Earth and Planetary Science Letters 198 (2002) 355^369

www.elsevier.com/locate/epsl

Osmium isotope binary mixing arrays in arc volcanism

Sophie Alves  , Pierre Schiano 1 , Franc°oise Capmas, Claude J. Alle'gre
Laboratoire de Ge¤ochimie et Cosmochimie, Institut de Physique du Globe de Paris, 4 place Jussieu, 75252 Paris Cedex 05, France
Received 17 July 2001; accepted 6 February 2002

Abstract

Os isotope ratios and Os and Re concentrations were measured in 56 lavas coming from 10 different subduction
zones. Samples span a large range of major element concentrations (from basalts to dacites) and Mg# (from 0.32 to
0.81). The 10 subduction zones, namely the Lesser Antilles, Java, Papua New Guinea, the Philippines, Izu^Bonin,
Kamchatka, the Aleutians, Mexico, Colombia and Peru^Chile, have a range of geodynamic settings. Measured 187 Os/
188
Os ratios range from 0.130 to 1.524 and Os concentrations range from 0.05 to 46 ppt. Re concentrations range
from 24 to 915 ppt. Os initial isotope ratios are systematically positively and linearly correlated with the inverse of Os
concentrations in arc lavas from a given volcano, indicating that the Os isotopic compositions always reflect a binary
mixing process. Similar mixing relationships are also seen at the sample scale. All trends converge towards
unradiogenic compositions similar to those of upper mantle peridotites. These mixing relationships might be ascribed
to a general contamination process; however, a single shallow-level process of crustal assimilation is hardly reconciled
with the diversity of basements (from oceanic crust to continental crust compositions) of the selected arc volcanoes,
the occurrence of the mixing lines for both primary and differentiated samples, and the absence of covariations
between Os contents, isotope ratios, and indices of contamination and differentiation. On the other hand, because
subducted components are very radiogenic and differ from one zone to another, the radiogenic components may be
explained by varying amounts and natures of oceanic crust and sediments in the source of arc lavas. However, this
explanation implies two disequilibrium processes, first during magma formation in order to produce heterogeneous
lavas, and second during magma ascent to the surface to preserve slab signatures. ß 2002 Elsevier Science B.V. All
rights reserved.

Keywords: subduction; lava; osmium; crust; recycling; mixing; rhenium; isotope ratios

* Corresponding author. Present address: Pheasant Memorial Laboratory, Institute for Study of the Earth’s Interior, Okayama
University at Misasa, Misasa, Tottori-Ken 682-0193, Japan. Tel.: +81-858-43-3826; Fax: +81-858-43-3795.
E-mail addresses: [email protected] (S. Alves), [email protected] (P. Schiano),
[email protected] (F. Capmas), [email protected] (C.J. Alle'gre).

1
Present address: Laboratoire Magmas et Volcans, Universite¤ Blaise-Pascal, 5 rue Kessler, 63038 Clermont-Ferrand Cedex,
France.

0012-821X / 02 / $ ^ see front matter ß 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 0 1 2 - 8 2 1 X ( 0 2 ) 0 0 5 2 4 - 1

EPSL 6160 26-4-02

356 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

1. Introduction Arc lavas may have very low Os concentrations

(as low as 6 1 ppt [7]) and so far this character-
Isotopic and geochemical studies on subduc- istic has been an important limitation for their
tion-related lavas aim at constraining the nature analysis. Yet, an increasing number of studies
of their mantle sources and the role of source het- on arc lava Os isotope characteristics [7,11,12]
erogeneity and petrogenetic processes in their show very radiogenic osmium signatures, which
compositions. Many components are potentially are interpreted in terms of either mantle source
involved in producing the geochemical signatures heterogeneity [7,11] or secondary crustal contam-
of arc lavas: depleted mantle, subducted materi- ination [11,12].
als, and arc basements. Moreover, the compo- We present Os and Re concentrations and iso-
nents involved contribute by complex mixing of tope ratio data for 56 arc lavas from 10 subduc-
their £uid and melt derivatives. tion zones selected to represent a wide range of
Isotopic tracers are particularly suitable to geodynamic characteristics, and the results are
study the sources of arc magmas (e.g., [1]). U/ discussed in terms of the two mentioned hypoth-
Th disequilibria and boron systematics suggest eses : potential magma sources or contaminants.
an oceanic crust £uid in£uence for most conver-
gent margins (e.g., [2,3]), and contribution of sub-
ducted sediments is largely recognized in stronti- 2. Geological background
um, neodymium and lead signatures of arc lavas
(e.g., [1]). Physical parameters of the 10 subduction zones
The Re^Os isotopic system has previously been selected for the study are presented in Table 1 and
used to assess recycled basaltic materials in man- summarized here. Subducted oceanic crust ranges
tle sources of ocean island basalts (OIB), mid- from 15 to 150 Ma, sediment thickness of the
ocean ridge basalts (MORB) (e.g., [4^6]), and accretionary prism varies from 125 to 1750 m
arc lavas [7]. Here, we undertook a systematic and crustal basements of the arcs are 15^70 km
study of Re^Os isotopic characteristics of arc thick. The nature of basements also varies from
lavas for various subduction zones all over the one zone to another, with ma¢c (Java), basaltic
world. Os isotope ratios are potential tracers of (Aleutians, Lesser Antilles), accreted arc terrane
slab contribution in arc lavas because: (1) sub- type (Kamchatka) and continental crust (Central
ducted sediments are very radiogenic in Os com- and South America) compositions. For most of
pared to the upper mantle (e.g., [8]), and (2) Re the samples, stratigraphic or absolute K^Ar ages
behaves as a moderately incompatible element are known and are less than a few hundred kyr
during mantle partial melting, whereas Os is (cf. Table 2). Moreover, samples from a given
highly compatible. Therefore, MORB have much volcano span a limited range of ages.
higher Re/Os ratios than peridotites (187 Re/ For some of the arcs (Lesser Antilles, Izu^Bo-
188
Os = 140^16 000 for MORB [6], 187 Re/ nin, Kamchatka, Aleutians, and Mexico), we
188
OsW0.5 for peridotites (e.g., [9])). Conse- measured Os contents and isotope ratios for sev-
quently, old oceanic crust is likely to be extremely eral samples of the same volcano (Table 2). For
more radiogenic than the depleted upper mantle, 35 of these 39 samples, major and trace element
so that recycled basaltic components should be data were obtained at CRPG Nancy (SiO2 , Mg#
identi¢ed by their elevated 187 Os/188 Os ratios. and Ni are given in Table 2). The samples span a
However, crustal materials are also likely to be large range of major element concentrations,
rather radiogenic in osmium (e.g., [10]), and con- mainly in the calc-alkaline series, from basalts to
tamination via assimilation of the arc basement is dacites, and all show normalized trace element
a potential way to impart high Os isotopic signa- patterns typical for subduction-related magma-
tures to arc magmas and consequently mask the tism (Fig. 1). Kamchatka (51^55 wt% SiO2 ,
source signatures acquired in the mantle melting Mg# 0.52^0.81) and the Lesser Antilles (49^62
zone. wt% SiO2 , Mg# 0.46^0.68) samples are mostly

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 357

Table 1
Geophysical characteristics of the 10 subduction zones
Subduction zone Convergence Age of Sediment thickness Basement Crustal basement
ratea AOCb on the AOCc thicknesse geochemical naturef
(cm/yr) (Ma) (m) (km)
Lesser Antilles: North/South 1.4 86/100^150 235/1750 25/30^35 Oceanic
Java 7.6^7.9 138 300 25^30 Ma¢c/ultrama¢c
rocks+Sedimentsg
Papua New Guinea 9^14 50 1500d 30 Oceanic
Philippines 9 50 120 ^ Transitional
Izu^Bonin (oceanic arc/Fuji) 9.6^6.7 146 600 15^18/30 Oceanic/continentalh
Kamchatka 8.9^9.2 90 364 25^45 Accreted arc terranes
+obducted oceanic crusti
Aleutians 8.7^7.0 54 350 18^25 Oceanic
Mexico 5.7^8.5 15 170 30 Continental
Colombia 8.9^8.4 15 270 66 Continental
Peru^Chile 10.3^10.8 26^82 125 40^70 Continental
a
[34], except for Papua New Guinea [35] and Philippines [36].
b
AOC: altered oceanic crust [37], except for Lesser Antilles [38].
c
[36].
d
[39].
e
[34], except for Papua New Guinea [40] and for Northern Lesser Antilles [30].
f
[37] unless speci¢ed.
g
[41].
h
[42].
i
[43].

undi¡erentiated to slightly di¡erentiated basalts 0.078 pg. The blank average contribution for Re
and andesites, whereas most of the rocks from was less than 20 pg. A total of 39 blank measure-
Izu^Bonin (49^66 wt% SiO2 , Mg# 0.39^0.63) ments were performed during the period of the
and the Aleutians (46^61 wt% SiO2 , Mg# 0.35^ study, which corresponds to one blank measure-
0.54) are di¡erentiated with basaltic to dacitic ment for every three sample analyses. Analyses
compositions (cf. Table 2). were performed by negative ion thermal ioniza-
tion mass spectrometry on a Finnigan Mat 262
at IPG-Paris. Twenty-three runs of the laboratory
3. Analytical techniques and results standard solution diluted at various Os concentra-
tions (loads of 0.5^10 pg) were performed
Os and Re were extracted using the chemical throughout the study period. The mean 187 Os/
188
separation previously described by J.-L. Birck et Os ratio of the standard was 0.1742 þ 0.0002.
al. [13]. This low blank Os extraction procedure In order to test sample homogeneity, we made
makes it possible to measure reliable Os isotope duplicate analyses using di¡erent aliquots from
ratios in rock samples with 1 ppt or less of total the same batch of 30^50 g of powder made
Os content. For each sample 0.5^5 g of powder from a single piece of rock.
was used. The mean 2c error (including propaga- The concentration data for all the arc lavas
tion of blank error due to correction) of the iso- analyzed are given in Table 2. Os concentrations
topic ratios was 1.5%, with three of 90 measured are extremely low and variable, ranging from 0.05
ratios having an error higher than 5% (maximum to 46 ppt, and Re contents range from 24 up to
11.3%). Because of improvement of the reagent 915 ppt. 187 Re/188 Os ratios range from 5 to 54 200.
puri¢cation processes, the total Os blank de- Although Os and Re contents of the samples
creased during the course of the study from might have been somehow a¡ected by fractional
0.220 pg to 0.005 pg, with a mean value of crystallization (see Section 4), it is interesting to

EPSL 6160 26-4-02

358 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

Table 2
Os and Re concentrations and Os isotopic ratios for arc volcanoes
Samplea Island/Volcano/ Rock type Age or date 187
Os/188 Os Os Re 187
Re 187
Os SiO2 e Mg#e Nie
Location of eruptionb 188 Os 188 Osd
init
(yr) ( þ 2c error)c (ppt) (ppt) (wt%) (ppm)
Lesser Antilles
SB9 I Saba/Mt. andesite V106 0.595 (0.004) 2.000 67 171 0.592 54.67 0.68 98
Sceney/
Principal Dome
SB9 II 0.540 (0.003) 1.418 139 498 0.531
SB7 I Saba/Well’s andesite 4.20U105 0.957 (0.013) 0.728 188 1 376 0.948 60.6 0.59 21
Bay Dome
SB7 II 1.450 (0.011) 0.905 164 1 023 1.443
SB53 Saba/Old andesite V106 0.404 (0.003) 1.054 150 706 0.392 53.7 0.66 50
Booby’s Dome
SB67 Saba/Airport andesite 6 5U105 0.215 (0.001) 2.090 150 350 0.212 55.8 0.68 85
2nd £ow
SB49 Saba/Upper andesite V106 0.220 (0.002) 1.729 464 1 306 0.198 59.55 0.61 35
Hell’s Gate,
Ancient Dome
SK 11 St. Kitts/Mt. basalt 6 104 0.179 (0.002) 0.940 301 1 551 0.179 50.3 0.54 9
Misery
SK9 I St. Kitts/Mt. andesite 6 104 0.407 (0.008) 0.248 184 3 763 0.407 56 0.54 12
Misery
SK9 II 0.446 (0.011) 0.241 187 3 862 0.445
SK9 III 0.432 (0.005) 0.257 185 3 601 0.431
SK6 St. Kitts/Mt. basalt 6 104 0.236 (0.004) 0.679 335 2 404 0.235 49 0.62 31
Misery
SK2 I St. Kitts/Mt. andesitic 6 104 0.382 (0.005) 0.519 345 3 339 0.381 53 0.52 10
Misery basalt
SK2 II 0.560 (0.017) 0.396 361 4 611 0.559
SK2 III 0.498 (0.003) 0.366 383 5 282 0.497
F802 Guadeloupe/ andesite 1^2U105 0.266 (0.030) 1.461 357 1 199 0.262 57.21 0.48 5
Soufrie're/Nue¤e
St. Vincent
F802 II 0.274 (0.003) 1.004 457 2 234 0.267
MG 802 Martinique/Mt. andesite 1902 0.149 (0.001) 1.911 103 262 0.149 62.04 0.46 4
Pele¤e
MI 902 I Martinique/Mt. andesite 13.5U103 0.743 (0.005) 0.375 101 1 413 0.742 60.94 0.46 2
Pele¤e, Aileron
Dome
MI 902 II 0.811 (0.025) 0.284
MF 902 Martinique/Mt. andesitic 1^5U105 0.229 (0.002) 0.837 144 842 0.222 55.17 0.53 7
Pele¤e, Morne basalt
Plume¤e
SV6 I St. Vincent andesitic 1979 0.624 (0.003) 2.530 393 798 0.624 54.1 0.55 22
basalt
SV6 II 0.571 (0.004) 3.096 430 708 0.571
SV8001 I St. Vincent/ andesitic 6 6U105 0.348 (0.001) 5.631 0.348 53.48 0.54 17
Soufrie're basalt
SV8001 II 0.465 (0.004) 3.514 310 445 0.461
SV8001 III 0.555 (0.002) 2.926 279 486 0.550
Java
MR 43 A Java/Merapi basalt 0.523 (0.004) 1.823 448 1 248
BL 36 (39) Java/Baluran basalt 0.262 (0.002) 5.416 246 223

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 359

Table 2 (Continued)
Samplea Island/Volcano/ Rock type Age or date 187
Os/188 Os Os Re 187
Re 187
Os SiO2 e Mg#e Nie
Location of eruptionb 188 Os 188 Osd
init
(yr) ( þ 2c error)c (ppt) (ppt) (wt%) (ppm)
Papua New Guinea
94-LIH-11 Papua New 6 1.6U106 0.1728 (0.0008) 19.60 220 54 0.1585
Guinea/Lihir
94-AMB-29 Papua New 6 1.6U106 0.1310 (0.0004) 38.75 37 5 0.1298
Guinea/Feni/
Ambittle
Philippines
KH920814-01 Pinatubo andesite 1991 0.258 (0.003) 0.755 399 2 587 0.258
Izu^Bonin
YH76 I Honshu/ basalt 6U105 0.164 (0.003) 1.146 328 1 378 0.151 50.69 0.63 56
Hakone
YH76 II 0.192 (0.013) 1.227 477 1 885 0.174
YH76 III 0.162 (0.001) 1.220 312 1 239 0.150
YH18 Honshu/ dacite 2.7U105 0.831 (0.031) 0.062 384 28 884 0.703 64 0.40 2
Hakone
YH18 II 1.181 (0.047) 0.048 396 54 197 0.941
YH18 III 0.567 (0.010) 0.109 490 22 891 0.466
K8 Honshu/ andesite 1.5U104 0.154 (0.002) 1.461 353 1 283 0.154 58
Hakone/
Komagatake
No. 14 I Izu Oshima/Mt. andesite 1986 0.430 (0.010) 0.196 727 17 885 0.430 57.38 0.39 4
Mihara
No. 14 II 0.419 (0.009) 0.205 726 18 250 0.419
No. 14 III 0.409 (0.006) 0.284 865 15 197 0.409
No. 28 Izu Oshima/Mt. basalt 1986 0.161 (0.004) 0.641 433 3 230 0.161 52.7 0.46 17
Mihara
No. 28 II 0.209 (0.007) 0.639 432 3 324 0.209
OM4 Honshu/ andesitic 104 0.144 (0.001) 3.615 347 463 0.144 53.3
Omuro/ basalt
Komuroyama
Hino I Honshu/Mt. basalt 6 1730 þ 30 0.163 (0.002) 1.705 416 1 182 0.163 50.08 0.52 37
Fuji/
Hinokimarubi
Hino II 0.208 (0.004) 1.446 662 2 238 0.208
F5 Honshu/Mt. basalt 864 0.140 (0.001) 3.425 373 525 0.140 50.41 0.52 40
Fuji/
Aokigahara
1707 Honshu/Mt. dacite 1707 0.155 (0.001) 7.781 522 325 0.155 66.06 0.43
Fuji/Hoei
crater
F2 Honshu/Mt. basalt 2U105 0.165 (0.001) 5.974 132 107 0.165 48.8
Fuji/Mishima
Kamchatka
K77381 I Klyuchevskoy basalt 6 104 0.1416 (0.0005) 10.48 387 181 0.142 51.59 0.76 211
K77381 II 0.141 (0.001) 9.581 353 178 0.141
K56P66 I Klyuchevskoy andesitic 1966 0.201 (0.004) 2.484 658 1 289 0.201 53.62 0.59 30
basalt
K56P66 II 0.203 (0.002) 2.186 463 1 030 0.203
T 1475 II Tolbachik basalt 1975 0.175 (0.001) 8.595 409 235 0.175 50.85 0.72 112
T 1475 II 0.176 (0.001) 8.390 377 218 0.176
PT30476 II Tolbachik basalt 1976 0.147 (0.001) 11.88 450 183 0.147 51.72 0.54 50
PT30476 II 0.147 (0.001) 11.15 506 217 0.147

EPSL 6160 26-4-02

360 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

Table 2 (Continued)
Samplea Island/Volcano/ Rock type Age or date 187
Os/188 Os Os Re 187
Re 187
Os SiO2 e Mg#e Nie
Location of eruptionb 188 Os 188 Osd
init
(yr) ( þ 2c error)c (ppt) (ppt) (wt%) (ppm)
4
P33182 I Tolbachik basalt 6 10 0.223 (0.005) 0.514 536 5 067 0.222 52.57 0.55 20
P33182 II 0.242 (0.006) 0.571 678 5 796 0.241
P33182 III 0.246 (0.004) 0.624 484 3 802 0.245
P33182 IV 0.153 (0.003) 1.400 426 1 470 0.153
651-1 I Krasheninnikov andesitic 3.96U104 (?) 0.230 (0.002) 0.311 53.14 0.54 20
basalt
651-1 II 0.210 (0.006) 0.324 680 10 312 0.203
651-1 III 0.134 (0.003) 1.285
651-1 IV 0.200 (0.007) 0.467 559 5 794 0.196
A29107 I Avachinsky andesitic 6 1.8U106 0.1366 (0.0005) 11.15 286 125 0.136 54.49 0.60 23
basalt
A29107 II 0.228 (0.008) 0.530 646 5 974 0.227
A29107 III 0.181 (0.010) 1.090 361 1 498 0.181
A8501f Avachinsky basalt 6 1.8U106 0.133 (0.001) 46.00 51.66 0.81 250
K80317 I Mutnovski basalt 6 104 0.236 (0.005) 0.747 757 4 967 0.235 51.49 0.52 14
K80317 II 0.231 (0.003) 0.832 915 5 376 0.230
Aleutians
B 1927 I Bogoslof basalt 1927 0.201 (0.001) 3.763 24 31 0.201 0.53
B 1927 II 0.319 (0.006) 1.618 166 507 0.319
B 1796 I Bogoslof andesite 1796 0.136 (0.001) 5.636 537 935 0.136 0.33
B 1796 II 0.155 (0.001) 6.979 589 408 0.155
ADK53 Adak andesite 1796 0.1378 (0.0005) 23.56 174 36 0.1378
UM21 I Umnak/Okmok basalt 1946 0.144 (0.001) 1.364 611 2 162 0.144
UM21 II 0.202 (0.001) 1.054 730 3 371 0.202
Mexico
VE21 Pico de Orizaba andesite 0.371 (0.004) 0.672 251 1 861
PU12 Popocatepetl/ andesite 6 23 000 0.250 (0.001) 6.263 287 225 0.250
Puebla
PO1 Popocatepetl basalt 6 23 000 0.289 (0.003) 4.296 179 206 0.289
PO6 Popocatepetl basalt 6 23 000 0.313 (0.004) 3.287 158 238 0.313
SOC9301-A Socorro Island basalt 1993 0.130 (0.001) 3.884 710 881 0.130
Colombia
KH 950727 Colombia/ andesite 0.279 (0.003) 1.647 187 559
Galeras
Peru^Chile
SHILA I Peru/Shila 0.521 (0.005) 1.207 170 713
16A Chile/Sierra andesite 1.524 (0.007) 2.141 100 265
del Lipez
6B Chile/Miscanti andesite 0.770 (0.003) 1.877 176 491
17L Chile/Antuco basalt 6 10 000 0.145 (0.003) 1.593 234 712 0.145
18N Chile/Villarrica basalt 6 14 000 0.133 (0.001) 4.761 165 168 0.133
19M Argentina/El basalt 0.215 (0.001) 21.01 104 24
Pedrero
a
Numbers I, II, III and IV are for duplicates of the same sample.
b
For Lesser Antilles, Kamchatka, Aleutians, Japan and Papua New Guinea, age references are respectively, R. Maury, personal
communication (and [44] for SV8001), [45], [28], [46] and I. Kaneoka, personal communication and [47].
c
Mass fractionation was corrected using 192 Os/188 Os = 3.08271. Measured ratios were corrected for blank contribution. 2c errors
take blank error propagation into account when several blanks are available for a short period of time.
d
Age correction uses VRe = 1.64U10311 yr31 [48].
e
SiO2 , Mg# and Ni compositions were measured at CRPG, Nancy, France, by emission-ICP and ICP-MS respectively, except
for Bogoslof [49] and Antilles (SiO2 and Mg# from R. Maury, personal communication).
f
Os data from [50].

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 361

Fig. 1. Spidergram of trace element concentrations normal-

ized to primitive mantle [32], for Izu^Bonin, Lesser Antilles
and Kamchatka arcs. Gray ¢elds show ranges for each sub- Fig. 2. 187 Re/188 Os ratio versus Os concentrations for arc
duction zone, and thick black lines are for MORB [32] and lavas, MORB and OIB (references in the text). The meteorite
continental crust [33]. Note that for Kamchatka, there are trend [14] is included for comparison.
no Ta data. The samples show trace element patterns typical
for subduction-related magmatism, that is, enriched in large account, the regression line for St. Kitts changes
ion lithophile elements, and depleted in high ¢eld strength el-
ements, and in heavy rare earth elements, relative to MORB.
to an almost parallel line with r2 = 0.98 (Fig. 4).
Because the presented isotope ratios were cor-
rected for sample ages, these relationships cannot
re£ect isochronal trends from the 187 Os/188 Os ver-
note that all arc lavas plot on the terrestrial trend sus 187 Re/188 Os diagram, where the 187 Re/188 Os
de¢ned on the 187 Re/188 Os versus Os diagram ratio would be dominated by the variations of
(Fig. 2 and [14]). Os contents. Instead, these linear trends must rep-
Os isotopic compositions of arc lavas span a
very large range of 187 Os/188 Os values, from abys-
sal peridotite-like ratios to extremely radiogenic
values (Fig. 3). In particular, initial 187 Os/188 Os
ratios (corrected for sample ages given in Table
2) range from 0.149 to 1.443 for the Lesser Antil-
les rocks, 0.140 to 0.941 for Central Japan (Izu^
Bonin arc), 0.137 to 0.245 for Kamchatka, 0.136
to 0.319 for the Aleutian volcanoes, and 0.130 to
0.313 for Mexico. On average, these ratios are
higher than those known for other oceanic basalts
(e.g., [4^6]). Fig. 4 shows initial 187 Os/188 Os ratios
plotted against 1/188 Os for arcs for which many
samples were analyzed per volcano, including in
Fig. 4f the data previously obtained for West Java
[7]. In such a diagram, each volcano appears to Fig. 3. Osmium isotopic ratios versus the inverse of Os con-
de¢ne one positive linear correlation, or two centrations in arc lavas (this study and [7]), MORB and OIB
trends for two di¡erent periods in a volcano’s (references in the text). Except for Mexican and South Amer-
life (Fig. 4f). All trends have high correlation co- ican samples and two samples from Java, all isotope ratios
e⁄cients (r2 s 0.79), except Saba (r2 = 0.60) and are corrected for post-eruption Re radio-decay according to
ages given in Table 2. Note that for South American samples
St. Kitts (r2 = 0.41) in the Lesser Antilles. Note, with no known age, the 187 Re/188 Os are relatively low (com-
however, that if the radiogenic sample SK2, with pared to the whole set of data) and therefore, the age correc-
intermediate Os concentrations, is not taken into tion would probably not greatly change the isotope ratio.

EPSL 6160 26-4-02

362 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

Fig. 4. 187 Os/188 Os versus 1/188 Os diagrams for the Lesser Antilles (a), Izu^Bonin (b), Kamchatka (c), the Aleutians (d), Mexico
(e) and West Java [7] (f). Each volcano is represented by one single shape of symbol. White symbols are for samples analyzed
once and shades of gray are for multiple analyses of one given sample (for example the three dark gray circles in b are for three
analyses of the Hakone YH18 sample). When larger than the symbol, error bars are added for isotope ratios. Each volcano dis-
plays one alignment, for which the regression line is shown. The intercept values (i) and correlation coe⁄cients (r2 ) of these re-
gression lines are as follows (i3r2 ): St. Vincent (0.128^0.96); Saba (30.132^0.60); Martinique (0.013^0.96); St. Kitts (0.210^0.40)
or (0.108^0.98) (dotted line if sample SK2 is not considered, see text); Izu Oshima (0.093^0.89); Hakone (0.130^0.998); Klyu-
chevskoy (0.123^0.99); Tolbachik (0.153^0.79); Avachinsky (0.133^0.998); Krasheninnikov (0.134^0.86); Bogoslof (0.093^0.97);
Popocatepetl (0.183^0.98); and Java Bajah Dome Miocene (0.334^0.89); Pliocene (0.235^0.82).

resent binary mixing relationships between di¡er- YH18, Hakone volcano, Japan). The Re distribu-
ent end-members. Moreover, all the correlations tion was much less heterogeneous, but still had
seem to converge on a similar unradiogenic Os- a maximum range of 286^646 ppt in sample
rich end-member. A29107. When reported on the 187 Os/188 Os versus
Duplicate analyses for half of the samples show 1/188 Os diagrams (Fig. 4), the multiple analyses of
that both Os concentrations and isotopic ratios a heterogeneous sample are in good agreement
are heterogeneous within samples. The largest (with the exception of two samples, B1796, Bogo-
ranges observed in a sample for Os concentrations slof, Aleutians and P33182, Tolbachik, Kamchat-
and 187 Os/188 Os initial ratios were respectively ka) with the mixing line de¢ned for the whole set
0.530^11.151 ppt (sample A29107, Avachinsky of samples from the same volcano (e.g., samples
volcano, Kamchatka) and 0.466^0.941 (sample from St. Vincent island, Lesser Antilles and sam-

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 363

ple YH18, Hakone, Japan), or even de¢ne them- The two main hypotheses that will be discussed
selves a mixing line (651-1, Krasheninnikov, are shallow-level fractionation and contamination
Kamchatka, r2 = 0.86). of the magmas and heterogeneity of the mantle
source due to subducted slab supply.

4. Discussion 4.1. Shallow-level processes

Important variations exist in Re and Os con- The arc lavas analyzed in this study have ex-
tents and Os isotope ratios of arc lavas. Those tremely low Os contents (mostly 9 10 ppt) when
variations are observed in subduction zones compared to mantle peridotites. These small con-
from all around the world and seem to re£ect centrations may re£ect the compatible behavior of
binary mixing relationships expressed by positive Os during mantle melting. However, since most of
linear correlations between the Os isotopic ratio the samples are di¡erentiated, fractional crystalli-
and the inverse of its concentration. zation may also play an important role in lava
All the correlations seem to converge on a sim- concentrations, although it alone cannot produce
ilar unradiogenic Os-rich end-member (Fig. 4). the linear relationships displayed in Fig. 4 because
Yet, the actual regression lines’ intercepts display isotopes are not fractionated by such a process.
a large range of 187 Os/188 Os values, from 0.013 to Nevertheless, the e¡ect of fractional crystalliza-
0.334 (except for Saba, Lesser Antilles : 30.132). tion on Os contents is still a matter of debate. As
This range includes upper mantle values, as de- Os is clearly a compatible element in the terres-
¢ned from both MORB glasses [6] and peridotites trial mantle, one would also expect its content in
(e.g.,[15]), and may indicate Os isotope heteroge- lavas to decrease with increasing silica content
neity of the depleted part of the mantle wedge, and to be concentrated in Fe^Mg minerals. How-
due, for example, to di¡erent metasomatism his- ever, recent studies have shown that Os is in-
tories. However, this explanation should be taken compatible in magmatic olivine (from oceanic
with caution because the range in the unradiogen- basalts), as opposed to mantle olivine [16,17].
ic component signature likely results only from Consequently, the authors inferred that fractional
poor determination of the origin of the mixing crystallization alone cannot explain the Os^Mg^
lines, as illustrated by the meaningless negative Ni variations observed in oceanic basalt, but
intercept for Saba. Because the study of the man- source heterogeneity or sul¢de/olivine parallel
tle wedge composition is beyond the scope of this crystallization could.
paper, the unradiogenic component will, from We have thus plotted Os contents as a function
here, be taken to represent the mantle wedge, re- of SiO2 , Mg# and Ni contents in Fig. 5. Taken as
gardless of its possible Os isotopic heterogeneity. a whole, the present analyses fail to reveal any
In this section, we will discuss the nature of clear correlation between Os contents and indices
materials, which can possibly represent the Os of di¡erentiation such as SiO2 (Fig. 5a) or La/Yb
radiogenic components identi¢ed in arc lavas. (not shown). However, for Japan, a negative cor-
These materials, and the processes by which relation appears with SiO2 , with the exception of
they impart their signatures to lavas, must ac- the most Os-rich and silica-rich sample. For Ja-
count for: (1) the systematic occurrence of binary pan and Kamchatka, positive relationships also
mixtures in arc volcanoes from various geody- appear between Os and Ni and Mg#, for the
namic settings; (2) a time-dependent variation in whole arc data set, as well as within each volcano
the radiogenic component for a given volcano [7]; sample set, although it should be noted that in the
(3) the radiogenic signatures of both primitive and latter case, such trends are based on two or three
di¡erentiated arc lavas; (4) the existence of a va- samples only. These observations suggest, at least
riety of Os radiogenic components in subduction- for these arcs, that Os behavior during fractional
related volcanism; and (5) the heterogeneity in Os crystallization is somehow related to that of Ni.
concentrations and isotopic ratios of the samples. However, it must be emphasized that from the

EPSL 6160 26-4-02

364 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

(SiO2 , Mg# and La/Yb) suggest a geochemical

behavior of osmium, di¡erent from that recently
inferred for Os-rich rocks from Western Mexico
[12], where Os, SiO2 and MgO contents show a
direct relationship.
In summary, it is not clear from our data how
osmium is a¡ected by fractional crystallization,
although the low contents in the lavas may be
partly due to this process.
Alternatively, due to these low Os concentra-
tions, arc lavas are potentially sensitive to con-
tamination and therefore, radiogenic Os isotope
ratios of arc lavas and the mixing trends observed
for each volcano could re£ect a systematic con-
tamination process during arc magma ascent (in-
teraction with and/or assimilation of the crustal
basement, e.g., [12,18,19]) or after eruption (hy-
drothermalism, e.g., [19]).
Oceanic or continental crusts on which arcs are
built are likely to be very radiogenic and could in
fact a¡ect the signature of the magma and possi-
bly create the observed rock heterogeneity. In ad-
dition, the nature, age and thickness of these
basements, from one zone to another, are very
variable and can explain the range of isotopic
signatures observed for the radiogenic component
of the mixing trends.
On the other hand, contamination via assimila-
tion of crustal basement is usually invoked when
volcanoes are built on continental crust or thick
layers of sediments [12,18,20]. In fact, the thermo-
dynamic conditions and e¡ect of assimilation of
continental crust materials on the chemical com-
Fig. 5. Os concentrations versus SiO2 (a), Mg# (b) and Ni
(c) contents. Because of sample Os heterogeneity, and for position of magmas have been well studied
simplicity, we have represented each sample by its average through the so-called assimilation^fractional crys-
value of Os content to avoid plotting several values of Os tallization model (AFC, e.g., [21,22]), but the case
for each Ni, SiO2 or Mg# value. Each arc is represented by of oceanic crust assimilation has rarely been con-
a single shade, and for each shade, each shape of symbol is
sidered. Nevertheless, among the subduction
for a di¡erent volcano of the same arc.
zones represented in Fig. 4, only Mount Fuji,
Japan, Popocatepetl, Mexico, and Martinique,
present data, Os^Ni co-behaviors do not appear Lesser Antilles lie on continental or sedimentary
to be systematic and that even the most primitive basements, while the Izu^Bonin arc, the Lesser
samples (with Mg# higher than 65) have very low Antilles, and the Aleutians are oceanic arcs
concentrations of Os ( 9 10 ppt, except for sample (Table 1). Yet, we observe mixing lines for all
A8501, Kamchatka), which are comparable to volcanoes, regardless of the nature of the base-
those of the di¡erentiated lavas. In addition, the ment, and we have not found any relationship
absence of systematic relationships between Os between the slope of the mixing line or the range
and the degree of di¡erentiation of the rocks of Os ratios for a given volcano and the nature,

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 365

thickness or age of the overlying plate, as could tion to the fact that hydrothermalism may be a
be expected for simple contamination via base- signi¢cant process overprinting initial Os isotope
ment assimilation. ratios of lavas, because of the low closure temper-
Moreover, even though assimilation of conti- ature of pyrrhotite and possibly other Os-rich
nental crust (and oceanic crust) would strongly sul¢de or oxide phases. However, as for the as-
a¡ect, through partial melting, the incompatible similation hypotheses, it is unlikely that hydro-
element contents of arc basalts, it is less clear thermal events occurred systematically in all the
what the e¡ect is on a compatible element such areas we looked at, regardless of the particular
as Os, which is likely to be very radiogenic but setting of each zone.
also very depleted in crustal materials. One could In summary, although some arc magmas are
thus expect that compatible elements are signi¢- contaminated by the crustal basement they travel
cantly less involved in the assimilation process. through (as shown by their trace element and Sr,
Yet, as mentioned above, the studied samples dis- O, Nd or Pb isotope signatures), and di¡erenti-
play very typical arc lava patterns in Fig. 1, which ated magmas are more likely to be exposed to
show no evidence of secondary enrichment from such a process, it seems unlikely that the high
crustal material. Os isotope ratios measured in arc lavas presented
Finally, the linear relationships between Os iso- here and the systematic observation of Os isotope
topes and the inverse of Os concentrations could mixing lines for a set of worldwide subduction-
be explained by a constant assimilation/di¡eren- related volcanoes can be explained by this single
tiation ratio (AFC model). This would further process. The lack of understanding of Os behav-
imply that the Os concentrations and isotope ra- ior during both crystallization and partial melting
tios of arc lavas decrease with increasing degree of of basement wall rocks in contact with hot mag-
di¡erentiation of the rocks. However, as already ma makes it di⁄cult at present to discuss more
mentioned, no clear relationship exists between precisely this hypothesis, and we believe that fur-
Os and di¡erentiation indices (SiO2 , Mg# (Fig. ther investigation is needed to assess its e¡ects on
5a,b), and trace elements) for the samples ana- Os contents of lavas.
lyzed (except, to some extent, for Os contents in
Japan and Kamchatka, but it is not true for Os 4.2. Mantle source heterogeneity
isotope ratios (not shown)) and primitive arc
lavas from Kamchatka show mixing lines and ra- An alternative explanation for the Os charac-
diogenic Os ratios similar to the ones for the more teristics of arc lavas is that the mixing lines docu-
silica-rich arc samples analyzed in this study (cf. mented in this study re£ect mixtures, in the
Fig. 4c). source, of Os from the mantle and from the
In the case of Java [7], contamination by base- slab. In this case, the unradiogenic components
ment rocks was dismissed for this reason and also can be accounted for by the mantle wedge and
because it was hard to reconcile with the two the radiogenic components may be either the al-
radically di¡erent mixing trends de¢ned by Mio- tered oceanic crust and/or the sediments that are
cene and Pliocene samples of the single Bajah subducted beneath the arc. A similar hypothesis
volcanic dome of Java, especially because it would has been proposed for the Miocene and Pliocene
imply that osmium isotopic characteristics of the trends de¢ned by samples from Bajah Volcano
basement have strongly changed with time. from Java [7], mainly because each of these trends
Os mobility has been observed in hydrothermal was formed by samples having a distinct and ho-
systems of mid-ocean ridge settings [23], and hy- mogeneous Pb signature, MORB-like Pb isotope
drothermal contamination after eruption has been ratios for Miocene samples and sediment-like Pb
invoked to explain some highly radiogenic signa- ratios for Pliocene samples.
tures of oceanic basalts [24] via secondary post- Sediments have 187 Os/188 Os isotope ratios of
entrapment alteration by seawater-derived radio- about 1 and 1.3, for marine and terrigenous sedi-
genic Os. Moreover, Brenan et al. [19] call atten- ments respectively (e.g., [8]). These high values are

EPSL 6160 26-4-02

366 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

in good agreement with the radiogenic end-mem- Applying the calculation adopted for Java sam-
bers of some of the mixing lines (for example, St. ples to the other subduction zones (taking into
Vincent, St. Kitts, or the Kamchatka trends). account the respective ages of the crust and the
However, sediments alone cannot account for all average 187 Re/188 Os ratios of MORB for the given
the trends we have determined, especially those ocean [6]), we obtain, in the case of Izu^Bonin,
that have isotopic ratios similar to or even higher Kamchatka, the Lesser Antilles, the Aleutians,
than those of sediments (Mt. Hakone in Japan, and Mexico, 187 Os/188 Os ratios of 13, 8, 1.4, 5
Martinique and Saba in the Antilles, Miocene and 1.4 respectively, for the subducting oceanic
lavas in Java). Moreover, in the case of Kamchat- crust. These values are all higher than the more
ka, previously published Pb isotope data seem to radiogenic arc lava from the same region, except
exclude sediments as a signi¢cant component of for the Lesser Antilles, where one sample is
the magma source (e.g., [25]). slightly more radiogenic than the estimated
The oceanic crust subducting at the trenches of down-going oceanic crust. Therefore, adding ra-
the considered subduction zones has high 187 Re/ diogenic Os from the subducted oceanic crust to
188
Os ratios (based on MORB values from [6]) the mantle wedge appears to be a potential way to
with a range of ages from 15 to 150 Ma. Conse- explain the binary mixtures observed for each vol-
quently, the subducted crust must be much more cano. As 187 Re/188 Os ratios have been shown to
radiogenic than the mantle wedge rocks and be quite variable along a single ridge [6], for the
therefore gives a potential end-member for the Lesser Antilles, the calculated altered oceanic
observed trends of Fig. 4. For example in Java, crust 187 Os/188 Os ratio (considering the average
187
where the highest Os isotope ratios were mea- Re/188 Os value for the Atlantic MORB) may
sured, an estimate of the oceanic crust Os ratio be lower than the actual ratio of the crust beneath
gives a value of 6 for the 187 Os/188 Os ratio [7], the Lesser Antilles arc.
which is much higher than the highest ratio mea- However, the case of Java shows that the crust
sured in an arc lava (3.15 [7]). Moreover, it is a alone cannot explain the existence of two linear
priori di¡erent for each subduction zone depend- trends for the Miocene and the Pliocene samples
ing on the age and 187 Re/188 Os ratio of the crust. from this volcano because it would imply either
Yet, Re and Os behaviors during hydrothermal- an important change in the mobilization of radio-
ism at ridges and during metamorphic processes genic Os or that the crust subducted below the
are poorly understood and the 187 Re/188 Os ratio volcanic dome did not have the same isotopic
of the subducting slab may be quite di¡erent from signature during the Miocene and Pliocene. More-
the ratios measured in MORB. Eclogites and over, some regions, Java, and also the Lesser An-
blueschists, which are thought to be the result of tilles, clearly display sediment-like signatures for
metamorphism of the oceanic crust during sub- Sr, Nd and Pb isotopes (e.g., [1,27]), and thus the
duction, have similar Os concentrations and sig- role of sediments in their Os signatures should be
ni¢cantly lower Re concentrations when com- considered.
pared to MORB [26]. In this case, calculated Combining oceanic crust and subducted sedi-
values for the Os isotope ratio of altered oceanic ments in the magma source could be another
crust may be overestimated when using 187 Re/ way to explain the unique characteristics of the
188
Os values of MORB. However, Becker [26] sug- source of each volcano [7]. Unfortunately, the
gested that Re is more likely removed from the lack of constraints on the actual Os isotope char-
oceanic crust during the subduction metamor- acteristics of the subducting materials at each par-
phism itself, rather than during hydrothermalism ticular trench and Os partition coe⁄cients during
at ridges. In this case, the oceanic crust can devel- partial melting of metasomatized sources and slab
op very high Os isotope ratios by radio-decay of dehydrating processes makes it di⁄cult at present
Re during its transport to the trench, before re- to quantitatively model the mixing processes. In
leasing Re in larger amounts than Os in the sub- any case, the observations suggest that a three-
duction zone. component mixture could occur in the melting

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 367

zone, between the depleted upper mantle rocks of the islands along the Lesser Antilles arc may be
and variable proportions of oceanic crust and along-arc 187 Re/188 Os ratio heterogeneity of the
sediments, a hypothesis supported by previous oceanic and/or di¡erent partition coe⁄cients for
arc studies (e.g., [2,28]). Furthermore, the linearity Os from one volcano source to another depending
of the trends obtained in the 187 Os/188 Os versus 1/ on the slab physical parameters or on slab water
188
Os diagram suggests that the ratio of recycled contents.
sediments to oceanic crust contributions is con- Another point is the possibility of preserving
stant for a given volcano at a given period of these source characteristics through magma ascent
time. The slab-derived H2 O-rich component is be- in the mantle and in the arc basement before
lieved to induce mantle melting at mantle wedge reaching the surface. Not only arc lavas but also
temperatures, by lowering the peridotite solidus HIMU OIB have radiogenic Os signatures which
(e.g., [29]). The Os results further suggest that may be interpreted as originating from crust re-
the radiogenic slab component, made of sedi- cycling in the mantle melting zone [4,5]. Several
ment-derived and oceanic crust-derived materials, disequilibrium hypotheses have been proposed to
was homogenized before initiating mantle melt- allow these magmas to reach the surface without
ing. A similar conclusion has been reached based having equilibrated with unradiogenic Os-rich
on boron and beryllium contents [28], and boron mantle peridotites, including shielding of mantle
isotope [3] data on arc lavas. We must note, how- Os in Os-rich phases within host minerals [31],
ever, that unlike our results, these studies infer slow equilibration kinetics between peridotites
that the homogeneous subducted component and melts [26], isolation of melts from peridotites
made from oceanic crust and sediments contribu- by reaction zones along conduit walls [26], and
tions is unique for a whole arc [28] or across an rapid ascent through a network of channels [4].
arc from the volcanic front to the back-arc [3]. Finally, another disequilibrium state, which is
Although source heterogeneity can account for inferred from arc lavas, is sample heterogeneity
the observed Os characteristics in arc lavas, it for Os contents and isotope ratios. Such hetero-
raises a few questions, which can be summarized geneity has also been observed in mantle rocks
as follows. and OIB (e.g., [15]), and led, for mantle rocks,
First, it has not been possible so far to observe to the ‘nugget e¡ect’ hypothesis, which is a pref-
any covariation between Os and either other sub- erential distribution of Os in refractory sul¢de or
duction contributions or mantle melting indices in oxide micro-inclusions. These Os-bearing phases
our data, which would con¢rm the deep origin of may be in isotopic disequilibrium with their host
Os signatures. This is likely due to the e¡ect of minerals, suggesting that di¡usion processes be-
later fractional crystallization on the trace element tween the inclusion and the host mineral have
contents of the fractionated rocks. been prevented [31]. For lavas, the observed het-
Second, the respective proportions of oceanic erogeneity implies that the magmas themselves are
crust and sedimentary Os cannot be the only fac- heterogeneous in concentration and isotopic com-
tors responsible for the slope of the mixing line. position, and that they are not homogenized dur-
Indeed, in that case, the Lesser Antilles arc should ing ascent to the surface. Although such a process
show decreasing slopes from northern to central is not well constrained at present, Os heterogene-
islands, because of the increasing thickness of the ity of the samples suggests that the radiogenic Os
sediment column in the Lesser Antilles trench transferred from lithospheric £uids or melts can-
(e.g., [30]). Yet, this is not the case (cf. St. Kitts not equilibrate, within the arc magmas, with the
and Saba, Fig. 4). A possible explanation may be unradiogenic Os coming from the mantle.
modi¢cation of the original Os contents by frac-
tional crystallization, without destroying the mix-
ing trends, but fractional crystallization more Acknowledgements
likely accounts only for scatter in some trends.
Another explanation for the varying mixing lines The authors wish to thank R. Maury, M. Se-

EPSL 6160 26-4-02

368 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369

met, I. Kaneoka, J.-L. Joron and J.-C. Komor- Lassen region of the Cascade arc, California, Earth Plan-
et. Sci. Lett. 177 (2000) 301^317.
owski for providing most of the samples. We are
[12] J.C. Lassiter, J.F. Luhr, Osmium abundance and isotope
very grateful to J.-L. Birck for technical assistance variations in ma¢c Mexican volcanic rocks: Evidence for
in measuring Os concentrations and isotope ra- crustal contamination and constraints on the geochemical
tios. We also thank K. Burton and B. Bourdon behavior of osmium during partial melting and fractional
for pro¢table discussions, and P. Foster for care- crystallization, Geochem. Geophys. Geosyst. 2 (2001)
ful reading and improving of the manuscript. C. 2000GC000116.
[13] J.-L. Birck, M. Roy-Barman, F. Capmas, Re-Os isotopic
Hawkesworth and an anonymous reviewer are measurements at the femtomole level in natural samples,
also thanked for their thorough reviews. This Geostand. Newslett. 20 (1) (1997) 19^27.
work was largely supported by a grant from the [14] J.L. Birck, C.J. Alle'gre, Contrasting Re/Os magmatic
Bureau de Recherches Ge¤ologiques et Minie'res fractionation in planetary basalts, Earth Planet. Sci.
Lett. 124 (1994) 139^148.
(BRGM) to S.A.[BW]
[15] S.R. Hart, G. Ravizza, Os partitioning between phases in
lherzolite and basalt, in: A. Basu, S.R. Hart (Eds.), Earth
Processes: Reading the Isotopic Code, Am. Geophys.
References Union, Washington, DC, 1996, pp. 123^134.
[16] K.W. Burton, P. Schiano, A. Gannoun, J.-L. Birck, C.J.
[1] D. BenOthman, W.M. White, P.J. Patchett, The geochem- Alle'gre, Contrasting behaviour of osmium in mantle and
istry of marine sediments, island arc magma genesis and magmatic olivine, in: Fall Meeting, EOS Trans. AGU,
crust-mantle recycling, Earth Planet. Sci. Lett. 94 (1989) San Francisco, CA, 2000.
1^21. [17] J.M. Brenan, W.F. McDonough, I. Horn, P. Sattari, Ol-
[2] C.J. Hawkesworth, S.P. Turner, F. McDermott, D.W. ivine-melt partitioning of Re and the PGEs: Experimental
Peate, P. van Calsteren, U-Th isotopes in arc magmas: constraints, in: Spring Meeting, EOS Trans. AGU, 2000.
Implications for element transfer from the subducted [18] J.P. Davidson, Crustal contamination versus subduction
crust, Science 276 (1997) 551^555. zone enrichment: Examples from the Lesser Antilles and
[3] T. Ishikawa, E. Nakamura, Origin of the slab component implications for mantle source compositions of island arc
in arc lavas from across-arc variation of B and Pb iso- volcanic rocks, Geochim. Cosmochim. Acta 51 (1987)
topes, Nature 370 (1994) 205^208. 2185^2198.
[4] E.H. Hauri, S.R. Hart, Re-Os isotope systematics of [19] J.M. Brenan, D.J. Cherniak, L.A. Rose, Di¡usion of os-
HIMU and EMII oceanic island basalts from the South mium in pyrrhotite and pyrite: implications for closure of
Paci¢c Ocean, Earth Planet. Sci. Lett. 114 (1993) 353^371. the Re-Os isotopic system, Earth Planet. Sci. Lett. 180
[5] P. Schiano, K.W. Burton, B. Dupre¤, J.L. Birck, G. Guille, (2000) 399^413.
C.J. Alle'gre, Correlated Os-Pb-Nd-Sr isotopes in the Aus- [20] W. Hildreth, S. Moorbath, Crustal contributions to arc
tral-Cook chain basalts: The nature of the mantle com- magmatism in the Andes of Central Chile, Contrib. Min-
ponents in plume sources, Earth Planet. Sci. Lett. 186 eral. Petrol. 98 (1988) 455^489.
(2001) 527^537. [21] D.J. DePaolo, Trace element and isotopic e¡ects of com-
[6] P. Schiano, J.-L. Birck, C.J. Alle'gre, Osmium-strontium- bined wall rock assimilation and fractional crystallization,
neodymium-lead isotopic covariations in mid-ocean ridge Earth Planet. Sci. Lett. 53 (1981) 189^202.
basalt glasses and the heterogeneity of the upper mantle, [22] W.A. Bohrson, F.J. Spera, Energy-constrained open-sys-
Earth Planet. Sci. Lett. 150 (1997) 363^381. tem magmatic processes II: Application of energy-con-
[7] S. Alves, P. Schiano, C.J. Alle'gre, Rhenium-osmium iso- strained assimilation-fractional crystallization (EC-AFC)
topic investigation of Java subduction zone lavas, Earth model to magmatic systems, J. Petrol. 42 (2001) 1019^
Planet. Sci. Lett. 168 (1999) 65^77. 1041.
[8] B. Peucker-Ehrenbrink, G. Ravizza, A.W. Hofmann, The [23] G.E. Bru«gmann, P. Herzig, A.W. Hofmann, Os isotopic
marine 187 Os/186 Os record of the past 80 million years, composition and Os and Re distribution in the active
Earth Planet. Sci. Lett. 130 (1995) 155^167. mound of the TAG-hydrothermal system, Mid-Atlantic
[9] T. Meisel, R.J. Walker, J.W. Morgan, The osmium iso- Ridge, Proc. ODP Sci. Results 158 (1998) 91^100.
topic composition of the Earth’s primitive upper mantle, [24] F. Marcantonio, A. Zindler, T. Elliott, H. Staudigel, Os
Nature 383 (1996) 517^520. isotope systematics of La Palma, Canary Islands: Evi-
[10] B.K. Esser, K.K. Turekian, The osmium isotopic compo- dence for recycled crust in the mantle source of HIMU
sition of the continental crust, Geochim. Cosmochim. ocean islands, Earth Planet. Sci. Lett. 133 (1995) 397^410.
Acta 57 (1993) 3093^3104. [25] A.B. Kersting, R.J. Arculus, Pb isotope composition of
[11] L.E. Borg, A.D. Brandon, M.A. Clynne, W.J. Walker, Klyuchevskoy volcano, Kamchatka and North Paci¢c
Re-Os isotopic systematics of primitive lavas from the sediments: Implications for magma genesis and crustal

EPSL 6160 26-4-02

 S. Alves et al. / Earth and Planetary Science Letters 198 (2002) 355^369 369

recycling in the Kamchatkan arc, Earth Planet. Sci. Lett. song, Seismic studies on the Ontong Java Plateau, Tecto-
136 (1995) 133^148. nophysics 34 (1976) 71^90.
[26] H. Becker, Re-Os fractionation in eclogites and blueschists [40] D.M. Finlayson, J.P. Cull, Structural pro¢les in the New
and the implications for recycling of oceanic crust into the Britain-New Ireland region, J. Geol. Soc. Aust. 20 (1973)
mantle, Earth Planet. Sci. Lett. 177 (2000) 287^300. 37^48.
[27] J.P. Davidson, Lesser Antilles isotopic evidence of the [41] W.B. Hamilton, Tectonics of the Indonesian Region, U.S.
role of subducted sediment in island arc magma genesis, Geol. Surv. Prof. Paper, 1979, 345 pp.
Nature 306 (1983) 253^256. [42] T. Matsuda, Collision of the Izu-Bonin arc with central
[28] J.D. Morris, W.P. Leeman, F. Tera, The subducted com- Honshu: Cenozoic tectonics of the Fossa Magna, Japan,
ponent in island arc lavas: Constraints from Be isotopes in: S. Uyeda, R.W. Murphy, K. Kobayashi (Eds.), Geo-
and B-Be systematics, Nature 344 (1990) 31^36. dynamics of the Western Paci¢c, Advances in Earth and
[29] E. Stolper, S. Newman, The role of water in the petro- Planetary Sciences, Vol. 6, Japan Scienti¢c Societies Press,
genesis of Mariana trough magmas, Earth Planet. Sci. Tokyo, 1979, pp. 409^421.
Lett. 121 (1994) 293^325. [43] B.F. Watson, K. Fujita, Tectonic evolution of Kamchatka
[30] R.C. Maury, G.K. Westbrook, P.E. Baker, P. Bouysse, D. and the Sea of Okhotsk and implications for the Paci¢c
Westercamp, Geology of the Lesser Antilles, in: G. Deng, Basin, in: D.G. Howell (Ed.), Tectonostratigraphic Ter-
J.E. Case (Eds.), The Geology of North America, H, The ranes, Circum-Paci¢c Council for Energy and Mineral
Caribbean Region, Geol. Soc. Am., Boulder, CO, 1990, Resources, Houston, TX, 1985, pp. 333^348.
pp. 141^165. [44] E. Heath, S.P. Turner, R. Macdonald, C.J. Hawkesworth,
[31] K.W. Burton, P. Schiano, J.-L. Birck, C.J. Alle'gre, Osmi- P. van Calsteren, Long magma residence times at island
um isotope disequilibrium between mantle minerals in a arc volcano (Soufrie're, St. Vincent) in the Lesser Antilles
spinel-lherzolite, Earth Planet. Sci. Lett. 172 (1999) 311^ Evidence from 238 U-230 Th isochron dating, Earth Planet.
322. Sci. Lett. 160 (1998) 49^63.
[32] S.S. Sun, W.F. McDonough, Chemical and isotopic sys- [45] F. Pineau, M.P. Semet, N. Grassineau, V.M. Okrugin, M.
tematics of oceanic basalts: Implications for mantle com- Javoy, The genesis of the stable isotope (O, H) record in
position and processes, in: A.D. Saunders, M.J. Norry arc magmas: The Kamchatka’s case, Chem. Geol. 153
(Eds.), Magmatism in the Ocean Basins, Vol. 42, Geol. (1999) 93^124.
Soc., Blackwell Scienti¢c, Oxford, 1989, pp. 313^345. [46] M. Tatsumoto, R.J. Knight, Isotopic composition of lead
[33] A.W. Hofmann, Chemical di¡erentiation of the Earth: in volcanic rocks from central Honshu ^ with regard to
The relationship between mantle, continental crust, and basalt genesis, Geochem. J. 3 (1969) 53^86.
oceanic crust, Earth Planet. Sci. Lett. 90 (1988) 297^314. [47] R.W. Johnson, D.A. Wallace, D.J. Ellis, Feldspathoid-
[34] J. Gill, Orogenic Andesites and Plate Tectonics, Springer- bearing potassic rocks and associated types from volcanic
Verlag, Berlin, 1981, 390 pp. islands o¡ the coast of New Ireland Papua New Guinea:
[35] J.W. Curtis, Plate tectonics of the Papua New Guinea- A preliminary account of geology and petrology, in: R.D.
Solomon Islands region, J. Geol. Soc. Aust. 20 (1973) Johnson (Ed.), Volcanism in Australia, Elsevier, Amster-
21^36. dam, 1976.
[36] T.A. Plank, C.H. Langmuir, The chemical composition of [48] M. Lindner, D.A. Leich, G.P. Russ, J.M. Bazan, R.J.
subducting sediment and its consequences for the crust Borg, Direct determination of the half-life of 187 Re, Geo-
and mantle, Chem. Geol. 145 (1998) 325^394. chim. Cosmochim. Acta 53 (1989) 1597^1606.
[37] R.D. Jarrard, Relations among subduction parameters, [49] M.T. McCulloch, M.R. Per¢t, 143 Nd/144 Nd, 87 Sr/86 Sr and
Rev. Geophys. 24 (1986) 217^284. trace element constraints on the petrogenesis of Aleutian
[38] G.K. Westbrook, Magnetic lineations and fracture zones, island arc magmas, Earth Planet. Sci. Lett. 56 (1981) 167^
in: R.C. Speed, G.K. Westbrook (Eds.), Lesser Antilles 179.
Arc and Adjacent Terranes: Ocean Margin Drilling Pro- [50] M. Roy-Barman, Mesure du rapport 187 Os/186 Os dans les
gram Regional Atlas Series, Marine Science International, basaltes et pe¤ridotites: Contribution a' la syste¤matique
187
Woods Hole, MA, 1984, Sheet 5. Re/186 Os dans le manteau, The'se en Ge¤ochimie, Uni-
[39] A.S. Furumoto, J.P. Webb, M.E. Odegard, D.M. Hus- versite¤ Denis Diderot, Paris VII, 1993.

EPSL 6160 26-4-02