JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
PAGES 25–37
2001
Re–Os Isotopes in the Horoman Peridotite:
Evidence for Refertilization?
A. E. SAAL1∗, E. TAKAZAWA2, F. A. FREY3, N. SHIMIZU1 AND
S. R. HART1
1
DEPARTMENT OF GEOLOGY AND GEOPHYSICS, WOODS HOLE OCEANOGRAPHIC INSTITUTION, WOODS HOLE,
MA 02543, USA
2
DEPARTMENT OF GEOLOGY, FACULTY OF SCIENCE, NIIGATA UNIVERSITY, 2-8050 IKARASHI, NIIGATA 950-2181,
JAPAN
3
DEPARTMENT OF EARTH, ATMOSPHERIC AND PLANETARY SCIENCES, MASSACHUSETTS INSTITUTE OF
TECHNOLOGY, CAMBRIDGE, MA 02139, USA
RECEIVED NOVEMBER 22, 1999; REVISED TYPESCRIPT ACCEPTED JUNE 27, 2000
Re–Os isotopic data for 20 samples from a well-characterized
140 m section across a layered sequence, ranging from plagioclase
lherzolite through lherzolite to harzburgite, of the Horoman peridotite
show: (1) a range in 187Os/188Os ratios (from 0·1158 to 0·1283)
similar to that reported for other peridotitic massifs, thereby suggesting
that the processes responsible for the Re–Os isotopic variation at
the meter-scale and the whole-massif scale are similar; (2) that the
Os isotopic ratio is controlled by the Re content through radiogenic
ingrowth over a period of >0·9 Gy. The ultramafic and some of
the mafic rocks (Type I layers) from the Horoman massif define an
‘apparent age’ of 1·12 ± 0·24 Ga in the Re–Os isochron diagram,
within error of the previously reported age of 833 ± 78 Ma based
on Sm–Nd isotopes. Although the Re–Os isotopic data do not define
an isochron, the consistency of the >900 Ma age defined by both
isotopic systems suggests that this age has a geologic meaning and
that mafic (Type I layers) and ultramafic rocks are genetically
related. A plausible explanation for the genetic relationship between
the mafic and ultramafic rocks, the meter-scale compositional variations from lherzolite to plagioclase lherzolite, the suprachondritic
187
Re/188Os ratios in some fertile peridotites, and the oldest Re
depletion model age of >1·86 Ga obtained for Horoman rocks is
a refertilization process involving reaction of a mid-ocean ridge
basalt-like magma with depleted lithospheric mantle at >900 Ma.
KEY WORDS:
Re–Os isotopes; Horoman peridotite
∗Corresponding author. Present address: Lamont–Doherty Earth Observatory, Room 58, Geochemistry Building, Columbia University, PO
Box 1000, Palisades, NY 10964-8000, USA. Telephone: (914) 3658712. Fax: (914) 365-8155. E-mail: [email protected]
INTRODUCTION
The Re–Os isotope system provides new insights into
understanding mantle processes. Parent and daughter
elements for other isotopic systems (Rb–Sr, Sm–Nd and
U–Pb) behave incompatibly during mantle melting, being
mainly controlled by silicate phases. In contrast, Re
behaves as a moderately incompatible element, whereas
Os is highly compatible during melting, and both Re
and Os have chalcophile and siderophile affinities (Hart
& Ravizza, 1996; Burton et al., 1998, 1999; Shirey &
Walker, 1998). Consequently, mantle peridotites have
higher Os concentrations than their partial melts, and the
Os isotope system is more resistant to young metasomatic
disturbances than are the other isotopic systems (Luck &
Allègre, 1991; Reisberg et al., 1991; Reisberg & Lorand,
1995; Hart & Ravizza, 1996; Roy-Barman et al., 1996;
Burton et al., 1998, 1999; Shirey & Walker, 1998).
In this paper, we report Re–Os isotope data for the
Horoman massif (Fig. 1). We selected a well-characterized
140 m section (the Bozu section) across this layered
peridotite with rock types ranging from plagioclase lherzolite through lherzolite to harzburgite; two mafic layers
from another part of the Horoman peridotite were also
studied. Takazawa and co-workers have studied this
section and mafic layers from the Horoman massif,
reporting major, trace element and isotopic compositions
of whole rocks and mineral separates (Takazawa et al.,
 Oxford University Press 2001
JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
JANUARY 2001
Fig. 1. Geological map of the Horoman peridotite (from Takazawa et al., 1999). Inset shows the location of the studied area.
1992, 1994, 1996, 1999, 2000; Takazawa, 1996). The
main objectives of this Re–Os study are to determine
the origin of the plagioclase lherzolites (do they represent
unmelted fertile peridotite or a region of melt impregnation and accumulation?), to evaluate the role of
mafic layers in the formation of the layered peridotites,
and to constrain the ages of magmatic processes that
affected this peridotitic massif.
the peridotites are increasingly depleted in a basaltic
component. The boundary between the compositional
layers is sharp on a megascopic scale, but transitional
over a few centimeters on a microscopic scale (Takazawa
et al., 2000).
The massif has been divided into two main zones, the
Upper and Lower Zones. Between these zones there is
a continuous transition with no evidence for a thrust or
shear zone (Niida, 1974, 1984). The Lower Zone, >2 km
thick, consists of cyclic layers of plagioclase lherzolite
through lherzolite to harzburgite, with subordinate dunite; the cycle repeats every 100–500 m. The abundance
of mafic layers in the Lower Zone is sparse, with the
exception of thin seams in the plagioclase lherzolite. In
these seams plagioclase occurs in a fine-grained aggregate
of plagioclase, olivine, spinel, minor orthopyroxene and
very rare clinopyroxene. This assemblage is interpreted
to have formed by a decompression reaction between
two pyroxenes and spinel, which in turn were derived
from a reaction between garnet and olivine (e.g. Ozawa
& Takahashi, 1995; Takahashi, 1997). The Upper Zone,
>1 km thick, is characterized by a centimeter-scale layering of plagioclase lherzolite and harzburgite as the
Geological background
The Horoman peridotite (Fig. 1) is a fault-bounded
mantle slice, 8 km × 10 km × 3 km, emplaced 23 ±
1·2 My ago (Rb–Sr isotopes on a phlogopite-bearing
spinel lherzolite, Yoshikawa et al., 1993) in the southern
end of the high-temperature low-pressure type Hidaka
Metamorphic Belt, Hokkaido, Japan (Niida, 1974, 1984).
The peridotite consists of several lithological sequences
of plagioclase lherzolite–lherzolite–harzburgite ± dunite–harzburgite–lherzolite–plagioclase lherzolite (Komatsu & Nochi, 1966; Niida, 1974; Obata & Nagahara,
1987; Takahashi, 1991; Takazawa et al., 2000). From
plagioclase lherzolite through lherzolite to harzburgite,
26
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
continuous ascent of a mantle peridotite diapir (Takahashi, 1992);
(4) polybaric melting of an upwelling fertile mantle
(plagioclase lherzolite), followed by subsolidus thinning
and folding of the residual harzburgite, lherzolite and
plagioclase lherzolite (Takazawa et al., 2000; Yoshikawa
& Nakamura, 2000).
In the Horoman peridotite, there are two dominant
types of mafic layers (Niida, 1984; Takazawa et al., 1999):
Type I (Al–Ti-augite type mafic granulites): plagioclase
+ Ti augite + olivine + orthopyroxene + Ti pargasite
(or kaersutite) + green spinel + titaniferous magnetite
+ ilmenite + sulfide;
Type II (Cr-diopside type mafic granulite): plagioclase
+ Cr diopside + olivine + orthopyroxene + pargasite
+ spinel + magnetite + sulfide.
The mafic layers vary from a few centimeters to several
meters in thickness, and are usually oriented parallel to
the foliation plane of the peridotite. The primary mineral
assemblages in the mafic layers (now transformed by
subsolidus reaction) indicate multiple episodes of melt
injection over a wide range of pressure from the garnet
stability field (Type I) to the plagioclase stability field
(Type II). Type I mafic layers have intralayer compositional heterogeneity, with the centers interpreted as
garnet clinopyroxenite cumulates and the margins as
mid-ocean ridge basalt (MORB)-like melt in equilibrium
with these cumulates, whereas Type II layers are interpreted as plagioclase-rich cumulates (Takazawa et al.,
1999).
Fig. 2. (a) Whole-rock mg-number [100Mg/(Mg + Fe)] along the
Bozu section, Horoman peridotite. Dashed vertical lines indicate the
contacts between lithologies. The shaded regions within the lherzolite
and plagioclase lherzolite layers indicate phlogopite-bearing lherzolite
and E-Type plagioclase lherzolite, respectively. Figure after Takazawa
(1996) and Takazawa et al. (2000). (b) Chondrite-normalized REE
contents for representative harzburgite (Bz-125L), lherzolite (Bz-143),
and E- and N-Type plagioclase lherzolite (Bz-260 and Bz-253, respectively) from the Bozu section. C1 chondrite values from McDonough & Sun (1995). Data from Takazawa et al. (2000). Β, N-Type
plagioclase lherzolite; Χ, E-Type plagioclase lherzolite; Φ, lherzolite;
crossed square, harzburgite.
The Bozu section
We chose the Bozu section for our study, as it is a
representative 140 m section from the Lower Zone. This
section is a continuously exposed lithologic sequence
ranging from harzburgite through lherzolite to plagioclase
lherzolite, and it has been well characterized, with petrologic and geochemical data obtained on scales ranging
from micrometers to tens of meters [Takazawa et al.
(2000) and references therein] (Fig. 2a). On the basis of
study of the Bozu section, Takazawa and co-workers
considered that the layered structure was produced by
melting of an upwelling fertile mantle [the N (normal)Type plagioclase lherzolite] at >0·9 Ga. Subsequently,
within the lithosphere and the garnet stability field, the
Horoman massif reacted with a light rare earth element
(LREE)-rich melt or fluid, producing enrichment of incompatible elements in harzburgite, lherzolite and in a
special type of plagioclase lherzolite that Takazawa et al.
(2000) defined as E (enriched)-Type (Fig. 2b). Although
the E-Type plagioclase lherzolite is enriched in incompatible elements, the abundances of major and moderately incompatible trace elements are similar to those
dominant rock types, with subordinate dunite. In the
Upper Zone, plagioclase lherzolite and mafic layers are
more abundant than in the Lower Zone (Takazawa et
al., 1999). On the basis of core compositions of pyroxenes,
the plagioclase lherzolite equilibrated at >20 kbar and
temperatures of 900–950°C in the Lower Zone and
1100–1150°C in the Upper Zone (Ozawa & Takahashi,
1995).
Several hypotheses have been proposed to explain the
formation of the lithologically layered peridotites:
(1) formation of cumulates through fractional crystallization (Niida, 1984);
(2) melting of lherzolite to create residual harzburgite
and complementary plagioclase lherzolite, which represents a region of melt accumulation (Obata & Nagahara, 1987);
(3) melt segregation processes governed by suction
associated with local melting along a fracture, during
27
28
127·0
120·0
117·0
115·0
109·0
103·0
101·0
97·7
94·7
89·3
87·9
86·2
84·3
N-Type Bz-262
N-Type Bz-261
E-Type Bz-260
E-Type Bz-259
E-Type Bz-258
E-Type Bz-257
N-Type Bz-256
N-Type Bz-255
N-Type Bz-254
N-Type Bz-253
N-Type Bz-252
N-Type Bz-251
N-Type Bz-250
39·9
3·530
G2
0·245
0·035
14·4
6·65
0·0014
2·010
0·788
0·121
86·4
34·8
0·0069
1·17
1·32
1·16
1·19
1·41
46·21
44·43
0·914
0·913
0·905
0·904
0·899
0·895
0·907
0·900
0·898
0·901
0·909
0·907
0·910
0·50
1·42
2·04
2·98
3·18
3·53
3·30
3·23
3·54
3·34
1·90
2·07
1·95
1·88
3·19
3·15
3·14
cpx
0
4·17
0·66
12·06
14·10
15·83
14·35
14·16
15·69
15·54
4·72
7·81
7·15
6·30
13·12
12·79
14·28
(wt %)
Yb
108·5
129·9
294
320
359
355
342
384
360
138
141
170
145
347
321
326
(ppb)
12·80
15·90
45·72
0·725
0·749
0·914
16·55
12·54
0·46
0
2730
1720
Sc
54·80
7·52
8·10
10·43
11·27
13·35
13·70
14·24
16·19
14·01
14·47
14·63
9·73
11·05
11·13
9·84
13·78
13·97
13·54
(ppm)
Pd
0·557
0·513
5·00
4·88
6·74
6·75
5·98
(ppb)
Ir
2·96
2·94
1·93
3·09
1·47
1·83
1·77
(ppb)
TMA are model ages calculated with respect to ‘primitive upper mantle’ 187Re/188Os = 0·428 and 187Os/188Os = 0·1290 ( McDonough & Sun, 1995; Meisel et al., 1996),
and a decay constant of 1·64 × 10−11 yr−1. TRD are Re depletion model ages assuming 187Re/188Os (sample) = 0. Major and trace element contents are from
Takazawa et al. (1999, 2000) with the exception of Pd and Ir contents, taken from Rehkämper et al. (1999). Modal proportion of clinopyroxene have been calculated
by mass balancing the mineral compositions with that of the whole rock (Takazawa et al., 2000). TMA model age is obtained using the measured Re/Os ratios of
the sample to extrapolate the present-day Os isotopic composition back in time until it equals that for a model of the isotopic evolution of the mantle. TRD model
age is calculated in the same way as the TMA model age, but assumes that the Re/Os ratio of the sample is equal to zero.
0·233
G9
Type 1 mafic layers
4·61
1·37
2·27
2·25
42·24
40·39
−24·93
1·41
39·43
38·70
40·06
40·18
3·40
2·07
3·51
9·66
39·58
38·96
−0·89
0·79
43·12
42·53
42·91
2·90
3·04
2·41
0·910
0·900
0·906
0·901
Al2O3
(wt %)
NUMBER 1
0·0066
0·0108
39·31
−4·12
43·03
40·94
1·61
40·74
1·44
−4·54
mg-no.
4·1
0·119
1·32
1·40
0·90
MgO
(wt %)
Bz-125
0·0022
0·180
0·163
0·155
0·63
0·78
0·43
0·48
0·90
0·10
0·32
1·86
1·25
1·14
1·27
0·44
0·53
0·71
TMA
(Ga)
13·1
3·86
0·120
0·119
0·123
0·441
0·332
0·341
0·371
0·391
0·373
0·585
0·156
0·254
0·228
0·092
0·476
0·480
0·217
(Ga)
Re/188Os TRD
11·6
0·0087
0·0373
0·0339
0·0322
0·125
0·124
0·126
0·126
0·123
0·128
0·127
0·116
0·120
0·121
0·120
0·126
0·125
0·124
187
9·1
4·31
4·24
3·98
0·092
0·069
0·071
0·077
0·081
0·077
0·121
0·0324
0·053
0·047
0·0191
0·099
0·100
0·045
Os/188Os
Bz-116
0·161
0·143
0·128
2·44
3·44
4·5
4·02
3·12
3·48
2·91
4·65
2·92
3·11
5·18
3·22
2·89
187
Harzburgite
Bz-134R
72·3
Bz-134
0·223
0·237
0·318
0·310
0·253
0·270
0·353
0·151
0·154
0·147
0·099
0·318
0·288
0·205
Re/Os
VOLUME 42
Bz-143
Lherzolite
132·0
N-Type Bz-263
4·57
(ppb)
(m)
(ppb)
Os
Distance Re
Plagioclase lherzolite
Sample no.
Table 1: Re and Os concentrations and Os isotopic composition of 20 samples from the Horoman peridotite
JOURNAL OF PETROLOGY
JANUARY 2001
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
fire-assay technique using a 4:1 flux-to-sample ratio.
The Os concentrations and isotopic compositions were
measured by negative thermal ionization mass spectrometry on NIMA-B [at Woods Hole Oceanographic
Institution (WHOI)], with oxygen-enhanced emission
and single-collector analog detection using an electron
multiplier and a dynamic collection routine. For the Re
analysis, 1 g of powder was dissolved in two steps using
first a HF–HNO3 mixture and later HCl; the Re was
extracted by a simple ion exchange technique. The Re
measurements were performed on a Finnigan MAT
Element ICP-MS (WHOI) by rapid peak hopping using
electrostatic scanning.
In-run precision is 2 <0·35% for 187Os/188Os, 2 <1%
for Os concentration and 2 <2% for Re concentration.
Analyses were corrected for total procedural blank of
4·8–5·6 pg for Os concentration and 2–3 pg for Re
concentration with an Os isotopic ratio of 0·2–0·26. The
correction is insignificant for Re and Os contents or Os
isotopic composition, with the exception of samples Bz116 and Bz-125, which have an estimated 2 error of
>25% for Re concentration as a result of the blank
correction. Re and Os replicates on separate powder
splits of Bz-134 differ by <10% for Re, <2% for Os
concentration, and <0·5% for 187Os/188Os ratios.
Fig. 3. (a) 187Os/188Os in Horoman peridotites and in peridotites from
different orogenic lherzolite massifs, compiled by Roy-Barman et al.
(1996). Compilation does not include mafic layers. (b) 187Os/188Os vs
distance along the Bozu section. It should be noted that the extreme
variation in Os isotopic composition occurs between E- and N-Type
plagioclase lherzolites separated by only 5 m. Symbols as in Fig. 2.
RESULTS
Results for 20 ultramafic and mafic samples analyzed
from the Horoman peridotite are listed in Table 1; wholerock major and trace element compositions have been
reported by Takazawa et al. (1999, 2000). In the Bozu
section, the 187Os/188Os ratios of the ultramafic rocks
range from 0·1158 to 0·1283; a similar range was reported
by Liu & Tanaka (1999), for another suite of samples
from the Horoman massif. The Re and Os concentrations
vary from 0·007 to 0·350 ppb and from 2·9 to 5·2 ppb,
respectively. Thus, whereas the Re content varies by a
factor of 50, the Os concentration varies by about a
factor of two; results that are typical of peridotite massifs
(Luck & Allègre, 1991; Reisberg et al., 1991; Reisberg &
Lorand, 1995; Roy-Barman et al., 1996).
Two major observations are: (1) the range of 187Os/
188
Os ratios in this 140 m section of peridotite is as large
as those reported for peridotite massifs, such as Beni
Boussera, Baldisero, Lanzo, Lherz and Ronda (Reisberg
et al., 1991; Reisberg & Lorand, 1995; Roy-Barman et
al., 1996) (Fig. 3a); (2) among the samples analyzed, the
lowest and highest Os isotopic ratios measured correspond to E-Type and a N-Type plagioclase lherzolites
separated by a distance of only 5 m (Fig. 3b).
The 187Os/188Os ratios and the Re contents define a
negative correlation with the MgO concentrations (Fig.
4). If the varying MgO content of these peridotites
of the lherzolite. During uplift of the massif, the peridotite
was transformed from the garnet- to the spinel- to
the plagioclase-facies mineralogy by subsolidus reaction
(Takazawa et al., 1996).
We selected two harzburgite, two lherzolites, four EType plagioclase lherzolites and 10 plagioclase lherzolites
for determination of Re–Os isotopic systematics. With
the exception of the plagioclase lherzolites, where we
analyzed all the samples available, the peridotite samples
are from the center of their respective layers. In addition,
we analyzed two Type I mafic layers; one each from the
Lower and Upper Zones.
ANALYTICAL TECHNIQUES
The analytical techniques for Re–Os isotopes were described by Ravizza & Turekian (1989) and Hauri & Hart
(1993). Samples weighing 200–500 g were powdered in
an agate shatterbox (Takazawa et al., 1999). Re and Os
were determined on separate powder splits. For the
analysis of Os, >1 g of powder was fused by the NiS
29
JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
JANUARY 2001
Fig. 4. 187Os/188Os ratios, Re and Os contents vs MgO content, and 187Os/188Os vs Al2O3 content from the Bozu section, Horoman peridotite.
Dashed line represents Re and Os content for primitive upper mantle (PUM) (McDonough & Sun, 1995).
reflects varying extents of melt extraction, the negative
correlation between Re and MgO contents suggests that
Re behaves as a moderately incompatible element during
melting. This interpretation is supported by the positive
correlation between Re and Al2O3, Sc and Yb contents,
and between Re and the modal proportion of clinopyroxene (see Fig. 5, below). On the other hand, the
small variation in Os concentration (a factor of about two;
Fig. 4) indicates that Os behaves as a highly compatible
element during mantle melting (Morgan, 1986; Hart &
Ravizza, 1996). However, the Os contents do not correlate with the 187Os/188Os ratios, or the abundances of
Re or MgO (Fig. 4). Possibly the expected positive
correlation between Os and MgO abundance is masked
by the nugget effect; that is, the difficulty in obtaining a
representative abundance of an element that is highly
concentrated in rare mineral phases.
The 187Os/188Os ratios for Horoman peridotites define
a strong positive correlation with the Re and Al2O3
contents (Figs 4 and 5) and a more scattered positive
correlation with 187Re/188Os ratios (Fig. 6a). Thus, there
is a gradual decrease in the Os isotopic composition from
plagioclase lherzolites through lherzolites to harzburgites.
The 187Os/188Os ratios for the ultramafic samples are
lower than the Os isotopic composition estimated for the
Primitive Upper Mantle (PUM, 0·1290 ± 0·0009; Meisel
et al., 1996). Moreover, most of the samples have 187Re/
188
Os ratios below that of PUM (0·4–0·428 for PUM;
McDonough & Sun, 1995; Meisel et al., 1996). However,
five N-Type plagioclase lherzolites plot to the right of
the geochron and three of the five samples have higher
187
Re/188Os ratios than the estimates for PUM (Fig. 6a).
It should be noted that the variation in 187Re/188Os ratios
in the N-Type plagioclase lherzolites is controlled by
both Re and Os contents, which vary by a factor of
about two, and range to higher and lower values than
estimates for PUM (Fig. 4).
The positive correlation defined by the Horoman peridotites in the isochron diagram ( 187Os/188Os vs 187Re/
188
Os) yields an apparent ‘age’ of 0·91 ± 0·35 Ga (Fig.
6a, excluding sample Bz-257). If the two mafic layers are
included in the regression, the apparent ‘age’ is 1·12 ±
0·24 Ga (Fig. 6b). Although these trends are not isochrons,
these apparent ‘ages’ are within error of the 833 ±
78 Ma inferred from whole-rock Sm–Nd isotopic data of
peridotites by Yoshikawa & Nakamura (2000). The age
of >850 Ma has been interpreted as the time of melt
extraction (Yoshikawa & Nakamura, 2000), but the Re
depletion model age (TRD) for the E-type plagioclase
lherzolite (Bz-257) with the lowest 187Os/188Os ratio
(0·1158) yields a minimum age for the depletion event
of 1·86 Ga (Table 1). Moreover, the trend in Fig. 6b
suggests that the Type I mafic layers are genetically
related to the peridotites, consistent with their Re–Os
model ages of >1·1–1·3 Ga (Table 1). Thus, the age of
the Type I mafic layers is much older than the 80 Ma
inferred from Nd isotopic data for core–margin pairs by
Takazawa et al. (2000). If the Type I mafic layers are
80 Ma, it would require an unrealistically high 187Re/
188
Os ratio of >1400 to produce the measured 187Os/
30
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
Fig. 5. 187Os/188Os ratios, Al2O3, Yb, Sc contents and modal clinopyroxene vs Re content. Modal proportions of clinopyroxene (in wt %) have
been calculated by mass balancing the mineral compositions with that of the whole rock (Takazawa et al., 2000). Dashed line represents Re
content for PUM; gray square represents the composition of the PUM (McDonough & Sun, 1995; Meisel et al., 1996); other symbols as in Fig.
4.
188
Os of about two, and a subsequent major modification
of the 187Re/188Os ratios, lowering it to the observed
value of >86 (Table 1). Therefore we consider the age
of the Type I mafic layers to be approximately the same
as that of the peridotites.
these elements reflect a varying proportion of a ‘basaltic
component’ in the Horoman peridotites. This observation
is consistent with the results presented by Burton et al.
(1999). They show that the Os budget in peridotites is
mainly controlled by sulfides, but a significant proportion
of the Re budget is located in the silicates. Thus, the
Re/Os ratio is low in sulfides but high in silicates; for
example, garnet has high Re/Os ratios (Roy-Barman et
al., 1996; Burton et al., 1998; Righter & Hauri, 1998).
Two alternatives have been proposed to explain the
variation in the proportion of a ‘basaltic component’ in
the Horoman peridotites: (1) variable extents of melt
extraction from a fertile mantle represented by the NType plagioclase lherzolite (Takahashi, 1992; Takazawa
et al., 2000; Yoshikawa & Nakamura, 2000); (2) melting
of lherzolite to create residual harzburgite and complementary plagioclase lherzolite, which represents a
region of melt accumulation; i.e. refertilization (Obata &
Nagahara, 1987).
DISCUSSION
What process controlled the Re–Os
systematics of the Horoman massif?
The correlation between the 187Os/188Os ratios and the
Re contents (Fig. 5) and the lack of correlation between
the Os isotopes and the Os concentrations (not shown)
suggest that the isotopic composition of the samples is
mainly controlled by variable Re content. The negative
correlation between Re and MgO and the positive correlation between Re and Al2O3, Sc and Yb contents, and
the modal proportion of clinopyroxene indicates that
31
JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
JANUARY 2001
Perhaps severe deformation has juxtaposed rock types
that were not adjacent when they formed (Ozawa & Takahashi, 1995; Yoshikawa & Nakamura, 2000; Takazawa et
al., 2000). However, a melting process provides no explanation for either the fertile plagioclase lherzolites with
high 187Re/188Os and low 187Os/188Os ratios relative to
those estimated for PUM and chondrites, or the sample
with the anomaluosly low 187Os/188Os ratios of 0·1156 (Bz257, Fig. 6a). Also, it is surprising that fertile plagioclase
lherzolite, 60% of the total outcrop (Takazawa et al., 1999),
is so abundant in lithospheric mantle represented by the
Horoman peridotite.
These obstacles for the simple melting hypothesis can
be overcome by a refertilization model, whereby a basaltic
component is added to a depleted peridotite. The addition
of a MORB-like component to a depleted peridotite
can explain the depleted Os isotopic signature, and the
suprachondritic 187Re/188Os ratios in the five plagioclase
lherzolites (Fig. 6a), as well as the the correlations between
Re, 187Os/188Os ratios, MgO, modal clinopyroxene, and
Al2O3, Sc and Yb contents (Figs 2 and 4). As refertilization
would be controlled by melt migration pathways, this
process could also explain meter-scale geochemical variations and the lowest 187Os/188Os ratios of 0·1156 (Bz257).
A variant of the refertilization model for the Horoman
peridotite proposed by Obata & Nagahara (1987) has
been advocated by Rehkämper et al. (1999), who measured platinum group elements (PGE) in six ultramafic
rocks from the Horoman massif. They found that the
three fertile Horoman lherzolites (>3% CaO) have suprachondritric Pt/Ir and Pd/Ir (as with Re and Os, the
elements in the numerator are incompatible relative to
the elements in the denominator). Although Pd–Ir data
are not available for the Horoman samples with suprachondritic 187Re/188Os, the four samples with suprachondritic Pd/Ir ratios do not have anomalously high
187
Re/188Os ratios, and the Os concentrations do not
correlate with the Ir content (Table 1). Rehkämper et al.
(1999) interpreted the high Pd/Ir ratios in the Horoman
peridotite by a combination of two processes: (1) the
refertilization of a depleted peridotite by sulfides derived
from percolating magmas, with or without the addition
of a basaltic component; (2) depletion of Ir abundance
by secondary processes such as progressive dissolution of
a trace carrier phase in percolating magma. The difficulty
in obtaining representative Os and Ir abundances (the
nugget effect), and the lack of correlation between Re/
Os and Pd/Ir indicate that the complexity of Re and
PGE abundance variations in mantle rocks is not well
understood (e.g. Snow & Schmidt, 1998; Rehkämper et
al., 1999). Nevertheless, suprachondritic Pd/Ir and Re/
Os abundance ratios in the Horoman and other peridotite
massifs (Reisberg et al., 1991; Roy-Barman et al., 1996;
Fig. 6. (a) 187Os/188Os vs 187Re/188Os. The ultramafic samples define
a rough positive correlation with an apparent ‘age’ of 0·91 ± 0·35 Ga
(excluding the sample Bz-257; dashed line with a 2 York regression).
This trend intersects the geochron at a lower 187Os/188Os than that for
present PUM, thereby indicating that the mantle source was a depleted
mantle. Five N-Type plagioclase lherzolites have high 187Re/188Os ratios
plotting to the right of the geochron. Values for PUM and geochron
from Meisel et al. (1996). Values for chondrite and geochron from
Smoliar et al. (1996). (b) 187Os/188Os–187Re/188Os isochron diagram for
Horoman samples including the two Type I mafic layers. All of the
samples define an apparent ‘age’ of 1·12 ± 0·24 Ga (2 York
regression), within error of the 833 ± 78 Ma age determined using
Nd isotopes (Takazawa, 1996; Yoshikawa & Nakamura, 2000). The
collinearity between the mafic layers and the ultramafic rocks suggests
that the mafic and ultramafic rocks are genetically linked. The dotted
lines and the open square show the region displayed in (a).
The first model, variable extraction of melt, can explain
the correlations between 187Os/188Os ratios, amount of
modal clinopyroxene and abundances of MgO, Al2O3, Re,
Sc and Yb (Figs 2 and 4). However, a major difficulty
with the interpretation that the lithologic variation, from
harzburgite to fertile lherzolite, formed as a result of partial
melting is the repetitive layering of rock types on the scales
of meters to hundreds of meters. These small-scale and
spatially systematic heterogeneities are not expected from
a melting process. Another example of local heterogeneity
is that plagioclase lherzolites with the highest and the lowest
Os isotopic ratios are located only 5 m apart (Fig. 3b).
32
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
Fig. 7. Primitive mantle normalized trace and major element contents for representative samples of the N-Type plagioclase lherzolite, E-Type
plagioclase lherzolite and Type I mafic layers. We have also plotted the average for both the depleted E-Type plagioclase lherzolite and the
mafic layers used as end-member in the mixing models (bold gray lines). It should be noted that in the depleted end-member (E-Type plagioclase
lherzolite), the elements more incompatible that Eu have been modified (enriched) by metasomatic processes. Data from Takazawa et al. (1999,
2000). Samples: Bz-260 for plagioclase lherzolite E-Type; Bz-250, Bz-254, Bz-256 and 62210 for N-Type; G-2 for mafic layer. Primitive mantle
values from McDonough & Sun (1995).
Burnham et al., 1998) stimulate us to evaluate a refertilization model using the major and trace element
composition of the Horoman mafic and ultramafic rocks.
mafic rocks (Takazawa et al., 1999, 2000) in a multiple
component analysis. The results indicated that the Type
II mafic layers are not a suitable basaltic component;
however, the Type I mafic layers are an adequate mafic
component for the mixing model.
Several observations indicate that Type I mafic layers
may have been involved in a refertilization process:
(1) the present mineralogy of the mafic layers and of
the seams in the plagioclase lherzolite do not represent
primary igneous phases; the present mineralogy in both
cases is the product of subsolidus reaction at low pressure (Takahashi & Arai, 1989; Ozawa & Takahashi,
1995; Takazawa et al., 1999). Using the major and trace
element contents of the cores and margins of Type I
layers, Takazawa et al. (1999) determined that they
formed as garnet pyroxenites. The seams of fine-grained
plagioclase, olivine, spinel, minor orthopyroxene and
very rare clinopyroxene in the plagioclase lherzolite from
the Lower Zone are interpreted to have formed by a
decompression reaction between two pyroxenes and
spinel, which in turn were derived from a reaction
between garnet and olivine (e.g. Ozawa & Takahashi,
1995; Takahashi, 1997). The presence of garnet in the
mafic layers and in the seams of the plagioclase lherzolite
shows that both the seams and Type I layers formed
within the garnet stability field. In contrast, the composition of the Type II mafic layers indicates that they
Can interaction between residual peridotite
and mafic layers explain the major and
trace element contents of the fertile
plagioclase lherzolite?
Takazawa et al. (2000) showed that the full compositional
range of the Horoman peridotites from fertile plagioclase
lherzolite to harzburgite cannot be explained by twocomponent mixing. However, a subset of these trends
are linear; namely, the compositional variation from the
N-Type plagioclase lherzolite to lherzolite. Therefore,
harzburgites are not considered in the following mixing
calculations. As proposed by Obata & Nagahara (1987),
harzburgites may have formed as residual rocks from
melting of lherzolite, whereas the compositional variation
from lherzolite to plagioclase lherzolite represents variable proportion of melt accumulation.
In the Horoman peridotite, there are two dominant
types of mafic layers, Type I and Type II (Niida, 1984;
Takazawa et al., 1999). To evaluate whether the mafic
layers represent the basaltic component responsible for
the formation of the N-Type plagioclase lherzolite, we
used the trace element abundances of the ultramafic and
33
JOURNAL OF PETROLOGY
VOLUME 42
NUMBER 1
JANUARY 2001
plagioclase lherzolite and the Type I mafic layers. We
used the average composition of Type I layers reported
by Takazawa et al. (1999); it is significant that this average
is probably a composite of a melt composition and its
associated garnet clinopyroxene cumulate. Hence this
process, which we will model by two-component mixing,
may be described more realistically as melt–peridotite
reaction with the melt reacting and precipitating clinopyroxene and garnet within the peridotite. As garnet is
an important phase that controls Re content of peridotites
(Roy-Barman et al., 1996; Burton et al., 1998; Righter &
Hauri, 1998), the addition of garnet pyroxenite to a
depleted peridotite can lead to high Re concentration
and Re/Os ratios in the N-Type plagioclase lherzolite.
For example, addition of garnet pyroxenite with high
Re/Os ratios, like the Type I mafic layers (Re/Os =
7–14), to a depleted peridotite can explain the N-Type
plagioclase lherzolites plotting to the right of the geochron
(Fig. 6a).
In the mixing model we do not use abundances of
incompatible elements, which may have been affected
by subsequent metasomatism. Specifically, we modeled
the major elements (including K and P), and the trace
elements from Gd to Lu, including Cr and Sc (Fig. 7).
It should be noted that K and P were affected by
metasomatism, and we include them only to illustrate that
the abundances of such elements are not well explained by
a two-component mixing model. First, the data are
normalized following the scheme of Allègre et al. (1995)
so that each element is given a comparable weight. Each
element used is normalized to its standard deviation,
multiplied by the concentration range and divided by
the mean concentration and the analytical error. Clearly,
the elements that have a larger concentration range will
be better indicators of the processes responsible for the
geochemical variation. To model the mixing process, we
consider the simple model presented by Cantagrel et al.
(1984); they expressed the mixing equation
Fig. 8. (Cm − Ca) vs (Cb − Ca) diagram to determine the proportion
of the end-member component in a mixture (Cantagrel et al., 1984).
Each point represents one element used in the mixing model. Cm, Ca
and Cb are the concentrations of each element in the mixture (N-Type
plagioclase lherzolite), the depleted peridotite (average of the E-Type
plagioclase lherzolite) and in the mafic layer (average of the Type I
mafic layers), respectively. Xb represents the slope of the line, which
defines the proportion of the mafic layer in the mixture. In this case
we used one variable regression.
are plagioclase-rich cumulates formed at low pressure
(Takazawa et al., 1999).
(2) The collinearity of the Type I mafic layers and the
ultramafic rocks in a Re–Os isochron diagram suggest
that these rocks are genetically related. The Re–Os
isotope systematics of these rocks defines an apparent
age ranging from >0·91 ± 0·35 Ga (peridotites) to 1·12
± 0·24 Ga (including the mafic rocks), similar to the
age reported for the Horoman peridotites based on
Nd isotopes (Takazawa, 1996; Yoshikawa & Nakamura,
2000). The similarity in the ages obtained from two
different isotopic system supports the hypothesis that the
mafic layers and the peridotites are genetically linked
(Fig. 6a and b).
(3) The higher proportion of plagioclase lherzolite and
mafic layers in the Upper Zone of the massif (Niida,
1984) is consistent with the hypothesis of refertilization
by the addition of a basaltic component (mafic layer) to
a depleted peridotite (lherzolite) as responsible for the
formation of the plagioclase lherzolite.
We evaluate a model whereby the N-Type plagioclase
lherzolite forms as a mixture of the depleted E-Type
Ca(1−Xb) + CbXb=Cm
as
(Cb−Ca)Xb = (Cm−Ca)
where Ca, Cb and Cm are the concentrations of an element
in the depleted peridotite, the mafic layer, and the mixture
(fertile peridotite), respectively; Xb is the proportion of
mafic layer in the mixture. We have defined the composition of the depleted lherzolite, the mafic layers and
the fertile peridotite (the mixture), so we can plot (Cm −
Ca) as a function of (Cb − Ca), where each point represents
one element. If the fertile peridotite is produced by
mixing between a depleted peridotite and a mafic layer,
the points (the elements) in the figure will define a line
passing through the origin with a slope Xb. The slope
represents the proportion of the mafic component in the
34
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
Fig. 9. Calculated trace and major element contents from the mixing model normalized by the measured value for four N-Type plagioclase
lherzolites. It should be noted that with the exception of P and K, the model reproduces the values for major and trace elements. We used the
same end-members for all the samples, changing only the mafic layer proportion in the mixture.
mixture (Fig. 8). Any other process, such as melting, will
produce strong dispersion in this type of plot where
compatible and incompatible elements are plotted together. Thus, we apply a simple linear regression and
obtained the proportion of the mafic component in the
mixture for each N-Type plagioclase lherzolite analyzed.
The uncertainty in the slope (proportion of mafic layers),
is obtained by propagating the analytical errors reported
by Takazawa et al. (1999, 2000) and the errors in the
linear regression.
Figure 9 shows calculated abundances normalized to
their measured values for four of the 10 N-Type plagioclase lherzolites. The mixing model reproduces the measured major and trace element composition for the fertile
plagioclase lherzolite fairly well by using the same two
end-members. As we previously mentioned, K and P are
the two elements that are not well reproduced by this
model. We believe that the K and P contents chosen for
the depleted peridotite end-member are too high, as a
result of late metasomatism (Fig. 7). We conclude that
the refertilization process is a plausible process for creating
the fertile plagioclase lherzolite as previously proposed
by Obata & Nagahara (1987).
The age of the melting and melt–rock
mixing and reaction events
The ages of the melting, mixing and reaction events
affecting the Horoman peridotite are not well constrained.
The oldest inferred age is 1·86 Ga, the Re depletion
model age for an E-Type plagioclase lherzolite (Bz-257,
Table 1). Based on Nd isotopic data for lherzolites and
plagioclase lherzolites, the age of a melting event is
considered to be >833 ± 78 Ma (Yoshikawa & Nakamura, 2000). The ultramafic samples (omitting Bz-257,
187
Os/188Os = 0·1156) define an apparent age of 0·91
± 0·35 Ga in the Re–Os isochron diagram. If both
mafic layers are included in the regression they define
an apparent age of 1·12 ± 0·24 Ga (Fig. 6). Although
the Re–Os isotope systematics does not define an isochron, the consistency of the >900 Ma age defined by
both isotopic systems suggests that this age has a geologic
meaning. If >900 Ma is the age of melting, it is difficult
to explain the 1·86 Ga Re depletion model age. If local
melting and refertilization of a lherzolite was responsible
for the >900 Ma age, we can explain the relatively old
model age of 1·86 Ga (Table 1) as representing an older
35
JOURNAL OF PETROLOGY
VOLUME 42
melting event. In this case, sample Bz-257 was unaffected
by the >900 Ma local melting and refertilization event.
NUMBER 1
JANUARY 2001
during the analytical work. This work has benefited
greatly from discussion with G. Ravizza and M. Rëhkamper, and the thoughtful reviews of J.-L. Bodinier,
J. Snow, M. Obata, E. Rampone and G. Piccardo.
CONCLUSIONS
In the 140 m Bozu section of the Horoman peridotite
massif the range in 187Os/188Os ratios is similar to that
reported for other peridotitic massifs, thereby suggesting
that the process responsible for the Re–Os isotope variation at the meter scale and the whole-massif scales are
similar. The Re–Os isotopic systematics of Horoman
peridotites indicates that the 187Os/188Os ratios are mainly
controlled by variation in the Re content, with Re
concentration governed by the amount of a ‘basaltic
component’.
The systematic layering of rock types in the Horoman
peridotite has been interpreted as resulting from varying
extents of melt extraction (e.g. Takazawa et al., 2000;
Yoshikawa & Nakamura, 2000). Stimulated by the suprachondritic 187Re/188Os ratios of some plagioclase lherzolites, we evaluated an alternative model; the
refertilization process proposed by Obata & Nagahara
(1987). The Re–Os isotopic systematics, major and trace
elements contents, and field and petrographic information
indicate that the N-Type plagioclase lherzolites can be
explained by a refertilization process involving mixing/
melt–rock reaction of a MORB-like rock and depleted
lithospheric mantle. This process can explain some of
the large compositional variations occurring over the
scale of meters. The consistency of the >900 Ma age
defined by Sm–Nd (Yoshikawa & Nakamura, 2000) and
Re–Os isotopic systems suggests that this age has a
geologic meaning, and that this age may represent the
time of local melting and refertilization. A Re depletion
model age of >1·86 Ga may represent an older melting
event that survived the local melting and melt–rock
reaction event at >0·9 Ga. The apparent age of >0·9
Ga defined by the peridotites in the Re–Os isochron
diagram and the >1·1–1·3 Ga Re–Os model ages for
the Type I mafic layers suggest that these mafic and
ultramafic rocks are genetically related. Although the
refertilization model of Obata & Nagahara (1987) is a
plausible process for explaining the meter-scale compositional variations from lherzolite to plagioclase lherzolite, the variable Re–Os model ages, and the
suprachondritic 187Re/188Os ratios of some plagioclase
lherzolites, we recognize that the strength of the constraints arising from Re and PGE abundances need
further evaluation with more detailed studies of Re and
PGE abundances in mantle rocks.
REFERENCES
Allègre, C. J., Schiano, P. & Lewin, E. (1995). Differences between
oceanic basalts by multitrace element ratio topology. Earth and
Planetary Science Letters 129, 1–12.
Burnham, O. M., Rogers, N. W., Pearson, D. G., van Calsteren, P.
W. & Hawkesworth, C. J. (1998). The petrogenesis of the eastern
Pyrenean peridotites: an integrated study of their whole-rock geochemistry and Re–Os isotope composition. Geochimica et Cosmochimica
Acta 62, 2293–2310.
Burton, K. W., Schiano, P., Birck, J. L. & Allègre, C. J. (1998). The
behaviour of Re and Os in mantle minerals with implications for
mantle melting. EOS Transactions, American Geophysical Union, 79(46),
Supplement, S373.
Burton, K. W., Schiano, P., Birck, J. L. & Allègre, C. J. (1999). Osmium
isotope disequilibrium between mantle minerals in a spinel-lherzolite.
Earth and Planetary Science Letters 172, 311–322.
Cantagrel, J. M., Didier, J. & Gourgaud, A. (1984). Magma mixing:
origin of intermediate rocks and enclaves in volcanism to plutonism.
Physics of the Earth and Planetary Interiors 35, 63–76.
Hart, S. R. & Ravizza, G. E. (1996). Os partitioning between phases
in lherzolite and basalts. In: Basu, A. & Hart, S. R. (eds) Earth Processes:
Reading the Isotopic Code. Geophysical Monograph, American Geophysical Union
95, 123–134.
Hauri, E. H. & Hart, S. R. (1993). Re–Os isotope systematics of HIMU
and EMII oceanic island basalts from the South Pacific Ocean. Earth
and Planetary Science Letters 114, 353–371.
Komatsu, M. & Nochi, M. (1966). Ultrabasic rocks in the Hidaka
metamorphic belt, Hokkaido, Japan. I. Mode of occurrence of the
Horoman ultrabasic rocks. Earth Sciences 20, 99–108 (in Japanese).
Liu, Y. Z. & Tanaka, T. (1999). Osmium-isotopic composition of
Horoman ultramafic complex in Hokkaido, Japan. Proceedings, Ninth
Annual V. M. Goldschmidt Conference, Boston, MA. Houston, TX: Lunar
and Planetary Institute, p. 175.
Luck, J. M. & Allègre, C. J. (1991). Osmium isotopes in ophiolites.
Earth and Planetary Science Letters 107, 406–415.
McDonough, W. F. & Sun, S.-s. (1995). The composition of the Earth.
Chemical Geology 120, 223–253.
Meisel, T., Walker, R. J. & Morgan, J. W. (1996). The osmium isotopic
composition of the Earth’s primitive upper mantle. Nature 383,
517–520.
Morgan, J. W. (1986). Ultramafic xenoliths: clues to Earth’s late
accretionary history. Journal of Geophysical Research 91, 12375–12387.
Niida, K. (1974). Structure of the Horoman ultramafic massif of the
Hidaka metamorphic belt in Hokkaido, Japan. Journal of the Geological
Society of Japan 80, 31–44.
Niida, K. (1984). Petrology of the Horoman ultramafic rocks in the
Hidaka metamorphic belt, Hokkaido, Japan. Journal of the Faculty of
Science, Hokkaido University, Series IV 21(2), 197–250.
Obata, M. & Nagahara, N. (1987). Layering of alpine-type peridotite
and the segregation of partial melt in the upper mantle. Journal of
Geophysical Research 92, 3467–3474.
Ozawa, K. & Takahashi, N. (1995). P–T history of a mantle diapir:
the Horoman peridotite complex, Hokkaido, northern Japan. Contributions to Mineralogy and Petrology 120, 223–248.
ACKNOWLEDGEMENTS
We thank K. Koga for help in modelling, and G. Ravizza,
B. Peucker-Ehrenbrink, J. Blusztajn and L. Ball for help
36
SAAL et al.
Re–Os ISOTOPES IN THE HOROMAN PERIDOTITE
Ravizza, G. & Turekian, K. K. (1989). Applications of the 187Re–187Os
system to black shale geochronology. Geochimica et Cosmochimica Acta
53, 3257–3262.
Rehkämper, M., Halliday, A. N., Alt, J., Fitton, J. G., Zipfel, J. &
Takazawa, E. (1999). Non-chondritic platinum-group element ratios
in oceanic mantle lithosphere: petrogenetic signature of melt percolation? Earth and Planetary Science Letters 172, 65–81.
Reisberg, L. C. & Lorand, J. P. (1995). Longevity of sub-continental
mantle lithosphere from osmium isotope systematics in orogenic
peridotite massifs. Nature 376, 159–162.
Reisberg, L. C., Allègre, C. J. & Luck, J. M. (1991). The Re–Os
systematics of the Ronda ultramafic complex of southern Spain.
Earth and Planetary Science Letters 105, 196–213.
Righter, K. & Hauri, E. H. (1998). Compatibility of rhenium in garnet
during mantle melting and magma genesis. Science 280, 1737–1741.
Roy-Barman, M., Luck, J. M. & Allègre, C. J. (1996). Os isotopes in
orogenic lherzolite massifs and mantle heterogeneities. Chemical Geology 130, 55–64.
Shirey, S. B. & Walker, R. J. (1998). The Re–Os isotope system in
cosmochemistry and high temperature geochemistry. Annual Review
of Earth and Planetary Sciences 26, 423–500.
Smoliar, M. I., Walker, R. J. & Morgan, J. W. (1996). Re–Os age
of group IIA, IIIA, IVA and IVB iron meteorites. Science 271,
1099–1102.
Snow, J. E. & Schmidt, G. (1998). Constraints on Earth accretion
deduced from noble metals in the oceanic mantle. Nature 391,
166–169.
Takahashi, N. (1991). Origin of three peridotite suites from Horoman
peridotite complex, Hokkaido, Japan: melting, melt segregation and
solidification processes in the upper mantle. Journal of Mineralogy,
Petrology and Economic Geology 86, 199–215.
Takahashi, N. (1992). Evidence for melt segregation towards fractures
in the Horoman mantle peridotite complex. Nature 359, 52–55.
Takahashi, N. (1997). Incipient melting of mantle peridotites observed
in the Horoman and Nikanbetsu peridotite complexes, Hokkaido,
northern Japan. Journal of Mineralogy, Petrology and Economic Geology 92,
1–24.
Takahashi, N. & Arai, S. (1989). Textural and chemical features of
chromian spinel–pyroxene symplectites in the Horoman peridotites,
Hokkaido, Japan. Science Report of the Institute of Geosciences, University
of Tsukuba, Section B 10, 45–55.
Takazawa, E. (1996). Geodynamic evolution of the Horoman peridotite,
Japan: geochemical study of lithospheric and asthenospheric processes. Ph.D. thesis, Massachusetts Institute of Technology, Cambridge.
Takazawa, E., Frey, F. A., Shimizu, N., Obata, M. & Bodinier, J. L.
(1992). Geochemical evidence for melt migration and reaction in
the upper mantle. Nature 359, 55–58.
Takazawa, E., Shimizu, N. & Frey, F. A. (1994). Development and
evolution of the lithospheric mantle: a case study from the Horoman
peridotite, Japan. Mineralogical Magazine 58A, 889–890.
Takazawa, E., Frey, F. A., Shimizu, N. & Obata, M. (1996). Evolution
of the Horoman peridotite (Hokkaido, Japan): implications from
pyroxene compositions. Chemical Geology 134, 3–26.
Takazawa, E., Frey, F. A., Shimizu, N., Saal, A. E. & Obata, M.
(1999). Polybaric petrogenesis of mafic layers in the Horoman
peridotite complex, Japan. Journal of Petrology 40, 1827–1851.
Takazawa, E., Frey, F. A., Shimizu, N. & Obata, M. (2000). Whole-rock
compositional variations in an upper mantle peridotite (Horoman,
Hokkaido, Japan): implications for melt segregation, migration and
reaction. Geochimica et Cosmochimica Acta 64, 695–716.
Yoshikawa, M. & Nakamura, E. (2000). Geochemical evolution of
the Horoman peridotite complex: implications for melt extraction,
metasomatism and compositional layering in the mantle. Journal of
Geophysical Research 105, 2879–2901.
Yoshikawa, M., Nakamura, E. & Takahashi, N. (1993). Rb–Sr isotope
systematics in a phlogopite-bearing spinel lherzolite and its implications for age and origin of metasomatism in the Horoman
peridotite complex, Hokkaido, Japan. Journal of Mineralogy, Petrology
and Economic Geology 88, 121–130.
37