Experimental investigations of the role of H2O in calc-alkaline ...

3 downloads 0 Views 2MB Size Report
Presnall DC (1966) The join forsterite-diopside-iron oxide and its bearing on the crystallization of basaltic and ultramafic magmas. Am J Sci 264:753–809Google ...
Contrib Mineral Petrol (1993) 113:143-166

Contributions to

Mineralogy and Petrology 9 Springer-Verlag1993

Experimental investigations of the role of H20 in calc-alkaline differentiation and subduction zone magmatism T.W. Sisson* and T.L. Grove

Earth, Atmospheric,and Planetary Sciences, Massachusetts Institute of Technology,Cambridge, MA 02138, USA Received August 19, 1991/Accepted July 2, 1992 Abstract. Phase relations of natural aphyric high-alumina

basalts and their intrusive equivalents were determined through rock-melting experiments at 2 kb, HzO-saturated withfO 2 buffered at NNO. Experimental liquids are lowMgO high-alumina basalt or basaltic andesite, and most are saturated with olivine, calcic plagioclase, and either high-calcium pyroxene or hornblende ( + magnetite). Crspinel or magnetite appear near the liquidus of wet highalumina basalts because H20 lowers the appearance temperature of crystalline silicates but has a lesser effect on spinel. As a consequence, experimental liquids follow calcalkaline differentiation trends. Hornblende stability is sensitive to the NazO content of the bulk composition as well as to H20 content, with the result that hornblende can form as a near liquidus mineral in wet sodic basalts, but does not appear until liquids reach andesitic compositions in moderate Na20 basalts. Therefore, the absence of hornblende in basalts with low-to-moderate Na20 contents is not evidence that those basalts are nearly dry. Very calcic plagioclase ( > An9o) forms from basaltic melts with high H2 O contents but cannot form from dry melts with normal arc Na 20 and CaO abundances. The presence of anorthite-rich plagioclase in high-alumina basalts indicates high magmatic H20 contents. In sum, moderate pressure HgO-saturated phase relations show that magmatic H20 leads to the early crystallization of spinel, produces calcic plagioclase, and reduces the total proportion of plagioclase in the crystallizing assemblage, thereby promoting the development of the calc-alkaline differentiation trend.

Introduction

Mafic through silicic arc volcanos often erupt explosively with the violent release of magmatic volatiles. This erupT.W. Sisson Center for Lithospheric Studies, University of Texas at Dallas, Richardson, TX 75083-0688, USA Correspondence to: * Present address:

tive style combined with the occasional presence of hydrous phenocrysts, as well as other more subtle features, have suggested to many investigators that H20 is important for subduction zone magmatism (see Gill 1981, p. 116-121 for a summary). Kennedy (1955) proposed that the calc-alkaline magma series, now known to characterize subduction environments, resulted by crystallization differentiation of hydrous basalts. H20-bearing basaltic magmas were thought to crystallize an iron oxide phase early and differentiate with silica and alkali enrichment and iron depletion (calc-alkaline series) whereas dry basaltic magmas fail to crystallize an iron oxide phase early and instead differentiate with iron enrichment and only modest increases in silica content (tholeiitic series). Osborn (1959) concurred with Kennedy's model, arguing that dissociation of dissolved magmatic HzO would produce a h i g h / 0 2 leading to early and abundant oxide precipitation and a caic-alkaline differentiation trend. Experimentation in both synthetic and natural systems as well as studies of the crystallization of historic lava lakes have confirmed that dry sub-alkaline basalts follow tholeiitic differentiation paths while crystallizing at geologically relevant oxygen fugacities and at pressures appropriate to the crust (Osborn 1959; Grove and Baker 1984; Helz 1987; Juster et al. 1989). Tholeiitic liquid lines of descent, temperatures, solid phase compositions, and solid-liquid exchange reactions are now well established for anhydrous basalts at low pressures. Calc-alkaline differentiation trends have proven considerably more difficult to reproduce experimentally. Estimates of thefO2 of subduction magmas fall in the range of below but close to the quartz-fayalite-magnetite (QFM) buffer to one log unit more oxidizing than the Ni-NiO (NNO) buffer (Gill 1981, p. 124). Anhydrous basalt melts have not crystallized along calc-alkaline differentiation paths over this range o f f O z conditions at crustal pressures (Grove et al. 1982; Juster at al. 1989). Early studies of the hydrous melting of basalts focussed on the influences of HeO-pressure andfO2 on phase appearance sequences and liquidus temperatures (Yoder and Tilley 1962; Holloway and Burnham 1972; Helz 1973, 1976; Allen et al. 1975). Some of this research was per-

144

formed before use of the electron microprobe had become routine and experimental phase compositions were incompletely characterized (Yoder and Tilley 1962; Allen et al. 1975). Moreover, those studies in which phases were most fully characterized were conducted over a range of temperatures where the basalts were highly crystalline and hydrous melts more primitive than silicic andesite were not produced (Helz 1973, 1976; Baker and Eggler 1987). Quench crystallization and loss of Fe to noble metal containers have plagued studies of the hydrous melting of basalts and andesites, hindering the extraction of quantitative information on melt compositions and mineral/melt exchange reactions (Helz 1973, 1976; Eggler and Burnham 1973; Baker and Eggler 1987). As a result of these difficulties, the crystallization behavior of HzO-bearing mafic arc magmas is very poorly understood and the H 2 0 contents of common mafic arc magmas cannot be estimated with precision. Accordingly, we undertook an experimental study to provide information on the hydrous, moderate-pressure crystallization paths followed by arc high-alumina basalts and their intrusive equivalents atfO 2s appropriate for arc magmas. A goal of this study has been to explore the possibility that magmatic H 2 0 is responsible for the calcalkaline differentiation trend. We discuss the effects of H 2 0 on both the crystallization sequences and phase compositions of hydrous arc basalts and show that magmatic H 2 0 can have a critical role in producing calcalkaline daughter liquids from high-alumina basalt parents. The crystallization behavior of hydrous arc basalts at lower crust or upper mantle depths are not addressed. A companion paper is concerned with the question of the origin of arc high-alumina basalts (Sisson and Grove 1992, this issue).

Table

I.

Experimental methods

Starting materials The experiments were performed using powdered natural igneous rocks, mixtures of rocks and minerals and synthetic oxide mixes plus added H 2 0 as starting materials. Natural aphyric high-alumina basalts with high (9.86 wt.%, sample 79-35g) to intermediate MgO contents (8.61, 7.47 wt.%, samples 82-62, 82-66) were collected from the Giant Crater-Chimney Crater lava flow system at Medicine Lake volcano, California. The high-alumina basalt 79-35g has a composition very similar to the Warner basalt used by Yoder and Tilley (1962) in their study of basalt melting. The phase relations of 79-35g at low and high pressures, anhydrous, are reported by Grove et al. (1982) and Bartels et al. (1991). Natural aluminous hornblende gabbro (sample 87S35a) and hornblende diorite (sample 85S52b), with compositions equivalent to low-MgO high-alumina basalt and basaltic andesite, were collected from quenched sills and dikes of the mafic intrusive complex at Onion Valley in the Sierra Nevada batholith, California. The intrusive rock samples have higher alkali and lower magnesia contents relative to the volcanic samples. Specific investigations of the stability of plagioclase and amphibole used combinations of rock, HzO, and either powdered natural plagioclase (An8o or albite) or NaOH. Experiments in the simple system Forsterite-Diopside-Anorthite-H20 (Fo-Di-An-H20) used a finely-powdered mix of reagent grade oxides and glass prepared from oxides. Rock, plagioclase, and mix compositions are presented in Table 1.

Experimental procedures Synthesis experiments were conducted H20-saturated at 2 kb in a TZM (titanium-zirconium-molybdenum) cold-seal pressure vessel using Au inner and outer sample capsules. Experiments in the system CaO-MgO-A1203-SiOz-H 2 O (CMAS-H20) were contained in welded Pt capsules. Fugacity of oxygen was buffered by a solid Ni-NiO assemblage. Because most experiments were hotter than the

Compositions of starting materials

Sample

SiO2

AI20 3 FeO

Medicine Lake Highland Basalts a 79-35g 48.2 18.2 82-62 49.3 17.7 82-66 51.2 17.3

8.39 9.02 8.66

Sierra Nevada Batholith Mafic Dikes c 87S35a 50.6 19.1 5.33 85S52b 53.3 17.5 5.95 Feldspars a Stillwater Bytownite e Amelia Albite f

48.1

32.8

0.48

68.6

19.4

0.0

Forsterite-Diopside-Anorthite Mix g 46.4 21.9

FezO 3 MgO

CaO

Na20

K20

TiO2

P205

MnO

H 2 0 + Ha O - Total

b --

9.86 8.61 7.47

12.0 11.3 10.2

2.29 2.58 3.11

0.09 0.30 0.70

0.66 0.95 1.01

0.05 0.11 0.16

0.16 0.17 0.16

--

--

99.9 100.0 100.0

3.74 2.56

4.32 4.78

8.85 7.27

4.23 3.68

1.00 2.02

1.27 1.12

0.37 0.33

0.17 0.18

1.13 1.07

0.15 0.16

100.3 99.9

-

0.04

2.17

0.08

.

11.8

0.16

0.0 -

12.8

16.0 0.02 18.9

-

.

. -

-

.

.

99.7

-

-

100.0

-

-

100.0

a Analyses by XRF, U.Mass., Amherst, all Fe as FeO. M.B. Baker-analyst b Dash indicates not analyzed or not present in mix c N a 2 0 by flame photometry, H 2 0 + / - and FeO by gravimetry and titration, rest by XRF. Staff of US Geological Survey, Branch of Analytical Chemistry Menlo Park, CA and Lakewood, CO - analysts d Averages of > 10 replicate analyses by electron microprobe - MIT, all Fe as FeO e Used in 85S52b + plagioclase mix in weight proportions 0.85 rock:0.15 plagioclase f Used in 82-66 + albite mix in weight proportions 0.84 rock:0.16 plagioclase g Calculated composition of oxide mix

145 Au-Ni minimum (945 955 ~ at I atm, Hansen 1958), the buffer was isolated from contact with Au by containment in one to three unsealed Pt capsules. The inner capsule holding the sample was twisted and crimped closed, but was not welded, to assure that the oxygen buffer and the sample were in contact with the same aqueous fluid. Experimental temperatures were measured with Pt-Ptg0Rhl0 thermocouples positioned in a well in the pressure vessel. Thermocouples were calibrated by comparison to a reference thermocouple calibrated to the melting points of NaC1, Au, and Pd on the IPTS (1968) scale. The difference between well temperature and sample temperature was calibrated numerous times, and reported experimental temperatures are precise to _+ 7 oC. Pressure was monitored with a Heise gauge and controlled to within 30 bar of the reported pressure. A pressure medium of mixed Ar and methane, in the proportion 2000 psi: 35 psi, reduced diffusive transport of hydrogen out of the Au outer capsule. The pressure vessel was positioned vertically in a Deltech DT31VT resistance furnace, held at pressure and temperature for the experiment, and then quenched. Experiments were begun by heating 10-25 ~ above the desired temperature for 1 hour to reduce the proportion of refractory crystals (plagioclase or plagioclase and olivine) to no more than a few percent. Melt plus any remaining crystals were then cooled over 1 2 hours to the intended level and held for the duration of the experiment (8 to 72 h). Experiments were quenched by extracting the vessel from the furnace, immediately inverting it, and rapping on the vessel with a pipe wrench. The sample capsule fell from the hot end of the vessel to the water-cooled pressure seal and quenched rapidly with virtually no growth of quench crystals. Early experiments (not reported) were quenched by blowing compressed air over the hot vessel, a procedure that resulted in common and extensive quench crystallization. Following quenching, the capsules were weighed to measure volatile gain or loss and punctured to verify the presence of water. The buffer capsules were examined for the presence of both Ni and NiO and to see if the sample had leaked into contact with the Pt buffer capsule. Experiments that passed these tests and did not contain abundant quench crystals (a sign that the capsule had not fallen, overall < 30% of all experiments passed) were considered successful and were mounted for electron probe analysis. Experimental duration. Duration of experiments varied with experimental temperature (Table 2) and was limited by diffusive loss of hydrogen through the outer Au capsule wall (0.010" thick) to the pressure medium and from the pressure medium out of the vessel to the atmosphere. Experiments at 1050 ~ could be continued for up to 36 h. Experiments at 965~ could be continued for up to 72 h. Loss of hydrogen leads to slightly decreased capsule weight, oxidation of the buffer and, if extensive, desiccation of the sample. Loss from the Au capsule is by grain boundary diffusion. Some lengths of Au tubing recrystallized and exchanged H 2 rapidly whereas other "good" lengths of tubing took longer to recrystallize and could be used for longer experiments. Analytical techniques. Glass and minerals were analyzed with the MIT four spectrometer JEOL-733 electron microprobe. This instrument uses automatic data reduction with Bence and Albee (1968) matrix corrections as modified by Albee and Ray (1970). Analytical conditions were varied to avoid Na migration during glass analysis. We settled on a 15 kV accelerating voltage, 10 nA beam current, a 20-25 micron beam diameter, and measured Na first for 5 seconds. Other elements were subsequently measured for up to 40 s depending on abundance. The good agreement between Na20 measured in the whole rock by atomic absorption .(Table 1) and by electron probe in supra-liquidus glasses of sample 87S35a (Table 3) shows that these analytical conditions are adequate for hydrous basalt and basaltic andesite melts. See Appendix 1 for discussion of analytical methods used for high SiO 2 glasses. Analyses of crystal margins were as mentioned, but the beam was focused to a diameter of 2 microns. Analytical precision is estimatecl by replicate measurements of andesite glass from a l-atm anhydrous experiment (sample 38b-129) of Grove and Juster (1989). One standard deviation of replicate glass

analyses expressed as relative percent of oxides are SiO2: 0.4%, A1203: 0.9%, CaO: 1.5%, MgO: 1.5%, FeO: 1.4%, MnO: 8.1%, P205: 5.6%, Na20: 1.9%, K20: 1.1%, based on 121 individual measurements over 13 analytical sessions. The mean sum of analyses of the anhydrous glass is 99.5 wt.%. Precision for the NiO analyses reported in Table 3 is poor, but these data show that Ni abundance in the silicate sample is low and therefore that transport of Ni from the buffer to the silicate sample was negligible. Melt Fe 3+/(Fe z+ + Fe 3+) was measured by M6ssbauer spectroscopy in excess glass from NNO-buffered supra-liquidus experiments on sample 87S35a using techniques presented in Burns and Dyar (1983). Because of the limited material available, glass was combined from runs 87S35a- no. 1 and 87S35a- no.5 performed at 1050 ~ and 1012~ respectively. Relative areas of absorption peaks indicate that Fe 3+ is 13.7% ( _+ 1%, 2-a) of total Fe in the combined sample. Ferric iron is calculated as 16% of total Fe in 87S35a - no. 1 and no. 5 using the expression of Kress and Carmichael (1988) and a melt H20 content of 6.2 wt.%. While the measured and calculated Fe 3+/total Fe are similar, we caution that the expression of Kress and Carmichael (1988) has not been calibrated on hydrous melts.

Results

Preliminary considerations The experimental c o n d i t i o n s a n d resulting phases a n d phase p r o p o r t i o n s are s u m m a r i z e d in Table 2. Average phase c o m p o s i t i o n s with s t a n d a r d deviations are reported in Table 3. Phase p r o p o r t i o n s are calculated by mass balance using bulk c o m p o s i t i o n s (Table 1), experimental phase c o m p o s i t i o n s (Table 3), a n d a least squares technique that incorporates analytical uncertainties for b o t h experimental phases a n d the bulk c o m p o s i t i o n (Albarede a n d P r o v o s t 1977). Iron-loss from the samples to A u capsules can be estimated from the mass balance calculations. Results show a p p a r e n t deficits a n d excesses ranging from - 6.1 to + 3.4% (relative) of the total iron (as F e O ) originally present. The similar a n d small m a g n i t u d e of a p p a r e n t deficits a n d excesses suggests that the a m o u n t of Fe lost to the Au capsules is low a n d certainly less t h a n 10% of the initial FeO. This was confirmed by electron probe analysis of the i n n e r walls of used Au sample capsules. I r o n c o n c e n t r a t i o n s in used A u capsules are below the detection limit of the electron probe even at high beam currents (200 n a n o a m p s , detection limit = 500 p p m Fe). Sisson a n d G r o v e (1992) show with a n o t h e r mass balance a p p r o a c h including a fictive F e O phase that negligible iron was lost in experiments performed in Au capsules at 1 kb. Examples of experiments in which oxygen buffers failed either by o x i d a t i o n or reduction are presented in Tables 4 a n d 5. The results are included as illustrations of the influence of oxygen fugacity o n liquid lines of descent of wet magmas.

Equilibrium E q u i l i b r i u m can be assessed with respect to phase appearance, composition, a n d morphology. All of the experiments in this study are direct syntheses in which crystals have grown from melt. Phase a p p e a r a n c e sequences a n d m i n eral c o m p o s i t i o n a l variations have n o t been reversed.

146 Table 2. Experimental conditions and products. See Table 3 for phase compositions. All experiments at 2kb pressure, pH O = ptotal, fO 2 buffered at NNO Run #

T~

Time (hours)

Phases (wt.% a, plus fluid)

K FeZ+'mg

1050 1050 1035 1025 1000

36 24 41 45 62

g1(93), g1(86), g1(81), gl(81), gl(51),

82-62#3 1012 #4 1000

20 30

82-66#3 #5 #7

1012 1000 965

Fe2+-mg

KD cpxb'liq

+ FeOC

0.27 0.30 0.27 0.30 0.33

0.27 0.25 0.28 0.26

+ + + + +

g1(72), o1(9), pl(10), cpx(9) g1(66), o1(10), pl(13), cpx(ll)

0.34 0.33

0.26 0.25

3.3 -4.9

32 31 72

g1(87), o1(5), pl(1), cpx(6) g1(88), o1(5), pl( < .5), cpx(6) g1(54), o1( < .5), pl(8), cpx(9), sp(3), hbl(26)

0.35 0.36 0.39

0.24 0.24 0.36 b

- 3.5 - 4.0 + 1.2

87S35a1 1025 #2 1000 #5 985 #3 970 # 11 965 # 10 950 # 13 925

8 10 18 8 63 71 66

gl(100) gl(100) gl(100) g1(96), o1( < .5), g1(87), o1(6), g1(67), gl(51),

-

-

85S52b# 11 1024 #12 998 :1t:13 968 # 14 940

8 8 12 20

g1(99), o1(1) g1(98), o1(2) g1(96), o1(4), q(tr) g1(86), o1(1), hbl (13)

79-35q#6 #4 #11 10 12

85S52b #2 #4 #6 # 1 # 14 #9

o1(7), o1(8), o1(8), o1(9) o1(12),

D~

sp( < .5) sp(1), pl(4), cpx(2) sp( < .5), pl(5), cpx(5) sp( < .5), pl(6), cpx(4) sp(1), p1(22),cpx(13)

hbl(3), hbl(5), hbl(17), hbl(26),

pl(1) pl(1), sp(2) pl(13), sp(3), ap(tr) pl(21), sp(3), ap(tr)

0.32 0.36 -

-

0.33 0.37 0.32 0.34

3.4 1.5 1.3 2.9 0.8

- 6.1 - 5.2 - 5.4 - 2.3 --0.1 + 0.2 0 -

1.4

2.2 0.3 + 2.4 -

(85%) + An80 (15%) 20 gl(100) 980 22 gl(100) 975 24 g1(96), pl(4) 965 25 g1(98), pl(.5), o1(1), hbld(.5), q(tr) 965 46 gl(80), pl(14), o1(2), hbl(4) 960 72 g1(73), pl(12), o1( < .5), hbl(15) 943

0.31 0.36 0.32

m

82-66 (83.7%) + Albite (16.3%) 44 g1(86), o1(3), pl( < .5), cpx(100), sp(1) # 1 965

0.32

0.27

+ 0.4

82-66 (98.1%) + NaOH (1.9%) 54 g1(67), cpx(10), hbl(20), sp(1), o1(1) #2 985 49 g1(58), cpx(7), hbl(34), sp(1) 1 965

0.31

0.26 0.26

0 + 0.3

82-66 (99.1%) + NaOH (0.9%) 49 g1(65), ol( < .5), pl(2), cpx(9), hbl(23), sp(1) # 1 965

0.32

0.31

+ 0.2

Fo-Di-An Mix #1 1132 ~2 1120

9

g1(91),

8

gl(4), o1(12), p1(54), cpx(29)

o1(3),

-

4.6

-

7.6

-

2.3

-

1.5

-

4.3

-

3.6

pl(6)

a Proportions calculated by mass balance incorporating analytical errors (Albarede and Provost 1977) for anhydrous phases normalized to 100%, glasses to 94%, and hornblende to 98% to account for H20, with all Fe as FeO and ignoring MnO and P205. Proportions are precise to 5% relative or better at abundances > 20%, -10% at abundances of 10 20%, 20% at abundances of 1 10%, and > 50% at abundances 930 ~ in which Fe-loss ranges from 8 to 37% (average a b o u t 22%). Melts m o r e mafic than intermediate andesite were not produced. Lower temperature, m o r e crystal-rich Q F M experiments that p r o d u c e d rhyolitic melts a p p e a r to have lost little Fe. Stern and Wyllie (1975) demonstrate that in three hours andesite melts lost 85% of their Fe to Pt capsules, 35% to Ag3oPd7o , and 25% to

150 Table 4. Experimental conditions and products, fO > or < NNO. See Table 5 for phase compositions. See Table 2 for abbreviations and notes Run

T~

79-35g 7 82-66 4~6

fO 2

Time (hours)

Phases (wt.%, plus fluid)

+ FeO

1000

> NNO

48

gl (46), ot(9), sp(5), cpx(18), p1(22)

- 1.0

975

< NNO

50

g1(65), o1(8), sp(.5), cpx(12), pl(14)

+ 0.8

Table 5. Electron microprobe analyses of experimental phases; f02 > or < Ni-NiO. See Table 4 for conditions. See Table 3 for abbreviations and notes Run ~ Phase

SiO 2

AI203

FeO

MgO

CaO

TiO 2

MnO

Na20

K20

P205

Cr203

NiO

Total

Fe203

79-35g g1(11) o1(11) sp(10)

4t:7 55.6(6) 19,6(2) 4.99(9) 5.23(10) 8.52(9) 0.90(2) 0.17(1) 4.63(15) 0.19(1) 0.12(2) 93.9 41.1(3) 0.08(7) 10.0(9) 48.4(7) 0.34(5) 0 0.34(3) 0 0.06(6) 100.3 0.28(2) 8.1(2) 17.9(5) 10.6(2) 0.07(2) 1.98(13) 0.43(3) 0.11(14) 0.15(3) 60.9 100.5 cpx(9) 47.4(6) 7.07(49) 7.6(5) 13.9(3) 22.1(1) 0.91(10) 0.16(3) 0.40(2) 0.12(2) 99.7 pl(8) 45.8(5) 34.0(4) 0 . 5 ( 1 ) 0.08(3) 17.7(34) 1.26(19) 0.03(1) 99.4 82-66 #6 g1(13) 54.9(2) 18.9(1) 7.91(12) 3.55(5) 7.87(5) 1.29(3) 0.15(2) 4.04(12) 1.11(3) 0.23(1) 93.0 o1(7) 37.7(2) 0.1(1) 24.5(5) 36.4(5) 0.35(8) 0.04(3) 0.39(5) 0.03(2) 99.5 0 sp(6) 0.18(1) 5.7(1) 34.9(6) 4.38(5) 0.10(3) 10.7(2) 0.33(2) 0.14(7) 0.05(1) 43.4 99.9 cpx(ll) 51.2(4) 3.4(5) 7.32(13) 15.3(1) 21.5(2) 0.73(9) 0.05(4) 0.16(2) 0.18(5) 99.8 pl(7) 46.3(3) 33.6(3) 0.60(9) 0.07(1) 17.0(2) 1.73(10) 0.03(0) 99.3

Ag75Pd25. Insufficient data are presented to assess Felosses in the andesite-melting experiments of Eggler and Burnham (1973) performed in AgsoPdso at Q F M and NNO, but Fe-loss is likely in light of the similar experiments listed above. These results show that Fe-loss to AgPd or Pt capsules can be substantial at high melt fractions, high temperatures, or low sample/container mass ratios at oxygen fugacity less than the magnetitehematite buffer. The experiments summarized here cannot have reached equilibrium unless and until the samples had lost sufficient Fe to be in equilibrium with their noble metal capsules. The rate of iron-loss is controlled by diffusion rates of Fe into the noble metal container and by the mass of the container, which often exceeds the mass of silicate material. The diffusion rates for Fe in noble metals are comparable to Fe-Mg intracrystalline diffusion rates for silicate minerals and are 10 4 times slower than diffusion rates in silicate melts (Grove 1981) Therefore, the silicate melt transports Fe to the noble metal capsule rapidly. The crystalline silicates change composition during the time of the experiment as their surrounding melt is continually depleted in Fe in its attempt to reach equilibrium with the metal capsule. Spulber and Rutherford (1983) and Rutherford (personal communication) state that significant iron was not lost from hydrous basalt samples melted at temperatures below 1100 ~ in thin-walled Ag 7o Pd 3o. We have tested and confirmed this observation by fusing sample 87S35a to a liquid for 8 h at 1050~ and 2 k b pressure, HzOsaturated, at the N N O buffer in AgvoPd3o tubing pro-

vided by M. Rutherford. The iron content of the quenched glass was within analytical error of that of the starting material and, as with Au, any iron lost to AgvoPdao is below our ability to measure. This differs from the result of Stern and Wyllie (1975) in Ag75Pd25 , likely due to a higher ratio of sample to container in our case. The 2 kb results of Spulber and Rutherford (1983) are therefore used later to supplement our experimental data. The large iron-losses in previous hydrous basalt melting studies contrast with minimal iron-losses in the present study and Spulber and Rutherford (1983). Low or minimal iron-losses provide a greater opportunity for a close approach to equilibrium. The cost of performing experiments in Au, and preventing iron-loss, has been that we are restricted to temperatures below the melting point of Au (1064 ~ at 1 atm). As will be shown later, this is not a serious limitation under HzO-rich conditions. Replicate glass analyses show that most experimental liquids were homogeneous, a prerequisite for equilibrium (but see note "f", Table 3). Synthesized crystals are euhedral and equant or tabular and individual crystals commonly reach sizes of several hundred microns across. Experiments were of sufficient duration that skeletal crystals were not produced. The crystals are largely newlygrown material and, unlike 1-atm experiments (Grove and Juster 1989, their Fig. 1), the quantities of unreacted starting materials are minimal or completely absent. Nevertheless, individual mineral-rim analyses vary outside of the practical reproducibility established earlier from glass analyses. Back-scattered electron imaging

151 shows that most crystals are weakly zoned, even along their rims. Sector zoning is common in the experimental clinopyroxenes, and has been found in some of the experimental hornblendes. The presence of sector zoning and the small heterogeneities in compositions of crystal rims are inconsistent with complete attainment of equilibrium. Experiments of far greater duration than are presently feasible would be required to produce true equilibrium phase compositions. The shortcomings of slightly heterogeneous solid phases are likely to have been present in all previous studies of the wet melting of basalts and andesites and were compounded in some of those studies by the additional problem of severe iron-loss. Heterogeneous solid phases may have gone unrecognized in those studies due to less precise analytical or electron beam imaging capabilities than are currently available, or heterogeneous minerals may simply have been ignored. Two kinds of additional evidence indicate a close approach to equilibrium. 1. The average compositions of crystal margins, discussed later, show regular and consistent partitioning of certain elements between minerals and liquids independent of whether the mineral in question was present in the starting material. 2. The mineral-liquid element partitioning for olivine, plagioclase, hornblende, and high-Ca pyroxene, expressed as exchange KDs and discussed later, are as expected from other phase equilibrium studies and natural samples.

Mineral appearance sequences and melt compositions Hydrous high-alumina basalts. Most of the HzO-saturated experiments on volcanic high-alumina basalts have produced high-alumina basalt or basaltic andesite liquids that are saturated with olivine, high-Ca pyroxene, and calcic plagioclase; many also contain a spinel mineral. Mineral appearance sequences are summarized in Table 2. The primitive high-alumina basalt, 79-35g, consists of > 90 wt.% liquid plus small amounts of olivine (Fo86.5) and a Cr-A1-Mg spinel at 1050~ A repeat experiment at 1050~ produced~85 wt.% liquid, plus olivine, high-Ca pyroxene, calcic plagioclase (An93), and a Cr-A1-Mg spinel. The liquid resembles high-alumina basalt with an intermediate MgO content (49.4 wt.% SiO2, 19.2 wt.% A1203, 6.58 wt.% MgO, normalized to 100 wt.% anhydrous). At lower temperatures 79-35g continues to crystallize olivine, high-Ca pyroxene, calcic plagioclase, and an increasingly Fe-Ti rich spinel. By 1000 ~ the lowest temperature investigated, the sample is 50 wt.% crystalline and the liquid is low-MgO highalumina basaltic andesite (52.5 wt.% Si02, 19.2 wt.% A1203, 4.99 wt.% MgO). The two lower-MgO high-alumina basalts, 82-66 and 82-62, also crystallize olivine, high-Ca pyroxene, and calcic plagioclase but a spinel mineral does not form until temperatures below 1000~ Hornblende appears by 965 ~ in sample 82-66 and coexists with olivine, calcic plagioclase, high-Ca pyroxene, and an Fe-Ti spinel. Liquid ranges from low-MgO high-alumina basalt (51.5 wt.~ SiO2, 19.2 wt.% AlzO 3, 4.98 wt.% MgO) to alumi-

nous andesite (59.1 wt.% SiO 2, 19.1 wt.% A1203, 3.25 wt.% MgO). Hornblende 9abbro and diorite. Crystallization of the hornblende gabbro and diorite differ from the volcanic high-alumina basalts in several respects. First, liquids with comparable silica contents exist at lower temperatures in the gabbro and diorite. Second, neither the hornblende gabbro nor the diorite crystallize high-Ca pyroxene. Third, the gabbro and diorite produce hornblende from liquids with low silica contents, equivalent to low-MgO high-alumina basalt or basaltic andesite. Hornblende gabbro 87S35a is saturated nearly simultaneously with olivine, hornblende, and calcic plagioclase between 985 and 970 ~ The liquid at 970 ~ is lowMgO high-alumina basalt (52.1 wt.% SiO2, 19.3 wt.% AlzO3, and 4.14 wt.% MgO). Magnetite appears at 965~ and olivine is lost by 950~ Apatite appears between 965 and 950~ and persists to lower temperatures. By 925 ~ the lowest temperature investigated, the sample is 50 wt.% crystalline and the liquid corresponds to corundum-normative aluminous andesite (60.0 wt.% SiO2, 18.9 wt.% A1203, 1.76 wt.% MgO). This is the only corundum-normative liquid produced in this study. Hornblende diorite 85S52b crystallizes olivine from a temperature higher than 1024~ down to 968~ Hornblende joins olivine between 998~ and 968~ Plagioclase did not crystallize over the temperature range investigated in this composition under HzO-saturated conditions. Plagioclase (An so) was mixed with 85S52b in proportions 85 wt.% rock, 15 wt.% plagioclase to force plagioclase saturation. The plagioclase-added composition first precipitates plagioclase between 965 ~ and 975 ~ Olivine and hornblende appear near 965 ~ and coexist with calcic plagioclase and low-MgO highalumina basaltic andesite liquid (53.7 wt.% SiO2, 20.0 wt.% A120 3, 3.75 wt.% MgO). Olivine, hornblende and plagioclase coexist through 943 ~ the lowest temperature investigated. By 943~ the sample is 27 wt.% crystalline, and the liquid corresponds to aluminous andesire (57.1 wt.% Si02, 19.1 wt.% Al20 3, 2.80 wt.% MgO). Forsterite-diopside-anorthite-HzO experiments. The experiments on high-alumina basalts produced basaltic through andesitic liquids coexisting with aqueous fluid, high-Ca pyroxene, calcic plagioclase, and olivine (with and without a spinel phase). Experiments were performed in the quaternary join Fo-Di-An-HzO to locate the HzOsaturated piercing point at which liquid coexists with fluid, diopside solid solution(ss) , forsterites~, and anorthite. At 2 kb, natural liquids saturated with olivine, highCa pyroxene, plagioclase, and aqueous fluid will lie at lower temperatures than that of the piercing point. The composition of the piercing point was estimated from the crystallizing proportions of olivine, high-Ca pyroxene, and plagioclase in the high-alumina basalt experiments, and a mix was prepared by combining synthetic Fo, Di, and An in the observed proportions. A piercing point temperature was estimated from a simple algorithm relating liquid composition to temperature presented elsewhere (Sisson and Grove 1992). An initial experiment at 1132 ~ produced liquid, Fo~, An, and fluid. A second experiment 12~ cooler was

152 largely crystalline, but contained small pools of liquid located next to Diss, FOss, An, and fluid. The liquid is slightly enriched in silica relative to the Fo-Di-An plane (1.5 wt.% excess SiO2) while the pyroxene is enriched in Ca-Tschermak's component. The true piercing point should lie a few degrees hotter than 1120 ~ a difference below our ability to control.

Experiments addressing hornblende stability Experiments were performed to investigate controls on hornblende stability in hydrous mafic liquids. Hornblende gabbro sample 87S35a has high alumina and a basaltic silica content (Table 1) and crystallizes hornblende as a liquidus mineral. Volcanic sample 82-66 also has high alumina and a basaltic silica content but has lower N a 2 0 and higher MgO than 87S35a. It crystallizes hornblende only at temperatures well below the liquidus when the liquid has reached an andesitic composition. Cawthorn and O'Hara (1976) and Cawthorn (1976) showed that melt N a 2 0 content exerts a major control on the stability of amphibole in the system CMAS-NaEO-H20 at an H20-pressure of 5 kb. Simplified hydrous basalt melts required at least 3 wt.% Na20 to form pargasite. We examined the influence of N a 2 0 on hornblende stability in natural basalts at an H20-pressure of 2 kb by artificially raising the sodium content of 82-66 to equal that of 87S35a (as analyzed by electron probe of fused samples). Sodium was first added by mixing powdered Amelia albite to 82-66. The mix was run at 965 ~ the same temperature at which 82-66 (plus H 2 0 ) produced 26 wt.% hornblende, plus olivine, high-Ca pyroxene, plagioclase, and titanomagnetite in andesitic melt. No hornblende grew from the albite-added mix. Instead, the addition of albite produced liquid plus olivine, high-Ca pyroxene, plagioclase, and titanomagnetite. Another mix was prepared by adding sodium to 82-66 as N a O H so that the bulk sodium content equaled that of 87S35a. This composition produced liquid and abundant hornblende, plus high-Ca pyroxene and titanomagnetite at 965 ~ Neither olivine nor plagioclase grew from the NaOH-enriched composition, and the liquid is nepheline-normative. Hornblende persisted at 985 ~ in the NaOH-enriched composition coexisting with liquid, titanomagnetite, highCa pyroxene, and olivine. The liquid has 54.8 wt.% SiO 2 (calculated anhydrous) and is nepheline-normative. One final mix was made from equal parts NaOH-enriched mix with pure 82-66. This composition produced the same solid phases as pure 82-66 at 965 ~ but the liquid has a lower silica content (55.4 vs. 59.1 wt.%). Differences between the albite-added and NaOHadded experiments can be reconciled by the model reaction: 1 Na2Oliq -]- 1 H20 + 2 anorthite + 2 forsterite + 3 SiO2u q = 1 pargasite + 1 albit%q Addition of N a 2 0 drives the reaction to the right, producing hornblende at the expense of olivine, calcic plagioclase, and silica. Addition of albite has the opposite effect, consuming hornblende.

Mineral compositions and mineral-liquid exchange reactions Pyroxene. The synthesized pyroxenes are high-calcium diopsides and salites with moderate to high alumina contents (2.9-7.9 wt.%). Pyroxenes do not have lower Ca or higher enstatite-ferrosilite contents at lower temperatures, consistent with the absence of a coexisting low-Ca pyroxene. Pyroxene rim compositions (Table 3) are somewhat heterogeneous and sector zoned grains are common. Aluminous sectors and grains are more iron and titaniumrich, whereas low-alumina sectors and grains are more magnesian. Crystals and sectors with moderate alumina have a well defined exchange K Fe-Mg ((Fe/Mg)pyx/ (Fe/Mg)liq) of 0.23 calculated with total Feliq or 0.26 calculated with FelZq = 0.86 total Fell q (based on the M6ssbauer measurement of melt ferric iron, Table 2). Grove and Bryan (1983) determined the same K~ e-Mg value, 0.23, using total melt Fe in 1 atm pressure QFMbuffered anhydrous experiments on mid-ocean ridge basalts. Applying the expression of Kress and Carmichael (1988) to the experiments of Grove and Bryan gives Fell+ as ~ 0.87 of total Fellq and KFDe2+ - ~ g ~ 0.26. Partitioning is more irregular between liquid and higher alumina crystals and sectors and the crystals are more iron-rich. The value of K w-Mg for the more aluminous sectors and grains is ~0.27 (total Feliq) or ~0.31 (estimated Fel]+), though poorly defined. Olivine. Iron and magnesium also show regular partitioning between olivine and liquid. Olivine-liquid pairs at NNO have an exchange K~~-Mg ((Fe/Mg)oliv/(Fe/Mg)liq) averaging 0.28 ( _ 0.05 2-a) for total Fenq or 0.33 for estimated Fel2+ (Table 2). This is close to the K~e-~ag values of 0.30 ( _+0.06 2-a) determined by Roeder (1974) using estimated melt Fe 2+ and 0.29 (Fetotal) o r 0.33 (Fe 2 + ) calculated from the MORB melting experiments of Grove and Bryan (1983) performed at QFM. Plagioclase. Compositions of coexisting plagioclase and liquid are shown in Fig. la. The exchange K ca-ha for plagioclase defined as (Ca/Na)plag/(Ca/Na)li q, has a value of ~ 5.5. for H20-saturated liquids at 2 kb pressure. We have included two plagioclase-liquid pairs produced in experiments on a granodiorite (Appendix 1), and find the exchange K ca-Na to hold for liquids ranging from highalumina basalt to low-silica rhyolite. Figure la also shows results for experiments on high-alumina basalts with 2 wt.% H 2 0 at 2 and 5 kb pressure (Baker and Eggler 1987) as well as high-alumina basalts at 1 kb H20saturated with about 4 wt.% H 2 0 in the melt (Sisson and Grove 1992). Exchange KCDa-Nas show a progressive increase with melt H 2 0 content from approximately 1.7 for melts with 2 wt.% H20, 3-4 for melts with 4 wt.% H 2 0 and 5.5 for melts with 6 wt.% H20. The 2 kb H20-saturated exchange K ca-Na is also higher than is found in dry experiments at low or high pressures (Fig. lb). Only one of fifty exchange KCa-Nas determined for anhydrous basalt liquids from 8 to 12 kb pressure has a value above 2.0. Data at higher pressures are scarce, but all values lie below 2.5 at pressures as high as 20 kb. Spinel. Synthesized spinel ranges from Mg-A1-Cr rich to magnetite and compositions cover much of the range

153 SPINELS

?lagioclase-Liquid Ca-Na Exchange

100

, '-}:

~:,: Olk

80 19 Wt.%(Byers

1959, 1961; Marsh 1976, 1982; Kay and Kay 1985; Brophy 1986; Myers et al. 1986;Nye and Reid 1986;Gust and Perfit 1987). Fuego samples include all eruptive products with SiO2 < 63 wt.% (Carr and Rose 1987)

160 extent. The effect of HzO-saturation at 2 kb is to lower the appearance temperature of silicates by over 150~ relative to their appearance at 1-atm anhydrous conditions. Oxides become stable at 1-atm at this low temperature in basaltic bulk compositions, but a significant interval of silicate mineral crystallization has already occurred at higher temperatures and liquids are enriched in FeO and to a minor extent SiO 2. The dry liquids at oxide-saturation are low-alumina ferrobasalt or low-alumina ferroandesite and do not resemble members of the calc-alkaline rock series (Grove et al. 1982; Helz 1987; Juster et al. 1989). At 2 kb under HzO-saturated conditions oxide and silicate minerals crystallized at close to the same temperature in basaltic or near-basaltic compositions. Since oxides are near-liquidus phases, FeO is removed from the liquid, and the resulting small change in FeO/MgO with increasing silica and alkalis is that characteristic of the calc-alkaline differentiation series (Figs. 5, 6). The first influence of HzO on the path followed during differentiation was anticipated by Yoder (1965, 1969a) and was developed by Grove and Baker (1984) by contrasting the differentiation paths followed during crystallization of basalt magma at 1 atm with the rock series trends defined by calc-alkaline lavas from subductionrelated volcanoes. At 1-atm the differentiation trend followed by basalt is the tholeiitic trend characterized by early FeO/MgO enrichment and little alkali enrichment. The crystal-liquid controls on the tholeiitic trend are crystallization dominantly of plagioclase, reduced proportions of Fe-Mg silicates, and appearance of oxides only at late stages and low temperatures. Plagioclase is enriched in Na20 and SiO2, and alkali enrichment is suppressed. Fe-Mg silicates crystallize in diminished proportions, and the residual liquid becomes FeO-enriched. The second effect of H20 on crystallization has not been widely recognized (but see Gill 1981, p 278). Osborn (1959) suggested that the calc-alkaline trend might be produced by the oxidizing effect of continental crust on differentiating magma. Osborn (1959), Presnall (1966), Roeder and Osborn (1966), and others carried out experimental investigations at 1-atm in simple systems (MgOFeO-Fe203-SiO 2 ___CaO, A120 3, or haplobasalt), and explored the effect of variations in fO2 on the FeO contents of liquids produced by fractional crystallization. From these experiments Osborn and co-workers concluded that any haplobasalt that crystallizes olivine, pyroxene, and spinel in a closed system shows a trend of ironenrichment at nearly constant SiO2 content. Osborn and coworkers found that FeO-depletion and SiO2-enrichment trends could be achieved during fractional crystallization if differentiation was carried out at elevated TO 2 that was externally controlled. The problem encountered by these investigators was that the extent of FeO-depletion and SiO2-enrichment were tied to fO2. The f O 2 required to cause pronounced SiO2-enrichment and FeOdepletion trends in haplobasalt are near the hematitemagnetite buffer and are more oxidizing than the crystallization conditions inferred for calc-alkaline magmas. The f O 2 conditions commonly estimated for subduction-related magmas range from one or two log units around the NNO buffer (Gill 1981). Only very modest S i O 2-

enrichment and FeO-depletion are possible in haplobasalts under these fO2 conditions (Roeder and Osborn 1966). Experiments carried out at 1-atm and variablefO 2 conditions using natural compositions have also shown that very highfO2s are required to stabilize an iron oxide phase near the liquidus of basaltic and andesitic liquids (Roeder 1974; Juster et al. 1989; Grove and Juster 1989). At fO2 conditions comparable to those inferred for arc environments, iron oxides were not stable near the liquidus at 1-atm unless melts were distinctly Fe and Cr-rich ( > 200 ppm Cr, Hill and Roeder 1974). The results of our HzO-saturated NNO-buffered experiments provide a mechanism for stabilizing an oxide phase nearer the liquidus in calc-alkaline series, thus promoting early FeO-depletion. H20 destabilizes silicate minerals relative to liquid, but has lesser effect on the stability of spinels. Other studies show this effect. Experiments in the system pyrope-diopside-H20 and pyropegrossular-H20 at 10 and 30 kb (Yoder 1969b; Sekine and Wyllie 1983) show that the addition of H20 expands the liquidus volume of spinel relative to silicates. Similarly, experiments on a lunar high-alumina basalt (Ford et al. 1972) show that the addition of H20 stabilizes spinel on the liquidus at low pressures (2 kb), whereas at anhydrous conditions spinel does not appear on the liquidus except at pressures in excess of 10 kb. Two other experimental studies on natural terrestrial basalts also demonstrate the ability of H20 to promote early iron oxide saturation. The experimental results of Baker and Eggler (1983) on Aleutian high-alumina basalt at 2 kb pressure, NNO-buffered, show that iron oxides appear approximately 200~ below the liquidus of the basalt if the melt is dry. H20 contents of 4.5 wt.% or more depress the appearance temperature of crystalline silicates by about 250 ~ whereas the appearance of an iron oxide phase is depressed 50 ~ or less. The result is that an iron oxide lies on or near the liquidus of the Aleutian highalumina basalt at H20 contents of 4.5 wt.% or more. Osborn (1963, 1969), in a discussion of unpublished experiments by Fudali and of work by Hamilton et al. (1964), shows that Columbia River Basalt exhibits a comparable effect. Magnetite appears about 60 ~ below the liquidus of Columbia River Basalt held in sealed capsules at 1 atm pressure (~NNO), but lies only about 15 ~ below the liquidus at 1 kb pressure H20-saturated at the QFM buffer and defines the liquidus at the magnetite-hematite (MH) buffer. Interpolation of the data of Hamilton et al. to the NNO buffer would place magnetite on or above the silicate liquidus of Columbia River Basalt at 1 kb pressure, HzO-saturated. Convergence of the liquidus with the appearance temperature of oxide for high-alumina basalts 79-35g and AT1, and Columbia River Basalt is illustrated in Fig. 11. The data for AT-1 are from Baker and Eggler (1983, their Figs. 3 and 4). On the assumption that the silicate liquidus is not strongly sensitive to fO2, the anhydrous oxide appearance temperature for 79-35g at NNO has been interpolated from 1-atm experiments on 79-35g and on high-alumina basaltic andesite sample 79-38b over the range QFM to NNO + 2 (Grove et al. 1982; Grove and Juster 1989). The data for Columbia River Basalt are from Osborn (1969) at 1 atm and Hamilton et al. (1964) at 1 kb

161 30O

&

'

i

,

i

,

\

250

i

,

F IAleutian

I

e~eine

i

i

HAB A T - l , Lake

The oxygen and hydrogen budget during magnetite fractionation

i NNO

HAB 7 9 - 3 5 g ,

NNO

200 I

150

i-2=

100 I

o F--

5o

O

,

0 0

1

2 wt.~.

3 H20

,

4 in

---- . - - ~ 5

6

.

7

melt

Fig. 11. Temperature differencebetween liquidus and appearance of Fe-oxide versus H20 content of melt H20-saturated and interpolated as described already. Figure 11 demonstrates that at fO2 appropriate for arc magmas, high wt.% H 2 0 in the melt leads to near liquidus crystallization of an iron oxide phase. Holloway and Burnham (1972) were the first to show that magnetite could coexist with hydrous basalt melt at the N N O buffer. Spulber and Rutherford (1983) melted similar basalts at comparable H 2 0 pressure but w i t h f O 2 in the wiistite stability field and found that ulv6spinel and ilmenite do not form until liquids attain SiO 2 contents > 55 wt.% (normalized anhydrous). Thus, a high melt H 2 0 content alone is insufficient to produce Fe-oxides with basalt melt. Direct comparisons are difficult to make with the experimental results of Helz (1973, 1976), Eggler (1972), and Eggler and Burnham (1973). Helz (1973, 1976) studied the melting of various low-alumina basalts including Columbia River Basalt at 5 kb H20-saturated at the Q F M and M H buffers. Magnetite appears between 1000 ~ and 970~ in the Columbia River Basalt at Q F M but never appears at H M and instead is replaced by hematite. Extrapolation of the magnetite appearance temperature to the N N O buffer is therefore very uncertain. The liquidus was not closely approached for any composition, which further precludes discussion of oxide crystallization from basalt or basaltic andesite liquids. Eggler (1972) and Eggler and Burnham (1973) studied melting of silicic andesites at various H 2 0 contents and total pressures, chiefly at the Q F M buffer. A major conclusion of their work was that magnetite was not a liquidus phase and therefore was unlikely to play a significant role in formation of calc-alkaline suites. Their experiments were, however, carried out with techniques similar to those of Baker and Eggler (1987) in which on average 44% of the total Fe was lost to noble metal capsules. Comparable Fe-losses are therefore probable in the Eggler and Burnham experiments, which suggests that the absence of early magnetite could result from Fe-loss. Wagner et al. (1991) and Wagner (written communication) have produced magnetite coexisting with silicic andesite liquids at 1 kb, H 2 0 saturated, in experiments performed in Au capsules.

One conclusion of haplobasalt crystallization experiments is that oxygen must be added if liquids are to become silica-enriched and iron-depleted (Osborn 1959; Presnall 1966; Roeder and Osborn 1966). This led to the proposal that either oxygen enters magma chambers or, in the case of hydrous magmas, that hydrogen leaves chambers. We have succeeded in producing silica enrichment and iron depletion at f O 2 s comparable to those observed in arc magmas and it is important to examine whether our results imply that hydrogen loss occurs in nature, and if so, how much must be lost. Fractional crystallization models starting with basaltic melt from experiment 82-66 no. 5, with Fe 3 § Fe of 0.14 and forming minerals in a closed system in the proportions necessary to produce calc-alkaline daughter liquids show that melt ferric iron would be completely exhausted after only 25-30 wt.% solidification unless some other processes operated. Presnall (1966) recognized that some of the HzO in a fluid-saturated magma would dissociate to Oz plus H 2 as it exsolved from the melt and that the free oxygen would be available to convert FeO to F%O 3. Figure 12 develops this model and gives the

100

,

,

,

25

80 -

20

~o 60 o

15

--

E 40

J

20F

10

7"

45 H2

0

\

expelled per lOOg magma

~

0

8 Fe0* 6 X3 X O

MgO

4 Na20 2

I

i

i

i

52

54

56

58

Si02

=

]

60

(wt.z)

Fig. 12. Lower figure, calculated variation of selected oxides (normalized anhydrous) during 40% crystallization of the assemblage 6% olivine, 35% pyroxene, 49% plagioclase, 10% magnetite beginning with melt from experiment 82-66 no. 5. Silicate mineral compositions determined by KDve-Mgand KDCa-Na, spinel from 8266 no. 6. Note silica and alkali enrichment and iron depletion. Upper figure, cumulative H z evolved in excess of that in equilibrium (NNO, 1000-950 ~ exsolved fluid (per 100g total magma). Melt H20 fixed at 6.2 wt.% (saturation at 2 kb), melt Fe 2+ - Fe 3§ from the expression of Kress and Carmichael (1988) but adjusted as in text, H20 and H 2 fugacity coefficientsfrom Burnham et al. (1969) and Shaw and Wones (1964), equilibrium constant for H20 from Labotka (1991, Table 11). Curve with squares shows change in excess H 2 production with increasing SiO2

162 amount of H z that must be lost to hold a high-alumina basalt and exsolved fluid of otherwise fixed total composition on the N N O buffer at 2 kb pressure as the basalt crystallizes over the temperature interval 1000~ to 950 ~ Olivine, pyroxene, plagioclase, and magnetite are formed in the proportions necessary to produce calcalkaline daughter liquids (from Table 6) and with compositions determined by experimental Fe z +-Mg and Ca-Na KDs. Amounts of Fe 2 + and Fe 3 + in the melt are calculated with the expression of Kress and Carmichael (1988) but with their f O z term adjusted to agree with our M6ssbauer-determined Fe 3+ in glass from sample 87S35a. After 40 wt.% crystallization, the melt has a SiO2 content of 59.4 wt.% (normalized anhydrous) and is andesitic. To remain on the N N O buffer to 40 wt.% solidification, each 100 g of magma must lose ~ 9 x 10-3 tools of H2 in excess of that in the equilibrium exsolved fluid. This corresponds to a loss of about 180 ppm by weight H z. Note that while hydrogen loss appears required, the amount necessary is only 2.6% of the total hydrogen initially contained in a melt with 6.2 wt.% HzO (approximately saturation concentration at 2 kb). Natural magmas also contain carbon and sulfur and these elements complicate determination of the H 2 budget. Heald et al. (1963) show for natural volcanic gasses that COz is the dominant carbon species in exsolved fluid or gas a t f O 2 near N N O and upper crustal pressures. Since carbon is dissolved in melt almost entirely as carbonate species (Stolper and Holloway 1988) exsolution of small amounts of COz will have little redox effect and thus will not influence the hydrogen budget significantly. The situation for sulfur is different. Heald et al. (1963) show that in an H20-dominated magmatic fluid withfOz near the N N O buffer, HzS is orders of magnitude more abundant than SO2 at pressures greater than 100 bar and temperatures of 1100 ~ or less. Therefore, a hydrous arc basalt that reaches fluid saturation at kilometer depths with f O e initially near N N O and then crystallizes will exsolve the majority of its sulfur as HzS. If sulfur were dissolved as $2, an initial melt sulfur concentration of 0.29 wt.% would be required to bind all of the excess H 2 as H2S. This is within the range (0.0-0.33 wt.% $2) measured by Anderson (1974, 1982) in arc basalt and basaltic andesite glasses. Lower initial sulfur concentrations are required if the sulfur is dissolved as sulfate species, as would be the case for an initial oxygen fugacity

greater than that of the N N O buffer (Katsura and Nagashima 1974). This consideration of the combined budgets for Hz and $2 suggests that hydrous arc basalts that reach aqueous fluid saturation at a n f O z initially near the N N O buffer can yield calc-alkaline daughter liquids by magnetite, pyroxene, olivine, and plagioclase crystallization with no preferential H 2 loss at all. The differentiating magma plus fluid system would likely follow a temperature and oxygen fugacity trajectory similar to that produced under oxygen-buffered conditions, as is observed in arc suites (Gill 1981, p. 124).

Calcic plagioclase in arc magmas It has often been observed that the plagioclase phenocrysts in arc magmas are unusually calcic. Plagioclase crystals, An9o , are common in arc high-alumina basalts and crystals as calcic as An95 are not unknown (Kuno 1950). Yoder (1969a) ascribed the presence of calcic plagioclase in arc magmas to the presence of H20, reasoning that since HzO depresses the temperature of the liquid + solid phase loop in the system albite-anorthite, at a given temperature a wet magma will contain more calcic plagioclase than an otherwise identical dry magma. Arculus and Wills (1980) criticized Yoder's model, observing that the liquidus plagioclase would have the same composition in a wet or a dry melt if the plagioclase plus melt twophase loop was simply depressed by addition of HzO with no change in its shape. Instead, they drew upon the experiments of Johannes (1978) in the systems albiteanorthite-H20 and albite-anorthite-silica-H20 that demonstrate that the plagioclase plus melt two-phase loop is broadened as well as depressed in temperature by the addition of H 2 0 , with the result that the liquidus plagioclase of a wet melt is markedly more calcic than the liquidus plagioclase of an otherwise identical dry melt. Our experiments on high-alumina basalts H20saturated at moderate pressure, combined with published plagioclase-basaltic or andesitic melt pairs, allow us to test how much H 2 0 is necessary to produce the calcic plagioclase observed in high-alumina basalts. We do this by employing the KCDa-Na that was determined in our H20-saturated experiments (Kt) = ~ 5.5), in experiments on high-alumina basalts with 2 wt.% H 2 0 in the melt (KD = 1.3-1.7; Baker and Eggler 1987), and in experiments on

Table 7. Predicted high-alumina basalt liquidus plagioclase compositions

KCDa-Na = Location Aleutians

n 45

Fuego, Guatemala

58

Lesser Antilles

64

(average) (range) (average) (range) (average) (range)

1.0

1.7

5.5

An 62.8 (An 52.0-70.4) An 60.6 (An 54.8-68.5) An 69.8 (An 67.1 76.3)

An 74.1 (An 64.8-80.2) An 72.3 (An 67.3-78.7) An 79.7 (An 77.6-84.6)

An 90.2 (An 85.6-92.9) An 89.4 (An 86.9-92.3) An 92.1 (An 91.8-94.7)

Compilation has been restricted to analyses with SiOz < 53 wt.% and A1203 > 18 wt.%. Data sources - Aleutians: Brophy (1986), Byers (1959, 1961), Gust and Perfit (1987), Kay and Kay (1985), Marsh (1976, 1982), Myers et al. (1986), Nye and Reid (1986). Fuego, Guatemala: CENTAM database of Carr and Rose (1987). Lesser Antilles: Brown et al. (1977), reported average compositions in that work have been weighted according to the number of analyses averaged

163

andesite liquids saturated with these minerals have high alumina contents and higher normative proportions of olivine: high-Ca pyroxene than similarly-saturated dry liquids. These differences are also observed between calcalkaline and tholeiitic magma series and this suggests that the fractional crystallization controls on these two magma series result from high and low H 2 0 contents, respectively. H 2 0 also depresses the appearance temperatures of crystalline silicates, but has lesser effect on spinel or titano-magnetite. At moderate-to-high H 2 0 contents and relevantfO2, an Fe-rich spinel phase can appear near or at the liquidus of basaltic melt. As a consequence H z O bearing high-alumina basalts can produce calc-alkaline daughter liquids by crystallization differentiation that requires no volatile exchange with wall rocks. The plagioclase that forms from high-H20 basalts is CaO- and Alz O 3-rich and alkali- and SiO z-poor- Crystallization of this phase further promotes the alkali and silica enrichment that is characteristic of the calc-alkaline series and is consistent with the calcic plagioclase found in many arc suites. Amphibole stability is sensitive to melt compositional features in addition to H 2 0 content. For basalt or basaltic andesite liquids with H 2 0 < 6 wt.%, amphibole will only appear as a phenocryst if the melts are also sodarich and low temperature. The absence of amphibole phenocrysts in most high-alumina basalts is therefore not evidence that those magmas are dry. In combination, these results demonstrate mechanisms that link the calcalkaline differentiation series of arcs to the subducted and recycled H20 that is so apparent in the eruption style of arc magmas.

dry high-alumina basalts from 1 atm to 8 kb pressure (KD = ~ 1.0; Grove et al. 1982; Baker and Eggler 1987; Meen 1987; Bartels et al. 1991) (Fig. lb). Calculated liquidus plagioclase compositions (average and both maximum and minimum An content) are presented in Table 7 for typical high-alumina basalts from the Aleutians, from the volcano Fuego in Central America, and from the Lesser Antilles, assuming that whole-rocks approximate melt compositions. In each locality, the calculated tiquidus plagioclase for dry or low H 2 0 content conditions are considerably less calcic than An9o. The calculated plagioclase for H20-saturated conditions at 2 kb are approximately An9o, matching that typically observed as phenocrysts. We conclude from this that the plagioclase in common high-alumina basalts grew from melts with HzO contents certainly greater than 2 wt.% and likely as high as 6 wt.%. This subject is expanded on in a companion paper (Sisson and Grove 1992). Conclusions

We have documented the effects of H 2 0 on the crystallization behavior of high-alumina basalts at moderate pressure and oxygen fugacities appropriate to subduction zone magmas. We find that many aspects of the calcalkaline differentiation series are explicable if arc basalt or basaltic andesite magmas contain approximately 4-6 wt.% of H20. H 2 0 shifts the relative sizes of the olivine, high-Ca pyroxene, and plagioclase primary liquidus volumes with the result that wet basalt and basaltic

Appendix 1. Additional experimental results Results are reported for two experiments on a granodiorite, 83S296b, collected from the Cretaceous granodiorite of Mitchell Peak in the south central Sierra Nevada batholith. Experiments were performed at 2 kb pressure, 800 __+5 ~ H20-saturated, at the NNO oxygen buffer. Rock powders were held in unsealed Pt96-Fe4 capsules, surrounded by Ni + NiO buffer material within sealed Au capsules, run in standard cold seal hydrothermal pressure vessels and were quenched at pressure with compressed air. Analytical procedures were as reported except that glasses were analyzed with a 40 micron diameter electron beam with samples held at liquid N2 temperature on a freezing stage to minimize Na migration Whole-rock

Experimental phases ( + ap, zr)

83S296b a SiO 2 a120 3 FeO d Fe203 MgO CaO TiO z MnO Na20 KzO

61.1 16.4 3.05 2.31 2.55 4.82 0.76 0.10 3.48 3.34 P205 0.25 H20 -]- 0.69 H200.15 COz 0.05 Total 99.05

Run # 1-23 days gl (12) b 73.1(5) c 15.0(3) 1.58(9) 0.19(4) 1.57 (20) 0.19(2) 0.08 (2) 3.74(27) 4.50(14) 0.05 (1)

91.3 e

pl (4) 55.3(4) 28.4(2) 0.36(4)

hbl (5) 44.3(9) 8.92(90) 18.2(9)

0 10.7 (3)

11.0(8) 10.6 (3) 1.63(25) 0.34 (6) 5.17(12) 1.59(17) 0.36(4) 0.56(6) 0.04 (2)

100.3

97.1

Run #10-18 days

bio (5) 37.2(4) 14.5(2) 18.8(3)

mt (1)

13.5(4) 0.05 (3) 2.66(22) 0,16 (4) 0.64(2) 8.42(9)

0.82 0.13 11.9

95.9

92.6

2.04 77.7

gl (7) 73.7(2) 14.7(3) 0.72(5) 0.22(2) 1.34 (3) 0.35(3) 0.10 (2) 4.09(8) 4.74(11)

92.0 e

pl(16) 57.1(7) 26.0(5) 0.22(4) 0.02(2) 8.56 (49) 6.27(29) 0.48(4)

98.7

hbl (5) 45.3(8) 8.11 (98) 12.9(9)

bio (6) 37.4(9) 15.4(7) 13.0(9)

13.4(8) 10.8 (4) 1.76(55) 0.43 (9) 1.49(27) 0.72(9)

15.3(6) 0.11 (3) 4.07(24) 0.19 (4) 0.72(7) 8.48(23)

94.9

94.7

mt(4) 2.8(3) 80.3(4) 0.56(6) 0.08(2) 9.4(3)

93.1

Abbreviations. gl-glass, pl-plagioclase, hbl-hornblende, bio-biotite, rot-magnetite, ap-apatite, zr-zircon Analysis by staff of the Branch of Analytical Chemistry, US Geological Survey (T.W. Sisson, unpublished data), b Number of replicate analyses, c Standard deviation of replicate analyses in units of the least significant digit reported, d FeO in whole rock measured by wet chemistry, all other analyses report all Fe as FeO. e Glass analyses are normalized to total 100% volatile-free with all Fe as FeO. Total reports the original sum of the electron probe analysis

164

Appendix 2. Coexisting natural hornblende rim and glass or matrix compositions Sample Material analy.

IC-B glass 9

IC-B hbP 7

AY-II-V matrix b 13

AY-II-V hbl 5

81-T-116 matrix 13

81-T-116 hbl 6

SiO 2 AlzO 3 FeO* MgO CaO Na 20 K zO TiO 2 MnO P205 Total

76.8 (4) c 12.3(1) 0.79 (4) 0.033 (6) 0.54 (3) 3.69 (11) 4.95 (9) 0.06 (1) 0.03 (1) < 0.01 99.2

46.4 (3) 6.94(13) 16.9 (2) 12.5 (2) 11.1 (1) 1.63 (2) 0.75 (2) 1.36 (9) 0.67 (2)

62.9 (2) 18,4(1) 4.25 (10) 1.78 (2) 4.71 (6) 4.95 (16) 2.03 (3) 0.72 (3) 0.15 (1) 0.14(1) 95.3 d

42.1 (1) 11.7(2) 12.6 (3) 14.0 (1) 11.4 (1) 2.81 (4) 0.47 (3) 2.92 (6) 0.32 (2)

55.8 (1) 18.9(2) 8.03 (10) 3.73 (4) 7.53 (2) 4.04 (18) 1.01 (2) 0.72 (1) 0.14 (1) 0.18(1) 93.8 d

40.9 (3) 13.1(2) 11.6 (2) 14.1 (2) 11.6 (1) 2.49 (6) 0.38 (1) 2.15 (2) 0.17 (2)

98.3

98.3

96.5

Sample origins: IC-B is porphyritic rhyodacite containing high-silica rhyolite glass, collected from Deer Mountain at the south end of the Inyo Volcanic Chain, eastern California. Sample AY-II-V is hornblende high-alumina basalt from Ayarza caldera, Guatemala. A whole-rock analysis of Ayarza hornblende basalt is presented by Peterson and Rose (1985). Sample 81-T-116 is hornblende high-alumina basaltic andesite from Santa Maria volcano, Guatemala. A whole-rock analysis of 81-T-116 is presented in Rose (1987). "Hornblende analyses are exclusively within 15 microns of the rims of euhedral phenocrysts, bMatrix was separated from phenocrysts by hand-picking under ethanol while viewing with a binocular microscope. Matrix separates are estimated to be better than 95% pure. Matrix was fused in Au capsules at 2kb with added doubly-distilled H20 and was quenched to glass with the rapid technique. Quenched glasses were analyzed by electron probe using the same techniques employed for experiments. Analyses are normalized to total to 100% volatile-free with all Fe as FeO. CNumbers in parentheses are one standard deviation of replicate analyses and are in units of the least significant digit reported, dOriginal total of hydrous glass analysis

Acknowledgements. We

thank Fred Frey, Mac Rutherford, and Nobu Shimizu for reviewing an early version of this manuscript, Rosalind Helz and an anonymous referee for constructive journal reviews, and Dean Presnall for an independent review. John Ferry performed very careful editing. Discussions with Tom Juster, Ro Kinzler, Tom Wagner, and Peter Keleman helped to formulate interpretations of the crystallization of hydrous magmas and conversations with Dean Presnall motivated treatment of the "oxygen problem". Roger Burns and D'Arcy Straub kindly performed the M6ssbauer measurement of glass and Bill Rose provided samples of hornblende high-alumina basalts from Central America collected under NSF grants DES 78-01190 and EAR 8205606. This research was supported by NSF grants EAR 8517327 and EAR 8721097. Final manuscript preparation was supported by Texas Advanced Research Program grant 009741-007 (to D.C. Presnall).

References Albarede F, Provost A (1977) Petrological and geochemical massbalance equations: an algorithm for least-square fitting and general error analysis. Computer Geosci 3 : 309-328 Albee AL, Ray L (1970) Correction factors for electron microprobe microanalysis of silicates, oxides, carbonates, phosphates and sulfates. Anal Chem 42:1408 1414 Allen JC, Boettcher AL (1983) The stability of amphibole in andesite and basalt at high pressure. Am Min 68:307-314 Allen JC, Boettcher AL, Marland G (1975) Amphiboles in andesite and basalt: I stability as a function of P-T-fO2. Am Min 60:1069-1085 Anderson AT (1974) Chlorine, sulfur, and water in magmas and oceans. Geol Soc Am Bull 85:1485-1492 Anderson AT (1982) Parental basalts in subduction zones: implications for continental evolution. J Geophys Res 87:7047 7060 Arculus RJ, Wills KJ (1980) The petrology of igneous blocks and inclusions from the Lesser Antilles island arc. J Petrol 21 : 143-168 Arculus RJ, Delong SE, Kay RW, Brooks C, Sun S-S (1977) The alkalic rock suite of Bogoslof Island, eastern Aleutian arc, Alaska. J Geol 85 : 177-186 Baker DR, Eggler D H (1983) Fractionation paths of Atka (Aleu-

tians) high-alumina basalts: constraints from phase relations. J Volc Geotherm Res 18 : 387-404 Baker DR, Eggler DH (1987) Compositions of anhydrous and hydrous melts coexisting with plagioclase, augite, and olivine or lowCa pyroxene from 1 atm to 8 kbar: application to the Aleutian volcanic center of Atka. Am Mineral 72:12-28 Bartels KS, Kinzler RJ, Grove TL (1991) High pressure phase relations of primitive high-alumina basalt from Medicine Lake volcano, northern California. Contrib Miner Petrol 108 : 253-270 Beard JS, Borgia A (1989) Temporal variation of mineralogy and petrology in cognate enclaves at Arenal volcano, Costa Rica. Contrib Mineral Petrol 103 : 110-122 Bence AE, Albee AL (1968) Empirical correction factors for the electron microanalysis of silicates and oxides. J Goel 76 : 382 403 Brophy JC (1986) The Cold Bay volcanic center, Aleutian volcanic arc I. Implications for the origin of hi-alumina arc basalt. Contrib Mineral Petrol 93 : 368-380 Brown GM, Holland JG, Sigurdsson H, Tomblin JF, Arculus RJ (1977) Geochemistry of the Lesser Antilles volcanic island arc. Geochim Cosmochim Acta 41:785 801 Burnham CW (1979) The importance of volatile constituents. In: Yoder HS (ed) The evolution of the igneous rocks. Princeton Univ Press, Princeton NJ:439-482 Burnham CW, Holloway JR, Davis N F (1969) Thermodynamic properties of water to 1000~ and 10000 bars. Geol Soc Am, Spec Pap 132:1-96 Burns RG, Dyar MD (1983) Spectral chemistry of green glassbearing 15426 regolith. Proc Lunar and Planet Sci Conf 14, J Geophys Res 88:B221-B228 Byers FM (1959) Geology of Umnak and Bogoslof Islands Aleutian Islands, Alaska, US Geol Surv Bull 1028-L:267 369 Byers FM (1961) Petrology of three volcanic suites, Umnak and Bogoslof Islands, Aleutian Islands. Geol Soc Am Bull 79:93 128 Carr M J, Rose WI (1987) CENTAM-a data base of Central American volcanic rocks. J Volc Geotherm Res 33 : 239-240 Cawthorn RG (1976) Melting relations in part of the system CaOMgO-AI203-SiO2-Na20-H/O under 5kb pressure. J Petrol 17:44 72 Cawthorn RG, O'Hara MJ (1976) Amphibole fractionation in calealkaline magma series. Am J Sci 276:309-329 Cawthorn RG, Curran EB, Arculus RJ (1973) A petrogenetic model

165 for the origin of the calc-alkaline suite of Grenada, Lesser Antilles. J Petrol 14 : 327 337 Conrad WK, Kay RW (1984) Ultramafic.and mafic inclusions from Adak island: crystallization history and implications for the nature of primary magmas and crustal evolution in the Aleutian arc. J Petrol 25:88-125 Eggler DH (1972) Water-saturated and undersaturated melting relations in a Paricutin andesite and an estimate of water content in the natural magma. Contrib Mineral Petrol 34:261 271 Eggler DH, Burnham CW (1973) Crystallization and fractionation trends in the system andesite-H20-CO2-O/ at pressures to 10 kb. Geol Soc Am Bull 84:2517-2532 Ford CE, Biggar GM, Humphries D J, Wilson G, Dixon D, O'Hara MJ (1972) Role of water in the evolution of the lunar crust; an experimental study of sample 14310; an indication of lunar calc-alkaline volcanism. Proc Third Lunar Sci Conf 1 : 207 229 Gill JB (198t) Orogenic andesites and plate tectonics. Springer, Berlin Heidelberg New York Green TH, Pearson NJ (1985) Experimental determination of REE partition coefficients between amphibole and basaltic to andesitic liquids at high pressure. Geochim Cosmochim Acta 49 : 1465-1468 Grove TL (1981) Use of FePt alloys to eliminate the iron loss problem in 1 atmosphere gas mixing experiments: theoretical and practical considerations. Contrib Mineral Petrol 84:298-304 Grove TL, Baker MB (1984) Phase equilibrium controls on the tholeiitic versus calc-alkaline differentiation trends. J Geophys Res 89 : 3253 3274 Grove TL, Bryan WB (1983) Fractionation of pyroxene-phyric MORB at low pressure: an experimental study. Contrib Mineral Petrol 84:293 309 Grove TL, Juster TC (1989) Experimental investigations of low-Ca pyroxene stability and olivine-pyroxene-liquid equilibria at latin in natural basaltic and andesitic liquids. Contrib Mineral Petrol 103 : 287-305 Grove TL, Kinzler RJ (1986) Petrogenesis of andesites. Ann Rev Earth Planet Sci 14 : 417 454 Grove TL, Gerlach DC, Sando TW (1982) Origin of calc-alkaline series lavas at Medicine Lake volcano by fractionation, assimilation and mixing. Contrib Mineral Petrol 80:160-182 Grove TL, Kinzler R J, Bryan WB (1990) Natural and experimental phase relations of lavas from Seroki volcano. Proc ODP, Sci Results, 106/109, College Station TX, pp 9-17 Grove TL, Kinzler RJ, Bryan WB (1992) Fractionation of midocean ridge basalt. In: Blackman D and Phipps-Morgan J, Proceedings of the 1st RIDGE institute: (in press) Gust DA, Perfit MR (1987) Phase relations of a high-Mg basalt from the Aleutian Island Arc: implications for primary island arc basalts and high-A1 basalts. Contrib Mineral Petrol 97:7-18 Hamilton DL, Burnham CW, Osborn EF (1964) The solubility of water and effects of oxygen fugacity and water content on crystallization in mafic magmas. J Petrol 5:21-39 Hansen M (1958) Constitution of binary alloys. McGraw-Hill, New York Hawthorne FC (1981) Crystal chemistry of the amphiboles. In: Veblen DR (ed) Amphiboles and other hydrous pyriboles-mineralogy. Rev Mineralogy, 9A, Mineral Soc Am Heald EF, Naughton JJ, Barnes IL (1963) The chemistry of volcanic gases. J Geophys Res 68 : 545-557 Helz RT (1973) Phase relations of basalts in their melting range at pH20 = 5 kb as a function of oxygen fugacity. I. Mafic phases. J Petrol 14 : 249-302 Helz RT (1976) Phase relations of basalts in their melting ranges at f, H g O = 5 kb. II. Melt compositions. J Petrol 17 : 139 193 Helz RT (t981) Phase relations and compositions of amphiboles produced in studies of the melting behavior of rocks. In: Veblen DR, Ribbe PH (eds) Amphiboles: petrology and experimental phase relations. Min Soc Am-Reviews in Mineralogy 9B : 279-353 Helz RT (1987) Differentiation behavior of Kilauea Iki lava lake, Kilauea Volcano, Hawaii: an overview of past and current work.

In: Mysen BO (ed) Magmatic processes: physicochemical principles. Spec Pub Geochem Soc 1 : 241-258 Hill R, Roeder P (1974) The crystallization of spinel from basaltic liquid as a function of oxygen fugacity. J Geol 82: 709 729 Holloway JR, Burnham CW (1972) Melting relations of basalt with equilibrium water pressure less than total pressure. J Petrol 13 : 1 - 2 9

Irvine TN, Baragar WRA (1971) A guide to the chemical classification of the common volcanic rocks. Can J Earth Sci 8 : 523-548 Johannes W (1978) Melting of plagioclase in the system Ab-AnH 2 0 and Qz-Ab-An-HaO at PH20 = 5 kbar, an equilibrium problem. Contrib Mineral Petrol 66:295-303 Johnson MC, Rutherford ML (1989) Experimentally determined conditions in the Fish Canyon tuff, Colorado, magma chamber. J Petrol 30:711-737 Johnston AD (1986) Anhydrous P-T phase relations of near-primary high-alumina basalt from the South Sandwich Islands. Contrib Mineral Petrol 92:368 382 Juster TC, Grove TL, Perfit MR (1989) Experimental constraints on the generation of Fe-Ti basalts, andesites, and rhyodacites at the Galapagos Spreading Center, 85~ and 95~ J Geophys Res 94 : 9251-9274 Katsura T, Nagashima S (1974) Solubility of sulfur in some magmas at 1 atmosphere. Geochim Cosmochim Acta 38:517 531 Kay SM, Kay RW (1985) Aleutian tholeiitic and calc-alkaline magma series; I. The mafic phenocrysts. Contrib Mineral Petrol 90 : 276-290 Kennedy GC (1955) Some aspects of the role of water in rock melts. In: Crust of the Earth-a symposium. Geol Soc Am Spec Paper 62, pp 489-503 Kilinic A, Carmichael ISE, Rivers ML, Sack RO (1983) The ferricferrous ration of natural silicate liquids equilibrated in air. Contrib Mineral Petrol 83 : 136-140 Kinzler RJ, Grove TL (1992) Primary magmas of mid-ocean ridge basalts 1. Experiments and methods. J Geophys Res 97 : 6885-6906 Kress VC, Carmichael ISE (1988) Stoichiometry of the iron oxidation reaction in silicate melts. Am Min 78 : 1267-1274 Kuno H (1950) Petrology of Hakone volcano and the adjacent areas, Japan. Geol Soc Am Bull 61:957-1020 Kushiro I (1969) The system forsterite-diopside-silica with and without water at high pressures. Am J Sci 267A:269-294 Labotka TC (1991) Chemical and physical properties of fluids. In: Kerrick DM (ed) Contact metamorphism. Min Soc Am-Reviews in Mineralogy 26 : 43-104 Marsh BD (1976) Some Aleutian andesites: their nature and source. J Geol 84 : 27-45 Marsh BD (1982) The Aleutians. In: Thorpe RS (ed) Andesites. Wiley, New York, pp 99-114 Meen JK (1987) Formation of shoshonites from calc-alkaline basalt magmas: geochemical and experimental constraints from the type locality. Contrib Mineral Petrol 97:333-351 Miyashiro A (1974) Volcanic rock series in island arcs and active continental margins. Am J Sci 274 : 321-355 Myers JD, Marsh BD, Sinha AK (t986) Geochemical and strontium isotopic characteristics of parental Aleutian arc magmas: evidence from the basaltic lavas of Atka. Contrib Mineral Petrol 94:1 11 Naney MT (1983) Phase equilibria of rock-forming ferromagnesian silicates in granitic systems. Am J Sci 283 : 993-1033 Nicholls IA, Harris KL (1980) Experimental rare earth element partition coefficients for garnet, clinopyroxene, and amphibole coexisting with andesitic and basaltic liquids. Geochim Cosmochim Acta 44 : 287 308 Nye CJ, Reid MR (1986) Geochemistry of primary and least fractionated lavas from Okmok volcano, central Aleutians: implications for arc magma genesis. J Geophys Res 91:271-287 Osborn EF (1959) Role of oxygen pressure in the crystallization and differentiation of basaltic magma. Am J Sci 257 : 609 647 Osborn EF (1963) Some experimental investigations bearing on the origin of igneous magmas of the earth's crust. Estud Geol 19 : 1-7

166 Osborn EF (1969) Experimental aspects of calc-alkaline differentiation. In: McBirney AR (ed) Proc Andesite Conf, Oregon Dept of Geol and Mineral Industries, Bull 65 : 33-42 Peterson PS, Rose WI (1985) Explosive eruptions of the Ayarza calderas, southeastern Guatemala. J Volc Geotherm Res 25 : 289-307 Presnall DC (1966) The join forsterite-diopside-iron oxide and its bearing on the crystallization of basaltic and ultramafic magmas. Am J Sci 264:753-809 Presnall DC, Dixon SA, Dixon JR, O'Donnell TH, Brenner NL, Schrock RL, Dyeus DW (1978) Liquidus phase relations on the join diopside-forsterite-anorthite from 1 atm to 20 kbar: their bearing on the generation and crystallization of basaltic magma. Contrib Mineral Petrol 66:203 220 Roeder PL (1974) Activity of iron and olivine solubility in basaltic liquids. Earth Planet Sci Lett 23:397 410 Roeder PL, Osborn EF (1966) Experimental data for the system MgO-FeO-Fe203-CaA12SizOs-SiO 2 and their petrological implications. Am J Sci 264:428-480 Rose WI (1987) Santa Maria, Guatemala: bimodal soda'rich calcalkalic stratovolcano. J Volc Geotherm Res 33 : 109 129 Rutherford MJ, Sigurdsson H, Carey S, Davis A (1985) The May 18, 1980, eruption of Mount St. Helens 1. Melt composition and experimental phase equilibria. J Geophys Res 90:2929-2947 Sekine T, Wyllie PJ (1983) Phase relations in the join grossularitepyrope-7.5 percent H20 at 30 Kb. Am J Sci 283:435 453 Shaw HR, Wones DR (1964) Fugacity coefficients for hydrogen gas between 0~ and 1000~ for pressures to 3000 arm. Am J Sci 262:918 929 Sisson TW, Grove TL (1993) Temperatures and H20 contents of low-MgO high-alumina basalts. Contrib Mineral Petrol 113:167-184 Spulber SD, Rutherford MJ (1983) The origin of rhyolite and plagiogranite in oceanic crust: an experimental study. J Petrol 24 : 1-25 Stern CR, Wyllie PJ (1975) Effect of iron absorption by noble-metal

capsules on phase boundaries in rock-melting experiments at 30 kilobars. Am Min 60:681-689 Stolper E, Holloway JR (1988) Experimental determination of the solubility of carbon dioxide in molten basalt at low pressure. Earth Planet Sci Lett 87:397-408 Tormey DR, Grove TL, Bryan WB (1987) Experimental petrology of normal MORB near the Kane Fracture Zone: 22 ~ ~ midAtlantic ridge. Contrib Mineral Petrol 96:121 139 Wager LR, Deer WA (1939) The petrology of the Skaergaard intrusion, Kangerlassuaq, East Greenland, Medd Groenl 105(4): 1-352 Wagner TP, Grove TL, Donnelly-Nolan JM (1991) Water saturated melting of Lake high-alumina basalts from Medicine Lake Volcano, northern California (abstract). Trans Am Geophys Union 72 : 548 Walker D, Shibata T, DeLong SE (1979) Abyssal tholeiite from Oceanographer Fracture Zone, II: phase equilibria and mixing. Contrib Mineral Petrol 70: 111-125 Yoder HS Jr (1965) Diopside-anorthite-water at five and ten kilobars and its bearing on explosive volcanism. Yearb Carnegie Inst Wash 64 : 82-89 Yoder HS Jr (1969a) Calcalkalic andesites: Experimental evidence bearing on the origin of their assumed characteristics. In: McBirney AR (ed) Proc Andesite Conf, Oregon Dept of Geol and Mineral Industries, Bull 65:77-89 Yoder HS Jr (1969b) The join diopside-pyrope-H20 at 10 Kb: its bearing on the melting of peridotite, the ACF metamorphic facies, and the gedrite-hornblende miscibility gap. Yearb Carnegie Inst Wash 69 : 176-181 Yoder HS Jr, Tilley CE (1962) Origin of basalt magmas: an experimental study of natural and synthetic rock systems. J Petrol 3 : 342-532

Editorial responsibility: J. Ferry