixes E, F, B and C were calculated from the composition of Mix A to obtain normative plagioclase compositions of An = 87.0, 61.4, 50.0 and 30.0, respectively, by fixing molecular proportions of normative minerals other than albite and anorthite, and varying CaO, Na20, A1203 and SiOz Compositions for mixes H, I and G were calculated by adding 20% plagioclase (Am94), 20% plagioclase (An6~ 8) and 30% plagioclase (Am6.3) to mixes A, B and C, respec- tively. The purpose of these additions was to obtain plagioclase as a liquidus phase lbr these compositions under hydrous conditions without changing bulk Ca#. These starting compositions (Table 1) almost cover the compositional range of high-Al basalts (Crawford et al. 1987) in terms of Ca# and AI# (AI# = 27 at Ca# = 35-69 to AI# = 34 at Ca# = 91).

Analytical techniques

Mineral and glass compositions in this study were determined at the University of Tasmania using a fully automated Cameca SX50 electron microprobe fitted with a WDS system at operating condi- tions of 15 kV and 10 nA. To minimize Na loss, Na was counted first (20 and 10 s for peak and background, respectively), and a defocused beam (20-30 gm in diameter) was used to analyze glass. Mineral analyses were done in spot mode (1-2 gm) or area mode depending on crystal sizes. X-ray intensities werc corrected for dead- time, background, and matrix effects using the Cameca ZAF rou- tine. All quoted analyses were made on adjacent plagioclase-glass pairs.

The H20 contents of glasses coexisting with liquidus or near- liquidus plagioclase in each experiment were determined via an

421

FTIR spectroscopic technique, exactly as described in detail by Danyushevsky et al. (1993). Each H20 analysis quoted is the aver- age of two determinations, and the accuracy is estimated to be 10 rel.%.

High pressure experiments

All anhydrous and hydrous experiments were carried out in the experimental petrology laboratory, Geology Department, University of Tasmania, by melting synthetic powder (melting experiments) in a high pressure, solid media 0.5 inch piston cylin- der apparatus using the 'piston in' tcchnique, similar to that of Boyd and England (1960) and Green and Ringwood (1967). Temperature was measured using a Eurothenn 818P programmable temperature controller with Pt/Ptg0Rh ~o thermocouple and was controlled within +2 ~ C of the set point. Accuracy in the pressure determination, which includes a minus 10% pressure correction for runs with talc/pyrex sleeves, is considered to be +0.5 kbar. Runs were quenched at run pressure by cutting the power to the furnace.

Sample containers, designed for 11 18 mg of sample powder. were specpure Fe for anhydrous experiments, and Ag75Pd2~, AgsoPds0 and Pt for hydrous experiments. Distilled water was added, in the amount required to each run, with a Hamilton microsyringe to the bottom of the capsule and the weight added was checked with a Mottler microbalance. The sample powder was then loaded into the capsule. Specpure Fe capsules were covered by snug-fitting lids and sealed by pressure, whereas the noble metal containers were welded shut and the welds checked for integrity. Anhydrous experiments were carried out at pressures of 5 and 10 kbar using a talc-pyrex assembly. Oxygen fugacity was not controlled to a particular buffer (e.g. iron-wustite, nickel-nickel oxide etc.). However, using specpure Fe caps ules, oxygen fugacity approximates iron-wustite buffer under anhydrous conditions. Hydrous experiments were performed only at 5 kbar using a NaCl-pyrex assembly. Conditions in these runs, particularly those where some iron is lost to alloy with the capsule, are relatively more oxidizing but the oxygen fugacity probably does not exceed quartz-fayalite-magnetite buffer.

At the end of each run, the capsule was opened and the run products were examined by microscope, electron microprobe cou- pled with backscattered electron imaging, and IR spectroscopy. Representative run products are shown in Fig. 1.

Compositions of synthetic mixes were checked by microprobe analysis of above liquidus glasses (Table 1). As expected, Fe gain occurred in anhydrous glasses using specpure Fe containers and in contrast, Fe loss occurred in hydrous glasses using Ags0 Pds0 and Pt containers. However, in all the above liquidus runs, bulk Ca# and AI# values are close to those for theoretical starting composi- tions.

Details of experimental conditions and products are listed in Table 2. Compositions of glasses, including their H20 contents, and plagioclase compositions produced in all our experiments are given in Table 3.

Proof of attainment of equilibrium

In this study, equilibrium between plagioclase and melt was checked by phase homogeneity. To minimize the possible effect of short- range quench modifications of the glasses (Fig. lb), the melt com- position was obtained as the average of large area scans on pools of glass. Almost all glasses show a 2or (95% confidence level) of Ca# values less than 1.5, with a few showing more dispersion (2c~ = 2-3). Similarly, Ca# values for plagioclases lie within a 2~ of 3.

It has been demonstrated that in hydrous systems the reaction rate between plagioclase and melt decreases exponentially with decrcasing temperatures, and that plagioclase-melt equilibria can be achieved within one hour at 1000~ (Johannes 1978). As all pla- gioclase-melt pairs reported herein are fiom experiments at tern-

422

peratures above 1000~ and run durations of at least 6 h, dis- equilibrium is unlikely to have occurred. Finally, using extremely fine-grained synthetic mixes as starting compositions favors attain- ing equilibrium within relatively short run times.

Experimental results

In discussing the results of our study of plagioclase-melt equilibria, we have chosen to express melt compositions in terms of the following parameters: 1. mole fraction of CaA1204 (XcaAi204) , a s used by Grove et al. (1992) in their plagioclase-melt equilibria model, and 2. 100*A1/(AI+Si) (AI#) and 100*Ca/(Ca+Na) (Ca#), as they reflect the dominant solid solution in plagioclase.

Fig. 1A Backscattercd electron image photograph of run T-2990. Dark gray is plagioclase and bright white is iron globule from specpurc Fe capsule. B Backscattered electron image photograph of run T-3085 showing large crystals of plagioclase (dark gray) with well-developed plagioclase quench evident as projections from the crystal corners into the surrounding glass. Bright white is iron glob- ule from specpure Fe capsule. C Backscattered electron image pho- tograph of run T-3137. Dark grav is plagioclasc, light gray is clinopy- roxene and bright white is vesicle infilled with polishing powder

Anhydrous experiments

Our experiments show a significant dependence of equi- librium plagioclase composition on melt Ca#, melt AI#, XC~A1204, and also pressure (Fig. 2A, B, C). The effect of increasing Ca#, AI#, and Xc~A~2o 4 is to increase An, whereas in contrast, increasing pressure tends to decrease the An content of equilibrium plagioclase. However, the effect of pressure is more clearly demonstrated by AI# (Fig. 2B) and Xc,AJ2o4(Fig. 2C) than Ca# (Fig. 2A). Interestingly, we note that Al# of the melt can describe plagioclase-melt equilibria for anhydrous conditions between 5 and 10 kbar at least as well as more complex parameters of melt composition such a s XCaA1204,

Despite this general strong compositional control, we note a significant range of plagioclase An contents in equi- librium with melts of the same Ca#, AI# or XCaAbO4 (10 mol % An variation tbr Ca# and up to 15 tool % An for AI# and XC~A1204) at a given pressure. We used the model of Grove et al. (1992) to test whether this variability is a result of inter-relationships of different parameters of melt composition (Fig. 2D). As this model reproduces our experiments with an accuracy always better than 5% An, which is the accuracy claimed for this model, we conclude that the variations observed reflect differences in melt compositions. However, as is evident from Fig. 2D, the model of Grove et al. (1992) tends to underestimate pla- gioclase An contents for highly anorthitic plagioclases (An>80) and overestimate An contents for low anorthitic plagioclases. This model also displays a systematic difference between our 5 and 10 kbar experiments, yield- ing higher An for 10 kbar runs. The most probable rea- son for these systematic deviations is that our experiments used compositions outside the range used by Grove et al, (1992) to derive their model.

To demonstrate that observed deviations of our exper- iments from the model of Grove et al. (1992) result from use of a different data set, we have derived for our anhy- drous experiments a simple empirical equation that links

423

Table 1 Starting compositions used in partial melting experiments. Analysis of rock standard BCR-1 (Gladney et al. 1990), fused to a glass on an iridium strip, obtained by the same analytical technique, is included in this table for comparison (anal. analyzed values, reeom. recommended values, n number of analyses) Mix 71047 E b A ~ H d F bx B ~ I ~

N g C h E X i C h E X i E X i C h C h E X i C h E X i E X i C h

r u n # 3094 j 2986 j 3017 k 3092 j 2945 j 2946 j

n 3 4 5 3 4 3

SiO2 48.27 46.15 45.15 48.25 47.15 51.00 48.24 5 0 . 0 2 50.37 5 1 . 7 8 51.19 51.30 51.95 TiO2 0.55 0.50 0.56 0.50 0.54 0.63 0.40 0.50 0.59 0.50 0.57 0.61 0.40 A1203 1 8 . 5 8 21.48 20.79 20.06 19.33 20.05 22.70 18.87 18.56 17.68 17.07 17.18 20.21 Fe203 1.16 1.04 1.04 - 0.83 1.04 - 1.04 - 0.83 FeO 9.40 8.46 10.93 8.46 11.03 4.92 6.77 8.46 9.02 8.46 10.50 10.29 6.77 MnO 0.19 0.17 0.19 0.17 0.17 0.11 0.14 0.17 0.18 0.17 0.25 0.29 0.14 MgO 6.87 6.18 6.17 6.18 6.17 6.82 4.94 6.18 6.21 6.18 6.17 6.15 4.94 CaO 12.62 14.65 14.91 12.99 13.37 14.00 13.63 ll.60 11.97 10.21 10.45 10.33 10.71 Na20 1.82 0.89 0.96 1.86 1.87 2.03 1.95 2.68 2.66 3.49 3.43 3.45 3.66 K~O 0.45 0.41 0.35 0.41 0.37 0.43 0.33 0.41 0.44 0.41 0.37 0.39 0.33 P 2 0 5 0.08 0.07 0.07 0.07 0.07 0.07 0.06 0.07 0.07 0.07 0.07 0.07 0.06 AI# 31.20 35.42 35.18 3 2 . 8 8 3 2 . 5 8 31.66 35.67 3 0 . 7 8 3 0 . 2 8 28.69 28.21 28.30 31.44 Ca# 79.30 90.10 89.56 79.42 79.80 79.21 79.42 70.56 71.32 6 1 . 7 8 62.74 62.33 61.79 ~Mix A = Sample 71047 + 10% plagioclase (Ans0) b Mixes B, C, E and F are designed to retain the same normative composition except for the composition of normative plagioclase (An73 = A, Ans0 = B, An30 = C, An87 = E, An6~.4 = F ~ 50% M i x A + 50% MixB d Mix H = Mix A + 20% plagioclase (Anvg.42)

Mix I = Mix B + 20% plagioclase (An~wa) fMix G = Mix C + 30% plagioclase (An4629) gN = natural sample of high A1 basalt from western Epi, Vanuatu (71047, Bardsell and Berry 1990) h C ---- calculated composition used to combine oxides during preparation of the mixes (see text for details) i E X = glass quenched from liquidus experiments J Anhydrous experiment, Fe Capsule

melt composi t ion in terms o f C a # and Al#, pressure and temperature, with composi t ion o f equilibrium plagioclase:

An 0nol%) = A/T(K) + B + C * P(kbar)/T(K) + D * In(Ca#reel0 + E * ln(Al#m~t0 (1)

for which statistical regression of our data for 21 exper- iments (Table 3) provides the following coefficients:

A = -63970, B = -164.10, C = -2575.3, D = 41.836, E = 33.434.

The s tandard deviation o f A n determined f rom this equat ion is 3 mol % (Fig. 3A), which is equivalent to the dispersion o f A n values in individual experiments, as noted above.

However , we emphasize that this equat ion is not sup- posed to describe plagioclase melt equilibria outside the P, T and composi t ional range, and relatively reduced con- ditions o f our experiments. To demonst ra te this we have plotted on Fig. 3B experimental results of Bender et al. (1978), Stolper (1980), Thy (1991), F r a m and Longhi (1992) and Grove et al. (1992), all performed in graphite capsules between 5 and 10 kbar on composi t ions broadly similar to ours. As can be seen f rom Fig. 3B, our equat ion can describe them with an accuracy better than 4 m o l % An without any systematic deviations. Those experiments o f Baker and Eggler (1987) on composi t ions falling within our b road composi t ional range (Al# >23, plagioclase composi t ions >An45) are well described by

our equation, whereas their composi t ions with lower AI# show significantly lower calculated An content than mea- sured. Similarly, two experiments o f D u n c a n and Green (1987) are no t well accounted for by our equation. These were performed on boninitic composi t ions which differ f rom the remainder o f the experimental database in hav- ing simultaneously high SiO2 (>52 wt%) and Mg O (>9 wt %).

In summary, our results are consistent with published studies and confirm that the composi t ion o f equilibrium plagioclase depends on both the melt A l# and Ca# ; fur- thermore, at a given melt Ca# , variable AI# can produce an An variat ion in equilibrium plagioclase o f up to 10 mol%.

Figure 4 presents a compila t ion o f 1 atm experimen- tal data f rom the literature, plus published experimental data at 5 to l0 kbar, and shows that increasing pressure leads to decreasing An contents in liquidus plagioclase relative to 1 atm, in agreement with conclusions o f Marsh et al. (1990) and F r a m and Longhi (1992). Fur thermore , as already noted above, the effect o f pressure on An con- tent o f plagioclase is less evident using C a # than on plots using AI# o r XcaA1204(Fig. 4).

HaO-bearing experiments at 5 kbar

T h e H 2 0 contents o f glasses in equilibrium with plagio- clase in our 5 kbar experiments as determined by F T I R

424

C b G f B C R - 1

C h E X i C h E X i ana l . /

3052 j 3073 j recom.

9 3 3

54.93 54.83 55.41 53.75 55.58/55.40 0.50 0.60 0.35 0 .35 2.19/2.24

15.56 15.05 19.19 18.39 13.84/13.65 1.04 - 0.73 - 8.46 9.85 5.92 9 .13 12.59/12.40 0.17 0.25 0.12 0 .06 0.14/0.18 6.18 6.17 4.33 4 .45 3.58/3.48 7.72 8.00 8.29 8 .55 7.20/6.95 4.95 4.84 5.32 5 .06 3.33/3.27 0.41 0.42 0.29 0 .25 1.55/1.69 0.07 0.07 0.05 0.07 -

25.02 24.44 28.98 28.74 - 46.29 47.74 46.29 48.29

vary between 1.3% and 5.8% (Table 3), which indicate that our runs were significantly H20-undersaturated. As for the anhydrous experiments, we note a significant range in An contents at any fixed Ca# value of the melt (Fig. 5A). In general, plagioclases in HzO-bearing experiments are more anorthitic than those crystallized at similar P-T conditions and melt Ca# in anhydrous experiments (Fig. 5A). However, there appears to be no direct corre- lation between H20 contents and this difference in pla- gioclase compositions. For our experiments the maximum An difference between hydrous and anhydrous experi- ments occurs around melt Ca# values from 60 to 70. At higher and lower Ca# values, the hydrous curve approaches the anhydrous curve, and for Ca# values below 50 there is no obvious difference between anhy- drous and hydrous runs.

Also shown with our data on Fig. 5A are H20-under- saturated experiments by Baker and Eggler (1987) at 2 and 5 kbar. These are restricted to melt compositions with Ca#<55, and in this compositional range overlap with our experiments, showing no significant increase in An content compared to 5 kbar anhydrous runs. In addition, we note that HzO-undersaturated experiments do not show any difference between 2 and 5 kbar plagioclase compositions, a point consistent with our earlier obser- vation that pressure effects on plagioclase composition are not evident on plots using melt Ca#.

Figure 5A also shows H20-saturated experiments by Sisson and Grove (1993) and Housh and Luhr (1991) at 2 and 4 kbar. These data define a coherent trend, and demonstrate a strong effect of H20 saturation on increas- ing plagioclase An content. As noted above for anhydrous and HzO-undersaturated experiments, any pressure effect on plagioclase composition is not shown by plots using melt Ca#. We argue therefore, that plots of An versus melt Ca# are valuable in their being able to demonstrate

the effect of 1-t20 o n plagioclase compositions indepen- dent of pressure.

Unlike H20-undersaturated conditions, H20 satura- tion causes a large increase in plagioclase An content over the entire range of melt Ca# < 80. The magnitude of this increase reaches its maximum of around 20 tool % An in the range -45<Ca#<~65 and gradually decreases with increasing Ca#. Although presently available data are insufficient to test fully the effect of increasing H20 con- tents towards saturation, we suggest that the actual achievement of H20-saturated conditions exerts a much stronger influence on plagioclase composition than increasing amounts of H20 in H20-undersaturated con- ditions. For example, for melt Ca#-60, H20-undersatu- rated 5 kbar experiments with ~5.5 wt % H20 show only half the increase in plagioclase An content as do H20- saturated experiments at 2 kbar with -6 wt % H20.

Housh and Luhr (1991) developed a model describ- ing plagioclase-melt equilibria for H20-saturated condi- tions. Their model accurately reproduces the H20-satu- rated experiments of Sisson and Grove (1993) (Fig. 5B). It fails, however, to match the H20-undersaturated exper- iments of this study, and Baker and Eggler (1987), which both plot between H20-saturated and anhydrous experi- ments (Fig. 5B). This may suggest, as we proposed above, that there is a difference in the effect of H20 on plagio- clase compositions dependent on H20-saturated or under- saturated conditions.

In contrast to the Ca# plot, both the very similar AI# and Xc~At2O4 plots (Fig. 5C, D) do not clearly show the effect of H20 on plagioclase compositions. Rather, as in the case of anhydrous melts, they demonstrate a general dependence of plagioclase compositions on pressure.

In summary, for hydrous melts a large (-20 mol % An) increase of plagioclase An content occurs only for H20-saturated melts at pressures above 2 kbar in the range ~45<Ca#<-65. For example, An90 can crystallize from a H20-saturated melt with Ca# as low as 60; the same composition anhydrous should crystallize plagio- clase with the maximum An content ~70 at a high Al# of 30, and - 60 at AI# of 25. For H20-undersaturated conditions the maximum observed effect is ~ 10 mol% An at 5 kbar and 5-6 wt % H20 in the melt. Lower H20 contents at either 5 or 2 kbar produce still less difference.

Application of experimental results to natural magmas

Mid-ocean ridge basalts (MORB)

As has been demonstrated by numerous experimental studies, and petrological studies of natural MORB, high- An plagioclase can crystallize from primitive MORB magmas. These magmas crystallize under anhydrous con- ditions, usually at pressures from 6 to 2 kbar (e.g. Kinzler and Grove 1992). Although our anhydrous 5 kbar

Table 2 Summary of experi- mental run conditions and

Run No T(~ P(kbar) H20(wt %) Duration (h) Capsule PRODUCTS

425

products. (L liquid, Pl plagioclase, Cpx clinopyrox- ene, Ol olivine, Amp amphibole, Tr trace, V vapor phase; H20 approx.% H20 added to experimental charges)

Mix E T-3094 1340 5 dry 9.5 Fe L T-3076 1320 5 dry 48 Fe L+PL T-3127 1200 5 2 24 Pt L T-3083 1180 5 2 24 Pt L+PI+Cpx T-3086 1095 5 5 24 Ags0Pds0 L+PI+V

Mix A T-2986 1300 5 dry 20.5 Fe L T-2990 1275 5 dry 20.7 Fe L+P1 T-3037 1300 10 dry 24 Fe L+P1 T-3040 1292 10 dry 48 Fe L+P1 T-3008 1285 10 dry 24 Fe L+P1 T-2997 1275 10 dry 20.5 Fe L+P1 T-3000 1265 10 dry 23.1 Fe L+P1 T-3017 1160 5 2 12 Agsc~Pdso L T-3035 1135 5 2 23 AgsoPdso L+PI+V T-3026 1060 5 5 16 AgsoPdso L+PI+Cpx+OI+V

Mix H T-3139 1305 5 dry 48 Fe L+P1

Mix F T-3092 1270 5 dry 20.5 Fe L T-3060 1250 5 dry 24 Fe L+P1 T-3064 1120 5 2 24 AgsoPdso L+PI+O1 T-3080 1020 5 5 49 AgwPd25 L+OI+V T-3107 1000 5 5 47 Ag75Pd25 L+OI+PI+V T-3103 980 5 5 33.5 Ag75Pd25 L+OI+PI+V

Mix B T-2945 1260 5 dry 17 Fe L T-2946 1235 5 dry 17.7 Fe L T-2951 1220 5 dry 18 Fe L+P1 T-2956 1205 5 dry 17 Fe L+P1 T-2953 1190 5 dry 17.3 Fe L+PI+O1 T-3038 1230 10 dry 24 Fe L+Pl+Cpx T-3003 1220 10 dry 24 Fe L+PI+Cpx T-2988 1080 5 2 10 AgsoPdso L+Cpx T-2999 1050 5 2 15.5 AgsoPdso L+Cpx+OI+V T-2998 1050 5 2 11.3 AgsoPdso L+Cpx+OI+V T-2972 1040 5 5 6.5 AgsiiPdso L+Cpx T-2978 990 5 5 24 Ag75Pd~5 L+OI+Cpx+V T-2981 940 5 5 24 Ag75Pd25 L+OI+Amp§

Mix I T-3147 1245 5 dry 24 Fe L+P1 T-3159 1275 10 dry 48 Fe L+P1 T-3250 1135 5 2 24 AgsoPd~o L+PI+V T-3156 1030 5 5 48 AgvsPd25 L+PI+Cpx+V T-3162 1020 5 5 48 AgT~Pd25 L+PI+Cpx+V

Mix C T-3052 1225 5 dry 24 Fe L T-3048 1210 5 dry 23.8 Fe L+O1 T-2957 1190 5 dry 17 Fe L+O1 T-2962 1182 5 dry 18 Fe L+OI+P1 T-2959 1175 5 dry 21.5 Fe L+OI+P1 T-3042 1225 10 dry 24 Fe L+PI+Cpx T-3010 1213 10 dry 24 Fe L+PI+Cpx T-3005 1200 10 dry 24 Fe L+PI+Cpx+O1 T-3019 1070 5 2 23 AgsoPdso L+PI+OI+Cpx+V

Mix G T-3073 1245 5 dry 24 Fe L T-3069 1220 5 dry 23.5 Fe L+PI T-3066 1180 5 dry 24 Fe L+PI+O1 T-3079 1255 10 dry 24 Fe L+P1 T-3085 1280 10 dry 24.5 Fe L+P1 T-3137 1140 5 2 24 AgsoPds0 L+Cpx+PI+V T-3078 1125 5 2 24 AgsoPdso L+Cpx+PI+V T-3104 1020 5 5 30 Ag75Pd25 L+Cpx+V T-3091 995 5 5 38.5 Ag75Pd2s L+Cpx+Pl+Amp(Tr)

426

Fig. 2 Correlations between plagioclase An content and A Ca# value, B AI# value, and C XCaAt204 in equilibrium melts lbr anhydrous experiments at 5 and 10 kbar. D Comparison of measured plagioclase An contents in our 5 and 10 kbar anhydrous experiments with values calculated using the model of Grove et al. (1992)

a 90 �84

0

~b

3 0 4 0

4 70

5 0 '

o

o - 5 k b - 10 kb

6'o 8'0 aoo c ~ # ( m e l t )

9 0 �84

~ 7 0

50

b

0

oOOA

0

0 0

O A

O OO &

A

23 25 2 7 29 31 33 35 37

Al#(melt)

gO �84

~ 7 0 '

5 0 �84

C

C9 O

@ o a A

O

OD A

O Lx

O 9O

~ 7 0 3 O

450

30 3 0 0.03 o.d7 0.il 0.15 30

X CaAl~04(melt)

d

lx

G r o v e e t a l . m o d e l

s'o 40 9'0 An observed

T a b l e 3 Compositions of quenched melts and plagioclases. (n number of analyses; number in parentheses 2 standard deviations)

Run 3076 ~ 3076 b 2990 3139 3060 a 3060 b 2953 2956 2951 3147 2959 2962 3066 3069 3003 3038 3159

Glasses n

SiO2 TiO2 A1203 FeO MnO MgO CaO Na20 K20 Total H20 Mg# c Ca#

AI#

3 3 4 4 3 3 3 3 3 8 9 8 3 3 5 4 6 44.44 44.65 46.27 46.49 48.61 47.99 50.60 50.48 50.83 51.63 54.83 55.61 54.77 53.66 49.40 50.09 50.54

0.62 0.61 0.60 0.46 0.59 0.56 0.69 0.56 0.60 0.51 0.59 0.71 0.56 0.38 0.61 0.59 0.50 20.21 20.06 18.38 22.42 17.10 17.24 14.50 16.02 16.43 18.76 15.45 14.37 14.58 17.81 15.90 16.75 18.99 11.42 11.06 11.51 9.43 10.86 10.82 12.61 11.74 10.19 8.83 10.60 10.05 9.88 10.00 13.58 11.25 9.48 0.11 0.15 0.21 0.11 0.23 0.19 0,21 0.22 0.21 0.11 0.15 0.14 0.10 0.18 0.21 0.16 0.13 6.73 6.90 6.17 4.91 6.57 6.49 5.58 6.04 6.49 5.33 4.71 4.89 5.03 4.59 5.83 5.73 5.00

14.75 15.13 12.85 14.07 11.92 11.76 10,18 9.94 10.40 10.55 7.82 8.36 7.92 8.15 9.26 9.70 10.82 0.97 0.92 1.81 1.82 2.66 2.54 3,60 3.48 3.50 3.66 5.03 4.92 4.96 4.95 3.55 3.69 3.52 0.40 0.33 0.40 0.28 0.41 0.41 0,60 0.50 0.49 0.40 0.45 0.39 0.43 0.27 0.58 0.52 0.32

99.66 99.81 98.19 99.99 98.95 98.00 98,57 98.98 99.13 99.79 99.63 99.45 98.23 99.98 98.93 98.47 99.29

51.23 52.65 48.86 48.13 51.88 51.67 44.09 47.84 53.17 51.83 44.20 46.45 47.57 45.00 43.35 47.59 48.46 89.37 90.10 79.68 81.02 71.26 71.89 60,96 61.19 62.17 61.43 46.21 48.42 46.88 47.61 59.02 59.23 62.98 (1.03) (0.42) (0.93) (1.05) (1.12) (0.84) (1.38) (0.75) (1.24) (1.19) (0.90) (1.83) (0.36) (0.53) (2.06) (0.73) (1.47) 34.90 34.62 31.89 36.24 29.31 29.74 25.25 27.22 27.59 29.99 24.93 23.35 23.88 28.11 27.50 28.27 30.69 (0.73) (0.41) (0.45) (0.34) (0.61) (0.34) (0.39) (0.16) (0.82) (0.35) (0.32) (l.09) (0.90) (0.39) (0.40) (0.43) (0.65)

Plagioclases n 5 4 3 4 4 3 4 4 4 11 4 8 4 6 6 7 3 SiO2 44.81 44.72 47.06 45.99 48.64 48.57 53.03 52.00 51.80 51.32 55.84 55.76 56.22 53.82 53.44 53.66 51.34 A1203 34.96 35.02 33.29 33.56 31.00 30.92 28.58 29.13 29.58 29.95 26.84 26.65 26.64 29.15 28.76 27.88 30.35 FeO 0.52 0.40 0.50 0.47 0.54 0.53 0.65 0.83 0.61 0.54 0.83 0.57 0.49 0.53 0.55 0.56 n.d. MgO 0.20 0.16 0.19 0.23 0.19 0.18 0.27 0.36 0.19 0.22 0.14 0.13 0.18 0.13 0.13 0.26 0.13 CaO 19.69 19.74 17.81 18.06 15.93 15.83 12.85 13.60 13.99 14.20 10.19 10.10 10.23 12.33 11.61 11.68 13.27 Na20 0.61 0.57 1.45 1.23 2.46 2.49 4.20 3.66 3.65 3.36 5.66 5.79 5.64 4.52 4.66 4.68 3.92 K20 0.04 0.04 0.09 0.05 0.07 0.09 0.14 0.12 0.10 0.08 0.08 0.09 0.10 0.05 0 . 2 1 0.19 0.10 Total 100.84 100.67 100.40 99.60 98.87 98.61 99.74 99.71 99.93 99.70 99.57 99.11 99.51 100.54 99.36 98.95 99.11 Ca# 94.69 95.01 87.17 89.06 78.16 77.82 62.83 67.24 67.92 69.99 49.89 49.07 50.05 60.10 57.96 57.99 65.14

(0.73) (1.27) (1.05) (1.08) (1.45) (I.67) (1.99) (2.39) (0.91) (1.96) (1.61) (1.44) (2.47) (3.04) (1.27) (3.46) (0.57) AI# 47.90 48.00 45.46 46.23 42.89 42.86 38.85 39.76 40.23 40.75 36.16 36.03 35.83 38.96 38.81 37.98 41.08

(0.68) (0.22) (0.47) (0.77) (0.50) (0.57) (0.84) (1.01) (0.36) (0.72) (0.91) (0.47) (0.53) (1.07) (0.72) (1.55) (0.08)

T a b l e 3 (continued) Run 3040 3005 3010 3042 3079 3085 3083 3086 3035 3026 3064 3250 3156 3019 3078 3137

427

Glasses n SiO: TiO2 Al203 FeO MnO MgO CaO NaaO K20 Total H20 Mg# Ca#

AI#

2 6 4 3 7 3 5 3 3 9 3 6 8 4 5 4 47.45 53.51 53.05 53.59 55.74 54.12 46.27 43.91 46.61 46.44 49.85 52.36 49.50 55.06 55.52 55.23

0.57 0 . 6 8 0 . 5 8 0 . 6 2 0 . 4 7 0 . 3 8 0 . 6 0 0.60 0 . 6 2 0 . 5 3 0 . 5 8 0 . 5 1 0.47 0.60 0 . 4 7 0.46 18.31 15.12 15.12 15.30 16.70 18.13 19.55 18.30 16.83 18.75 17.89 18.66 19.17 16.61 18.81 17.76 10.66 11.89 11.83 10.54 8 . 3 9 8 . 2 8 6 . 6 9 8 . 6 0 8 . 5 4 6 . 5 0 7 . 2 6 4.52 6 . 6 9 7 . 6 5 5 . 1 3 5.25 0.12 0.20 0 . 1 7 0.14 0 . 1 3 0 . 1 5 0 . 1 9 0 . 0 9 0 . 1 8 0 . 1 7 0 . 1 8 0 . 1 9 0 . 1 3 0.16 0 . 0 7 0.08 6.51 4 . 7 8 5 . 1 4 5 . 0 9 5 . 2 5 4 . 2 8 6 . 7 7 6 . 4 7 6 . 9 0 5.57 6 . 6 8 5 . 5 9 4 . 3 9 4.26 4.09 4.40

13.37 7 .01 7 . 2 0 7 . 4 0 8 .11 8 .52 13.48 13.81 12.20 11.72 11.46 10.69 9 . 4 5 6 . 9 6 7 . 8 1 7.87 1.89 5 . 1 8 5 . 2 5 5 . 0 6 5 .13 5 . 1 6 1 .06 0.89 1.95 2.04 2 . 6 3 3 . 6 2 3 . 4 8 4 . 5 5 5 . 2 0 4.92 0.42 0 . 5 9 0.50 0 . 5 0 0 . 3 3 0 . 2 6 0 . 4 8 0 . 4 0 0 . 4 5 0 . 4 2 0 . 3 9 0 . 3 6 0.36 0 . 4 5 0 . 2 8 0.29

99.30 98.96 98.86 98.24 100.25 99.27 95.09 93.08 94.29 92.14 96.92 96.51 93.65 96.29 97.39 96.26 . . . . 1.3 3.9 1.7 5.8 2.1 2.1 5.4 2.5 1.9 2.2

52.12 41.74 43.64 46.26 52.73 47.95 64.33 57.28 59.02 60.43 62.12 68.79 53.91 49.81 58.70 59.90 79.62 42.77 43.13 44.69 46.66 47.72 87.53 89.51 77.52 76.04 70.68 61,98 60.02 45.86 45.36 46.92 (1.26) (2.68) (0.52) (1.24) (1.26) (0.78) (0.64) (0.27) (1.17) (0.88) (0.69) (0.80) (1.90) (2.63) (2.63) (0.64) 31.26 24.99 25.15 25.18 26.10 28.31 33.24 32.94 29.86 32.24 29.72 29.58 31.34 26.23 28.54 27.48 (0.14) (0.34) (0.42) (0.20) (0.88) (0.34) (0.41) (0.66) (0.62) (0.69) (0.23) (0.48) (0.38) (2.41) (1.02) (0.70)

Plagioclases n 1 7 7 3 5 5 2 3 5 3 2 1 6 7 5 5 SiO2 48 .79 57.82 57.70 58.57 56.01 56.21 44.56 43.57 44.93 44.99 46.37 50.30 48.57 55.31 53.8l 53.94 AlaO3 30.41 26.04 25.50 25.30 26.97 26.31 34.16 34.63 32.71 33.74 31.86 31.21 31.17 26.79 28.95 28.25 FeO 0.57 0 . 5 8 0 . 6 5 0 . 4 8 0 . 3 9 0 . 5 7 0 . 8 4 0 . 8 4 0.92 0 . 6 2 0 . 5 9 n.d. 0.62 0 . 8 3 0.50 0.62 MgO 0.34 0 . 1 5 0.14 0 . 1 3 0 . 1 5 0 .11 0 . 1 6 0 .11 0 . 4 5 0 . 0 5 0 . 3 3 0 . 1 6 0 . 2 2 0 .11 0 . 1 9 0.34 CaO 15.60 8 . 1 9 8 . 3 8 8 . 4 9 9 .97 10.39 19.26 19.93 18.62 18.68 17.49 14.76 16.36 10.51 12.35 11.88 Na20 2.61 6 . 6 4 6 . 5 9 6 . 5 2 5 . 8 3 5 . 6 7 0 . 6 7 0 . 3 2 0 . 9 3 0.79 1.61 3.31 2.21 5.48 4.54 4.69 K20 0.16 0 .21 0 . 1 8 0 . 1 8 0 . 0 9 0 . 0 8 0 . 0 3 0.02 0 . 0 5 0 . 0 3 0 . 0 5 0.06 0 . 0 3 0 . 0 7 0 . 0 4 0.05 Total 98.48 99.65 99.17 99.69 99.41 99.36 99.71 99.45 98.63 98.94 98.31 99.81 99.20 99.13 100.40 99.76 Ca# 76.75 40.53 41.24 41.87 48.56 50.31 94.12 97.18 91.67 92.77 87.75 71.17 80.37 51.45 60.06 58.29

(-) (2.48) (2.34) (1.07) (2.99) (2.68) (0.40) (0.68) (1.92) (2.06) (1.49)(~ (3.82) (1.27) (2.59) (2.71) AI# 42.35 34.67 34.25 33.73 36.21 35.55 47.46 48.36 46.18 46.75 44.74 42.24 43.06 36.34 38.80 38.16

(-) (0.80) (1.36) (0.24) (0.74) (0.56) (0.59) (0.09) (1.19) (1.70) (1.47)(-) (0.88) (0.45) (0.69) (1.56)

~AREA 1 bAREA 2 ~Mg = 100mg/(mg+Fe 2+)

experiments do not cover the composi t ional range of reported in some MORB. Note, however, that because of primitive MORB, having only a m a x i m u m of 6.5 wt % the documented negative correlat ion between crystalliza- MgO, the plagioclase-melt equi l ibr ium Equa t ion 1 t ion pressure and A n conten t of plagioclase, our 5 kbar derived from our study successfully describes the 5-10 data predict less anorthi t ic plagioclase composi t ions than kbar primit ive (up to 9.5 wt % MgO) M O R B magma-p la - those that may have crystallized at lower pressure (5-2 gioclase pairs produced in experimental studies of Bender kbar). F o r example, using Eq. 1, for a melt with C a # = et al. (1978) and Grove et al. (1992). This allows us to 80 and AI# = 30, at 1250~ the difference between the use our 5 kbar anhydrous experimental results (Fig. 2) 5 kbar and 2 kbar equi l ibr ium plagioclase composi t ions to examine the origin of h igh-An (> 85) plagioclases is est imated to be approximately 5 tool % An. The Grove

Fig. 3 Comparison of mea- sured plagioclase An content with values calculated using 90 our Eq.l for A our 5 and 10 kbar anhydrous experiments 09 (dashed lines bracket +_ 3 tool% An uncertainty), and B for 5 ~ 7 0 to 10 kbar anhydrous expcri- O ments of Bender et al. (1978), Stolper (1980), Thy (1991), o Fram and Longhi (1992), ~ 50 Grove et al. (1992). -~ B&E = Baker and Eggler (1987), D&G = Duncan and Green (1987). See text for discussion 3 0 "

30

/ / / / ~

a

/ /j "

t j . ~ " t

U,/ A - 10 kb i

50 7'0 9'0 A n o b s e r v e d

90

r

~70 o

o

30 30

b

~/ / / / r /

~o

,+'/~, Y a n h y d r o u s /~+.;" �9 (5-iOkb)

,'7,# 4 + B&E (Skb) //i + r D&G (5kb)

50 70 90 A n o b s e r v e d

428

Fig. 4A-C A comparison of anhydrous experimental data from 5 to 10 kbar (references listed in the caption to Fig. 3) with experimental data at 0.101 Mpa from Bender et al. (1978), Walker et al. (1979), Grove et al. (1982), Grove and Bryan (1983), Baker and Eggler (1987), Duncan and Green (1987), Sack et al. (1987), Tormey et al. (1987), Grove and Juster (] 989), Juster et al. (1989), Thy (1991), Thy et al. (1991), Fram and Longhi (1992), Gaetani et al. (in press). See text for dis- cussion

gO

~ 7 0 - .,~

50

30

90

~ 7 0

50

38.

a

.. : o �9 �9 ~ 1 4 9 o | . , �9

�9 �9 � 9 �9 . -....'~o~

0 % ~I. ~�9 ~ �9 ~ ~ ~ ."

O~ 0 0 0

" s 7'5 8~ C a # ( m e l t )

C

* ".'36"0 t *o .J

�9 . =e~ ' . o , . ~

�9 ."$~:~. e .~ " ".'~L'~I~ 0

.'. '~,; o

o

~0

o

o

' o . 6 4 ' o . 6 a ' o . i ~ ' o . t 8

x CaAl=O,(melt)

b 9O :o.

:., ;. ~ ' . ' . . w~ 'o . j . . : . l . . ~ ^ �9 -

~ o e ~ ' ~ , ~ 5 " ~ o '~ �9 k~'~:A". ~ �9 o o

~ ~ 0 0 0

50 t o 3 ~ 2'2 ' 2'~ ' 3b

#(mezt)

~ 7 0

a n h y d r o u s o ( 5 - t O k b )

�9 1 a r m

o

' 311 '

et al. (1992) model predicts a difference o f - 2 tool % An over the same pressure difference.

A large database of MORB glasses (Melson catalogue; T. J. Falloon, personal communication) shows that although magmas as calcic as Ca# = 84 (CaO/Na20 = 10) do contribute to the construction of oceanic crust, most MORB glasses are characterized by Ca# values from 64 to 78. According to our experimental data, this range of typical MORB Ca# values corresponds with plagioclase compositions from An 60 to 85, which is indeed the main compositional range of phenocrystic plagioclase in MORB (e.g. Dmitriev et al. 1985; Green et al. 1979; Natland 1989). Also, we conclude that there are two fac- tors which can explain the occasional existence of pla- gioclases of Anss to An94in MORB (Natland 1989; Sinton et al. 1993; Haskell et al. 1993).

The first factor, as suggested by Fisk (1984), is high melt Ca# (>78, and <90), which corresponds with CaO/Na20 values of 10-15. The existence within the oceanic lithosphere of low volume ultra-depleted refrac- tory (CaO/Na20 > 12) melts has been proposed by a num- ber of authors (Duncan and Green 1980; Natland 1989; Grove et al. 1992; Sobolev and Shimizu 1993). Although, as noted above, no MORB glasses have suitable compo- sitions to match the proposed refractory melts, such melts have been found as melt inclusions in some MORB olivines (Danyushevsky et al. 1988), spinels (Donaldson and Brown 1977) and plagioclases (Natland 1989), which

demonstrates emphatically that these melts do contribute to the ocean crust. However, they are rarely preserved due to effective mixing and homogenization of magmas in transit to, or within, sub-ridge magma chambers (Duncan and Green 1980; Natland 1989; Grove et al. 1992; Sobolev and Shimizu 1993).

The second factor is high melt Al# (>30). As can be inferred from our experimental data, and as shown exper- imentally by Haskell et al. (1993), melts with AI# ~30.5 (A1203 contents o f - 18.6%) can crystallize plagioclase as calcic as An90-93 at Ca# values of ~84 at low pressure. The existence of high-A1 MORB magmas (A1203 > 18%) has been demonstrated from studies of melt inclusions in olivine and plagioclase phenocrysts from MORB dredged from near the VEMA Fracture Zone, Atlantic ocean (Sobolev et al. 1989) and the Galapagos Spreading Center, Pacific ocean (Sinton et al. 1993).

Arc basalts

As clearly shown by Beard and Borgia (1989), plagio- clases crystallizing in arc basalts and gabbros are usually much more anorthitic than those crystallizing in MORB. Typical arc high-A1 basalts have abundant plagioclase phenocrysts (Crawford et al. 1987). For oceanic arcs, these plagioclase phenocrysts generally have relatively wide, unzoned and inclusion-free, anorthitic An85=95 cores sur- rounded by sieved and inclusion-rich rims with more sodic

429

compositions, usually from Arts0 to Ans0 (Vanuatu arc: Barsdell 1988; Tongan arc: Ewart 1976; New Britain arc: Heming 1977; Kurile arc: Bailey et al. 1989; dredged and subaerial Mariana arc: Meijer and Reagan 1981; Crawford et al. 1986; Lesser Antilles: Arculus 1978). Plagioclase phenocrysts in continental margin arc high- A1 basalts (e.g. Central America: Rose et al. 1978; Walker and Carr 1986; Japan: Fujimaki 1986; Southern Andes: Lopez-Escobar et al. 1981; western USA: Gerlach and Grove 1982; New Zealand: Cole 1978) and gabbroic plu- tons (Smith et al. 1983) and their cognate inclusions (Beard and Borgia 1989) can also be highly anorthitic, with core compositions from Ang0 to An95 not uncommon. However, in general, continental margin high-A1 basalts have slightly less anorthitic phenocrysts (Ansi, s) than those in intra-oceanic arc lavas.

It has been claimed that H20 present in arc magmas can force crystallization of more calcic plagioclase than would crystallize from the same composition under anhy-

Fig. 5 Correlation between plagioclase An content and Ca# value A, AI# value C, and XC,A~,O 4 D in equilibrium melts for anhydrous experiments at 5 kbar andHzO-saturated and undersaturated exper- iments at 2 to 5 kbar from Baker and Eggler (1987) ( = B&E), Housh and Luhr (1991) ( = H&L), Sisson and Grove (1993) ( = S&G), and our data. B Application of the model of Housh and Luhr (1991) to anhydrous and H20-bearing experiments from sources listed in captions to Figs. 3, 4, and 5A,C,D. See text for discussion

drous conditions (e.g. Arculus and Wills 1980). The H20- saturated experiments of Sisson and Grove (1993) do indeed show that H:O exerts the most pronounced effect on An content in plagioclase in the melt composition range from CaO/Na:O from 2 to 5 (Ca# from 50 to 70, Fig. 5). In this compositional range, H20 at a level of -6 wt % can increase the An content in plagioclase by 20 tool%. However, as we have shown here, the same amount of H:O for significantly undersaturated condi- tions can increase An content by only -10 tool%. This implies that crystallization pressure plays a major role in controlling the An content of equilibrium plagioclase compositions, due to the well-known pressure dependence of HzO solubility in magmas.

Natural aphyric high-A1 basalts with CaO/Na:O up to 5.5 (Ca# up to 75) are recorded from several localities (Sisson and Grove 1993). Sisson and Layne (1993) have demonstrated that the H:O contents of at least some high- A1 arc melts are as high as 6 wt%. Data from our exper- imental study and that of Sisson and Grove (1993) show that under H20-saturated conditions with 6% H:O, such melts can crystallize plagioclases as calcic as Ang0 9> In other magmas with significantly lower H20 contents (1 3 wt%, e.g. Baker 1987; Sekine et al. 1979), or at significantly H20-undersaturated conditions, the calcic nature of plagioclase may be related to notably higher melt CaO/Na20 values (up to at least 8, Ca# 80-85). Furthermore, since for most arc magmas there is

a .~

~d~@ �9 o ~ B 5

~o

6s i " . - . �9

. / " . , i

go so vb 9o Ca#(melt)

C Q o

85- Do ~ O

%

~, , "o

#o. 45 . . . . . . . . . . . . . .

22 24 26 28 3,0 32 34 36 Al#(m lt)

O

0

i

O

Hs0-saturated [] S&G ( 2 k b ) A H&L ( 2 k b )

H&L ( 4 k b )

H z 0 - undersaturated B&E (2kb) B&E (5kb) this study

anhydrous (5kb) this study

19

17

11

9

�9 anhydrous (tatm-20kb) j +Hs0-saturated (2-5kb U Hz0-undersaturated 7

5 6 8 9 10 1 0 0 0 0 / T ( K )

85

65

45 0 .02

' !

d o, B ~0

i ~ ' ~ O O 13 15131:3 �9

DO) A A ~

~ c ~ A d

o.b8 o.io 0.i4 XeaAl~04(melt)

430

abundant evidence of extended clinopyroxene crystal- lization prior to plagioclase appearance (e.g. Merelava volcano in Vanuatu crystallized An93 after extensive clinopyroxene fractionation; Barsdell 1988), then the CaO/Na20 values of parental arc magmas are likely to have been much greater than 8.

The thick crust beneath arc axial chains forces exten- sive pooling, mixing and fractionation of parental arc magmas. Little evidence is preserved in the evolved mafic phenocryst assemblages in erupted high-A1 arc lavas for the existence beneath such arc volcanoes of highly depleted magmas. However, plagioclase phenocrysts, which are more buoyant than the mafic phenocrysts, may continually concentrate by flotation in the upper parts of arc magma chambers and accumulate in fractionated residual liquids ranging from basaltic andesite to rhyolitic compositions. As argued by Crawford et al. (1987), this mixture is likely to be erupted as typical arc high-A1 basalts, in which the entrained plagioclase phenocrysts are unlikely to be in equilibrium with the host liquids.

The existence of refractory parental magmas for "nor- mal" arc basalts needs to be taken into account in any petrogenetic model for these magmas.

Boninitic lavas from intra-oceanic forearcs

Plagioclase phenocrysts in some intra-oceanic forearc lavas compositionally transitional from arc tholeiites to high-Ca boninites are often as calcic as An95 and have groundmass plagioclase laths commonly in the range An~54j (Crawford et al. 1986). Plagioclase phenocrysts from high-Ca boninite lavas from the northern Tongan forearc are mainly in the range An95_~00 (Falloon and Crawford 1991), and the most anorthitic plagioclases known from analogous ophiolitic boninites and associ- ated gabbros from the Troodos ophiolite, Cyprus, are up to An98 (Duncan and Green 1987; Thy et al. 1989; Malpas et al. 1989).

These melts are characterized by moderate H20 con- tents (2-3 wt %, Sobolev and Danyushevsky, in press; Sobolev et al. 1993). According to our experimental results, the only plausible factor which can force crystal- lization of such calcic plagioclases is the abnormally high CaO/Na20 values of these melts, indeed, melt inclusions in Fo92 94 phenocrysts in high-Ca boninites from the Tongan forearc with An9s_~00 phenocrysts, have CaO/NazO from 18~9 (Ca# - 90-93) (Falloon and Green 1986).

Conclusions

This experimental study of plagioclase-melt equilibria has demonstrated that melt Ca# and AI#, H20 content and

crystallization pressure all control the composition of liq- uidus plagioclase. Increasing melt Ca#, AI# and H20 con- tent all increase An content of plagioclase, whereas increasing pressure leads to crystallization of more albitic plagioclase compositions. However, the importance of the role played by each of these factors during crystallization of natural magmas varies. Melt Ca# has the strongest control on plagioclase An content, but melt AI# also exerts a significant control. H20 content can notably increase the An content of plagioclase, up to 10 mol% for H20-undersaturated melts, and 20 tool% for H20-satu- rated melts.

The presence of the exceptionally calcic plagioclases (up to An~00) in primitive subduction-related boninitic and related rocks cannot be attributed to the presence of the demonstrated amounts of H20 (up to 3 wt%). Rather, they must be due to the involvement of extremely refrac- tory (CaO/Na20> 18) magmas in the petrogenesis of these rocks.

Despite the refractory nature of some primitive MORB glasses, none are in equilibrium with the most cal- cic plagioclase (An94) found in MORB. These plagioclases were likely produced from more refractory melts with CaO/Na20 = 12 15, or from melts with exceptionally high A1203(> 18%).

We have shown that refractory melts with CaO/Na20 values much higher than 10 are required to produce the most calcic plagioclases in both MORB and boninitic sub- duction-related rocks. Magmas of appropriate composi- tions to crystallize these most calcic plagioclases are only found as melt inclusions in near liquidus phenocrysts from these rocks, but are not known among wholerock or glass compositions. Available models for magma generation in mid-ocean ridge and intra-oceanic forearc settings invoke a spectrum of magma compositions which includes small volume, highly refractory melts. The fact that the latter melts are not erupted as discrete magma batches (but are occasionally preserved in melt inclusions in near-liq uidus phenocrysts) indicates that they are effectively mixed and homogenized with volumetrically dominant, less refrac- tory magmas.

The high H20 contents (-6 wt%) in some high-A1 basaltic arc magmas may be responsible for the existence of plagioclases up to An95 in arc lavas. Although occa- sional examples of very hydrous (H20 to 6 wt%) high-Al basaltic arc magnnas are known, an alternative possibil- ity, the involvement of melts with abnormally high CaO/Na20 values (> 8) in petrogenesis of these rocks, may account for the presence of highly anorthitic plagioclases.

In conclusion, highly anorthitic plagioclase phe- nocrysts are the best evidence that very depleted, high CaO/Na20 magmas are involved in the petrogenesis of typical arc basalts, MORB and boninites, and that exten- sive mixing of near primary melts with a range of CaO/Na20 values is an important process in the genera- tion of these magma suites.

431

Acknowledgements This work formed part of Y. Panjasawatwong's PhD project, initiated by Prof D.H. Green and supervised by AJC and DHG at the Department of Geology, University of Tasmania. This project was funded by the Australian International Development Aid Bureau (AIDAB), and Australian Research Council grants to AJC and DHG. The latter also provided financial support for LVD. The technical assistance of Wieslaw Jablonski with the electron microprobe and of Graham Rowbottom with the FTIR analyses are acknowledged. Thorough reviews by Peter Meyer and Roger Nielsen are much appreciated, as is constructive input and editorial handling by Tim Grove.

References

Arculus RJ (1978) Mineralogy and petrology of Grenada, Lesser Antilles Island Arc. Contrib Mineral Petrol 65: 413424

Arculus RJ, Wills KJA (1980) The petrology of plutonic blocks and inclusions from the Lesser Antilles island arc. J Petrol 21 : 43 799

Autio LK, Rhodes JM (1984) Costa Rica rift zone basalts: geo- chemical and experimental data from a possible example of mul- tistage melting. Initial Rep Deep Sea Drill Proj 60:729-745

Bailey JC, Frolova TI, Burikova IA (1989) Mineralogy, geochem- istry and petrogenesis of Kurile island-arc basalts. Contrib Mineral Petrol 102: 265-280

Baker DR (1987) Depths and H20 content of magma chambers in the Aleutian and Mariana island arcs. Geology 15:496-499

Baker DR, Eggler DH (1987) Compositions of anhydrous and hydrous melts coexisting with plagioclase, augite, and olivine or low-Ca pyroxene from 1 atm to 8 kbar: application to the Aleutian volcanic center of Atka. Am Mineral 72:12-28

Barsdell M (1988) Petrology and petrogenesis of clinopyroxene-rich tholeiitic lavas, Merelava Volcano, Vanuatu. J Petrol 29 : 927@64

Barsdell M, Berry RF (1990) Origin and evolution of primitive ankaramitic lavas from western Epi, Vanuatu. J Petrol 31: 747-777

Beard JS, Borgia A (1989) Temporal variation of mineralogy and petrology in cognate gabbroic enclaves at Arenal volcano, Costa Rica. Contrib Mineral Petrol 103:110-122

Bender JF, Hodges FN, Bence AE (1978) Petrogenesis of basalts from the project FAMOUS area: experimental study from 0 to 15 kbars. Earth Planet Sci Lett 41:277-302

Boyd FR, England JL (1960) Apparatus for phase-equilibrium mea- surements at pressures up to 50 kbar and temperatures up to 1750~ J Geophys Res 65:741-748

Brophy JG (1986) The Cold Bay volcanic centre, Aleutian volcanic arc. Contrib Mineral Petrol 93:368 380

Cole J-W (1978) Andesites of the Tongariro volcanic centre, North Island, New Zealand. J Volcanol Geotherm Res 3:121-153

Crawford AJ, Beccaluva L, Serri G, Dostal J (1986) Petrology, geo- chemistry and tectonic implications of volcanics dredged from the intersection of the Yap and Mariana trenches. Earth Planet Sci Lett 80:265 280

Crawford A J, Falloon TJ, Eggins S (1987) The origin of island arc high-alumina basalts. Contrib Mineral Petrol 97: 417~430

Danyushevsky LV, Sobolev AV, Dmitriev LV (1988) Orthopyroxene-bearing low-Ti tholeiites - the new type of ceanic ridge tholeiite. Doklady Akad Navk SSSR 292:102-105

Danyushevsky LV, Falloon TJ, Sobolev AV, Crawford AJ, Carroll M, Price RC (1993) The H20 content of basalt glasses from Southwest Pacific back-arc basins. Earth Planet Sci Lett 117:347 362

Dmitriev LV, Sobolev AV, Suschevskaya NM, Zapunny SA (1985) Abyssal glasses, petrologic mapping of the oceanic floor and "geochemical leg" 82. Initial Rep Deep Sea Drill Proj 82:509-518

Donaldson CH, Brown RW (1977) Refractory megacrysts and mag- nesium-rich melt inclusions within spinel in oceanic tholeiites: indicators of magma mixing and parental magma composition. Earth Planet Sci Lett 37:81 89

Duncan RA, Green DH (1980) Role of multistage melting in the formation of oceanic crust. Geology 8:2~26

Duncan RA, Green DH (1987) The genesis of rel?actory melts in the formation of oceanic crust. Contrib Mineral Petrol 96: 326-342

Ewart A (1976) A petrological study of the younger Tongan andesites and dacites, and the olivine tholeiites of Niua Fo'ou Island, S.W. Pacific. Contrib Mineral Petrol 58:1 21

Falloon TJ, Crawford AJ (1991) The petrogenesis of high-calcium boninite lavas dredged from the north Tonga ridge. Earth Planet Sci Lett 102:375 394

Falloon TJ, Green DH (1986) Glass inclusions in magnesian olivine phenocrysts from Tonga: evidence for highly refractory parental magmas in the Tongan arc. Earth Planet Sci Lett 81:95 103

Fisk MR (1984) Depth and temperatures of mid-ocean-ridge magma chambers and the composition of their source magmas. In: Gass IG, Eippard S J, Shelton AW (eds) Ophiolitcs and oceanic lithos- phere. Blackwell Scientific, Oxford, pp 17-23

Fram MS, Longhi J (1992) Phase equilibria of dikes associated with Proterozoic anorthosite complexes. Am Mineral 7 : 605-616

Fujimaki H (1986) Fractional crystallization of the basaltic suite of Usu volcano, southwest Hokkaido, Japan, and its relationships with the associated felsic suite. Lithos 19:129-140

Gaetani GA, Grove TL, Bryan WB (1994) Experimental phase rela- tions of basaltic andesite from hole 839B under hydrous and anhydrous conditions. Proc Ocean Drilling Proj Res Leg 135, in press.

Gerlach DC, Grove TL (1982) Petrology of Medicine Lake Highland volcanics characterization of end members of magma mixing. Contrib Mineral Petrol 80:147 159

Gladney ES, Jones EA, Niekell EJ, Roelandts I (1990) 1988 com- pilation of elemental concentration data for USGS basalt BCR- 1. Geostandards Newsletter 14:209-359

Green DH, Ringwood AE (1967) The genesis of basaltic magmas. Contrib Mineral Petrol 15:103-190

Green DH, Hibberson WO, Jaques AL (1979) Petrogenesis of mid- ocean ridge basalts. In: McElhinny MW (ed) The Earth: its ori- gin, structure and evolution. Academic Press, London, pp 265~90

Grove TL, Bryan WB (1983) Fractionation of pyroxene-phyric MOR at low pressure: an experimental study. Contrib Mineral Petrol 84:293-309

Grove TE, Gerlach DC, Sando TW (1982) Origin of calc-alkaline series lavas at Medicine Lake volcano by fractionation, assimi- lation and mixing. Contiib Mineral Petrol 80:160-182

Grove TL, Juster TC (1989) Experimental investigation of low-Ca pyroxene stability and olivine-pyroxene-liquid equilibria at l- atin in natural basaltic liquids. Contrib Mineral Petrol 103:28%305

Grove TL, Kinzler RJ, Bryan WB (1992) Fractionation of mid- ocean ridge basalt (MORB). Geophys Monogr Am Gcophys Union 71:281-310

Haskell K, Forsythe L, Nielsen RL, Fisk MR (1993) Experimental constraints on the parental magma for the high An feldspar bear- ing Lamont Seamount lavas. EOS Trans Am Geophys Union 74 Supplement No 43:357

Heming R F (1977) Mineralogy and proposed P-T paths of basaltic lavas from Rabaul Caldera, Papua New Guinea. Contfib Mineral Petrol 61:15 33

Housh TB, Luhr JF (1991) Plagioclase-melt equilibria in hydrous systems. Am Mineral 76:477-492

Johannes W (1978) Melting of plagioclase in the system Ab-An- H20 and Qz-Ab-An-H20 at PH20=5 kbars, an equilibrium prob- lem. Contrib Mineral Petrol 66:295-303

432

Juster TC, Grove TL, Perfit MR (1989) Experimental constraints on the generation of FeTi basalts, andesites and rhyodacites at the Galapagos Spreading Center, 85~ and 95~ J Geophys Res 94B:9251~274

Kinzler RJ, Grove TL (1992) Primary magmas of mid-ocean ridge basalts 1. Experiments and methods. J Geophys Res 97: 6885--6906

Koyagnchi T (1986) Evidence for two-stage mixing in magmatic inclusions and rhyolitic lava domes on Niijima Island, Japan. J Volcanol Geotherm Res 29:71-98

Kudo AM, Weill DF (1970) An igneous plagioclase thermometer. Contrib Mineral Petrol 25 : 5~65

Lopez-Escobar L, Vergara M, Frey FA (1981) Petrology and geo- chemistry of lavas from Antuco Volcano, a basaltic volcano of the southern Andes (37~ ' S). J Volcanol Geotherm Res 11:329 352

Malpas J, Brace T, Dunsworth SM (1989) Structural and petro- logic relationships of the CY-4 drill hole of the Cyprus Crustal Study Project. Geol Surv Canada Pap 88-9, pp 39-68

Marsh BD, Fournelle J, Myers JD, Chou I-M (1990) On plagio- clase thermometry in island arc rocks: experiments and theory. In: Spencer R J, Chou I-M (eds) Fluid-mineral interaction (Geochem Soc Spee Pub 2). Geochemical Society, Penn State University, pp 65-83

Mcijer A, Reagan M (1981) Petrology and geochemistry of the island of Sarigan in the Mariana Arc; calc-alkaline volcanism in an oceanic setting. Contrib Mineral Petrol 77:337-354

Natland JH (1989) Partial melting of a lithologically heterogeneous mantle inferences from crystallization histories of magnesian abyssal tholeiites from the Siqueiros Fracture Zone. In: Saunders AD, Norry MJ (eds) Magmatism in the ocean basins. Blackwell Scientific, Oxford, pp 41-70

Price RC, Kennedy AK, Riggs-Sneeringer M, Frey FA (1986) Geochemistry of basalts from the Indian Ocean triple junction: implications for the generation and evolution of Indian Ocean ridge basalts. Earth Planet Sci Lett 78:379-396

Rose WI Jr, Anderson AT Jr, Woodruff LG, Bonis SB (1978) The October 1974 basaltic tephra fiom Fuego Volcano - description and history of the magma body. J Volcanol Geotherm Res 4:3-53

Sack RO, Walker D, Carmichael ISE (1987) Experimental petrol- ogy of alkalic lavas: constraints on cotecties of multiple satura- tion in natural basaltic liquids. Contrib Mineral Petrol 96 : 1 23

Sekine T, Katsura T, Aramaki S (1979) H20 saturated phase rela- tions of some andesites with application to the estimation of the initial temperature and H20 pressure at the time of eruption. Geochim Cosmochim Acta 43:1367-1376

Sinton CW, Christie DM, Coombs VL, Nielsen RL, Fisk MR (1993) Near primary melt inclusions in anorthite phenocrysts from the Galapagos Platform. Earth Planet Sci Lett 119:527-537

Sisson TW, Grove TL (1993) Experimental investigations of the role of HaO in calc-alkaline differentiation and subduction zone magmatism. Contrib Mineral Petrol 113 : 143-166

Sisson TW, Layne GD (1993) 1-t20 in basalt and basaltic andesite glass inclusions from four subduction-related volcanoes. Earth Planet Sci Lett 117:619 635

Smith TE, Huang CH, Walawender M J, Cheung P, Wheeler C (1983) The gabbroic rocks of the peninsular ranges batholith, southern California Cumulate rocks associated with calc- alkalic basalts and andesites. J Volcanol Geotherm Res 18:249 278

Sobolev AV, Danyushevsky LV (1994) Petrology and geochemistry of the high-Ca boninites from the north termination of the Tonga Trench: constraint on high-Ca boninite primary magma gener- ation conditions. J Petrology (in press)

Sobolev AV, Shimizu N (1993) Ultra-depleted primary melt included in an olivine from the Mid-Atlantic Ridge. Nature 363:151 154

Sobolev AV, Danyushevsky LV, Dmitriev LV, Suschevskaya NM (1989) High-alumina magnesian tholeiite as the primary basalt magma at Midocean ridge. Geochem Int 26:128 133

Sobolev AV, Portnyagin MV, Dmitriev LV, Tsamerian OP, Danyushevsky LV, Kononkova NN, Shimizu N, Robinson PT (1993) Petrology of ultramafic lavas and associated rocks of the Troodos Massif, Cyprus. Petrology 1:331 361

Stakes DS, Shervais JW, Hopson CA (1984) The volcanic-tectonic cycle of the FAMOUS and AMAR Valleys, mid-Atlantic Ridge (36047 ' N): evidence from basalt glass and phenocryst composi- tional variations for a steady state magma chamber beneath the valley midsections, AMAR 3. J Geophys Res 89:6995-7028

Stolper E (1980) A phase diagram for mid-ocean ridge basalts: pre- liminary results and implications for petrogenesis. Contrib Mineral Petrol 74: 13-27

Stolz A J, Varne R, Wheller GE, Foden JD, Abbott MJ (1988) The geochemistry and petrogenesis of K-rich alkaline volcanics from the Batu Tara volcano, eastern Sunda arc. Contrib Mineral Petrol 98 : 374-389

Thy P (1991) High and low pressure phase equilibria of a mildly alkalic lava from the 1965 Surtsey eruption: Experimental results. Lithos 26:223 243

Thy P, Schiffman P, Moores EM (1989) Igneous mineral stratigra- phy and chemistry of the Cyprus Crustal Study Project drill core in the plutonic sequences of the Troodos ophiolite. Geol Surv Canada Pap 88-9, pp 147-186

Thy P, Lofgren GE, Imsland P (1991) Melting relations and the evolution of the Jan Mayen magma systems. I Petrol 32 : 303-332

Tormey DR, Grove TL, Bryan WB (1987) Experimental petrology of normal MORB near the Kane Fracture Zone - 22~176 mid-Atlantic ridge. Contrib Mineral Petrol 96:121-139

Walker DA, Shibata T, DeLong SE (1979) Abyssal tholeiites from the Oceanographer Fracture Zone. II Phase equilibria and mix- ing. Contrib Mineral Petrol 70:111-125

Walker JA, Carr MJ (1986) Compositional variations caused by phenocryst sorting at Cerro Negro volcano, Nicaragua. Geol Soc Am Bull 97:1156-1162

Wilkinson J FG (1982) The genesis of mid-ocean ridge basalts. Earth Sci Rev 18:1 57