1999gc000032p04 - SCHIggg00.pdf

Primitiv

eCaO-rich

,silica-undersaturate

dmelt

sin

islan

arcs

:Evidenc

efortheinvolvemen

tof

clinopyroxene-ric

hlithologie

sinthe

petrogenesi

sofarcmagmas

P.Schiano

Departmen

tde sScience

sdelaTerre(UMR6524),Universit

eÂBlaise-Pascal

,5rue Kessle

r,63038Clermont-Ferrand

cedex, France (schiano@opgc.univ-bpclermont.fr)

Divisio

nofGeologica

landPlanetar

yScience

s(170-25)

,Californi

aInstituteofTechnolog

y,Pasadena

,California

91125

J.M.Eiler

Divisio

nofGeologica

landPlanetar

yScience

(

170-25)

,Californi

aInstituteofTechnolog

y,Pasadena

,California

91125 (eiler@gps.caltech.edu)

I.D.Hutcheon

Analytica

landNuclearChemistr

yDivision

,Lawrenc

eLivermor

eNationa

lLaborator

y,P.O.Box808,Livermore,

California 94551 (hutcheon@llnl.gov)

E.M.Stolper

Divisio

nofGeologica

landPlanetar

yScience

(

170-25)

,Californi

aInstituteofTechnolog

y,Pasadena

,California

91125 (ems@gps.caltech.edu)

[1]

Abstract

Onthebasisofth estudyofolivine-hoste

dmeltinclusion

sinacalc-alka

linebasaltfrom

Bata nIsland

(

Philippines)wedefin eadistinctiv

etyp eofprimitive

,nepheline-no

rmativeislan darc

magmacharacterizedbyunusuall

yhighCaOcontent

(

upto19.0wt%)tha tcannotbesimplyexplained

bymeltingofth emetasomatize

dperidotiti

cmantl ewedgeabovesubductin

goceaniclithosphere

.CaO-

richmeltinclusion

swit hthesecharacteristicsar epreserve

dinFo

85±90

olivine ,andcompositional

variation

samongtheinclusion

sareinterpre

tedtoreflec tmixingbetweenmelt ssuchasthosefoundinthe

mostCaO-ric

hinclusion

(

presentinFo

olivine )andmeltssimilartoprimitiv

e``normal

''islandarc

magma

(

trappedinFo

olivine)

.Compilatio

nofprimi tiveislan darcmagma

sfromtheliteratureshows

thatwholerock sandolivine-hoste

dmeltinclusion

swit hCaOcontent

s>13wt%arefoundinmanyarc

volcanoe

sfromallovertheworldinadditiontoBatan .Thes einclusion

soccurinlavasrangin gfrom

CaO-ric

hankaramitestobasalticandesite

swit hlow-CaOcontent

(

i.e.,<13wt%) .Theglobally

occurrin

gCaO-ric

hinclusion

sandwholerock scompris

eagroupthatalthoug

hdefine donth ebasisof

theirCaOcontent

siscompositiona

llydistinctiv

ewhencompare

dtoislandarclavasthathavelowerCaO

contents

;forexample

,the yhavelowerFeOatagivenSiO

contenttha nmostarclavas ,andtheyareall

nephelin

enormative

,wit hnormativ

enephelin

econtent

spositivel

ycorrelate

dwithCaOcontents.

Variation

sinCaOcontentandnormativ

ecomposition

sofexperimenta

lpartialmeltsoflherzolit

erelated

tochange sofpressure

,temperature

,andsourcecompositio

nsuggesttha ttherearenocondition

sunder

whic hpartialmeltingofperidotit

ecangenerat

emelt shavingCaOcontent

sandotherproperties

comparabl

etothos eobserve

dfo rtheprimitive

,CaO-ric

harc-derivedmelt sidentifie

dhere.Although

meltin gofperidotit

eathig hpressur einthepresenc

eofCO

canproduceCaO-rich

,silica-poorliquids,

bytheAmerica

nGeophysica

lUnion

ochemistry

Geophysics

Geosystems

Published by

AGU and the Geochemical Society

AN ELECT

RONIC JOUR

NAL OF

THE EA

RTH SCIENCES

Article

olume 1

May 30, 2000

per

umber

1999GC0000

32

ISSN:

1525-2027

Geochemistry

Geophysics

Geosystems

weconsid

eritu

nlikel ytha tthi sis

espo

nsiblefo rproducin

gtheCaO-rich

,silica-undersaturate

dmelts

considere

dinthisstudybecausetherearesignifican

tdifference

sinnearlyallothe rcompositional

characteristic

sbetwee ntheCaO-ric

har cmagma

sandmeltsknownorthough ttobeproduce

dbymelting

ofcarbonate

dperidotite

.Modelmajorelementcomposition

sofpartialmeltsofclinopyroxene-rich

lithologies

(

mantl epyroxenites

,lowercrusta lpyroxenites

,andeclogite

s)calculatedusingtheME LTS

algorith

msuggesttha tthemostCaO-rich

,nepheline-norm

ativemeltinclusion

sandwholerocks

identifie

dherecouldrepres entintermediat

etohig hdegree

(

10±40wt%)partialmeltsofpyroxenites

atlowercrusta ltouppermantlepressures

.Suchahypothesi

sissupporte

dbythecompariso

nbetweenthe

traceelemen tcomposition

sofmodelpyroxenit

esourcesofth eBatanCaO-ric

hmeltinclusion

sand

naturall

yoccurrin

gpyroxenites

.Themostlikel ysourceoftheprimi tiveCaO-rich

,silica-undersatura

ted

arcmelt sidentifiedhereislowe rcrusta landshallowuppermantlepyroxene-ric

hcumulate

sfromarc

environment

sbecausethesecumulate

shaveCaOconcentration

sattheupperendoftherang eobserved

formantl epyroxenites

.Theyar etherefor

emorelikel ytoyiel dpartialmelt swiththerestricte

drang eof

remarkabl

yhig hCaOcontent

softhemostCaO-ric

hinclusion

sandwholerock sidentifiedhere.

Moreove

r,thes ecumulate

softencontainamphibole

,whichwouldlowe rtheirsolidu stempera

tures

relativetoth eanhydrou

spyroxenit

eequivalent

stovaluesmoreconsisten

twiththoseexpecte

dindeep

crusta lorshallowsubarcenvironments.

Keywords

Meltinclusions

;ar clavas.

Inde

xterms

Chemica

levolution

;geochemica

lcycles;majorelementcomposition.

Receive

Novembe

r16,1999;

Revise

March9,2000;

Accepte

March15,2000;

Published

May 30, 2000.

Schiano, P., J. M. Eiler, I. D. Hutcheon, and E. M. Stolper, 2000. Primitive CaO-rich, silica-undersaturated melts

in island arcs: Evidence for the involvement of clinopyroxene-rich lithologies in the petrogenesis of arc magmas,

Geochem.

Geophys.

Geosyst

., vol. 1, Paper

number

1999GC000032

[15,276

words,

11 figures,

2 tables].

1. Int

roduction

[2] K

Mostislandarclava sarebelievedtobe

derivedfrommagma

sformedbypartialmelt-

ingofth emantleoverlyin

gsubductin

gocea-

niclithosphere

,wheremeltingisfluxedby

metasomati

cenrichmen

tinwater

(

andother

elements

)carriedbyaqueou

sfluid sand/or

silicatemelt sreleasedfromth esubducted

oceaniccrust[e.g.,

Nicholl

sandRingwoo

1973].Themantlesourcesonwhichthis

coupledenrichment-meltin

gproces soperates

arethough ttobeper idotitessimi lartoor

somewha

tmoredepletedthanthesourcesof

mid-ocea

nridgebasalts

(

MORBs

)[

Tatsum

iet

.,1986;

Davidso

,1987;

Nohd

aandWas-

serbu

,1981;

Ryerso

nandWatso

,1987].

Thetraceelementcharacteristic

sofar clavas

e.g. ,thei rhi ghconcentrati onsoflarge- ion

lithophil

eelement

(

LILE)andthei rrelatively

lowconcentration

sofhigh-fiel

dstrengthele-

ments

(

HFSE))an dtheirmajorelementcom-

position

sareinfluence

dbythemetasomatism,

andhighthermodynami

cactivitie

sofH

stronglyaffectthecomposition

sofprimary

magma

sandtheirsubsequen

tdifferentiation

sequence

s[see

Stolpe

randNewma

,1994,

andreference

stherein].

[3]K

Althoug

hcouple

dmetasomatis

mand

fluxedmeltin gofperidotite

sinth emantle

wedgeisbelievedtobethedominan

tprocess

bywhichprimaryar cmagma

sform,somearc

magma

shavealsobeenpropose

dtoformby

May 30, 2000.

SCHIANO

L.:

CaO-RICH MELTS IN ISLAND ARC

1999GC000032

Geochemist

Geophysics

Geosystems

onperidotitic sources: for example,

partial melting of subducted crust [

Defant and

Drummond

, 1990;

Drummond and Defant

1990;

Kay

,1978],meltingatdeepcrustal

levels [

Pichler and Zeil

, 1972], and melting

of an ocean island basalt-type source [

Ito and

Stern

, 1985] have all been proposed. In this

article, we document a distinctive type of

primitive, nepheline-normative island arc mag-

ma characterized by unusually high CaO con-

tents

(

up to 19 wt %) that cannot be simply

reconciled with the standard model of arc

petrogenesis described above. Although arc-

derived samples with these characteristics

have been previously reported [

Metrich et

., 1999;

Sisson and Bronto

, 1998;

Della-

Pasqua and Varne

, 1997;

Gioncada et al

1998;

Metrich and Clocchiatti

, 1996;

Foden

and Varne

, 1980;

Foden

, 1983;

Kennedy et

., 1990;

Carr and Rose

, 1984;

Thirwall and

Graham

, 1984;

Shimizu and Arculus

, 1975;

Arculus

, 1976], they have not been recognized

as a coherent, globally occurring magma type

or as distinct from CaO-rich magmas found in

oceanic island [

Tronnes

, 1990], mid-ocean

ridge [

Kamenetsky et al

., 1998;

Sours-Page

et al

., 1999], and back arc [

Kamenetsky et al

1997] settings. After documenting the exis-

tence of this distinctive magma type, empha-

sizing in particular a well-defined group from

Batan Island

(

northernmost Philippines, Lu-

zon-Taiwan arc), but also showing that such

magmas occur in arcs from around the world,

we examine the compositions of CaO-rich,

silica-undersaturated magmas in subduction

zone settings with the following objectives in

mind:

(

1) to provide a compilation of the

occurrences and compositions of such mag-

mas;

(

2) to examine their relationship to more

typical magmas in subduction zone settings;

(

3) to estimate plausible chemical and miner-

alogical characteristics of their sources; and

(

4) to evaluate the implications of their ex-

istence for models of melt generation in sub-

duction zones.

2. Definition of a Distinctive

CaO-Rich, Silica-Undersaturated

Magma Type From Subduction

Zone Settings

2.1. CaO-Rich, Silica-Undersaturated

Melt Inclusions in Olivine Phenocrysts in

Lavas From Batan Island

(

Northernmost

Philippines, Luzon-Taiwan Arc)

2.1.1. Sample description, analytical

techniques, and results

[4]

The geologic setting of Batan Island and the

petrography of its principal rock types

(

basalts,

basaltic andesites, and andesites) are described

Richard et al

. [1986]. Plagioclase, olivine,

and clinopyroxene dominate the phenocryst

assemblages of the basalts, while basic and

acid andesites contain phenocrysts of clinopyr-

oxene, hornblende, plagioclase, and titanomag-

netite. Sample B45 from Mount Iraya, the

youngest

(

2.32 to <0.1 myr) volcano on Batan

Island [

Richard et al

., 1986], is a high-MgO

(

10.8 wt %) calc-alkaline basalt [

Metrich et al

1999;

Richard et al

., 1986]. Compositions of

primary melt inclusions in olivine phenocrysts

(

90±75

) from this sample were determined on

glasses produced after heating each inclusion

up to its homogenization temperature

(

i.e., the

temperature at which the inclusion contents

(

gas, glass, and crystals) homogenized visually

to a uniform melt phase). Heating experiments

were performed with a high-temperature, opti-

cal heating-stage apparatus [

Sobolev et al

1980] that allowed visual monitoring of indi-

vidual melt inclusions during heating. Homo-

genization temperatures were 1220 20

8

C, and

effective times of quenching were <1 s. Com-

parison between the compositions of homoge-

nized and unheated glass inclusions trapped in

olivine phenocrysts with similar forsterite con-

tent shows that the amount of host olivine

needed to be added to the unheated glass

inclusions to get the heated ones is relatively

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

mall

(

%in

ight). Consequently, the

effect of the heating procedure on the concen-

trations of elements other than MgO

(

and to a

lesser extent FeO and SiO

) is relatively small,

and therefore the main conclusions of the paper

would be unchanged if unheated inclusions

were considered. It should be noted that this

statement applies not only to the Batan samples

but also to those in the compilation that will be

presented later in this article.

Table 1.

Representative Major and Trace Element Compositions of Homogenized Melt Inclusions in

Olivine in a Calc-Alkaline Basalt

(

B45) From Batan Island

(

8

N±121

8

E, Luzon-Taiwan Arc)

Sample Identification

B7-1 B7-2 B2-1

B1-2 B1-1 B5-1 B8-1 B6-1

SiO

, wt %

44.02 45.05 44.00 45.61 44.10 44.04 45.89 49.42 54.73

TiO

, wt %

0.70

0.64

0.79

0.67

0.94

0.76

0.74

0.91

0.98

, wt % 16.01 16.87 14.92 16.58 16.13 16.02 16.43 15.39 14.44

FeO*, wt %

7.31

6.80

8.86

8.02

8.82

8.85 12.27 11.11

8.84

MnO, wt %

0.12

0.10

0.15

0.14

0.15

0.20

0.18

0.19

MgO, wt %

8.30

8.27

8.96

8.36

7.67

7.72

6.71

7.17

5.63

CaO, wt %

18.54 17.08 17.20 16.27 17.20 17.06 11.57 10.22

7.48

O, wt %

2.58

2.56

2.43

2.53

2.29

2.60

2.75

3.05

3.29

O, wt %

0.93

1.01

0.86

1.01

0.90

1.01

1.40

1.43

2.17

, wt %

0.05

0.04

0.05

0.08

0.04

0.03

0.08

0.03

SUM

98.57 98.42 98.20 99.26 98.23 98.24 98.02 98.95 97.77

HOST

Ba, ppm

219

214

185

237

136

245

215

323

423

Th, ppm

2.12

2.18

1.57

2.32

1.04

2.87

2.06

3.74

6.75

U, ppm

0.63

0.57

0.38

0.49

0.31

0.67

0.39

0.91

1.37

Nb, ppm

2.20

2.26

4.40

2.35

2.09

2.83

2.68

4.00

6.17

Sr, ppm

537

539

498

490

398

571

536

663

561

Zr, ppm

56.1

60.7

52.6

58.6

47.0

66.9

62.7

90.2 124.7

Y, ppm

17.3

17.0

19.9

14.1

13.4

17.1

17.9

26.7

25.2

Sc, ppm

N.D. N.D. 26

La, ppm

9.21

9.39

8.31

9.10

7.45 10.52

9.11 13.13 18.40

Ce, ppm

21.4

20.9

18.4

21.2

15.7

24.9

19.8

31.5

43.3

Nd, ppm

14.3

14.2

12.3

13.6

10.7

19.3

13.2

22.4

26.2

Sm, ppm

3.15

2.99

2.61

3.00

1.95

3.49

2.86

4.33

4.99

Eu, ppm

0.95

0.92

0.67

0.85

0.57

0.99

0.83

1.30

1.26

Dy, ppm

2.81

2.95

2.47

2.58

1.94

3.01

2.46

4.56

Er, ppm

2.09

2.01

1.85

1.82

1.21

2.07

1.93

3.42

3.32

Yb, ppm

2.34

2.26

2.09

1.39

2.31

2.25

3.77

3.37

Major element compositions were measured with the Caltech 5-spectrometer JEOL 733 microprobe using an accelerating voltage of

15Kv, a sample current of 10nA anda defocused beam

(

size

m). Data were reduced using a modified ZAF procedure

(

CITZAF

[

Armstrong

, 1988]). The reproducibility and accuracy of the Na measurements were tested on glass standards. The abundances of Sr, Y,

Zr, Nb, Ba, rare earth elements, Th, and U were determined with the Lawrence Livermore National Laboratory

(

LLNL) ion microprobe, a

modified Cameca IMS-3f instrument, using a 17 keV,

primary ion beam, focused into a

20-

m-diameter spot. Positive secondary

ions were extracted and accelerated, nominally to 4500 V. A field aperture inserted in the sample image plane allowed only ions from the

central 30-

m-diameter area of the imaged field to enter the mass spectrometer. Isobaric molecular interferences were minimized by

energy filtering, using a 40-eV window and offsetting the accelerating voltage by 100 V from the voltage at which the energy distribution

dropped to 10% of its maximum value. Trace element concentrations were determined from

Ca-normalized ion intensities

using sensitivity factors established from analyses of a suite of silicate glass and mineral standards. Each analysis consisted of 8±10

cycles over the masses of interest. On the basis of analyses of NIST glass standards

(

NBS-610, -612, and -614, using the concentrations

given by

Pearce et al

. [1997], the accuracy varies between 3 and 10% for Sr, Y, Zr, Ba, La, Ce, Nd, Er, and Yb, between 10 and

15% for Nb, Th, Sm, and Dy, and between 15 and 25% for U and Eu. Sc analyses were carried out at a mass resolving power of 4000

and zero energy offset, using a 2-nA

primary ion beam. The

Ca-normalized ion intensities were converted to concentrations using

sensitivity factors established from analyses of NIST standard glass SRM 612.

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

100

0.0

0.4

0.8

1.2

1.6

200

300

400

Normative

nepheline(%)

Normative

hypersthene(%)

(ppm)

CaO (wt%)

Ba (ppm)

Sc (ppm)

trend B

CaO (wt%)

FeO* (wt%)

MgO (wt%)

Host forsterite (%)

SiO

(wt%)

trend A

LEGEND

Batan samples

< 13 % CaO melt inclusions (i.e., trend A)

> 13 % CaO melt inclusions (i.e., trend B)

< 13 % CaO whole rocks (i.e., trend A)

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

[

]

enize

d inclusions were analyzed for

major and minor elements using the electron

microprobe at Caltech and for trace elements

(

Sc, Sr, Y, Zr, Nb, Ba, rare earth elements

(REE), Th, and U) using the ion microprobe

at the Lawrence Livermore National Labora-

tory

(

Table 1; see footnote for analytical pro-

cedures). The inclusions range in composition

from nepheline-normative basalt to quartz-nor-

mative dacite; major element variations in these

glasses are illustrated in Figure 1. The CaO and

MgO contents decrease from 18.5 to 5.0 wt %

and 9.5 to 1.3 wt %, respectively, as SiO

increases from 44.3 to 63.7 wt %. Variations

in the compositions of melt inclusions are well

correlated with the Mg# of their host olivines,

with the most SiO

-poor and CaO- and MgO-

rich inclusions trapped in the most magnesian

olivines

(

) and the most SiO

-rich and

CaO- and MgO-poor inclusions trapped in the

least magnesian olivines

(

[6]

Trace element compositions of olivine-

hosted melt inclusions are also well correlated

with major element composition: for example,

Sc content decreases from 65 to 20 ppm and

incompatible trace elements increase regularly

as CaO decreases

(

Figures 1f±1h). The mid-

ocean ridge basalt

(

MORB)-normalized trace

element patterns of the inclusions

(

Figure 2a)

aretypicalofarcmagmas

(

i.e.,enrichedinLILE

and depleted in HFSE and in high REE

(

HREE)

relative to MORB; enriched in low REE

(

LREE) relative to HREE); these trace element

features are thought to reflect melting of mantle

sources depleted with respect to the sources of

MORBthathavebeenmetasomatizedbyaslab-

derived, H

O-rich component [

Stolper and

Newman

, 1994]. Note that although the incom-

patible element concentration variations are

correlated with major element composition,

the La/Yb ratio does not increase as the CaO

content decreases

(

Figure 2a).

[7]

Although in most variation diagrams the

melt inclusions and previously analyzed whole

rocks from Batan [

Richard et al

., 1986;

Maury

et al

., 1998;

McDermott et al

., 1993] define a

single trend, with the CaO-rich inclusions fall-

ing on a continuous extension of the trend

defined by the inclusions with lower CaO con-

tents, Figure 1d shows that this is not the case

for FeO: that is, the CaO-rich inclusions define

a population offset toward lower FeO contents

from the trend of whole rocks and lower-CaO

inclusions. For the purposes of later discussion,

we refer to the trend defined by the Batan whole

rocks and lower-CaO inclusions as trend A and

the higher CaO, lower FeO trend as trend B.

Note that inclusions from trends A and B do not

occur in the same olivine phenocrysts. The

distinction between trends A and B is also

illustrated in Figure 3, in which the composi-

tions of the melt inclusions and Batan whole

rocks were first recalculated to equivalent CaO-

MgO-Al

-SiO

(

C-M-A-S) [

O'Hara

, 1968;

Basaltic Volcanism Study Project

, 1981] and

then projected from alumina onto the S-C-M

plane. Also shown on this diagram are experi-

mentally determined trends for olivine + clin-

Figure 1.

CaO variation diagrams comparing melt inclusions

(

circles) in olivine phenocrysts from B45

basalt from Batan Island

(

Luzon-Taiwan arc)

(

this study and

Metrich et al

. [1999]) with Batan whole rocks

(

triangles) [

Richard et al

., 1986;

Maury et al

., 1998;

McDermott et al

., 1993]. Trend A

(

open symbols)

comprises low-CaO melt inclusions and all Batan whole rocks and is interpreted as due to progressive

fractionation of olivine + clinopyroxene + amphibole

(

see Figure 3); trend B

(

dotted circles) comprises the

CaO-rich inclusions and is interpreted as due to magma mixing between liquids similar to the most CaO-rich

inclusions in trend B

(

i.e., in the Fo

olivines) and melts near the most primitive end of trend A. Normative

compositions are calculated assuming Fe

=0.15

P

Fe. Half-solid circles and triangles in Figure 1e are

quartz normative samples.

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

a (p

pm)

O = 17.2 wt%

CaO = 17.1 wt%

CaO = 16.3 wt%

CaO = 11.6 wt%

CaO = 10.2 wt%

CaO = 7.5 wt%

Batan melt inclusions

}

trend A

trend B

10000

1000

100

1000

Zr Eu

0.1

100

< 13 % CaO melt inclusions (i.e., trend A)

> 13 % CaO melt inclusions (i.e., trend B)

Oceanic and continental nephelinites

Oceanic and continental melilitites

Carbonatites

Nephelinites from SW Japan

Figure 2.

(a) Normal mid-ocean ridge basalt (NMORB)-normalized diagram showing trace element data

for melt inclusions in olivine phenocrysts in B45 basalt from Batan Island. The order of elements is based on

the incompatibility sequence during MORB petrogenesis [

Hofmann

, 1988]. Compositions are normalized to

the average NMORB composition of

Hofmann

[1988]. (b) Plot of La/Yb ratio versus La concentration

comparing data for Batan melt inclusions with data for primitive melilitites [

Wilson et al

., 1995], nephelinites

from oceanic and continental regions [

Wedephol et al

., 1994;

Hoernle and Schmincke

, 1993], nephelinites

from SW Japan [

Tatsumi et al

., 1999], and carbonatites from Africa, Australia, Europe, and North and South

America [

Nelson et al

., 1988]. Symbols are given in the accompanying legend, and the distinction between

trend A and trend B for the Batan samples is the same as in Figure 1.

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

roxe

e+p

lagioclase-saturated liquids at 1

atm and olivine + clinopyroxene + amphibole-

saturated liquids at 2 kbar

(

O saturated)

[

Tormey et al

., 1987;

Grove and Bryan

, 1983;

Walker et al

., 1979;

Grove et al

., 1982;

Caw-

thorn and O'Hara

, 1976;

Sisson and Grove

1993] and the trend of whole rocks from the

Tonga arc

(

SW Pacific), a typical tholeiitic arc

series [

Ewart et al

., 1973, 1977;

Gill

, 1981].

2.1.2. Compositional evolution of Batan melt

inclusions

[8]

The observation that the most CaO-rich and

SiO

-poor melts are trapped in the most for-

LEGEND

Batan samples

> 13 % CaO melt inclusions (i.e., trend B)

<13 % CaO melt inclusions (i.e., trend A)

<13 % CaO whole rocks (i.e., trend A)

Tonga arc

<13 % CaO tholeiites

Ol-CPx-Pl

cotectic trend

Ol-Cpx-Amp

cotectic trend

olivine

diopside

amphibole

plagioclase

trend A

trend B

Projection from Al

wt.%

Tonga

wollastonite

enstatite

Figure 3.

Compositions of Batan melt inclusions

(

circles) and whole rocks

(

triangles) recalculated to

equivalent CaO-MgO-Al

-SiO

(

C-M-A-S) and then projected from alumina onto the S-C-M plane. The

distinction between trend A

(

open symbols) and trend B

(

dotted circles) for the Batan samples is the same as

in Figure 1. Also shown on this diagram are experimentally determined trends for olivine-clinopyroxene-

plagioclase-saturated melts at 1 atm and olivine-clinopyroxene-amphibole-saturated liquids at 2 kbar

(

O-

saturated) [

Tormey et al

., 1987;

Grove and Bryan

, 1983;

Walker et al

., 1979;

Grove et al

., 1982;

Cawthorn

and O'Hara

, 1976;

Sisson and Grove

, 1993] and the trend of whole rocks

(

shaded triangles) from the Tonga

arc

(

SW Pacific), a typical tholeiitic arc series [

Ewart et al

., 1973, 1977;

Gill

, 1981]. Plagioclase and

amphibole compositions are from [

Tormey et al

., 1987;

Grove and Bryan

, 1983;

Walker et al

., 1979;

Grove et

., 1982;

Cawthorn and O'Hara

, 1976;

Sisson and Grove

, 1993]. Analyses have been recalculated as

CMAS components by following a modified version [

Basaltic Volcanism Study Project

, 1981] of the

procedure used by

O'Hara

[1968]. The procedure is as follows: C =

(

CaO

/3 + K

O + 2Na

O)56.08;

(

+Cr

+Fe

+Na

O)101.96; M =

(

MgO + FeO + MnO)40.30; S =

(

SiO

2Na

O)60.08. All oxides are in molecular proportions.

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

ritic oliv

ne crystals

(

)

ca-undersaturated melts are

relatively primitive. In this section, we evaluate

the origin of the compositional variation from

this primitive end of the array of melt inclusion

compositions to the CaO-poor, SiO

-rich com-

positions contained in the least forsteritic oli-

vines.

[9]

One possible explanation for the relation-

ship between melt inclusion composition and

the Mg# of the host olivine is that composi-

tional variations in melt inclusions represent a

trend of fractional crystallization. However,

although the correlated decreases in the Mg# of

thehostolivinesandtheMgOandCaOcontents

of the inclusions would be consistent with frac-

tional crystallization of a clinopyroxene oli-

vine-bearing assemblage, such a fractionating

assemblage

(

assuming the clinopyroxene to be

compositionally similar to the diopside, augite,

or salite phenocrysts in Batan lavas [

Richard et

., 1986]) could not readily account for the

decreaseinnormativenephelinewithdecreasing

CaO content shown in Figure 1e

(

i.e., crystal-

lization of such an assemblage from nepheline-

normative liquids typically results in increasing

rather than decreasing normative nepheline

[

Sack et al

., 1987]). Likewise, although frac-

tional crystallization of assemblages dominated

by nepheline-normative minerals such as am-

phibole, jadeite, or acmite could, in principle,

decrease the normative nepheline contents of

residual liquids, it would not decrease their

CaO contents.

[10]

Figures 1b±1d show that trend A

(

defined

by melt inclusions) is typical of the composi-

tional trend of whole rock analyses of lavas

from Batan

(

shown as triangles in Figure 1).

As indicated above, Figure 1d also demon-

strates the important point that trend B, de-

fined by the CaO-rich inclusions, is not

simply an extension or continuation of the

compositional trends of arc lavas from Batan;

this point is also made in Figure 3, where the

CaO-poor melt inclusions and whole rocks

defining trend A follow the olivine-clinopyr-

oxene-amphibole cotectic but the trend B melt

inclusions fall off this cotectic. The correspon-

dence of trend A to the olivine-clinopyroxene-

amphibole cotectic suggests that the composi-

tional variations of Batan whole rocks could

primarily reflect relatively low pressure crys-

tallization of an amphibole-bearing assem-

blage from low-CaO, high-FeO parental

melts comparable to those with

12 wt %

CaO trapped in Fo

olivine. This consistency

of major element variations in Batan lavas

with fractional crystallization of an amphi-

bole-bearing assemblage agrees with the con-

clusions of several previous studies [

Richard

et al., 1986; Maury et al., 1988; McDermott et

., 1993]. Notethatthetholeiitictrendofwhole

rocklavasfromtheTongaarcshowninFigure3

contrasts with the Batan trend and suggests the

Tonga trend is produced by low-pressure frac-

tionation of an olivine-clinopyroxene-plagio-

clase assemblage. This hypothesis is supported

by mass balance calculations [see

Gill

, 1981,

page 272] that indicate that fractionation of an

olivine-clinopyroxene-plagioclase assemblage

can account quantitatively for major element

variation within the Tonga volcanic suite.

[11]

We thus conclude that trend A, comprising

the low-CaO Batan melt inclusions and the

Batan whole rock lavas, can be interpreted

relatively simply in terms of progressive ex-

traction of olivine + clinopyroxene + amphi-

bole from a parental basaltic liquid with

wt % CaO and 8±10 wt % MgO, but that

trend B cannot be explained as readily by

crystallization differentiation. Likewise, pro-

gressive partial melting of a peridotitic source

is unlikely to account for trend B, since this

would produce a positive correlation between

CaO and FeO if the source contains residual

clinopyroxene [e.g.,

Kinzler and Grove,

1992a,

Baker and Stolper

, 1994;

Hirose and Kush-

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

Table 2.

Representative Major Element Compositions of the CaO-Rich, Silica-Undersaturated Melt Inclusions and Whole Rocks in Island Arcs

Identified in This Study

Host Arc

Volcano Sample SiO

TiO

FeO* MnO MgO CaO Na

Sum Reference

Inclusion Fo

Java Galunggung J-1 45.00 1.20 19.20 6.04 0.09 4.39 18.70 3.07 0.53 0.14 N.D. 98.45 S&B

(

98)

Inclusion Fo

F-1 43.90 1.28 18.40 7.53 0.09 4.99 18.40 3.49 0.66 0.18 N.D. 99.03 S&B

(

98)

Inclusion Fo

I-E 44.20 1.30 19.20 8.60 0.10 4.10 17.70 2.70 0.42 0.12 N.D. 98.57 S&B

(

98)

Inclusion Fo

Vanuatu Merelava 31551 46.20 0.49 13.68 8.87 0.09 12.47 15.53 1.55 0.56 0.41 N.D. 99.98 D&V

(

97)

Inclusion Fo

31551 45.88 0.46 14.07 8.87 0.02 12.18 15.89 1.59 0.52 0.37 N.D. 99.98 D&V

(

97)

Inclusion Fo

E. Sunda Rindjani 48002 43.39 1.13 15.35 9.36 0.05 11.77 15.24 2.43 0.98 0.16 N.D. 100.00 D&V

(

97)

Inclusion Fo

48001 45.49 0.93 14.23 8.57 0.14 9.54 17.76 1.85 1.33 0.16 N.D. 100.13 D&V

(

97)

Lava

48001 48.32 0.69 10.53 9.05 0.17 14.02 14.38 1.50 0.90 0.15 N.D. 99.85 F

(

83)

Inclusion Fo

Eolian Stromboli ol5/b118 48.88 0.75 12.97 6.51 N.D. 8.16 13.78 1.85 1.59 0.32 N.D. 94.91 M&C

(

96)

Inclusion Fo

Vulcano IV88.1 44.46 0.55 11.69 7.82 0.15 8.50 15.58 2.63 3.67 0.43 0.05 95.65 G

(

98)

Inclusion Fo

IV88.3.1 45.34 0.80 9.99 7.92 0.11 8.94 15.30 2.34 3.04 0.44 0.06 94.40 G

(

98)

Lava

N. Guinea Lihir

L3 47.35 0.76 12.46 10.36 0.21 9.12 14.73 2.90 0.44 0.36 N.D. 98.85 K

(

90)

Lava

L6 47.24 0.90 12.69 10.37 0.21 7.47 13.12 2.15 2.23 0.49 N.D. 97.03 K

(

90)

Lava

Nicaragua N.D.

CR-T 43.74 1.75 13.81 9.39 0.15 8.87 15.26 2.66 1.03 0.90 N.D. 97.70 C&R

(

84)

Lava

Lesser Grenada 6155 46.93 0.97 16.16 9.39 0.17 7.57 14.53 1.82 1.01 0.20 N.D. 98.88 T&G

(

84)

Lava

Antilles

450 47.17 0.96 15.58 10.01 0.18 7.06 14.43 2.12 0.52 0.12 N.D. 98.31 T&G

(

84)

Lava

503 45.60 0.95 15.91 8.53 0.18 10.95 13.50 1.88 1.01 0.40 N.D. 99.04 S&A

(

75)

Inclusion Fo

Luzon

Batan B45/53 44.33 0.73 16.74 7.19 0.12 8.18 18.07 2.72 0.91 0.37 N.D. 99.91 M

(

98)

Inclusion Fo

B45/43a 46.98 0.64 14.03 7.82 0.14 8.95 17.50 2.51 0.96 0.17 N.D. 100.14 M

(

98)

Inclusion Fo

B45/43b 44.92 0.77 15.68 8.93 0.15 8.79 17.36 2.44 0.83 0.64 N.D. 101.05 M

(

98)

FeO* is total iron reported as FeO. N.D., not determined. References: S&B

(

98),

Sisson and Bronto

[1998]; D&V

(

97),

Della-Pasqua and Varne

[1997]; F

(

83),

Foden

[1983]; M&C

(

96),

Metrich and Clocchiatti

[1996]; G

(

98),

Gioncada et al

.[1996];K

(

90),

Kennedy et al

. [1990]; C&R

(

84),

Carr and Rose

[1984]; T&G

(

84),

Thirwall and Graham

[1984]; S&A

(

75),

Shimizu and Arculus

[1975]; M

(

98),

Metrich et al

. [1998].

SCHIANO

AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

iro

, 1993], contrary to what is observed for

trend B

(

Figure 1d). Moreover, as discussed

below

(

see section 4.1), trend B probably

cannot be explained by progressive melting of

a pyroxene-rich lithology at a constant pressure,

because current estimates suggest that progres-

sive melting of pyroxenite results in liquids

with increasing CaO and decreasing normative

nepheline, contrary to the observed trend.

[12]

Our preferred explanation for the composi-

tional variation among the CaO-rich inclusions

(

i.e., those defining trend B) is that it represents

a mixing trend between the most CaO-rich

inclusions

(

i.e., those in the Fo

olivines) and

melts near the most primitive end of the ``nor-

mal'' Batan trend that we have defined as trend

A. Figures 1 and 3 demonstrate that this can

indeed explain the trend B. Although it is

difficult to test this explanation for trend B

using only the chemical compositions of melt

inclusions, in principle, isotopic differences

between inclusions along the compositional

array or detailed petrographic investigation of

whole rocks on this trend could be used to

evaluate this hypothesis.

2.2. Compilation of the Occurrences and

Compositions of CaO-Rich,

Silica-Undersaturated Melts in

Subduction Zone Settings

[13]

We compiled from the literature the com-

positions of arc-related whole rock samples

Galunggung

Batan

Stromboli

Lihir

Nicaragua

Grenada

Lombok

Rindjani

Epi

Merelava

Vulcano

LEGEND

> 13% CaO arc lavas

> 13% CaO melt inclusions

> 13% CaO arc lavas and melt inclusions

Figure 4.

Map showing the locations of the CaO-rich, silica-undersaturated whole rocks and melt

inclusions mentioned in the text. Solid circles indicate locations where CaO-rich inclusions have been found,

open squares indicate locations where CaO-rich whole rocks have been found, and solid squares indicate

locations where both have been found.

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems

oli

e-hosted melt inclusions having

CaO contents >13 wt %

(

Table 2). In addi-

tion to the Batan inclusions (this study and

Metrich et al

. [1999]), whole rocks and melt

inclusions with CaO contents >13 wt % are

found in volcanoes from many convergent

margins

(

Figure 4), including Galunggung

(

Sunda arc), Stromboli and Vulcano (Aeolian

arc), Lihir

(

New Guinea), Nicaragua

(

Central

America), Grenada

(

Lesser Antilles), Lombok

(

Sunda arc), Rindjani

(

Sunda arc), and Epi and

Merelava

(

Vanuatu arc) [

Sisson and Bronto

1998;

Della-Pasqua and Varne

,1997;

Gioncada

et al

., 1998;

Metrich and Clocchiatti

, 1996;

Foden and Varne

, 1980;

Foden

, 1983;

Kennedy

et al

., 1990;

Carr and Rose

, 1984;

Thirwall and

Graham

, 1984;

Shimizu and Arculus

, 1975;

Ar-

culus

, 1976].

[14]

All the inclusions reported in Table 2 are

hosted by Mg-rich olivine

(

84±94

) in basalts,

basaltic andesites, or ankaramites. As was

done for the Batan samples reported here, they

were homogenized by high-T experiments to

reverse postentrapment crystallization of the

host crystal [

Roedder

, 1984;

Schiano and

Bourdon

, 1999], with the exception of inclu-

sions from Galunggung [

Sisson and Bronto

1998], which were instead corrected by oli-

Figure 5.

Variation diagrams comparing Batan

CaO-rich melt inclusions

(

i.e., trend B) and CaO-

rich melt inclusions and whole rocks from other arc

environments

(

all of which have CaO > 13 wt %)

with ``normal'' melt inclusions and whole rocks

from Batan and other arcs

(

all of which have CaO

contents < 13 wt %)

(

shaded triangles without dot,

1350 basalts with MgO > 4.5 wt % from 30 arcs,

compiled by

Plank and Langmuir

[1988]. Shaded

circles without dot, melt inclusions in olivine

phenocrysts from

Gioncada et al

. [1998] and

Sisson

and Layne

[1993]). Symbols are given in the

accompanying legend. FeO* is total iron reported

as FeO. Major element concentrations have been

normalized to 100%.

FeO* (wt%)

CaO (wt%)

SiO

(wt%)

LEGEND

Batan samples

> 13 % CaO melt inclusions (i.e., trend B)

< 13 % CaO melt inclusions (i.e., trend A)

< 13 % CaO whole rocks (i.e., trend A)

other arcs

> 13 % CaO melt inclusions

< 13 % CaO melt inclusions

> 13 % CaO whole rocks

< 13 % CaO whole rocks

CaO (wt%)

SiO

(wt%)

SCHIANO

ET AL.:

CaO-RICH MELTS IN ISLAND ARCS

1999GC000032

Geochemistry

Geophysics

Geosystems