Hydrated sub-arc mantle: a source for the Kluchevskoy
volcano, Kamchatka/Russia
Frank Dorendorf a, Uwe Wiechert a;b;1, Gerhard Wo«rner a;*
a Geochemisches Institut, University of Go«ttingen, Goldschmidtstr. 1, 37077 Go«ttingen, Germany
b Geophysical Laboratory, Carnegie Institiution of Washington, 5251 Broad Branch Rd. NW, Washington, DC 20015, USA
Received 23 April 1999; received in revised form 15 July 1999; accepted 16 November 1999
Abstract
Oxygen isotope ratios of olivine and clinopyroxene phenocrysts from the Kluchevskoy volcano in Kamchatka have
been studied by CO2 and ArF laser techniques. Measured N18O values of 5.8^7.1x for olivine and 6.2^7.5x for
clinopyroxene are significantly heavier than typical mantle values and cannot be explained by crustal assimilation or a
contribution of oceanic sediments. Positive correlations between N18O and fluid-mobile elements (Cs, Li, Sr, Rb, Ba, Th,
U, LREE, K) and a lack of correlation with fluid-immobile elements (HFSE, HREE) suggest that 18O was introduced
into the mantle source by a fluid from subducted altered oceanic basalt. This conclusion is supported by radiogenic
isotopes (Sr, Nd, Pb). Mass balance excludes simple fluid-induced mantle melting. Instead, our observations are
consistent with melting a mantle wedge which has been hydrated by 18O-rich fluids percolating through the mantle
wedge. 18O-enriched fluids are derived from the subducted oceanic crust and the Emperor seamount chain, which is
responsible for a particularly high fluid flux. This hydrated mantle wedge was subsequently involved in arc magmatism
beneath Kluchevskoy by active intra-arc rifting. ß 2000 Elsevier Science B.V.
Keywords: subduction zones; Kamchatka Peninsula; isotopes; slabs; Emperor Seamounts; £uid phase; hydration; trace elements
1. Introduction
Arc magmatism is triggered by lowering the
solidus in the mantle wedge through ingress of
slab-derived £uids. The mantle wedge is the
main source of subduction zone magmatism, but
a signi¢cant contribution of trace elements comes
from subducted sediments and/or £uids from al-
tered oceanic crust. It has been shown [1] that
some trace element abundances in depleted arc
magmas are correlated with the speci¢c subduc-
tion £uxes beneath arcs and Sr and Be isotopes to
permit tracing and quanti¢cation of sediment
£uxes into the wedge source. More controversial
is the nature and amount of the slab £uid. Recent
estimations range from less than 1% [2] to values
up to 20% [3]. These di¡erences stem from limited
knowledge of £uid composition and the nature of
interaction with the mantle.
Oxygen isotopes are a powerful tool for tracing
0012-821X / 00 ß 2000 Elsevier Science B.V.
PII: S 0 0 1 2 - 8 2 1 X ( 9 9 ) 0 0 2 8 8 - 5
* Corresponding author. Tel. : +49-551-393971/72;
Fax: +49-551-393982; E-mail: gwoerne@gwdg.de
1 Present address: Institute for Isotope Geology and Min-
eral Resources, ETH-Zentrum, Sonneggstr. 5, CH-8092 Zu«-
rich, Switzerland.
EPSL 5329 31-12-99
Earth and Planetary Science Letters 175 (2000) 69^86
www.elsevier.com/locate/epsl
Open access under CC BY-NC-ND license.
Open access under CC BY-NC-ND license.
crustal material in basalts [4]. This assumes the
mantle oxygen isotope composition to be constant
[5] and that rocks studied represent magmatic O
isotope values. Glasses and/or feldspar in whole
rock samples are very sensitive to alteration and
ma¢c minerals have been di⁄cult to analyze con-
ventionally. However, new laser ablation techni-
ques have been developed and give reliable O iso-
tope compositions on olivine and pyroxene [6].
The accepted range of O isotope composition
for mantle rocks based on these studies is
5.7 þ 0.5x (2c) relative to SMOW [7]. Previous
work [8,9] reported conventional N18O data and
elevated values for Kluchevskaya group lavas in
the Central Kamchatka, the implications of which
have been controversial.
We present new ArF and CO2 laser ablation
oxygen isotope data for olivine and clinopyroxene
phenocrysts from a suite of 16 basalts and basaltic
andesites from the Holocene Kluchevskoy volca-
no. Our results exclude alteration and problems
with Mg-rich samples in conventional studies. In
addition, the high precision of our data shows
N18O to be closely correlated with some major
and trace elements and Sr and Pb isotopes.
From these correlations we show that the source
of the 18O-enriched oxygen is hydrothermally al-
tered oceanic crust introduced by source contam-
ination rather than assimilation of arc crust. We
argue for a special role of the subducted Emperor
seamount chain and Quaternary intra-arc rifting
in Kamchatka.
2. Geological setting and sampling
Quaternary volcanism on Kamchatka in Far
Eastern Russia is related to the subduction (9
cm/yr) of the Paci¢c below the Eurasian plate.
Volcanic centers are concentrated in three zones
parallel to the trench (Fig. 1A): (1) the Eastern
Volcanic Front (EVF), (2) the Central Kamchat-
ka Depression (CKD) and (3) the western Sre-
dinny range (SR). The EVF consists of about 20
active and many more Holocene to Pleistocene
volcanoes. Quaternary volcanism in the SR is of
Fig. 1. (A) Kamchatka Peninsula, tectonic structure and active volcanism in the Eastern Volcanic Front (EVF), Central Kam-
chatka Depression (CKD) and Sredinny Ridge (SR). Modi¢ed after [12]. (B) Cinder cones, historic eruptions and sample loca-
tions at Kluchevskoy volcano.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8670
the back arc type. The Kluchevskaya Group in
the CKD, which represents a 200 km wide graben
structure, comprises 12 volcanoes all less than
50 000 years old. The largest and highest is the
recently active Holocene Kluchevskoy volcano
(4750 m), one of the most productive arc volca-
noes in the world. Eruption products are high-Mg
basalts and basaltic andesites (HMB) and high-Al
basaltic andesites (HAB). This suite is explained
by polybaric fractionation under hydrous condi-
tions starting at 19 kbar and 1350‡C [10,11]. High
magmatic water contents suppress plagioclase
crystallization resulting in strong Al2O3 enrich-
ment in the melt.
Geophysical studies show that Cenozoic vol-
canic deposits extend to 8 km below Kluchevskoy
volcano [12]. Greenschist to amphibolite facies
rocks of Mesozoic (?) age constitute the crust
mantle down to 30 km. A funnel-like zone of
magma ascent between 20 and 50 km connects
the Kluchevskaya group volcanoes with their
mantle source [13]. The top of the Wadati^Benio¡
zone below Kluchevskoy is at a depth of about
160 km [12].
Samples studied here were on the NE and N
base of Kluchevskoy (Fig. 1B, EPSL Online
Background Dataset2 Table 1). The historic sam-
ples are HMB from the eruptions of 1932 and
1938 and HAB from the eruptions of 1951, 1953
and 1956. The remaining samples (all HMBs)
Fig. 2. Plots of MgO versus SiO2, TiO2, K2O and Na2O for Kluchevskoy samples. HMB and HAB are distinguished by open
squares and circles, respectively. Closed symbols are calculated fractionation-corrected, primitive compositions for Mg#70. Vec-
tors for 10% olivine (ol), 20% clinopyroxene (cpx) and 10% olivine+20% clinopyroxene (ol+cpx) fractionation are shown in each
diagram. It is obvious from MgO^K2O^SiO2 and trace element relations that additional processes other than fractional crystalli-
zation are necessary to explain the compositional range. Dashed lines in the plots of MgO versus SiO2 and K2O represent con-
tours of N18O (x). The gray ¢eld represents the data for Kluchevskoy of [10].
2 http://www.elsevier.nl/locate/epsl; mirrorsite: http://www.
elsevier.com/locate/epsl
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 71
were derived from undated Holocene cinder
cones and lava £ows. Great care was taken to
avoid any trace of alteration. Samples were large
enough (1^s 10 kg) to allow separation of phe-
nocrysts.
3. Analytical procedures
Major and trace elements (Ba, Co, Cr, Ga, Sc,
Sr, Ni, V, Zn, Zr) were measured on glass fusion
discs by X-ray £uorescence (XRF). From interna-
Table 1
Trace element data for samples from Kluchevskoy volcano. Regression coe⁄cients for trace elements versus N18O of the raw and
fractionation corrected data are also shown. Strong correlations (rs 0.6 or r630.6) are emphasized in bold.
KLU-01 KLU-02 KLU-03 KLU-04 KLU-05 KLU-06 KLU-07 KLU-08 KLU-09 KLU-10 KLU-11 KLU-12 KLU-13
Li 10.9 13.2 8.3 7.8 9.1 8.1 8.6 8.9 8.3 7.5 7.9 13.9 13.3
Be 0.55 0.63 0.50 0.58 0.62 0.48 0.47 0.62 0.50 0.49 0.54 0.66 0.68
Sc 33.0 29.9 35.9 35.2 35.1 34.0 36.0 30.9 35.0 35.1 36.9 21.1 27.0
V 255 259 239 249 235 245 244 246 250 256 261 260 267
Cr 256 45 424 285 241 341 500 220 343 319 320 31 6
Co 32.0 36.9 36.9 39.2 39.2 40.0 40.0 34.9 38.0 39.1 35.9 33.1 30.0
Ni 93 39 150 100 106 107 166 88 104 108 109 31 14
Cu 79 84 60 79 74 59 64 72 78 74 73 91 88
Zn 85 88 79 84 80 81 78 80 78 81 81 87 88
Ga 19.0 17.9 15.9 15.1 18.1 16.0 16.0 17.0 17.0 19.0 18.0 18.0 19.0
Rb 15.1 17.8 12.2 10.2 12.3 11.3 11.6 11.9 9.5 9.4 9.4 19.0 19.0
Sr 341 392 323 295 316 299 312 318 290 291 286 398 403
Y 18.5 21.2 14.9 19.5 17.9 15.9 15.3 18.5 17.8 18.3 19.0 22.0 21.8
Zr 87 104 81 101 99 81 77 98 87 85 85 104 105
Nb 1.46 1.75 1.15 1.70 1.67 1.17 1.31 1.78 1.43 1.46 1.39 1.99 1.92
Cs 0.43 0.47 0.42 0.36 0.46 0.28 0.39 0.39 0.33 0.31 0.32 0.51 0.48
Ba 344 396 289 224 260 283 280 260 243 224 228 435 428
La 6.09 7.55 5.41 6.12 6.69 5.03 5.31 6.72 5.54 5.32 5.36 8.43 8.20
Ce 15.1 19.3 13.6 16.3 17.1 13.0 13.1 17.5 14.3 14.2 14.6 21.1 21.3
Pr 2.49 3.16 2.17 2.62 2.77 2.14 2.13 2.83 2.38 2.34 2.46 3.39 3.54
Nd 11.9 14.9 10.4 12.4 12.7 10.2 10.3 12.9 11.2 11.4 11.6 15.8 15.6
Sm 3.35 4.07 2.94 3.54 3.41 2.91 2.99 3.50 3.07 3.21 3.25 4.14 4.11
Eu 1.06 1.26 0.90 1.07 1.04 0.94 0.93 1.07 1.01 1.02 0.97 1.27 1.29
Gd 3.32 3.75 2.89 3.52 3.35 3.01 2.87 3.35 3.18 3.20 3.20 3.77 3.76
Tb 0.54 0.62 0.45 0.57 0.54 0.48 0.46 0.56 0.53 0.55 0.56 0.63 0.65
Dy 3.42 3.82 2.87 3.66 3.27 3.07 2.91 3.46 3.36 3.42 3.42 3.84 3.82
Ho 0.71 0.77 0.61 0.79 0.69 0.64 0.59 0.71 0.68 0.70 0.70 0.82 0.81
Er 1.99 2.22 1.68 2.27 1.98 1.82 1.73 2.06 1.95 2.06 2.04 2.36 2.34
Tm 0.29 0.33 0.26 0.33 0.30 0.27 0.25 0.30 0.28 0.31 0.31 0.34 0.36
Yb 1.94 2.13 1.61 2.10 1.91 1.67 1.61 1.98 1.84 1.98 1.96 2.15 2.16
Lu 0.29 0.34 0.25 0.32 0.29 0.25 0.24 0.30 0.29 0.30 0.30 0.34 0.32
Hf 2.04 2.49 1.88 2.35 2.25 1.98 1.86 2.32 2.08 2.07 2.03 2.55 2.56
Ta 0.09 0.09 0.09 0.12 0.11 0.07 0.08 0.11 0.09 0.10 0.10 0.14 0.16
Tl 0.07 0.06 0.05 0.03 0.05 0.03 0.05 0.04 0.03 0.03 0.04 0.07 0.06
Pb 2.94 3.32 3.06 2.76 3.21 2.88 2.90 2.95 2.84 2.51 2.38 3.42 3.38
Th 0.65 0.77 0.61 0.49 0.56 0.55 0.59 0.59 0.48 0.48 0.44 0.80 0.78
U 0.42 0.49 0.41 0.33 0.40 0.40 0.40 0.38 0.34 0.32 0.31 0.49 0.46
La/Y-
b
3.14 3.55 3.36 2.92 3.50 3.01 3.30 3.39 3.01 2.69 2.74 3.93 3.79
Ba/Sr 1.01 1.01 0.90 0.76 0.82 0.95 0.90 0.82 0.84 0.77 0.79 1.09 1.06
U/Th 0.64 0.64 0.68 0.68 0.72 0.72 0.68 0.64 0.72 0.66 0.69 0.62 0.59
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8672
tional standards (JA2, JB3) and duplicated meas-
urements the precision (relative 2c error) for ma-
jor elements is estimated at 6 1% (except FeO
and Na2O = 2%) and 6 5% for trace elements
with abundances s 20^30 ppm. Other trace ele-
ments were determined by ICPMS. The interna-
tional standards JB3 (n = 19) and JA2 (n = 8) were
measured as unknowns in the experiments. From
this the 2c error of the method is estimated to be
6 20% for Nb and Ta, 6 10% for Be, Cs, Cu, Hf,
Table 1 (continued)
KLU-14 KLU-15 KLU-16 r r70 recom.
values
JB3
(n = 19)
Stabwn
(%)
dev.
(%)
recom.
values
Stabwn
(%)
Stabwn
(%)
dev.
(%)
13.5 9.8 0.86 0.89 7.21 7.09 5.4 1.6 27.3 27.2 7.5 0.3
0.71 0.51 0.48 0.29 0.81 0.60 10.3 34.3 2.05 2.09 7.2 1.7
27.1 38.1 41.5 30.60 30.03 33.8 32.5 4.0 19.6 20.0 2.0
259 246 244 0.47 0.21 372 382 2.6 126 124 1.6
33 432 857 30.52 30.16 58 62 6.3 436 455 4.2
33.1 40.1 47.6 30.54 30.16 34.3 35.5 3.4 29.5 29.0 1.7
21 110 206 30.62 0.04 36.2 33.5 8.1 130 137 4.8
83 77 0.49 0.27 194 178 8.5 9.2 29.7 28.7 4.8 3.4
89 77 56 0.47 0.24 100 104 3.4 65 62 5.2
18.0 15.0 15.2 0.24 30.18 19.8 19.0 4.2 16.9 15.5 9.0
19.3 15.8 0.94 0.95 15.1 15.4 6.6 1.9 73 76 5.0 3.7
375 334 246 0.85 0.81 403 396 1.8 248 252 1.4
23.1 17.6 0.50 0.40 26.9 24.3 4.4 10.8 18.3 16.5 8.0 11.2
107 75 67 0.35 30.05 98 95 3.5 116 116 0.4
2.00 1.30 0.37 0.16 2.47 2.04 8.6 21.3 9.47 9.29 10.2 1.9
0.58 0.44 0.72 0.63 0.94 0.93 5.6 1.5 4.63 5.01 7.9 7.5
403 308 229 0.92 0.93 245 241 1.7 321 319 0.8
8.42 6.80 0.71 0.70 8.81 8.40 3.3 4.9 15.8 16.2 6.6 2.7
21.5 17.0 0.65 0.63 21.5 21.3 3.7 0.9 32.7 34.3 4.9 4.6
3.61 2.71 0.64 0.63 3.26 3.42 3.4 4.7 3.84 4.14 2.8 7.4
16.7 12.3 0.64 0.65 15.6 16.4 3.6 4.9 13.9 15.7 2.5 11.5
4.27 3.36 0.64 0.65 4.27 4.56 2.9 6.3 3.11 3.39 3.7 8.2
1.33 1.02 0.66 0.72 1.32 1.32 4.2 0.3 0.93 0.92 3.4 1.2
3.88 3.04 0.50 0.36 4.67 4.46 3.7 4.8 3.06 3.06 4.8 0.1
0.64 0.51 0.44 0.24 0.73 0.73 3.2 0.6 0.44 0.48 5.4 7.5
4.04 3.05 0.39 0.12 4.54 4.38 2.7 3.7 2.80 2.78 5.2 0.8
0.83 0.66 0.43 0.21 0.80 0.89 2.8 10.3 0.50 0.60 4.0 16.2
2.38 1.84 0.35 0.05 2.49 2.57 3.2 3.0 1.48 1.70 3.3 13.2
0.34 0.28 0.38 0.10 0.42 0.38 3.0 11.5 0.28 0.25 3.6 11.8
2.26 1.69 0.30 30.04 2.55 2.42 2.8 5.5 1.62 1.60 3.6 1.1
0.33 0.26 0.23 30.16 0.39 0.36 2.6 8.4 0.27 0.24 3.8 10.5
2.57 1.89 0.41 0.15 2.67 2.61 5.6 2.4 2.86 2.95 6.3 3.1
0.14 0.09 0.34 0.19 0.150 0.157 60.9 4.4 0.80 0.72 24.6 10.5
0.05 0.02 0.63 0.52 0.048 0.047 22.6 2.9 0.32 0.33 11.4 3.8
3.66 2.86 0.72 0.43 5.58 5.05 4.8 10.6 19.2 19.2 9.4 0.0
0.79 0.71 0.94 0.87 1.27 1.32 9.3 4.1 5.03 4.84 6.8 4.0
0.49 0.39 0.86 0.70 0.48 0.47 4.4 1.8 2.21 2.19 5.8 0.9
3.73 4.03 0.75 0.71
1.07 0.92 0.93 0.90 0.89
0.62 0.55 30.74 30.74
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 73
Li, Y, Pb, Rb, Tl, Th and U and V5% for the
rare earth elements (REE). Water contents were
determined by Karl Fischer titration (2c around
10%).
Oxygen isotope ratios of olivine were measured
at the Geophysical Laboratory, Carnegie Institu-
tion using a Synrad Series 48 CO2 laser. The tech-
nique is similar to that in [14]. Oxygen, however,
was not converted to CO2 after reaction of sam-
ples with BrF5, but frozen on a molecular sieve.
The isotopic composition of oxygen is determined
on masses 32, 33, and 34 on a Finnigan MAT-252
mass spectrometer. The oxygen standard gas was
calibrated to the SMOW scale using NBS-28
(N18O = 9.60x). Analyses of NBS-28 were per-
formed by rapid heating with a defocused laser
beam [15], in order to avoid grain size e¡ects.
Daily variations were corrected by an in-house
olivine standard (N18O = 5.2x). The external er-
ror of this method is estimated to be better than
þ 0.2x based on ¢ve analyses of NBS-28
(9.57 þ 0.18x, 2c), a mean of 5.85 þ 0.06x for
two analyses of UWG-2, a garnet standard used
at the University of Wisconsin, and an average
reproducibility (2c) better than þ 0.2x for mul-
tiple runs. Single 1^2 mm large Ca-rich pyroxene
crystals were analyzed using an ultraviolet (UV)
ArF laser in Go«ttingen and a Finnigan MAT-251.
On average the reproducibility of pyroxene ana-
lyses is less than analyses of olivines. This may be
due to inclusions of matrix minerals and surface
contamination rather than true heterogeneity. Py-
roxene data have been corrected using the same
olivine standard as described above, ensuring ac-
curate and comparable calibration to the SMOW
scale.
The isotope ratios of Sr and Nd on correspond-
ing whole rocks were measured at a Finnigan
MAT262-RPQIIplus in Go«ttingen. The Sr and
Nd isotope ratios were corrected for mass fractio-
nation to 86Sr/88Sr = 0.1194 and 144Nd/
146Nd = 0.7219 and referenced to NBS987
(0.710245) and La Jolla (0.511847). Measured val-
ues of these standards over the period of the study
were 0.710262 þ 24 (9 analyses) and 0.511847 þ 20
(11 analyses). Lead isotopes were corrected to
NBS981 [16]. Thirteen analyses of this standard
gave a mean of 206Pb/204Pb = 16.90 þ 0.01, 206Pb/
204Pb = 15.44 þ 0.02 and 206Pb/204Pb = 37.53 þ 0.05.
Blanks for Sr, Nd and Pb are 6 1 ng, 6 0.03 ng
and 6 0.5 ng, respectively. From continuous
measurement of standards and repeated measure-
ments of samples, total errors (2c) less than
0.004% for Sr and Nd, and less than 0.1% for
Pb isotopes were determined.
4. Analytical results
4.1. Major and trace elements
Our samples represent HMB and HAB with
SiO2 ranging from 52 to 54 wt% and MgO from
4.6 to 11.7 wt% (Fig. 2, EPSL Online Background
Dataset, Table 1, see footnote 2). The trace ele-
ment pattern is typical for Quaternary volcanic
rocks of Kamchatka and for island arcs in gener-
al, i.e. with relative depletions in HFSE and en-
richment in LILE (Table 1, Fig. 3). From the N-
MORB-normalized NbN/YbN range of 0.8^1.0 we
argue for a mantle source similar to that of the N-
MORB. Low HREE concentrations can be ex-
plained by relatively high melting degrees
(V20%). Sr isotope ratios of Kluchevskoy volca-
no have a narrow range from 0.70351 to 0.70366
and are slightly enriched compared to other Qua-
ternary lavas from Kamchatka (Table 2, Fig. 4).
Fig. 3. N-MORB-normalized trace element pattern of Klu-
chevskoy volcano/Kamchatka (thin lines). Note the typical
arc signature with strong enrichment in LILE and depletion
in HFSE, typical for rocks from Kamchatka (own unpub-
lished data of the CKD are used as reference) and average
arc basalts in general [38]. N-MORB composition and the
order of incompatible elements (enlarged by Cs and all REE)
after [39].
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8674
HAB are more radiogenic in Sr isotopes com-
pared to HMB. Pb and Nd isotopic ratios are
MORB-like (Fig. 4). These results are consistent
with [17].
Major and trace element trends can be approxi-
mated by olivine and clinopyroxene (1:2 mass ra-
tio) fractionation from a parent composition such
as sample KLU-16 (Fig. 2). Compared to HMBs,
the HABs form a tight compositional group,
probably caused by limited sample numbers and
eruptions from the same magma batch close in
time (1951, 1953, and 1966). Combining our
Table 2
Mineral laser oxygen data (in x relative to SMOW) and radiogenic isotope ratios for corresponding whole rocks from Kluchev-
skoy volcano. Melt oxygen isotope ratios (N18O) were calculated from the mean of olivine data (N18Ool), assuming an olivine melt
fractionation of 30.4x
Sample Rock
type
Phenocrysts N18Ool N18Ocpx N18O 87Sr/86Sr 143Nd/144Nd 206Pb/204Pb 207Pb/204Pb 208Pb/204Pb
KLU-01 HMB plag, cpx, ol 6.90 6.51 7.27 0.70356 0.51309 18.31 15.50 37.97
6.90 7.40
6.82 7.43
KLU-03 HMB ol, cpx 6.40 6.68 6.80 0.70357 0.51310 18.28 15.50 37.97
6.36 6.45
6.32
6.50
KLU-04 HMB plag, cpx, ol 5.97 6.21 6.32 0.70353 0.51307 18.28 15.50 37.97
5.86 6.43
KLU-05 HMB plag, cpx, ol 5.93 6.29
5.84
KLU-06 HMB ol, cpx 6.36 6.80 6.80 0.70355 0.51309
6.43 6.44
KLU-07 HMB ol, cpx 6.28 7.15 6.64 0.70355 0.51307 18.30 15.50 37.97
6.19 7.07
6.77
KLU-08 HMB plag, cpx, ol 6.00 5.73 6.41 0.70352 0.51308 18.30 15.52 38.01
6.01 6.49
KLU-09 HMB plag, cpx, ol 5.82 6.40 6.23 0.70351 0.51311
5.83 6.43
6.27
KLU-10 HMB plag, cpx, ol 5.85 6.57 6.22 0.70352 0.51311 18.29 15.51 38.00
5.78 6.46
KLU-11 HMB plag, ol, cpx 5.84 6.28 6.23 0.70353 0.51310 18.29 15.50 37.97
5.82 6.56
KLU-15 HMB ol, cpx 6.95 7.26 7.34 0.70355 0.51312 18.30 15.49 37.94
6.97 6.72
6.89
KLU-16 HMB ol, cpx 5.97 6.43
6.09
KLU-02 HAB plag (ol, cpx) 7.07 7.15 7.43
6.99 7.36
KLU-12 HAB plag (ol, cpx) 6.89 7.66 7.29 0.70366 0.51309 18.30 15.51 38.00
7.22
7.21
KLU-13 HAB plag (ol, cpx) 7.13 7.53 0.70366 0.51309
KLU-14 HAB plag (ol, cpx) 7.05 7.63 7.47 0.70362 0.51309
7.09 7.56
7.73
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 75
own with published data [10,11,18] the HABs
show some scatter in major elements, LILE,
K2O/Na2O, Ba/Zr, La/Yb. This scatter and lack
of correlation between LILE and HFSE do not
allow for a single parental magma. Therefore we
have to assume that processes additional to crys-
tallization are required to explain the major and
trace elements, and isotopic data.
4.2. Oxygen isotopes
The oxygen isotope ratios of single olivines
from di¡erent rocks vary from 5.8x to 7.1x
(Table 2). Replicate samples are reproducible to
within 0.2x. Results for pyroxene give a range
from 5.7 to 7.7x. Pyroxene is on average 0.4x
heavier than olivine from the same rock, consis-
tent with cpx-ol fractionation [19]. Melt values are
calculated from the mean N18O of olivine, using
olivine melt fractionation of 30.4x at 1100^
1200‡C [19]. The resulting range of 6.2^7.5x is
clearly larger and values are distinctly higher than
for typical mantle-derived melts (e.g., [4]).
Fig. 4. Plots of Sr, Nd and Pb isotopic ratios versus N18O.
Symbols as in Fig. 2. Fields for MORB, Iceland, CAB (con-
tinental arc basalts) and IAB (island arc basalts) are derived
from a compilation for fresh basalts of [4]. The Hawaii data
were taken from [21]. The Kamchatka ¢eld (K) includes un-
published isotope data of the authors for the Eastern Vol-
canic Front, Central Kamchatka Depression and Sredinny
Ridge. The altered oceanic crust is assumed to be similar in
143Nd/144Nd and 206Pb/204Pb to the MORB ¢eld. For N18O
and 87Sr/86Sr we used the bulk composition of AOC from
[25]. From all oxygen data, only the N18O values of Hawaii
and Kamchatka were determined by laser ablation.
6
Fig. 5. Correlations of N18O with £uid-mobile elements like
K (shown as oxide), Ba, Rb and Pb. Measured data: open
symbols; fractionation-corrected values: closed symbols. An-
alytical error, regression lines and regression coe⁄cients are
shown for the crystal fractionation-corrected data. Strong
positive correlations for K, Ba and Rb and only a weak pos-
itive trend for Pb with N18O are notable.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8676
Having established this unusual range in O iso-
tope composition, we will now explore correla-
tions between N18O and other geochemical and
isotopic parameters (Figs. 4^7) in order to under-
stand the origin of these variations.
A steep trend is observed between O isotopes
and 87Sr/86Sr (Fig. 4). Historic rocks tend to have
higher N18O and Sr isotopic ratios compared to
Holocene samples. There are also positive corre-
lations between N18O and some trace elements
(Cs, Li, Sr, Ba, Rb, Pb, Th, U, LREE) and K,
i.e. elements which are known to be enriched in
slab-derived £uids (named ‘£uid-mobile’ elements
hereafter). Strong positive correlations are also
observed between O isotopic ratios and oxide or
elemental ratios such as K2O/Na2O, Ba/Zr and
La/Yb (Fig. 7). A negative correlation of N18O
was found with U/Th.
Some of the elemental variability is due to crys-
tal fractionation. To correct for this, all concen-
trations were recalculated to Mg#70 by adding
olivine and clinopyroxene in proportions of 1:2
(Table 1 and EPSL Online Background Dataset,
Table 1, see footnote 2, Figs. 2, 5 and 6; ¢lled
symbols). Mineral major element compositions
were derived from the microprobe data of [10],
trace elements were calculated using melt/liquid
distribution coe⁄cients in [20]. This approach
provides an e¡ective fractionation correction for
all incompatible trace elements and some major
elements (K, Ti, P and Al). A maximum of 30%
fractionation is calculated for most evolved
HABs. Corrected £uid-mobile element concentra-
tions still strongly correlate with N18O. Interest-
ingly, weak correlations between those elements
which are immobile in £uids but can be trans-
ported in melts like Al and HFSE (named ‘melt-
mobile’ elements hereafter) disappear after the
fractionation correction.
5. Discussion
Fractionation of olivine will increase the N18O
of the melt by less than 0.1x and pyroxenes do
not have any detectable e¡ect [19]. Fractionation
of magnetite is insigni¢cant relative to other
phases because depletion in FeO or TiO2 is not
observed. Separation of other minerals such as
plagioclase or quartz would even lower N18O in
the magma. It is obvious that separation of ma¢c
minerals is not responsible for the increased oxy-
gen isotope ratios in the Kluchevskoy lavas.
Remaining options to explain a substantial
change in oxygen isotopes in Kluchevskoy mag-
Fig. 6. Plot of N18O versus melt mobile elements Ti, Na, Al
(shown as oxides), and Zr. Symbols as in Fig. 5. The analyti-
cal error, regression lines and regression coe⁄cients are
shown for the fractionation-corrected data. There are no cor-
relations of this group of elements with N18O.
Fig. 7. Correlations of N18O with trace element ratios. The
analytical error (2c), regression lines and regression coe⁄-
cients are also shown. The ratios were calculated from the
original measured data, because fractionation should have no
in£uence on incompatible element ratios.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 77
mas are: (1) crustal assimilation; (2) source con-
tamination by subducted sediments ; or (3) £uids
from the altered oceanic crust in the mantle
source. These possibilities are evaluated in the
following sections.
5.1. Crustal assimilation
Assimilation of old crust is known to cause 18O
enrichment of volcanic rocks in continental arcs
(e.g., [4]). Even OIBs erupting through thin oce-
anic crust can be strongly contaminated by al-
tered basalts [6]. Therefore assimilation of 18O-
enriched crust needs to be tested to explain the
O isotopic characteristics of Kluchevskoy lavas.
The crust below the CKD consists of Creta-
ceous to Tertiary accreted arc and oceanic mate-
Fig. 8. AFC model for Sr, 87Sr/86Sr and N18O, calculated for
assimilation of primitive island arc basalt magmas by hy-
drated 18O-enriched island arc crust (Table 3) with variable
N18O (8, 10, 12x) but similar Sr and 87Sr/86Sr composition.
Tick marks express F, the remaining melt fraction. Plotted
Sr data are measured uncorrected values. The gray ¢eld rep-
resents the data for basalts from all Kamchatka (unpublished
data of the authors).
Fig. 9. Source contamination model, showing melt composi-
tions derived from a mantle source modi¢ed by variable ad-
ditions of sediment and/or slab £uid (mixing endmembers de-
rived from Table 3) shown in (A) 87Sr/86Sr-Sr and (B) 87Sr/
86Sr^18O space. Lines in A and B represent melts, derived
from 10% melting of such contaminated mantle, the degree
of source contamination is shown by tick marks. Fractiona-
tion-corrected data are plotted in A, to approximate primary
melt composition.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8678
rial [12]. Such young volcanic rocks or volcani-
clastic sediments can be close to mantle values
in Pb and Nd isotopes. Sr and oxygen isotopic
compositions may be more variable, because of
low T exchange with seawater. From dacitic rocks
of the Bonin Islands N18O values up to +15x are
known, whereas pyroxene phenocrysts of the
same rocks give mantle values [22]. Obviously,
arc basalts and OIBs erupting into a submarine
environment can become 18O-enriched by sea-
water alteration.
In Fig. 8 we try to model the apparent correla-
tion between N18O and Sr by AFC (Table 4). The
relative rate of assimilation of arc crust to the rate
Table 3
Parameters used in the mixing and AFC models
Sedimenta DMb KMc IABd Slab £uide Arc crustf
N18O (x) 20 5.7 5.8 5.8 12 8; 10; 12
O (%) 50 50 50 50 85 50
87Sr/86Sr 0.707 0.7028 0.7033 0.7033 0.7036 0.7046
Sr (ppm) 250 16.6 19 200 550 200
aSediment composition assumed to be in equilibrium with Cretaceous sea water, Sr after [24].
bDepleted mantle (DM): N18O and 87Sr/86Sr after Ito et al. (1987), Sr after [28].
cMantle source below Kluchevskoy volcano (KM) calculated as 99% DM+1% sediment.
dPrimitive island arc basalt melt (IAB) used in the AFC modeling, Sr were empirically derived to approximate the observed
trends.
eSlab £uid derived from altered oceanic crust [25,26].
f The N18O value for arc crust corresponds to AOC, 87Sr/86Sr and Sr were empirically derived to approximate the observed
trends.
Table 4
Rock and £uid compositions used in the slab^wedge interaction model (all concentrations in ppm). Di¡erences between the com-
positions of the calculated £uids and those from the literature data [26,40] are caused by slightly di¡erent input parameters in
the modeling
EMa Slabb Slab £uid (SF)c Wedge £uid (WF)d SF/WF SF/EM
[40] [26] [40] [26]
Rb 0.38 9.58 71 1.09 65 186
Ba 10.47 22.6 2210 173 13 211
Th 0.036 0.7 1.69 0.33 5.1 47
U 0.017 0.3 0.59 0.33 1.8 36
Pb 0.090 0.3 15 4.80 3.1 168
Sr 17.8 115 180 575 47 100 4^6 32
Nb 0.36 1.22 0.04 0.36 0.23 2.55 0.1^0.2 0.1^1.0
Zr 8.62 66.5 1.70 22 0.1 0.2
La 0.41 1.84 0.84 0.81 1.0 2.1
Sm 0.40 2.5 0.01 0.04 0.3 0.03
Tm 0.30 0.43 0.001 0.005 0.3 0.005
Y 22.6 26.9 0.03 0.41 0.1 0.001
U/Th 0.46 0.43 0.35 0.99 0.4 0.8
La/Tm 1.36 4.28 612 170 3.6 450
Ba/Sr 0.59 0.20 3.85 1.73 2.2 6.6
aEnriched mantle calculated from depleted mantle [28], contaminated by 0.5% sediment [24].
bSlab composition (upper part of altered oceanic crust) after [25], Pb after [38].
cCalculated £uid (1%) in equilibrium with the slab (59:40:1 by mass garnet:clinopyroxene:rutile), using Dmineral=fluid values of [40]
and [26].
dCalculated £uid (1%) composition in equilibrium with the EM (60:30:10 by mass olivine:orthopyroxene:amphibole), using
Dmineral=fluid values of [40] and [26].
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 79
of crystallization (r) is taken at 0.5. A low Sr
content of only 200 ppm for the primitive magma
(IAB in Fig. 8) is consistent with the lowest Sr
concentrations in most ma¢c Kamchatka lavas
(215 ppm, Bakening volcano, Dorendorf et al.,
submitted). Given the model parameters as in Ta-
ble 3, the AFC model can, in principle, explain
the data only if the Sr isotopic composition of the
parent is signi¢cantly more radiogenic than a de-
pleted MORB source (Figs. 8 and 9).
To produce the most 18O-enriched HAB from a
mantle-derived magma with 18O = 5.7x by AFC,
the magmas need to be extensively fractionated,
between F = 0.9 and F = 0.65 for the di¡erent
AFC curves. This translates into a proportion of
10^23% assimilated ma¢c crust. At this propor-
tion, the assimilant forms up to 45% of the result-
ing magma. Even the most primitive sample
(KLU-16) with N18O = 6.4x would need s 20%
fractionation and s 10% assimilation, which is
not consistent with its primitive Mg# number of
71. An AFC process should also strongly increase
the contents of SiO2 and all incompatible trace
elements. In addition, from thermal balance argu-
ments, it is impossible for a basaltic magma to
assimilate large volumes of basaltic wall rock.
Similar arguments argue against assimilation of
accreted oceanic sediments. We conclude that,
although fractional crystallization has a¡ected
the bulk rock composition, assimilation of crustal
rocks cannot explain the observed high oxygen
isotope ratios and Sr^O isotope trend.
5.2. Source contamination by sediments
It is well established by geochemical and iso-
topic composition in arc rocks that sediments
contribute to arc volcanism (e.g., [1]). Sediments
also have the potential to enrich the mantle in
N18O. Northwest Paci¢c sediments are rich in sili-
ceous ooze and reach N18O values of up to 30x
[23]. At subduction, £uids or melts derived from
such sediments might introduce an 18O-rich com-
ponent into the mantle. In order to test the role of
such sediment component, we calculated simple
binary mixing between oceanic sediments (87Sr/
86Sr = 0.707 and N18O = 20x) and a depleted
MORB mantle [7]. Sediment Sr concentrations
are from [24], the 87Sr/86Sr and N18O are assumed
to be in equilibrium with Cretaceous seawater.
The model (Fig. 9) calculates Sr, 87Sr/86Sr, and
N18O compositions in melts from a mantle source
(DM) that has been modi¢ed by the addition of
subducted sediment. A minor sediment contribu-
tion to the mantle source would drastically change
Sr, Nd, and Pb isotope ratios but not signi¢cantly
change N18O and Sr contents in resulting melts
(almost horizontal dotted lines in Fig. 9). Involv-
ing melts or £uids from sediments would give
similar results. A sedimentary component to ex-
plain the 18O enrichment in Kluchevskoy magmas
is thus ruled out.
5.3. Source contamination by £uids from the
altered oceanic crust
The upper oceanic crust is 18O-enriched by hy-
drothermal alteration and exchange with seawater
(e.g., [25]). Changes in Nd and Pb isotopes in
altered MORB are small due to low Nd and Pb
abundances in seawater. The upper oceanic crust
is further characterized by 87Sr enrichment, which
strongly correlates with N18O [25]. From these
data, a maximum of N18O = 12x is assumed for
the upper (low T) altered oceanic crust, subducted
below Kluchevskoy. Similar values can be in-
ferred for subducted seamount material. Aqueous
£uids released by the breakdown of hydrous min-
erals have been shown experimentally (e.g.,
[26,27]) to be enriched in alkalis, alkaline earths,
U, Th, and Pb but to be essentially free of HFSE
and HREE.
We calculated melt compositions from a mantle
source that is variably enriched in a sediment
component (no sediment = DM; 1% sediment =
KM, see discussion above) and a variable £uid
component with 18O = 12x (1, 2, 5, 10, 20%,
dashed and solid lines in Fig. 9). For parameters
used see Tables 3 and 4.
A large contrast in Sr concentration between
slab-derived £uids (550 ppm Sr [26]) and depleted
mantle (16.6 ppm Sr [28]) results in a strong lev-
erage of the £uid on the mantle Sr isotope com-
position for low £uid proportions (up to 5%).
Beyond 5% £uid addition, this process will cause
a steep trend between Sr and O isotopes because
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8680
of the relatively unradiogenic Sr isotope compo-
sition of the £uid. The calculated trends fall close
to the observed compositions, however the
amount of £uid in the source (5^20 wt%) is rather
high. Such large amounts of hydrous £uid would
cause extensive and unrealistically high degrees of
melting [2]. It was argued for Kamchatka that the
18O-rich component is a hydrous melt rather than
a hydrous £uid [8]. However, hydrous melts are
quite di¡erent in trace elements to hydrous £uids,
particularly in their high HFSE [26]. From trace
element patterns of Kluchevskoy samples and the
correlation between N18O and trace elements (or
lack thereof for HFSE) a role for a hydrous melt
component is ruled out.
Time elapsed between £uid enrichment in the
arc magma source, melting and eruption is in
the order of 30 000^s 150 000 years [29,30].
Therefore in our model, the increase of U/Th,
connected with the depletion in 18O, should be
expressed in a measurable U^Th isotope disequi-
librium. However, in a preliminary study no 238U/
230Th disequilibrium could be found in Kluchev-
skoy samples which argues against a recent
(6 300 000 years) £uid-induced melting event.
6. A N18O-enriched mantle source?
In the previous section we calculated a mixture
between the mantle and a slab-derived £uid, with
the result of unrealistically high amounts of £uid
in the magma source. However, if we allow for
time, the amount of in¢ltrating £uid at any given
stage can be much less than that estimate for the
bulk rock/water ratio. In the next sections we de-
velop the following model :
1. Small amounts of hydrous 18O-enriched £uid
liberated from the slab migrate into the over-
lying non-convecting cold lithosphere of the
mantle wedge. Such small volumes of £uid
can (over time of some m.y.) produce a mantle
that has exchanged at a high £uid/rock ratio. A
proportion of 20% £uid (see above) will then
be possible without extensive melting.
2. The £uid forms hydrous minerals in the man-
tle. Its type and quantity depend on the P^T
conditions in the mantle wedge and the chem-
ical composition (i.e. availability of K, Na, Al)
from the mantle and £uid. Only a small
amount of £uid will be instantly ¢xed in hy-
drous phases, the remaining part will migrate
further into the mantle until it is completely
reacted.
3. Equilibration with mantle mineralogy and the
newly formed hydrous minerals will result in a
concentration pro¢le of 18O and certain trace
elements from the slab into the mantle. In an
extreme case, the mantle close to the slab is
completely equilibrated with the slab £uid
and the £uid far away from the slab is com-
pletely equilibrated with the mantle minerals.
The composition of the mantle and the two
£uids can be modeled by percolation [31] or
zone re¢ning [3]. To simplify the model, we
only discuss the endmember compositions.
Co-variation of trace elements and radiogenic
isotopes with N18O is then produced by mixing
these endmembers.
4. Such metasomatized mantle becomes involved
in arc magmatism due to intra-arc rifting
of the CKD. Mixing of the di¡erently meta-
somatized mantle regions creates the observed
geochemical and isotopic correlations with
N18O.
This model represents a chromatographic pro-
cess which implies the possibility of di¡erent
length scales for the pro¢les of di¡erent elements.
However, we have no independent constraint as
to where melting takes place within the spectrum
of percolation-modi¢ed compositions. To simplify
the model, we only discuss the endmember com-
positions and co-variations of trace elements and
radiogenic isotopes with N18O are then assumed to
be produced by mixing these endmembers.
6.1. Mantle composition
NbN/YbN of Kluchevskoy samples shows that
their source was similar to or even more depleted
than the MORB source (Fig. 3). From the trend
in Sr and O isotopes, it is shown that this source
was contaminated by some limited amounts of
sediment. We calculated the mantle wedge source
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 81
below Kamchatka (KM) as a mixture of depleted
mantle and 1% maximum of sediment (Table 4).
6.2. Slab £uid composition
A critical parameter in our model is the com-
position of the two endmember £uids. The in¢l-
trating slab £uid depends on pressure, tempera-
ture, and mineralogical composition of the slab.
The P^T conditions in the uppermost part of the
slab in the region underlying the lithospheric
mantle can be estimated using the model of [32].
The situation for Kamchatka is similar to the
Aleutians, i.e. fast subduction of an old slab,
which results in a relatively cold P^T path
(T = 200^600‡C at 1^3 GPa). At these conditions,
the slab £uid should be a hydrous £uid. Experi-
ments on trace elements in hydrous £uids were
performed exclusively at temperatures exceeding
900‡C. Fluid/mineral partition data are thus
only available for phases in the deeper part of
the slab, such as garnet, clinopyroxene, amphi-
bole, rutile, orthopyroxene and olivine [3,26,27].
The £uid/mineral partition behavior for other
phases like apatite, lawsonite, chlorite, chloritoid,
epidote, zoisite^clinozoisite, phengite and serpen-
tine, likely to be residual in shallower regions, is
unknown. However, even minor amounts of these
phases may strongly in£uence trace elements in
the £uid. For example K, Rb, Ba and Cs are
carried essentially by phengite [33], Y and REE
by lawsonite and apatite [34]. A more reliable
estimate of the £uid composition is be derived
from studies of high-pressure metamorphic ter-
ranes [33,35]. From these studies it is concluded
that slab £uids released at shallow pressures will
also be enriched in LILE and LREE but poor in
HREE and HFSE and thus be similar to the
deep-derived £uids.
The slab £uid composition was then calculated
from Dmineral=fluid data for 2^3 GPa and 900^
1100‡C of [3,26] assuming 1% £uid dehydration
from an eclogite composed of 40% clinopyroxene,
59% garnet and 1% rutile. These assumptions are
identical to those of [26]. We use the estimates of
[25,36] for elemental compositions of altered
MORB. The N18O of the altered MORB is taken
as +12x, consistent with N18O in water-rich melt
inclusions from mantle xenoliths [37].
6.3. Wedge £uid composition
Fluids derived from the slab will percolate into
the mantle wedge and react with mantle minerals
to form hydrous phases such as serpentine, chlor-
ite, talc, amphibole or phlogopite depending on P
and T. If no further mantle minerals can be
formed, because of a depletion in Na, K, and
Al, the remaining £uid will migrate in unreacted
mantle regions. The trace element content of this
£uid migrating through the mantle depends on its
initial composition and extent of exchange with
mantle minerals. The mantle-equilibrated £uid
composition can then be calculated from partition
coe⁄cients between the mantle minerals and
the £uid [3,26]. Missing data for Dmineral=fluid in
amphibole and orthopyroxene were calculated
from Dcpx=fluid data in [3,26] and combined with
Dcpx=melt, Damph=melt and Dopx=melt from [20]. The
model mantle contains 60% olivine, 30% ortho-
pyroxene, and 10% amphibole; its composition
is shown in Table 4. The limited availability
of Dmineral=fluid restricts the considered elements
to Rb, Ba, Th, U, Pb, Sr, Nb, Zr, Y, La, Sm
and Tm (missing DTm calculated from DTm =
(DEr+DYb)/2).
6.4. Elemental variations with N18O
Combining the three endmembers outlined
above, we can now test the model in explaining
the observed elemental and isotopic correlations.
The slab and wedge £uid (Table 4) interacts
with and creates two di¡erent mantle composi-
tions that may be taken as endmember magma
sources. The degree of mantle enrichment or de-
pletion of a £uid-mobile element by the migrating
£uid is described by the ratio between the concen-
tration of that element in the slab-equilibrated
£uid (CSF) and the concentration in the wedge-
equilibrated £uid (CWF). If CSFsCWF, the trace
element content will decrease with distance to the
slab. These elements should positively correlate
with N18O. We can thus predict the direction of
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8682
elemental variations based on this model for dif-
ferent elements.
6.5. Fluid-mobile elements
From CSFsCWF (Table 4) enrichments can be
predicted for Ba, Sr, Rb, La, Th, U and Pb in the
slab £uid-modi¢ed mantle which (at 5^20% £uid
interaction) also has the appropriate 18O-enriched
isotope composition. Mixing magmas from these
sources (or mixing the distinct mantle domains
before melting) will cause the observed positive
correlations between N18O and slab £uid-enriched
elements. Our model predicts that Rb and Ba,
which have the highest CSF/CWF, should be par-
ticularly well correlated with N18O. This is in fact
what we see (Table 1, Fig. 5). Residual phlogopite
in the wedge would even increase CSF/CWF for Rb
and Ba and retain these elements from the wedge
£uid. This is supported by positive correlations of
ratios of K, Ba and Rb with all other £uid-mobile
elements versus N18O (shown by Ba/Sr in Fig. 7).
All mantle minerals including amphibole have
very low Dmineral=fluid values for Pb. Therefore
the reacting £uid remains undepleted in Pb
(CSFVCWF), which explains that Pb only weakly
correlates with N18O.
U/Th of the £uid migrating though the mantle
is increased because Th is more compatible in
clinopyroxene than U [26]. Garnet would cause
the opposite e¡ect, which suggests that garnet is
not a stable phase in the mantle source. The rel-
atively high U/Th for Kluchevskoy compared to
other volcanoes of Kamchatka may be an indica-
tion that melts derived from a strongly metasom-
atized mantle dominate, as is supported by the
general correlation between U/Th and Ba/Nb in
Kamchatka lavas (unpublished data).
6.6. Melt-mobile elements
The HFSE (Nb, Zr), Y and all REE except La
have CSF6CWF resulting in their relative enrich-
ment in the wedge £uid and a depletion in these
elements in the reacted mantle. However, the pre-
dicted di¡erence is small. Garnet, which has a
major in£uence on the LREE/HREE, is present
in the slab but is rare or absent in the mantle
source. This causes the four times higher La/Tm
in the calculated slab £uid compared to the wedge
£uid and corresponds to the observed positive
correlation of LREE/HREE with N18O (Fig. 7)
and a positive correlation between La/Yb and
Nb/Yb.
6.7. Sr, Pb and Nd isotopes
Sr and Pb contents are low in the mantle but
have high concentrations in both calculated £uids.
Therefore the Sr and Pb isotope ratios of the
metasomatized mantle will be strongly controlled
by the isotope composition of the slab. 87Sr/86Sr
and Pb isotopic ratios show almost no co-varia-
tions with N18O in Kluchevskoy samples. Interest-
ingly, the ¢eld for Pb isotopes from Kluchevskoy
and for Kamchatka as a whole overlaps with the
¢elds of the Hawaii plume (Fig. 4). If we assume
that the Hawaiian plume component has not sig-
ni¢cantly changed isotopically in Pb over the last
100 Ma, we can explain the Pb isotopic ratios in
the Kluchevskoy rocks to be largely derived from
£uids of this altered plume material.
Nd will behave similarly to Sm, which has low
concentrations in the slab £uid compared to the
mantle. Therefore the Nd isotope ratio will be
de¢ned almost entirely by the wedge and a corre-
lation with N18O is not expected. This is observed
in our data.
6.8. Special conditions for Kluchevskoy?
Several speci¢c plate tectonic conditions exist
for Kluchevskoy volcano, which may be the rea-
son that the process of lithospheric 18O enrich-
ment by slab £uids can be detected here:
1. Slab dehydration at shallow levels prior to
melting or 18O-enriched £uids in subducted
oceanic crust are not unique to Kamchatka.
A special feature of subduction below the Klu-
chevskaya Group, however, is the subduction
of the Emperor seamount chain, one of the
largest of its kind on the globe. This chain is
composed essentially of low T hydrothermal
altered pillow lavas and volcaniclastics. Fast
subduction of the relatively old and cold Pacif-
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 83
ic plate then should favor the £ux of 18O-en-
riched slab £uids into deeper parts of the man-
tle [32]. The large magma production rate and
the observed 18O-enriched compositions in the
CKD may all be a consequence of both high
£uid input over time and intra-arc extension
just where the seamount chain is subducted.
2. It should be easier to trace the process of
wedge enrichment for Kamchatka than in oth-
er arcs on thick continental crust or with sig-
ni¢cant sediment subduction.
3. In most arc systems, the magma is mostly de-
rived from the convecting wedge, and a contri-
bution from the subcontinental lithospheric
mantle should be minor. Intra-arc rifting in
Kamchatka (CKD) may cause that lithospheric
mantle regions becomes involved in arc mag-
matism (Fig. 10).
7. Conclusions
Oxygen isotope ratios of olivines and clinopyr-
oxenes from the Kluchevskoy volcano show dis-
tinctly higher values than for MORB. Because
assimilation of signi¢cant arc crust is ruled out,
the observed geochemical and isotopic features
must be a characteristic of the arc magma source.
This source is quite di¡erent from the MORB-
type mantle. Strong positive correlations between
N18O and £uid mobile major (K, Si) and trace
elements (Cs, Li, Sr, Rb, Ba, Th, U, LREE) argue
for a depleted mantle source that has exchanged
with an isotopically heavy £uid. Sr and Pb are
mainly £uid-derived while Nd isotopic composi-
tions represent unmodi¢ed mantle. Similarities be-
tween Kluchevskoy rocks and the Hawaii plume
component in Pb isotopic compositions show that
no additional sediment is necessary to explain the
slight radiogenic enrichment compared to MORB
and that the source for the ‘heavy’ £uid may be
the dehydrating seamount chain.
The large amount of £uid needed clearly argues
against a simple £uid-induced melting. Melt-
ing and £uid transfer from the slab is de-coupled
in time and space. Enrichment of the non-con-
vecting sub-arc lithosphere over at least 10 m.y.
is necessary to produce a su⁄ciently 18O-enriched
mantle.
Acknowledgements
Tim Elliot and Chris Hawkesworth provided
very helpful reviews, Richard Arculus’ comments
and criticism were extremely constructive, quite
substantial, and ^ in fact ^ benevolent. To all go
our sincere thanks for the time and consideration
they invested in their review. Among our Russian
friends, Tanya Churikova and Alexander Kolos-
kov were instrumental in obtaining this represen-
tative sample set from Kluchevskoy and other
volcanoes in Kamchatka. This work was funded
by the Deutsche Forschungsgemeinschaft (Wo
365/15-1,-2).[AH]
Fig. 10. Model of sub-arc lithospheric metasomatism which
can explain the strong 18O and £uid compatible element en-
richment in primitive basaltic andesites of Kluchevskoy vol-
cano. Special features of the subducting slab (old low T al-
tered ocean island chain, fast subduction) led to an in£ux of
large amounts of £uid in the mantle. The lithospheric part of
the mantle is too cold to melt but is enriched in £uid com-
patible elements and 18O. A metasomatized mantle can be
presently involved in arc magmatism by: (1) assimilation of
enriched lithospheric mantle by melts from the asthenospher-
ic mantle or (2) partial melting of the enriched lithospheric
mantle, caused by the recent intra arc rifting. Cases 1 and 2
suppose an enriched lithospheric mantle below the CKD. A
mass balance for 18O, the lack of a strong U^Th disequili-
brium and the primitive Pb, Sr and Nd isotopic ratios de-
mand that the time of the mantle metasomatism is at least
some Ma but not more than 100 Ma.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8684
References
[1] T. Plank, C.H. Langmuir, Tracing trace elements from
sediment input to volcanic output at subduction zones,
Nature 362 (1993) 739^743.
[2] E. Stolper, S. Newman, The role of water in the petro-
genesis of Mariana Trough magmas, Earth Planet. Sci.
Lett. 121 (1994) 293^325.
[3] J. Ayers, Trace element modeling of aqueous £uid-peri-
dotite interaction in the mantle wedge of subduction
zones, Contrib. Mineral. Petrol. 132 (1998) 390^404.
[4] R.S. Harmon, J. Hoefs, Oxygen isotope heterogeneity of
the mantle deduced from global 18O systematics of basalts
from di¡erent geotectonic settings, Contrib. Mineral. Pet-
rol. 120 (1995) 95^114.
[5] D. Mattey, D. Lowry, C. Macpherson, Oxygen isotope
composition of mantle peridotite, Earth Planet. Sci.
Lett. 128 (1994) 231^241.
[6] J.M. Eiler, K.A. Farley, J.W. Valley, E. Hauri, H. Craig,
S.R. Hart, E.M. Stolper, Oxygen isotope variations in
ocean island basalt phenocrysts, Geochim. Cosmochim.
Acta 61 (1997) 2281^2293.
[7] E. Ito, W.M. White, C. Goepel, The O, Sr, Nd and Pb
isotope geochemistry of MORB, Chem. Geol. 62 (1987)
157^176.
[8] F. Pineau, M.P. Semet, N. Grassineau, V.M. Okrugin, M.
Javoy, The genesis of the stable isotope (O, H) record in
arc magmas: the Kamtchatka’s case, Chem. Geol. 135 (1-
2) (1999) 93^124.
[9] A.B. Kersting, R.J. Arculus, T.K. Kyser, V.M. Okrugin,
A.P. Khrenov, S.A. Fedotov, Klyuchevskoy Volcano,
Kamchatka, USSR; geochemical evidence for crustal con-
tamination?, in: W. Compston (Ed.), Seventh Interna-
tional Conference on Geochronology, Cosmochronology
and Isotope Geology, Abstracts, Geological Society of
Australia 27 (1990) 55, Geological Society of Australia,
Sydney, N.S.W.
[10] A.B. Kersting, R.J. Arculus, Klyuchevskoy Volcano,
Kamchatka, Russia; the role of high-£ux recharged,
tapped, and fractionated magma chamber(s) in the genesis
of high-Al2O3 from high-MgO basalt, J. Petrol. 35 (1994)
1^41.
[11] A.A. Ariskin, G.S. Barmina, A.Y. Ozerov, R.L. Nielsen,
Genesis of high-alumina basalts from Kluchevskoi volca-
no, Petrology 3 (1995) 496^521.
[12] S.A. Fedotov, Y.P. Masurenkov, Active Volcanoes of
Kamchatka, Moscow, 1991.
[13] V.B. Piip, Y.A. Ye¢mova, Seismic sections of the Earth’s
crust under volcanoes of Kamchatka, Int. Geol. Rev. 35
(1993) 170^177.
[14] Z.D. Sharp, A laser-based microanalytical method for the
in situ determination of oxygen isotope ratios of silicates
and oxides, Geochim. Cosmochim. Acta 54 (1990) 1353^
1357.
[15] M.J. Spicuzza, J.W. Valley, M.J. Kohn, J.P. Girard, A.M.
Fouillac, The rapid heating, defocused beam technique; a
CO2-laser-based method for highly precise and accurate
determination of delta 18O values of quartz, Chem. Geol.
144 (1998) 195^203.
[16] W. Todt, R.A. Cli¡, A. Hanser, A.W. Hofmann, 202Pb-
205Pb spike for Pb isotope analysis, Terra Cognita 4
(1984) 209.
[17] A.B. Kersting, R.J. Arculus, Pb isotope composition of
Klyuchevskoy Volcano, Kamchatka and North Paci¢c
sediments; implications for magma genesis and crustal
recycling in the Kamchatkan arc, Earth Planet. Sci.
Lett. 136 (1995) 133^148.
[18] S.A. Khubunaya, S.O. Bogoyavlenskiy, T.Y. Novogor-
odtseva, A.I. Okrugina, The mineralogy of the Kluchev-
skoi magnesian basalts depicting the fractionation in the
magma chamber, Volcanol. Seismol. 15 (1994) 315^338.
[19] Y.F. Zheng, Calculation of oxygen isotope fractionation
in anhydrous silicate minerals, Geochim. Cosmochim.
Acta 57 (1993) 1079^1091.
[20] G. Website, http://www-ep.es.llml.gov/germ/partitioning.
html#Ds
[21] J.M. Eiler, K.A. Farley, J.W. Valley, A.W. Hofmann,
E.M. Stolper, Oxygen isotope constraints on the sources
of Hawaiian volcanism, Earth Planet. Sci. Lett. 144 (1996)
453^468.
[22] P.F. Dobson, J.R. O’Neil, Stable isotope compositions
and water contents of boninite series volcanic rocks
from Chichi-jima, Bonin Islands, Japan, Earth Planet.
Sci. Lett. 82 (1987) 75^86.
[23] D. Rea, I.A. Basov, L.A. Krissek, Leg 145 Scienti¢c
party, Scienti¢c results odf drilling the North Paci¢c
transect, Proceedings of the Ocean Drilling Program, Sci-
enti¢c Results 145 (1995) 577^595.
[24] J.C. Bailey, Role of subducted sediments in the genesis of
Kurile-Kamchatka island arc basalts; Sr isotopic and el-
emental evidence, Geochem. J. 30 (1996) 289^321.
[25] H. Staudigel, T. Plank, B. White, H.U. Schmincke, Geo-
chemical £uxes during sea£oor alteration of the basaltic
upper oceanic crust; DSDP sites 417 and 418, in: G.E.
Bebout, D.W. Scholl, S.H. Kirby, J.P. Platt (Eds.), Sub-
duction Top to Bottom, Geophysical Monograph 96
(1996) 19^38, American Geophysical Union, Washington,
DC.
[26] J.M. Brenan, H.F. Shaw, F.J. Ryerson, D.L. Phinney,
Mineral-aqueous £uid partitioning of trace elements at
900 degrees C and 2.0 GPa; constraints on the trace ele-
ment chemistry of mantle and deep crustal £uids, Geo-
chim. Cosmochim. Acta 59 (1995) 3331^3350.
[27] R. Stalder, S.F. Foley, G.P. Brey, I. Horn, Mineral-aque-
ous £uid partitioning of trace elements at 900‡C^1200‡C
and 3.0 GPa to 5.7 GPa new experimental data for garnet
clinopyroxene and rutile and implications for mantle
metasomatism, Geochim. Cosmochim. Acta 62 (1998)
1781^1801.
[28] M.T. McCulloch, V.C. Bennett, Progressive growth of the
Earth’s continental crust and depleted mantle; geochem-
ical constraints, Geochim. Cosmochim. Acta 58 (1994)
4717^4738.
[29] T. Elliott, T. Plank, A. Zindler, W. White, B. Bourdon,
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^86 85
Element transport from slab to volcanic front at the Ma-
riana Arc, J. Geophys. Res. B Solid Earth Planets 102
(1997) 14991^15019.
[30] S. Turner, F. McDermott, C. Hawkesworth, P. Kepezhin-
skas, A U-series study of lavas from Kamchatka and the
Aleutians: constraints on source composition and melting
processes, Contrib. Mineral. Petrol. 133 (1998) 217^
234.
[31] O. Navon, E. Stolper, Geochemical consequences of melt
percolation; the upper mantle as a chromatographic col-
umn, J. Geol. 95 (1987) 285^307.
[32] S.M. Peacock, Large-scale hydration of the lithosphere
above subducting slabs, in: J.L.R. Touret, A.B. Thomp-
son (Eds.), Fluid-Rock Interaction in the Deeper Conti-
nental Lithosphere, Chem. Geol. 108 (1993) 49^59, Else-
vier, Amsterdam.
[33] S.S. Sorensen, J.N. Grossman, M.R. Per¢t, Phengite-
hosted LILE enrichment in eclogite and related rocks;
implications for £uid-mediated mass transfer in subduc-
tion zones and arc magma genesis, J. Petrol. 38 (1997) 3^
34.
[34] R. Tribuzio, B. Messiga, R. Vannucci, P. Bottazzi, Rare
earth element redistribution during high-pressure-low-
temperature metamorphism in ophiolitic Fe-gabbros
(Liguria, northwestern Italy); implications for light REE
mobility in subduction zones, Geology (Boulder) 24
(1996) 711^714.
[35] G.E. Bebout, The impact of subduction-zone metamor-
phism on mantle-ocean chemical cycling, in: H. Staudigel,
F. Albarede, D. Hilton, T. Elliott (Eds.), The Mantle-
Ocean Connection, Chem. Geol. 126 (1995) 191^218,
Elsevier, Amsterdam.
[36] G.E. Bebout, M.D. Barton, Metasomatism during sub-
duction; products and possible paths in the Catalina
Schist, California, in: J.L.R. Touret, A.B. Thompson
(Eds.), Fluid-Rock Interaction in the Deeper Continental
Lithosphere, Chem. Geol. 108 (1993) 61^92, Elsevier, Am-
sterdam.
[37] J.M. Eiler, B. McInnes, J.W. Valley, C.M. Graham, E.M.
Stolper, Oxygen isotope evidence for slab-derived £uids in
the sub-arc mantle, Nature 393 (1998) 777^781.
[38] M.T. McCulloch, A.J. Gamble, Geochemical and geody-
namical constraints on subduction zone magmatism,
Earth Planet. Sci. Lett. 102 (1991) 358^374.
[39] A.W. Hofmann, Chemical di¡erentiation of the Earth;
the relationship between mantle, continental crust, and
oceanic crust, in: E. Welin (Ed.), Isotope Geochemistry;
The Crafoord Symposium, Earth Planet. Sci. Lett. 90
(1988) 297^314, Elsevier, Amsterdam.
[40] J.C. Ayers, S.K. Dittmer, G.D. Layne, Partitioning of
elements between peridotite and H2O at 2.0^3.0 GPa
and 900^1100 degrees C, and application to models of
subduction zone processes, Earth Planet. Sci. Lett. 150
(1997) 381^398.
EPSL 5329 31-12-99
F. Dorendorf et al. / Earth and Planetary Science Letters 175 (2000) 69^8686